Page i
Two peas in a pod... or not?
Investigating Lexical and Phonological Attrition Effects in
Monozygotic Twins
Mina Solanki
S2392364
Nieuwe Boteringestraat 82b
9712 PR,
Groningen,
The Netherlands
0628953383
M.R.Solanki@Student.Rug.nl
MA in European Linguistics
Faculty of Liberal Arts
Rijksuniversiteit Groningen
Thesis Supervisor: Dr. M. Keijzer
Second Reader: Dr. H. Loerts
Page ii
i. Acknowledgements
Page iii
ii. Table of Contents
1. Introduction……….. 1.
2. Theoretical framework: First language attrition…………. 3.
2.1. What is L1 attrition? ………... 3.
2.2. Why does attrition occur? ……….. 5.
2.2.1. Psycholinguistic Theories………. 5.
2.2.2. Linguistic Theories………. 8.
2.3. Predictors of L1 attrition……… 11.
2.3.1. The age of L1 attrition onset………. 11.
2.3.2. Educational Level……… 12.
2.3.3. Gender………... 13.
2.3.4. Length of Residence………. 14.
2.3.5. Motivational Factors………... 14.
2.4. Which domains are most affected by attrition…………... 15.
Page iv
2.4.2. Phonology………. 17.
2.4.3. Syntax and Grammar………... 22.
2.4.4. Morpho-syntax……….………. 24.
2.5. What is the time course of L1 attrition? ………. 25.
2.6. Attrition studies involving siblings……… 27.
2.6.1. Attrition studies involving monozygotic twins…….. 27.
3. Research questions and predictions……….. 29.
4. Methodology……… 31. 4.1. Participants………. 31. 4.2. Materials………. 32. 4.2.1. Lexical Measures……… 32. 4.2.2. Phonological Measures……….... 35. 4.3. Procedure……… 37.
4.4. Design and Analysis……….. 37.
Page v
6. Discussion……… 47.
6.1. Revisiting the research questions………... 54.
7. Conclusion………. 57.
8. References………. 59.
9. Appendices……… 66.
Appendix A. Cloze Tests……….... 66.
Appendix B. Dutch Cloze tests……… 72.
Appendix C. Story Re-telling Transcript, Participant MSS……….. 78.
Appendix D. Story Re-telling Transcript, Participant MRS……….. 80.
Appendix E. MSS’s Type/Token Use……….. 82.
Appendix F. MRS’s Type/Token Use……….. 88.
Appendix G. MSS’s VOT Values……….. 92.
Appendix H. MRS’s VOT Values………...……… 94.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 1
1. Introduction
Monozygotic twins remain a fascination of society at large: two seemingly identical individuals who are still two separate persons are an intriguing phenomenon and topics such as telepathy feature prominently in the public debate. But this fascination extends to research, where past studies have used monozygotic twins to reveal the effects of nature versus nurture on behavior, or neural development, and monozygotic twins also feature widely in genetic research. What is typically under researched, however, is language development in twin pairs.
There are a number of studies that have shown language development in especially monozygotic twins to deviate from the typical path shown in singleton children. Not only is language development typically claimed to be slower (e.g. Tomasello, 1986; Bishop, 2006), but the phenomenon of twin language, where twins allegedly have a unique language that only they understand also features in language development studies involving twins (e.g. Bishop et al., 1998; Thorpe et al., 2001). Most studies also show that, as twins get older, initial language lags or deviations drastically diminish, so that they are no longer different from their non-twin peers (Tomasello, 1986). But while all these studies describe the nature and development of twin language, twin language itself has only very rarely been used to answer long-standing linguistic questions, reminiscent of the way that twins have been used in genetic research. One area of linguistic enquiry where twin language can be used to study previously unanswered question is language attrition. Language attrition describes the non-pathological language erosion that typically follows a break from the mother tongue environment and an immersion in a second language setting. What happens in situations where one of the twins moves to another language environment and the other one stays behind? Monozygotic twins are an ideal way of looking more closely at questions that have defined the research field of attrition for as long as it has existed: questions such as how fast a process attrition is, which domains are most easily affected, and whether it is an irreversible phenomenon or merely temporary.
On a practical level, such studies are hard to carry out, as ‘separated’ monozygotic twins (where one has moved internationally and the other one has not) are hard to come by and subject recruitment is almost by definition accidental.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 3
2. Theoretical framework: First language attrition 2.1 What is L1 attrition?
First language (L1) attrition is generally understood to be the erosion of (parts of) the mother tongue following a disruptive event such as a move. In other words, a person moves from the L1 environment and becomes immersed in an L2 setting. He or she then experiences a dramatic decrease of L1 use and a steep increase of L2 use patterns. At the same time, the exact definition of language attrition is much debated and no one definition of the phenomenon is uniformly adopted across studies. Pavlenko (2004) argues that whilst certain research outcomes may look like L1 language attrition they may not be permanent and thus cannot be categorised as such. This is controversial; as such a restrictive definition excludes cases where loss has certainly taken place, albeit only temporary. In addition, very few longitudinal studies exist in the realm of L1 attrition, making a definition along the lines of Pavlenko’s hard to investigate: studies often do not follow subjects for long enough to see if once lost language parts resurface later on in the trajectory. Pavlenko, then, defines L1 language attrition as “a more or less permanent restructuring, convergence, or loss of previously available phonological and morphosyntactic rules, lexical items, concepts, classification schemas, categorical distinctions and conversational and narrative conventions, exhibited not only in the L2 but also in a monolingual L1 context, and not only in production but also in perception and comprehension” (Pavlenko, 2004: 54). L2 influences on the L1 which cannot be grouped under one of these described dimensions are labeled not as attrition by Pavlenko, but as Crosslinguistic Influence. Examples include borrowing, restructuring, convergence and shift. Pavlenko therefore reasons that attrition is the step which (in some cases) follows Crosslinguistic Influence and is marked not only by shift but by the inability to “produce, perceive, or recognise particular rules, lexical items, concepts, or categorical distinctions due to L2 influence” (Pavlenko, 2004: 47). Pavlenko (2004) adds that L1 attrition would mean that subjects would not be recognised by peers as native speakers of their mother tongues.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 4 [this is] caused in some way by the interference from the L2” (Schmid, 2011: 38). It is thus important to view attrition as a sub-phenomenon within the broader research agenda of bilingualism studies.
