Oktatás | Andragógia » Hannah De Mulder - Second Language Acquisition of Gender, Italian Gender in Native Speakers of Dutch and English

Source: http://www.doksinet Hannah De Mulder Second Language Acquisition of Gender: Italian gender in native-speakers of Dutch and English Source: http://www.doksinet Table of Contents Acknowledgements 4 Introduction 5 Chapter 1 Gender 1.1 What is Grammatical Gender? 1.2 The Syntax of Gender 1.3 The Gender System in Italian, Dutch and English 1.31 Italian 1.32 Dutch 1.33 English 1.34 Summary 1.4 Acquisition and Processing of Gender in the L1 1.41 Acquisition of Gender in the L1 1.42 Processing of Gender in the L1 1.5 Acquisition and Processing of Gender in L2 1.51 Acquisition of Gender in the L2 1.52 Processing of Gender in the L2 1.53 The L1 in L2 Acquisition and Processing of Gender 1.6 Grammatical Gender: a Broader Perspective 6 6 7 8 8 10 11 11 12 12 16 21 21 24 25 33 Chapter 2 The Study 2.1 Outline of the Study 2.2 Research Questions 2.21 Predictions 2.3 The Experiments 2.31 Subjects 2.32 Design 2.33 Materials 2.34 Procedure 35 Chapter 3 Results 3.1 Accuracy

Comprehension Questions 3.2 Results SPR Experiment 3.21 General Differences between Groups 3.22 Influence of the DP 3.3 Results AJT 3.31 General Differences between Groups 3.32 Influence of the DP 3.33 Consistency 3.4 Relation SPR and AJT Data 3.5 Individual Data SPR and AJT 3.6 Summary 52 35 37 37 43 43 45 46 50 2 52 52 53 54 56 57 58 61 63 64 65 Source: http://www.doksinet Chapter 4 Discussion 4.1 Results and Research Questions 4.11 Research Question 4: Influence of the DP 4.12 Research Question 1: Natives vs Non-Natives 4.13 Research Question 2: English vs Dutch 4.14 Research Question 3: Individuals 4.15 Summary of Research Questions 4.2 The Results in a Broader Perspective 4.21 FFFH vs FTFA 4.22 Processing and Knowledge in Language Acquisition 4.3 Directions for Future Research 66 66 70 72 73 74 75 75 77 78 References Appendix A: Language Background Questionnaires Appendix B: Individual Subject Information Appendix C: Italian Proficiency test Appendix D: Training

Sentences, Items, Fillers and Questions Appendix E: Accuracy Rate Comprehension Questions Appendix F: Instructions SPR and AJT Appendix G: Means, Standard Deviations and Ranges SPR Appendix H: Means, Standard Deviations and Ranges AJT 80 83 87 89 98 108 109 110 114 3 66 Source: http://www.doksinet Acknowledgements First of all, I would like to thank my supervisor, Sharon Unsworth, for her copious comments and suggestions both in setting up the experiments and in writing this thesis. Her input made this thesis better and considerably more reader friendly than it would have been otherwise. Many thanks also go to Antonella Sorace Without her help and support in creating and performing the experiments, this thesis would not have been possible. My time spent at the University of Edinburgh under her supervision was a valuable experience in many ways, but it was especially useful for this thesis in that it gave me the opportunity to test English learners of Italian as well as the Dutch

learners tested at the University of Utrecht. I am also very grateful to Francesca Filiaci for checking my Italian. Mara Mari and Dave Jones were of great help to me in finding the English learners of Italian and Manuela Pinto helped me find the Dutch learners. Many thanks! I would also like to thank Frank Wijnen for being the second reader of this thesis. Last, but definitely not least, I would like to thank my subjects. Without their participation, this study would not have been possible 4 Source: http://www.doksinet Introduction Faced with the job of acquiring a language with gender, the language learner has to determine the gender of an entity that does not always have very obvious gender characteristics: the noun. Take the French noun monde (world), for example Somehow, the learner has to figure out that this word is masculine and that this fact entails that it has to occur with the masculine form of the determiner and adjective. Considering how many nouns there are in French

and that the gender of each noun has to be determined and stored in the lexicon, this may seem a pretty tough task to do. Nonetheless, all native speakers eventually acquire the gender system of their language and they rarely make mistakes in a word’s gender once they have a mature linguistic system. What about second language learners though? How do they deal with acquiring the gender system of a new language? Are they capable of acquiring gender in the same flawless manner in which first language learners acquire gender or is the acquisition of gender more problematic in a second language? There is a relatively large body of research on the (second language) acquisition of gender (which will be discussed extensively in following chapters), but the results are somewhat mixed. Most studies find that the acquisition of gender remains problematic for many learners, but conversely, there are some studies that find nativelike standards of gender in second language learners. The

theoretical accounts of the acquisition of gender are also quite diverse. According to some, it is only possible to acquire gender in a second language if there is gender in the first language; following others, the amount of exposure to the second language is considered to be much more important for the learner than the nature of the first language. Given this multitude of points of view and data, it is quite hard to draw any firm conclusions regarding the second language acquisition of gender. This MA thesis aims to shed some light on the second language acquisition of gender in adults, looking specifically at the acquisition of gender in Dutch and English learners of Italian. In the first chapter, I will review the literature on the acquisition of gender and I will discuss the theoretical frameworks in which the syntax of gender is usually embedded. The second chapter presents the study that I have conducted on second language acquisition of gender. In this chapter, my research

questions will be addressed and design, materials and procedure of the study will be described. Chapter 3 consists of the results obtained from the two experiments conducted for this study and chapter 4 interprets and discusses these results. It will be shown that Dutch and English learners of Italian can acquire the Italian gender system quite successfully. Both groups of learners show that they know the gender of nouns and that they are aware of the effects that the noun’s gender has on the form of the determiner and the adjective. Both non-native groups show similar processing of gender as well, although there are some differences between the natives and the non-natives in this respect. In both groups there are individuals that perform completely within the native range in the two experiments that have been conducted for this thesis. The conclusion of this study is thus that it is possible for learners from different first language backgrounds to become successful, possibly even

nativelike, users of the Italian gender system. 5 Source: http://www.doksinet Chapter 1 Gender In this chapter, a general overview of the nature, acquisition and processing of grammatical gender is given. In the first half, grammatical gender is defined and the relevant theoretical background is discussed. As the experiment that lies at the heart of this thesis concerns the acquisition of Italian gender by English and Dutch learners of Italian, a short overview of the gender system of all three languages is provided as well. In the second half of this chapter, the acquisition of gender in first and second languages will be discussed, as will recent findings on the processing of gender. 1.1 What is Grammatical Gender? Grammatical gender is not a universal element of language, but it is relatively common in the languages of the world. There are many definitions of grammatical gender in the literature, but the following one by Matthews (1997, cited in Comrie 1999) is relatively

general and uncontroversial: ‘grammatical gender is a system in which the class to which a noun is assigned is reflected in the forms that are taken by other elements syntactically related to it’. This definition makes it clear that even though gender is in essence a feature of the noun, a language cannot truly be said to have gender if gender is not reflected in other elements of the language besides the noun (see also Corbett 1991). In French, for example, the noun table (table) is feminine, but this only really becomes clear when one considers it in combination with a determiner and/or an adjective: (1) la table the fem table fem the white table blanche white fem In this case, the feminine gender of the noun dictates that it has to occur with the feminine determiner (la instead of the male determiner le) and with the feminine form of the adjective (blanche instead of the masculine form blanc). This behaviour of ‘syntactically related elements’ is referred to as gender

agreement in the literature. The gender system of a language thus has two components: gender assignment (which is responsible for assigning a particular gender to each noun) and gender agreement (which dictates the morphophonological form of the gender-marked items other than nouns). Example (1) also clearly shows that grammatical gender and natural gender are generally unrelated 1; there is nothing inherently feminine about a table that entails that it should have feminine gender. This arbitrary nature of grammatical gender can also be seen in the fact that the particular gender assigned to a noun can vary across languages. In French, for example, the noun for flower (fleur) is feminine, whereas in Italian it is masculine (fiore). In all languages, however, there are some links between natural gender and grammatical gender. Nouns that clearly refer to female or male people or animals (like mother, father, mare or bull) generally have a gender feature that is congruent with their

natural gender. Even this rule has its exceptions, however, 1 But see section 1.3 for examples of languages for which this statement is not true 6 Source: http://www.doksinet as is evidenced by the fact that the Dutch and German nouns for girl (meisje in Dutch, Mädchen in German) are both neuter. 1.2 The Syntax of Gender In this overview of the syntax of gender, I will assume a generative model of gender found in many of the main publications on the acquisition of gender (cf. White 2003 for a basic model, shown in (2) and (3), and Franceschina 2005 for a more elaborate version). According to the generative model (cf Chomsky 1995 and Carstens 2000), gender is a grammatical feature (one of the so-called phi-features) that is specified in the lexical entry of each noun. The gender feature on the noun is considered to be interpretable at LF (i.e it includes information that is required for semantic interpretation and hence is readable at the LF interface). In the lexical entry of the

noun, the gender feature is always valued; that is, the particular gender of the noun is already specified (e.g table [gender: f]) when it enters the derivation Gender features on lexical items other than nouns are considered to be uninterpretable at LF and unvalued. Gender features on, for example, adjectives or determiners are thus not legible at the LF interface and are not specified for any particular gender. As these features are not legible at the LF interface, they have to be deleted by means of feature checking before they reach the LF interface. The valued gender feature of the noun thus checks any unvalued gender feature it encounters within an appropriate configuration (i.e a head/head or a specifier/head configuration, see White et al 2004), so that no unchecked features reach LF. (2) (3) The position of the adjective can vary depending on the language. In some languages (like French, Italian and Spanish, see (3)), the noun generally precedes the adjective, whereas in

other languages (like Dutch, German and English, see (2)) the adjective precedes the noun. The position of the adjective is determined by the strength of the features in Num. The Det-N-Adj order (French, Italian, Spanish) is derived if the features in Num are strong; the Det-Adj-N order (Dutch, German, English) is obtained 7 Source: http://www.doksinet if the features in Num are weak. As can be seen in (3), in languages with a Det-N-Adj order, the noun is raised to Num. This movement means that the noun can value the gender features of the adjective and the determiner, as it is now in a specifier-head relation with both the determiner and the noun. If this movement did not occur, the unvalued gender features of the determiner and the adjective would not be checked, resulting in uninterpretable features at LF which cause the derivation to crash. This raises the question why a Det-Adj-N order (example 2) is possible in some languages. After all, in this order the noun is not in a

specifier-head or a head-head relation with the determiner and the adjective, so this should entail that the derivation crashes at LF. The reason why the derivation does not crash is that the noun moves to Num covertly. The gender features of the determiner and the adjective are thus checked before they reach the LF interface, but after spell-out, which means that the Det-AdjN order is what is heard, but not what is processed at LF. 2 To summarise: gender is specified in the lexical entry of the noun, but not in the lexical entry of other gender-marked items. The gender feature of these other items has to be valued in the derivation so that gender agreement can take place. This valuation happens when the noun moves to Num (either overtly in languages with a Det-N-Adj order or covertly in languages with a Det-Adj-N order). 1.3 The Gender System in Italian, Dutch and English Following Corbett (1991) and Comrie (1999), a distinction can be made between languages that assign gender to

nouns based on semantic principles and languages that assign gender based on formal principles. In languages that assign gender based on semantic principles (like Tamil and Kannada), the meaning of the noun is the key determinant of the noun’s gender. In Tamil, for example, all male human or godly entities are denoted by nouns of masculine gender, all female human or godly entities are denoted by feminine nouns, and all the other nouns are denoted by neuter gender. Many other languages are much more formal in their principles for assigning gender. 3 In languages with formal gender systems, nouns are assigned gender based on rules that take the form of the noun as the prime criterion for gender assignment. In Hausa, for example, nouns ending in -aa are feminine; practically all nouns with a different ending are masculine. The three languages considered in the following subsections can be placed at varying points in the formal-semantic range of possible gender systems. 1.31 Italian

Like most Romance languages, Italian has a rich formal gender system, in which all nouns are marked for masculine or feminine gender. In most cases the gender of the noun is clearly visible in the word ending, as masculine singular nouns usually end in –o and feminine singular nouns usually end in –a (plural nouns generally end in –i and –e respectively). As in all formal gender systems, there are also some semantic 2 It is not entirely clear, however, how the correct form of the determiner is realised in the Det-Adj-N order. As feature checking occurs after spell-out (ie at LF), the gender feature of the determiner is unchecked at PF and hence the determiner is not valued for gender. How the correct morphophonological form of the determiner is realised at PF is thus unclear. Whilst this issue is relevant, it is not discussed any further as it is beyond the scope of this thesis. 3 Note that whereas strictly semantic gender systems do exist, strictly formal systems do not.

According to Corbett (1991), all formal systems have some semantic principles for determining the gender of a noun as well. Nouns with clearly female or male referents (like mother or father) will thus generally be assigned a congruent gender even in very formal systems. 8 Source: http://www.doksinet clues to gender (see footnote 3); nouns denoting clearly female or male entities are thus generally assigned a congruent gender (the noun ragazzo (boy) is masculine, ragazza (girl) is feminine). There are, however, some exceptions to the formal rules: the noun programma (program) is masculine, for example, even though it ends in –a and the noun mano (hand) is feminine irrespective of its –o ending. Apart from these exceptions to the –o/-a rule, Italian also has a relatively large number of singular nouns that end in –e. The gender of these nouns can thus not be predicted from the ending of the noun (the noun luce (light), for example, is feminine, but the noun fiume (river) is

masculine). The most reliable clues for the gender of an Italian noun are the lexical items that are also marked for gender: determiners, pronouns, adverbs and adjectives. Usually, the adjectives follow the noun that they are related to (a Det-N-Adj order; cf. (3)) as in (4): (4) il sasso the masc stone masc the white stone bianco white masc Some adjectives can precede the noun, however, resulting in a Det-Adj-N order (cf. (2)): (5) la brutta the fem ugly fem the ugly sculpture scultura sculpture fem The form of the adjective, pronoun and adverb generally follows the –o/-a rule. The adjective/adverb red is thus realised as rosso if it occurs with a masculine noun and rossa if it occurs in combination with a feminine noun; the pronoun my is realised as mio in relation to a masculine noun and mia with a feminine noun. The singular definite determiner occurs in two forms: il for masculine 4 and la for feminine. Even these clues to the gender of a noun can backfire occasionally,

however, as not all adjectives and adverbs end in –o or –a; difficile (difficult), for example, is invariant with respect to gender. In addition, if the noun following the definite determiner starts with a vowel, the definite determiner becomes l’ irrespective of the noun’s gender. A phrase like l’atlante verde (the green atlas) thus gives no clues with respect to the gender of the noun even though both a determiner and an adjective are present. In general though, Italian has a very transparent overt gender system that can be summarised relatively easily in a few phonological rules. Table 1 presents an overview of the facts that are relevant for this thesis: 4 Actually, the masculine definite determiner has two forms: il and lo. Which form is used depends on the initial phoneme of the noun or adjective that follows it. Only the form il is relevant for this thesis, however. Hence, the lo-form will be excluded from any subsequent discussion 9 Source: http://www.doksinet

Table 1: Relevant elements of the Italian gender system Singular definite determiner Ending singular adjective Ending singular noun Masculine il Feminine la -o -o (regular)/ -e (irregular) -a -a (regular)/ -e (irregular) 1.32 Dutch Dutch has a formal gender system in which nouns can be assigned either common gender (i.e masculine or feminine) or neuter gender Most Dutch nouns have common gender (about 75% of Dutch nouns are common gender, cf. Van Berkum 1996 cited in Hulk and Cornips 2006). There is no direct morphological expression of gender on the noun 5; gender is manifested indirectly by the form of other gendermarked words. Various word categories give clues to the gender of the noun: determiners, adjectives, quantifiers and pronouns can all be used in the identification of the right gender. The easiest way to determine the gender of a noun is by looking at the form of the singular definite determiner. Common gender nouns occur with the determiner de 6 (the), neuter gender

nouns with het (the). The form of the singular demonstrative pronoun is similarly determined by the gender of the noun. Common gender nouns take the demonstrative deze (this) if the object referred to is close by or die (that) if the object is far away. Neuter nouns, on the other hand, go with dit (this) for close by and dat (that) for far away. Quantifiers have a schwa added to the stem if the noun they agree with has common gender, but just the stem if the noun is neuter. A common noun like hond (dog) thus becomes iedere hond (every dog) in combination with the universal quantifier every, whereas a neuter noun like huis (house) becomes ieder huis (every house). The prenominal adjective shows a more complex alternation. If the adjective modifies a singular, indefinite and neuter NP, the prenominal adjective does not have a schwa added to its stem. In all other cases, it gets a schwa-suffix In example (6), the neuter singular noun huis (house) in combination with the indefinite article

een (a) dictates that the adjective mooi (beautiful) does not receive the schwa-suffix: (6) een mooi a beautiful a beautiful house huis house neut In example (7) on the other hand, the substitution of the indefinite determiner for a definite determiner means that the schwa-suffix is required: (7) het mooie the neut beautiful the beautiful house 5 6 huis house neut The diminutive suffix –je and the prefix ge- do reliably indicate that the noun is neuter, however. De is also the form of the plural definite determiner for both common and neuter nouns. 10 Source: http://www.doksinet Example (8) shows that a combination of the common singular noun hond (dog) and the indefinite determiner een also leads to a schwa-suffix on the adjective: (8) een mooie a beautiful a beautiful dog hond dog com Apart from being significantly less transparent than the Italian gender system, the Dutch gender system also differs in the word order of the DP. In Dutch only the DetAdj-N order is

possible; a Det-N-Adj order is never allowed 1.33 English Unlike Italian and Dutch, English has hardly any gender system to speak of, as there are only very few lexical items that show overt gender characteristics. The elements of English in which gender is present are the third person pronouns (he, she and it), anaphors (himself, herself and itself) and the endings of some nouns that clearly refer to male or female entities (actor vs. actress, for example) In these limited areas of English grammar in which gender plays a role, the assignment principles are thus purely semantic, in contrast to the predominantly formal principles of the gender systems of Dutch and Italian. Apart from the cases described above then, English does not have a gender feature on nouns or gender agreement and hence, as there is no feature to check, it does not require feature checking for gender. The order of the DP in English is Det-Adj-N. 1.34 Summary The three languages considered in this section and the

previous two sections thus differ considerably in their gender characteristics. English, unlike Dutch and Italian, has a very impoverished gender system in which the gender assignment rules are purely semantic. Dutch and Italian, on the other hand, both assign gender to nouns on the basis of formal principles. These formal principles differ quite significantly between the two languages, however, as the gender system is transparent in Italian and opaque in Dutch. In Italian, the gender of the noun is thus generally obvious from the noun’s ending, whereas in Dutch the noun’s ending rarely gives any clues to the noun’s gender. The word order of the DP is also different in the three languages Dutch and English both only allow a Det-Adj-N order, whereas in Italian the Det-NAdj order is the canonical word order (but the Det-Adj-N order is also possible for certain adjectives). Table 2 gives an overview of these differences Table 2: Overview characteristics gender systems

Predominantly Semantic Gender Predominantly Formal Gender Transparent Gender Det-N-Adj Order Det-Adj-N Order Italian X Dutch X English √ √ √ X √ √ √ X X √ N/A X √ 11 Source: http://www.doksinet 1.4 Acquisition and Processing of Gender in the L1 In this section, a general overview will be given of first language acquisition and processing of gender. Given the fact that the acquisition of gender is language dependent to some extent, the developmental trajectory for the acquisition of gender in Italian, Dutch and English will also be summarised briefly. 1.41 Acquisition of Gender in the L1 1.411 General Overview The acquisition of gender by L1 children is relatively well documented in the literature. A general trend that can be observed is that gender emerges quite early and in a relatively error-free manner, although there is some developmental variation depending on which language is being acquired. In principle, there are three different types of

information that the child could use in determining the gender of a noun: semantic clues, morphophonological clues and syntactic clues. If the child considers semantic clues, she will look at whether the noun refers to something with a clear natural gender and assume that the noun’s gender is congruent with the natural gender of the referent. Italian nouns like madre (mother) and padre (father) could thus be assigned gender on the basis of semantic clues even though the noun ending is ambiguous with respect to gender. If morphophonological information plays a role in the child’s gender assignment, the child will consider the form of the noun in determining its gender. A noun ending in something that is typically associated with a specific gender will then be assigned the gender that is congruent with that ending. An Italian child encountering nouns like pianta (plant) or finestra (window) for the first time could thus correctly assign feminine gender to these nouns purely on the

basis of their morphophonological form (i.e the –a ending) If the child focuses on syntactic clues, on the other hand, she will consider the forms of lexical items related to the noun like determiners and adjectives. On encountering the noun fiore (flower) for the first time in combination with the masculine determiner il or an adjective ending in -o, for example, the Italian child could thus assign masculine gender to the noun even though the form of the noun itself is ambiguous with respect to gender. Various studies have tried to show how the child uses these clues and whether there is a hierarchy in the importance that is assigned to each of these three clues. Intuitively, it seems plausible to assume that children start off by classifying nouns on the basis of their natural gender, realising only later on in their linguistic development that morphophonological and syntactic clues are also a valuable (and more reliable) source of gender information. Indeed, one of the first

significant studies in this area, conducted by Karmiloff-Smith (1979), was set up on this assumption: children start off by assuming that grammatical gender is based on natural gender and only at later stages in development do they become capable of appreciating morphophonological and syntactic cues. In order to test this assumption, Karmiloff-Smith presented her subjects (341 French children between 3 and 12 years old) with pictures of imaginary animate and inanimate objects that were given novel French-sounding names and asked them questions about the pictures. The questions had been created so that children were forced to use determiners and adjectives in their answers, so that it could be ascertained which gender the child had assigned to the novel objects. The pictures and names were modified so that the relative importance of the three different types of clues could be tested. This was done in various different ways, for example: 12 Source: http://www.doksinet • • •

• The gender of the determiner and the noun suffix were consistent (as in un bicron in which the masculine indefinite determiner un is used in combination with the masculine noun suffix –on). Both the determiner and the suffix can thus be used as clues. No determiner was provided, so the children had to base their decisions purely on the phonological form of the noun, as in deux bicrons (two bicrons), in which the word deux does not give any clues to the gender of bicrons. The gender of the determiner and the noun suffix were inconsistent, as in une bicron, in which the feminine indefinite determiner une is used in combination with the masculine noun suffix –on. In this case, the child thus has to make a choice between the gender information on the determiner and the information on the noun suffix. The sex of the person depicted in the picture and the noun suffix were inconsistent with no gender marked determiner. To give an example: a picture with clearly female persons is

referred to as deux bicrons. The child thus has to choose between the natural gender clue and the noun suffix clue. It turned out that L1 French children are predisposed towards using phonological clues over all other clues. In situations in which there is an inconsistency between the phonological clue and either or both the syntactic and the semantic clue, the information from the phonological clue is preferred until the age of nine. If the child is thus presented with a picture with a clearly feminine referent and the picture is described as une bicron (so the determiner is feminine, but the noun ending is masculine), the child will generally assign masculine gender to the noun. This does not mean that syntactic and semantic clues play no role in the young child’s gender assignment, however. If children are presented with the phrase un coumile, in which the masculine definite determiner is used and the noun ending –ile is not specific to any gender (i.e it can occur with both

feminine and masculine nouns), they tend to assume that coumile is masculine. Similarly, if the child is presented with a picture with clearly feminine referents and hears the phrase deux coumiles (in which no gender is present at all), she will generally assign feminine gender to the noun. After the age of nine, syntactic and semantic clues start to become more important. Evidence for this is the observation that children start to choose the syntactic or the semantic information over phonological information in inconsistent situations. une bicron, for example, would then be assigned feminine gender on the basis of the feminine determiner, irrespective of the masculine noun ending. In short: phonological gender clues are the most important clues to children younger than 9 years old, although syntactic and semantic clues do play some role in gender assignment. From the age of nine onwards, syntactic and semantic clues become more important, but phonological information is still

considered. Karmiloff-Smith’s results for French children have been replicated for different languages. Pérez-Pereira (1991 cited in Franceschina 2005) shows that the same results are obtained for children acquiring Spanish; Mills (1986 cited in Oliphant 1998) found that German children, although aware of semantic clues, also considered morphophonological clues to be more important in assigning gender to a noun, and Levy (1983 cited in Oliphant 1998) confirmed these findings for children acquiring Hebrew. The most striking neglect of semantic clues by young children comes from Koehn’s (1994) and Müller’s (1994, both cited in Oliphant 1998) studies, which show that two French-German bilingual children occasionally use the incorrect determiner 13 Source: http://www.doksinet with nouns that very obviously refer to male or female referents: le dame (the masc lady) and die Papa (the fem dad). Considering the results from these studies, various generalisations can be made

regarding the acquisition of gender in the L1 (see also Franceschina 2005). In younger children, morphophonological clues appear to carry more weight than semantic and syntactic clues. This does not mean that younger children do not pay attention to semantic and syntactic clues, however, as they can use these clues if the phonological clue is ambiguous (i.e if the noun ending is not specific to any gender, see discussion above on Karmiloff-Smith 1979). For older children, on the other hand, syntactic clues are generally the most important clue in gender assignment. 7 This difference is presumably due to the fact that younger children have not yet learnt which lexical items can be gender marked and hence must focus on the morphophonological realisation of the noun. 1.412 Acquisition of Gender in Italian, Dutch and English as an L1 As stated at the beginning of section 1.41, the developmental trajectory of gender in the L1 is dependent, to some extent at least, on the specific language

being learnt. In this section, I will thus look briefly at the acquisition of gender in Italian, Dutch and English monolingual children. The acquisition of gender is generally relatively unproblematic for normally developing children acquiring a Romance language and Italian is no exception to this rule (cf. Bottari et al 1993/1994, Bottari et al 1998 and Kupisch et al 2002) As section 1.31 shows, gender is overt and generally quite regular in Italian Once the child has realised that an –a ending on nouns denotes feminine gender and an –o ending denotes masculine gender, the gender of many nouns can be determined relying solely on morphophonological clues. Nouns ending in –e are potentially more problematic, but the gender-marked forms of the determiner, the adjective and the adverb generally clearly indicate what gender these nouns belong to. Once the child has acquired which items besides the noun are marked for gender and which forms belong to which gender, gender assignment

should thus be relatively easy. This is reflected in the pattern of normal acquisition in monolingual Italian children. Several months before their second birthday, they are already producing article-like phonetic forms (see Pizzuto and Caselli 1992, cited in Bottari et al. 1993/1994) and by 3 years old, Italian children use gender reliably in their speech. A very different picture is obtained if we consider the acquisition of gender in Dutch. Due to the lack of transparency in the Dutch gender system (see section 1.32), Dutch monolingual children have a significantly harder time at figuring out how gender is instantiated in their language. Determiner-noun combinations do occur relatively early in the child’s development (generally in the first half of the third year), but overgeneralisation of the common gender definite determiner occurs very frequently (cf. Zonneveld 1996 and Van der Velde 2003 and 2004, cited in Hulk and Cornips 7 The age difference between what has been termed

‘younger children’ and ‘older children’ depends, to some extent, on the particular language being acquired. In Karmiloff-Smith’s (1979) study on French gender acquisition, children from about 9 years old started to consider syntactic clues to be the most important. In Pérez-Pereira’s (1991, cited in Franceschina 2005) study on Spanish gender acquisition, on the other hand, there was not such a clear progression towards a greater weighting of syntactic clues. The terms ‘older children’ and ‘younger children’ are thus left vague intentionally. 14 Source: http://www.doksinet 2006). Not until children are about 6 years old do they stop overgeneralising the common gender definite article de in contexts in which the neuter definite determiner het is target-like. Interestingly, the overgeneralisation is always in one direction The common gender definite determiner de is thus used in contexts in which the neuter gender definite determiner het should be used, but het is

never used in contexts in which de should be used. Dutch children also have problems acquiring the adjectival inflection of the Dutch gender system (see Weerman et al. 2003), using the adjectival inflection for common gender nouns in cases where neuter gender adjectival inflection is required. Noun phrases such as (9) thus occur quite frequently in the speech of young Dutch children: (9) *een mooie huis a beautiful house neut a beautiful house (targetlike: een mooi huis) Only once children are about 7 years old are they fully target-like in this respect. In both determiner-noun combinations and in determiner-adjective-noun combinations, overgeneralisation of the common gender is thus a problem. This overgeneralisation strategy is understandable, however, as about 75% of Dutch nouns are common gender (and all plural nouns are also preceded by the definite determiner de). Setting common gender as the default when encountering novel nouns (or when the gender of a known noun cannot be

retrieved quickly enough) is thus a productive strategy that is apparently hard to abandon. The reasons why there is not such a clear default strategy in Italian children may be because the two genders are more or less equally divided over nouns (but the fact that the Italian gender system is more transparent than the Dutch system may also mean that Italian children are less in need of a default strategy). On the whole, then, it can be stated that the acquisition of gender is a longer and more laborious process for Dutch children than it is for Italian children. As English does not really have a grammatical gender system, not much can be said about the acquisition of gender in English. As gender only plays a role in very limited areas of English (i.e pronouns and anaphors, see section 133) and the rules are relatively simple to formulate (‘if the pronoun or anaphor refers to a man, use he/himself, if it refers to a woman use she/herself, use it/itself for anything else’) one might

expect children to acquire English gender very easily and early on in linguistic development. Interestingly though, it turns out that this is not the case According to Corbett (1991), English children do not reliably use the correct gender in these contexts until they are about 10 years old (although he notes that there is not much research on this topic). It is possible that this late acquisition of the correct use of gender in pronouns and anaphors is related to the fact that young children generally prefer morphophonological clues to semantic clues (see previous section), although 10 year olds are generally expected to be able to deal with semantic clues as well. More research thus has to be done on the matter, but, for Dutch and Italian at least, the generalisation seems to be that the more overt and transparent the gender system of a language is, the easier it is for the first language learner to acquire. 15 Source: http://www.doksinet 1.42 Processing of Gender in the L1 In

the previous section, the acquisition of gender in the L1 was discussed. That section was primarily concerned with how children acquire knowledge of gender or, in other words, how children acquire what the gender of each noun is, which items other than nouns are gender-marked and in which contexts gender-marking is required. This knowledge is a key element in the process of mastering the gender system of a language. After all, if the child never learns what gender a certain noun has, all further stages in fully mastering the gender system will fail as well. In this section, another key element in the acquisition and use of gender is discussed: the processing of gender in the L1. Only if a child (or an adult) can process all the required knowledge fast enough (i.e realising in each context what the gender of a noun is and which form the other gender-marked items have to take) in real time, will normal use of the gender system be possible. If the processing of gender is hindered (due to

problems in retrieving the gender of a noun fast enough, for example), abnormal usage of gender will be the result, even if there is nothing wrong with the underlying knowledge of gender. In this overview of processing of gender, there are two subsections: processing in gender assignment and processing in gender agreement. In the gender assignment subsection, the role of gender assignment in the lexical retrieval of nouns will be discussed. This section is thus concerned with the processing of gender for nouns in isolation. The gender agreement subsection, on the other hand, looks at the processing of gender agreement, and hence, at the processing of the noun in combination with other gender-marked items. 1.421 Processing in Gender Assignment In the literature, there is some debate as to how gender assignment is processed (see Schriefers and Jescheniak 1999 and Hohlfeld 2006 for a review). Two main routes for gender assignment have been proposed: the lexical route and the form-based

route. 8 The assumption underlying the lexical route of gender assignment is that gender information is stored at a different level in the mental lexicon (the lemma level) than phonological information (which is presumed to be stored at the lexeme level). Following this model of gender assignment, it is claimed that gender information is retrieved earlier than phonological information. This means that information regarding the noun’s ending cannot be used in assigning gender to a noun (as the phonological information that would give a clue to gender is accessed after the valued syntactic gender feature has been retrieved). One piece of evidence in favour of this hypothesis comes from the so-called ‘tip of the tongue’ phenomenon. In a ‘tip of the tongue state’, the speaker is incapable of retrieving the exact form of the word that they want to use, but they are generally capable of retrieving certain features of the word. Vigliocco et al (1997) show that 84% of Italian

speakers who had been brought into a ‘tip of the tongue state’ were capable of retrieving the gender of the word they were looking for. This means that phonological information regarding the word’s ending cannot be the sole way of determining a word’s gender (after all, the word’s ending is not accessible to these speakers, but its gender is). 8 The proponents of these two different routes each claim that their route is the route in gender assignment. It is not at all improbable, however, that the two routes are not mutually exclusive This is also the conclusion Hohlfeld (2006) draws regarding her own data. 16 Source: http://www.doksinet According to the form-based route, on the other hand, gender assignment is established primarily by considering the phonological information of the word. One version of the form-based route is called the ‘reliable cue hypothesis’ by Gollan and Frost (2001, cited in Hohlfeld 2006), which claims that nouns that are reliably marked for

gender will be processed more quickly than nouns that are not clearly gendermarked. Evidence in favour of this hypothesis comes from a study by Desrochers and Paivio (1990 cited in Hohlfeld 2006). In their gender identification task participants were shown to be quicker in identifying the gender of a particular noun if it contained a reliable gender clue than when this was not the case. Another instantiation of the form-based route is termed the ‘postlexical checking hypothesis’ as advanced by Bates, Devescovi and Pizzaniglio (1995) and Bates, Devescovi, Hernandez and Pizzaniglio (1996 both cited in Hohlfeld 2006). According to this hypothesis, gender assignment consists of two stages. In the first stage, gender is retrieved from the mental lexicon; in the second stage the information from the lexicon is compared to the phonological information regarding the noun’s ending (which is termed ‘postlexical checking’). If the two types of information converge, gender assignment is

facilitated; if the two types of information are incongruent, gender assignment is predicted to be inhibited. This hypothesis would thus also predict that nouns with irregular endings take longer to process (because the phonological information is not conclusive with respect to gender) than nouns with regular endings. Evidence for this hypothesis comes from a gender priming study by Bates et al. (1996 cited in Hohlfeld 2006). In this study, participants were shown to react faster to Italian nouns that were gender-marked by regular endings (i.e –o or –a, see section 131) than when they were marked by irregular endings. It may not be necessary to choose between a lexical route and a form-based route account of gender assignment, however, as there is evidence that both routes can be used depending on the nature of task and the language under investigation. The data from Hohlfeld (2006), for example, show that the nature of the task can influence which route is chosen. In Hohlfeld’s

first experiment, native German speakers had to determine the gender of various transparently and opaquely gender-marked bare nouns. Hohlfeld’s materials consisted of transparent masculine nouns ending in –ling (Häftling, prisoner), -er (Leuchter, candlestick) or –eur (Friseur, hairdresser), transparent feminine nouns ending in –ung (Festung, fortress), -heit (Weisheit, wisdom) or –ei (Malerei, painting) and transparent neuter nouns ending in –chen (Päckchen, parcel), -nis (Hindernis, hindrance) or starting with ge- (Gebäck, pastries). The non-transparent nouns had various different endings that are not generally related to any particular gender (like Unterschied, difference and Tabak, tobacco). The results of this experiment showed that the transparency of the noun made no difference to the speed with which the gender was determined. (Taraban and Kempe 1999 show similar results for native Russian speakers.) In this task then, the participants seem to be using the

lexical route for gender assignment (as the noun endings do not seem to influence the response times). In Hohlfeld’s second experiment, participants were required to assign gender to transparent and opaque words and nonwords (Zenkling, for example, is a transparent nonword, Podun is an opaque nonword). In this task, as in the first experiment, the participants did not show a difference in processing between the transparent and opaque words. They did base their answers on reliable gender cues for the nonwords, however, indicating that a form-based route can be induced if there is no entry in the 17 Source: http://www.doksinet mental lexicon, that is, if the lexical route is impeded. This finding also makes sense from the point of view of the acquisition data discussed in the previous section. After all, young children do not have an entry in the mental lexicon for nouns that they have never encountered before, so they tend to focus on phonological clues more than on other clues

