Language attrition

Language attrition is the loss of a first or second language or a portion of that language. Speakers who routinely speak more than one language may use their languages in ways slightly different from a single language speaker, or a monolingual. The knowledge of one language may interfere with the correct production or understanding of another. The study of these interference phenomena is the field of Applied Linguistics. Interference can work two ways. A person who acquires a second language (L2) after the first (L1), may be inhibited in the acquisition of this second language by the first language. However, interference can also work the other way: the second language can interfere with the correct use of the first. More recently, research has started to investigate linguistic traffic containing L2 interferences and contact phenomena evident in the L1. Such phenomena are probably experienced to some extent by all bilinguals. They are, however, most evident among speakers for whom a language other than the L1 has started to play an important, if not dominant, role in everyday life (Schmid and Köpke, 2007). This is the case for migrants who move to a country where a language is spoken which, for them, is a second or foreign language. The L1 change and L2 interference that can be observed in such situations is considered language attrition.

Academic study origins

Only in the past few decades has the study of language attrition become a sub-field of linguistics beginning with a 1980 conference at the University of Pennsylvania titled Loss of Language Skills ( Lambert and Freed, 1982 ). The aim of this conference was to discuss areas of second language (L2) attrition and to ideate on possible areas of future research in L2 loss. The conference revealed that attrition is a wide topic covering different types of language loss and that there are many possible reasons for the loss. A related phenomenon is the loss of language due to contact with other, more dominant languages, possibly leading to language death.

The field gained new momentum with two conferences held in Amsterdam in 2002 and 2005, as well as a series of graduate workshops and panels at international conferences such as the International Symposium on Bilingualism (2007, 2009), the annual conferences of the European Second Language Association and the AILA W0rld Congress (2008). The outcome of some of these meetings have been published in edited volumes (Schmid et al. 2004; Köpke et al. 2007) and special issues of journals, such as the ' 'Journal of Neurolinguistics (Vol. 17:1, 2004) , the ' 'International Journal of Bilingualism (Vol. 8:3, 2004) and ' 'Bilingualism: Language and Cognition (Vol. 13:1, 2010) .

First Language Attrition (FLA)

The term 'First Language Attrition' (FLA) refers to the gradual decline in native language proficiency among migrants. As a speaker uses their L2 frequently and becomes proficient (or even dominant) in it, some aspects of the L1 can deteriorate or become subject to L2 influence.

L1 attrition is a process which is governed by two factors: the presence and development of the L2 system on the one hand, and the diminished exposure to and use of the L1 on the other (Schmid & Köpke, 2007); that is, it is a process typically witnessed among migrants who use the later-learned environmental language in daily life. The current consensus is that attrition manifests itself first and most noticeably in the vocabulary knowledge of speakers (in their lexical access and their mental lexicon) (e.g. Ammerlaan, 1996 ; Schmid & Köpke, 2008) while grammatical and phonological representations appear more stable among speakers for whom emigration took place after puberty (Schmid, 2009) .

Attrition research has often wrestled with the problem of how to establish the border between the ‘normal’ influence of the L2 on the L1, which all bilinguals probably experience to some degree (as is suggested by, among others, Cook 2003 ), and the (consequently to some degree ‘abnormal’) process of L1 attrition, which is confined to migrants. It has recently been suggested that this distinction is not only impossible to draw, but also unhelpful, as “bilinguals may not have one ‘normal’ language (in which they are indistinguishable from monolinguals [...]) and one ‘deviant’ one (in which knowledge is less extensive than that of monolinguals, and also tainted by interference from L1 in SLA and from L2 in attrition)” (Schmid & Köpke 2007:3) . Rather, while L1 attrition may be the most clearly pronounced end of the entire spectrum of multicompetence, and therefore a more satisfying object of investigation than the L1 system of a beginning L2 learner (which may not show substantial and noticeable signs of change), attrition is undoubtedly part of this continuum, and not a discrete and unique state of development.

Like second language acquisition (SLA), FLA is mediated by a number of external factors, such as exposure and use (e.g. Hulsen 2000 ; Schmid, 2007 ; Schmid & Dusseldorp, 2010 ), attitude and motivation ( Ben-Rafael & Schmid, 2007 ; Schmid, 2002 ) or aptitude ( Bylund, 2008 ). However, the overall impact of these factors is far less strongly pronounced than what has been found in SLA.