Indeed, the interaction between the L1 and L2 is further explored by Köpke (2004:23), who claims that the cause of L1 attrition cannot solely be attributed to lack of exposure and competition from another language system , the way it has often been viewed in the past, but it can also be seen as the outcome of the lack of use of the L1, as the frequency of use of the L1 has “direct consequences on [activation levels], (Köpke, 2004: 23), of linguistic structures in the brain and thus the ease with which such structures are later retrieved; Or more simply put, it is “[t]he loss of skills in L1 due to non-use” (Scherag et al. 2004:98).
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 5 The debate about the primary cause of attrition, it being L1 non- use or L2 attrition has not been resolved so far. However, as has been pointed out above, L1 non-use also plays a role. The foundations underlying this debate are more closely examined below.
2.2 Why does attrition occur? 2.2.1 Psycholinguistic Theories
Many theories within the realm of linguistics and beyond have been hypothesised in order to explain L1 attrition. This section summarises a few possible (psycho)linguistic theories which have been evoked to explain the underlying cause of attrition.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 6 on the ERP waveform in this study suggests that cognate status can “be used as a marker investigating relative activation of the non-target language in real time” (Christoffels, Firk & Schiller, 2007: 203). It is thus the effect of facilitation in using cognates which evidences the activation of both languages in bilingual speakers. Kroll et al.’s study also shows that when the L2 is in use the L1 needs to be suppressed to avoid interference. This finding has been taken up by follow-up studies that have targeted language attrition more specifically, such as Linck et al. (2009) and Levy et al. (2007).
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 7 phonologically related word that an effect can be witnessed. Semantically related words always elicited a facilitation effect, regardless of the number of times words had been named in the L2. In sum, this study can only provide illustration of retrieval induced forgetting theory effects in phonologically related words after ten or more repetitions, perhaps casting doubt on the validity of the conclusions. From these results it can also be gleaned that those most vulnerable to this phenomenon are less fluent speakers who require more inhibition in the native language in order to retrieve a phonologically related word in the second language.
Linck et al. (2009) also make a case for the role of inhibition related to retrieval induced forgetting. In their study (Linck et al. 2009), attention was focused on the effect of L2 immersion on the L1. The participants in this study were native (American) English speakers who studied in Spain for a semester; they were contrasted with a control group who took part in a university Spanish language course but did not have any immersion experience. Before immersion of the experimental group, all participants (in both groups) were tested and were found to have similar scores on measures of cognitive ability and proficiency. After three months, all participants were re-tested using a translation recognition task (tapping receptive skills) and verbal-fluency task (tapping production skills). The results from this study show attenuation to the L1 in the immersed environment, as the L2 immersed students suffered less distractor interference in the translation recognition task and were able to name more Spanish exemplars, but fewer English exemplars than their classroom peers as part of the verbal fluency task. When returning home after the immersion experience the participants were tested again in both tasks. After re-test, participants performed in the L2 to the same degree as on previous tests and verbal fluency in English rebounded. These results are compatible with the inhibitory account, whereby the L1 is inhibited in the L2 context in order to facilitate L2 use, making L1 items more difficult to retrieve, after returning to the L1 the need to inhibit is gone and thus exemplars are as easily retrieved as before the immersion experience (Linck et al, 2009). From this study it then follows that attrition effects are particularly noticeable directly after a move, as it is then that the L1 needs to be most strongly inhibited in order not to interfere with the L2 which is not yet very strong.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 8 languages and that; therefore, these links are stronger in monolinguals. “By this view bilinguals’ relatively decreased ability to activate lexical representations (even in the dominant language) arises because bilinguals split their time using word forms in two different languages and, therefore, spend less time strengthening the connections particular to each language” (Gollan et al. 2005: 1231). Using this hypothesis, language attrition is the result of bilingualism as the links between semantical and lexical representations become weaker when they are shared between two languages, making access to the lexical representation weaker than in a monolingual, as “lexical representations in each language have reduced functional frequency” (Linck et al. 2009:1512). This theory is difficult to apply in an attrition context as it already assumes that from only being a bilingual the ability to activate lexical representations is decreased. It also assumes that when bilinguals return to an L1- dominant context they will still have reduced access to their L1, which was not found in the results of Linck et al. (2009) when participants were re-tested after six months after their return to America. Lastly, another frequently invoked psycholinguistic theory used to explain attrition data and that has in one form or another founds its way into many existing accounts of attrition is the activation threshold hypothesis (Paradis, 1993). This hypothesis holds that linguistic items (which can be words, but also smaller units like phonemes) have activation thresholds. The higher an activation threshold, the harder it is to access a word and vice versa. Frequent and recent use of particular words (or phonemes) lowers the threshold of those items and they will therefore be easier to access in subsequent attempts. This relates to language attrition, as it can be hypothesised that the use of L2 items where L1 items used to be selected raises the thresholds of the L1 items. In other words, if a speaker has been immersed in the L2 environment for some time and uses The L2 as opposed to the L1 on a daily basis, the thresholds for L1 items are raised and L1 items thus become more difficult to access. It is not only the use of the L2, but the inactivity of the L1 which raises these thresholds. In order to re-activate a word with a high threshold, more impulses are required (Paradis, 1997). L1 language attrition can therefore be attributed to the prolonged absence of impulses in the activation of a lexical item.
2.2.2 Linguistic Theories
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 9 have been invoked to explain attrition, the most pertinent of which will be addressed in this section.
The theory of regression has attracted a great deal of attention in theories for language attrition and the exact ‘route’ it takes. The regression hypothesis (originally formulated by Jakobson, 1941, but see also Hyltenstam & Viberg, 1993) entails that the language which attrition is subject to, follows a particular pattern, namely a reversal of the order in which certain features were learnt. For this reason, the regression hypothesis is sometimes summed up as ‘last in, first out’. Thus, the features which were last acquired in a language and often the more complex features, tend to be lost first, whereas those easier features which are acquired early tend not to be affected so much. Whether it is indeed possible to predict such a process has been researched by scholars such as Ohlstain (1989). Hyltenstam & Viberg (1993) posit that a language may regress due to the replacement of said language with another which becomes the dominant language used for communication. This thus links the regression hypothesis with attrition as it is the second language which ensures a change (regression) in the first. Although many theorists have researched whether regression can indeed be applied to first language attrition, the results have been very mixed (see Keijzer, 2007 for an overview).
