Subject Anaphora in the L2 Dutch
of Native Italian Speakers
Alma Veenstra, s1328891
MA in Applied Linguistics
Faculty of Arts
University of Groningen
Supervisors:
Dr. Charlotte Koster
Prof. dr. Petra Hendriks
Table of Contents 0. Abstract p. 2 1. Introduction p. 3 2. Theoretical Background p. 8 2.1 Linguistic Theories p. 8 2.2 Previous Studies p. 19
2.3 Second Language Acquisition p. 22
2.4 Hypotheses p. 31
3. Method p. 34
3.1 Participants p. 34
3.2 Materials p. 35
3.3 Procedure p. 44
3.4 Design and Analyses p. 46
4. Results p. 49
4.1 Production Task Results p. 49
4.2 Comprehension Task Results p. 57
4.3 Language Questionnaire and Proficiency Test Results p. 59
4.4 Correlations p. 61
5. Discussion p. 66
5.1 Topic Shift Production p. 66
5.2 Topic Shift Comprehension p. 69
5.3 Bidirectional Optimization p. 73
5.4 Italian Constraint Ranking p. 74
5.5 Other Factors p. 75
6. Conclusion p. 80
Acknowledgments p. 82
0. Abstract
Children and elderly adults have difficulties producing and comprehending subject anaphora (Hendriks et al, 2008; Wubs et al., 2009). Their cognitive resources are too limited to take into account the opposite perspective, causing a deviant performance.
This study investigates subject anaphora by Italian speakers of Dutch within an Optimality Theory framework. Twenty Italian speakers of Dutch and twenty native speakers of Dutch took part in a production and a comprehension task. Two hypotheses were tested:
1. Because of cognitive demands of using a second language, Italian L2 adults are unable to take into account the opposite perspective, and fail to interpret NPs marking a topic shift and they produce null subjects or pronouns for non-topics.
2. Italian L2 adults use their Italian constraint ranking for subject anaphora in Dutch; they incorrectly interpret pronouns as not referring to the topic of the discourse, and produce null subjects for topics.
There was no evidence to support hypothesis 1, but there was a negative correlation between age and successfulness in subject anaphora, similar to the L1 adults. In addition, the Italian L2 adults’ production was similar to that of the Dutch native speakers, but they incorrectly
interpreted pronouns in comprehension, supporting hypothesis 2. (200 words)
1. Introduction
A story structure is rather complex, which makes telling a story more difficult than it may seem to be at first glance. A speaker has to keep in mind what he or she wants to tell, what he or she has already told, what the hearer knows, and what their shared knowledge is. In
addition to that, the speaker has to refer to the characters in such a way that the hearer is able to interpret these referring expressions. Consider examples (1) and (2):
(1) An Indian sees an apple in a tree. The Indian tries to reach for the apple, but then the Indian falls. A cowboy wants to help the Indian, so the cowboy brings the Indian a ladder. The cowboy picks the apple and the cowboy gives the apple to the Indian.
(2) An Indian sees an apple in a tree. He tries to reach for it, but then he falls. A cowboy wants to help him, so he brings him a ladder. The cowboy picks the apple and gives it to the Indian.
these discourse anaphora can take, either a type of NP or pronoun, are distributed along the degree of ‘givenness’ between the speaker and the hearer (Gundel, Hedberg, & Zacharski, 1993). A character that is new to the conversation is usually introduced with an indefinite NP. As soon as this referent is established or ‘given,’ and both speaker and hearer know who it is, it can be referred to with a pronoun. Application of these rules to example (1) results in a linguistically more economical version, example (2).
Nevertheless, there is yet another group of language users who can be expected to experience difficulties in their performance: second language speakers. In contrast to children and elderly, adult second language speakers are in the possession of both a developed Theory of Mind and a sufficiently large working memory capacity. The factor that might pose a problem is that they are not speaking their first language; this could pose such a high demand to their cognitive resources, that the use of discourse anaphora might be compromised. Another possibility is the interference of structures from the L1 into the L2.
This study focuses on the performance of second language speakers regarding production and comprehension of discourse anaphora. More specifically, second language speakers using Dutch with Italian as their mother tongue will be studied. Dutch, similar to English, is a non pro-drop language, whereas Italian is a pro-drop language. This entails that in Italian, when a referent is ‘given,’ the subject anaphor may be omitted, or dropped. Consider the following example (3):
(3) Un uomo cammina nella via. Porta un cappotto blu.
A man walks in-the street. Wears a coat blue
‘A man is walking in the street. He is wearing a blue coat.’
Notice that in the second Italian sentence, there is no overt pronoun marking the subject of the verb porta ‘wear’ - this phenomenon is referred to as a ‘null-subject.’ When speakers of Italian learn Dutch, they have to learn that in Dutch every finite verb has to have an overt subject1. The fact that Italian L2 speakers of Dutch must keep in mind this new rule, along
1
This is however not true for the second verb in a conjunct, which is similar to English. Consider example (i): (i) De man loopt op straat en ziet een hond.
with all other differences between Italian and Dutch, could be a factor influencing their successfulness of the discourse structure.
The question this study seeks to answer is: Are adult second language speakers able to take the perspective of others in a discourse into account with regard to discourse anaphora? In other words: Are they able to produce resolvable subject anaphora and can they
successfully interpret them? Hendriks et al. (2008) found an asymmetry in the performance of production and comprehension of elderly Dutch native speakers: The elderly were not able to take into account the opposite perspective in the production task, whereas their performance in comprehension was similar to that of the young adults. Wubs et al. (2009), who studied subject anaphora in children, found they did not take into account the other’s perspective in both the production task and the comprehension task. The present study will explore subject anaphora in second language speakers and whether they display an asymmetry in production and comprehension as well.
Researchers are debating whether L2 proficiency is related to the transfer of pragmatic competence from the native language. Takahashi (1996), studying request strategies, found no evidence for this relation, and neither did Pennock Speck & Suau Jiménez (1998) for text coherence. It is argued that discourse coherence is a pragmatic competence, rather than a linguistic competence, the question remains what factors do influence the successfulness of second language speakers with regard to discourse structure. Therefore, this study will investigate which factors are important for the appropriate use of subject anaphora by second language speakers.
Theories. Previous studies that used these discourse rules in an OT framework to investigate
the asymmetries in production and comprehension of subject anaphora in children and elderly (Wubs, 2008; Wubs et al., 2009; Hendriks et al., 2008) are presented in section 2.2. As this study concerns second language speakers instead of native speakers from the previous studies, a section on second language acquisition is in order. In section 2.3 relevant studies on second language acquisition are presented, as well as some more specific studies on speakers with a pro-drop first language learning a non-pro-drop second language. The two hypotheses distilled from this theoretical background are stated in section 2.4.