L1 attriters, like L2 learners, may use language differently from native speakers. In particular, they can have variability on certain rules which native speakers apply deterministically ( Sorace, 2005 ; Tsimpli et al., 2004 ). In the context of attrition, however, there is strong evidence that this optionality is not indicative of any underlying representational deficits: the same individuals do not appear to encounter recurring problems with the same kinds of grammatical phenomena in different speech situations or on different tasks ( Schmid, 2009 ). This suggests that problems of L1 attriters are due to momentary conflicts between the two linguistic systems and not indicative of a structural change to underlying linguistic knowledge (that is, to an emerging representational deficit of any kind).

This assumption is in line with a range of investigations of L1 attrition which argue that this process may affect interface phenomena (e.g. the distribution of overt and null subjects in pro-drop languages) but will not touch the narrow syntax (e.g. Tsimpli et al., 2004 ; Montrul, 2004 ; Montrul, 2008 ).

Manifestations of language attrition

L1

Lexical L1 attrition

It has often been pointed out that L1 attrition usually first manifests itself in the lexicon (Schmid & Köpke, 2008). It is possible for lexical representations in the L1 to be influenced by the semantic potential of corresponding items in the L2. Instances of such interlanguage effects are reported by e.g. Pavlenko (2003, 2004), who concludes that for her L1 Russian speakers, a number of Russian terms appear to have gained a different meaning by semantic extension from their L2, English. Secondly, it has been noted that among attriters, lexical access can become impaired, resulting in poorer performance on picture naming tasks ( Ammerlaan, 1996 ; Hulsen, 2000 ; Montrul, 2008 ) and reduced lexical diversity in free speech ( Schmid, 2002 ).

Grammatical L1 attrition

Generative approaches to L1 attrition often focus on the possibility that the developing linguistic system may show evidence of irrevocable structural changes to the actual grammar of a native language. This was highlighted early on in the history of attrition research: “It is crucial to know whether a given example of language loss can be attributed to a change in how the relevant language is represented in the mind of the user or to a change in the way stable knowledge (competence) is being used.” ( Sharwood Smith 1983:49, his emphasis). In a similar vein, Seliger and Vago (1991) define the object of investigation as “the disintegration or attrition of the structure of a first language (L1) in contact situations with a second language (L2)” ( Seliger and Vago 1991:3 ). However, most research appears to indicate that attrition does not affect uninterpretable features, but that variability may be observed in features that are interpretable at the interface levels ( Tsimpli et al., 2004 :274; Tsimpli, 2007 : 85). There therefore seems to be little evidence for an actual restructuring of the language system: the narrow syntax remains unaffected, and the observed variability may be ascribed to the cognitive demands of bilingual processing.

L2

Lambert and Moore (1986) attempted to define numerous hypotheses regarding the nature of language loss, crossed with various aspects of language. They envisioned a test to be given to American State Department employees that would include four linguistic categories (syntax, morphology, lexicon, and phonology) and three skill areas (reading, listening, and speaking). A translation component would feature on a sub-section of each skill area tested. The test was to include linguistic features that are the most difficult, according to teachers, for students to master. Such a test may confound testing what was not acquired with what was lost. Lambert, in personal communication with Köpke and Schmid (2004), described the results as 'not substantial enough to help much in the development of the new field of language skill attrition'.

The use of translation tests to study language loss is inappropriate for a number of reasons: it is questionable what such tests measure; too much variation; the difference between attriters and bilinguals is complex; activating two languages at once may cause interference. Yoshitomi (1992) attempted to define a model of language attrition that was related to neurological and psychological aspects of language learning and unlearning. She discussed four possible hypotheses and five key aspects related to acquisition and attrition. The hypotheses are:

According to Yoshitomi (1992), the five key aspects related to attrition are: neuroplasticity, consolidation, permastore/savings, decreased accessibility, and receptive vs. productive abilities.

The regression hypothesis (last thing learned, first thing lost)

The regression hypothesis, first formulated by Roman Jakobson in 1941, goes back to the beginnings of psychology and psychoanalysis. Generally speaking, it states that that which was learned first will be retained last, both in 'normal' processes of forgetting and in pathological conditions such as aphasia or dementia. As a template for language forgetting, the regression hypothesis has long seemed an attractive paradigm. However, as Keijzer (2007) points out, regression is not in itself a theoretical or explanatory framework. Both order of acquisition and order of attrition need to be put into the larger context of linguistic theory in order to gain explanatory adequacy.