(i.e they use the form-based route) Older children, on the other hand, who do have ready access to many nouns in the mental lexicon use syntactic clues more than other clues (i.e they use the lexical route) 1.422 Processing in Gender Agreement In the previous section, the processing of gender on bare nouns was discussed. In this section, the processing of gender in the DP as a whole will be considered. Two issues regarding gender agreement processing that are debated in the literature will be discussed. The first subsection looks at whether semantic information plays a role in processing gender agreement or whether gender agreement is a purely syntactic process. The second subsection considers to what extent the gender marking on the determiner and the adjective help or hinder in the processing of gender on the noun. 1.4221 Gender Agreement: the Role of Semantic Information Is gender agreement a purely syntactic process or does semantic/conceptual information enter the agreement

procedure as well? On the one hand, grammatical gender seems to be largely unrelated to semantic gender. As stated in section 1, in many languages the gender of most nouns has absolutely nothing to do with natural gender, so from that point of view it would not be expected that semantic information plays a large role in the gender agreement process. On the other hand, we see that all formal systems are based, to some extent at least, on natural gender for a certain number of nouns. Given this fact then, it might not be so strange if semantic information does play some role in the gender assignment process, at least for nouns that refer to something with an obvious natural gender. In the literature, the evidence for the role of semantic information in the processing of gender agreement is somewhat mixed. In an ERP study by Hagoort and Brown (1999), Dutch subjects were shown sentences with congruent and incongruent determiner-noun combinations (i.e combinations in which the determiner

and the noun had the same gender and differing gender, respectively). The results of this study showed clearly that gender agreement incongruencies lead to so-called P600/SPS effects. These effects are well documented in the neurolinguistic literature and occur when subjects are processing syntactic (as opposed to semantic) anomalies (cf. Osterhout and Holcomb 1992, Hagoort et al 1993, both cited in Sabourin and Haverkort 2003, and Hagoort et al. 1999, cited in Hagoort and Brown 1999) Another study on the processing of gender in German speakers by Gunter et al. (2000) yielded similar results. In this study, participants were shown sentences with DPs with a high cloze probability (i.e a high likelihood of following the verb) and a low cloze probability (a low likelihood of following the verb) in which the determiner and the noun were either congruent or incongruent. Example sentences can be found in Table 3: 18 Source: http://www.doksinet Table 3: Example sentences Gunter et al.

(2000) Cloze probability High Gender congruent High incongruent Low congruent Low incongruent Sentence Sie bereist das Land She travels the neut land neut Sie bereist den Land She travels the masc land neut Sie befährt das Land She drives the neut land neut Sie befährt den Land She drives the neut land neut At the early stages of processing, Gunter et al. (2000) found two different effects that occurred separately (i.e they did not interact): a so-called LAN effect (left-anterior negativity, which occurs when syntactic anomalies are processed) and a N400 effect (which occurs when semantic anomalies are processed). The LAN effect was only present when the gender of the noun and the determiner were incongruent and the N400 only occurred when the DP was an unlikely continuation of the sentence (i.e in the low cloze sentences). This finding was taken to mean that both semantic and syntactic processes operate separately at early stages of processing. At later levels of

processing, semantic and syntactic processes do interact, however, as can be seen by the nature of the P600 effect that was observed in this experiment. The P600 effect was present in all cases in which there was gender disagreement, but this effect was diminished in cases where the semantic expectancy was low (i.e in the low cloze sentences). The authors conclude that this indicates that at later processing stages semantic processes do influence the way in which syntactic anomalies are processed. Vigliocco and Franck (1999) also show that at some point in the process of gender agreement, semantic information about the natural gender of the noun enters the procedure. In their study, French and Italian speakers had to complete a sentence with the masculine or the feminine form of a prescribed adjective. The test sentences consisted of nouns with either conceptual or grammatical gender (so-called ‘head nouns’) followed by an incongruent noun (so-called ‘local nouns’). An example

of a test sentence with a head noun with conceptual gender followed by an incongruent local noun can be seen in (10): (10) lo sposo the masc groom masc the groom in the church in in chiesa church fem In this example, the head noun is thus conceptually masculine (sposo) and it is followed by a feminine local noun (chiesa). An example of a test sentence with a head noun with grammatical gender followed by an incongruent local noun can be seen in (11): (11) il cero the masc candle masc the candle in the church in in 19 chiesa church fem Source: http://www.doksinet In (11), the head noun (cero) is thus grammatically, but not conceptually, masculine and again, it is followed by the feminine local noun, chiesa. Before these two sentences were shown, the subjects were given the masculine and the feminine form of a certain adjective, brutto and brutta (ugly) for example, which they had to use to complete the sentence. A correct completion of the sentences in (10) and (11) would thus

be ‘lo sposo/il cero in chiesa e’ brutto’ (the groom/the candle in the church is ugly); an incorrect completion would be ‘lo sposo/il cero in chiesa e’ brutta’. It turns out that agreement errors are more common when the head noun has grammatical gender than when it has conceptual gender. Subjects were thus more likely to use the incorrect adjective brutta following a head noun with masculine grammatical gender (like cero in (11)) than following a head noun with masculine conceptual gender (like sposo in (10)). According to Vigliocco and Franck, this means that semantic information, and not just syntactic information, is processed during gender agreement. At the current state of knowledge, it thus seems like gender agreement is a purely syntactic process in the early stages of processing, but that semantic information can enter the procedure at a later point in the derivation (cf. Vigliocco and Frank 1999, Hagoort and Brown 1999 and Gunter et al. 2000) More research is

necessary, however, to determine exactly when and how semantic information influences gender agreement. 1.4222 Gender Agreement: The Gender Congruency Effect The previous section considered the effect of semantic information on gender processing; this section will consider the effect that gender-marked items other than nouns have on gender processing. Various studies have found that congruent gender marking speeds up and incongruent marking inhibits the processing of a following noun in native speakers (see Gurjanov et al. 1985 for this effect in Serbo-Croatian, Garnham et al. 1995 for Spanish, Van Berkum 1997 for Dutch, Jacobsen 1999 for German and Guillelmon and Grosjean 2001 for French). In example (12), the noun treno (train) is thus processed more quickly due to the presence of the correct masculine determiner il than in example (13), where the process is inhibited due to the presence of the incorrect feminine determiner la: (12) il the masc the train treno train masc *la the

fem the train treno train masc (13) This effect is quite robust for L1 speakers in both reading and spoken word recognition. The presence of gender agreement in a language thus seems to help speakers to process sentences (or at least those parts of sentences that contain gender agreement) quickly and accurately. 20 Source: http://www.doksinet 1.5 Acquisition and Processing of Gender in L2 In this section, a general overview of the acquisition and processing of gender in a second language will be given. The differences between the acquisition and processing of gender in the L1 and the L2 will be discussed, as will the potential causes of these differences. The two main theories of second language acquisition that make specific contrasting predictions regarding the acquisition and processing of gender will be related to this debate. 1.51 Acquisition of Gender in the L2 The most striking thing about the acquisition of gender in the L2 is that it tends to remain problematic for many

learners, even for those at high levels of proficiency (cf. Holmes and Dejean de la Bâtie 1999, Bartning 2000, Bruhn de Garavito and White 2000, Dewaele and Véronique 2001, Weerman et al. 2003, Hawkins and Franceschina 2004 and Hulk and Cornips 2006 9), although there are some studies that find nativelike gender proficiency for some groups of learners (cf. White et al 2001 and White et al. 2004) This problematic development and incomplete mastery of grammatical gender in the L2 thus sharply contrasts with the relatively error-free manner in which gender is fully acquired in the L1. Apart from these substantial differences, there are, however, some similarities in the way that first and second language learners acquire gender. Both first and second language learners find definite determiner-noun combinations easier to acquire than indefinite determiner-noun combinations, which in turn are more easily acquired than noun-adjective or adjective-noun combinations (cf. Bartning 2000, Bruhn

de Garavito and White 2000 and Dewaele and Véronique 2001). In the next two subsections, two factors that may influence the path of L2 gender development will be discussed. In the first subsection, the role of formal instruction in the acquisition of gender will be considered; in the second subsection, an overview will be given of how differences in proficiency relate to accuracy in the use of L2 gender. In the literature, it has also been suggested that the gender characteristics of the L1 (i.e whether or not the L1 has gender) play a role in a learner’s ability to acquire grammatical gender in an L2. As the nature of the L1 potentially plays a role both in acquisition and in processing, the discussion of this factor is dealt with in section 1.53, after an overview of gender processing in the L2 has been given in section 1.52 1.511 Formal Instruction It is possible that formal instruction regarding the nature of the L2 gender system may help the learner in assigning the correct

gender to nouns and in using correct gender agreement. Especially in form-oriented types of instruction, emphasis is laid on the rote learning of the gender of nouns and on learning the syntactic contexts in which gender agreement is required. The question is, however, to what extent this kind of overt formal instruction has an effect on the learner’s underlying knowledge of the gender system. Most studies have not found a great effect of formal instruction on successful acquisition of gender. In Dewaele and Véronique’s (2001) study, for example, the gender assignment and agreement of two groups of 18 to 21 year old 9 Strictly speaking, Hulk and Cornips’ (2006) study looks at children that have been exposed to two languages from birth (2L1). They are thus not the pure second language learners that are considered in the other studies. As their findings are also interesting from the point of view of L2 acquisition of gender, the study is mentioned here and will be referred to in

later sections. 21 Source: http://www.doksinet Dutch learners of French were compared. The first group had chosen French as their second language at secondary school and had received formal instruction in French for five hours a week over six years; the second group had chosen French as their third language (English was their second language) and had received a more limited input of three hours a week over four years. Even though the amount of formal instruction was thus quite different for the two groups, the results of this study showed that there was no correlation between the amount of formal instruction and the accuracy rates for gender assignment or agreement. According to Harley (1998 cited in Franceschina 2005), even for young learners, the accuracy rate for gender assignment and agreement is not significantly affected by the amount of formal teaching. In her study of 7 and 8 year olds in early French immersion programmes, she found that formal instruction was better than

no instruction, but that the children were not capable of extending their knowledge of gender to new words. Formal instruction thus only seemed to be useful for learning the gender of individual words, but not for learning the gender system as a whole. As neither the amount of formal instruction nor the age at which formal instruction takes place seem to correlate with success in the acquisition of gender, the conclusion seems to be that formal instruction does not play an important role in explaining individual differences in the correct use of gender. 1.512 Level of Proficiency At first sight, it may seem a bit strange to consider the level of proficiency of the L2 learner as a factor in explaining individual differences. After all, the more proficient the learner is, the more accurate they are expected to be on tasks of gender assignment and gender agreement. Whilst this is true, it is still interesting to note that there is some debate in the literature on whether it is possible

for second language learners to become indistinguishable from native speakers in their use of grammatical gender. According to some researchers (cf. White et al 2001 and White et al 2004), even intermediate learners can be indistinguishable from native speaker controls, whereas other researchers (cf. Franceschina 2005) claim that otherwise highly proficient learners can be non-nativelike in their use of gender. In White et al. (2001), low, intermediate and advanced proficiency English and French learners of Spanish are compared to native Spanish speakers on a gender comprehension test. In Spanish, nouns can be omitted, as can be seen in (14): (14) un negro está encima a masc black masc is on-top there is a black one on the table de of la the mesa table It is thus possible to drop the noun after negro if it is clear in the context which black thing is being referred to. Note that the gender of the dropped noun is visible from the form of the determiner and the adjective. In (14),

the dropped noun thus has to have masculine gender, as the determiner and the adjective are both marked for masculine gender. In White et al’s test, this characteristic of Spanish is used to test participants’ sensitivity to gender. Participants were given a sentence with a null nominal accompanied by three pictures. One of the pictures was the target picture; the gender 22 Source: http://www.doksinet of the object in this picture was thus congruent with the gender of the null nominal. One of the other pictures was the non-target picture; the gender of the object in this picture was thus opposite to the gender of the null nominal. The third picture was the distractor picture; this object was different in number from the null nominal. White et al. hypothesised that if the participants were sensitive to gender, they would choose the picture with the object that has the same gender as the null nominal. It turned out that French learners at all three levels of proficiency were

indistinguishable from native controls. The English speakers, on the other hand, were comparable to controls in the intermediate and advanced proficiency levels, but not at the low proficiency level. White et al. (2004) tested the acquisition of Spanish in intermediate and advanced French L1 and English L1 learners on a similar comprehension test and a production test. This study also found that intermediate and especially advanced learners of Spanish in both languages performed very similarly to native controls. The conclusion of both of these studies is thus that overall proficiency level is the key factor in explaining individual differences in the acquisition of gender. To start with, learners have problems in retrieving the noun’s gender and in choosing the correct morphophonological form for the other gender-marked items, but as learners are exposed to and use the second language more (and hence become more proficient), the retrieval of the noun’s gender and gender agreement

can proceed much more automatically. 10 In contrast to the two studies by White et al. discussed above, the studies in Franceschina (2005) and Sabourin et al. (2006) find that proficiency is not a clear predictor of correct gender use. Both studies look at the use and comprehension of grammatical gender in highly proficient L2 learners and conclude that nativelike attainment is not found in all learners. A more extensive discussion of these studies will be delayed until section 1.53 (because both of these studies consider the role of the L1 to be the key determinant of successful gender use), but for now it suffices to say that there is some evidence that high proficiency does not necessarily entail nativelike gender use. To summarise: L2 gender acquisition is quite dissimilar from L1 gender acquisition in various respects. In contrast to the acquisition of gender in an L1, the acquisition of gender in an L2 is often a laborious process that does not always lead to complete mastery of

the system. There is some debate on which factors influence the capacity to fully master the gender system of a second language. From the available studies, it seems clear that the amount of formal instruction is not a relevant factor in this issue, but the status of other potentially explanatory factors is less clear. Proficiency must play some role in the level of a learner’s gender system, as low proficiency learners are generally significantly less accurate in gender assignment and agreement than 10 A similar idea is developed in Hulk and Cornips (2006). They claim that the acquisition of gender is problematic because it is a so-called interface phenomenon (cf. Tsimpli et al (2004), Serratrice et al (2004) and Sorace (2005) for more information on interface phenomena in acquisition). The idea lying behind this claim is that any linguistic phenomenon that requires the integration of various sources of information is acquired more slowly than linguistic phenomena for which this

integration of information is not necessary. According to Hulk and Cornips (2006), then, gender is a prime candidate for acquisitional problems because it requires integration of lexical, morphophonological and syntactic information. 23 Source: http://www.doksinet more advanced learners. Franceschina (2005) and Sabourin et al (2006), on the other hand, show that high proficiency in a language does not necessarily entail nativelike sensitivity to gender. As stated above, a full discussion of these two studies and the role of the L1 in the acquisition of L2 gender will be given in section 1.53, after gender processing in the L2 is discussed in the following section. 1.52 Processing of Gender in the L2 The previous section has shown that the acquisition of gender in the L1 is quite dissimilar from the acquisition of gender in the L2. In this section, it will be shown that the processing of gender in the L1 and the L2 is also different, even at high levels of proficiency. Sabourin and

Haverkort (2003), for example, show that advanced German L1 learners of Dutch do not process gender in the same way as native Dutch speakers. In the first experiment of their study, the German learners were required to assess the grammaticality of various sentences with correct and incorrect definite determiners and adjectival agreement. In this off-line task, the German learners showed a nativelike knowledge of gender in the determiner-noun combinations and a relatively poor performance in the adjectival agreement cases. The conclusion for this off-line task thus seems to be that the German learners are nativelike in their gender assignment, at least in the case of determiner-noun combinations. The second experiment of this study shows that this is a premature conclusion. In this experiment, an ERP version of the grammaticality judgement task in the first experiment was used. The ERP output of this task clearly shows that even though the German learners may be behaviourally very

similar to the native speakers, they do not process gender in the same way. For the gender agreement errors in the definite determiner-noun combinations (in which the learners displayed nativelike knowledge), for example, a clear P600 effect was obtained for native speakers (a response to syntactic anomalies, see section 1.422) A P600 effect was also visible for the German learners in this condition, but it was qualitatively different to the one for the native speakers in that it was much more restricted and started later than the one found in the native speakers. According to Sabourin and Haverkort (2003), this finding indicates that even though second language learners may be behaving very similarly to native speakers in their use and comprehension of gender, that is, in their knowledge of gender, they are not necessarily processing gender in the same way. A difference in the L1 and L2 processing of gender is also found in Taraban and Kempe (1999). Native speakers of Russian and

advanced L2 learners were asked to read Russian sentences and to choose one of two verbs to complete the sentence they had read. In Russian, gender is also marked on the verb, so of the two verb options, one verb had the correct gender marking for the subject NP and the other verb had incorrect gender marking. In the test sentences, the subject noun was either transparent or opaque with respect to gender and occurred either with or without a transparently gender-marked adjective. The results of the study showed that the L2 learners were generally more error-prone than the native speakers in choosing the correct verb form, but this effect was especially clear in the case of opaque nouns. For transparent nouns, the reading times were more or less comparable between native speakers and non-native speakers. In the case of opaque nouns, on the other hand, the native speakers were significantly faster than the non-natives, although this effect was diminished if a gender-marked adjective was

present in the subject NP. This suggests that, unlike the L1 subjects, the L2 learners did not have immediate access to the gender feature of the noun when retrieving the noun from the mental lexicon and that 24 Source: http://www.doksinet they rely on other gender-marked items in retrieving an opaque noun’s gender. The processing of gender in these L2 speakers thus relies to a greater extent on the disambiguating effect of other gender-marked items than it does for L1 speakers, who can rely purely on the gender information present in the lexical entry of the noun. The final study to be discussed in this section is Guillelmon and Grosjean (2001). In this study, the gender congruency effect (see section 1.4222) was explored for native and very proficient English speakers of French. The participants in this study were presented aurally with a determiner-adjective-noun sequence in which the gender of the determiner and the adjective could be congruent (as in (15)) or incongruent (as

in (16)) with the gender of the noun. (15) la the fem jolie nice fem glace icecream fem the nice ice-cream (16) *le the masc joli 11 nice masc glace icecream fem the nice ice-cream Once the DP had been heard, the participants were required to repeat the noun as quickly as possible. The monolingual French speakers showed a clear gender congruency effect: their repetition of the noun was facilitated if the determiner and the adjective preceding it had the same gender; their repetition was inhibited if the determiner and the adjective had a different gender than the noun. The non-native group, on the other hand, showed absolutely no congruency effect. Their reaction times were thus not affected either positively or negatively by the gender marking on the determiner and the adjective. Taken together, these three studies show that even highly proficient learners may not be processing gender in the same way as native speakers. Thus, whilst proficient second language learners will

generally show that they have the knowledge of the gender system of an L2 (i.e they know which gender to assign to which noun and what the gender agreement rules are), their problem may lie in processing this knowledge. 1.53 The L1 in L2 Acquisition and Processing of Gender The influence of the L1 in second language acquisition is heavily debated in the literature. In this debate, two positions are generally taken: the L1 plays either a major role or a minor role in the acquisition of the second language. The first position, that the L1 plays a major role, is held by the Failed Functional Features Hypothesis (FFFH) of second language acquisition proposed by Hawkins and Chan (1997 cited in White 2003). The second position, that the L1 plays a minor role, is held by the Full 11 Note that the stimuli were presented aurally to the subjects and that the acoustic characteristics of joli and jolie are the same. The form of the adjective thus gives the subject no clues to the gender of the

noun. 25 Source: http://www.doksinet Transfer Full Access hypothesis (FTFA) proposed by Schwartz and Sprouse (1994, 1996 cited in White 2003). These two theories of second language acquisition are not specific to gender, nor do they make any specific statements on the processing of language in general or gender in particular. Nonetheless, given the basic assumptions underlying these theories, it is possible to extrapolate some clear predictions regarding the acquisition and processing of gender in learners from different L1 backgrounds. 1.531 The L1 as a Major Factor in L2 Gender 1.5311 Franceschina (2005) An important study on the influence of the L1 in gender acquisition is Franceschina (2005). In her study, Franceschina looks at the production and comprehension of grammatical gender in Spanish in 53 learners with an L1 with gender (+gen L1) and 15 learners with an L1 without gender (-gen L1). 12 All of Franceschina’s subjects had received multiple years of naturalistic

exposure to Spanish and scored within the native range of the University of Wisconsin Spanish Placement test; the subjects tested in this study were thus highly proficient speakers of Spanish. Franceschina then conducted six off-line tests on these subjects, requiring written, oral and metalinguistic production and written and oral comprehension. Here, the results of Franceschina’s first and fourth test will be discussed as together, they require the subjects to use various different types of comprehension and production. Franceschina’s first test is called the ‘guessing game’, which was intended to determine whether the non-natives were capable of using gender agreement clues in sentence comprehension in the same way that native speakers can. In this test, subjects simultaneously heard and read sentences that referred to certain objects or concepts indirectly. The subjects’ task was to guess which object or concept was being referred to. An example sentence is given in (17):

(17) Los trajo Martín y dijo que son para them masc brought Martín and said that are for Martín brought them and said that they were for you usted you In this example, the subjects have to choose what los (them masc ) refers to from the following possibilities: flores (flowers fem ), joyas (jewels fem ) or chocolates (chocolates masc ). As los is masculine, the only possible answer here is chocolates, as the two other possibilities are feminine nouns. By choosing the correct noun, the subjects thus demonstrate that they can use gender clues to comprehend sentences. The results of this test show that there is a significant difference between the native speakers and the –gen L1 group, but no difference between the native speakers and the +gen L1 group. The +gen L1 group is thus apparently capable of using gender clues in their comprehension of sentences in the same way that native speakers can, whereas the –gen L1 group are not as good as the natives at using such clues.

Franceschina’s fourth test is a grammaticality judgement task that was designed to find out whether subjects were sensitive to gender mistakes. The subjects simultaneously heard and read sentences and had to indicate whether or not they were grammatical. Once they had identified an ungrammatical sentence, they were asked to 12 The +gen L1s are: Arabic, French, German, Greek, Italian and Portuguese; the –gen L1 was English for all participants. 26 Source: http://www.doksinet correct the mistake. Subjects were given sentences with grammatical gender in various structural contexts, so that a general picture of the subjects’ understanding of Spanish gender could be obtained. In (18)-(20) some (grammatical) examples of the possible contexts are given: (18) Det-N-Adj la the fem cultura culture fe catalana Catalan fem m the Catalan culture (19) Det-Adj-N un nuevo a masc new masc a new scheme esquema scheme masc (20) pronoun/adjective me lo dieron to-me it masc they-gave It

was given to me open abierto open masc Again, the results of this test show that whilst there was a significant difference between the performance of the –gen L1 group and the native speakers, there was no difference between the +gen L1 group and the native speakers. 13 In contrast to the +gen L1 group, the –gen L1 group thus do not show nativelike sensitivity to gender mistakes in the input. The results of the remaining four tests showed the same trend: although the performance of the –gen L1 group is good, it is consistently significantly different from the performance of the native speakers, whereas the performance of the +gen L1 group is indistinguishable from that of the native speakers. Franceschina thus concludes that the nature of the L1 is the decisive factor in whether or not learners will be able to acquire nativelike mastery of an L2 gender system. If the L1 has gender, nativelike attainment will be possible, if the L1 does not have gender, nativelike attainment

will be impossible to achieve. Franceschina claims that these results are evidence for the validity of the FFFH account of second language acquisition. According to the FFFH, adult L2 learners are incapable of acquiring features that are not found in their L1. The idea behind this assumption is that Universal Grammar makes all the functional features that determine parametric differences between languages available for a limited amount of time. Once the critical period for the acquisition of functional features has passed (as would be the case for adult second language learners), it is not possible to acquire new functional features. This assumption has clear implications for the acquisition of gender. As gender is a functional feature (see section 12), gender features can only be acquired during this so-called ‘critical period’, according to the FFFH. The first language learner thus has no problems in acquiring the gender system of their 13 Franceschina does not specify in which

elements of the test the –gen L1 group differs from the +gen L1 and the native group. She gives each group a general score in which the subjects’ accuracy in rating the sentences as grammatical and ungrammatical and the accuracy of the corrections in the ungrammatical cases is combined, but the scoring criteria are not defined. It is thus not possible to see in which areas of the test the –gen L1 group performed worse than the other two groups. 27 Source: http://www.doksinet language, but the post-puberty second language learner is ‘too late’ to be able to acquire the gender feature if this feature is not present in her L1. By showing that the +gen L1 group, but not the –gen L1 group, perform within the native range, Franceschina’s study (2005) thus does indeed seem to fall very neatly in line with the FFFH account of second language acquisition. However, before Franceschina’s conclusions can be firmly accepted, a few problems with her study should be considered. In

the first place, if we consider the FFFH’s views on the acquisition of grammatical gender in –gen L1 learners, one might expect that their use of grammatical gender would be rather random. If, as is claimed by Franceschina and the FFFH, they have no conception of what a gender feature is, they might be expected to use one and only one gender all the time or they might be completely random in their gender assignment, using feminine gender markers and gender agreement one time and masculine markers and agreement the other. This is not what is observed, however, as the –gen L1 group in Franceschina (2005) do perform quite successfully on the various tasks 14, just not as well as the +gen L1 group. How are the –gen L1 group capable of this performance if it is biologically impossible for them to acquire a gender feature? To answer this question, Franceschina and the other proponents of the FFFH view of L1 influence would have to reply that these learners are using compensatory

strategies that enable them to use the L2 gender system more or less correctly. The nature of these strategies is not clear, as neither Franceschina nor other proponents of the FFFH say very much about how it is possible that the –gen L1 learners are still capable of performing quite successfully on L2 gender. Maybe the successful –gen L1 learners are able to create strong memory links between the L2 nouns and their gender. If this is so, it is possible that a –gen L1 learner’s general memory capacities are a good measure of their capacity to acquire L2 gender. It may also be possible that the –gen L1 learners use other agreement relations in the L1 to model gender agreement in the L2 on. An English learner might then be able to model Spanish gender agreement on English subject-verb agreement, for example, which may result in imperfect but relatively successful gender agreement. Or maybe the English learners are capable of extending the gender agreement necessary for pronouns

and anaphors in English to the Spanish forms of gender agreement. This hypothesis could be considered by looking at the L2 gender performance of –gen L1 learners who do not have any agreement in the L1 (or at least no gender agreement whatsoever between any elements of the language). At the moment, however, this is mere speculation, so it is not clear how the FFFH can deal with the relatively successful performance of high proficiency –gen L1 learners on L2 gender. Another point to bear in mind with respect to Franceschina’s (2005) study is that only 15 subjects were tested in the –gen L1 group (all of whom had English as the L1) compared to 53 learners in the +gen L1 group (who came from different L1 backgrounds). This difference in number and in population (members from one L1 are compared to members of a number of different L1s) may mean that the conclusions are not valid. After all, as all the –gen L1 subjects had English as their L1, it is equally possible that it is

something specific to English (other than its –gen nature) that caused the –gen L1 group’s problems. It could also be the case that 15 subjects 14 In the various tasks presented to the subjects, the –gen L1 group have accuracy rates over 81% (compared to accuracy rates over 92% for the +gen L1 group and 95% for the native group). 28 Source: http://www.doksinet are not enough to cancel out individual differences in the performance. Maybe a few of the subjects had a very bad performance and as there were only 15 subjects in total, this dragged down the performance of the whole group (whereas a few bad performances in a group of 53 learners would have a less drastic effect on the performance of the whole group). This point also relates to another problem with Franceschina’s study: Franceschina does not report any individual data. As the assumption underlying the FFFH is that it is impossible for any –gen L1 learner to fully acquire the gender system of a language, it is

crucial to know whether any member of the –gen L1 group performed completely in the native range for all tests. As this data is not given, it is impossible to assess whether it is truly the case that no –gen L1 learner can acquire gender. To summarise: Franceschina’s study (2005) shows that +gen L1 learners of Spanish are consistently nativelike in their use and comprehension of the Spanish gender system, whereas the –gen L1 group consistently performs slightly worse than the native speakers. Franceschina thus concludes, following the FFFH, that the L1 is a very important, even crucial, factor in the L2 acquisition of gender. The validity of this conclusion is, however, undermined by the various problems in the study. 1.5312 Sabourin et al (2006) Another study that shows that the nature of the L1 is an important factor in the acquisition of L2 gender comes from Sabourin et al. (2006) In this study, Sabourin et al. compare the knowledge of Dutch grammatical gender in three

groups of advanced learners: L1 German, English and Romance subjects. The idea underlying this subject selection is that the three languages have different grammatical gender characteristics. The gender system of German is very similar to the Dutch system; even though German has three genders and Dutch only two, generally German feminine and masculine nouns are common gender in Dutch and German neuter nouns are also neuter in Dutch. Like Dutch, the German gender system is also opaque, as the gender of the noun is generally not clear from the morphophonological form of the noun itself (although there are more morphological cues in German than in Dutch). Romance languages also have a gender system, and hence are comparable to Dutch in that respect, but the two gender systems are not very similar. Romance languages have masculine and feminine gender instead of common and neuter gender, so the gender feature of a noun in Romance is no help in determining which gender the Dutch equivalent

will have. Another difference is that Romance languages often have transparent gender systems, whereas Dutch has an opaque gender system. English has no grammatical gender system and thus is the furthest removed from Dutch with respect to grammatical gender. The gender system of these three groups of learners was tested by looking at their gender assignment and gender agreement capacities. All groups were relatively good at assigning gender to nouns, although there was a clear hierarchy in performance: the German group performed best, followed by the Romance group, followed by the English group. Gender agreement proved to be more problematic for all groups, but the hierarchy in performance was still visible: again, the German group performed best, followed by the Romance group, followed by the English group. The results of this study thus show that the nature of the L1 does influence a learner’s capacity to acquire gender in an L2. 29 Source: http://www.doksinet Sabourin et al.