To investigate the use of referring expressions by Italian speakers of Dutch, twenty Italian adults who had learned Dutch after childhood and twenty Dutch adults participated in two experiments that were modifications of those used by Wubs (2008) and Hendriks et al. (2008). The participants’ performance on the production and the comprehension of subject anaphora was tested, as well as their Dutch proficiency. The Italian participants were also given a questionnaire, to retrieve background information. In chapter 4, following chapter 3 on methodology, the results of the experiments will be discussed in relation to the hypotheses presented in section 2.4. Chapter 5 will discuss these results and give an answer to the
2. Theoretical Background
From the previous section it has become clear that groups of speakers such as children and elderly people have difficulty producing and interpreting the correct referring expression. It was argued that it is worthwhile to investigate the performance of second language speakers. In this section the relevant literature is reviewed, with section 2.1 Linguistic Theories, dealing with Optimality Theory, the Givenness Hierarchy and Centering Theory. Section 2.2 Previous
Studies describes the previous studies on production and comprehension of subject anaphora,
whereas section 2.3 Second Language Acquisition reviews the phenomena of L1 influence and the pro-drop parameter in Italian, and earlier studies in this area. The background section is concluded by section 2.4, Hypotheses, predicting the behaviour of the second language speakers in this study.
2.1 Linguistic Theories
First, it is necessary to define what discourse and discourse anaphors are. Discourse can be understood as written or oral communication. Discourse anaphors are the expressions that are used to refer to entities within this discourse. Malt (1985) gives an example of how a sentence can be more informative within a discourse than on its own: “Yes, he did” implies only that a male character has done something. If the previous utterance had been: “Did John ever return your book?” it becomes clear that he refers to John, and the thing he did is returning the book. This shows how previously uttered information is important for the determining of the
functions as well (he did). These subject anaphora are bound to certain sets of rules, among others the Givenness Hierarchy (Gundel, Hedberg & Zacharski, 1993) and Pro-Top from Centering Theory (Grosz, Joshi & Weinstein, 1995).
Gundel and her colleagues claim that “different determiners and pronominal forms conventionally signal different cognitive statuses (information about location in memory and attention state), thereby enabling the addressee to restrict the set of possible referents” (1993: 274-275). They propose a Givenness Hierarchy which prescribes what form should be used for which cognitive status. By cognitive status of a referent, they mean the hearer’s
knowledge of and the attention towards the referent. This hierarchy is presented in Table 1:
Table 1
The Givenness Hierarchy (Gundel et al., 1993)
in focus activated familiar uniquely identifiable referential type identifiable
it that this this N
that N the N indefinite this N a N
When a referent is in focus, that is, when the referent is at the current centre of attention, a personal pronoun can be used. When the referent is activated, which is a little less given than
in focus but still represented in current short-term memory, the referent can be referred to with
a demonstrative pronoun or a demonstrative NP. When the referent is familiar, a demonstrative NP can be used. Then, again less given, when a referent is uniquely
identifiable, it can be referred to with a definite NP. A referential referent can be described
the “addressee is able to access a representation of the type of object described by the expression” (1995: 276).
In predicting the actual distribution of referring expressions in discourse, the Givenness Hierarchy interacts with Grice’s maxim of quantity. Grice’s maxim of quantity states that a speaker should conform to the following rules (Grice, 1989):
1. Make your contribution as informative as required.
2. Do not make your contribution more informative as required.
A referent, for example, in focus also entails all lower statuses, as it is also activated and
familiar and so on. As a result the associated forms with these lower statuses can also be used
for the higher status, which means that with different referring expressions the same cognitive status can be described. Gundel et al. (1993, p. 294) give the following example (4) to
illustrate this:
(4) These incredibly small magnetic bubbles are the vanguard of a new generation of
ultradense memory-storage systems. These systems
Those systems The systems
New generation ultradense memory-storage systems are extremely rugged: they are resistant to radiation and are non-volatile.
instance, that a pronoun is preferred over a NP. Wubs et al. (2009) derive from this a rule for situations in which the speaker can choose either a pronoun or a NP: Avoid full NPs. To get an impression of how the Givenness Hierarchy works in simple discourse, consider the short story in the following example (5):
(5) A boy and a girl walk in the street. The boy is getting cold. He puts on his hat.
If the indefinite NP, the definite NP, and the pronoun were in any other place in the story, it would no longer make any logical sense: the discourse structure would be incomprehensible.
Whereas the Givenness Hierarchy mostly deals with the cognitive status of referring expressions in a conversation more or less in isolation, Centering Theory proposed by Grosz et al. (1995) models the interaction of cognitive status (or focus of attention), choice of referring expression, and the coherence of a discourse. They consider discourse coherence in the context of the relation within a sequence of sentences. Similar to the ideas of Gundel et al. (1993), they claim that pronouns and NPs have different effects on a hearer; a wrong choice on the part of the speaker may mislead the hearer in his or her interpretation. A referent has to be introduced with a NP before it can be referred to with a pronoun, and pronouns should be retraceable to the intended referent. Consider the following example (6) by Grosz et al. (p. 8):
(6) a. Terry really goofs sometimes.
b. Yesterday was a beautiful day and he was excited about trying out his new sailboat. c. He wanted Tony to join him on a sailing expedition.
d. He called him at 6AM.
The pronoun in (6e) is ambiguous: Semantic plausibility points at Tony to be the referent of
he, but hearers initially interpret the pronoun as a referring expression for Terry. Terry has
been the focus of attention throughout the previous sentences and the use of a pronoun suggests he still is. Centering Theory suggests that the cognitively least demanding way to continue the focus on a referent is by using a pronoun. However, since the pronoun leads to an incorrect referent, sentence (6e) is not very coherent. Centering Theory proposes a guide by which coherence of a discourse can be measured, and part of this concerns the use of pronouns. Every utterance has a backward looking center and a forward looking center. A forward looking center (Cf) is any referent in an utterance. These are ranked according to prominence: a subject is more prominent than a direct object, which in turn is more prominent than an indirect object etcetera. The most prominent referent is the preferred center (Cp). The anaphor in an utterance that is connected to the (Cp) in the previous utterance is the backward
looking center (Cb). Others refer to this (Cb) as topic (see Wubs, 2008). An important rule
Grosz et al. derive is that a pronoun, rather than a NP according to the pronoun rule, should always refer to the topic. We can therefore conclude that when an utterance contains a single pronoun, this pronoun is topic. Example (7) illustrates the assignment of topic hood to anaphora:
(7) a. An Indian sees an apple.
(Cf) = {Indian, apple}; (Cp) = an Indian; (Cb) = - b. He tries to reach for the apple. (he = Indian)
The forward looking centers in utterance (7a) are an Indian and an apple. These are both referring expressions. An Indian is the preferred center, because of the two (Cf)s it is the most prominent center, namely the subject. Utterance (7a) has no topic (Cb), because there is no previous utterance. Utterance (7b) has two forward looking centers, he and the apple. He is the preferred center, because it is the subject and therefore more prominent than the direct object the apple. The backward looking center, or the topic of this utterance, is he, because it was the preferred center of utterance (7a). Utterance (7c) has only one forward looking center, which is then automatically also the preferred center, and as this referent was already
preferred center in (7b) it is the backward looking center in this utterance. The relationship between these three centers is as follows (Kehler, 1997):
1. An anaphor that is the (Cb) in the current utterance, (Cb) in the previous utterance, and by definition the (Cp) in the current utterance, will most likely be the (Cb) in the next utterance. This is referred to as Center Continuation.