Another theory which could be seen as means of explaining L1 attrition is Dynamic Systems Theory. De Bot & Makoni (2005: 5) define a dynamic system as “a system of interacting variables that is constantly changing due to interaction with its environment and self reorganization”. This definition holds that it is not only those factors within the linguistic system which have an effect on how language changes, but also external factors such as the environment. This theory of an interaction of variables through networks can be applied to language attrition, as language attrition also assumes an interaction of external and internal factors such as age at emigration (as well as testing), motivation and education level [see 2.3 Predictors of L1 Attrition], but all in relation to the environment speakers now find themselves in.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 10 attrition as an interaction of many variables and in particular the L2 causing the reorganization of the L1, it is extremely difficult to test. Meara (2004) created a computer simulation of a simple network with 2,500 items, all items with a connection to two other words. He tested the consequences of the activation or deactivation of one word. By deactivating a differing numbers of words he was able to study the whole network and the outcomes, which were very varied. After many attrition effects (deactivations) were carried out the outcomes of the networks varied from clear effects of attrition to networks which, despite the deactivations that were applied, remained relatively stable. Although Meara’s study does not address the complexity of the natural lexicon, it does give insights into the reasons why L1 attrition provides so many different outcomes in attriters. It also makes it clear that the system and variables which affect L1 attrition are very complex and multiple.
The last theory to be addressed in this section is the linguistic theory of simplification. Simplification entails the loss of marked forms in preference for those unmarked forms which can be more regularly applied or constitute less complex rules than their marked counterparts. In terms of attrition, the L1 undergoes a simplification from marked to unmarked features and this therefore constitutes loss or attrition as the marked forms are no longer used or available to the speaker. Seliger (1989: 173) provided a definition of simplification which is applicable to L1 attrition in regards to the rules which have effect on the L1. “If both languages contain a rule which serves the same semantic function, that version of the rule which is formally less complex and has a wider linguistic distribution (i.e. can be used in a greater variety of linguistic environments) will replace the more complex more narrowly distributed rule. That is, in the case of L1 attrition, simpler, more widely distributed rules from L2 will replace those from L1.”
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 11 the simplification theory, as the L1 was shown to be affected by the less complex and more linguistically distributable rules of the L2. Although this theory can be applied to grammatical attrition it cannot be applied to other domains such as lexicon, where word families and lexical fields are not necessarily governed by the clear rules. In sum, simplification postulates attrition to be a reduction of the more complex mother tongue that was before, under the influence of extending L2 categories to the L1.
2.3 Predictors of L1 attrition
Over the years one of the primary aims of L1 attrition research has been to uncover the factors that cause attrition in the first place. One of the most defining characteristics of L1 attrition is that it is caused by a break from the L1 environment and a subsequent immersion in an L2 environment. And yet, which of these two is the primary cause of attrition is still a matter of debate, as the discussion above illustrates. In addition, past investigations have tried to uncover predictor variables that are at work in shaping L1 attrition.
2.3.1 The age of L1 attrition onset
The age of onset of L1 attrition and thus reduced contact with the L1 appears to exert great influence on how much of the L1 is maintained and the ultimate attainment level of the L2. There is a distinction between children and adults with regards to the effect of L1 attrition, which can be related to the critical period hypothesis (Lenneberg, 1967) [see 2.1 What is L1 attrition?]. Children seem to be more flexible when it comes to language, which has its advantages as the L2 is often learnt to full proficiency. At the same time, however, this can also have disadvantages as due to the flexibility regarding language in children, the L1 can be completely forgotten (Pallier et al. 2003).
Although intuitively obvious, only a limited number of studies have systematically varied the effects of age in both post-puberty and pre-puberty attriters (Schmid, 2011). There is, however, evidence regarding the effects of attrition on the L1 for both groups. The age of onset of attrition ultimately determines the degree to which the L1 will experience attrition.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 12 relatively young (3- 8 years old) when they became immersed in a second language environment, with the result that the L1 seemingly underwent complete attrition and even recognition of the phonological system appeared to have become inaccessible in later testing. This contrasts with the Köpke (1999) study where the native German participant in question had emigrated to an English speaking environment later in life but still had access to his L1 which suffered relatively little attrition in comparison. As touched upon before, the factor of age of emigration heavily ties in with the critical period hypothesis, which assumes that the ability to learn a second language shows a gradual decline after puberty, which DeKeyser & Larson-Hall (2005) conclude is due to a change in learning capacities and the brain becoming less flexible through the years. Similarly, L1 attrition experiences a decline in effect, which has also been posited to be around the age of puberty, to be precise, around the age of 12 (Schmid, 2011; Bylund (2009)). The study by Bylund (2008) provides evidence for this claim. In this study, participants (all L1 Spanish speakers were immersed in an L2 Swedish setting). Participants differed in their ‘ages of reduced contact’ (ARC) with the L1 and it was discerned through testing that those participants who had an ARC of less than 12 diverged from monolingual patterns whereas all participants with an ARC of over12 converged with monolingual Spanish behaviour (Bylund, 2009). The reason for such effects of age has been claimed by Köpke (2004) to be due to the stabilisation of linguistic knowledge as age increases. Linguistic knowledge, therefore, does not become stable until the age of 12, which is when it becomes less susceptible to L1 attrition. In the stages leading up to 12, the L1 may become fully attrited and replaced by the L2.
In conclusion, age has a significant effect on the processes of attrition: those below the age of 12 who experience reduced contact with the L1 may experience complete attrition of the L1, whereas those with an age above 12 already have a stable linguistic system and therefore do not experience such drastic effects. It should be stressed once more that it is uncertain whether the loss of the first language below the age of 12 can be seen as a consequence of L1 attrition or incomplete acquisition and this issue remains delicate.
2.3.2 Educational Level
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 13 level of education (Köpke & Schmid, 2004). It is therefore wise to control for this variable when testing for language attrition as comparing those who have a higher level of education and those with a lower level may provide results which are misleading as they are due to the level of education and not due to the effect of language attrition per se.
Previous studies such as Jaspaert & Kroon (1989) and Yağmur (1997) have controlled for education levels in their research. Jaspaert & Kroon tested 30 Italian immigrants in The Netherlands. The outcome of these tests suggested that where attrition was present, education was the strongest significant factor. In Yağmur’s (1997) study, education was treated as one of the main independent variables and what was found was that performance in verbal fluency was affected in attriters and non-attriters. As would be expected, those with a lower level of education performed worse than those with a higher level. Köpke (1999) also found that those with a higher level of education were able to produce target structures more often than those with a lower level of education. Schmid & Dusseldorp (2010) found education to have an impact on tasks which involve free speech and cloze tests. Those speakers who are highly educated outperform those with a lower level of education on both of these tasks.