(2006) relate the difference in the performance of the three groups to the extent to which gender can be transferred from the L1 to Dutch. In the case of German, so-called ‘surface transfer’ is claimed to be possible: because the two systems are quite similar, surface features may be transferred from one language to the other. 15 An outright translation strategy may thus explain the German group’s good performance on the gender assignment test. Gender agreement was difficult for all groups, but because German learners are used to an opaque gender system (as in Dutch), their L1 gender agreement processing may help them out to some extent. The Romance learners, on the other hand, will not be able to use surface transfer, because the two gender systems are too different. So-called ‘deep transfer’ (ie the transfer of abstract features from one language to another) may be possible, however, as the abstract gender feature present in Romance languages could be transferred to Dutch.

The English group performed worst then, because they could use neither surface transfer nor deep transfer, as English does not have a gender feature to transfer. So, if we assume that surface transfer and deep transfer is more useful to a learner than deep transfer on its own, it is possible to account for the hierarchy. The German group perform best because they can use surface transfer and deep transfer, the Romance group perform moderately well because they can use deep transfer and the English group perform worst because they can use neither surface transfer nor deep transfer. Like Franceschina (2005), the study by Sabourin et al. (2006) would seem to uphold the FFFH account of L2 gender acquisition. There are, however, some differences between Sabourin et al. and Franceschina’s findings Contrary to Franceschina’s (2005) study, even the best learners in Sabourin et al.’s study (ie the German group) still perform significantly worse than the native Dutch speakers on gender

agreement 16. The difference between the three experimental groups is thus not that the two groups with gender (i.e the German and the Romance group) perform as the native speakers do and the group without gender (i.e the English group) performs worse than the three other groups; all three non-native groups perform significantly worse than the native group. For all three groups of highly proficient learners then, Dutch gender remained a persistent problem, contra to what was found in Franceschina (2005) and what might be expected following the FFFH. It should be noted here that although the FFFH does not make specific claims about the state of L2 ultimate attainment, it follows from FFFH basic assumptions that highly proficient learners from a +gen L1 background should perform very similarly to native speakers. After all, as these learners already have a gender feature in their L1 which they can transfer to the L2, they do not have to acquire anything fundamentally new; they only have

to acquire the specific gender-values for the nouns in the L2. If these learners have had enough exposure to the L2 and hence have had enough time to acquire the gender of each noun, there is no reason why they should perform any worse on tasks of gender assignment and agreement than native speakers. 15 It is not entirely clear in Sabourin et al. (2006) how surface transfer works or which features qualify for surface transfer. 16 Sabourin et al. (2006) don’t have a native speaker control group for the gender assignment task, only for the gender agreement task. Whether the L2 groups perform within the native range for gender assignment can thus not be determined (although the accuracy rate in the gender assignment task is above 80% for all groups, so all groups are successful on this task). 30 Source: http://www.doksinet Sabourin and Haverkort (2003, cf. §152) corroborate the finding that +gen L1 learners are not completely nativelike in L2 gender by showing that the processing

of gender in Dutch by advanced German speakers is not the same as the processing of gender by Dutch native speakers. Again, it should be noted that although Franceschina and the FFFH do not make specific predictions regarding the processing of L2 gender, the basic assumptions of the FFFH lead to certain predictions regarding L2 processing. As the +gen L1 group has nothing fundamentally new to acquire, they should show nativelike processing of L2 gender if they are highly proficient users of the L2. This should be especially clear in the case of German learners of Dutch, as the gender systems of German and Dutch are very similar. 17 To summarise: Sabourin et al. (2006) show that the nature of the L1 influences a learner’s capacity to acquire L2 gender. If either surface transfer or deep transfer is possible, the learner will be better at acquiring the L2 gender system than when these modes of transfer are not possible. In contrast to Franceschina (2005), however, Sabourin et al.

(2006) find a difference between the performance of the non-native learners (irrespective of the gender characteristics of the L1) and the native speakers. 1.532 The L1 as a Minor Factor in L2 Gender Various studies support the idea that whereas the L1 may play a facilitating role in acquiring L2 gender, it is not the crucial factor. The study by White et al(2001; see section 1.512), for example, shows that while initially there is an effect of the L1, eventually the performance of all learners becomes indistinguishable from the performance of the native speakers. This is shown by the fact that initially the French (+gen) low proficiency group outperform the English (-gen) low proficiency group, which suggests that the gender feature in French helps in acquiring Spanish gender in the first phases of acquisition. At higher levels of proficiency (intermediate and high proficiency), this difference disappears and the performance of the English group is as good as the performance of the

French group, both performing similarly to the native Spanish group. A similar study by White et al (2004, described in section 1512) also shows that speakers of both +gen and –gen languages can attain nativelike performance in a gender comprehension task, although in this study, in contrast to White et al. (2001), an effect of the L1 at low levels of proficiency is not found A study by Bruhn de Garavito and White (2000), which is somewhat in contrast to the two studies by White et al. mentioned above, shows that both French and English groups of high proficiency learners experienced persistent problems with acquiring Spanish gender. According to this study, gender thus remains problematic for learners of Spanish, but the problems are not related to the gender features of the L1 background of the learners. 18 The results of the studies mentioned above are thus 17 Guillelmon and Grosjean (2001) also show that English learners of French process gender differently than native

speakers, as the learners did not show any gender congruency effect in the recognition of nouns. Whether the lack of a gender congruency effect is typical for all non-natives or specific to learners from a –gen L1 background is not clear,however, as this study did not look at the gender congruency effect in learners from a +gen L1 background. 18 Hulk and Cornips’ study (2006) similarly finds no effect of the L1 in the acquisition of Dutch gender in their 2L1 subjects. As in the study by Bruhn de Garavito and White, all learners were relatively poor in their sensitivity to the gender system, but the nature of the L1 did not correlate with their performance. Hulk and Cornips attribute their subjects’ poor performance to the fact that they are exposed to a relatively impoverished version of Dutch. They may thus not have been able to establish 31 Source: http://www.doksinet somewhat inconsistent in whether or not nativelike sensitivity to gender can be attained in a second

language. They are consistent, however, in finding that the L1 is not the primary factor influencing the capacity to acquire an L2 gender system. White et al. (2001, 2004) and Bruhn de Garavito and White (2000) claim that their results are in line with the FTFA account of second language acquisition mentioned in section 1.53 According to the FTFA hypothesis of second language acquisition, L2 learners start off by adopting the L1 grammar for the L2. The L1 grammar is thus the initial state of the second language for beginning learners. As the L2 learner uses the L2 and is exposed to the L2, the L1 grammar settings are substituted for the L2 settings. Initially, then, the L2 learner will show a very clear influence of the L1 in her L2 usage, but this effect diminishes as the learner becomes more proficient in the L2. FTFA thus predicts that low proficiency learners coming from a –gen L1 will have problems with gender in the L2, as they cannot transfer the gender feature from their L1

to the L2. Low proficiency learners coming from a +gen L1, on the other hand, will have considerably less difficulty acquiring gender, as they can transfer the L1 gender feature and feature checking mechanisms to the L2. 19 However, as all learners are presumed to still have access to UG, the differences between +gen and –gen learners should disappear, as exposure to and use of the L2 allow even the –gen L1 learners to figure out the workings of the L2 gender system. 20 The studies discussed in this section are thus generally in line with the predictions of the FTFA, as the L1 does have some influence on the gender acquisition process in the early stages of acquisition, but eventually learners from all L1 backgrounds become equally good (or bad 21) in their mastery of the L2 gender system. There is, however, one issue that is not really dealt with in the FTFA framework, namely when the resetting of the L1 grammar to encode L2 properties is expected to take place. Can learners

incorporate gender into the L2 at an early or a late stage in their L2 development? In other words: at what stage in linguistic development does the FTFA predict that learners will show proficient use of an L2 gender system? Another related issue regards the difference between –gen L1 and +gen L1 learners. Following the FTFA, initially, the +gen L1 learners are expected to be more advanced users of the L2 gender system, because the +gen L1 learners can transfer their gender feature to the L2. It is probable though that the +gen L1 learners also have to change their L1 settings to incorporate L2 gender. The +gen L1 learners do not have to create a gender feature like the –gen L1 learners (as they can transfer the L1 gender feature), but they may have to reset their L1 settings in order to deal with, for example, gender- the correct gender feature on nouns because they have not received the right input that is necessary for this connection to be made. 19 Although the extent of the

transfer advantage may depend on the similarity of the gender system of the L1 and the L2 (cf. the discussion on surface transfer and deep transfer in section 1531) 20 Note that nativelike final attainment is not necessarily the outcome. A learner may converge on a representation that they presume is appropriate for the L2, but which is not exactly the same as the native speakers’ representation (cf. White 2003) 21 It should be noted, however, that very unsuccessful ultimate attainment of L2 gender is only expected under FTFA if the learner faces severe learnability problems. If the input received by the learners is qualitatively very poor, for example, or the exposure is very limited learners may not be able to aquire L2 gender successfully. 32 Source: http://www.doksinet marking on items that are not gender marked in the L1. 22 Again, the question is when this resetting is likely to occur. From a theoretical point of view then, it would be desirable if the FTFA hypothesis made

some predictions as to when the resetting of the L1 grammar to encode L2 properties is expected to take place. It seems likely that the FTFA would predict that the resetting moment is dependent on the amount of exposure received by a learner and on the frequency and transparency of the phenomenon in the input. If we consider gender, we could thus assume that a learner acquiring Dutch gender might need more input for full mastery than a learner acquiring Italian gender. After all, in Italian, gender is very prominently and transparently present in the input, whereas in Dutch, gender is less prominent and opaque. One might thus expect that proficient learners would have had enough input to have fully acquired transparent and frequent elements of the language, but possibly not enough to have acquired the more obscure elements of the L2. In the case of gender, we might thus offer the following predictions following FTFA: if gender is transparent and/or frequent in the L2, learners can

reset their L1 settings at an early stage in L2 acquisition. If gender is opaque and/or infrequent in the L2, learners will not reset their L1 settings until a late stage in their L2 development. Assuming these predictions, it may be insightful to briefly reconsider some of the studies on the acquisition and processing of gender discussed in previous sections. Franceschina (2005) looks at the acquisition of Spanish gender, a transparent and predominant feature of Spanish. 23 Franceschina finds that all the high proficiency learners are successful in their acquisition of gender, as expected by the FTFA. What is not expected by the FTFA in this case is that the –gen L1 group should perform worse than the +gen L1 group. This could be explained, however, by the problems with Franceschina’s study discussed in section 1.5311 Sabourin et al (2006) and Sabourin and Haverkort (2003) look at the acquisition of Dutch gender, which is opaque and relatively infrequent in comparison to a Romance

language like Spanish (i.e there are less contexts in which gender is marked in Dutch than in Spanish) Their results are as expected under the FTFA: even learners who have had a lot of exposure to Dutch still have problems with the acquisition and processing of the Dutch gender system. It is also expected that the German learners will perform better than the Romance learners and that the Romance learners will perform better than the English learners. After all, if all the learners are using their L1 settings (because the nature of Dutch gender has not allowed them to reset the L1 settings to L2 settings), it is not surprising that German learners (whose L1 gender system is similar to the Dutch gender system) will perform better than Romance learners (whose L1 gender is not very similar to Dutch gender) who, in turn, will perform better than English learners (who do not have L1 gender). To summarise: existing studies on L2 gender show that the nature of the L1 influences the ease and

speed with which L2 gender is acquired to some extent. It seems unlikely that the L1 is the only relevant factor, however, as, in some studies at least, learners 22 This would also explain why +gen L1 learners do not necessarily become entirely nativelike in their L2 gender. After all, if they too have to change their L1 settings, it is possible that the result is different from the native speaker settings for gender. 23 As Spanish and Italian are both Romance languages, Spanish gender is quite similar to Italian gender: they have similar morphological forms and gender agreement is present on similar categories. 33 Source: http://www.doksinet from both +gen and –gen L1 backgrounds seem to be capable of acquiring gender to high levels of proficiency. These findings are in line with the FTFA account of second language acquisition (provided the extra assumptions discussed above are added to it) and in contrast to the FFFH account in that the FFFH does not predict that–gen L1

learners will be able to reach a high level of proficiency in L2 grammatical gender. 1.6 Grammatical Gender: a Broader Perspective As can be seen from the previous sections, there is a considerable body of research on grammatical gender. When applied to gender, two of the main theories on second language acquisition make quite different predictions regarding the status of grammatical gender in the second language learner. More information on the acquisition and processing of gender can thus potentially give information about the acquisition and processing of many other features in a second language as well. If it turns out that gender features can only be mastered by those with gender in their first language, this has potentially far reaching consequences for other predictions related to the acquisition of grammatical features besides gender (case and number features, for example, to name but two), not to mention educational practice. After all, if it is simply biologically impossible

to acquire certain linguistic features after puberty, maybe more effort should be made to acquire languages with features alien to the L1 at a young age. On the other hand, if it turns out that gender features can be acquired throughout the lifetime, then it becomes even more important to find out what triggers the resetting of the L1 grammar to L2 values. A further comparison of the processing of gender in first and second language could also offer insights into another important issue in linguistics more generally. One of the main questions in linguistics is how knowledge of language interacts with the processing of language or, to put it in more classical terms, how competence interacts with performance. How is language affected by problems in linguistic knowledge? How is language affected by processing problems? These questions can be considered by looking at experimental populations with ‘non-normal’ language, like young children, aphasics and second language learners. Can

their non-normal language be attributed to lack of knowledge of the relevant linguistic concepts? Or is the required speed of processing and integration of knowledge the bottleneck? To reformulate these questions with respect to L2 gender: do second language learners not know which gender to assign to which categories and in what contexts? Or are they just incapable of integrating all of this knowledge fast enough? Some of the studies discussed in the previous sections referred to this question (be it somewhat indirectly), but more research on grammatical gender in the L2 considering both on-line and offline data would help to further illuminate this issue. The next chapter presents two new experiments which are designed to do just that. Many of the issues raised in this chapter will form the basis for the new experiments and hence will be revisited in the next chapter: the influence of the L1 on the L2 (with the FFFH/FTFA debate in the background), the capacity of learners to reach

nativelike proficiency, the difference in processing opaquely and transparently gender-marked nouns and the use of deep transfer in L2 acquisition. 34 Source: http://www.doksinet Chapter 2 The Study This chapter describes two new experiments on the L2 acquisition and processing of gender. The first two sections give a general overview of the experiments; further sections will give more specific information on the design (section 2.32), the materials (section 2.33) and the procedure (section 234), as well as the subjects (section 2.31) 2.1 Outline of the Study The study consists of two experiments that have been designed to test subjects’ sensitivity to gender mistakes in the input: a self-paced reading (SPR) experiment and an acceptability judgement task (AJT). In both experiments, the performance of a control group of native Italians is compared to the performance of a group of English learners of Italian and a group of Dutch learners of Italian. The selection of these three

groups is based on the gender characteristics of the three languages: Italian has an overt, transparent gender system, Dutch has a covert, opaque gender system and English has practically no gender system. The effect of the gender characteristics of English and Dutch on the acquisition and processing of Italian gender can thus be considered in the two experiments. The first experiment is an SPR experiment in which subjects are required to read Italian sentences at their own pace and answer simple comprehension questions about them. Each sentence is divided into three parts; in the target sentences, the second part of the sentence always consists of a Det-N-Adj DP with correct or incorrect gender agreement (either the determiner and the adjective both have correct gender marking with respect to the noun or one or both of these elements has incorrect gender marking). As well as the gender marking on the determiner and the adjective, the nature of the noun is also varied: the target

sentences consist of equal numbers of feminine, masculine, regular and irregular nouns. The reading time for each of the three segments of the sentence is measured during the experiment. The idea behind the SPR task is that the reading times of the sentences will give an insight into the way gender is processed in grammatical, ungrammatical, masculine, feminine, irregular and regular contexts by the native and non-native speakers of Italian. Potential differences between the natives and the non-natives in reading times (and hence in gender processing) should thus be uncovered in the SPR experiment. The second experiment, which was conducted immediately after the SPR experiment, is an acceptability judgement task. In this experiment, subjects were required to read sentences and give them a mark out of six for acceptability (a mark of 1 signifying a completely unacceptable Italian sentence, 6 a perfectly acceptable one). The target sentences in the SPR task and the AJT are the same, so

sentences containing various different DPs were presented to the subjects in order to assess the influence of the nature of the DP on the acceptability of the sentence. The idea behind this experiment was to test the participants’ knowledge of gender assignment and agreement. The assumption is that if participants assign high marks to sentences with correct gender marking and low marks to sentences with incorrect gender marking, they know what the gender of the noun is and what forms the determiner and the adjective have to take to reflect that gender. On the other hand, if participants assign a low mark to a 35 Source: http://www.doksinet sentence with correct gender marking or a high mark to a sentence with incorrect gender marking, then they either do not know what the gender of the noun is or how this gender should be instantiated on the determiner and the adjective. These two experiments build upon the previous research described in Chapter 1 in various ways. Both experiments

follow the set-up of the studies in Franceschina (2005) and Sabourin et al. (2006) in that the gender acquisition of a group of learners from a –gen L1 (English) is compared to that of learners from a +gen L1 background (Dutch). In contrast to the studies by Franceschina and Sabourin et al, this study will also include an online measure of gender processing (i.e the reading times in the SPR experiment). This online measure should give some insight into the possible differences in processing induced by the –gen or +gen L1 background in comparison to native speaker processing. The SPR experiment also adds to the study by Sabourin and Haverkort (2003, also discussed in chapter 1). Sabourin and Haverkort (2003) do have an online measure of gender processing (they look at participants’ ERPs), but they do not compare the performance of a –gen L1 group to a +gen L1 group (they only consider the performance of German learners of Dutch). The SPR study thus combines the two set-ups:

online data on the processing of gender is collected and a comparison is made between learners from –gen L1 and +gen L1 backgrounds. The AJT also offers new information on L2 gender in various ways. Most studies that test learners’ knowledge of gender assignment and agreement require learners to specify what the gender of a certain noun is or whether a certain sentence containing gender marking is grammatical or not. The set-up chosen in this study gives a more fine-grained picture of the learner’s knowledge of gender, as the use of a scale allows the learner more options to rate a sentence. In a grammaticality judgement task (in which the learner has a binary choice on whether the sentence is grammatical or not), it is not possible for the learner to rate a sentence with a DP with one incorrectly gender-marked element (a DP with an incorrect determiner, but a correct adjective for instance) differently to a sentence with incorrect gender marking on both the determiner and the

adjective. It is also not possible for the learner to indicate what type of gender mistake they consider worse; whether an incorrect determiner is considered to be worse than an incorrect adjective or vice versa cannot be determined if a standard grammaticality judgement task is used. Because a scale is used in the AJT, subtle differences in acceptability caused by the nature of the noun may surface in this test. If, for example, sentences containing irregular nouns are consistently marked lower or higher than sentences with regular nouns, this could indicate that regular and irregular nouns are represented differently in the participant’s lexicon. Again, these kinds of subtleties cannot be inferred from standard binary choice grammaticality judgement tasks. As the learners are not specifically asked to focus on gender in their judgements (as they would be in a gender identification task) and because they are not forced to make a yes/no choice with respect to the grammaticality of

the sentence (as they would be in a binary choice grammaticality judgement task), it is expected that this set-up will tap into the learner’s intuitive knowledge of Italian gender rather than their metalinguistic knowledge (i.e the explicit knowledge about gender they have acquired in the classroom) in grading the sentences. This is desirable because the aim of the AJT is not to assess which learners have paid more attention in the classroom, but which learners have a more nativelike underlying knowledge of gender. 36 Source: http://www.doksinet As the same target sentences are used in both the SPR and the AJT study, the set-up of these two studies also allows the reading time of the DPs to be related to the mark given to the sentences containing the DPs. The learner’s processing abilities for gender can thus be compared to their knowledge of gender. This study will also consider individual data from the non-native groups, so that it will be possible to determine whether any

individuals are capable of performing within the native range on the two experiments. 2.2 Research Questions The two experiments have been designed to shed light on various issues in second language acquisition of gender. The main research questions of this thesis are the following: 1. Do non-native learners of Italian differ with respect to native speakers of Italian in: a) their knowledge of gender? b) their processing of gender? 2. Are there differences between the English and the Dutch learners of Italian with respect to: a) their knowledge of gender? b) their processing of gender? 3. If there are differences between the native speakers and the non-native learners of Italian, are there any individuals in the Dutch or the English group who perform in the native range with respect to: a) their knowledge of gender? b) their processing of gender? The influence of the nature of the DP on the processing and knowledge of gender in the natives and the non-natives will also be considered.

The following questions will thus be addressed: 4. To what extent do the native and the non-native subjects’ reading times and acceptability judgements depend on: a) the grammaticality/pattern type of the DP b) the regularity of the noun c) the gender of the noun 2.21 Predictions 2.211 Research Question 1 Will the native speakers of Italian differ from the non-native learners of Italian in their knowledge and processing of gender as judged in the SPR experiment and the AJT? Many studies on L2 acquisition of gender have found that it is very hard for learners to attain nativelike processing and knowledge of gender (cf. Holmes and Dejean de la Bâtie 1999, Bartning 2000, Bruhn de Garavito and White 2000, Dewaele and Véronique 2001, Weerman et al. 2003, Hawkins and Franceschina 2004) On the other hand, there are also a number of studies that find nativelike gender in certain groups of learners (White et al. 2001 and 2004) As different studies have had 37 Source: http://www.doksinet

different outcomes, previous research thus does not really lead to a firm prediction for this research question. The predictions that the FTFA would give for this study are also not entirely clear-cut: in theory, both non-native groups should be capable of attaining nativelike gender, but this is dependent on whether the learners in these groups have had enough of the right opportunities to acquire L2 gender (enough exposure, good quality input etc.) The FFFH, on the other hand, is very clear in its predictions for the English group: the English learners will definitely differ from the native speakers in their performance on both tests. Following the FFFH, the Dutch learners should perform quite similarly to the native group. After all, the Dutch learners have nothing new to acquire, so as long as they have had enough exposure to Italian to determine the gender of the nouns in the input, their performance should be very similar to that of the natives. Given all these different points

of view, what should the predictions for this research question be? Most studies find that learners are capable of acquiring the gender of nouns in isolation (even the FFFH would claim that it is possible for –gen L1 learners to memorise the gender of nouns in isolation). As the learners in this study are advanced, it seems likely then, that they will have at least some knowledge of gender. In other words: they will probably know that Italian nouns have gender and that this gender is realised on various other items, like determiners and adjectives. Furthermore, as Italian has a transparent gender system, the learners should also be able to recognise the gender of regular nouns fairly easily, as they will have been taught that nouns ending in –a are generally feminine and those ending in –o are generally masculine. The irregular nouns, on the other hand, may prove to be more difficult for the non-natives. As the gender of these nouns is not immediately visible from the noun’s

ending, the learners may have problems in retrieving the gender of the noun. If the learners encounter an irregular noun with incorrect gender marking (as they do in the ungrammatical patterns presented to them in this study), they may thus be ‘tricked’ into thinking that this gender marking is correct. Regarding the nonnatives’ knowledge of gender then, the following prediction can be made: learners will not differ in their knowledge of gender with respect to regular nouns, but they will differ when irregular nouns are considered. In other words: the marks given on the AJT to sentences with regular nouns will not differ from the natives’ marks, but the marks given to sentences with irregular nouns will differ from those of the natives. The predictions regarding the learners’ processing of gender are somewhat different. The FTFA and the FFFH do not say anything explicit about processing of gender in an L2, but given the basic assumptions of these theories it should not be

impossible for non-natives to become nativelike in their processing of gender (although the FFFH would rule out –gen L1 learners from becoming nativelike). Nonetheless, nativelike processing of gender in non-natives has not been found in any of the studies discussed in chapter 1 (Taraban and Kempe 1999, Guillelmon and Grosjean 2001 and Sabourin and Haverkort 2003). It thus seems unlikely that the non-natives will show nativelike processing in this study. The predictions for the non-natives processing of gender are thus the following: the non-natives will differ from the natives in their processing of gender (i.e their reading times in the SPR experiment) 2.212 Research Question 2 Research question 2 considers whether there will be a difference in the performance of the non-native groups in the two experiments. FTFA and FFFH make very clear 38 Source: http://www.doksinet predictions with respect to this research question. Proponents of the FFFH would predict that the Dutch group

will outperform the English group on all fronts, as Dutch has gender, but English does not. The Dutch group will thus perform more like the natives than the English group in both the AJT and the SPR experiment. FTFA makes quite different predictions regarding this research question. FTFA predicts that there will only be a difference between +gen L1 learners and –gen L1 learners at the early stages of language acquisition. As this study considers the performance of advanced learners of Italian, proponents of FTFA would thus predict that both groups should perform very similarly. If this study shows that the Dutch group outperforms the English group on both tests, FFFH is supported. If, on the other hand, this study finds that there is no difference between the two non-native groups on either of the tasks, this would be in favour of FTFA. Previous studies show that there is support for both of these options. As discussed in chapter 1, some studies find that there is a clear influence

of the L1 in the acquisition of L2 gender, whereas other studies do not find significant differences in the performance of +gen and –gen L1 learners in tests of L2 gender. It is thus somewhat of an open question which direction the results of this study will go in. The gender feature in Dutch may give the Dutch group an advantage over the English group even though only advanced learners are tested in both groups. On the other hand, the Dutch and the Italian gender systems are quite dissimilar, both in their nature (opaque vs. transparent) and in the developmental trajectory in L1 acquisition (slow, laborious acquisition of Dutch gender vs. quick, effortless acquisition of Italian gender) This study will thus have to show whether the Dutch group are at an advantage in their acquisition of gender in comparison to the English group or whether the performance of the two groups is similar. 2.213 Research Question 3 The third research question considers whether any individuals in the two

non-native groups will fall in the native range in either or both of the experiments. The predictions for this question are also directly related to the FFFH/FTFA debate. The FFFH would predict that whilst it is perfectly possible that there will be individual Dutch learners that perform within the native range, it is fundamentally impossible for any of the English learners to be nativelike in their performance. FTFA, on the other hand, would predict that learners from both language backgrounds should be able to perform within the native range, provided the learners have had enough exposure to Italian. As stated in the predictions for research question 1, it is probable that the non-native groups will generally know the gender of the regular nouns. The irregular nouns, on the other hand, were predicted to cause problems for the non-native speakers. Nonetheless, a number of non-native speakers may know the gender of the irregular nouns used in this study. These individual learners may

thus show nativelike knowledge of gender throughout the AJT, not just for the regular nouns. Whether there will be any non-native individuals that fall in the native range on the SPR experiment as well is more of an open question. Previous studies on the processing of gender in the L2 have found that, when group scores are considered, non-natives tend to process gender differently to native speakers. Given this fact then, it seems unlikely that this study will find a large number of individuals that fall within the native range on the SPR experiment. None of the previous studies on non-native gender processing 39 Source: http://www.doksinet give individual data, however, so it is unclear whether some of the non-natives in those studies were in fact nativelike in their gender processing. It is thus possible that this study will find a small number of non-natives that fall in the native range for the SPR experiment. So, if there are indeed individuals whose performance is nativelike

in this study, will they be in the Dutch, the English or in both groups? If this study finds that only individuals from the Dutch group perform in the native range, this would support the FFFH account. If, on the other hand, individuals from both the English and the Dutch group are nativelike in their performance, this would be along FTFA lines. The results of this study will have to show whether there are individuals that perform within the native range and if so, whether these individuals come from a +gen and/or a –gen L1 background. 2.214 Research Question 4 Research question 4 looks at the nature of the DP and the noun on the subjects’ acceptability judgements and reading times. Three different factors are considered: the grammaticality/pattern type of the DP, the regularity of the noun and the gender of the noun. 2.2141 The Nature of the DP and the Noun in the AJT The grammaticality of the DP is expected to have a large effect on the marks given by the native speakers. As the

native speakers will know the gender of all of the nouns and what form the determiner and the adjective should have as a result of this gender, it is expected that the natives will assign significantly lower marks to sentences with ungrammatical DPs in comparison to sentences with grammatical DPs. This prediction can be further specified if the pattern type of the DP is considered. The hierarchy in marking in native speakers is predicted to be as follows: completely grammatical DPs will receive the highest marks, followed by DPs with one gender marking error, followed by DPs with two gender marking errors. As the natives are expected to know the gender of all of the nouns in the AJT, whether they are regular or irregular, the regularity of the noun is not expected to influence the natives’ judgements on the AJT. After all, there is no reason why incorrect gender marking in DPs with regular nouns should be considered worse or better than incorrect gender marking in DPs with irregular

nouns. As there is no evidence in prior research that indicates that nouns of any particular gender are more easily recognised or that there is a default strategy for nouns in L1 acquisition of Italian, it is not expected that the gender of the noun will influence the judgements given by the native speakers. The predictions for the non-natives are slightly more complicated. As the learners tested in this study are advanced, it is expected that they will have at least some knowledge of grammatical gender in Italian. They will probably recognise the gender of the regular nouns and be able to determine whether the determiner and the adjective have the correct form given the gender of the noun. For DPs with regular nouns then, the prediction is that the non-natives will give significantly higher marks to the sentences with grammatical DPs than to sentences with ungrammatical DPs. The learners may not always be aware of the gender of the irregular nouns, however, as they cannot use the

gender information on the noun to guide them. This may mean that the non-native learners do not always recognise incorrect gender marking in DPs with irregular nouns. The prediction is then that ungrammatical DPs with irregular nouns may not receive significantly lower marks than grammatical DPs with irregular nouns. In the case of regular nouns then, the pattern type hierarchy will be the same 40 Source: http://www.doksinet for the non-native speakers as for the native speakers: completely grammatical DPs will receive the highest marks, followed by DPs with one gender marking error, followed by DPs with two gender marking errors. Given that the non-natives are predicted to be unsure about the gender of irregular nouns, there will probably not be a clear pattern hierarchy for the irregular nouns. The gender of the noun presumably will not play a great role in the non-natives’ judgements, as there is no evidence in prior research that nouns of any particular gender are more easily