2. An anaphor that is the (Cb) in the current utterance, (Cb) in the previous utterance, but not the (Cp) in the current utterance, will not likely be the (Cb) in the next utterance. This is referred to as Center Retaining.
3. When an anaphor is the (Cb) in the current utterance, but is not the (Cb) in the next, this is referred to as Center Shifting.
In this paper centers will be referred to as topics, therefore the terms topic continuation, topic retaining and topic shifting will be used. Sentence (7a) does not have such a relation; there is no (Cb) established since there is no previous sentence. The same applies to (7b), because this sentence has no (Cb) in its previous sentence. (7c), on the other hand, is an example of topic
continuation: both the current and previous (Cb) and the (Cp) refer to Indian.
subject anaphors are generated by the rules suggested by the Givenness Hierarchy and Centering Theory. To investigate the production and comprehension of subject anaphora, Hendriks et al. (2008) and Wubs (2008)/Wubs et al. (2009) studied their participants’ behaviour on Topic Shifting in short stories. More on these previous studies can be found in section 2.2.
Centering Theory provides rules, according to which speakers continue, retain or shift the topic. Beaver (2004), who reformulates the Centering Theory rules in the terms of
Optimality Theory, summarizes these topic hood rules as follows: “The topic of a sentence is the entity referred to in both the current and the previous sentence, such that the relevant referring expression in the previous sentence was minimally oblique” (p. 14). He also proposes a ranking of constraints that apply to topicalization, the most relevant being:
1. Pro-Top: The topic is pronominalized. 2. Align: The topic is in subject position.
These constraints, together with the rule that the (Cb) is chosen from the (Cf)s of the previous sentence, were used by Wubs (2008) to formulate a guide for defining topics and topic shifts in the stories produced by her participants. This guide will be further elaborated on in section 3.4 of the next chapter.
It is a common observation that linguistic rules are regularly violated. Prince and Smolensky (2004) propose Optimality Theory, which deals with conflicting constraints in phonetics, but can also be applied to grammatical issues. Optimality Theory states that the output should “satisfy the conflicting constraint set as well as possible” (2004, p. 3). In order to achieve the most optimal, well-formed structures, the conflicting constraints are ranked in a strict domination hierarchy. In this hierarchy, each constraint has priority over those that are ranked below it. It is therefore necessary to satisfy the higher ranked constraint rather than the lower ranked constraint. In an Optimality Theory view, the rules that were derived from the Givenness Hierarchy and Pro-Top from Centering Theory are regarded as conflicting constraints in the use of subject anaphors (Hendriks et al., 2008; Wubs, 2008; Wubs et al., 2009).
In Optimality Theory, two conflicting constraints each present a different choice for a certain form to express the intended meaning. With regard to subject anaphora, when a speaker wants to express a topic, he or she can choose between using a NP or a pronoun. The meanings, forms and constraints of this production dilemma are presented in an OT (syntax) tableau. This mechanism will be explained by the tableaux for subject anaphora from Wubs et al. (2009).
Table 2
The OT Tableau for Unidirectional Optimization of a [+Topic] Meaning
Input: + topic Avoid full NPs Pro-Top
☞ <pronoun>
<full NP> *!
The input (+ topic) is the meaning the speaker wants to express, the forms (<pronoun>, <full NP>) are the possible linguistic choices he or she has for expressing this meaning, and the violable constraints (Avoid full NPs, Pro-Top) are ranked from most important to the left to least important to the right. An asterisk denotes a violation of a constraint, and when the violation is fatal, an exclamation mark is added. One can see that for the form <full NP> the first and most important constraint, Avoid full NPs, is violated. For the form <pronoun> no constraints are violated. This results in the acceptance of <pronoun>: This is the optimal form to express the intended meaning, indicated by the pointing finger. A similar tableau can be designed for the expression of a (– topic) meaning, see Table 3:
Table 3
The OT Tableau for Unidirectional Optimization of a [-Topic] Meaning
Input: - topic Avoid full NPs Pro-Top
☞ <pronoun> *
<full NP> *!
The optimal choice for a speaker to express a referent with a (- topic) meaning is a
pronoun when he or she wants to express the topic, and to use a pronoun to refer to a referent that is not the topic as well.
A conversation usually consists of both a speaker who produces utterances and a hearer who has to interpret them. To illustrate the process of interpreting utterances, Table 4 is presented, the OT tableau for the interpretation of pronouns, from Wubs et al. (2009):
Table 4
The OT Tableau for Unidirectional Optimization of Pronoun Interpretation
Input: pronoun Avoid full NPs Pro-Top
☞ <+ topic>
<- topic> *
The hearer hears a pronoun (input) and based on the constraint ranking this should be interpreted as referring to a topic, <+ topic>. The <-topic> option is ruled out because it violates the Pro-Top constraint. The same can be done for the interpretation of a full NP, as presented in Table 5:
Table 5
The OT Tableau for Unidirectional Optimization of NP Interpretation
Input: full NP Avoid full NPs Pro-Top
☞ <+ topic> ☞ <- topic>
applicable. A pronoun is predicted to be interpreted as the topic, whereas a NP can be equally well be interpreted as referring to the topic as to a referent that is not the topic.
Discourse is a joint activity of a speaker and a hearer; the speaker has to consider how a hearer will interpret his or her form when trying to pass on a meaning and the hearer has to take the speaker’s intention into account when interpreting the form. In this sense it is also informative what the speaker does not choose to say. The four previous tableaux only reason from either one of the two perspectives. That is what in OT terms is referred to as
unidirectional optimization. When a speaker does take the conversational partner into consideration, one can speak of bidirectional optimization (Blutner, 2000). Adult native speakers and hearers are argued to be able to take into account each other’s perspective in a discourse, and select the optimal form-meaning pair through bidirectional reasoning. In a bidirectional OT tableau, the bidirectionally optimal form-meaning pair is selected, see Table 6, from Wubs et al. (2009):
Table 6
The OT Tableau for Bidirectional Optimization of Subject Anaphora
<form, meaning> Avoid full NPs Pro-Top
<pronoun, + topic>
<pronoun, - topic> *
<full NP, + topic> * <full NP, - topic> *
the two pairs <pronoun, - topic> and <full NP, + topic>. The second optimal pair is then <full NP, - topic> even though that pair violates the first constraint equally to the other candidate <full NP, + topic>. The optimal pairs are indicated with the upwards pointing fingers. Adults can take both the speaker’s and the hearer’s perspective into account, and produce and
interpret pronouns when the referent is the topic, and produce and interpret NPs when the referent is not the topic.