L1

Keijzer (2007) conducts a study on the L1 attrition of Dutch in anglophone Canada. Her study compares language acquisition in children with non-pathological language attrition found in emigrant populations and contrasts it with language use in non-attrited, mature speakers. In particular, it examines the parallels and divergences between advanced stages of L1 Dutch acquisition (in adolescents) and the L1 attrition of Dutch émigrés in Anglophone Canada, as opposed to control subjects. She finds some evidence that later-learned rules, for example with respect to diminutive and plural formation, do indeed erode before the earlier learned information. However, there is also considerable interaction between the first and second language, so a straightforward 'regression pattern' cannot be observed.

L2

Citing the studies on the regression hypothesis that have been done, Yukawa (1998) says that the results have been contradictory. It is possible that attrition is a case-by-case situation depending on a number of variables (age, proficiency, literacy, the similarities between the L1 and L2, and whether the L1 or the L2 is attriting). The threshold hypothesis states that there may be a level of proficiency that once attained, enables the attriting language to remain stable.

The age effect

L1 attrition

While attriters are reliably outperformed by native speakers on a range of tasks measuring overall proficiency there is an astonishingly small range of variability and low incidence of non-targetlike use in data even from speakers who claim not to have used their L1 for many decades (in some cases upwards of 60 years, e.g. de Bot & Clyne 1994, Schmid 2002 ), provided they emigrated after puberty: the most strongly attrited speakers still tend to compare favourably to very advanced L2 learners (Schmid, 2009). If, on the other hand, environmental exposure to the L1 ceases before puberty, the L1 system can deteriorate radically (Schmid, 2012).

There are few principled and systematic investigations of FLA specifically investigating the impact of AoA. However, converging evidence suggests an age effect on FLA which is much stronger and more clearly delineated than the effects which have been found in SLA research. Two studies which consider pre- and postpuberty migrants ( Ammerlaan, 1996, AoA 0-29 yrs; Pelc, 2001, AoA 8–32 years) find that AoA is one of the most important predictors of ultimate proficiency, while a number of studies which investigate the impact of age among postpuberty migrants fail to find any effect whatsoever ( Köpke, 1999, AoA 14-36 yrs; Schmid, 2002, AoA 12-29 yrs; Schmid, 2007, AoA 17-51 yrs). A range of studies conducted by Montrul on Spanish heritage speakers in the US as well as Spanish-English bilinguals with varying levels of AoA also suggests that the L1 system of early bilinguals may be similar to that of L2 speakers, while later learners pattern with monolinguals in their L1 (e.g. Montrul, 2008 ; Montrul, 2009). These findings therefore indicate strongly that early (pre-puberty) and late (post-puberty) exposure to an L2 environment have a different impact on possible fossilization and/or deterioration of the linguistic system.

A recent investigation, focussing specifically on the age effect in L1 attrition, lends further substantiation to the assumption of a qualitative change around puberty: Bylund (2009) investigates the L1 of 31 Spanish speakers who emigrated to Sweden between the ages of 1 and 19 years and concludes that "there is a small gradual decline in attrition susceptibility during the maturation period followed by a major decline at its end (posited at around age 12)" (Bylund 2009:706).

The strongest indication that an L1 can be extremely vulnerable to attrition if exposure ceases before puberty, on the other hand, comes from a study of Korean adoptees in France reported by Pallier (2007) . This investigation could find no trace of L1 knowledge in speakers (who had been between 3 and 10 years old when they were adopted by French-speaking families) on a range of speech identification and recognition tasks, nor did an fMRI study reveal any differences in brain activation when exposing these speakers to Korean as opposed to unknown languages (Japanese or Polish). In all respects, the Korean adoptees presented in exactly the same way as the French controls.

All available evidence on the age effect for L1 attrition therefore indicates that the development of susceptibility displays a curved, not a linear, function. This suggests that in native language learning there is indeed a Critical Period effect, and that full development of native language capacities necessitates exposure to L1 input for the entire duration of this CP.