recognised or that there is a clear default strategy for Italian nouns in L2 learners of Italian. It is thus predicted that there will be no significant difference between the marks given to DPs with feminine nouns and those assigned to DPs with masculine nouns. 2.2142 The Nature of the DP and the Noun in the SPR Experiment The processing of gender in the natives will presumably also be affected by the grammaticality of the DP. As ungrammaticality generally leads to longer reading times 24, it seems likely that grammatical gender-marking on the DP should lead to faster reading times than ungrammatical gender-marking. This outcome would also be in line with the gender congruency effect, according to which correctly gendermarked nouns are processed more quickly than incorrectly gender-marked nouns (see section 1.4222 for discussion of this effect and references) Gender processing is thus predicted to be affected by the grammaticality of the DP for the natives. The predictions for the

effect of pattern type are similar: completely grammatical DPs are expected to take the shortest amount of reading time, followed by DPs with one gender marking error, followed by DPs with two gender marking errors. The gender processing of the natives is not expected to be influenced by the regularity of the noun. As research by Taraban and Kempe (1999) and Hohlfeld (2006) shows (see chapter 1, section 1.421), natives do not take longer to process familiar irregular nouns than regular nouns. As the gender of the irregular nouns should be immediately available to the natives, it is thus not expected that the natives will show any differences in processing time given DPs with regular nouns in comparison to DPs with irregular nouns. Like the regularity of the noun, the gender of the noun is not expected to influence the natives’ performance in the SPR experiment either. Prior research gives no reason to believe that nouns are processed differently depending on their gender, so the

reading times are not expected to be affected by the gender of the noun. The predictions regarding the gender processing of the non-natives are less clear. It is not certain whether the non-natives will also show processing differences for grammatical and ungrammatical sentences. The one study of the gender congruency effect in non-native learners (by Guillelmon and Grosjean 2001, see section 1.4222) failed to show a gender congruency effect. Non-native learners thus did not process nouns faster if they were preceded by the correct determiner, nor was their processing slowed down if the noun was preceded by an incorrect determiner. The study by Guillelmon and Grosjean only considered learners from a –gen L1 background, however, so it is possible that learners from a +gen L1 background will show a gender congruency effect in their gender processing. If FTFA and FFFH are applied to 24 See Pearlmutter et al. 1999 for an example of a study in which sentences with a subject-verb mismatch

lead to longer reading times than the same sentences with subject-verb agreement. 41 Source: http://www.doksinet processing, opposing predictions for this part of research question 4 can be offered. According to FFFH, the English group will not show nativelike differences in reading time for grammatical and ungrammatical gender marking, but it is possible that the Dutch group will show a nativelike difference. FTFA does not predict there to be a difference between the two non-native groups. Following FTFA both groups could show nativelike processing of gender if the learners in the non-native groups have had enough qualitatively good exposure to Italian. Given the paucity of research in this area and the contrasting predictions given by FTFA and FFFH, it is hard to determine in advance which direction the results for this part of the study are likely to go in. The results of this study will thus have to show whether the FTFA or the FFFH points of view are upheld. As the influence

of the DP pattern on the reading times is related to the influence of the grammaticality of the DP, it is also not clear what the predictions for the pattern hierarchy should be. If non-native learners show a gender congruency effect, it would be expected that the pattern hierarchy would be the same as for the natives: the completely grammatical DPs take the shortest amount of reading time, followed by the DPs with one gender marking error, followed by the completely ungrammatical DPs. If there is no gender congruency effect in the non-natives, it is also not expected that there will be a clear pattern hierarchy in the reading times. The regularity of the noun may also influence the reading times of the non-natives. The regular nouns are not expected to cause the non-natives great processing problems: it is assumed that the non-natives will be able to retrieve the gender of regular nouns relatively easily, so it is also expected that the processing of regular nouns will be

unproblematic. Irregular nouns may cause the non-natives trouble, however, as they may not be able to retrieve the gender of irregular nouns very easily (as information from the noun’s ending cannot be used in determining the noun’s gender). It is thus expected that non-natives will take longer to process DPs with irregular nouns than DPs with regular nouns (see Taraban and Kempe 1999 for this finding in L2 learners of Russian). As there is no evidence in prior research that masculine nouns are more easily processed than feminine nouns or vice versa, the gender of the noun is not expected to influence the reading times of the non-natives in the SPR experiment. 2.215 Summary of Predictions The predictions can be summarised as follows: Research Question 1: 1. No difference between natives and non-natives for regular nouns in the AJT 2. Native speakers and non-native learners differ in judgements for irregular nouns in the AJT. 3. Natives and non-natives differ in their reading times

in the SPR experiment Research Question 2: 4. Whether the performance of the English group will differ from the performance of the Dutch group is unclear. The FFFH predicts that the Dutch group will outperform the English group; FTFA predicts that the performance of the two groups will be similar. Research Question 3: 5. A small number of non-natives will fall into the native range regarding their performance on the AJT and the SPR experiment. Whether these nativelike nonnatives will only be in the Dutch group, as the FFFH would claim, or whether 42 Source: http://www.doksinet there will also be English individuals who are nativelike, as FTFA would predict, remains to be seen. Research Question 4: 6. Natives: the grammaticality/pattern type of the DP should influence both the reading times in the SPR experiment (ungrammatical DPs are expected to take more time than grammatical DPs) and the marks given in the AJT (ungrammatical DPs are expected to receive lower marks than

grammatical DPs). 7. Natives: the regularity and gender of the noun are not expected to influence the reading times or the marks given by the native speakers. 8. Non-natives: it is expected that the grammaticality/pattern type of the DP will influence the marks if the noun in the DP is regular. In case of regular nouns, the non-natives will thus give higher marks to grammatical sentences than to ungrammatical sentences. 9. Non-natives: the grammaticality of the DP is not expected to influence the marks when the noun in the DP is irregular. 10. Non-natives: it is not clear whether the grammaticality of the DP will influence the reading times of the non-natives. Following the FFFH, the English group will not show an influence of grammaticality, but the Dutch group might. FTFA would claim that both groups may show an effect of grammaticality. 11. Non-natives: the regularity of the noun is predicted to influence the marks and the reading times. Sentences with regular nouns are predicted to

receive high marks and short reading times if the gender marking is grammatical and low marks and long reading times if it is not. This difference in marking and reading times is not expected to occur for sentences with irregular nouns. 12. Non-natives: the gender of the noun is not predicted to influence the marks or the reading times. 2.3 The Experiments 2.31 Subjects In both experiments, one control group and two experimental groups were tested. The control group consisted of ten native speakers of Italian. Five out of ten subjects in this group were male, five were female. Seven of the Italian native speakers were recruited in Edinburgh (Scotland) where they were living temporarily for study or work purposes; the three other native speakers were recruited in Utrecht (The Netherlands) where they were living temporarily for study purposes. The native speakers were all between 22 and 37 years old, with a mean age of 27.9 years All of the native speakers had last lived in Italy between

five months and thirteen years ago 25, with 30.5 months as the mean number of months since last permanent stay in Italy. All participants in this group go back to Italy on a regular basis and all of them reported frequently speaking Italian whilst abroad. Other than occasional word finding difficulties, none of the native speakers reported any changes in their Italian as a result of living abroad. 26 The two experimental groups consisted of advanced Dutch and English learners of Italian. Eleven English learners of Italian were tested; one of these learners had to be excluded, however, as she had acquired some Italian as a child and hence no longer qualified as an adult L2 learner of Italian. The remaining ten English subjects had all 25 There were no great differences in the performance of this individual and the other natives that had last lived in Italy more recently. 26 See appendix B for individual details of both the control and the experimental groups. 43 Source:

http://www.doksinet acquired Italian after puberty. Of these ten learners, five were female and five were male. The English learners were between 20 and 59 years old with a mean age of 30,5 years. All but one of the English learners had spent at least four years formally studying Italian and had lived in Italy for at least 4 months (range = 4-96 months; mean = 18.9 months) The remaining learner had not lived in Italy consecutively for any significant period of time and had formally studied Italian for one year only, but this learner had had a large amount of exposure to Italian as he has been married to an Italian for 17 years and regularly visits Italy for short periods of time. All of the English learners had learnt at least one language with grammatical gender other than Italian (usually French) after puberty. None of the learners had acquired a language with grammatical gender before puberty. All of the non-native learners (and a subset of the natives) were asked to complete an

Italian proficiency test which tested the subject’s grammar, vocabulary and reading skills in Italian (see Appendix C). For the grammar part of the test, the subjects had to select missing words (a cloze test) and incorrect words; the subjects’ vocabulary was also tested via a cloze test. For the reading part of the test, subjects had to read short pieces of text and answer questions about them. All parts of the test were multiple choice with four possible answers The English subjects performed well on this test with the number of errors ranging between 0 and 9 (out of 65, mean number of errors 3,3). Twelve Dutch learners of Italian were tested, but two of these learners had to be excluded. One had learnt some Italian as a child; the other learner had not studied Italian formally at all nor had he spent a significant amount of time in Italy (which meant that he was not comparable to the other non-native learners). The remaining ten Dutch learners had all learnt Italian after

puberty and had studied Italian formally for at least three years. Of the ten Dutch learners, two were male and eight were female The ages of the Dutch group ranged between 21 and 47 years old with a mean age of 30,1 years. All of the Dutch learners had spent at least three years studying Italian and seven of the Dutch learners had spent a significant amount of time living in Italy (range = 4-36 months; mean = 14.1 months) The other three learners had not been to Italy for any long continuous period of time, but they had been visiting Italy regularly for many years. Like the English group, all of the Dutch learners knew another gender language other than Italian which they had learnt after puberty (usually German and/or French). None of the Dutch learners had learnt a gender language other than Dutch before puberty. The Dutch group also performed well on the proficiency test, with the number of errors ranging between 0 and 5 (out of 65, mean number of errors 2.1) Statistical tests show

that the three groups do not differ significantly from each other in their performance on the proficiency test. Results from a Kruskal-Wallis test 27 indicated that there was no significant difference between the three groups (p=.946) Mann-Whitney tests further specified this result to show that there was no significant difference between the native group and the English group (p= .884; z= -145), the native and the Dutch group (p= .941; z= -074) or the Dutch and the English group (p=.697; z=389) Given these results, it can be assumed that the three groups have a 27 A Kolmogorov-Smirnov test showed that the results of the proficiency test were not normally distributed. Parametric tests like an ANOVA or a t-test could thus not be performed on the data The tests that have been used (i.e the Kruskal-Wallis test and the Mann-Whitney test) are both nonparametric tests that do not assume normal distribution 44 Source: http://www.doksinet comparable level of proficiency. For an overview

of the subject information, see Table 4: Table 4 Subject information Group N Age range Age mean Native English Dutch 10 10 10 22-37 20-59 21-47 27.9 30.5 30.1 Range errors proficiency test 1-3 (n=4) 0-9 (n=10) 0-5(n=10) Proficiency test mean 1.75 3.3 2.1 2.32 Design The SPR experiment and the AJT have various between- and within-subjects variables. The first between-subjects independent variable is the subject’s first language. This variable has three levels: Italian, English and Dutch The second between-subjects independent variable is related to the first variable, namely the learner group. This variable has two levels: native (for the subjects with Italian as their L1) and non-native (for the subjects with English or Dutch as their L1). As all subjects received the same set of sentences, there are four within-subjects independent variables which are related to the nature of the noun and the DP. The first withinsubjects variable is the pattern of the DP This variable has

four levels: Det-N-Adj (in which the determiner and the adjective are both correctly gender-marked); *Det-NAdj (in which the determiner and the adjective are both incorrectly gender-marked); *Det-N-Adj (in which the determiner is incorrectly gender-marked, but the adjective has correct gender-marking ); Det-N-*Adj (in which the determiner has correct gender-marking, but the adjective is incorrectly gender-marked). The second withinsubjects variable is an alternative formulation of the first variable: the grammaticality of the DP. This variable has two levels: grammatical (ie the Det-N-Adj pattern) and ungrammatical (i.e the *Det-N-Adj, Det-N-Adj and the Det-N-Adj pattern combined). The third within-subjects variable is the regularity of the noun, which has two levels: regular (i.e the noun ends in –a for feminine nouns and –o for masculine nouns) and irregular (i.e the noun ends in –e) The fourth within-subjects variable is the gender of the noun, which also has two levels:

masculine and feminine. A free independent variable in the test design was the non-native subjects’ knowledge of other languages with gender. As can be seen in appendix B, most subjects did know at least one other language with gender (usually French or German). This is potentially problematic in that the aim of this thesis is to consider the adult second language acquisition of gender in Dutch and English learners of Italian. If the subjects have learnt another language before Italian, then, strictly speaking, they are no longer second language learners and predictions from theories of second language acquisition (like FTFA and FFFH) do not hold for them. This is because the acquisition of a language prior to the acquisition of Italian may have influenced the way in which the learners acquired Italian gender. Especially learners who have acquired both opaque and transparent gender systems may thus be at an advantage in comparison to learners who do not have as much experience with

the acquisition of gender. In an ideal situation then, subjects would have no knowledge of a gender language other than Italian. This proved to be an impossible criterion, however, as hardly any Italian learner had not studied another gender language at some point in time. Care was taken, however, to ensure that all learners had acquired the gender 45 Source: http://www.doksinet language after puberty. In the literature, this post-puberty acquisition of a language is assumed to have a much smaller effect on the development of the linguistic faculty than pre-puberty (or child) L2 acquisition (cf. Unsworth 2005 for a review of the differences between child and adult L2 acquisition). Knowledge of another gender language thus may play some role in the observed results, but as the acquisition of this language had occurred after puberty and because most subjects did know another language with gender (so within the pool of subjects there were no great differences in this respect) the

effect may not be that great. Whereas the independent variables are the same for both the SPR experiment and the AJT, the dependent variables do differ in these two tasks. The dependent variable in the SPR experiment is the time it takes the subjects to read each sentence (i.e reading time); the dependent variable in the AJT is the mark given to each sentence. 2.33 Materials Materials consisted of 4 training sentences, 192 experimental stimuli and 62 comprehension questions all presented using the program E-prime (see appendix D for a list of all stimuli). 96 of the 192 sentences were items, the remaining 96 sentences were fillers. For the SPR experiment, all of the sentences were divided into three segments which the subjects could read through at their own pace by pressing the spacebar. The first letter of each first segment was a capital letter; the last letter of the third segment was followed by a full stop. In the AJT, subjects were shown the target sentence as a whole with the

scale visible in the screen at all times, so the subjects did not have to remember which ends of the scale were ‘acceptable’ and ‘unacceptable’. The target sentences were the same in the SPR experiment and the AJT, but different fillers were used in both tasks. Aside from the SPR experiment and the AJT, a language background questionnaire (with one version for the native speakers and another for the non-natives, see appendix A) and a proficiency test (see appendix C) were also completed by the subjects. 2.331 Items In the items in the SPR experiment, the second segment always consisted of the target DP (consisting of a determiner, a noun and an adjective) and nothing else. The first segment usually consisted of a clause or a subject and a verb; the third segment consisted of a few words to wrap up the sentence. In the AJT, the same items were presented as in the SPR experiment, but the items were not divided into three segments. An example of a typical target sentence is shown

in (21), (the three segments present in the SPR experiment are divided by a vertical line): (21) L’alunno ha dato la risposta esatta The pupil has given the fem answer fem correct fem The pupil gave the correct answer to the question alla to the domanda question The DP is the key region in all the target sentences. In order to determine the influence of the nature of the DP on the reading times in the SPR experiment and the marking in the AJT, the DP was presented in four different patterns. The four patterns in which the DP occurred in each sentence are summarised in Table 5: 46 Source: http://www.doksinet Table 5 Pattern types items Pattern Type Det-N-Adj Example la fem risposta fem esatta fem *Det-N-Adj il masc risposta fem esatto masc *Det-N-Adj il masc risposta fem esatta fem Det-N-*Adj la fem risposta fem esatto masc Description determiner and adjective correct determiner and adjective incorrect determiner incorrect, adjective correct determiner correct,

adjective incorrect Out of the 96 items, 24 items were grammatical sentences (with the pattern Det-NAdj) and 72 were ungrammatical (with 24 items in each of the other three patterns). Each target sentence was presented to the subjects in all four forms. Every subject would thus read (22)- (25) at some point in the SPR experiment: (22) L’alunno The pupil ha has dato given la the fem risposta answer fem esatta correct fe alla to the domanda question m The pupil gave the correct answer to the question (23) L’alunno ha dato il risposta esatto The pupil has given the masc answer fem correct masc The pupil gave the correct answer to the question alla to the domanda question (24) L’alunno The pupil ha has dato given il risposta the masc answer fem esatta correct fe alla to the domanda question m The pupil gave the correct answer to the question (25) L’alunno ha dato la risposta esatto The pupil has given the fem answer fem correct masc The pupil gave the correct

answer to the question alla to the domanda question To determine whether learners are consistent in their processing and knowledge of gender (i.e whether the DPs with correct or incorrect gender marking would consistently take the same amount of time to read in the SPR experiment and receive the same mark in the AJT), each DP was presented in two different contexts. Alongside (21), subjects were thus also presented with the following target sentence: (26) Paolo non ha dato la risposta 47 esatta al presentatore Source: http://www.doksinet Paolo not has given the fem answer fem correct fe to the presenter m Paolo did not give the correct answer to the presenter This target sentence was also presented to the subject in all four forms. The order in which the four forms occurred was randomised and interspersed with other target sentences and fillers. Subjects would thus rarely see the same DPs in consecutive sentences. Apart from the pattern of the DP, the nature of

the noun was also varied. In (21)-(26), a feminine regular noun is used (a noun ending in –a). Two other feminine regular nouns were also used (finestra, window and pianta, plant). As well as the three feminine regular nouns, three masculine regular nouns (sogno, dream, treno, train and gatto, cat), three feminine irregular nouns (luce, light, tradizione, tradition and chiave, key) and three masculine irregular nouns (fiore, flower, fiume, river and viaggiatore, traveller) were also used. (See Table 6 for an overview of all the nouns) All of the feminine regular nouns thus end in –a, all of the masculine regular nouns end in –o and all of the irregular nouns end in –e. Again, each of these 12 nouns was used in all four forms and in two different sentences (i.e with two different first and third segments). The nouns used in this experiment are all relatively frequent nouns taken from an Italian frequency dictionary (Juilland and Traversa 1973). One of the three nouns in each of

the four categories (i.e feminine/masculine regular/irregular) came from the second five hundred cohort of most frequent nouns, the other two nouns came from the third and the fourth five hundred cohort respectively. Table 6 Item nouns specified for frequency and regularity Regular Nouns Irregular Nouns Feminine Nouns 2nd 500 3rd 500 finestra risposta (window) (answer) luce tradizione (light) (tradition) 4th 500 pianta (plant) chiave (key) Masculine Nouns 2nd 500 3rd 500 sogno treno (dream) (train) fiore fiume (flower) (river) 4th 500 gatto (cat) viaggiatore (traveller) The selection of the nouns occurred on the basis of frequency and various other factors. No nouns were chosen that had a clear cognate in either English or Dutch The nouns that do bear some resemblance to English words, are also similar to Dutch words (like treno which is train in English and trein in Dutch). Any advantage in understanding the meaning of the nouns was thus equal for both experimental groups

(although it should be noted that the nouns were relatively frequent, so both experimental groups were expected to be familiar with the nouns anyway). No nouns that start with a vowel or the consonant s impure (i.e s followed by a consonant), z, gn, ps or x were chosen to avoid unwanted consequences in the morphological form of the determiner. 28 The determiner in the Det-N-Adj DP was thus always il or la Each of the nouns occurred with a different adjective. All the adjectives that were used were regular (i.e they ended in –a or –o) and relatively frequent in Italian 2.332 Fillers 28 The determiner becomes l’ if it is followed by a word starting with a vowel and lo if it is followed by any of the consonants mentioned above (s impure, z, gn, ps or x). 48 Source: http://www.doksinet Filler sentences were created to resemble the target sentences to some extent, but the filler sentences never contained Det-N-Adj phrases. Because the target sentences are very similar (only the

gender marking is different in each set of four sentences), the fillers also resemble each other closely so that the distinction between fillers and items is somewhat blurred. Different fillers were used in the SPR experiment and the AJT, but the same selection criteria were applied to both. In order to avoid biasing subjects towards expecting grammatical or ungrammatical sentences, subjects were presented with equal numbers of grammatical and ungrammatical sentences. Because three out of four of the items were ungrammatical (only the Det-N-Adj pattern is grammatical, all the other patterns contained at least one incorrectly gender-marked element), most of the fillers had to be grammatical. As 72 of the items were ungrammatical, only 24 of the fillers could be ungrammatical (there were 192 stimuli in total, 96 of which were thus ungrammatical and 96 were grammatical). This means that there were also 72 grammatical fillers and 24 grammatical items. 29 (27) is an example of a typical

grammatical filler: (27) Nel cortile il figlio del calciatore ha In the courtyard the son of the footballer has In the courtyard, the son of the footballer played with the ball giocato played a with The ungrammatical fillers did not contain gender marking errors (apart from the one mentioned in footnote 29), but they did contain a number of other errors like incorrect subject-verb agreement, wrong word order, incorrect tense, regularisation errors and incorrect prepositions. (28) is an example of a typical ungrammatical filler in which the past participle deciduto is a regularisation error; the correct form should be deciso: (28) Lì per lì Mario ha deciduto di Then and there Mario had decided to Then and there, Mario decided to stop smoking smettere stop di to fumare smoke 2.333 Training Sentences Both the SPR experiment and the AJT were preceded by instructions (in Dutch for the Dutch group, in English for the Italian and the English group, as the Italians did not speak any

Dutch) and four training sentences, so that the subjects knew what they were expected to do before the actual experiments started. The training sentences were created on the same basis as the fillers and thus did not contain any Det-N-Adj combinations. For both the SPR experiment and the AJT, two out of four of the training sentences were grammatical and two were ungrammatical. 2.334 Comprehension Questions 29 After the experiments had been conducted, it turned out that a typing error had caused one of the fillers to be ungrammatical. In the actual experiment there were thus 71 grammatical fillers and 25 ungrammatical fillers. Presumably, this extra ungrammatical filler is not enough to bias the subjects towards expecting ungrammatical sentences. A more problematic element of this mistake is that a gender-marking error was made. The masculine noun salotto is preceded by the feminine determiner la instead of the masculine determiner il (although an adjective is not present, so this

filler sentence was not completely like the items). Whilst this was not intended, presumably, this one gender error in the fillers did not significantly alter the processing of the other gender errors in the items. 49 pallone ball Source: http://www.doksinet During both the SPR experiment and the AJT, comprehension questions (in Italian) would occasionally follow grammatical sentences. The comprehension questions would only concern the sentence directly preceding the question (subjects thus did not have to memorise the sentences) and could be answered by yes or no. The comprehension questions were meant to check whether subjects were paying attention to the sentences they were reading. An accuracy measure of these comprehension questions for each subject can be found in appendix E. In order to avoid biasing subjects towards one answer, the answers to the comprehension questions were equally divided between yes and no. For the English and the Italian group, yes could be answered by

choosing the y-key on the keyboard and no could be answered by choosing the n-key on the keyboard. For the Dutch group, yes could be answered by the j-key on the keyboard (for ja in Dutch) and no by the n-key (for nee in Dutch). Two of the four training sentences in the SPR experiment were followed by yes/no comprehension questions, one of which should be answered by yes, the other by no. In the actual SPR experiment, 30 comprehension questions were presented to the subjects. 5 of these comprehension questions followed items, the remaining 25 followed filler sentences. Of the 5 item comprehension questions, 3 should be answered no and 2 should be answered yes. Of the 25 filler comprehension questions, 12 should be answered no and 13 should be answered yes. The number of yes and no answers is thus equally balanced throughout the SPR session. The AJT had a similar set-up: two of the four training sentences were followed by comprehension questions, one of which had to be answered yes, the

other by no. In the actual AJT, 5 out of 30 comprehension questions followed items (but different items than in the SPR experiment), of which 3 should be answered yes and two should be answered no. Of the 25 other questions following fillers, 12 should be answered yes and 13 should be answered no. Throughout both experiments, the answers to the comprehension questions are thus balanced. See Table 7 for an overview: Table 7 Distribution yes/no answers comprehension questions Training Item Filler SPR Yes 1 2 13 No 1 3 12 AJT Yes 1 3 12 No 1 2 13 2.34 Procedure Prior to the actual experiments, subjects were asked to complete a language background questionnaire (see appendix A) before they were given verbal instructions about the two experiments. As this questionnaire was emailed to the participants once they had agreed to participate in the experiment, some participants had already completed the questionnaire beforehand. These subjects would thus immediately receive verbal

instructions from the experimenter about the two experiments. After the verbal instructions, the subjects were asked to read written instructions for the SPR experiment (see appendix F). Once the subjects had read the instructions, they were presented with 4 SPR training sentences, which had exactly the same lay-out as the sentences in the actual experiment. If the subjects had no further questions after the training sentences, they could press the spacebar to start the SPR experiment. 50 Source: http://www.doksinet 2.341 Procedure Self-Paced Reading Subjects were tested on the experimenter’s laptop in a quiet place; the experimenter was present in the same room, but was not facing the participant. Each sentence was presented in three segments; each segment was shown in the middle of the screen. The presentation of the material was self-paced: subjects had to press the spacebar every time they had read one segment and were ready to proceed to the next segment. As soon as the

participants proceeded to the next segment, the previous segment would disappear and the following segment would appear in the middle of the screen (where previously the first segment was present). Subjects thus could not look back at previous segments once they had proceeded on to a following segment. The presentation of the items and the fillers was random. Occasionally, comprehension questions would appear after the subject had read a sentence. The comprehension question would appear in the middle of the screen (in Italian), followed by the English sentence (for the English and the Italian group): please press y for yes and n for no (the Dutch version of this sentence was on the screen for the Dutch group). Subjects could not proceed to the following sentence until they had answered the question and they could not change their answer once they had pressed one of the two answer keys. 2.342 Procedure Acceptability Judgement Task After all the SPR sentences had been read, the subjects

were required to press the spacebar to proceed to the instructions for the AJT (see appendix F). Once the subjects had read the instructions, they could press the spacebar to go to the 4 AJT training sentences. The training sentences had exactly the same lay-out as the sentences in the actual experiment. If the subject had no further questions after the training sentences, the subjects could press the spacebar to go on to the test sentences. Each test sentence was presented in the middle of the screen. At the bottom of the screen, under each test sentence, the following English sentence was present: please assess this on acceptability from 1-6. Under this sentence, a graphic representation of the scale was shown (with the number 1 and the words very unacceptable at the left of the screen and the number 6 and the words very acceptable at the right of the screen). A Dutch version of this sentence was presented to the Dutch group. Again, the distribution of items and fillers was random.

The subjects could not proceed to the following sentence until a mark from one to six had been given. It was also not possible for the subjects to go back to previous sentences once they had arrived at a new sentence. As in the SPR experiment, comprehension questions would occasionally appear in the middle of the screen, followed by the sentence: please press y for yes and n for no. Again, subjects could not proceed to the following sentence until they had pressed one of the two answer keys. Once an answer had been given, the subject could not go back to change it at a later stage. After the AJT had been completed, the non-native speakers (and some of the native speakers) were asked to fill in a proficiency test (see appendix C). This test had to be done with the experimenter present, no subject was allowed to fill in the test at a later point in time. Once the test had been filled in, subjects were either paid 5 euros for their participation or they were given a small present. 51

Source: http://www.doksinet In the next chapter, the results of the experiments will be presented. 52 Source: http://www.doksinet Chapter 3 Results In this chapter, the results of the SPR experiment and the AJT will be provided. The first section presents the subjects’ accuracy on the comprehension questions for both the SPR experiment and the AJT; the next two sections give the results for the SPR experiment and the AJT. 3.1 Accuracy Comprehension Questions As can be seen from Table 8, all groups had an overall accuracy rate of over 80%, with most averages over 90%. Participants thus generally paid attention to the sentences they were reading during the two experiments. Table 8 Accuracy rates comprehension questions Group Native Mean SPR Accuracy 91.7% English 89% Dutch 91.7% Range SPR Accuracy 76.7%100% 83.3%90% 76.7%967% Mean AJT Accuracy 94% 96.7% 97.3% Range AJT Accuracy 80%100% 93.3%100% 93.3%100% Mean Overall Accuracy 92.8% 92.8% 94.5% Range Overall Accuracy

81.7%983% 88.3%967% 85%98.3% 3.2 Results SPR Experiment For the results of the SPR experiment, three different reading times are considered: the reading time of the second segment (RT2), the reading time of the third segment (RT3) and the sum of the reading times for all three segments (RTsum). Analysis of the RTsum is carried out in order to determine whether the differences between the three groups and between the different DP types have an overall effect on reading times. Analysis of the second segment is relevant, because the second segment contains the target region of the experiment (the DP with correct or incorrect gender marking). If there is a direct effect of the nature of the DP on the reading time, it should thus show in the analysis of this segment. It may also be possible, however, that the effect of the nature of the DP is delayed. This would mean that the difference in processing time of the DP is not visible until the third segment. (See Pearlmutter et al. 1999 for an

example of a study in which the ungrammaticality of a segment is not visible until the following segment). The reading times of the third segment are thus also analysed so that any delayed effects can be considered as well. Some of the data in the SPR experiment had to be excluded from analysis due to technical or subject error. In a limited number of cases, the reading times of the segments were improbably long or short. Because it was deemed impossible for any subject to spend less than 300 milliseconds on a segment 30, all data points under 30 Pressing the spacebar as fast as possible without reading the segment takes about 100ms; for reading and understanding a segment a minimum of 200ms was thus maintained. In practice, most subjects did not have any reading times lower than 400ms for a segment and the removed data points were due to 53 Source: http://www.doksinet 300ms were excluded from analysis. This criterion lead to 5 items being removed Based on visual analysis of the

data, excessively long reading times (i.e reading times of more than 6000ms per segment) were also excluded from the data because it was assumed that they reflected temporary inattention of the subject (i.e coughing, asking a question etc.) rather than actual reading time This criterion resulted in 50 deletions In total, these criteria meant 55 items (1.9% of the data) had to be deleted In order to determine which types of tests could be done on the SPR data, Kolmogorov-Smirnov and Levene’s tests were performed. The Kolmogorov-Smirnov tests show that the reading times are not normally distributed, as they always resulted in a p-value of .000 The data from the RTsum, the RT2 and the RT3 are thus not normally distributed either within the three groups or over all three groups combined. Levene’s test for homogeneity of variances shows the same picture: however the data are analysed, within all segments and all groups, the significance level is .000 Again, this means that the data from

the RTsum, the RT2 and the RT3 do not have homogeneous variances either within the three groups or over all three groups combined. The raw data from the SPR experiment can thus not be used in any parametric tests (like an ANOVA or a t-test), as two of the basic assumptions of these tests are that the data is normally distributed and that the variances are homogeneous. A possible solution to the problem of non-normal distribution and variances is to transform the data. The raw data from the SPR experiment was thus transformed into logarithms and the Kolmogorov-Smirnov test and Levene’s test were run again. The transformed data show a more normal distribution and variance than the raw data, but even the transformed data are not completely normal. The Kolmogorov-Smirnov test shows a non-normal distribution for all groups and all RTs except for the logRTsum of the English group (p = .200) and the logRT3 for the English group and the Dutch group (p = .200 and 116 respectively) Levene’s

test of homogeneity of variance shows a similar picture for the transformed data: the variances are only normally distributed for logRT3 (p =.059); the variances are not normally distributed for the other two reading times (p= .041 for logRTsum and p= 022 for logRT2) The overall non-normality of the distribution and variance of the data even in the transformed version thus means that parametric tests on the SPR data cannot be used. Nonparametric tests (namely the Mann-Whitney test) will be used instead in the analysis of the SPR results. 3.21 General Differences Between Groups The Mann-Whitney test shows that there are significant differences between the reading times of the native group, on the one hand, and the combined data of the two non-native groups, on the other, at all three points of analysis (p=.000 and z= -1523 at RT2, p=.000 and z= -1487 at RT3 and p=000 and z= -1869 at RTsum), with the native group consistently processing the segments faster than the two non-native groups.