Although it is true that those with a higher level of education outperform those with a lower level on attrition tasks, the way in which education is measured in such studies differs. Although one may have completed a certain level of education, this does not mean that the person is currently at that level. Köpke & Schmid (2004) mention the study by Pelc (2001) as an example of how different criteria are used to measure education. Pelc (2001) in his study measured education as the number of years of L1 education before emigration and the number of years of education in the L2 environment. The outcome of this study provided a correlation with the number of years of L1 education and the linguistic performance of participants (Köpke & Schmid, 2004).
Despite the seemingly abundant evidence that education plays a role in attrition, it is important to remember that education as a factor may also interact with other factors such as age.
2.3.3 Gender
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 14 Schmid & Dusseldorp (2010). The results of Köpke’s (1999) study showed no effect of sex on any of the tasks. In the study by Schmid & Dusseldorp the effect of sex was only significant for errors, as “female participants had fewer errors in the free speech task than males” (Schmid & Dusseldorp, 2010: 139).
2.3.4 Length of Residence
The length of residence (or LOR for short) has an effect on the amount of L1 attrition experienced (Schmid, 2009).There are unfortunately few studies that address attrition and the amount of erosion undergone in a short period of residence in direct comparison to those which investigate L1 attrition after 8-10 years. The study by Laufer (2003) investigated the knowledge of collocations in L1 Russian speakers who had immigrated to Israel and thus had acquired Hebrew as their L2. These participants were contrasted with a Russian monolingual control group. When immigrants in this study were divided into two groups according to their length of residence in Israel: 2-6 years or 17-40 years, a significant difference between the groups was found. This revealed the largest amount of attrition in collocational knowledge after a long period of residence in the L2 country. “[T]he longer the residence, the less uniform people become in their L1 knowledge” (Laufer, 2003: 25). This study thus provides evidence which suggests that the longer a period of residence is the more L1 attrition will be undergone. Studies such as Chang (2012) contradict this, as they show results of L1 attrition after a very short period of time. Similar findings have been revealed in Linck et al. (see– [2.2.1. Psycholinguistic Theories] thus it is not the case that L1 attrition effects are not measurable after short periods of time; however, greater effects can be seen after a longer period of residence.
2.3.5 Motivational Factors
Motivation has been found to play a major role in second language achievement, which can reflect the sociopolitical, economic and cultural circumstances of the particular subjects in question (Paradis, 2007). Inasmuch as motivation can be a factor for success or failure in second language learning, it can also have an effect on the amount and rate of attrition a first language can undergo (Paradis, 2007).
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 15 threshold (Paradis, 1993). The opposite effect can also found, namely a positive attitude towards the L1 can lower the threshold (Paradis, 1993) and thus will lessen the effect of L1 attrition.
The ability to express oneself in another language may also provide motivation for the usage of that particular language. The study by Kim & Starks (2008) shows attrition in Korean- English bilinguals who have learnt New Zealand English in adolescence. These participants displayed a preference for the use of English when expressing emotions such as rage. The reason for such a preference is posited to be a consequence of learning the L2 English in adolescence when they have to “deal with their emotions without the support of the L1” (Kim & Starks, 2008: 316). This then leads to a change in attitude concerning their L1 as they may experience a change in favour of the conceptual representations of emotions in the L2 (Pavlenko, 2002). This could then be the cause or a factor towards L1 attrition, or in any case have an impact on the rate of attrition.
It is possible to conclude from this section that motivation indeed plays a role in the attrition or maintenance of the L1, and studies would do good to take this into account in a more controlled manner
2.4 Which domains are most affected by attrition?
The studies investigating the domains affected by language attrition are abundant. Previous studies have clearly shown that language attrition can be highly selective in the areas it affects and the degrees to which these areas are affected (Gürel, 2004; Scherag et al., 2004; Schmid, 2002, 2009; Goral, 2004; Kasparian et al., 2014; Seliger, 1991).
2.4.1 Lexicon and Semantics
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 16 (fault) in an ERP investigation. In carrying out this experiment they wished to gain insight into neurocognitive mechanisms underlying L2 processing and whether these are maturationally constrained. Participants were given sentence judgment tasks in which they had to deem a sentence correct or incorrect whilst certain words in the sentence had been swapped for similar words (minimal pairs) or outright mismatch violations in which a noun would be replaced with an unrelated noun. Results in this ERP study revealed that those participants who had emigrated did not automatically identify when a minimal pair had been switched in a sentence context but did identify when an unrelated word was used. This study therefore gives evidence to the neurological basis of attrition. Lexical attrition is, of course, difficult to test when based solely on free speech and other tasks without brain imaging evidence, as any word a participant cannot retrieve in the mental lexicon may not be due to its non-existence but may be due to retrieval problems. However, “if you cannot retrieve a word in [a]very short space of time, it is effectively gone, and communication becomes impeded” (Schmid, 2011: 45). Schmid & Beers Fagersten (2010) analyse results of L1 German Jews who immigrated to Anglophone and Dutch countries and posit that disfluency markers found in the data, such as empty and filled pauses and other hesitations may be due to the inaccessibility of lexical representation and also to the shift towards the L2 norm regarding filled and empty pauses.
Another way of measuring lexical attrition is through frequency profiling: Schmid (2011) addresses the data she collected in Schmid (2002) and illustrates how attriters are more likely to use high-frequency words. Words with low frequencies are vulnerable to attrition, explained by the activation threshold (Paradis, 1993, 2001, see also [2.2 Why does it occur?]). Although the results of Schmid’s (2002) data thus do point in the direction of lexical attrition, Schmid (2009) reports that in comparison with the control group, more errors were made by the attriters, and by and large “speech samples appeared largely unaffected by the attritional process” (Schmid, 2009: 40). This ratifies hypotheses that language attrition is selective and consequently may also be selective in individuals.
Whilst the lexicon generally reveals effects of attrition in studies to date, the domain of semantics provides results which contradict each other.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 17 When faced with their answers they reacted with disbelief (Schmid & Köpke, 2009). This result provokes the awareness that the L2 can influence the L1 to the extent that L1 attrition takes place in the semantic domain and produces speakers who diverge from their L1 norms with uncertainty about their own correctness. Whilst this study provides evidence for semantic attrition, Scherag et al.’s (2004) study gives a different result [see 2.4.4. Morpho-syntax]. The study by Scherag et al. (2004) provides evidence illustrating that the semantic domain is not affected in attrition as, within the experiment they carried out, semantic processing was unaffected. They speculate that the reason for the unaffected semantic-associative processing may lie in the fact that this processing relies on “abstract conceptual representations that, at least partly, are not linked to a particular language or language at all” (Scherag et al. 2004: 107).