In this first section of chapter 2 the relevant theories, the Givenness Hierarchy, Centering Theory, and Optimality Theory were explained and presented in an OT constraint ranking with regard to subject anaphora. In section 2.2 an overview will be given of two studies that applied these theories to investigate the production and comprehension asymmetry in children and in elderly people.
2.2 Previous Studies
In the introduction of this paper two studies were mentioned, one by Hendriks et al. (2008) on the production and comprehension of subject anaphora by elderly people, and one by Wubs et al. (2009) on the same issue by children. The experiments of this present paper are based on their experiments and findings. This section will give an overview of their methodology and the most important findings.
Hendriks et al. (2008) were the first to study the asymmetry between the production and comprehension of subject anaphora in the elderly. They wanted to know whether
limited cognitive resources, it was conceivable that elderly participants with a limited working memory capacity would show similar results with regard to subject anaphora. In order to study these possible difficulties with bidirectional optimization, Hendriks et al. (2008) conducted a production experiment and a comprehension experiment, the results of which they correlated with working memory test scores.
Twenty-five elderly and twenty-five young Dutch adults participated in the production experiment. It consisted of eight picture storybooks, each containing six pictures. The stories were all of the same structure and depicted two characters of the same gender. There was a main character and a secondary character. The secondary character was only present in pictures 2, 3, and 4, and was the actor only in picture 4. Then the main character was present and actor again in pictures 5 and 6. The participants were presented the pictures one-by-one and asked to briefly describe what was happening in each picture. This storybook structure was designed to elicit a discourse containing a mid-story Topic Shift from main character to secondary character and then a re-introduction of the main character towards the end of the story. The researchers identified the topics (cf. Grosz et al., 1995) in each utterance and then concluded whether the participant had produced a successful topic shift. In the case of a successful topic shift, they examined the referring expression used for the later re-introduction of the main character: A NP or a pronoun. The elderly produced almost ten times more
pronouns than the younger adult group (elderly 31%; young adults 3.5%) at this point, leading the researchers to conclude that elderly have trouble with bidirectional optimization in
production. They credited this difficulty to their limited working memory, as the memory test scores were strongly correlated with number of NPs.
potentially ambiguous pronoun, of which the participants were asked who it referred to. Assuming that pronouns preferably refer to topics, it was expected that the participants would choose the main character if they considered that the topic was continued, or the secondary character if they interpreted the old topic (the original first actor) as shifted to the second character. The researchers concluded that both groups chose the main character half of the time (elderly 48.5%; young adults 46.5%), and the secondary character the other half.
Hendriks et al. (2008) concluded that elderly adults displayed a deficient production of subject anaphora, whereas their comprehension of pronouns was similar to that of the young adults. This might have been due to the structure of the stories, where the production of a topic shift did not seem to be unavoidable. Being a speaker, the elderly adults optimize only unidirectionally, but the data were too sensitive to plausibility effects to conclude that as a hearer the elderly (or the young adults, for that matter) are able to perform bidirectionally.
Wubs (2008) conducted a similar study with Dutch children (aged 4;3 – 6;5) and an adult control group as participants. The experiments were adapted from Hendriks et al. (2008) but improved. The production experiment consisted of four picture story books, each
containing six pictures and designed to elicit a topic shift. Similar to the elderly, the children produced significantly more pronouns when reintroducing the main character (children 51%; adults 3%). When the children wanted to express a referent that is not the topic, they often used a pronoun, because for the speaker that is the optimal form for that situation, exactly as the unidirectional optimization tableaux predict.
the adults did. The unidirectional OT tableau (Table 5) predicts that when children encounter a NP, this leads them to interpret the NP as being either a topic or not a topic. In the topic shift stories, they have a 50% chance of not interpreting the NP signalling a topic shift. The
potentially ambiguous pronoun in the last sentence of the stories would in their eyes then still refer to the first character. And indeed, in the topic shift stories the children interpreted 50% of the time the pronoun referring to the first character, whereas adults did so 35% of the time. Therefore it was concluded that also in comprehension, children were not able to perform bidirectionally. Wubs (2008) gave the children a working memory test and found high correlations between pronoun production and working memory capacity. She concluded that the children’s performance is affected by their smaller working memory capacity, but she also argues that an underdeveloped Theory of Mind might be a factor in their unidirectional
optimization.
Table 6 in section 2.1 showed the bidirectional optimization of forms and functions in subject anaphora, which guides the selection of referring expressions in adult native speakers. To explain why it is conceivable that adult second language speakers might not be able to optimize bidirectionally, section 2.3 will deal with the characteristics of second language learning, the Italian language, and previous research in bilingualism.
2.3 Second Language Acquisition
Researchers agree that first language acquisition differs from second language acquisition (SLA) in terms of the ease with which the language is learnt. Regardless of how first language acquisition is believed to take place, either through Chomskyan Universal Grammar or Emergent Grammar (Hopper, 1998) or yet differently, there is a difference reported between first and second language acquisition, especially after a certain age. The critical period hypothesis, that claims that language learning becomes more difficult as the speaker’s age increases, is not only valid for first language acquisition, but also for SLA (Pinker, 1994). Bialystok (1997), however, concluded that maturation of the brain, closely related to age, is not the only important factor for success in SLA; also the length of residence in the country of the second language and the time speaking it are strong predictors. She also found that the “correspondence between language structures in the first and second language” (p. 116) is an important factor. Birdsong and Molis (2001) confirm these findings. Therefore these factors will be taken into account in this study.
Learning a second language is different from learning a first language, in that there is already knowledge present of another completely acquired language. In addition to this native language, new second-language grammar features, vocabulary, morphology, and phonology have to be learned, and kept separate from the first language. Sometimes L1 knowledge can be helpful in learning a second language, for example cognates in vocabulary learning. The English word ‘mathematics’ would be easier to learn for a French person who knows the word ‘mathématique’ in the first language, than it is for a Dutch person who knows the word
In Optimality Theory, first language acquisition is similar for speakers of different languages, in that the constraints are there. They only need to be re-ranked according to the target language (Blutner, Hendriks, & De Hoop, 2006). Based on the language input, the speaker will adjust his or her constraint ranking. When learning a second language, a new ranking needs to be learned, this can be difficult when one only has the constraint ranking from the first language. Picco (2005, p. 2) defines transfer as “transfer of structure, arising when the learner uses some L1 feature, be it phonological, lexical, grammatical or pragmatic, rather than that of the TL,” in which TL stands for target language, the second language to be learnt. Applying this to Optimality Theory, transfer of a L1 structure might be explained by a constraint ranking that is different from the target language. The speaker still uses the
constraint ranking from his or her L1 in the L2.
(8) Un uomo cammina nella via. Porta un cappotto blu.
A man walks in-the street. Wears a coat blue
‘A man is walking in the street. He is wearing a blue coat.’