(See Table 9 for an overview of the p-values and Table 10 for an overview of the raw data. See also appendix G for more information on the raw data of the SPR experiment.) Further Mann-Whitney tests between the native group and each of the two non-native groups confirm this finding. The native group is significantly faster than the English group in processing RTsum, RT2 and RT3 (p=.000 for all three comparisons, z= -17.71 at RTsum, z= -1390 at RT2 and z= -1373 at RT3) The technical problems (holding the spacebar down for too long and thereby skipping a segment, for example). 54 Source: http://www.doksinet native group is also significantly faster than the Dutch group in processing the RTsum, the RT2 and the RT3 (again, p=.000 for all three comparisons, z= -1459 at RTsum, z= -12.47 at RT2 and z= -1202 at RT3) Whereas the native and the non-native groups differ in their processing times, there is hardly any difference between the two non-native groups in their speed of processing.

Mann-Whitney tests show that the Dutch group process the whole sentence faster than the English group (RTsum p=.003, z= -297) There is no significant difference in processing between the two groups when RT2 (p=.164, z= -139) or the RT3 (p=051, z= -1.95) is considered, however, even though the raw data show that the Dutch group process these segments faster as well (see Table 10). The difference between the two non-native groups in the processing of sentences is thus small. Table 9 Group differences SPR, p-values (significant p-values in boldface) Native vs. English Native vs. Dutch Dutch vs. English RT2 p=.000 p=.000 p=.164 RT3 p=.000 p=.000 p=.051 RTsum p=.000 p=.000 p=.003 Table 10 Group differences per group, raw data in ms (M= mean, SD= standard deviation) Native English Dutch Non-Native (combined) RT2 M 1061 1437 1417 1427 SD 601 772 799 786 RT3 M 1108 1550 1481 1516 SD 668 902 850 877 RTsum M 3269 4679 4432 4557 SD 1451 2049 1968 2013 3.22 Influence of the DP There

are several variables within the DP that may influence the participants’ reading times: the grammaticality of the DP (or, more specifically, the pattern type of the DP), the regularity of the noun, and the gender of the noun. In order to test the overall influence of grammaticality on the reading times, a MannWhitney test was conducted in which the data from the grammatical Det-N-Adj pattern was compared with the combined data from the ungrammatical DP patterns *Det-N-Adj, Det-N-Adj and Det-N-Adj (see Table 11 for raw data). The native group showed no effect of grammaticality in their overall reading times (the p-value of RTsum was .624, z= -491), nor was there a delayed effect of grammaticality visible in their reading times for the third segment (the p-value for RT3 was .375, z= -.887) The native group did show a significant effect of grammaticality in their reading times for the second segment (i.e the segment that contained the ungrammaticality) with p=.002 and z= -306 If the

second segment contained an ungrammatical DP pattern, it was thus read more slowly (mean reading time for an ungrammatical second segment is 1079ms) than a second segment with a grammatical DP pattern (mean reading time is 1006ms). The results for the two non-native groups are relatively similar to the native group’s results. In both the non-native groups, there 55 Source: http://www.doksinet is no effect of grammaticality in the RTsum or the RT3, but there is nearly a significant effect of grammaticality in the RT2 (p=.058 and z= -189 for the English group and p=.051 and z= -195 for the Dutch group) The mean reading time for ungrammatical items for the English group is 1465ms compared to 1356ms for grammatical items; the mean reading time for ungrammatical items for the Dutch group is 1447ms compared to 1326ms for grammatical items. Both the Dutch and the English group thus also show a near-significant tendency to read incorrectly gendermarked segments more slowly than correctly

gender-marked segments. Table 11 Mean reading times per group for grammatical and ungrammatical items at RTsum, RT2 and RT3 (SD between brackets) Native English Dutch RTsum gram 3258 (1451) 4758 (2024) 4262 (1782) ungram 3272 (1452) 4653 (2058) 4489 (2024) RT2 gram 1006 (628) 1356 (675) 1326 (726) ungram 1079 (591) 1465 (800) 1447 (821) RT3 gram 1158 (784) 1635 (837) 1459 (786) ungram 1091 (623) 1522 (922) 1489 (871) This overall effect of grammaticality can be analysed more closely by considering the differences in processing times between each different pattern (see appendix G for raw data). Six different comparisons can be made using Mann-Whitney tests: Det-NAdj vs *Det-N-Adj, Det-N-Adj vs. *Det-N-Adj, Det-N-Adj vs. Det-N-*Adj, DetN-Adj vs. *Det-N-Adj, Det-N-Adj vs. Det-N-*Adj and Det-N-Adj vs. Det-N*Adj. Because the same data set is used repeatedly in all six comparisons, the p-value has to be lowered in order to avoid making a type I error. In this case, the normally

significant p-value .05 has to be divided by 6 (a Bonferroni adjustment), which entails that a significant p-value for the comparisons is now .008 Mann-Whitney tests show that within the native group, there are no significant differences between any of the comparisons at RTsum or RT3. The processing times for RT2 do show interesting differences, however (see Table 12 for an overview of the p-values and Table 13 for an overview of the raw data). Within the native group, the difference between Det-NAdj and *Det-N-Adj is significant (p=.000, z= -449), as are the differences between *Det-N-Adj and Det-N-Adj (p=.005, z= -282) and *Det-N-Adj and Det-N-Adj (p=.004, z= -291) The differences between the remaining patterns are not significant The differences in processing times are somewhat different in the two non-native groups (see Tables 12 and 13). The only significant difference in the Dutch group is the comparison between the Det-N-Adj pattern and the *Det-N-Adj pattern in RT2 (p=.008 and

z= -264) No other comparisons are significant in either of the two nonnative groups The effect of grammaticality is thus clearest for RT2, although even in that segment the English group does not show a significant difference. 56 Source: http://www.doksinet Table 12 P-values RT2 per pattern, per group (significant p-values in boldface) Pattern Comparison Det-N-Adj vs. *Det-N-Adj Det-N-Adj vs. *Det-N-Adj Det-N-Adj vs. Det-N-*Adj *Det-N-Adj vs. *Det-N-Adj *Det-N-Adj vs. Det-N-*Adj *Det-N-Adj vs. Det-N-*Adj Native p=.000 English p=.038 Dutch p=.008 p=.130 p=.228 p=.193 p=.133 p=.175 p=.408 p=.005 p=.322 p=.215 p=.004 p=.393 p=.103 p=.916 p=.831 p=.664 Table 13 Raw data RT2 per pattern, per group (M= mean, SD= standard deviation, R= range) Native Det-N-Adj M SD 1006 628 English 1356 675 Dutch 1326 726 R 3444622 4914715 4845747 *Det-N-Adj M SD 1204 719 1524 858 1528 894 R 3014090 4285624 4935596 *Det-N-Adj M SD 1027 530 1435 779 1410 769 R 5533907

4334812 3525304 Det-N-*Adj M SD R 1007 481 3233598 1435 760 4214270 1403 791 4095881 Another aspect of the DP that may influence the processing times is the regularity of the noun. Non-native learners of Italian may take more time processing irregular nouns if they cannot retrieve the gender of the irregular noun quickly enough. Results of the Mann-Whitney test show, however, that neither the native group nor the two non-native groups show a significant difference in processing irregular and regular nouns at any of the three analysis points. The regularity of the noun thus does not affect the processing time in the RTsum, RT2 or RT3 in the native, the English or the Dutch group (p> .116 for all measure points in all groups) Mann-Whitney tests also show that the gender of the noun does not affect the processing of the SPR sentences at any of the three points of analysis for any of the three groups (p> .133 for all measure points). In other words, regular nouns are no more

difficult to process than irregular nouns or vice versa, and masculine nouns are no more difficult to process than feminine nouns or vice versa. (See appendix G for raw data of the processing times of regular, irregular, masculine and feminine nouns at RTsum, RT2 and RT3.) 3.3 Results AJT Kolmogorov-Smirnov tests for normal distribution and Levene’s test for homogeneity of variance were conducted on the AJT data in order to determine whether parametric tests could be performed. The results for both were highly significant, whichever way the data were grouped together (p≤.021 for all analyses) It can thus be concluded that the data from the AJT is not normally distributed and that the variances are not homogeneous. Transformation of the data also did not change the situation: both when the marks were transformed into logarithms and when the scale was changed to 57 Source: http://www.doksinet an ordinal scale (instead of the 6-point scale of the raw data), all p-levels remained

highly significant (p=.000 for all analyses) Parametric tests can thus not be conducted on the AJT data; non-parametric tests must be used instead. 3.31 General Differences Between Groups In order to determine whether the native and the two non-native groups differed in their judgements on the AJT, Mann-Whitney non-parametric tests were used (for the means, standard deviations and ranges of the AJT data see appendix H). As an assumption of this test is that the dependent variable is measured on an ordinal scale, the raw data from the AJT had to be transformed. An ordinal scale was thus created for the marks given to both grammatical and ungrammatical items (MarkAll) and two scales were created for the grammatical (MarkGram) and the ungrammatical (MarkUngram) items separately. These three scales were necessary so that the differences between the groups and the influence of the DP could be considered across all of the data and separately for the grammatical and the ungrammatical items.

Mann-Whitney tests showed that there was a significant difference between the marks assigned by the native group and the marks assigned by the two non-native groups in all conditions (see Table 14 for p-values and Table 16 for raw data). When all the marks of the grammatical and the ungrammatical items were combined, the native group gave significantly lower marks than the non-native groups (mean mark 3.32 vs 4.26 with p=000 and z= -1297) The same result is observed when only the ungrammatical items are considered: the natives give significantly lower marks than the non-natives (mean mark 2.48 vs 377 with p=000 and z= -1998) If only the grammatical items are considered, there is also a significant difference between the natives and the non-natives, but this time the natives assign significantly higher marks than the non-natives (mean mark 5.83 vs 573 with p=003 and z= -297) The natives thus seem to differentiate more clearly between the grammatical and ungrammatical items. Further

Mann-Whitney tests allow the difference between each of the three groups to be compared as well (see Table 15 for p-values and Table 17 for raw data). If the marks from the grammatical and the ungrammatical items are combined, the marks assigned by the native group differ significantly from both the English group and the Dutch group. The native group assigned significantly lower marks than the English group (native mean mark is 3.32, English mean mark is 421 with p= 000 and z= 1111) The marks assigned by the native group were also significantly lower than the marks assigned by the Dutch group (Dutch mean mark is 4.31 with p= 000 and z= 1137) The difference between the two non-native groups is not significant, however (p=.200 and z= -128) An analysis of the ungrammatical items only presents a similar picture: the native group differs from both the English and the Dutch groups 31, with the native group assigning significantly lower marks than the English or the Dutch group (native mean

mark is 2.48; English mean mark is 375; Dutch mean mark is 3.80) Again, there is no significant difference between the English and the Dutch group with respect to the marks assigned to the ungrammatical items (p=.182 and z= 134) The marks from the grammatical items indicate that there is a significant difference between the native group and the English group (p=.000 and z= -381; native mean mark 5.83, English mean mark is 562) There is no difference between 31 Native~English comparison: p=.000, z= -1708 ; native~Dutch comparison: p=000, z= -1756 58 Source: http://www.doksinet the native group and the Dutch group (p=.154 and z= -143, Dutch mean mark is 5.84), however, and there is a difference between the English and the Dutch group (p=.010 and z= -258) Table 14 Native vs. non-native, p-values (significant values in boldface) Native vs. NonNative MarkAll p=.000 MarkGram p=.000 MarkUngram p=.003 Table 15 Group differences AJT, p-values (significant values in boldface) Native

vs. English Native vs. Dutch Dutch vs. English MarkAll p=.000 p=.000 p=.200 MarkGram p=.000 p=.154 p=.010 MarkUngram p=.000 p=.000 p=.182 Table 16 Native vs. non-native, mean mark (SD between brackets) Native Non-Native All 3.32 (193) 4.26 (142) Grammatical 5.83 (070) 5.73 (075) Ungrammatical 2.48 (141) 3.77 (124) Table 17 Group differences specified per group, mean mark (SD between brackets) Native English Dutch All 3.32 (193) 4.21 (144) 4.31 (140) Grammatical 5.83 (069) 5.62 (092) 5.84 (051) Ungrammatical 2.48 (141) 3.75 (127) 3.80 (121) 3.32 Influence of the DP As in the SPR experiment, there are several variables that may influence the participants’ marking in the AJT: grammaticality/pattern type of the DP, regularity of the noun and the gender of the noun. Mann-Whitney tests show that there is a very clear effect of grammaticality of the DP in all three groups’ judgements (see Table 17 for raw data): ungrammatical items receive significantly lower marks than

grammatical items (p=.000 and z= -2180 for the native group, p=000 and z= -1809 for the English group and p=.000 and z= -2181 for the Dutch group) The precise nature of this effect can be specified further if the four different pattern types are considered as well. Mann-Whitney tests show that within the native group, the Det-N-Adj pattern is consistently given a significantly higher mark than the *Det-N-Adj pattern (p=.000; z= -19.53), the *Det-N-Adj pattern (p=.000; z= -1692) and the Det-N-*Adj pattern (p=.000; z= -1920) The native group give a significantly lower mark to the *Det-NAdj pattern in comparison to the Det-N-Adj pattern (p=.000; z= -387) and the DetN-*Adj pattern (p=.000; z= -330) The only comparison that does not have a significant difference in mark-attribution is the comparison between the *Det-N-Adj 59 Source: http://www.doksinet pattern and the Det-N-*Adj pattern (p=.290; z= -106) (See Table 18 for an overview of the p-values and Table 19 for details of the raw

data). The influence of pattern is somewhat different for the English group, however. The English group are like the native group in that the Det-N-Adj pattern is consistently marked higher than the *Det-N-Adj pattern (p=.000; z= -1497), the *Det-N-Adj pattern (p=.000; z= -1565) and the Det-N-*Adj pattern (p=.000; z= -1506) Like the native group, the English group also show no significant difference in their marking of the *Det-N-Adj pattern and the Det-N-Adj pattern (p=.222; z= -122) The main difference between the native group and the English group is in the comparison between the *Det-N-Adj pattern and the Det-N-Adj and the Det-N-Adj pattern. Whereas the native group give the patterns with one gender error higher marks than the pattern with two gender errors, the English group do not make a significant difference between the two (p=.262 and z= -112 for the *Det-N-Adj vs. *Det-N-Adj comparison and p=.918 and z= -10 for the *Det-N-Adj vs. Det-N-*Adj comparison). The Dutch group is

quite similar to the native group with respect to the influence of pattern type on their marking. Like the native group, the Det-N-Adj pattern receives significantly higher marks than the *Det-N-Adj pattern (p=.000; z= -1846), the *Det-N-Adj pattern (p=.000; z= -1845) and the Det-N-*Adj pattern (p=.000; z= -18.21) The *Det-N-Adj pattern also receives significantly lower marks than the *Det-N-Adj (p=.008; z= -267) and the Det-N-*Adj pattern (p=.000; z= -432) Again, like the native group, the difference in mark given to the *Det-N-Adj and the Det-NAdj pattern is not significant (p=.104; z= -163) Table 18 P-values specified per group for the pattern comparisons AJT (significant values in boldface) Pattern Comparison Det-N-Adj vs. *Det-N-Adj Det-N-Adj vs. *Det-N-Adj Det-N-Adj vs. Det-N-*Adj *Det-N-Adj vs. *Det-N-Adj *Det-N-Adj vs. Det-N-*Adj *Det-N-Adj vs. Det-N-*Adj Native p=.000 English p=.000 Dutch p=.000 p=.000 p=.000 p=.000 p=.000 p=.000 p=.000 p=.000 p=.262 p=.008

p=.001 p=.918 p=.000 p=.290 p=.222 p=.104 60 Source: http://www.doksinet Table 19 Means (M), standard deviations (SD) and ranges (R) per pattern, per group AJT Det-N-Adj M SD Native 5,83 0,69 English 5,62 0,92 Dutch 5,84 0,51 R 1-6 1-6 2-6 *Det-N-Adj M SD R 2,13 1,11 1-5 3,78 1,31 1-6 3,58 1,17 1-6 *Det-N-Adj M SD 2,82 1,72 3,66 1,22 3,81 1,26 Det-N-*Adj M SD R 2,49 1,23 1-6 3,80 1,28 1-6 4 1,18 2-6 R 1-6 1-6 2-6 In most cases, the regularity of the noun does not influence the mark given by any of the three groups. Within the native group, there is no overall effect of the regularity of the noun (MarkAll, p=.193 and z= -130) and neither is there an effect of regularity if only the marks given to grammatical items are considered (MarkGram, p=.627 and z=-.49) If only ungrammatical items are considered (MarkUngram), however, the influence of the regularity of the noun is nearly significant (p=.051 and z= -196), with irregular nouns receiving higher marks (mean is 2.58)

than regular nouns (mean is 2.39) The results for the English group are similar to those of the native group: there is no effect of regularity if both grammatical and ungrammatical items are considered (MarkAll p=.121, z= -155) or if only grammatical items are considered (p=.135; z= -150) The effect of regularity is significant if only the ungrammatical items are taken into account though (p=.013; z= -248) with irregular nouns receiving higher marks (mean is 3.87) than regular nouns (mean is 362) The Dutch group shows no effect of regularity of the noun at all (MarkAll p=.992, z= -01; MarkGram p=.774, z= -29 and MarkUngram p=885, z= -14) The p-values of these results are summarised in Table 20; the raw data is presented in Table 21: Table 20 Effect regularity of noun (significant p-values indicate a difference in marking for regular vs. irregular nouns) MarkAll MarkGram MarkUngram Native p=.193 p=.627 p=.051 English p=.121 p=.135 p=.013 Dutch p=.992 p=.774 p=.885 Table 21 Mean

mark and SD of regular and irregular nouns per group (SD between brackets, boldface represents a significant difference) Native English Dutch MarkAll Reg 3.25 (1.94) 4.14 (1.43) 4.31 (1.40) MarkGram Reg Irreg 5.85 5.82 (.67) (.72) 5.72 5.53 (.76) (1.05) 5.86 5.82 (.40) (.61) Irreg 3.39 (1.91) 4.29 (1.45) 4.31 (1.40) MarkUngram Reg Irreg 2.39 2.58 (1.37) (1.44) 3.62 3.87 (1.20) (1.33) 3.79 3.80 (1.22) (1.21) The gender of the noun does not have a great influence on the mark in any of the three groups. The only significant effect of gender is present in the native group when only the grammatical items are considered (p=.047; z= -199), with feminine nouns receiving higher marks (mean is 5.93) than masculine nouns (mean is 573) None of 61 Source: http://www.doksinet the other groups show a significant effect of gender in their marking. For an overview of the p-values results, see Table 22; for raw data, see Table 23: Table 22 Effect of gender per group (significant p-values

indicate differences in reading time for masculine and feminine nouns) MarkAll MarkGram MarkUngram Native p=.337 p=.047 p=.313 English p=.793 p=.552 p=.925 Dutch p=.812 p=.964 p=.681 Table 23 Mean mark and SD of feminine and masculine nouns per group (SD between brackets, boldface represents a significant difference) Native English Dutch MarkAll Fem 3.36 (1.91) 4.23 (1.42) 4.29 (1.41) MarkGram Fem Masc 5.93 5.73 (034) (.91) 5.68 5.57 (.82) (1.01) 5.85 5.83 (.46) (.56) Masc 3.27 (1.95) 4.20 (1.46) 4.32 (1.39) MarkUngram Fem Masc 2.51 2.45 (1.37) (1.45) 3.74 3.75 (1.24) (1.30) 3.78 3.82 (1.22) (1.21) 3.33 Consistency As the same DPs were always embedded in two different sentences (see section 2.33), it was possible to determine to what extent the subjects were consistent in their rating of the DP. Consider examples (29)- (32): (29) L’alunno ha dato la risposta esatta alla The pupil has given the fem answer fem correct fem to the The pupil gave the correct answer to the

question domanda question (30) Paolo non ha dato la risposta esatta al Paolo not has given the fem answer fem correct fem to the Paolo didn’t given the correct answer to the presenter presentatore presenter (31) *La bambina ha dato la fiore gialla a The girl has given the fem flower masc yellow fem to The girl gave the yellow flower to her mother sua her madre mother (32) *Il coniglio ha mangiato la fiore gialla nella sua The rabbit has eaten the fem flower masc yellow fem in its The rabbit ate the yellow flower in its cage 62 gabbia cage Source: http://www.doksinet In (29) and (30), the same Det-N-Adj DP is presented in its two different versions; in (31) and (32) the two versions of the same *Det-N-Adj DP are shown. In order to determine the level of consistency for each of the three groups, the data was examined for inconsistent judgements with respect to the two versions of each sentence. If, for example, a subject rated the grammatical (29) with a 6, but the equally

grammatical (30) with a 1, this was considered an inconsistent judgement. Another example of an inconsistent judgement would be if the subject rated the ungrammatical (31) with a 5 and the equally ungrammatical (32) with a 2. As the scale has 6 points, an inconsistent judgement was defined as follows: one sentence of a DP pair receives a mark of three or lower, whilst the other sentence receives a mark of four or higher. There was one exception to this rule: if one sentence received the mark four and the other was marked three, this was not considered an inconsistent judgement. The following combinations are thus inconsistent: 6-1, 6-2, 6-3, 5-1, 5-2, 5-3, 4-2 and 4-1. No other combinations were considered to be inconsistent judgements. As there were 10 subjects per group and each subject could make 48 inconsistent judgements (there were 96 items in total, so 48 sentence pairs), 480 inconsistencies were possible per group. For the native group, 38 of these inconsistencies were counted.

This means that 79% (38/480 x 100) of the native group’s judgements were inconsistent, so their consistency measure was 92.1% For a more specific breakdown of these inconsistencies, we can consider the pattern types in which these inconsistencies occurred. Five of the 38 (132%) inconsistencies occurred in the Det-N-Adj pattern, 1 (2.6%) occurred in the *Det-N-Adj pattern, 22 ( 57.9%) in the *Det-N-Adj pattern and 10 (26.3%) in the Det-N-*Adj pattern. We can also consider the regularity and gender of the noun in the inconsistencies: 18 (47.4%) out of the 38 inconsistencies occurred when the DP contained a regular noun; 20 (52.6%) inconsistencies occurred when the noun in the DP was irregular. 12 (316%) out of the 38 inconsistencies occurred when the noun in the DP was feminine and 26 (68.4%) when the noun was masculine. The English group was quite similar to the native group in their inconsistencies: they had 37 inconsistencies in total, which entails an overall inconsistency rate of

7.7% (37/480 x 100) and a consistency measure of 92.3% 6 (162%) of the 37 inconsistencies occurred in the Det-N-Adj pattern, 9 (24.3%) in the *Det-N-Adj pattern, 8 (21.6%) in the *Det-N-Adj pattern and 14 (37.8%) in the Det-N-*Adj pattern. 18 (486%) out of the 37 inconsistencies occurred when the noun in the DP was regular; 19 (51.4%) inconsistencies were counted when the noun was irregular 22 (59.5%) of the 37 inconsistencies occurred when the noun in the DP was feminine; 15 (40.5%) when the noun was masculine The Dutch group was quite different from both the native group and the English group in its inconsistencies: this group only had 12 inconsistencies, so their inconsistency rate was 2.5% (12/480 x 100) and their consistency measure was 975% Of these 12 inconsistencies, 2 (16.7%) were present in the Det-N-Adj pattern, the *Det-N-Adj pattern and the *Det-N-Adj pattern, respectively. The remaining 6 (50%) inconsistencies occurred in the Det-N-*Adj pattern. 2 of the inconsistencies

were counted when the DP contained a regular noun; 10 (83.3%) when the DP had an irregular noun. The pattern for the inconsistencies regarding gender of the noun is the 63 Source: http://www.doksinet same: 2 inconsistencies occurred when the DP contained a feminine noun, the remaining 10 occurred when the DP had a masculine noun. For an overview of this data, see Table 24: Table 24 Number of inconsistencies per group specified for pattern, regularity and gender Det-N- *DetAdj N*Adj Native 5 1 English 6 9 Dutch 2 2 *DetN-Adj Det-N- Regular Irregular Feminine Masculine *Adj N N N N 22 8 2 10 14 6 18 18 2 20 19 10 12 22 2 26 15 10 3.4 Relation SPR and AJT Data As the items are the same in the SPR experiment and in the AJT, it is possible to investigate whether there is a relation between the processing time for an item and the mark that is given to that item. A Spearman’s correlation test 32 was carried out to determine whether the marks given in the AJT and the reading

times were correlated. The test considered the relation between the RTsum and the mark (because the mark was given for the whole sentence) and the RT2 and the mark (because the RT2 is the crucial reading time in the processing of gender). The relation between the RTsum and the mark was only significant for the native group (p=.000), but the correlation was very weak (Spearman’s rho= -.114) As Spearman’s rho is negative, this means that the total reading time increased when the marks decreased: lower marks thus entailed higher overall reading times. The correlation is very weak, however, so the total reading time and the marks are not clearly related. The two non-native groups did not show any relation between RTsum and their marks. The Spearman’s rho for RT2 was significant for both the native group (p= .000) and the English group (p=.016), but not for the Dutch group (p=784) As the correlation coefficient is only -.202 for the native group and -078 for the English group, the

correlation between the RT2 and the mark is weak for the native group and very weak for the English group. The marks and the reading times thus do not seem to correlate strongly in any of the groups with respect to the RTsum or the RT2. See Table 25 for an overview of the p-values and the Spearman’s rho for each group: Table 25 Correlation reading times and marks per group (significant p-values in boldface) Native English Dutch RTsum p-value .000 .373 .849 RT2 Spearman’s rho p-value -.114 .000 -.029 .016 -.006 .784 32 Spearman’s rho -.202 -.078 .009 As the data were not normally distributed (see sections 3.2 and 33 for further information), a Pearson’s correlation test could not be performed. The Spearman’s correlation test was used instead as this test does not assume that the data are normally distributed. This test assumes that the variables are on an ordinal scale, so the marks were transformed accordingly (the same transformation was performed for the Mann-Whitney

and Kruskal-Wallis tests previously). 64 Source: http://www.doksinet 3.5 Individual Data SPR and AJT One question that has not been discussed in the previous sections is whether there are any individuals in the non-native groups who are indistinguishable from the natives. In order to determine this, Mann-Whitney tests were done in which the performance of the group of natives was pitted against the performance of each individual nonnative. These Mann-Whitney tests showed that there was one Dutch individual (subject number 2 in the Dutch group 33) whose performance was very similar to the performance of the native group in all the SPR and the AJT analysis points. The pvalues and z-scores of this individual were as follows (the dependent variable is thus compared between the native group and the one Dutch subject): RTsum p= .042, z= -2.04; RT3 p= 183, z= -133; RT2 p= 036, z= -209; MarkAll p=544, z= -61; MarkGram p=.146 , z= -145 and MarkUngram p=517, z= -65 As all of the comparisons

between this individual and the native group have a p-level higher than .01 (and most have are higher than 05), the differences between the natives and this individual were generally not significant and hence the performance of this subject can be considered practically nativelike. No other individual is indistinguishable from the natives in both the SPR experiment and the AJT, but there are a number of individuals that do not differ significantly from the natives in one of the two tasks. One of the English subjects (subject number 5) and one of the Dutch subjects (subject number 4) do not differ significantly from the natives in the AJT. The p-values of these two subjects are as follows: MarkAll p=.241, z= -117 for the English individual and p= 538, z= -62 for the Dutch subject; MarkGram p=.525, z= -64 for the English subject and p=937, z= -08 for the Dutch; MarkUngram p=.124, z= -154 for the English subject and p=264, z= -112 for the Dutch individual. As all of the comparisons have

p-values above 05, these two subjects can thus be considered nativelike regarding their performance on the AJT. There are two other subjects (one from the English group, number 7, and one from the Dutch group, number 6) that are very nearly in the native range for the AJT. These two subjects both have one analysis point that is just significant and thus excludes them from complete nativelike performance. One of the English subjects (number 11) does not differ significantly from the native group with respect to the SPR experiment. The p-values for this subject are as follows: RTsum p=.320, z= -100; RT3 p=447, z= -76; RT2 p=671, z= -42 As all the p-values are above .05, this subject can be considered within the native range for the SPR experiment. There is one other Dutch subject (number 9) that very nearly falls into the native range for the SPR experiment as well, but this subject has a pvalue of .013 (z=-247) for RT3 The remaining individuals do show a nativelike performance for some

of the data points for some of the tasks, but overall their performance is quite different from that of the natives. In Table 26 an overview is given of the native means and standard deviations for all analysis points of the SPR experiment and the AJT and the means and standard deviations for the non-native subjects that fall within the native range: 33 See appendix B and E for subject information for this particular subject and other subjects mentioned in the rest of this section. 65 Source: http://www.doksinet Table 26 Mean and standard deviations natives and individual learners (SD between brackets, N/A = not applicable) Native English #1 English #2 Dutch #1 Dutch #2 RT2 1061 (601) N/A RT3 1108 (668) N/A RTsum 3269 (1451) N/A 1238 (916) 1019 (658) N/A 1314 (1020) 1101 (774) N/A 3767 (2186) 3138 (1668) N/A MarkAll 3.32 (1.93) 3.43 (1.78) N/A MarkGram 5.83 (.70) 5.96 (.20) N/A MarkUngram 2.48 (1.41) 2.58 (1.16) N/A 3.08 (1.71) 3.24 (1.62) 5.79 (.51) 5.92 (.28) 2.18

(.72) 2.35 (.51) 3.6 Summary In the SPR data, there is a significant difference between the processing times of the native group compared to the non-native group, but there is not much difference between the processing times of the two non-native groups. The influence of the grammaticality of the DP on the processing times is quite similar for both the native and the non-native group. All groups tend to take more time to process the ungrammatical segments, especially if both the determiner and the adjective are incorrectly gender-marked. The regularity and gender of the noun do not greatly influence the processing times of the items in any of the three groups. A similar picture was obtained for the AJT data. There is a significant difference in the judgements of the native and the non-native groups, but there is not much difference between the two non-native groups. All of the groups assign higher marks to the completely grammatical items in comparison to the completely ungrammatical

items, but there is some difference between groups for the other comparisons. As in the SPR experiment, the regularity and gender of the noun do not greatly influence the marks given to the items in any of the three groups. There are four non-native subjects that fall in the native range for either the AJT or the SPR experiment and there is one Dutch learner who falls into the native range for both experiments. The other learners are in the native range in their performance on some elements of the two tasks, but overall their performance is quite different to the performance of the natives. In the next chapter, these results will be interpreted and discussed more extensively. 66 Source: http://www.doksinet Chapter 4 Discussion This chapter will interpret and discuss the results presented in the previous chapter. The first section consists of a summary of the results in relation to the research questions posed in Chapter 2; in the second section, the results are interpreted in a

broader perspective, and in the third section directions for further research are explored. 4.1 Results and Research Questions The answers to the research questions posed in Chapter 2 can now be considered. The research questions of this thesis are the following: 1. Do non-native learners of Italian differ with respect to native speakers of Italian in: a) their knowledge of gender? b) their processing of gender? 2. Are there differences between the English and the Dutch learners of Italian with respect to: a) their knowledge of gender? b) their processing of gender? 3. If there are differences between the native speakers and the non-native learners of Italian, are there any individuals in the Dutch or the English groups who perform in the native range with respect to: a) their knowledge of gender? b) their processing of gender? 4. To what extent do the subjects’ reading times and acceptability judgements depend on: a) the grammaticality/pattern type of the DP b) the regularity of the

noun c) the gender of the noun In the next sections, each research question will be discussed in light of the predicted and actual outcomes. The discussion of the research questions will start with the fourth research question, so that the specific differences between the natives and the non-natives regarding the different DP types can be incorporated in the discussion of the more general research questions 1, 2 and 3. 4.11 Research Question 4: Influence of the DP The fourth research question looks at the influence of the nature of the DP in the reading times and marks of the subjects. Does the grammaticality of the DP, the pattern type of the DP, the regularity of the noun and the gender of the noun influence the results of the two experiments? 67 Source: http://www.doksinet 4.111 The Influence of Grammaticality and Pattern The predictions for this part of research question 4 were as follows (repeated from section 2.215): 6. Natives: the grammaticality/pattern type of the DP

should influence both the reading times in the SPR experiment (ungrammatical DPs are expected to take more time than grammatical DPs) and the marks given in the AJT (ungrammatical DPs are expected to receive lower marks than grammatical DPs). 8. Non-natives: it is expected that the grammaticality/pattern type of the DP will influence the marks if the noun in the DP is regular. In case of regular nouns, the non-natives will thus give higher marks to grammatical sentences than to ungrammatical sentences. 9. Non-natives: the grammaticality/pattern type of the DP is not expected to influence the marks when the noun in the DP is irregular. 10. Non-natives: it is not clear whether the grammaticality of the DP will influence the reading times of the non-natives. Following the FFFH, the English group will not show an influence of grammaticality, but the Dutch group might. FTFA would claim that both groups may show an effect of grammaticality. The results for the natives were generally in line

with the predictions. Native speakers assigned significantly lower marks to ungrammatical items (mean mark was 2.48) than to grammatical items (mean mark was 5.83), showing that they clearly distinguished between grammatical and ungrammatical gender marking on DPs. For the reading times, the picture is slightly more nuanced: the natives did not show an effect of grammaticality when the reading times of the whole sentence were considered. The RTsum thus was not significantly larger if the sentence contained an ungrammatical DP in comparison to a grammatical DP. There was also no delayed effect of grammaticality in the reading times of the sentences. In other words: the RT3 was not larger if the preceding segment contained an ungrammatical DP. There was an effect of grammaticality if the reading times for the second segment were considered, however. There was thus a local effect of ungrammaticality of the DP in the reading times of the natives, with incorrectly gender-marked DPs taking

longer to read than correctly gender-marked DPs. The results for the non-natives differed from the predictions in some places. The nonnatives assigned significantly lower marks to ungrammatical items (mean mark is 3.77) than to grammatical items (mean mark is 573) It was predicted that this difference would only be visible in the case of regular nouns and that the marking of grammatical and ungrammatical sentences would not differ when sentences contained irregular nouns. This was not the case, however, as the non-natives distinguished between grammatical and ungrammatical items, irrespective of the regularity of the noun. Like the natives, the non-natives thus also clearly distinguish between correct and incorrect gender marking in DPs. The results of the non-natives in the SPR experiment are relatively similar to the natives’ results. Like the natives, the non-natives did not show an effect of grammaticality when the reading times of the whole sentence were taken into account, nor

did they show any delayed effects of grammaticality in their reading times for the third segment. The picture becomes slightly less clear when the reading times for the second segment are considered. Neither of the two non-native groups showed a 68 Source: http://www.doksinet significant difference between their reading times of grammatical and ungrammatical segments, but the difference was close to significance, especially for the Dutch group (the Dutch group had p=.051, the English group had p= 058) The non-natives thus tend to read ungrammatical items more slowly than grammatical items and hence show something of a gender congruency effect, but the difference is not significant. The effect of grammaticality on the reading times and the marks can be considered more closely if the four different patterns are compared. The reading times and marks can be compared six different ways (Det-N-Adj vs. *Det-N-Adj, Det-N-Adj vs. *Det-N-Adj, Det-N-Adj vs. Det-N-*Adj, Det-N-Adj vs.