The domain of semantics provides a dilemma in determining whether it is a domain which always suffers the effects of L1 attrition, as case studies provide contradictory results. The semantic domain is, however, intrinsically tied to the lexicon, which as mentioned does undergo L1 attrition. It can therefore be concluded that whilst some studies reveal no change in the semantic domain after a period of non-use, if thought to be inherently related to lexicon, the semantic domain it should indeed be affected by L1 attrition, as evidenced by the previously mentioned study by Altenberg (1991) and those in relation to it in the category of the lexicon.
2.4.2 Phonology
L1 language attrition has also been quite abundantly attested in the domain of phonology. Chang (2012) reports the results of his study, which produced an interesting finding regarding the influence of the L2 on the L1 in participants who were novice learners of their L2: native speakers of American English who were taking a six week intensive course in Korean at a South Korean University.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 18 would have been the case for a native speaker. Assimilation may also occur when a vowel category in the L1 already exists and can be used in the L2, even if the categories in both languages are slightly different. With the assimilation and dissimilation to the L1 or L2 and addition of vowel categories to maintain distance, a shift can be witnessed that affects the whole system as the vowel space becomes more crowded and distance is created in other to maintain the distinction between vowel categories. This shift is also applicable in cases of assimilation where the L1 is modified to become more like the L2 (Chang, 2012).
Participants in Chang’s (2012) study reported mostly to function in English outside of their Korean classroom. The Participants were tested in Korean and English production, using words which were maximally similar to each other in pronunciation in order for cross-language acoustic comparison to take place. The results of this study show that the participants underwent phonetic drift in their L1, particularly in English stop consonants and vowels, restructuring the vowel space to assimilate to the Korean vowel space. Chang (2012: 265) concluded that while “phonetic drift may provide the basic mechanism of phonological attrition, it does not fully determine the occurrence of attrition”. This is to say that, whilst phonetic drift may be indicative of attrition or quite possibly the step towards it, it does not entail it.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 19 plosive VOT values of MZ, as she produced longer VOT values, which were in fact in between Dutch and English norms. Whilst the voiceless VOT values were affected, the voiced plosive values showed no signs of L1 language attrition as they were in line with the Dutch norm. They did, however, become pre-voiced in English.
The second experiment reported in Mayr et al.’s study investigated the vowel system in English and Dutch in the twins, as the vowel space is differently divided in both of these languages. More specifically, standard Dutch distinguishes between phonologically short and long vowels, closing diphthongs and schwa, whereas English has short vowels, long vowels, closing diphthongs, centring diphthongs and schwa (Mayr et al. 2012). These vowels are inventoried in the tables below, one for Dutch and one for English respectively.
(Mees & Collins, 1983:64)
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 20 (Roach, 2004: 243).
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 21 /aʊ/ respectively” (Mayr et al. 2012: 696). MZ did, however, distinguish between remaining diphthongs. Again categories with no corresponding form in English still showed a shift towards a more open position. From these results it can be suggested that cross-linguistic influence affects a shift in the vowel system and does not only affect individual sounds. MZ’s vowel system does not, however, represent a full shifted L1; it rather shows a unique composition, suggesting “complex interaction between token-level and system-level shifts in bilingual systems” (Mayr et al. 2012: 698).
In summary, the result of Mayr et al.’s study showed consistent use of longer VOT values by the emigrated sister and indeed L1 language attrition in her vowel system, which had shifted towards the L2 English, therefore allotting this sister, who lives in England, with an accented L1. However, it is important to stress that not all areas or sounds were affected, as voiced plosives were produced within the Dutch norm and Dutch /a/ was not affected. Looking at two siblings with identical vocal tracts and otherwise similar experiences, this study thus provides compelling evidence to the effect L1 language attrition can have in the domain of phonology.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 22 that some participants were clearly and confidently rated as native speakers. 14 consecutive bilinguals in this study were rated as having a foreign accent in their native language. The result that a number of L1 speakers of German were assessed by monolingual German speakers as possessing a foreign accent in the L1 is in line with the earlier discussed phonological attrition research (Mayr et al., 2012; Change, 2012), thus providing further evidence that the L2 can have a significant effect on the L1 phonological system, compelling it to undergo a restructuring, which ultimately makes the L1 accented.
2.4.3 Syntax and Grammar
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 23 in particular in the overt pronoun o. L1 attriters did, however, manage to maintain referential properties of the overt pronominal and null pronoun. These examples serve as evidence that the domain of syntax can indeed be affected by L1 attrition. Contradicting these studies is the example of Köpke’s 1999 study, illustrated in Köpke (2001). In this example a native German speaker who had been living in Canada for 39 years was interviewed. This speaker asserted that he only spoke English as he felt more comfortable using this language in any situation and that he never went back to Germany. He was tested using picture description tasks, sentence generation and grammaticality judgement. The results from these tasks revealed lexical retrieval difficulties and problems in the correct use of prepositions. However, they showed hardly any errors in grammar on any of the tasks and an above average success rate in grammaticality judgments compared to the other speakers tested in the experiment. This is surprising, indicating that although this subject had not used German for a substantial amount of time, the underlying grammatical structures were still available for use and thus were still present in mental structures. This is an exceptional case when compared to the other test subjects. This thus shows that attrition is extremely variable within groups and that the results of one test subject cannot be generalised to a whole group. One exception, however, does make one question the rule. Whilst many participants may undergo slight attrition in the syntactic domain, the evidence provided by exceptions makes this notion also not generalisable to the whole.
Lastly, the study carried about by Keijzer (2007) provides evidence for attrition in the syntactical domain. Keijzer (2007) used a grammaticality judgement test, looking at different aspects of syntactical features such as negation, passive constructions, verb second, subordination and discontinuous word order in the L1 Dutch of 45 Dutch immigrants in Anglophone Canada. This test yielded significant results in that her 45 Dutch Canadians produced significantly slower scores on the grammaticality judgment task that a group of 45 matched control subjects in the Netherlands, with attriters and learners making more mistakes than controls. These results are then explained by Keijzer through education level [see 2.3.2. Educational level], as this was the only factor which had predictive power over the results. The subjects with a higher level of education consistently outperformed those who were less well educated. This study gives evidence to the effect of language attrition in the syntactic domain.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 24 attrition. The domain of syntax is less likely to be affected to the same extent as the lexical domain but attrition effects are most definitely also felt in the syntactic domain, be it with time.