Applying the rules from Centering Theory (Grosz et al., 1995), explained in section 2.1, the subject of the second sentence is the discourse topic, and is therefore allowed to be omitted. This is not allowed for English (or Dutch): There, the overt pronoun ‘he’ is obligatory. Italians can interpret null subjects because of the inflectional system of the language. To illustrate this, the following example is taken from Di Eugenio (1990, p. 273):
(9a) Mariai voleva andare al mare. ‘Mariai wanted to go to the sea.’ (9b) (proi) Telefono a Giovannij.
‘Shei called Giovannij up.’
(9c) 1. (proi) Si arrabbio perche (proi) non loj trovo a casa. ‘Shei got angry because shei did not find himj at home.’ 2. (proi/?j) Si arrabbio perche (proj) stava dormendo. ‘Shei/?Hej got angry because hej was sleeping.’ 3. Luij si arrabbio perche (proj) stava dormendo. ‘Hej got angry because hej was sleeping.’
4. (proj) Si e'arrabbiato perche (proj) stava dormendo. ‘Hej has gotten angry(-masc.) because hej was sleeping.’
can only be interpreted as referring to Maria, as she was the topic of (9b) and also of (9c.1) , hence an example of topic continuation. Notice that Giovanni is referred to by an overt pronoun lo, because this referent had not been assigned the topic status in (9b). The null subject in (9c.2) could refer either to Maria or Giovanni, but the discourse structure makes it more conceivable that Maria is the intended referent, remember the Pro-Top constraint that says that a pronoun (a null pronoun is also a pronoun) refers to the topic. But when applying this reasoning to the second null subject of the sentence, the semantics (Maria is sleeping) clashes with the logical truth (Maria was making the phone call). Also, the masculine morphology on the past participle stava dormendo indicates the subject should be male. The null subject and therefore also the previous null subject would most logically refer to
Giovanni. In (9c.3) the overt pronoun lui is used, to mark a topic shift, from Maria to Giovanni. Would Giovanni be a girl, the topic shift might still be marked with a pronoun, because if it were to refer to Maria, a null subject would have been used. In both hypothetical cases, the past participle morphology should then be feminine. The null subject in (9c.4) can only refer to Giovanni, because of the masculine morphology on the past participle. In terms of discourse coherence, (9c.3) is a better continuation of (9b) than (9c.2) is.
As was explained before, languages differ only in the ranking of constraints. Hendriks, De Hoop, Krämer, De Swart, and Zwarts (2010) give an example of how the constraints are ranked for English and Italian with regard to the realization of subjects. In English, which is similar to Dutch in this respect, all clauses must have a subject (SUBJECT), and all
Table 7
OT Tableau for Subjects in English
Input: “It is raining” SUBJECT FULL-INTERPRETATION
☞ It is raining. *
Is raining. *!
For English, the SUBJECT constraint is ranked above FULL-INTERPRETATION. Although the subject of “it is raining” does not bear any meaning, it is the most optimal choice, since “is raining” violates the more important constraint SUBJECT. For Italian, however, these
constraints are ranked differently (see also Grimshaw and Samek-Lodovici, 1998). For Italian, the constraint FULL-INTERPRETATION is ranked highest. Hendriks et al. (2010) present the following OT tableau for Italian, Table 8:
Table 8
OT Tableau for Subjects in Italian
Input: “It is raining” FULL-INTERPRETATION SUBJECT
EXPL piove. *!
☞ Piove. *
Italian does not have expletives, because all constituents in a sentence must have a meaningful interpretation. Therefore “piove” in which the subject is dropped, is the optimal form.
Also, subjects that are dropped should always refer to the topic, which is a small adaptation of the constraint for the purpose of this study. In Dutch it is more important to realise the subject, than to drop the topic. Consequently, the SUBJECT constraint is ranked higher than Drop-Top. In addition to the basic set of constraints, the tableaux show the subject anaphors this study is concerned with: Null subjects, pronouns, and NPs. Furthermore, there are the two meanings being studied: + topic and – topic, leading to the six form-meaning candidates. Table 9 shows the OT tableau for bidirectional optimization of subject anaphora in Dutch:
Table 9
OT Tableau for Subject Anaphora in Dutch
DUTCH SUBJECT Avoid NP Pro-Top Drop-Top
<null, +topic> * <null, -topic> * * A<pronoun, +topic> * <pronoun, -topic> * <NP, +topic> * * A<NP, -topic> *
Table 10
OT Tableau for Subject Anaphora in Italian
ITALIAN Avoid NP Pro-Top Drop-Top SUBJECT
A<null, +topic> * <null, -topic> * * <pronoun, +topic> * A<pronoun, -topic> * <NP, +topic> * * <NP, -topic> *
<Null, + topic> is the first optimal pair, because it violates only the SUBJECT constraint. Ruling out all other pairs containing either a null or + topic, <pronoun, - topic> is the second optimal pair. One nevertheless has to keep in mind that this is a simplified representation of the use and interpretation of subject anaphora in Dutch and Italian, for illustration purposes only. An exhaustive representation of subject anaphora involves many more constraints and more complex form-meaning pairs, but moves beyond the scope of this paper. In Italian for example, to allow for the use of NPs the – topic meaning needs to be further categorized to distinguish situations in which pronouns and NPs are used. One could also think of a further distinction between definite and indefinite NPs, depending on the givenness of a referent.
When comparing the Dutch and Italian tableaux, it becomes visible how the re-ranking of the constraint SUBJECT causes the difference between the two languages. If Italian
speakers would apply the Italian constraint ranking in Dutch, one would see dropped topics, and pronouns that are not necessarily referring to a topic.
subjects were omitted. The participants judged many of the incorrect sentences being correct, a phenomenon White attributes to interference from the pro-drop L1. In her UG view, the parameters for pro-drop in the participants’ mother tongue were set and caused interference when they tried to perform in the second language. She found that proficiency in L2 English was highly correlated with the number of errors.
In contrast to White, Galasso (2002) does not believe in full-accessible Universal Grammar. He studied the L1 Spanish influence on L2 English not in terms of parameter resetting, but in terms of language specific problem solving mechanisms. He gathered spontaneous speech of Spanish students of English in a classroom setting and analysed the occurrence of pro-drop constructions. He found a strong correlation between duration of contact with the second language and correct overt pronoun use, which entails a correlation between proficiency and correct pronoun use.
Sorace and Filiaci (2006) studied the comprehension of overt and null subject
pronominals in English near-native speakers of Italian. They found that both native speakers and near-native speakers correctly interpreted a null subject in a complement clause to refer to the subject of the main clause, but that the near-native speakers significantly more often than the native speakers interpreted an overt pronoun in a complement clause to refer to the subject of the main clause. This indicates that the different constraint ranking between English and Italian also works the other way around: The English constraint SUBJECT is higher ranked in the near-natives’ Italian, than in the natives’ Italian.
2.4 Hypotheses
This study investigates whether Italian speakers of Dutch correctly produce and comprehend subject anaphora in Dutch. If they display a non-target performance, the question is whether this is caused by either the inability to optimize bidirectionally or by the influence of their Italian constraint ranking. Hendriks et al. (2008) claim that
Because bidirectional optimization is more complex and hence requires more cognitive resources than unidirectional optimization, it is conceivable that under certain circumstances also adults may fail to optimize bidirectionally and produce a unidirectionally optimal output instead (p. 451).