*Det-N-Adj, Det-N-Adj vs. Det-N-*Adj and Det-N-Adj vs. Det-N-*Adj). The native group show a significant difference in marking for all comparisons except the *Det-N-Adj vs. DetN-*Adj comparison. DPs with no gender-marking errors are thus consistently given higher marks than DPs with gender-marking errors and DPs with one gender-marking error are consistently given higher marks than DPs with two gender-marking errors. This result also shows that native speakers are not more sensitive to gender-marking errors on the determiner than errors in the adjective or vice versa (as there is no significant difference between the marks given to the *Det-N-Adj and the Det-N-Adj pattern). The Dutch group falls exactly in line with the native group with respect to their marking: all of the comparisons that are significant for the native group are also significant for the Dutch group. The sensitivity to gender marking mistakes in the input is thus qualitatively similar for the Dutch and the native

group, even though there is a quantitative difference between the marks assigned by the Dutch group and the native group. In other words, although the natives give significantly lower marks to ungrammatical items and significantly higher marks to grammatical items than the Dutch group, the Dutch group nevertheless make the same relative distinctions in their marking of grammatical and ungrammatical items. The English group, on the other hand, is somewhat different. For the English group, only the following comparisons are significant: Det-N-Adj vs. *Det-N-Adj, Det-NAdj vs. *Det-N-Adj and Det-N-Adj vs. Det-N-*Adj. The English group thus gives significantly higher marks to DPs without gender errors than to DPs that do contain gender-marking errors, but they do not make a distinction between DPs with two gender-marking errors in comparison to DPs with one gender-marking error. As the English group is sensitive to correct vs. incorrect gender marking, this difference between the English

group and the two other groups does not entail that the English group is insensitive to gender mistakes in the input or that their knowledge of gender is deficient. After all, the key element in a learner’s knowledge of gender is that they know what constitutes correct and incorrect gender marking. As the correct sentences do receive significantly higher marks than the incorrect sentences then, the English group does know what constitutes correct gender marking. If we look at the effect of pattern type in the reading times of the natives and the nonnatives, a clearer difference between the native and the non-native groups can be observed. The native group show a significant difference in reading times for the following comparisons: Det-N-Adj vs. *Det-N-Adj, Det-N-Adj vs. *Det-N-Adj and *Det-N-Adj vs. Det-N-*Adj. The direction of the difference is always the same: DPs with no gender-marking errors are read faster than DPs with two gender-marking errors and DPs with one gender-marking

error are read faster than DPs with two gender-marking errors. Interestingly, DPs with no gender-marking errors are not read 69 Source: http://www.doksinet any faster than DPs with one gender-marking error. It is not clear why there is no effect of grammaticality in these comparisons, but possibly the native speakers were reading the segment too fast to notice one mistake. DPs with one gender-marking error may then have been mistaken for grammatical DPs, which would explain why there is a difference between DPs with one grammatical error and DPs with two grammatical errors. The non-native groups performed quite differently to the natives in their reading times. For the Dutch group, only the difference between the Det-N-Adj and the *DetN-Adj pattern was significant (DPs with no gender mistakes were read faster than DPs with two gender mistakes). There was no difference in reading times for any of the other comparisons. The English group did not have any significant comparisons The

non-native groups thus do not show as clear a difference in their processing of grammatical and ungrammatical DPs as the natives. Nonetheless, the trend in the data is that completely grammatical DPs are processed more quickly than completely ungrammatical DPs, so the difference between the natives and the non-natives may be quantitative more than qualitative. This trend goes against what is found in the study by Guillelmon and Grosjean (2001). In that study, the non-native learners do not show any difference in the processing of DPs that are correctly or incorrectly gender marked. It should be noted, however, that Guillelmon and Grosjean only consider the gender congruency effect in –gen L1 learners, not in +gen L1 learners. In the experiments in this thesis, a tendency towards processing grammatical sentences faster than ungrammatical ones is found in the English group, but significant results are only found for the Dutch group. In order to determine whether the non-natives, and

especially English non-natives, show a proper effect of grammaticality in their reading times, a larger study should be conducted. Possibly, ten non-native participants in each group were not enough to find a significant result for the nonnatives. 4.112 The Influence of Regularity and Gender of the Noun The predictions for the influence of the nature of the noun on the marks and reading times were the following: 7. Natives: the regularity and gender of the noun are not expected to influence the reading times or the marks given by the native speakers. 11. Non-natives: the regularity of the noun is predicted to influence the marks and the reading times. Sentences with regular nouns are predicted to receive high marks and short reading times if the gender marking is grammatical and low marks and long reading times if it is not. This difference in marking and reading times is not expected to occur for sentences with irregular nouns. 12. Non-natives: the gender of the noun is not predicted

to influence the marks or the reading times. The results of the experiments were somewhat in contrast to the predictions regarding the regularity of the noun. The regularity of the noun was shown not to have a great influence on any of the groups in either of the experiments. In the SPR experiment, none of the groups showed any differences in reading times for DPs with regular nouns in comparison to DPs with irregular nouns. The non-natives thus did not have longer reading times for irregular nouns compared to regular nouns in contrast to the predictions and to the findings in Taraban and Kempe (1999) for L2 learners of Russian. It seems then that regular and irregular nouns (or at least the regular and 70 Source: http://www.doksinet irregular nouns used in this study) are not processed differently by the non-natives. In contrast to Taraban and Kempe’s assumption that L2 learners do not have immediate access to the gender feature of irregular nouns, the learners do not take

longer to process irregular nouns and hence do seem to be able to access the gender feature of irregular nouns as fast as the gender feature of regular nouns. It may be possible, however, that this result was obtained because the irregular nouns used in this study are quite frequent in Italian. Less familiar irregular nouns may indeed lead to the processing problems found in Taraban and Kempe (1999). Further testing with different and less frequent irregular nouns could show whether this claim is correct. In the AJT, the Dutch group did not show any effect of regularity in their judgements. The natives and the English group, on the other hand, tended to give irregular nouns higher marks than regular nouns if the sentence was ungrammatical (this effect was significant for the English group and nearly significant for the natives). It is not clear why this effect was only present in the ungrammatical items or why it is present only in these two groups. One possible explanation for this

finding is that these two groups were somewhat unsure about the gender of the irregular nouns and hence gave them more borderline marks (i.e marks of 3 and 4) than the ungrammatical regular nouns There are several reasons to believe that this explanation does not hold, however. In the first place, it seems unlikely that the natives would be unsure about the gender of the irregular nouns, especially as the irregular nouns were relatively frequent in Italian. That this explanation is also invalid for the English group, becomes clear when the distribution of inconsistencies for the two groups are considered. The number of inconsistent judgements (which would indicate uncertainty about the gender of the noun) are not greater for irregular nouns than for regular nouns in either of the two groups, so it seems unlikely that the difference in marks is down to confusion about the true gender of the irregular noun. It seems then that the native and the English group are better at recognising

ungrammaticality in regular nouns than in irregular nouns, but why this should be the case is not clear. A replication of this study could show whether this finding is robust or incidental in this study. The prediction that the gender of the noun would not influence the performance of the subjects was upheld. In the SPR experiment, there was no effect of the gender of the noun on the reading times for any of the groups. Feminine nouns and masculine nouns are thus evidently equally easy to process. In the AJT, the native group did give significantly higher marks to feminine nouns if only the grammatical items were taken into account. The non-natives did not differentiate between feminine and masculine nouns at all in their marking. Again, why the native group would give grammatical sentences with feminine nouns higher marks than grammatical sentences with masculine nouns is not clear. Presumably, this result does not indicate general differences between the status of feminine and

masculine nouns, as otherwise the difference should have been present in the ungrammatical items and in the SPR experiment as well. On the other hand, the consistency measure does show that the natives are a lot less likely to give inconsistent judgements when the noun is feminine (12 inconsistent judgements) than when the noun is masculine (26 inconsistent judgements). Again, a replication of this study would have to show whether there is a structural difference between masculine and feminine nouns in native speakers. 71 Source: http://www.doksinet 4.12 Research Question 1: Natives vs Non-Natives The following predictions were made regarding the difference between the native and the non-native performance in the two experiments: 1. No difference between natives and non-natives for regular nouns in the AJT 2. Native speakers and non-native learners differ in judgements for irregular nouns in the AJT. 3. Natives and non-natives differ in their reading times in the SPR experiment The

results of the AJT show that the marks that the native speakers give to the sentences differ significantly from the marks that the non-native speakers give in all categories (i.e MarkAll, MarkGram and MarkUngram) This is not entirely in line with the predictions as it was thought that the marks would only differ when sentences with irregular nouns were considered. As it turned out, the regularity of the noun did not influence the marking, so the difference between the native and the non-native marks cannot be attributed to problems with the irregular nouns. The general picture is that native speakers give grammatical items higher marks and ungrammatical items lower marks than the non-natives do. The natives thus differentiate more clearly between the grammatical and the ungrammatical items than the non-natives. It may be premature to assume that this means that the non-natives did not know the gender of the nouns or the related forms of the determiner and the adjective, however. What

is clear is that the natives consider gender-marking errors to be much more serious errors than the non-natives do. If the natives encounter a sentence with incorrect gender marking, they will generally give it a mark in the unacceptable range of the scale (i.e 3 or below, the mean mark for incorrectly gender-marked sentences is 2.48) When non-natives encounter incorrect gender marking, however, they often give it a mark at the low end of the acceptable range (i.e a mark around 4, the mean mark is 3.77 for ungrammatical items) That the non-natives are not completely insensitive to gender marking is also shown by the fact that the difference between the marks given to the grammatical items and the ungrammatical items is significant for both the non-native groups (with ungrammatical items receiving lower marks than grammatical items). The non-native speakers thus do systematically differentiate between correct and incorrect gender marking, but they do not consider a sentence with gender

mistakes to be at the ‘very unacceptable’ end of the scale. This conclusion is supported by the consistency measure for the two non-native groups. If the non-natives were confused about the gender of the nouns and gender marking, their judgements should have been very inconsistent. This is not what is found, however; in fact, it is the natives who have the largest number of inconsistent judgements, not the non-natives. The knowledge of gender (i.e knowing the gender of a noun and which forms the determiner and the adjective should take as a result of that gender) is generally similar in the natives and the non-natives, but whereas the natives consider gender-mistakes to be serious errors, the non-natives consider them to be more minor mistakes. The results of the SPR experiment show that the reading times of the natives and the non-natives are significantly different. The natives are faster than the non-natives at all points of analysis. From a general point of view then, it can be

concluded that the natives are faster at processing the segments than the non-natives. This does not necessarily say anything fundamental about the nature of the gender processing of the non-natives, however. If DPs with gender marking are only processed faster because 72 Source: http://www.doksinet sentences are just generally processed faster by natives, then this is not saying anything specific about the processing of gender. The interesting results for this study concern the way the second segment is processed by the native and the non-native subjects. The natives clearly show that segments with incorrect gender marking are read more slowly than segments with correct gender marking; a finding that is in line with the gender congruency effect. If the non-natives show this effect as well, it would be plausible to conclude that their processing of gender is similar to native gender processing. It is predicted, however, that native processing of gender differs from non-native

processing, so this effect is not expected. The results of this study give a somewhat ambivalent picture of this issue: if the reading times from all the incorrectly gender-marked items are grouped together and pitted against the reading times from the correctly gender-marked items, the nonnatives show a near-significant effect of grammaticality. If the reading times for the completely correct pattern (Det-N-Adj) are compared to the reading times of the completely incorrect pattern (*Det-N-Adj), we see that the Dutch group do show a significant difference, but the difference is not significant for the English group. In other words, the non-natives have a tendency to read incorrectly gender-marked items more slowly than items with correct gender marking, but this tendency does not always reach significance in the results. One way to determine whether non-natives are comparable to natives in their processing of gender would be to replicate the present study with a larger number of

non-native participants, as it is possible that the relatively small number of participants was not enough to obtain significant results. If a study with a larger number of participants does not show an effect of gendermarking on the reading times, then it is likely that there is no gender congruency effect for non-natives. 4.13 Research Question 2: English vs Dutch The prediction for this research question is the following: 4. Whether the performance of the English group will differ from the performance of the Dutch group is unclear. The FFFH predicts that the Dutch group will outperform the English group; FTFA predicts that the performance of the two groups will be similar. In the AJT, the Dutch group and the English group generally gave similar marks, both differentiating clearly in their marks between grammatical and ungrammatical items. The average mark for all the items did not differ significantly between the two groups and neither did the mark if only ungrammatical items were

considered. Both nonnative groups tended to give higher marks than the native group in these two points of analysis. If only the grammatical items were considered, however, the Dutch group did differ from the English group. Like the native group, the Dutch group assigned significantly higher marks to the grammatical items than the English group did. The performance of the native group and the Dutch group did not differ when only grammatical items were considered. When the pattern comparisons are considered, another difference between the two non-native groups appears. Whereas the patterns in five out of six of the comparisons receive significantly different marks in both the native group and the Dutch group, the English group shows a significant difference only in three out of the six comparisons. 34 34 For the Dutch and the native group the following comparisons receive significantly different marks: Det-N-Adj vs. *Det-N-Adj; Det-N-Adj vs. *Det-N-Adj; Det-N-Adj vs. Det-N-*Adj;

Det-N-Adj vs. 73 Source: http://www.doksinet The only other difference between the two groups is that the English group shows an effect of the regularity of the noun in the marking of ungrammatical items. The Dutch group shows no effect of regularity. On these three fronts then (ie the marking of grammatical items in general, the marking of regular ungrammatical items and in the pattern comparisons), the performance of the Dutch group differs from that of the English group. In all other points of analysis in the AJT, the performance of the Dutch and the English group is not significantly different. In the SPR experiment, the performance of the two non-native groups is also quite similar. The reading times for both the second and the third segment do not differ significantly between the two groups. The Dutch group does show a significantly faster reading time for the whole sentence, however. As mentioned in the previous section, any interesting differences in gender processing are

likely to turn up in the analysis of the reading times in the second segment. Neither of the two non-native groups show a significant effect of grammaticality in their processing of the second segment, but the p-value of the Dutch group is closer to significance than the p-value of the English group (p=.051 vs p=058 respectively) Whereas both groups thus show a near-significant tendency to read incorrectly gender marked segments more slowly than correctly gender marked segments, this tendency is closer to significance for the Dutch group than for the English group. This difference becomes clearer when the pattern types are compared. The Det-N-Adj pattern is read significantly faster than the *Det-N-Adj pattern by the Dutch group, but this difference is not significant for the English group. The two groups are also similar in that the regularity and gender of the noun do not influence reading times at all. Overall then, there is no great difference between the performance of the two

groups; the marks given by the two groups only differ in four places 35 and the reading times only differ in two places 36. The two groups thus differed in six places, out of a total of 44 possible differences 37.The fact that the Dutch group can potentially use deep transfer of the gender feature whereas English cannot thus does not seem to cause any fundamental differences between the two groups. After all, if there was a fundamental difference between the two groups, this should have manifested itself in significant differences in most points of analysis in both tasks. *Det-N-Adj and Det-N-Adj vs. Det-N-*Adj (with patterns with more gender errors receiving lower marks than patterns with less gender errors). For the English group, the following comparisons received significantly different marks: Det-N-Adj vs. *Det-N-Adj; Det-N-Adj vs. *Det-N-Adj and Det-N-Adj vs. Det-N-*Adj (with patterns with gender errors receiving lower marks than patterns without gender errors). 35 Namely: in

the marking of grammatical items, in the pattern comparison *Det-N-Adj vs. *Det-NAdj, in the pattern comparison Det-N-Adj vs. Det-N-*Adj and the influence of regularity in the marking of ungrammatical items. 36 Namely: in the reading time for the RTsum and in the significant difference for the Dutch group between the Det-N-Adj vs. *Det-N-Adj comparison. 37 Differences were possible in 22 different places in the AJT: 3 in the overall marking (in markall, markgram and markungram), 1 for the influence of grammaticality, 6 for the influence of pattern, 3 for the influence of regularity (in markall, markgram and markungram) and 3 for the influence of gender (in markall, markgram and markungram). In the SPR experiment, 22 differences were also possible: 3 in overall reading times (RTsum, RT2 and RT3), 1 for the influence of grammaticality, 6 for the influence of pattern, 3 for the influence of regularity (in RTsum, RT2 and RT3) and 3 for the influence of gender (RTsum, RT2 and RT3). 74

Source: http://www.doksinet 4.14 Research Question 3: Individuals The predictions for this question are stated in 5.: 5. A small number of non-natives will fall into the native range regarding their performance on the AJT and the SPR experiment. Whether these nativelike nonnatives will only be in the Dutch group, as the FFFH would claim, or whether there will also be English individuals who are nativelike, as FTFA would predict, remains to be seen. The results of the study show that there is one Dutch subject who was very similar to the natives in her performance. Her marks in all analysis points of the AJT did not differ at all from the native marks (with p-values ≥ .146 for the three points of analysis); her reading times were also very similar to the natives’ times although the RTsum and the RT2 times were significant at the .05 level (042 and 036 respectively). As the analysis of the RT2 is the most important from the point of view of gender processing, there is some doubt as

to whether her gender processing is completely nativelike (as the RT2 p-value is significant at the .05 level), but her scores are definitely very similar to the natives. If not completely nativelike then, her knowledge and processing of gender are very similar to that of the natives. None of the other individuals come close to performing within the native range for both of the experiments, but there are a number of individuals that fall in the native range for one of the two tasks. Two of the non-native subjects perform completely in the native range on the AJT. At none of the analysis points do their marks differ from the native marks (p-values are all ≥ .124), so their knowledge of gender, or, more accurately, the extent to which they find gender errors acceptable, is completely nativelike. One of the English subjects performs completely in the native range on the SPR experiment. None of her reading times in the three measure points differ from the reading times of the natives

(p-values all ≥ .320), so her processing of gender is nativelike. There is one Dutch subject who is also very similar to the natives in his gender processing. One of the analysis points is significantly different from the natives (his RT3 has a p-value of .013), but his reading times in the second segment, the crucial segment for gender processing, do not differ from the natives. His gender processing is thus also very similar to native gender processing. As predicted then, there are a few natives that fall in the native range for at least one of the tasks. The results of this study also show that these individuals can come from both the English and the Dutch group. Nativelike knowledge of gender and/or processing of gender, as considered in this study, can thus be attained by advanced learners of Italian. Interestingly though, nativelike sensitivity to gender errors does not necessarily entail nativelike processing or vice versa. These capacities are apparently capable of developing

somewhat independently. That a dissociation between marking and reading times is possible is also shown by the correlation scores for the marks and the reading times: in both non-native groups, higher marks do not necessarily entail faster reading times and low marks do not entail slower reading times. 4.15 Summary of Research Questions In short, then, the research questions can be answered as follows: the grammaticality and pattern type of the DP influence the reading times and marks of the natives very clearly. The natives consistently show significant differences in their reading times 75 Source: http://www.doksinet and marks for correctly gender marked items in comparison to incorrectly gender marked items. The grammaticality and pattern type of the DP also clearly influence the marks of the non-natives, but whether their reading times are also affected by this factor is less clear. Both non-native groups tend to take longer to read ungrammatical items in comparison to

grammatical items, but this effect is not always significant. The regularity and gender of the noun do not generally influence the reading times and marks of any of the three groups to a great extent, although in the AJT some effects were found for some of the analysis points. The non-natives do not differ significantly from the natives in their knowledge of gender. They generally know what the gender of the noun is and how this should be reflected in the forms of the determiner and the adjective, as shown by their differentiation between grammatical and ungrammatical items and by their consistency measure. The natives and the non-natives do differ in the extent that they consider gender-marking mistakes to be unacceptable, however. Whereas the nonnatives do not consider gender mistakes to be serious errors (sentences with gender errors were generally given marks at the low end of the acceptable part of the scale), the natives felt that sentences with gender errors were highly

unacceptable. Whether the natives and the non-natives differ with respect to the processing of gender is less clear. The natives show a significant difference in their reading times of correctly and incorrectly gender marked items; the non-natives generally only showed a nearsignificant tendency to read items with gender errors more slowly than items without gender errors. Further research with more non-native participants thus will have to determine whether the near-significant difference will end up significant or not significant. The two non-native groups generally do not differ in their performance on the two tests. There are, however, some elements in which the two groups differ The Dutch group differentiates slightly more clearly between correct and incorrect gender marking in both the AJT and the SPR experiment. If the difference between the two groups is significant at any points of analysis then, the Dutch group’s performance tends to be more similar to the performance of

the native group. It is not the case, however, that the individuals that perform within the native range on the tests are only in the Dutch group. Both the Dutch and the English group have a more or less equal number of individuals that perform within the native range on the AJT or the SPR task. 4.2 The Results in a Broader Perspective Apart from answering the research questions, the results can also be considered in a broader perspective. In this section, the results will be presented in light of the FTFA vs. FFFH debate and general issues in second language acquisition that are related to this discussion. This section also considers the role of processing and knowledge in language acquisition in general and in second language acquisition more specifically. 4.21 FFFH vs FTFA As discussed in Chapters 1 and 2, two of the main theories of second language acquisition, the FFFH and the FTFA, make opposing predictions regarding the outcome of this study. The FFFH predicts that the Dutch

group will consistently outperform the English group on both of the tests. As the Dutch group can transfer the gender feature from Dutch to Italian, the FFFH expects advanced Dutch learners of 76 Source: http://www.doksinet Italian to have no problems in acquiring Italian gender. The English group, on the other hand, is expected to experience considerable problems in the acquisition of Italian gender, as the lack of a gender feature in their L1 makes it impossible for them to acquire gender properly. Whereas –gen L1 learners with a good memory may be able to acquire the gender of nouns in isolation, the lack of a gender feature in the L1 will make it impossible for them to process and use gender agreement in a nativelike manner. As learners who come from a +gen L1 background are capable of acquiring gender, the FFFH also predicts that it is possible for individual subjects in the Dutch group (and potentially even the Dutch group as a whole) to perform within the native range for

both the tests. As the –gen L1 English learners of Italian are incapable of acquiring the gender feature in Italian, it is also impossible that any individual subject in the English group would be able to perform within the native range on either of the tests. FTFA, on the other hand, does not predict any difference in the performance of the English and the Dutch group. According to FTFA, the gender feature of Dutch may indeed put the Dutch learners at an advantage in comparison to the English learners at the initial stages of language learning, but this difference should disappear as the exposure to Italian allows the English learners to create a gender feature for Italian. As long as the learners have had enough exposure to Italian then, they should in principle be able to acquire Italian gender. This does not necessarily entail that all advanced learners of Italian will be completely indistinguishable from native Italians in their use of gender, however. After all, learners from

both –gen and +gen L1 backgrounds have to create new settings for the L2 gender. The –gen L1 learners have to create a new grammatical feature and the +gen L1 learners will have to change elements of their L1 gender system to be in line with the L2 gender system (to give an example: adverbs may be gender marked in the L2 but not in the L1, a +gen L1 learner would thus have to change her settings in this respect to be in line with the L2). Only if the learner has had enough qualitatively good exposure, so that these new settings can be created, will it be possible for the learner to be nativelike in L2 gender. Whereas FTFA thus does not predict that all learners will necessarily be nativelike in their use of L2 gender, it does not preclude that there will be individual subjects that have nativelike L2 gender. The results of this study are generally in line with the FTFA predictions. Contrary to FFFH prediction, the difference between the non-native groups was not significant for

most points of analysis. Generally, the two non-native groups performed remarkably similarly on both tests. It should be noted, however, that if the difference between the two groups was significant, it was generally the Dutch group that would perform more like the natives. In this sense then, the FFFH is correct in predicting that the Dutch group’s performance is more similar to that of the native group than the English group’s performance. The FFFH prediction that it is impossible for –gen non-natives to perform within the native range is not upheld, however. Even though there are no individuals in the English group that perform completely in the native range in both experiments, one of the English subjects performs within the native range on the AJT and another English subject performs in the native range in the SPR task. As the FFFH would deem nativelike gender agreement and processing of gender to be impossible for –gen L1 learners, the fact that there are English subject

that perform within the native range is completely against FFFH predictions. The results regarding the individual performances are in line with FTFA predictions, however. The results 77 Source: http://www.doksinet of this study thus clearly favour the FTFA theory as applied to gender over the FFFH theory. Within the scope of this study (i.e gender as considered via acceptability judgements and reading times), the FFFH take on L2 acquisition of gender is thus falsified. This may mean that other predictions offered by the FFFH with respect to gender and other grammatical features will turn out to be false as well. If it is not impossible for learners from a –gen L1 background to acquire the elements of gender considered in this study, then it may also not be impossible for –gen L1 learners to master the complete grammatical gender system (i.e comprehension and production of gender) to a nativelike level. If a new feature like gender can be acquired in an L2, then it seem

plausible that other new grammatical features can be acquired as well. Learners from an L1 without number or subject-verb agreement may thus also be able to acquire number or subject-verb agreement in an L2. The idea that puberty cuts off whole linguistic populations from being able to acquire any new language to a nativelike level may thus be going too far. This debate relates to a more general issue in linguistics: is it possible to become nativelike in a language learnt after puberty or is nativelike attainment only possible if the language is acquired during a certain developmental period? The FFFH assumes that nativelike acquisition is only possible during a limited timeframe, the FTFA assumes that nativelike acquisition is possible if the learner receives enough good quality input. If, as this study would seem to show, there is no biological reason that would make nativelike second language acquisition impossible, then it becomes interesting to consider the differences between

individuals that lead one individual to be nativelike in an L2 and the other to fall outside the native range. Presumably, both internal and external factors will play a role: the amount and quality of exposure to the L2 will be important, but a learner’s motivation to learn and general verbal ability will undoubtedly also play an important role (see Hyltenstam and Abrahamsson 2003 for a review). If a biological explanation can be discarded, research into which factors can control nativelike attainment will become increasingly important. 4.22 Processing and Knowledge in Language Acquisition The results of this study have direct bearing on another issue in linguistics more generally: the importance of processing and knowledge in language acquisition. In acquiring any element of language, it is essential to have the relevant knowledge for that language: knowing which words express which concepts and how to put the words together in order to express ideas in a grammatical way is a key

element of successful language acquisition. Having the necessary lexical, syntactic, morphological and phonological knowledge is not enough, however, to be fully proficient in a language. It is also necessary to be able put all these different elements of knowledge together in real time. The acquisition of gender is a case in point in this respect: it is not enough just to know the gender of the noun and how this is reflected in the forms of other words; the lexical, morphological and syntactic information has to be integrated fast enough so that correct gender agreement can be achieved in the use of the language. The results of this study show that whereas the knowledge of gender is not that problematic for most of the non-natives (i.e the non-natives generally know what the gender of the noun is and how this should be reflected in the determiner and the 78 Source: http://www.doksinet adjective), their processing of gender is not nativelike in all respects. The native speakers

show a clear difference in the reading times of DPs with grammatical and ungrammatical gender marking. This difference is also seen in the non-natives, but it is less pronounced and only visible when completely correct DPs are compared to completely incorrect DPs (i.e the Det-N-Adj vs the *Det-N-Adj pattern). The nonnatives are thus showing some kind of gender congruency effect, but it is not as clear as the one shown by the natives. This study thus fits in with a number of other studies that have been mentioned in previous chapters (Serratrice et al. 2004, Tsimpli et al 2004, Sorace 2005 and Hulk and Cornips 2006) that claim that problems in second language acquisition are primarily due to processing problems, not problems in knowledge. This result is also reminiscent of the study by Sabourin and Haverkort (2003) discussed in Chapter 1: like the German learners of Dutch in their study, it is not the case that the non-natives are radically different from the natives in their processing

of gender mistakes. The Germans show a P600 effect when processing the gender marking anomalies, just like the natives do, but the difference lies in the nature of the P600 effect (the Germans had a more restricted P600 that started later than the natives’ P600 effect). A similar result is found in this study: it is not that the non-natives do not process or realise the differences between DPs with grammatical and ungrammatical gender marking, their response to gender mistakes just isn’t as clear as the natives’ response. In contrast to what is implied by theories like the FFFH and other theories which assume there is a critical period for language acquisition, there does not seem to be a defining moment in the child’s development at which it becomes impossible to acquire a second language to native speaker standards. The difference between native speakers and advanced non-native speakers is not radical, it is one of degrees. Still, the question remains: why is it so hard to

become nativelike in a second language? If there is no absolute biological reason why this is not possible, why then are there so few individuals that become indistinguishable from natives in their use of language? Does the problem lie in a gradual decrease of processing skills? Is this an across-theboard phenomenon, in other words: are all our processing skills affected by this deterioration or is the deterioration specific to language processing? If the deterioration is specific to language processing, why is it specific to language? 4.3 Directions for Future Research The questions posed in the previous paragraph could be addressed in future research. Especially if other studies like this study show that the difference between first and second language acquisition is not absolute, then it will be interesting to see what factors hinder the second language learner in acquiring the second language to native standards. Before that can be done, however, it is necessary to replicate this

study with a larger number of subjects. In the first place, it would be interesting to know if some of the near-significant results in this study will become significant if more subjects are considered. Will the non-natives show an equally clear effect of grammaticality as the natives if more non-natives are tested? Will the difference between the reading times of the Det-N-Adj and the *Det-N-Adj pattern become significant for the English group if more English subjects are considered? Will the Dutch group still tend towards the natives more than the English group? The existing study could also be changed and extended in other ways in order to gain more insights into the nature of L2 acquisition of gender. The role of transfer in 79 Source: http://www.doksinet acquiring a second language could be considered even better if a third group of nonnatives was added to the test. In the present study, only deep transfer was possible: the Dutch group may have been able to transfer their

gender feature to Italian, but the two systems are too dissimilar to make surface transfer possible. In a new study, the performance of, for example, Spanish learners of Italian could be considered as well, so that the possible effects of surface transfer on the acquisition of gender could be considered. Would these subjects, who can use both surface transfer and deep transfer, perform significantly better than the Dutch subjects? Would this group show a clear gender congruency effect like the natives in this study? The result of this study would also be interesting from the point of view of the FFFH/ FTFA debate: if there is still a clear distinction between the native speakers of Italian, on the one hand, and the three groups of non-natives, on the other, then this would cast doubt on the FFFH. After all, if even learners with language backgrounds very similar to Italian are not performing any better than learners from –gen L1 backgrounds, then it seems unlikely that transfer of

features has much to do with a learner’s capacity to acquire a language. If, on the other hand, the Spanish and the Dutch learners of Italian differ significantly from the English learners, then the transfer of features would seem a plausible explanation. Another interesting modification to the present experiments would be to consider the acquisition of Dutch gender in English and Italian learners of Dutch. Will there be a difference in the capacities of learners to acquire the gender feature of a language with an opaque gender system in comparison to a language with a transparent gender system? This new study would also show whether the extra assumptions added to the FTFA (discussed in Chapter 1) are valid: is it really the case that the moment of resetting in the acquisition of a second language depends on the nature of the feature that is being acquired? Are transparent, predominant features acquired at an earlier stage than opaque features? This study could also show what is

easier for learners: acquiring a transparent system coming from an opaque system or the other way round. Are Dutch learners of Italian thus comparatively better at acquiring Italian gender than Italian learners of Dutch or vice versa? This may in turn also shed light on the processing problems in second language acquisition discussed in the previous section. Does it matter whether the different elements of knowledge that have to be integrated are transparent or opaque? Is it easier to process transparent elements of a language or is it purely the process of integration that is problematic, regardless of the nature of the system underlying it? A less profound change in the study might be to look at –gen L1 learners of a language with gender that do not have English as their L1. To my knowledge, all studies that have considered the acquisition of gender in an L2 by –gen L1 learners have always looked at English learners. This choice may be skewing the results, however. After all, if

only English learners are ever looked at, it is impossible to determine whether the results can be generalised to other learners from a –gen L1 background. Further studies might thus look at the performance of, for example, Turkish learners of gender languages. Many questions thus remain to be answered in the study of L2 acquisition of gender. Within the limited scope of this study, however, the research questions have been answered. Although there are significant differences between the natives and the nonnatives, both groups of non-natives seem to have been able to acquire the elements of 80 Source: http://www.doksinet Italian gender as considered in this study relatively successfully. 81 Source: http://www.doksinet References Bartning, I. (2000) Gender Agreement in L2 French: Pre-advanced vs Advanced Learners. Studia Linguistica 54, pp 225-237 Bottari, P., P Cipriani and AM Chilosi (1993/1994) Protosyntactic Devices in the Acquisition of Italian Free Morphology. Language