2.4.4 Morpho-syntax
The domain of morpho-syntax has also been taken up in L1 attrition research. Scherag et al.’s (2004) study demonstrates the lack of effects of attrition in this domain in German immigrants to the U.S. The study focuses on participant groups of native German immigrants to the U.S., and compares these to Native German speakers and Native English speakers who were immigrants to Germany. Groups were tested using an auditory semantic and morpho-syntactic priming paradigm where German adjectives, which were used as primes for the target noun, were inflected for gender, either incorrectly or correctly. These adjectives were also either semantically associated or not associated to the target noun. The categories that thus arose from these distinctions were: (sem+/syn+); (sem+/syn-); (sem-/syn+); (sem-/syn-). For example, ‘faltiges (wrinkled)’ would be the prime for the target nouns, with the previously mentioned conditions respectively: ‘gesicht (face)’, ‘haut (skin)’, ‘gerücht (rumour)’ and ‘lohn (reward)’. Results from this experiment show that all native speakers were primed in sem+/syn+ and had the highest error rates in sem-/syn-. Native English speakers had the higher error rates in the sem-/syn+ condition. As all German participant groups showed processing gains in the semantic and morpho-syntactic conditions, no attrition could be evidenced. Scherag et al. (2004) posit that this result may be due to underlying structures that once acquired are not effected by non-use in an extended period of time.
From this study in the domains of syntax it can be concluded that morpho-syntax in general, is a domain which is less susceptible to L1 attrition, although not entirely free of any attrition effects either.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 25 of L1 attrition and does so within the shortest time period, quite possibly followed by the domain of phonology
2.5 What is the time course of L1 attrition?
The time course of attrition is difficult to determine as many individual factors play a role. The majority of studies in the field are cross-sectional group studies, which focus on the amount of attrition which has taken place after generally ten to fifteen years, as this gives relatively stable effects which are not only due to the active development of the L2 (Schmid, 2011). Not only does Schmid (2011) posit this as a good time frame after which to research attrition effects, De Bot, Gossmans & Rossing, (1991) also report this as a generally accepted baseline. Olshtain & Barzilay (1991) also report effects after eight years of emigration. Studies have, however, been done in shorter time frames after emigration, mostly in more recent years but the majority of studies to date set a minimum threshold stay of 10 years in the L2 environment (Keijzer, 2014). The previously illustrated Chang (2012) study provides results which suggest attrition effect can be witnessed after as short a time span as six weeks of immersion in Korea. This study thus contradicts the need to wait eight to ten years for measurable attrition effects in participants.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 26 the newly to be acquired L2. After L2 proficiency levels pick up, speakers can become more relaxed regarding their L1 and attrition effects may level off.
Whilst L1 attrition can occur in a short period of time, it is not possible to say when it stops or even if it reaches a point where the L1 no longer experiences attrition. A study by Hütz (2004) reviews letters by a native German speaker who had emigrated to the U.S. His analysis revealed an ongoing process of L1 attrition that produced noticeable effects after five years, but persisted even after fifty years.
The studies mentioned indicate that L1 attrition can occur in a short and over a long space of time. It is therefore difficult to define the time course of attrition with specific reference points. Longitudinal studies to date, in addition to being very scare compared to cross-sectional designs, do not test participants with the same variables and thus these studies cannot be compared with each other to reach such a mapped out time line of L1 attrition. It is the variables which interact with L1 attrition that can have substantial effects on the rate at which it influences the L1. Variables such as intrinsic and extrinsic motivation and similarities between the languages play a large role in how the L1 is affected (Köpke, 2005, 2007). Schmid (2002) also found that negative attitude was a significant predictor of L1 attrition.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 27
2.6 Attrition studies involving siblings
As can be attested from the previous discussion of past attrition work, attrition work can be categorized along different dimensions: longitudinal work vs. cross-sectional studies and group comparisons vs. more case-based analyses. Within the latter category, a subsection of studies focuses on siblings. An example is Kuhberg (1992) who studied two Turkish children (sisters) who moved to Germany and Ohlstain (1986) who studied 3 Israeli children who moved from Israel to the US and then back after 1-8 years. However, studies which compare siblings who are late acquirers of a second language, where one sibling immigrates whilst the other stays in the country where the mother tongue is spoken, are few and far between. But this does allow for a direct comparison and more insights into the precise mechanisms of attrition. This lack of work can be attributed to the unique category which these participants fall into. Participants who satisfy the requirements for such a comparison are unlikely to be recruited, especially if a study requires participants to be monozygotic twins or siblings, where only one sibling has immigrated and the other has stayed in the country where the mother tongue is spoken.
Another important factor to mention in terms of studies on siblings is the comparison between siblings which is often not addressed as participants in attrition studies may often be siblings but will not be matched and analysed in order for comparison between them, they are often analysed within the group.
There is, however, one study that specifically compares monozygotic twin regarding L1 language attrition; this is the Mayr et al. (2012) study. This study will be illustrated briefly in the following section. For the Mayr et al. (2012) study in detail, see [2.4.2 Phonology].
2.6.1 Attrition studies involving monozygotic twins
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 28 “two offspring born of the same pregnancy and developed from a single fertilized ovum that splits into equal halves during an early cleavage phase in embryonic development, giving rise to separate fetuses. Such twins are always of the same sex, have the same genetic constitution, possess identical blood groups, and closely resemble each other in physical, [psychological], and mental characteristics.”
Interesting in this definition is the mention of similarity in ‘psychological and mental characteristics’. This is supported by Stafford (1966) who posits that monozygotic twins possess the same level of aptitude and therefore would be expected to achieve the same scores on aptitude tests (including language aptitude). Any difference in such scores would thus be theorised as due to environmental factors and error effects.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 29
3. Research questions and predictions
This study examines the lexical and phonological attrition of one member of two monozygotic twin sisters who had had the same upbringing until 19 years of age, after which one of the twins moved to the Netherlands and learnt Dutch fluently as a second language. This twin (MRS) can therefore be classified as a late bilingual in an immersion context [see 4.2 Participants]. Lexical and phonological attrition form the focus of this study, as these are the two domains which have been identified as showing the earliest signs of attrition. This study thus closely follows the design laid out in Mayr et al., but with two crucial differences: 1) the length of residence of the sister who moved away is only very brief (2 years as opposed to the 30 years in Mayr et al.’s study), and 2) the age at testing by definition is considerably younger. Through exploring previous research the following research questions and predictions form the basis of the current study:
1) Can lexical and/or phonological attrition be attested in one of the twin sisters?