Children and elderly adults were shown to have problems with bidirectionality, which may be also true for Italian speakers of Dutch. Speaking and hearing a second language may be a circumstance that imposes additional load on the cognitive resources available for language production and comprehension. Along the same line of reasoning, Sorace and Filiaci (2006) describe how a situation of processing overload may lead to a default strategy in which a speaker resorts to overt pronouns, even when null subjects would be allowed. Additionally, Sorace et al. (2009) showed that being bilingual (studying all possible combinations of pro-drop and non-pro-pro-drop language pairs) does cause problems in use of pronouns, and is not necessarily related to the influence of pro-drop. One can therefore predict that Italian L2 speakers of Dutch will not have sufficient cognitive resources to use subject anaphora bidirectionally.
proficiency, length of residence and duration of speaking the L2. A correlation between these factors and performance on the tasks is possible. In an OT framework, pro-drop induced problems with subject anaphora are caused by a difference in constraint ranking: For the Italian speakers pro-drop, formulated in the constraint Drop-Top, is considered a constraint in the optimization for subject anaphora, whereas the Dutch speakers do not have this constraint ranked so prominently. In production, Italian speakers of Dutch who apply the Italian
Table 11
Predictions for the Successfulness of Subject Anaphora use by Italian L2 Speakers of Dutch
Target Dutch Hypothesis 1:
Unidirectional Dutch
Hypothesis 2: Italian constraint ranking
Production Will use pronouns for topics, and NPs for the
reintroduction of the second character
Will use pronouns for both + topics and – topics, including the
reintroduction of the second character
Will use null subjects for topics,
pronouns for the reintroduction of the second character
Comprehension Will interpret a NP
signalling a topic shift, and interpret pronouns as topic continuation
Will interpret pronouns as referring to a topic, and NPs as referring to both + topics and - topics
Will interpret both NPs and pronouns signalling a topic shift
In addition to the NP-pronoun distinction in the production task, this study will also examine more specifically the types of NPs and pronouns used. The Givenness Hierarchy in section 2.1 prescribes the forms corresponding to the different cognitive statuses, and the question is whether the Italian and Dutch speakers follow this in their production of subject anaphora.
3. Method
The research question is whether Italian speakers of Dutch are able to optimize
bidirectionally, and how much their performance is related to their pro-drop background. To test which of these hypotheses is true, two experiments have been conducted, which will be explained in this chapter. This chapter will provide information on the participants who were included, the tests that were used for the production and comprehension of topic shifts, the language questionnaire and the proficiency test that were administered, and how these will be analysed.
3.1 Participants
In this study, 21 Italian adults and 20 Dutch adult controls participated. The Italian
male and 5 female participants in the Italian group and their education ages (age at which the participant had their last schooling) range from 15 to 50. The two highest values were from two participants taking the official Dutch finals after part time schooling at a considerably late age. Leaving them out, the education ages range from 15 to 37.
Also a control group was tested. The participants in the control group were 20 Dutch adults, with ages ranging from 20 to 61. There were 8 male and 12 female participants in the Dutch group and their education ages varied from 16 to 34, see Table 12. The control subjects did not take the language questionnaire or the Dutch proficiency test, as it was safe to assume that they all had a native proficiency. None of the participants reported any language
disorders.
Table 12
Age and Education Age of the Italian and Dutch Participants
Participants Italian Dutch
Age 24-64 µ = 38.55; sd = 11.77 20-61 µ = 36.55; sd = 14.13 Education Age 15-50 µ = 27.2; sd = 8.28 16-34 µ = 22.5; sd = 4.38 3.2 Materials Production task
The materials for the production and comprehension task were based on the Wubs
less meaningful objects were in black and white. This colour scheme should make it more likely for the participants to focus on the actions of the story characters. See Figure 1 for the 2-picture policeman/criminal practice story:
Figure 1. A practice story from the production experiment.
Figure 2. A test story from the production experiment.
The structure for all four stories is presented in Table 13 based on Wubs (2008):
Table 13
Structure of the Production Experiment Stories
Main character Sec.character
Picture In picture? Actor? In picture? Actor?
1 Yes Yes No -
2 Yes Yes No -
3 Yes Yes Yes No
4 Yes No Yes Yes
5 Yes No Yes Yes
6 Yes Yes No -
the discourse topic, after which the first character is re-introduced to become the discourse topic again.
Story 1 shows a dancer watering flowers, a nurse picking the flowers and the dancer wearing one of the flowers. Story 2 shows a pirate kicking a ball in the water, a knight fishing the ball out of the water, and the pirate reunited with the ball. Story 3 shows a princess buying an ice-cream cone, giving the cone to a witch, the witch eating the cone, and the princess buying a new one. Story 4 shows an Indian trying to reach for an apple, a cowboy picking the apple, and the Indian again with the apple. This last story is different from the cowboy/Indian story in Wubs (2008)/Wubs et al. (2009), in which the cowboy took the ladder away from the Indian half-way through the story and left the Indian stuck in the tree. This previous version overly concentrated upon the Indian’s dilemma and resulted in an aberrant discourse structure in the narratives of both the Dutch children and adult control participants. For the pictures of the production task stories, see Appendix 1.
Comprehension task
Hij wil een toren van de blokken bouwen.
‘He wants to build a tower with the blocks.’
Hij gaat bij een timmerman aan tafel zitten.
‘He sits at the table with a carpenter.’
De timmerman helpt mee de toren van de voetballer te bouwen.
‘The carpenter helps building the soccer player’s tower.’
Per ongeluk stoot de timmerman de toren van de voetballer om.
‘Accidentally the carpenter pushes the soccer player’s tower.’
Hij schrikt er een beetje van.
‘He is a bit startled by it.’
Question: Wie is er geschrokken?
‘Who is startled by it?’
Table 14
Structure of the Topic Shift Stories in the Comprehension Experiment
Main character
Secondary character
Sentence Mentioned Actor Referent Mentioned Actor Referent
1 Yes Yes Subject NP No - -
2 Yes Yes Subject
pronoun
No - -
3 Yes Yes Subject
pronoun
Yes No Pp object or
NP
4 Yes No Pp object or NP Yes Yes Subject NP
5 Yes No Pp object or NP Yes Yes Subject NP
6 ? ? Ambiguous
pronoun
? ? Ambiguous
pronoun
The stories without a topic shift also contain two characters, but the first character continues to be the topic throughout the story. An example of such a story without a topic shift and with a concluding question is the hairdresser/singer story, example (11):
(11) Een kapster wil graag muziek maken.
‘A hairdresser wants to make music.’
Ze koopt een blokfluit van haar zakgeld.
‘She buys a recorder from her pocket money.’
Ze gaat met haar blokfluit naar het huis van een zangeres.
‘She takes her recorder to the house of a singer.’
En dan leert de kapster een aantal liedjes van de zangeres.