L2 attrition

In Hansen & Reetz-Kurashige (1999), Hansen cites her own research on L2-Hindi and Urdu attrition in young children. As young pre-school children in India and Pakistan, the subjects of her study were often judged to be native speakers of Hindi or Urdu; their mother was far less proficient. On return visits to their home country, the United States, both children appeared to lose all their L2 while the mother noticed no decline in her own L2 abilities. Twenty years later, those same young children as adults comprehend not a word from recordings of their own animated conversations in Hindi-Urdu; the mother still understands much of them.

Yamamoto (2001) found a link between age and bilinguality. In fact, a number of factors are at play in bilingual families. In her study, bicultural families that maintained only one language, the minority language, in the household, were able to raise bilingual, bicultural children without fail. Families that adopted the one parent - one language policy were able to raise bilingual children at first but when the children joined the dominant language school system, there was a 50% chance that children would lose their minority language abilities. In families that had more than one child, the older child was most likely to retain two languages, if it was at all possible. Younger siblings in families with more than two other brothers and sisters had little chance of maintaining or ever becoming bilingual.

Attrition and frequency of use

One of the basic predictions of psycholinguistic research with respect to L1 attrition is that language loss can be attributed to language disuse (e.g. Paradis, 2007 ; Köpke, 2007 ). According to this prediction, attrition will be most radical among those individuals who rarely or never speak their L1 in daily life, while those speakers who use the L1 regularly, for example within their family or with friends, will to some degree be protected against its deterioration. This assumption is based on the simple fact that rehearsal of information can maintain accessibility. The amount of use which a potential attriter makes of her L1 strikes most researchers intuitively as one of the most important factors in determining the attritional process (e.g. Cook, 2005 ; Paradis, 2007 ). Obler (1993) believes that less-frequently used items are more difficult to retrieve. The speed of retrieving a correct form or the actual production of an incorrect form is not indicative of loss but may be retrieval failure instead. In other words, what appears to be lost is in fact difficult to retrieve.

There is, however, little direct evidence that the degree to which a language system will attrite is dependent on the amount to which the language is being used in everyday life. Two early studies report that those subjects who used their L1 on an extremely infrequent basis showed more attrition over time ( de Bot, Gommans & Rossing, 1991 and Köpke, 1999 ). On the other hand, there is also some evidence for a negative correlation, suggesting that the attriters who used their L1 on a daily basis actually performed worse on some tasks (Jaspaert & Kroon 1989).

However, more recent, larger-scale studies of attrition which attempt to systematically elicit information on language use in a large range of settings have failed to discover any strong links between frequency of L1 use and degree of L1 attrition ( Schmid, 2007 ; Schmid & Dusseldorp, 2010 .)

Motivation

Paradis (2007:128) predicts that "Motivation/affect may play an important role by influencing the activation threshold. Thus attrition may be accelerated by a negative emotional attitude toward L1, which will raise the L1 activation threshold. It may be retarded by a positive emotional attitude toward L1, which will lower its activation threshold." However, the link between motivation and attrition has been difficult to establish. Schmid & Dusseldorp (2010) fail to find any predictors for individual performance among a large number of measures of motivation, identity, and acculturation. Given Paradis' prediction, this finding is surprising, and may be linked to methodological difficulties of measurement: by definition, studies of language attrition are conducted a long time after migration has taken place (speakers investigated in such studies typically have a period of residence in the L2 environment of several decades). The operative factor for the degree of attrition, however, is probably the attitude towards both L1 and L2 at the beginning of this period, when massive and intensive L2 learning is taking place, affecting the overall system of multicompetence. However, attitudes are not stable and constant across a person's life, and what is measured at the moment that the degree of individual attrition is assessed may bear very little relation to the original feelings of the speaker.

It is difficult to see how this methodological problem can be overcome, as it is impossible for the attrition researcher to go back in time, and impractical to measure attitudes at one point in time and attrition effects at a second point, decades later. The only study which has made an attempt in the direction of assessing identity at the time of migration is Schmid (2002), who investigated attrition in historical context. Her analysis of oral history interviews with German Jews points strongly towards an important correlation of attitude and identity at the moment of migration and eventual L1 attrition.