Here the prediction is that MRS will experience language attrition in her first language but many changes in the first language will also be due to cross-linguistic influence and not necessarily language attrition.
2) Which of the two domains under investigation is most affected by initial attrition?
Here the prediction is that the lexical domain will be the most affected, as this has been observed in other studies [see Lexicon 2.4.1]. As the time of testing is only relatively recent after the move, however, the prediction is that lexical effects are noticeable in production but not yet in reception. As such, the prediction is that the fluency of language in regards to the lexicon will be greater in the monolingual twin, as noticeable in free speech. In addition, effects are expected in production but not in comprehension tasks.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 31
4. Methodology
4.1 Participants
This study investigated two monozygotic twins, both mother tongue speakers of British English. (MRS and MSS). Both participants were born in London but raised in Boston, England and therefore speak the East Midlands variety of English. Participant MRS had moved to the Netherlands on an exchange year when she was 19.5 years old and in that year had gained a high proficiency in Dutch. She had then moved back to England for a year before returning to the Netherlands at 21,5 years of age. In total MRS’s amount of time spent in the Netherlands was calculated at 521 days at the time of testing, spread over 19 months. By contrast, MRS’s twin sister had not spent any substantial period in the Netherlands. MSS had, however, spent a year in Australia at 20 years of age which she reported had affected her accent. MRS and MSS had both had experience with Spanish as a school subject, but rated their level of proficiency as poor, a 1 on a scale of 1 to 5.
Both participants came from a higher social economic background. At the time of testing, MRS was completing her Master’s in European linguistics, whilst MSS was enrolled in a Bachelor degree program in equine science and equestrian psychology. Both twins (MSS and MRS) rated their level of English proficiency at 5 out of 5, although MRS reported some difficulties in word retrieval when speaking and slower processing times, supported by MSS’s views on MRS’s level of proficiency. MRS also reported a change in her grammar use; supported by the comment made by MSS that MRS used ‘Dutch grammar’ in written work. MSS also reported that MRS had a perceivable accent other than her native British accent, which she subsequently described as European. Listening capacity was not reported to have been affected by either of the monozygotic twins.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 32 members. MSS used only English on a day-to-day basis. The monozygotic twins were comparable in height and weight, both with very similar measurements. MRS measured 162 cm, whilst MSS measured 158 cm. MRS weighed 56kg, whereas MSS weighed 51kg. From these measurements it can be assumed that the vocal tract length in both twins is very similar.
4.2 Materials
This study consisted of two parts, each tapping into a different language domain. In the first half of the experiment, the focus was on lexical attrition. The tests included cloze procedures; letter and category fluency tasks; the Lognostics productive vocabulary test Lex 30 V4.0 (Fitzpatrick & Meara, 2008); the Lognostics receptive vocabulary test V_YesNo v1.0 (Meara & Miralpeix, 2014) and a film re-telling of Michael Dudok de Wit’s Father and Daughter (Roccotroco, 2009). In the second half of the experiment, tests were used to analyse features pertaining to the phonological domain, namely, a global accent perception test and a VOT analysis of the voiceless initial plosives used in the story re-telling task. Each of these tasks is briefly described below.
4.2.1 Lexical measures
Cloze test
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 33 (Oller & Streif, 1975); entailing that the deletion of part of a word increases the processing demands for non-native speakers, whilst leaving native speakers unaffected (Gradman & Spolsky, 1975: 59). Native speakers are thus expected to perform better on such tasks as they have the assumed ability to expect or predict the following grammatical or lexical items. The texts used in this thesis covered differing topics, from history to health [see Appendices A, B for the English and Dutch version respectively].
Verbal fluency task
The verbal fluency task is used to determine the approximate size of the lexicon in participants. This is done through asking subjects to productively generate as many exemplars of a given category as possible, typically within a set time limit (of usually 30 to 60 seconds). The score (number of exemplars named) is then taken to indicate the approximate vocabulary size of the participant.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 34 beginning with the letter A, although Proper nouns could not be named. In the letter fluency task the naming of other parts of speech, such as grammatical function words, was permissible. The following task was another category fluency task which involved the naming of fruit and vegetables. Lastly in the category and letter fluency tasks, was the letter fluency task, which elicited words beginning with B. Again these words could be any part of speech other than Proper nouns. A stopwatch app was used on a Samsung Galaxy S3 mobile phone in order to time one minute for each category (the set time limit for this task). An Olympus Digital Voice Recorder WS-200S was used in order to record the speech production of participant MRS and an Apple iPhone 5c voice recording app was used in order to record production from participant MSS.
Lognostics tasks
For the Lognostics tests, online tests, presented on a computer screen, were, used as these are standardised and therefore would not encounter any bias from the participants. The tests can be found at http://www.lognostics.co.uk/tools/index.htm. Specifically, in this experiment, the Lognostics productive vocabulary test Lex 30 V4.0 (Fitzpatrick & Meara, 2008) and the receptive vocabulary test V_YesNo v1.0 (Meara & Miralpeix, 2014) were used. The productive vocabulary test featured an online test in which participants had to produce 4 items related to each of the 30 items given in the test. For example, participants had to produce 4 items related to the word ‘potato’. No restrictions applied as to which words were coined, and could range from nouns to adjectives. When all fields were filled in and entered, the test gave a score which could be interpreted as the productive vocabulary level at which the participant was at. This test took around 15 minutes to complete. The score given also allowed for comparison to native speaker ranges, as were described in Meara (2009), where a score of 60 out of the available 120 point was given as the threshold for a native speaker.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 35 was compared to the thresholds regarding proficiency in the manual (Meara & Miralpeix, n.d.), which stated that a ‘good’ non-native speaker who could be recognised as native speaker would achieve a score of 6,000 to 10,000, 10,000 being the highest score possible.