‘Then the hairdresser learns a couple of songs from the singer.’
Ze maakt heel veel mensen blij met haar mooie muziek.
‘She makes many people happy with her beautiful music.’
Question: Wie maakt de mensen blij met haar mooie muziek?
‘Who makes the people happy with her beautiful music?’
These stories without a topic shift all have the same structure, as presented in Table 15, from Wubs (2008):
Table 15
Structure of the Stories without a Topic Shift in the Comprehension Experiment
Main character
Secondary character
Sentence Mentioned Actor Referent Mentioned Actor Referent
1 Yes Yes Subject NP No - -
2 Yes Yes Subject
pronoun
No - -
3 Yes Yes Subject
pronoun
Yes No Pp object or
NP
4 Yes Yes Subject NP Yes No Pp object or
NP
5 Yes Yes Subject NP Yes No Pp object or
NP
6 ? ? Ambiguous
pronoun
? ? Ambiguous
pronoun
The characters in each story are depicted on an A4-format picture, two by two, but randomly placed on the left or on the right. In this respect the present study differs from Wubs (2008). She did not have these pictures and found that the children often gave other responses than one of the two characters. In our study, the pictures of the characters are shown to the participant prior to his or her hearing of the story and are shown again immediately after the comprehension question has been asked. See Figure 3 for the hairdresser and the singer:
Figure 3. The pictures accompanying the hairdresser/singer story from the comprehension
experiment.
Language questionnaire
specific topics, subdivided in a part on Dutch language experience in general, a part on language use in the home, and a part on language use in the workplace. Questions concerning language in the workplace were allowed to be interpreted as language in the educational setting if applicable (i.e. for students and teachers). A sample 5-point scale on a separate sheet of paper was used for the participants to visualise the possible answers to the self-evaluation questions about their Dutch language skills (bad-weak-reasonable-good-excellent). Topics that might prove to be important are for instance the length of the participants’ stay in the Netherlands, the age of arrival, and the amount of current use of both languages. For the entire questionnaire, see Appendix 3.
Proficiency test
Establishing a proficiency score is useful for determining whether proficiency and
participants. The original test was thoroughly pre-tested and proved to yield a success rate of 80% to 90% amongst “non-attrited native speakers” (Schmid, 2005, p. 11). This result shows that even the near-native Dutch speaking Italians will not easily score at ceiling. The texts and the correct missing words are reproduced in Appendix 4.
3.3 Procedure
The experiments were conducted individually either in a quiet room at the university, or the participants’ home or workplace. The entire experiment was recorded with an Olympus digital voice recorder WS-200S so that transcriptions of the participant’s answers in the production task could be made at a later moment. First the researcher gave a summary of what the participant could expect from the session: They would be telling stories with pictures and answering questions about stories they would be listening to. The participant was asked for permission to make the recordings, after which the researcher explained that a colleague would be listening to the tapes at a later stage without access to the pictures. Therefore, the participant would have to be as explicit as possible in telling the stories.
First, the Italian participant was interviewed on the first part of the language
room for instruction or correction. Then the participant told all four test stories. Occasionally, when participants could not think of the correct Dutch word, the researcher helped out. The researcher made sure not to provide any determiner in front of a noun, only the noun itself.
After the production task, the remainder of the language questionnaire was dealt with. The questionnaire was given in the form of an interview to allow for clarification and
elaboration of the answers. For the self report proficiency scale questions, the researcher showed the participant a sample scale.
Next, the participant did the comprehension task. He or she was instructed not to think too long, and to provide the answer that first came into mind. Before each story, the
researcher shortly showed the picture with the two characters corresponding to the story, and named the characters. The recordings were played from a cd on an AKAI disc man PD-P4000T. The comprehension experiment started with the practice story, after which the pause-button was pushed and the researcher asked the accompanying question. Immediately
thereafter, the two-character picture was presented again and the participant gave an answer. Up to this point there was room for instruction or correction. Then the researcher moved on to the test stories in a similar manner.
Finally, the participant took the proficiency test, which was introduced as a small puzzle. The participant was instructed on how to fill in the c-test. The participant tried to complete each text within five minutes. The three texts were provided on separate sheets of paper and after five minutes the next text was given. Notes were taken by the researcher of the actual time it took the participant to fill in each text. This concluded the test session.
The Dutch control group did not take the proficiency test, neither were they
interviewed according to the language questionnaire. Their test sessions were nevertheless equal in length to the sessions of the Italian participants, as the Dutch participants also served as controls for a parallel study that was conducted at the same time, which included some additional tests. Half of the control group was consequently tested by another researcher.
3.4 Design and Analyses
The production experiment was designed to elicit topic shifts. According to the Optimality Theory analysis, discussed in chapter 2.1, successful topic shifts and appropriate reference to the reintroduced character indicate bidirectional linguistic performance. Wubs (2008)
developed guidelines for the scorer to determine topics, topic shifts and target items in language production, based on Grosz et al. (1995) and Beaver (2004). First, the scorer has to determine the discourse topic:
1. Take the utterance of which you want to determine the topic. 2. List all referring expressions in that utterance.
3. Discard all referring expressions not referred to in the previous utterance. 4. If there is one pronoun among the residual expressions, this is the topic. 5. If there are no or more than two pronouns left, the subject is the topic. For all utterances, the topics are determined, and noted. Subsequently, it needs to be established whether a first topic shift takes place:
1. If the topic in the description of Picture 3 differs from the topic in the description of Picture 4, the topic has shifted.
3. When there are multiple utterances for a picture, a topic shift is coded when at least one of the utterances from Picture 3 differs from one at Picture 4 or 5.
Following these rules, determine for each story whether a topic shift has taken place. Finally, to determine the target item (i.e., the referring expression used to reintroduce the first
character in each story), the following rules must be applied:
1. Code NP, Pronoun or other, for the first referring expression used to describe the first character in Picture 6.
2. If the first character is described as the subject in Picture 5, code this referring expression.
The results of the production experiment were entered in Excel first, and later in SPSS for further analysis, such as ANOVAs and correlations.
The comprehension experiment was designed to study the comprehension of topic shifts. Through the addition of pictures shown before each story, this was a forced choice task. The participants’ interpretation of the pronoun was determined on the basis of their answer to the comprehension question following the story. In the non-topic shift stories, the potentially ambiguous pronoun was expected to refer to the first character. In the topic shift condition, the potentially ambiguous pronoun refers to the second character. For both
conditions, answers were coded as either ‘Character 1’ or ‘Character 2.’ These answers were entered in Excel, and later in SPSS to perform analyses such as ANOVAs and correlations.