The reasons for this pattern of L1 attrition probably lie in a situation where the persecuted minority had the same L1 as the dominant majority, and the L1 thus became associated with elements of identity of that dominant group. In such situations, a symbolic link between the language and the persecuting regime can lead to a rejection of that language. Discovering such a link and its impact on the language attrition process is extremely complicated for most groups of migrants, since measurements are usually applied a long time after migration took place. However, the findings presented here suggest that it is the attitude at the moment of migration, not what is assessed several decades later, that impacts most strongly on the attritional process.

Other studies on bilingualism and attrition

Gardner, Lalonde, & Moorcroft (1987) investigated the nature of L2-French skills attriting by L1-English grade 12 students during the summer vacation, and the role played by attitudes and motivation in promoting language achievement and language maintenance. Students who finished the L2 class highly proficient are more likely to retain what they knew. Yet, interestingly, high achievers in the classroom situation are no more likely to make efforts to use the L2 outside the classroom unless they have positive attitudes and high levels of motivation. The authors write: "an underlying determinant of both acquisition and use is motivation”(p. 44).

In fact, the nature of language acquisition is still so complex and so much is still unknown, not all students will have the same experiences during the incubation period. It is possible that some students will appear to attrite in some areas and others will appear to attrite in other areas. Some students will appear to maintain the level that they had previously achieved. And still, other students will appear to improve.

Murtagh (2003) investigated retention and attrition of L2-Irish in Ireland with second level school students . At Time 1, she found that most participants were motivated instrumentally, yet the immersion students were most likely to be motivated integratively and they had the most positive attitudes towards learning Irish. Immersion school students were also more likely to have opportunities to use Irish outside the classroom/school environment. Self-reports correlated with ability. She concludes that the educational setting (immersion schools, for example) and the use of the language outside the classroom were the best predictors for L2-Irish acquisition. Eighteen months later, Murtagh finds that the majority of groups 1 and 2 believe their Irish ability has attrited, the immersion group less so. The results from the tests, however, do not show any overall attrition. Time as a factor did not exert any overall significant change on the sample's proficiency in Irish (Murtagh, 2003:159).

Fujita (2002) , in a study evaluating attrition among bilingual Japanese children, says that a number of factors are seen as necessary to maintain the two languages in the returnee child. Those factors include: age on arrival in the L2 environment, length of residence in the L2 environment, and proficiency levels of the L1. Furthermore, she found that L2 attrition was closely related to another factor: age of the child on returning to the L1 environment. Children returning around or before 9 were more likely to attrite than those returning later. Upon returning from overseas, pressure from society, their family, their peers and themselves force returnee children to switch channels back to the L1 and they quickly make effort to attain the level of native-like L1 proficiency of their peers. At the same time, lack of L2 support in the schools in particular and in society in general results in an overall L2 loss.

Conclusion

The loss of a native language is often experienced as something profoundly moving, disturbing or shocking, both by those who experience it and by those who witness it in others: “To lose your own language was like forgetting your mother, and as sad, in a way”, because it is “like losing part of one’s soul” is how Alexander McCall Smith puts it (The Full Cupboard of Life, p. 163). This intuitive appeal of the topic of language attrition can be seen as both a blessing and a curse.

In order to achieve a better understanding of the process of language attrition, scientific investigations first have to identify those areas of the L1 system which are most likely to be affected by influence from the L2. Initially, in the absence of experimental data and evidence of L1 attrition itself, a profitable approach is to look to neighbouring areas of linguistic investigation, such as language contact, creolisation, L2 acquisition, or aphasia. In the early years of L1 attrition research, the 1980s, many researchers made valuable contributions of this nature, often augmented by small-scale experiments and/or case studies (see Köpke & Schmid, 2004 ). A number of strong predictions with intriguing theoretical implications were made during this period of L1 attrition research (ibid).

There are a number of factors which will impact in different ways on the process of L1 attrition. Frequent use (interactive or receptive) of a particular language may help to maintain the native language system intact, and so may a positive attitude towards the language or the speech community. On the other hand, none of these factors may be enough in themselves, and not all exposure to the language may be helpful. A small, loose-knit L1 social network may even have a detrimental effect and accelerate language change. Most importantly, however, the opportunity to use a language and the willingness to do so are factors which interact in complex ways to determine the process of language attrition. As yet, our understanding of this interaction is quite limited.

See also

References