Story Re-telling task
To elicit free speech, the subjects were asked to watch a short animated film: Michael Dudok de Wit’s Father and Daughter and then retell the main events of this film. Film retellings have frequently been used in past linguistic investigation because they offer a controlled way of recording free speech (cf. Perdue, 1993). The re-telling of Michael Dudok Wit’s Father and Daughter required the use of a computer in order to view the film and an Olympus Digital Voice Recorder WS-200S to record the re-telling. The Olympus Digital Voice Recorder WS-200S was used with participant MRS and an Apple iPhone 4 voice recording app was used in order to record the re-telling test completed by participant MSS. This test comprised the watching of the video, which did not contain any speech, and the subject subsequently being asked to retell the story in as much detail possible. Participants were not given a set time to re-tell the story, thus eliciting free speech which was more natural than the singular items elicited in the letter and category fluency tests. As the video also did not contain speech, participants were free to describe the events in the video as they liked, thus not being influenced by priming effects. This re-telling was recorded and used for subsequent analyses. The recorded sound files were used in order to analyse the complexity of the speech and vocabulary used through MLU and type/token ratio analyses. The sound files were also used in the Global perception task, where near native speakers of English were asked to judge the participants MRS and MSS on their ‘nativeness’. The story re-telling task sound files were also used in VOT analyses in order to reveal underlying phonological differences in the participants’ speech.
4.2.2. Phonological measures
Global Accent Perception Rating Task
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 36 performed. A similar experiment in the context of L1 attrition has been carried out by de Leeuw et al. (2010). In order to carry out this task, 13 participants (11 near natives and 2 native speakers of English) were asked to listen to both recordings of the story re-telling, which had been renamed file A and B, without any given previous background information. Participants where then asked to fill in a questionnaire [Appendix I.] and rate the speakers A and B in terms of nativeness and give information as to why they had come to their conclusions. Participants were encouraged to give linguistic reasons and factors as to why a speaker would be considered native or not. Participants were also asked to rate the certainty of their judgements on a scale of 1 to 5, 1 being extremely uncertain and 5 being extremely certain. Sound file A was the recording of the re-telling by MRS and sound file B was the recording of the re-telling by MSS.
VOT Measures Task
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 37 voiceless plosive had thus ended. The VOT of each initial voiceless plosive was inserted, along with the corresponding word, into an Excel document in order to keep track of the VOTs and to later allow for statistical analyses. Once all of the VOTs had been inserted separately for MRS and MSS into two separate Excel documents, the means and standard deviations for /p/, /t/ and /k/ were calculated.
4.3 Procedure
In the first part of the experiment, where participants were recorded, tests were implemented in a set order for both monozygotic twin sisters in order to avoid any bias that could arise from testing in other orders. For MRS specifically, all tests were first administered in English and then in Dutch to avoid priming effects.
Firstly participants MRS and MSS completed the cloze task, before moving on to the verbal fluency tasks, which were timed (1 minute) on a stopwatch app used on a Samsung Galaxy S3 mobile phone for each category. Speech was recorded in these tasks by an Olympus Digital Voice Recorder WS-200S for the speech production of participant MRS and an Apple iPhone 5c voice recording app for participant MSS. After these tasks, the Lognostics tasks were completed online and the scores were noted down. Following the Lognostics tasks, the story re-telling task was completed and recorded using a Olympus Digital Voice Recorder WS-200S for participant MRS and an Apple iPhone 4 voice recording app for participant MSS.
4.4 Analysis
This section describes the method of scoring the various tasks that formed part of the test battery as well as the analyses that were subsequently performed. It should be emphasized that this is a case study rather than a group design, by definition disallowing the use of inferential statistics. Instead, the data are examined using visual inspection.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 38 essentially formed the second half of words) was judged according to grammatical, semantic and orthographic correctness. This then resulted in scores for each of the five cloze tests, leading to a total possible score of 100. This was then taken to reflect the overall language proficiency of both MRS and MSS in English. This procedure was also followed for MRS’s Dutch C-test. In the letter fluency and category fluency tasks the data were analysed by transcribing the recordings [for transcripts see Appendix A.] which were made and examining the number of words produced in each task, after elimination of possible duplicates and proper nouns.
From the Lognostics productive vocabulary test Lex 30 V4.0 (Fitzpatrick & Meara, 2008) and the receptive vocabulary test V_YesNo v1.0 (Meara & Miralpeix, 2014), scores were automatically generated. These scores were then compared against the scores showing levels of ‘native-ness’ in the respective guide manuals available through the online tests. The levels presented in the guide manual differentiate levels of proficiency typically pertaining to non-native speakers based on the scores attained. The highest threshold in this test indicates a good level of proficiency. In the Lognostics productive vocabulary test Lex 30 V4.0 (Fitzpatrick & Meara, 2008); the threshold for the recognition of native-speaker-like productive vocabulary according to the guide for interpretation of the results (Meara, 2009) is a score of 60 or higher from the available 120 points. In the receptive vocabulary test V_YesNo v1.0 (Meara & Miralpeix, 2014) the manual (Meara & Miralpeix, n.d.) states the threshold for a ‘good non-native speaker or recognition as a non-native-speaker is 6,000 to 10,000, with 10,000 being the highest score obtainable. Other thresholds which are stated are: 3,500 to 6,000 for intermediate level learners, 2,000 to 3,500 for competent beginners and scores below this being unreliable.
The film-retelling recordings were transcribed according to Jeffersonian transcription conventions (Jefferson, 2004) [for transcripts see Appendix C, D.]. From the film re-telling transcriptions it was possible to calculate the mean length of the utterances (MLU) and the type/token ratio [see appendix E, F.]. The difference between the outcomes of the type/ token ratios in participants MRS and MSS was analysed by calculating a z ratio. The use of differing grammatical structures such as SVO and SOV was also examined.
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 39 The mean certainty of the judgement as to whether MRS (sound file A) was non-native was also analysed. The same was done for MSS (sound file B).
Mina Solanki S2392364 M.r.solanki@student.rug.nl Page 40
5. Results
Lexical tasks C-test
The cloze task (C-test) yielded scores out of 100, with each text having a maximum score of 20. Table 1. Shows the scores obtained by MSS and MRS, in English.
Table 1. C-test English (General Language Proficiency.)
Text 1 Text 2 Text 3 Text 4 Text 5 Total
MRS 17 13 17 18 13 78
MSS 16 18 17 16 16 83
As can be seen MRS and MSS both produced highly similar results with MSS achieving a score of 83 and MRS achieving a score of 78. In other words, both subjects performed at the higher end of the scale. MRS also performed the test in Dutch, as shown by Table 2.
Table 2. C-test Dutch.
Text 1 Text 2 Text 3 Text 4 Text 5 Total
MRS 19 17 15 14 9 74
Interestingly, MRS attained a score of 74, very much similar to her score when the test was carried out in English. This indicates a balanced level of overall proficiency in both L1 and L2, which is striking as the L2 had only been acquired less than two years ago.
Verbal fluency task