The language questionnaire was designed to provide general personal information and more specific information on language use. The answers were collected in Excel to calculate means and standard deviations for the numerical data. These were also entered in SPSS for correlations.
test seems to focus on reading and writing skills. Nevertheless, the c-test is claimed to “tap into lower-level overall language proficiency, such as knowledge of vocabulary, grammar and idioms” (Keijzer, 2007). C-tests can be scored in a binary manner, with ‘1’ for a correct answer and ‘0’ for an incorrect answer. With this scoring, minor spelling errors are scored as being just as incorrect as an entirely different word. For the present study the more elaborate 0-9 scoring system was used (cf. Keijzer, 2007; Schmid, 2005), to obtain a score as distinctive as possible. The answers were scored using the following criteria to assign points to each item:
0 = empty
1 = incorrect lexical stem and incorrect word class 2 = incorrect lexical stem but correct word class 3 = correct lexical stem but incorrect word class
4 = correct lexical stem, correct word class, agreement error 5 = all of above correct, but still slightly wrong
6 = acceptable variant with spelling error 7 = correct word spelling error
8 = acceptable variant 9 = correct word
The proficiency score is the total score for all items, with a maximum of 540. These scores were used to calculate correlations with the production and comprehension data.
4. Results
4.1 Production task results
In the production task, the Italian participants and the Dutch control participants were asked to describe pictures in a short story. The stories were designed to elicit a topic shift, encouraging the speaker to use a NP, rather than a pronoun which would indicate unidirectional
optimization) to (re-)introduce a character. Their use of NPs would indicate bidirectional optimization. The results of the production experiment were analysed in two data sets: A general set, including NPs and pronouns throughout the entire discourse, and a specific set, including NPs and pronouns at the three moments in the story where a character is (re-) introduced.
In Figure 4, the overall use of pronouns and NPs is presented. Both groups use more NPs than pronouns. Note that the Italian participants used more utterances to describe the pictures than the Dutch controls, as can be seen in Table 16.
Overall use of subject anaphors
0% 20% 40% 60% 80% 100% Control Italian NP Pronoun
A z test showed that these differences were significant (z = -6.93; p <.001). In addition, in Table 16 the frequencies and percentages of the overall use of subject anaphors are presented.
Table 16
Overall Use of Subject Anaphors
Control Italian
Subject Anaphor Frequency Percentage Frequency Percentage
NP 496 70% 440 52.8%
pronoun 212 30% 394 47.2%
To check whether the differences in proportion of pronouns and NPs are due to the difference in total utterances or to other factors, also the proportions of pronouns and NPs without the introductory referring expressions are compared. This resulted in new percentages, presented in Table 17:
Table 17
Use of Subject Anaphors for Continuations
Control Italian
Subject Anaphor Frequency Percentage Frequency Percentage
NP 265 57% 168 32%
pronoun 207 43% 365 68%
demonstrative and personal pronouns, and definite, indefinite, demonstrative and bare NPs. Within the category of NPs, the participants mostly use definite NPs over other NPs, and within the pronouns, they use personal pronouns over other pronouns, see Figure 5:
The overall specific use of subject anaphors
0% 20% 40% 60% 80% 100% Control Italian bare NP dem. NP ind. NP def. NP null P rel. P dem. P pers. P
Figure 5. Graph representing the overall specific use of subject anaphors.
A x2 test showed a significant difference between the Italian and Dutch participants (x2 (7) = 76.3; p <.001). Follow-up z tests for each category shows that this difference is largely due to the percentage of definite NPs (z = 7.019; p <.001) and personal pronouns (z = -6.75;
Table 18
Overall Specified Use of Subject Anaphors
Control Italian
Subject Anaphor Frequency Percentage Frequency Percentage
bare NP 11 1.6% 18 2.2% dem. NP 16 2.3% 12 1.4% ind. NP 90 12.7% 112 13.4% def. NP 379 53.5% 298 35.7% rel. P 1 0.1% 2 0.2% dem. P 53 7.5% 61 7.3% pers. P 158 22.3% 309 37.1% null P 0 0 % 22 2.6%
Before the second analysis of the specific set (NPs and pronouns at the three moments in the stories in which a character is being introduced), could be carried out, the participants’ productions were examined for successful topic shifts. Every story consisted of two
The second analysis investigated the production of pronouns or NPs as a means of introducing or reintroducing a story character in these topic shift productions. For this analysis, the referring expressions at three moments in each story were distinguished. In the first picture of the stories, the participant must introduce the first character, hereafter referred to as Ref 1; in the third or fourth picture the second character should be introduced, Ref 2; and in the fifth or sixth picture the first character is reintroduced, Ref 3. A repeated measures ANOVA was performed on the basis of mean percentages per participant. To guard against departures from normality, the proportions were transformed using the arcsine transformation.
Position (Ref 1 vs. Ref 2 vs. Ref 3) was treated as a within-participants factor with as the
dependent variable pronoun or NP. Group (Italians vs. Controls) was treated as a between-participants factor. To prevent violations of the assumption of sphericity, the Huyn-Feldt correction was applied.
There was only a marginally significant main effect of Position (F (2.37) = 3.24; p <.07), with NPs being produced 97% of the time at Ref 1, NPs produced 98% at Ref 2, and NPs produced 92% at Ref 3. The results are presented in Table 19:
Table 19
Subject Anaphors at the three Referent Points for all Participants
Pronoun NP
Ref 1 3% 97% (SE = 1.6%)
Ref 2 2% 98% (SE = 1.4%)
Ref 3 8% 92% (SE = 3.2%)
Pronouns and NPs for Control Group 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Ref 1 Ref 2 Ref 3
Pronoun NP
Figure 6. The subject anaphora used by the Dutch control participants at Ref 1, 2, and 3.
Pronouns and NPs for Italians
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Ref 1 Ref 2 Ref 3
Pronoun NP
to investigate the proportions of indefinite vs. definite NPs in these three sentences. Because the number of NPs other than definite or indefinite NPs was negligibly small, these were left out of the analysis. Again, an arcsine transformation was used, as well as a Huyn-Feldt correction. Position (Ref 1 vs. Ref 2 vs. Ref 3) was treated as a within-participants factor, as well as Definiteness (indefinite vs. definite). Group (Italians vs. controls) was treated as a between-participants factor.
Main effects were found for Position (F (2.76) = 3.509; p <.05), Definiteness (F (1.38) = 17.33; p <.00), and for interactions between Definiteness * Group (F (1.38) = 8.237;
p <.01), Position * Definiteness (F (2.76) = 43.302; p <.00), and Position * Definiteness *
Group (F (2.76) = 4.925; p =.01). The percentages of NPs used (in complementary
distribution with pronouns used, therefore only NPs are mentioned) are presented in Table 20.
Table 20
NPs used at the three Referent Points
Control Italian
def NP indef NP def NP indef NP
Ref 1 56% 44% (SE = 8%) 40% 60% (SE = 8%)
Ref 2 67% 33% (SE = 7%) 32% 68% (SE = 7%)
Ref 3 97% 3% (SE = 2%) 97% 3% (SE = 2%)
NPs for Control Group 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Ref 1 Ref 2 Ref 3
Ind. NP Def. NP
Figure 8. NPs used by the control participants at Ref 1, 2, and 3.
NPs for Italians 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Ref 1 Ref 2 Ref 3