Acquisition 3, pp 327-369 Bottari, P., P Cipriani, AM Chilosi and L Pfanner (1998) The Determiner System in a Group of Italian Children with SLI. Language Acquisition 7, pp 285-315 Bruhn de Garavito, J. and L White (2000) L2 Acquisition of Spanish DPs: The Status of Grammatical Features. In S Catherine Howell et al (eds) BUCLD 24 Proceedings, Cascadilla Press: Somerville, MA. Carstens, V. (2000) Concord in Minimalist Theory Linguistic Inquiry, 31 pp 319355 Chomsky, N. (1995) The Minimalist Program MIT Press, Cambridge Comrie, B. (1999) Grammatical Gender Systems: A Linguist’s Assessment Journal of Psycholinguistic Research, Vol. 28, pp 457-466 Corbett, G.C (1991) Gender Cambridge University Press, Cambridge Dewaele J. and D Véronique (2001) Gender Assignment and Gender Agreement in Advanced French Interlanguage: a Cross-sectional Study. Bilingualism: Language and Cognition 4, pp. 275-297 Franceschina, F. (2005) Fossilized Second Language Grammars The acquisition of grammatical

gender. John Benjamins PLC, Amsterdam/ Philadelphia Garnham, A., J Oakhill, MF Ehrlich and M Carreiras (1995) Representations and Processes in the Interpretations of Pronouns: New Evidence from Spanish and French. Journal of Memory and Language, 34 pp 41-62 Guillelmon, D. and F Grosjean (2001) The Gender Marking Effect in Spoken Word Recognition: The Case of Bilinguals. Memory and Cognition 29, pp 503-511 Gunter, T.C, AD Friederici and HJ Schriefers (2000) Syntactic Gender and Semantic Expectancy: ERPs Reveal Early Autonomy and Late Interaction. Journal of Cognitive Neuroscience 12, pp. 556-568 Gurjanov, M., G Lukatela, J Moskovljevic, M Savic and M Turvey (1985) Grammatical Priming of Inflected Nouns by inflected Adjectives. Cognition, 19 pp. 55-71 Hagoort, P. and CM Brown (1999) Gender Electrified: ERP Evidence on the Syntactic Nature of Gender Processing. Journal of Psycholinguistic Research, Vol. 28, pp 715-728 Hawkins, R. and F Franceschina (2004) Explaining the acquisition and

nonacquisition of determiner-noun gender concord in French and Spanish In P Prévost and J. Paradis (eds), The acquisition of French in different contexts Focus on functional categories, pp. 175-205 John Benjamins, Amsterdam Hohlfeld, A. (2006) Accessing Grammatical Gender in German: The Impact of Gender-marking Regularities. Applied Psycholinguistics, 27 pp 127-142 Holmes, V.M and B Dejean de la Bâtie (1999) Assignment of Grammatical Gender by Native Speakers and Foreign Learners of French. Applied Psycholinguistics 20, pp. 479-506 Hulk, A. and L Cornips (2006) Neuter Gender Determiners and Interface Vulnerability in Child L2/2L1 Dutch. In: S Unsworth, T Parodi, A Sorace and M. Young-Scholten (eds), Paths of Development in L1 and L2 acquisition Amsterdam: John Benjamins. 82 Source: http://www.doksinet Hyltenstam, K. and N Abrahamsson (2003) Maturational Constraints in SLA In C Doughty and M.H Long (eds) The Handbook of Second Language Acquisition, pp. 539-588 Oxford: Blackwell

Jacobsen, T. (1999) Effects of Grammatical Gender on Picture and Word Naming: Evidence from German. Journal of Psycholinguistic Research 28, pp 499-514 Juilland, A.G and VP Traversa (1973) Frequency Dictionary of Italian Words Mouton, Den Haag. Karmiloff-Smith, A. (1979) A Functional Approach to Child Language Cambridge University Press, Cambridge. Kupisch, T., N Müller and KF Cantone (2002) Gender in Monolingual and Bilingual First Language Acquisition: Comparing Italian and French. Lingue e Linguaggio 1, pp. 107-150 Oliphant, K. (1998) Acquisition of Grammatical Gender in Italian as a Foreign Language. The Canadian Modern Language Review 54, pp 239-261 Pearlmutter, N.J, SM Garnsey and K Bock (1999) Agreement Processes in Sentence Comprehension. Journal of Memory and Language, 41 pp 427-456 Sabourin, L. and M Haverkort (2003) Neural substrates of representation and processing of a second language. In R van Hout, A Hulk, F Kuiken and R Towell (eds.), The lexicon-syntax interface in

second language acquisition, pp 175-195. John Benjamins, Amsterdam Sabourin, L., LA Stowe and GJ de Haan (2006) Transfer Effects in Learning a Second Language Grammatical Gender System. Second Language Research 22, pp. 1-29 Schriefers, H. and JD Jescheniak (1999) Representation and Processing of Grammatical Gender in Language Production: A Review. Journal of Psycholinguistic Research, Vol. 28, pp 575-600 Serratrice, L., A Sorace and S Paoli (2004) Crosslinguistic Influence at the SyntaxPragmatics Interface: Subjects and Objects in English-Italian Bilingual and Monolingual Acquisition. Bilingualism: Language and Cognition 7, pp 183-205 Sorace, A. (2005) Selective optionality in language development In L Cornips and K. Corrigan (eds), Syntax and variation Reconciling the biological and the social. John Benjamins, Amsterdam Taraban, R. and V Kempe (1999) Gender Processing in Native and Nonnative Russian Speakers. Applied Psycholinguistics 20, pp 119-148 Tsimpli, I., A Sorace, C Heycock

and F Filiaci (2004) First Language Attrition and Syntactic Subjects: A Study of Greek and Italian Near-native Speakers of English. International Journal of Bilingualism, Vol 8, pp 257-277 Unsworth, S. (2005) Child L2, Adult L2, Child L1: Differences and similarities: A study on the acquisition of direct object scrambling in Dutch. PhD dissertation, Utrecht University. Van Berkum, J.JA (1997) Syntactic Processes in Speech Production: The Retrieval of Grammatical Gender. Cognition 64, pp 115-152 Vigliocco, G., T Antonini and MF Garrett (1997) Grammatical Gender is on the Tip of Italian Tongues. Psychological Science 8, pp 314-317 Vigliocco, G. and J Franck (1999) When Sex and Syntax Go Hand in Hand: Gender Agreement in Language Production. Journal of Memory and Language 40, pp 455-478. Weerman, F., J Bisschop and L Punt (2003) L1 and L2 Acquisition of Dutch Adjectival Flexion. Paper presented at Generative Approaches of Language Acquisition 2003, Utrecht. 83 Source:

http://www.doksinet White, L. E Valenzuela, M Kozlowska-Macgregor, I Leung and H Ben Ayed (2001). The Status of Abstract Features in Interlanguage: Gender and Number in L2 Spanish. In Anna HJ Do et al (eds) BUCLD 25 Proceedings, pp 792-802 White, L. (2003) Second Language Acquisition and Universal Grammar Cambridge University Press, Cambridge. White, L., E Valenzuela, M Kozlowska-MacGregor and YI Leung (2004) Gender and Number Agreement in Nonnative Spanish. Applied Psycholinguistics 25, pp 105-133. Zonneveld, W. (1996) De Verwerving van Morfologie” In: J Don, Inleiding Morfologie. Coutinho, Muiderberg 84 Source: http://www.doksinet Appendix A: Language Background Questionnaires Questionnaire Natives: 1) Name: 2) Sex: 3) Date of birth: 4) Nationality: 5) Telephone number: 6) Email address 7) Education (specify highest level): Language Background 1) Have you been able to speak any language(s) other than Italian from birth? If so, please specify which language(s). 2) How long ago

did you last live in Italy? 3) Do you ever go back to Italy? If so, how often and for how long? 4) At what age did you start studying English? 5) How well do you: Very well Fairly well Not well Not at all i understand spoken English? ii speak English? iii read English? iv write English? 6) Do you think living abroad has influenced your Italian? If so, in which way? For instance, do you have word finding difficulties when you speak Italian? Do you use English constructions in your Italian? Do you have a bit of an English accent when you speak Italian? etc. 7) Do you know any language(s) other than Italian or English? Please specify. 8) How did you acquire these languages? In a classroom setting, from native speakers or both? Please specify for each language. 9) For each language, please fill in the following table: 85 Source: http://www.doksinet Other Language 1 (OL1): How well do you: Very well Fairly well Not well Not at all Very well Fairly well Not well Not at all

Very well Fairly well Not well Not at all i understand spoken OL1? ii speak OL1? iii read OL1? iv write OL1? Other Language 2 (OL2): How well do you: i understand spoken OL2? ii speak OL2? iii read OL2? iv write OL2? Other Language 3 (OL3): How well do you: i understand spoken OL3? ii speak OL3? iii read OL3? iv write OL3? Questionnaire Non-Natives: 1) Name: 2) Sex: 3) Date of birth: 4) Nationality: 5) Telephone number: 6) Email address 7) Education (specify highest level): Language Background Italian 1) Mother tongue (if bilingual or multi-lingual from birth please specify all languages): 86 Source: http://www.doksinet 2) How well do you: Very well Fairly well Not well Not at all i understand spoken Italian? ii speak Italian? iii read Italian? iv write Italian? 3) Do you have any Italian friends, relatives or flatmates (or any other acquaintances)? How often do you speak Italian with them? 4) At what age did you start studying Italian? 5) For how long have you been

studying Italian? 6) Have you ever lived in Italy? If so, please specify for how long: 7) Do you: i watch Italian-language television? ii. read Italian-language newspapers? iii. read Italian-language books iv. listen to Italian-language radio yes yes yes yes no no no no Language Background Other Languages 8) Do you know any language(s) other than your mother tongue or Italian? Please specify. 9) How did you acquire these languages? In a classroom setting, from native speakers or both? Please specify for each language. 10) For each language, please fill in the following table: Other Language 1 (OL1): How well do you: Very well Fairly well i understand spoken OL1? ii speak OL1? iii read OL1? iv write OL1? 87 Not well Not at all Source: http://www.doksinet Other Language 2 (OL2): How well do you: Very well Fairly well Not well Not at all Very well Fairly well Not well Not at all i understand spoken OL2? ii speak OL2? iii read OL2? iv write OL2? Other Language 3 (OL3):

How well do you: i understand spoken OL3? ii speak OL3? iii read OL3? iv write OL3? 88 Source: http://www.doksinet Appendix B: Individual Subject Information Individual information native speakers Italians Sex Age 1 2 3 4 5 6 7 8 9 10 M F F F M M F F M M 25 26 22 34 26 30 37 25 28 26 Time since last lived in Italy in months 5 12 6 30 18 36 156 6 24 12 Errors proficiency test N/A N/A N/A 2 N/A N/A N/A 1 1 3 Individual information English learners English Sex Age Years studying Italian Other gender languages 1 2 3 4 5 M M F F M 21 21 22 21 59 4 5 4 4 1 French French/German French/Greek French French/German 6 7 8 M F M 40 21 58 15 6 4 9 11 F F 20 22 4 4 French French French/German/ Spanish French French/German/ Portuguese 89 Time spent in Italy in months 9 12 14 4 regular visits 96 12 12 Errors proficiency test 9 1 0 6 0 6 5 1 7 1 8 0 Source: http://www.doksinet Individual information Dutch learners Dutch Sex Age Years studying Italian Other

gender languages 2 4 5 6 7 F F F F F 26 22 45 25 33 6 5 4 4 4 N/A French German German French/German 8 F 22 4 9 M 21 3 10 11 12 F F M 28 32 47 3 8 7 French/German/ Spanish German/French/ Russian German/French French/German German/French 90 Time spent in Italy in months 36 4 9 17 regular visits 8 Errors proficiency test 5 1 1 1 5 regular visits 12 13 regular visits 0 4 3 0 1 Source: http://www.doksinet Appendix C: Italian Proficiency Test (answers in bold) Select the best answer: 1) I miei nonni mi dei racconti quando ero piccola. A B C D leggere ha letto leggeva leggevano 2) Quando eravamo a Roma, a vedere la Cappella Sistina. A B C D abbiamo andati siamo andato abbiamo andato siamo andati 3) Io il compito adesso. A B C D sta facendo facendo sto facendo state facendo 4) Cinzia una matita gialla. A B C D è hanno ha ho 5) Voi inglese? A B C D parla parlate parlano parlo 6)

Davide tanto il fine-settimana. A B C D dorme dormiamo dormo dormi 7) Nelle pizzerie molte pizze. A si vende 91 Source: http://www.doksinet B C D si vendono si vendo vendere 8) I nostri amici con noi in vacanza se avessero avuto abbastanza ferie quest’anno. A B C D avrebbero venuti sarebbero venuti avrebbe venuto sarebbero venuto 9) L’estate prossima io in Italia con la mia famiglia. A B C D viaggerò viaggio viaggerà viaggerete 10) Cristina seguire un corso d’italiano, ma non ha il tempo per farlo. A B C D vorrebbe vorreste vorrei vorresti 11) Giacomo, di fare la valigia? A B C D hai finito ho finito finivo finito 12) Se un milione di dollari, cosa faresti? A B C D avevi hai avrai avessi 13) Michele, è importante che tu quello che dicono i tuoi genitori. A B C D fai faccia fare fa 14) Se tu , saresti potuto venire

in macchina con me alla festa ieri sera. 92 Source: http://www.doksinet A B C D ha voluto avesse voluto avessi voluto avevi voluto 15) Alla fine di questa settimana, io di studiare per l’esame. A B C D avrei finito avessi finito avremo finito avrò finito In each Italian sentence, select the one underlined word or phrase that is incorrect: 1) Sara non studia più da quando ha deciso a lavorare a tempo pieno. A B C D più da ha a 2) Io voleva comprare quella macchina per mio figlio. A B C D voleva quella per mio 3) In questo negozio non si vende costumi da bagno. A B C D in si vende da 4) Quando vedrò la mia professoressa, la dirò quanto ho studiato per l’esame. A B C D vedrò la quanto l’ 5) Mio sorella abita in un paese piccolo vicino a Firenze. A B C D mio in paese a 93 Source: http://www.doksinet 6) Se io fossi ricco, viaggerò in tutto il mondo con la mia famiglia. A B C D fossi ricco viaggerò famiglia 7) Se io avevo avuto

l’opportunità di studiare all’università, sarei diventato avvocato. A B C D avevo avuto sarei diventato 8) Per il mio compleanno voglio un bicicletta nuova. A B C D per mio voglio un 9) Abbiamo mangiato molto bene al matrimonio di Federico sabato scorsa. A B C D mangiato al di scorsa 10) Io vado agli Stati Uniti dopo l’estate. A B C D io agli dopo l’ 11) Laura no ha mai visto il film "I Vitelloni." A B C D no mai visto il 12) La torta di mirtilli che fa tua mamma è la meglio di tutte. A B C D di tua meglio tutte 94 Source: http://www.doksinet 13) L’altro giorno Luigi ha andato a Roma in gita. A altro B ha C a D gita 14) I giocatori della squadra del Milan sanno giocare meglio che quelli del Inter. A B C D della sanno che quelli 15) La chiesa foste costruita nel 1546, e ancora oggi è in buone condizioni. A B C D foste costruita è condizioni Select the best answer: 1) vai per Natale? A B C D Cosa Quale Dove Qualunque 2) Lavorando

nei campi, Mario ha preso molto sole sulle . A B C D braccie braccia bracce bracci 3) Mi hanno rubato il portafoglio a Roma. A B C D Che bello! Che sfortuna! Che tipo! Che fico! 4) L’autobus arriva in ritardo durante l’inverno. A B C D stamattina mai ieri sempre 95 Source: http://www.doksinet 5) Se non per il mio professore, non parlerei così bene l’italiano. A era B fossi C è D fosse 6) Tua cugina ha chiamato e vuole che dai un colpo di telefono più tardi. A B C D la lo le gli 7) Lui ha cominciato costruire la casa. A B C D a di che da 8) Voglio a parlare l’italiano. A B C D di imparare fare essere 9) Ci sono biscotti e del latte per colazione. A B C D dei delle gli uno 10) Pietro dice che non può il nostro invito. A B C D accettare accertare accendere accennare Read the Italian text and select the best answers for the questions:

Indicazioni per casa mia: Prendi via Australia fino all’incrocio con via Mazzini. Gira a destra Dopo il secondo semaforo, gira a sinistra su via Garibaldi. C’è un panificio all’angolo e una gelateria dall’altra parte della strada. Vai dritto un chilometro e la mia casa è sulla destra, numero 28. 1) Qual è la mia via? 96 Source: http://www.doksinet A B C D via Australia via Mazzini via Garibaldi via numero 28 2) La mia casa è: A B C D accanto alla gelateria un chilometro dopo il panificio dopo il semaforo su via Mazzini Pasta alla Norma Ingredienti: melanzane, pomodori, prezzemolo, ricotta salata Sbucciare e friggere le melanzane. Tagliare a pezzettini i pomodori Cuocere i pomodori a parte per 15-20 minuti o finché non diventano asciutti. Poi aggiungere le melanzane, il prezzemolo, e il formaggio, e servire caldo con la pasta. 3) Le melanzane devono essere: A B C D grandi amare cotte con burro sbucciate 4) Che cosa NON serve per fare questa ricetta? A B C D il

rosmarino la pasta i pomodori le melanzane 3 gennaio 1998 Cara nonna, Ciao, come stai? È stato bello vederti a Natale. Ti volevo ringraziare per il bellissimo maglione che mi hai regalato. Come hai fatto a sapere che il verde è il mio colore preferito? Per Capo d’anno vado in montagna con degli amici e il maglione mi terrà calda. Spero di venirti a trovare a Pasqua con la mamma. Ti chiamerò più avanti per metterci d’accordo. Un bacione, Chiara 5) Che cosa ha regalato la nonna a Chiara? 97 Source: http://www.doksinet A B C D un viaggio in montagna un maglione verde una torta della nonna un uovo di Pasqua 6) Quanto spesso si vedono Chiara e la nonna? A B C D 3 o 4 volte allanno tutti i giorni quasi mai ogni sabato Ieri il sindaco di Trieste ha dichiarato il suo sostegno per una iniziativa giovanile che cercherebbe di mettere fine all’uso di droghe pesanti tra i giovani di Trieste. Questo piano è stato escogitato dagli studenti del liceo Marchesi con l’aiuto di un

loro professore. Gli studenti propongono di lanciare una serie di riunioni con altri studenti e con dei tossico-dipendenti con lo scopo di parlare dei pericoli dell’abuso di droghe pesanti. Queste riunioni si terrebbero subito dopo la scuola Sono invitate anche le famiglie degli studenti. 7) Cosa vogliono fare gli studenti del liceo Marchesi? A B C D parlare con il sindaco eliminare luso di droghe pesanti tra i giovani aiutare il loro professore educare dei tossico-dipendenti sulle loro possibilità lavorative 8) Il sindaco di Trieste: A B C D oppone liniziativa degli studenti usa droghe pesanti insegna al liceo Marchesi sostiene il progetto degli studenti Quest’anno c’è la possibilità che i giocatori di calcio facciano sciopero, mettendo in pericolo le prime partite della stagione e mettendo in crisi i tifosi in tutto il paese. Marco Carelli, centravanti dell’Udinese, ha spiegato in un’intervista,"Io amo il calcio. Se non amassi questo gioco non sarei mai arrivato

a questo punto nella mia carriera. Però credo che sia giusto che i giocatori vengano pagati in merito a come giocano, e che non ci debbano essere dei limiti sugli stipendi." La settimana prossima ci saranno delle riunioni tra i capitani di tutte le squadre di Seria A e i capi della Federazione Italiano Gioco Calcio per cercare un accordo. 9) Perchè i giocatori stanno pensando di fare sciopero? A Perchè si sentono in pericolo 98 Source: http://www.doksinet B C D Perchè i giocatori dellUdinese predono più soldi degli altri giocatori Perchè protestano contro la violenza negli stadii Perchè non vogliono avere limiti sugli stipendi 10) Perchè i tifosi sono in crisi? A B C D Perchè Marco Carelli non gioca più Perchè le prime partite della stagione possono saltare Perchè i capi della Federazione Italiano Gioco Calcio vanno in sciopero Perchè non vengano pagati 99 Source: http://www.doksinet Appendix D: Training Sentences, Items, Fillers and Questions Training

Sentences and Questions SPR Grammatical Ogni tanto i gatti mangiano Question: I gatti mangiano gli uccelli? A casa sua, Carlo mi prestò dei calzini Question: Carlo mi prestò delle scarpe ? Ungrammatical Il martedì Elena ha visto che le donne compriamo Luisa e’caduta gli uccelli dal suo armadio ortaggi dal negozio dagli scale Training Sentences and Questions AJT Grammatical Ogni mese, Cecilia compra molti vestiti al mercato. Question: Cecilia compra vestiti al mercato ? Qualche volta, prendiamo il treno per la spiaggia. Question: Prendiamo il tram per la spiaggia ? Ungrammatical Claudia ha detto che i sculturi in questo paese sono molto brutte. Ieri, ho troviamo un quadro di Rembrandt nell’attico. Grammatical SPR Fillers Alla mostra Quando erano a Firenze A scuola Ogni mattina A casa sua Se saremo in ritardo Dopo Natale Ogni volta che Appena Maria suonò il campanello Dopo che i nostri amici Domani Ti ho parlato Di quando in quando Lho pagato più di mille euro A casa di

Luisa Sebbene fosse in ritardo Lei ha mandato Il medico di campagna Gianni ha detto che Secondo Roberto Elena ha detto che Ieri Allo spettacolo teatrale Davanti al pubblico il quadro era stato ammirato né Maria né Giulia il professore leggeva lentamente andava parecchie volte al bar Gianni mi mostrò dei libri prenderemo lautobus ci siamo guardati allo specchio era andato a farle visita la cameriera aprì furono partiti dovresti andare a vederla dello scrittore il cui libro prendo il tram ma questo orologio la camera da letto per la riunione linvito al suo fidanzato aiuta ogni persona se fosse restata a casa gli abitanti degli Stati Uniti e caduta mi sono svegliata ho riso molto hai cantato la canzone 100 da tutti. visitarono il museo. e spiegava ogni parola. a prendere il caffè. dalla sua libreria. per il teatro. e non eravamo contenti. laveva trovata a casa. la porta. pulimmo il salotto. perché sta male. era stato premiato. per il centro commerciale. non funziona. è più

lunga che larga. camminava lentamente. senza che sua sorella lo sapesse. che ne abbia bisogno. sarebbe guarita più presto. sono ricchi. dalle scale. alle quattro di mattina in punto. quando è entrato il comico. molto bene. Source: http://www.doksinet Nel cortile In genere Al ballo Allopera Ieri sera Immediatamente il figlio del calciatore i cani mangiano carne Luisa ha baciato Gianni siamo arrivati tardi ho visto il pittore i cui quadri i pompieri sono partiti circa venticinque persone sono Ad un tratto apparse Ogni tanto Carlo va a Roma Lì per lì Mario ha deciso Per molto tempo ho preso il treno Dora in poi Elena terrà il portafogli È probabile che le tue scarpe si trovino Lo comprerò domani a meno che Il sabato i bambini comprano dolci Tutti e due i ragazzi hanno rubato le mele In ogni parte del mondo il vino dItalia Al più presto i malati vennero portati Al campeggio si raccontavano molte storie Nel diciannovesimo secolo le donne di solito portavano Per paura dei topi

Giulia non va giù Avendo dormito solo due ore Mario sbadigliava Il politico ha detto che parlare in pubblico richiede Senza dire una parola Roberta ha dato Alla mostra la scultura era stata ammirata Quando erano a Firenze né Giulia né Maria A scuola il professore di fisica Ogni mattina andava parecchie volte al bar A casa sua mi Gianni mostrò dei libri Se saranno in ritardo prenderanno lautobus Dopo Natale vi siete guardati allo specchio Al ballo Gianni ha baciato Luisa Appena Claudia suonò il campanello la cameriera aprì Dopo che i nostri amici furono partiti Questo pomeriggio dovresti andare a vederla Ti ho parlato del poeta la cui poesia Di quando in quando prendo lautobus Per questo orologio ho pagato più di mille euro Camminava lentamente sebbene fosse in ritardo Lui ha mandato linvito alla sua fidanzata Il medico di campagna non aiuta nessuno che Gianni pensava che se fosse restata a casa Secondo Roberto gli abitanti degli Stati Uniti Elena ebbe un ecchimosi perché era

caduta Ieri alle quattro di mattina in punto Allo spettacolo teatrale non ho riso di Davanti al pubblico hai cantato la canzone Il figlio del calciatore ha giocato a pallone 101 ha giocato a pallone. quando sono affamati. sulla bocca. a causa di un ingorgo stradale. sono famosi. dalla caserma. sul palcoscenico. con lautobus. di smettere di fumare. ogni mattina. nella borsa. in camera da letto. non labbiano venduto. dal negozio. dal frutteto. è famoso. allospedale. sui cavalli. le gonne. in cantina. tutto il tempo. prontezza di spirito. mille euro a Carlo. da tutti. visitarono il museo. leggeva assai lentamente. a prendere il tè. dalla sua libreria. per il teatro. e non eravate contenti. sulla bocca. la porta. pulimmo la cucina. perché sta male. era stata premiata. per il centro commerciale. ma non funziona. per la riunione. senza che sua sorella lo sapesse. ne abbia bisogno. sarebbe guarita più presto. sono poveri. dalle scale. mi sono svegliata. nessuna barzelletta. piuttosto

male. nel cortile. Source: http://www.doksinet Ungrammatical SPR Fillers In genere Al ballo Allopera Ieri sera Immediatamente Ad un tratto Ogni tanto Lì per lì Per molto tempo Dora in poi Siamo probabile che Lo comprerò domani Il sabato E due tutti i ragazzi Nella ogni parte del mondo Più al presto Al campeggio Nel diciannovesimo secolo Per paura dei topi Avendo dormito solo due ore Il politico ha dirito che Senza dicendo una parola Se saremo in ritardo Nel cortile A casa di Luisa i cani mangiano carne quando sono affamato. Luisa ha baciato Gianni sul bocca. siamo arrivati tardi a causa da un ingorgo stradale. ho visto il pittore la cui quadri sono famosi. i pompieri siamo partiti dalla caserma. circa venticinque persone sono apparire sul palcoscenico. Carlo va alla Roma con lautobus. Mario ha deciduto di smettere di fumare. Claudia e Gianni prendevo il treno ogni mattina. Elena tenerà il portafoglio in borsa. le tue scarpe si trovino in camera da letto. a meno che non

labbiano vendere. i bambini compri dolci dal negozio. hanno rubato le mele dal frutteto. il vino dItalia è famoso. i malati vennero portati allospedale. si raccontavano molte storie sui cavalla. le gonne di solito portavano le donne. Giulia non anda giù in cantina. Mario sbadigliavi tutto il tempo. parlare in pubblico richiede prontezza di spirito. Roberta ha dato mille euro a Carlo. prenderemo lautobus degli teatro. il figlio di calciatore ha giocato a pallone. la salotto è più lungo che largo. Grammatical AJT Fillers Tre settimane fa ho perso un anello di mia madre nel giardino In cucina, la cuoca offrí del tè al postino Ieri mattina, il direttore portò una rosa alla sua segretaria Gianni ha detto che i teatri in questa città sono molto belli Quando era a Firenze, andava tutti i giorni in biblioteca Quando venderò la casa, comprerò una macchina per mio fratello Carlo pensa che i viaggi in treno non siano sempre comodi Sono sicuro che mia madre non comprerebbe quel cappello

Questo pomeriggio, l’avvocato manderà la lettera al giudice La segretaria pensava che la cuoca fosse andata al mercato Il cameriere ha già apparecchiato per dodici Poco tempo fa, Anna ha perduto la borsa nell’ autobus Uscendo dalla posta, Stefano ha incontrato l’insegnante d’arte Mentre apriva la scatola tutti la guardavano con curiosità Quelle barche erano cariche di pesce quando arrivarono Si dice che quell’attrice abbia venti paia di scarpe La strada che cerco è vicina a questa piazza Parlando aprì un cassetto e tirò fuori una carta geografica Ieri stava un po’ meglio, ma oggi sta peggio 102 Source: http://www.doksinet La cattedrale di Pisa fu cominciata nell’anno 1063 Ogni volta che vado lì, quell’uomo mi segue fino alla porta È un peccato vendere quel negozio di fiori Poco fa, Anna è caduta dall’albero Pose il pacchetto sul tavolino vicino alla finestra Hanno una villa in una di quelle stradine di campagna Mi diede quel libro prima di partire per

Roma Invece di leggere un romanzo, dovrebbe studiare quella lezione Ti proibisco di andare in quella casa La bottiglia di vino è sulla tavola nella sala da pranzo Milano è una delle piú grandi città d’Italia Ieri sera, m’accompagnarono a teatro per sentire l’opera La battaglia di Waterloo ebbe luogo il 18 giugno 1815 Ogni settimana, Antonio compra molte pere al mercato Il clima di questo paese è malsano in estate Benedetto ha detto che ama molto sua madre Domani mattina, porterò il libro dal libraio Ieri pomeriggio, il fornaio aveva venduto tutto il pane Clara laverà queste due macchine nel cortile Ieri mattina, Anna si alzò alle undici in punto La lezione è piú facile di quanto si possa pensare Il banchiere non è ricco come il farmacista Stefano non parla l’inglese bene come l’italiano Sceso dalla torre, mi sedetti un momento sull’ erba Anna cerca una giacca che vada bene con la sua gonna Antonio ha scritto che devi tornare immediatamente Quel ragazzo danza

meglio di quanto pensi Dissero che il re di Spagna era morto Due settimane fa ho perso un anello di mia madre nel giardino In cucina, la cuoca offrì al postino del tè La settimana scorsa, il professore portò una rosa alla sua segretaria Gianni ha detto che in questa città i teatri sono molto belli Quando era a Roma, andava tutti i giorni in biblioteca Comprerò una macchina per mio fratello, quando venderò la casa Carlo pensa che i viaggi in tassì non sono sempre comodi È sicuro che sua madre non comprerebbe quel cappello Questo pomeriggio, il giudice manderà la lettera all’avvocato La segretaria pensava che il direttore fosse andato al mercato Il cameriere ha apparecchiato per dodici Ieri mattina, Anna ha perduto la borsa nell’autobus Stefano ha incontrato l’insegnante d’arte uscendo dalla posta Mentre apriva il regalo tutti la guardavano con curiosità Quelle barche erano cariche di pesce quando partirono Si dice che quell’attrice abbia cento paia di orecchini La

strada che cerco è vicina a questa chiesa Aprì un cassetto parlando e tirò fuori una carta geografica Lunedì stava un po’ meglio, ma oggi sta peggio La torre di Pisa fu cominciata nell’anno 1063 Ogni volta che ci vai, quell’uomo ti segue fino alla porta È un peccato vendere quel negozio di vestiti 103 Source: http://www.doksinet Poco fa, Giulia è caduta dall’albero Pose il pacchetto sul tavolino vicino alla scrivania Hanno una casa in una di quelle stradine di campagna Ungrammatical AJT Fillers Mi diede libro quel prima di partire per Roma Invece da leggere un romanzo, dovrebbe studiare quella lezione Ti proibo di andare in quella casa La bottiglia di vino siamo sulla tavola nella sala da pranzo Milano è una degli piú grandi città d’Italia Ieri sera, accompagnarono al teatro per sentire l’opera La battaglia di Waterloo avesti luogo il 18 giugno 1815 Ogni settimana, compra Antonio molte pere al mercato Il clima questo paese è malsano in estate Benedetto ha

detto che amare molto sua madre Domani mattina, porterò il libro dalla libraio Ieri pomeriggio, il fornaio aveva venditi tutto il pane Clara laverà due queste macchine nel cortile Ieri mattina, Anna si alzò alle undici al punto La lezione è piú facile di quanto si possa pensiamo Il banchiere è non ricco come il farmacista Stefano non parlavi l’inglese bene come l’italiano Sceso dalla torre, mi sedetti un momento sulla erba Anna cercare una giacca che vada bene con la sua gonna Antonio scritto ha che devi tornare immediatamente Quel ragazzo danza buono di quanto pensi Dissero che il re da Spagna fosse morto In cucina, la cuoca offrire del tè al postino Gianni ha detto che i teatri in questa città molto belli sono SPR and AJT Items FR= feminine regular noun, MR= masculine regular noun, FI= feminine irregular noun, MI= masculine irregular noun Det-N-Adj FR Maria era troppo bassa per chiudere la finestra piccola in fondo al corridoio Il ladro pensava che la finestra piccola

sarebbe stata difficile da scavalcare L’alunno ha dato la risposta esatta alla domanda Paolo non ha dato la risposta esatta al presentatore Il giardiniere ha potato la pianta nuova con le forbici Luisa ha comprato la pianta nuova dal vivaio in centro *Det-N-Adj FR Maria era troppo bassa per chiudere il finestra piccolo in fondo al corridoio Il ladro pensava che il finestra piccolo sarebbe stata difficile da scavalcare L’alunno ha dato il risposta esatto alla domanda Paolo non ha dato il risposta esatto al presentatore Il giardiniere ha potato il pianta nuovo con le forbici Luisa ha comprato il pianta nuovo dal vivaio in centro 104 Source: http://www.doksinet *Det-N-Adj FR Maria era troppo bassa per chiudere il finestra piccola in fondo al corridoio Il ladro pensava che il finestra piccola sarebbe stata difficile da scavalcare L’alunno ha dato il risposta esatta alla domanda Paolo non ha dato il risposta esatta al presentatore Il giardiniere ha potato il pianta nuova con le

forbici Luisa ha comprato il pianta nuova dal vivaio in centro Det-N-*Adj FR Maria era troppo bassa per chiudere la finestra piccolo in fondo al corridoio Il ladro pensava che la finestra piccolo sarebbe stata difficile da scavalcare L’alunno ha dato la risposta esatto alla domanda Paolo non ha dato la risposta esatto al presentatore Il giardiniere ha potato la pianta nuovo con le forbici Luisa ha comprato la pianta nuovo dal vivaio in centro Det-N-Adj MR Da quando Pietro ha avuto il sogno infausto non vuole più’ andare a dormire Giulia pensa che il sogno infausto tornerà domani Il macchinista pensa che il treno lungo sia difficile da manovrare Mario ha visto che il treno lungo non entrava tutto nella stazione Roberto ha visto che il gatto persiano si rincorreva la coda Il cane ha rincorso il gatto persiano fin sull’albero *Det-N-Adj MR Da quando Pietro ha avuto la sogno infausta non vuole più’ andare a dormire Giulia pensa che la sogno infausta tornerà domani Il

macchinista pensa che la treno lunga sia difficile da manovrare Mario ha visto che la treno lunga non entrava tutto nella stazione Roberto ha visto che la gatto persiana si rincorreva la coda Il cane ha rincorso la gatto persiana fin sull’albero *Det-N-Adj MR Da quando Pietro ha avuto la sogno infausto non vuole più’ andare a dormire Giulia pensa che la sogno infausto tornerà domani Il macchinista pensa che la treno lungo sia difficile da manovrare Mario ha visto che la treno lungo non entrava tutto nella stazione Roberto ha visto che la gatto persiano si rincorreva la coda Il cane ha rincorso la gatto persiano fin sull’albero Det-N-*Adj MR Da quando Pietro ha avuto il sogno infausta non vuole più’ andare a dormire Giulia pensa che il sogno infausta tornerà domani Il macchinista pensa che il treno lunga sia difficile da manovrare Mario ha visto che il treno lunga non entrava tutto nella stazione Roberto ha visto che il gatto persiana si rincorreva la coda Il cane ha

rincorso il gatto persiana fin sull’albero Det-N-Adj FI 105 Source: http://www.doksinet Il vampiro pensa che la luce intensa sia insopportabile Uscendo dal cinema, la luce intensa fa male agli occhi degli Paolo e Angela I giovani pensano che la tradizione religiosa non sia interessante Secondo il magistrato, la tradizione religiosa era contro la legge Nella fiaba, la chiave dorata apre la porta alla terra delle meraviglie La madre di Gianni tiene la chiave dorata sotto il cuscino *Det-N-Adj FI Il vampiro pensa che il luce intenso sia insopportabile Uscendo dal cinema, il luce intenso fa male agli occhi degli Paolo e Angela I giovani pensano che il tradizione religioso non sia interessante Secondo il magistrato, il tradizione religioso era contro la legge Nella fiaba, il chiave dorato apre la porta alla terra delle meraviglie La madre di Gianni tiene il chiave dorato sotto il cuscino *Det-N-Adj FI Il vampiro pensa che il luce intensa sia insopportabile Uscendo dal cinema, il luce

intensa fa male agli occhi degli Paolo e Angela I giovani pensano che il tradizione religiosa non sia interessante Secondo il magistrato, il tradizione religiosa era contro la legge Nella fiaba, il chiave dorata apre la porta alla terra delle meraviglie La madre di Gianni tiene il chiave dorata sotto il cuscino Det-N-*Adj FI Il vampiro pensa che la luce intenso sia insopportabile Uscendo dal cinema, la luce intenso fa male agli occhi degli Paolo e Angela I giovani pensano che la tradizione religioso non sia interessante Secondo il magistrato, la tradizione religioso era contro la legge Nella fiaba, la chiave dorato apre la porta alla terra delle meraviglie La madre di Gianni tiene la chiave dorato sotto il cuscino Det-N-Adj MI La bambina ha dato il fiore giallo a sua madre Il coniglio ha mangiato il fiore giallo nella sua gabbia Elena ha detto che il fiume calmo era molto riposante I bambini pensavano che il fiume calmo andava bene per nuotare Durante la crociera, il viaggiatore grasso

ha mangiato tanto cibo Nella cabina della nave, il viaggiatore grasso non entrava nel letto *Det-N-Adj MI La bambina ha dato la fiore gialla a sua madre Il coniglio ha mangiato la fiore gialla nella sua gabbia Elena ha detto che la fiume calma era molto riposante I bambini pensavano che la fiume calma andava bene per nuotare Durante la crociera, la viaggiatore grassa ha mangiato tanto cibo Nella cabina della nave, la viaggiatore grassa non entrava nel letto *Det-N-Adj MI La bambina ha dato la fiore giallo a sua madre Il coniglio ha mangiato la fiore giallo nella sua gabbia 106 Source: http://www.doksinet Elena ha detto che la fiume calmo era molto riposante I bambini pensavano che la fiume calmo andava bene per nuotare Durante la crociera, la viaggiatore grasso ha mangiato tanto cibo Nella cabina della nave, la viaggiatore grasso non entrava nel letto Det-N-*Adj MI La bambina ha dato il fiore gialla a sua madre Il coniglio ha mangiato il fiore gialla nella sua gabbia Elena ha detto

che il fiume calma era molto riposante I bambini pensavano che il fiume calma andava bene per nuotare Durante la crociera, il viaggiatore grassa ha mangiato tanto cibo Nella cabina della nave, il viaggiatore grassa non entrava nel letto Questions Questions SPR Items Il ladro pensava che la finestra piccola sarebbe stata difficile da scavalcare Q: Il ladro pensava che la finestra sarebbe stata facile da scavalcare? N Mario ha visto che il treno lungo non entrava tutto nella stazione Q: Il treno entrava tutto nella stazione? N Uscendo dal cinema, la luce intensa fa male agli occhi degli Paolo e Angela Q: La luce fa male agli occhi degli Paolo e Luisa? N Elena ha detto che il fiume calmo era molto riposante Q: Elena pensa che il fiume era riposante? Y I giovani pensano che la tradizione religiosa non sia interessante Q: I giovani pensano che la tradizione è non interessante? Y Fillers Per paura dei topi, Giulia non va giù in cantina Q: Giulia va giù in cantina? N Al campeggio, si

raccontavano molte storie sui cavalli Q: Si raccontavano molte storie sui cani? N Nel cortile, il figlio del calciatore ha giocato a pallone Q: La figlia del calciatore ha giocato a pallone? N Avendo dormito solo due ore, Mario sbadigliava tutto il tempo Q: Mario ha dormito dieci ore? N L’ho pagato più di mille euro, ma questo orologio non funziona Q: L’orologio funziona? N Elena ha detto che e’ caduta dalle scale Q: Elena e’caduta dall’altalena? N Senza dire una parola, Roberta ha dato mille euro a Carlo Q: Roberta ha dato cento euro a Carlo? N Appena Maria suono’ il campanello, la cameriera aprì la porta Q: Il maggiordomo aprì la porta? N Dopo Natale, ci siamo guardati allo specchio e non eravamo contenti Q: Erano contenti di vedersi allo specchio? N Il sabato, i bambini comprano dolci dal negozio Q: I bambini comprano dolci il lunedì? N Quando erano a Firenze, né Maria né Giulia visitarono il museo Q: Maria e Giulia visitarono il museo? N 107 Source:

http://www.doksinet Ad un tratto, circa venticinque persone sono apparse sul palcoscenico Q: Venti persone sono apparse sul palcoscenico? N Sebbene fosse in ritardo per la riunione, camminava lentamente Q: Camminava lentamente? Y È probabile che le tue scarpe si trovino in camera da letto Q: È probabile che le scarpe siano in camera da letto? Y In genere, i cani mangiano carne quando sono affamati Q: I cani mangiano la carne? Y Lì per lì, Mario ha deciso di smettere di fumare Q: Mario ha deciso di smettere di fumare? Y Al ballo, Luisa ha baciato Gianni sulla bocca Q: Luisa ha baciato Gianni? Y Tutti e due i ragazzi hanno rubato le mele dal frutteto Q: Tutti e due i ragazzi hanno rubato la frutta? Y Alla mostra, il quadro era stato ammirato da tutti Q: Tutti hanno ammirato il quadro? Y Il medico di campagna aiuta ogni persona che ne abbia bisogno Q: Il medico di campagna aiuta ogni persona malata? Y Immediatamente, i pompieri sono partiti dalla caserma Q: I pompieri reagirono

immediatamente? Y In ogni parte del mondo, il vino d’Italia è famoso Q: Il vino d’Italia è noto? Y Secondo Roberto, gli abitanti degli Stati Uniti sono ricchi Q: Gli Americani sono ricchi secondo Roberto? Y A casa sua, Gianni mi mostrò dei libri dalla sua libreria Q: Gianni mostrò i libri? Y A casa di Luisa, la camera da letto è più lunga che larga Q: La camera da letto di Luisa è più lunga che larga? Y Questions AJT Items Il cane ha rincorso il gatto persiano fin sull’ albero Q: Il gatto ha rincorso il cane? N Il vampiro pensa che la luce intensa sia insopportabile Q: Il vampire piace la luce? N Il macchinista pensa che il treno lungo sia difficile da manovrare Q: Il macchinista pensa che il treno sia difficile da manovrare? Y La bambina ha dato il fiore giallo a sua madre Q: La bambina ha dato il fiore a sua madre? Y L’alunno ha dato la risposta esatta alla domanda Q: L’alunno ha risposto correttamente? Y Fillers Ieri mattina, il direttore portò una rosa alla sua

segretaria Q: La segretaria ha ricevuto una rosa? Y Quelle barche erano cariche di pesce quando arrivarono Q: C’era del pesce sulle barche? Y Si dice che quell’attrice abbia venti paia di scarpe Q: L’attrice ha venti paia di scarpe? Y Ieri stava un po’ meglio, ma oggi sta peggio 108 Source: http://www.doksinet Q: Sta peggio di ieri? Y Milano è una delle piú grandi città d’Italia Q: Milano è una grande città? Y Pose il pacchetto sul tavolino vicino alla finestra Q: Pose il pacchetto sul tavolino? Y Ieri mattina, Anna si alzò alle undici in punto Q: Anna si alzò alle undici in punto ieri? Y Benedetto ha detto che ama molto sua madre Q: Benedetto ama sua madre? Y Anna cerca una giacca che vada bene con la sua gonna Q: Anna cerca una giacca? Y Ogni settimana, Antonio compra molte pere al mercato Q: Antonio compra molte pere? Y Poco tempo fa, Anna ha perduto la borsa nell’autobus Q: Anna ha perduto la borsa nell’autobus? Y La battaglia di Waterloo ebbe luogo il 18

giugno 1815 Q: La battaglia di Waterloo ebbe luogo nell’ anno 1815? Y Dissero che il re di Spagna era morto Q: Dissero che il re di Spagna era vivo? N Poco fa, Anna è caduto dall’albero Q: Anna è caduta dal tetto? N Carlo pensa che i viaggi in treno non siano sempre comodi Q: Carlo pensa che i viaggi in treno siano piacevoli? N Clara laverà queste due macchine nel cortile Q: Clara laverà cinque macchine? N In cucina, la cuoca offrí del tè al postino Q: La cuoca offrí del vino al postino? N Il cameriere ha già apparecchiato per dodici Q: Il cameriere ha già apparecchiato per venti? N La cattedrale di Pisa fu cominciata nell’anno 1063 Q: La cattedrale fu cominciata nell’anno 1578? N Questo pomeriggio, l’avvocato manderà la lettera al giudice Q: L’avvocato manderà la lettera all’imputato? N Parlando aprí un cassetto e tirò fuori un carta geografica Q: Ha preso una torta dal cassetto? N Stefano non parla l’inglese bene come l’italiano Q: Stefano parla

l’inglese bene come l’italiano? N Uscendo dalla posta, Stefano ha incontrato l’insegnante d’arte Q: Stefano ha incontrato l’insegnante di matematica? N La bottiglia di vino è sulla tavola nella sala da pranzo Q: La bottiglia di vino è sul pavimento? N Tre settimane fa ho perso un anello di mia madre nel giardino Q: Ho perso l’anello di mia zia? N 109 Source: http://www.doksinet Appendix E: Accuracy Rate Comprehension Questions Subject SPR SPR AJT errors Accuracy errors Italian 1 0 100,0% 2 5 83,3% 3 1 96,7% 4 3 90,0% 5 2 93,3% 6 1 96,7% 7 3 90,0% 8 0 100,0% 9 3 90,0% 10 7 76,7% AJT Accuracy Total errors Overall Accuracy 1 6 2 0 1 1 0 1 3 3 96,7% 80,0% 93,3% 100,0% 96,7% 96,7% 100,0% 96,7% 90,0% 90,0% 1 11 3 3 3 2 3 1 6 10 98,3% 81,7% 95,0% 95,0% 95,0% 96,7% 95,0% 98,3% 90,0% 83,3% English 1 2 3 4 5 6 7 8 9 11 5 2 2 3 2 5 3 4 3 4 83,3% 93,3% 93,3% 90,0% 93,3% 83,3% 90,0% 86,7% 90,0% 86,7% 1 1 1 1 0 2 0 1 2 1 96,7% 96,7% 96,7% 96,7% 100,0% 93,3% 100,0% 96,7%

93,3% 96,7% 6 3 3 4 2 7 3 5 5 5 90,0% 95,0% 95,0% 93,3% 96,7% 88,3% 95,0% 91,7% 91,7% 91,7% 2* 4 5 6 7 8 9 10 11 12 7 2 3 2 2 2 2 3 1 1 76,7% 93,3% 90,0% 93,3% 93,3% 93,3% 93,3% 90,0% 96,7% 96,7% 2 0 1 0 0 1 1 1 0 2 93,3% 100,0% 96,7% 100,0% 100,0% 96,7% 96,7% 96,7% 100,0% 93,3% 9 2 4 2 2 3 3 4 1 3 75,0% 96,7% 93,3% 96,7% 96,7% 95,0% 95,0% 93,3% 98,3% 95,0% Dutch * For subject number 2 in the Dutch group, there was a slight misunderstanding regarding how to answer the questions. The low score in the comprehension questions thus reflects this misunderstanding, not lack of attention or understanding of the sentences. 110 Source: http://www.doksinet Appendix F: Instructions SPR and AJT Instructions SPR • In this part of the experiment, you will be required to read sentences and answer questions about these sentences. Please just read through the sentences even if the sentence seems strange to you. • The sentence will not appear as a whole on the screen, it will appear

in three chunks. To see the first chunk, you will have to press the spacebar Once you have read the first chunk, you press the spacebar again to see the second chunk and so on. • BUT: once you press the space bar, the previous chunk will disappear and the new chunk will appear, so please make sure you have read the first chunk properly before you move on. • Every now and then, you will be asked a simple question about the sentence that can be answered by ‘yes’ (the y-key on the keyboard) or ‘no’ (the n-key on the keyboard). • The next four sentences are training sentences to show you what to do. These sentences are not part of the actual experiment. Instructions AJT • In this part of the experiment, you will be required to judge sentences according to their acceptability and answer questions about them. • The whole sentence will appear on the screen. Once you have read the sentence, please give the sentence a mark out of 6, using the number keys on the keyboard. •

Marking the sentence with a 1 indicates that you find the sentence very unacceptable; marking the sentence with a 6 indicates you find the sentence very acceptable. • As I am interested in the different degrees of acceptability that the sentences have, please use the whole range of the scale. You will find that some sentences are very bad, whereas others may be less bad without being entirely acceptable. So please do not limit your answers to just 1 and 6 • The next four sentences are training sentences to show you what to do. These sentences are not part of the actual experiment. 111 Source: http://www.doksinet Appendix G: Means, Standard Deviations and Ranges SPR Mean, SD and Range per pattern type, RTsum Native English Dutch Det-N-Adj M SD R 3258 1448 10859692 4758 2024 176811472 4262 1782 158310818 *Det-N-Adj M SD R 3492 1624 125111310 4739 2159 164117564 4453 1966 162612348 *Det-N-Adj M SD R 3169 1430 10409513 4479 1917 196412173 4447 1927 158612588 Det-N-*Adj M SD R

3157 1263 10377731 4625 1916 192716878 4566 2175 161414915 Mean, SD and Range per pattern type, RT2 Native Det-N-Adj M SD 1006 628 English 1356 675 Dutch 1326 726 R 3444622 4914715 4845747 *Det-N-Adj M SD 1204 719 1524 858 1528 894 R 3014090 4285624 4935596 *Det-N-Adj M SD 1027 530 1435 779 1410 769 R 5533907 4334812 3525304 Det-N-*Adj M SD R 1007 481 3233598 1435 760 4214270 1403 791 4095881 R 3344355 4055204 4054142 Det-N-*Adj M SD R 1063 570 3563568 1575 933 4515547 1556 939 3845300 Mean, SD and Range per pattern type, RT3 Native Det-N-Adj M SD 1158 784 English 1635 837 Dutch 786 1459 R 3944042 4084774 4035952 *Det-N-Adj M SD R 1147 683 3723500 1548 995 5743909 1470 880 4375832 *Det-N-Adj M SD 1062 612 1442 828 1440 785 Mean, Standard Deviation and Range regular nouns RTsum Det-N-Adj Reg M SD R Native 1363 1387 10859692 English 4645 1867 207711421 Dutch 4237 1721 158310110 *Det-N-Adj Reg M SD R 3635 1575 134610437 4866 2000 191111163 4306 1669

14698964 112 *Det-N-Adj Reg M SD R 3283 1559 11899513 4455 1827 196410577 4452 1951 187312588 Det-N-*Adj Reg M SD R 3206 1313 10377731 4775 1925 195411918 4708 2336 177014915 Source: http://www.doksinet Mean, standard deviation and range regular nouns RT2 Det-N-Adj Reg M SD R Native 951 474 3442848 English 1295 607 4912991 Dutch 1334 762 5065747 *Det-N-Adj Reg M SD R 1268 758 3573591 1536 787 5534744 1482 826 4935596 *Det-N-Adj Reg M SD R 1058 594 3763907 1405 730 4334733 1410 830 6165304 Det-N-*Adj Reg M SD R 1040 452 3232699 1426 694 4214066 1448 896 4775881 Mean, Standard Deviation and Range Regular Nouns RT3 Det-N-Adj Reg M SD R Native 1161 910 3944042 English 1617 873 4084774 Dutch 1388 673 5304047 *Det-N-Adj Reg M SD R 1148 660 3723500 1549 799 5743909 1488 915 4435832 *Det-N-Adj Reg M SD R 1046 602 3864355 1429 819 4054784 1429 744 4053785 Det-N-*Adj Reg M SD R 1045 558 3563568 1586 902 4645221 1539 893 4205300 Mean, Standard Deviation and Range Irregular Nouns

RTsum Native English Dutch Det-N-Adj Irreg M SD R 3353 1507 12008567 4872 2174 176811472 4288 1849 165910818 *Det-N-Adj Irreg M SD R 3352 1665 125111310 4613 2307 164117564 4593 2212 162612348 *Det-N-Adj Irreg M SD R 3055 1285 10407833 4502 2011 200212173 4443 1911 15869584 Det-N-*Adj Irreg M SD R 3106 1212 13116997 4705 2246 192716878 4429 2008 161411697 Mean, Standard Deviation and Range Irregular Nouns RT2 Native English Dutch Det-N-Adj Irreg M SD R 1061 750 3704622 1417 735 5014715 1317 691 4844401 *Det-N-Adj Irreg M SD R 1140 674 3014090 1512 926 4285624 1574 957 5754745 *Det-N-Adj Irreg M SD R 997 457 3533478 1464 826 4404812 1409 708 3524569 Det-N-*Adj Irreg M SD R 975 507 3373598 1445 824 5654270 1358 672 4093889 Mean, Standard Deviation and Range Irregular Nouns RT3 Native Det-N-Adj Irreg M SD R 1155 638 3893942 *Det-N-Adj Irreg *Det-N-Adj Irreg M SD R M SD R 1147 707 414- 1078 624 3343378 3997 113 Det-N-*Adj Irreg M SD R 1081 583 3903344 Source:

http://www.doksinet English 1653 802 Dutch 1531 883 4214321 4035952 1546 1160 1451 847 467- 1455 5916 437- 1452 4465 841 827 5215204 4794142 1563 968 1573 987 4515547 3845070 Mean, Standard Deviation and Range Feminine Nouns RTsum Native English Dutch Det-N-Adj Fem M SD R 3221 1490 10859692 4644 1916 176811472 4360 1971 168410818 *Det-N-Adj Fem M SD R 3522 1572 125911310 4838 2289 191117564 4460 2044 146912348 *Det-N-Adj Fem M SD R 3144 1442 10408042 4514 2033 196412173 4278 1797 158612367 Det-N-*Adj Fem M SD R 3048 1152 10377204 4558 1875 193211735 4718 2395 174314915 Mean, Standard Deviation and Range Feminine Nouns RT2 Native English Dutch Det-N-Adj Fem M SD R 941 525 3444364 1350 713 4914715 1340 679 5554401 *Det-N-Adj Fem M SD R 1232 715 3574090 1524 810 4285208 1572 976 4935596 *Det-N-Adj Fem M SD R 1031 578 3533578 1413 758 4333858 1278 546 3523570 Det-N-*Adj Fem M SD R 956 388 3232439 1358 654 4213788 1456 882 4095881 Mean, Standard Deviation and

Range Feminine Nouns RT3 Native English Dutch Det-N-Adj Fem M SD R 1199 939 3944042 1616 818 4083700 1509 890 5665952 *Det-N-Adj Fem M SD R 1162 673 3723334 1509 867 4675916 1490 868 4434465 *Det-N-Adj Fem M SD R 1052 655 3344355 1456 895 4675204 1387 772 4054142 Det-N-*Adj Fem M SD R 1025 482 3682595 1521 963 4515519 1520 835 3844915 Mean, Standard Deviation and Range Masculine Nouns RTsum Native English Dutch Det-N-Adj Mas M SD R 3295 1411 12138567 4869 2127 204411421 4161 1567 15839042 *Det-N-Adj Mas M SD R 3463 1679 125110465 4642 2029 164111163 4445 1897 162610753 114 *Det-N-Adj Mas M SD R 3194 1424 12289513 4444 1801 200211011 4617 2042 187312588 Det-N-*Adj Mas M SD R 3267 1361 13037731 4922 2272 192716878 4413 1930 161411402 Source: http://www.doksinet Mean, Standard Deviation and Range Masculine Nouns RT2 Native English Dutch Det-N-Adj Mas M SD R 1071 713 3624622 1361 639 5013246 1311 772 4845747 *Det-N-Adj Mas M SD R 1176 724 3013591 1525 907 5295624 1486 806

5244445 *Det-N-Adj Mas M SD R 1024 481 4113907 1457 801 4404812 1498 822 5755304 Det-N-*Adj Mas M SD R 1058 555 3373598 1513 850 5654270 1348 688 4773757 Mean, Standard Deviation and Range Masculine Nouns RT3 Native English Dutch Det-N-Adj Mas M SD R 1117 592 3892920 1653 858 4214774 1410 667 4033822 *Det-N-Adj Mas M SD R 1133 694 4153500 1510 738 4713761 1449 896 4375832 115 *Det-N-Adj Mas M SD R 1073 568 4283033 1428 760 4054784 1494 798 5263998 Det-N-*Adj Mas M SD R 1102 646 3563568 1628 903 4645547 1553 954 4575300 Source: http://www.doksinet Appendix H: Means, Standard Deviations and Ranges AJT Mean, SD and Range per pattern type Det-N-Adj M SD Native 5,83 0,69 English 5,62 0,92 Dutch 5,84 0,51 Range 1-6 1-6 2-6 *Det-N-Adj M SD Range 2,13 1,11 1-5 3,78 1,31 1-6 3,58 1,17 1-6 *Det-N-Adj M SD 2,82 1,72 3,66 1,22 3,81 1,26 Range 1-6 1-6 2-6 Det-N-*Adj M SD Range 2,49 1,23 1-6 3,80 1,28 1-6 4 1,18 2-6 *Det-N-Adj Reg M SD Range 2,66 1,67 1-6 3,66 1,22 1-6 3,79 1,27

2-6 Det-N-*Adj Reg M SD Range 2,52 1,22 1-6 3,73 1,24 1-6 4,04 1,17 2-6 Mean, Standard Deviation and Range regular nouns Det-N-Adj Reg M SD Range Native 5,85 0,67 1-6 English 5,72 0,76 1-6 Dutch 5,86 0,4 4-6 *Det-N-Adj Reg M SD Range 1,98 1,07 1-5 3,48 1,13 1-6 3,53 1,16 2-6 Mean, Standard Deviation and Range Irregular Nouns Det-N-Adj Irreg M SD Range Native 5,82 0,72 1-6 English 5,53 1,05 1-6 Dutch 5,82 0,61 2-6 *Det-N-Adj Irreg M SD Range 2,28 1,13 1-5 4,08 1,41 1-6 3,63 1,17 1-6 *Det-N-Adj Irreg M SD Range 2,98 1,77 1-6 3,67 1,23 2-6 3,83 1,25 2-5 Det-N-*Adj Irreg M SD Range 2,47 1,25 1-5 3,87 1,32 1-6 3,95 1,19 2-6 Mean, Standard Deviation and Range Feminine Nouns Det-N-Adj Fem M SD Range Native 5,93 0,34 4-6 English 5,68 0,82 1-6 Dutch 5,85 0,46 3-6 *Det-N-Adj Fem M SD Range 2,18 1,09 1-5 3,63 1,24 1-6 3,59 1,18 1-6 *Det-N-Adj Fem M SD Range 2,78 1,61 1-6 3,79 1,2 1-6 3,79 1,24 2-5 Det-N-*Adj Fem M SD Range 2,57 1,29 1-6 3,82 1,28 1-6 3,94 1,22 2-6 Mean, Standard

Deviation and Range Masculine Nouns Det-N-Adj Mas M SD Range Native 5,73 0,91 1-6 English 5,57 1,01 1-6 Dutch 5,83 0,56 2-6 *Det-N-Adj Mas M SD Range 2,08 1,13 1-5 3,93 1,36 1-6 3,57 1,15 2-6 116 *Det-N-Adj Mas M SD Range 2,86 1,83 1-6 3,53 1,23 1-6 3,83 1,29 2-6 Det-N-*Adj Mas M SD Range 2,42 1,17 1-5 3,78 1,27 2-6 4,05 1,14 2-6