First language attrition
Monika S. Schmid, University of Groningen
Abstract:
Speakers who live in an L2 environment for an extended period of time often experience
change in the way in which they use their L1, a process referred to as L1 attrition. The present
article provides an overview of language attrition phenomena at various linguistic levels.
However, attrition cannot be trivially or linearly related to factors such as the frequency of use
of the L1. It is argued here that attrition phenomena are not the outcome of a change to the
underlying linguistic system nor of access problems due to an increase in activation
thresholds, but of crosslinguistic influence in online speech production.
One of the most fundamental and most interesting characteristics of linguistic data of any kind
is its inherent variability. Even mature, monolingual native speakers will usually not perform
according to the 'target standard' one hundred percent of the time. In free spoken data, there
are slips of the tongue, grammatical errors, variance in pronunciation and so on. Any kind of
formal test on which all participants achieve the maximum score can be deemed to show a
ceiling effect due to being too simple. In this context, it is interesting to ask what factors will
have an impact on this variance, and in what way. If there are distracting factors, skills that
rely on procedural memory and are therefore largely automatic, such as grammar or
phonology, are less prone to interference than the accessing of information represented in
declarative memory, such as lexical words (Paradis, 2004; 2009). Fatigue can also play a role,
and it has recently even been suggested that linguists might do well to take into account
insights from chronobiology on what have been called 'Circadian Rhythms', that is, the
optimal time of day for an individual to perform at his or her best (a factor that varies
considerably within populations) (de Bot, forthc.).
One of the strongest factors that determine variability in linguistic performance relates to
the way in which the language under observation has been acquired, as well as other
languages that the speaker might know. Comparisons of speakers who have been exposed to
the target language from birth (native speakers or L1ers) with others for whom it was not the
first language (L2ers) will usually find that the L1ers perform 'better', and that there is a wider
range of scores represented among the L2ers. Factors that can contribute to this distribution
include the age at which the L2ers started to learn the language in question, the manner of
acquisition (naturalistic vs. formal), the amount of time devoted to language learning and
input received, and individual characteristics such as language learning aptitude, attitude and
motivation.
While investigations into these types of variability and cross-linguistic influence (CLI) in
a speaker's L2 are well-established, there is much less research into CLI affecting the
bilingual's L1. This process, commonly known as L1 attrition, can become highly noticeable
in speakers who have migrated to another country and henceforth lived in an L2-dominant
environment, but is by no means confined to such relatively extreme settings (Schmid &
Köpke, 2007). For example, it has been shown that some phonetic categories in the first
language of bilingual speakers exhibit phonetic drift, that is, the assimilation of the properties
of a sound towards the value specified by the other language. This phenomenon is strongest
for very experienced long-term immersed bilinguals (Flege, 1987) but can even be observed
in novice learners, as Chang (2010) demonstrates for a group of native English speakers
enrolled in an intensive six-week (beginner's) course of Korean. Syntactic processing in the
L1 appears to be similarly affected. For example, Dussias and Sagarra (2007) investigate
relative clause attachment in English and Spanish. They point out that the sentences in (1)
have different meanings for monolingual speakers of these two languages:
(1a) An armed robber shot the sister of the actor who was on the balcony.
(1b) Un ladrón armado le disparó a la hermana del actor que estaba en el balcón. (p.101,
their examples (1) and (2)
While English monolinguals tend to interpret (1a) to mean that the actor was on the balcony
(low attachment), in Spanish, most speakers would take it to be the sister in (1b). These
tendencies can be investigated by introducing a disambiguating element (e.g. 'the sister of the
actor who was pregnant' or 'the brother of the actress who was pregnant') and studying the
way in which monolinguals and bilinguals resolve the potential conflict, e.g. by means of self-
paced reading tasks or recording eye-movements. Dussias and Sagarra show that parsing
strategies can, to some extent, be transferred from the L2 to the L1 among highly proficient
L2 speakers, but that this transfer is contingent on the amount of exposure that the speaker has
to the L2.
These findings show that processes of transfer that are to some extent similar to the
crosslinguistic influence we witness in second language learners can also come to affect the
L1 of bilingual speakers. This transfer becomes more intensive and noticeable among
speakers who are highly proficient in another language, receive extensive input in that
language and (presumably) use it frequently (although none of the studies discussed above
have quantified linguistic output as a factor) and it is noticeable across all linguistic levels. In
particular where the lexicon is concerned, L2-to-L1 transfer appears to be very common (for
an overview see Schmid & Jarvis, submitted; Schmid & Köpke, 2008) but at least some
aspects of grammatical and phonetic categories also appear to be open to L2 impact (see e.g.
the studies collected in Schmid, 2010, Schmid & Köpke, 2011)
The scope and limitation of attrition effects
The situation of linguistic drift sketched above, where a migrant achieves a high level of
proficiency in the language of his or her new environment, uses this language on a daily basis,
and consequently experiences an increase in variability in the way some lexical, grammatical
or phonetic properties of the language are applied, has been termed language attrition.
1
The
metaphor underlying this label is not an entirely felicitous one, partly due to the many
inherently negative collocations using the term attrition (e.g. 'war of attrition') but also
because of the strong presumption it evokes that the process will be one of some sort of
linguistic 'reduction' due to a constant 'grinding away' at the substance or fabric of the attriting
language caused by the use of another (as is the implication of the term 'attrition' in other
areas, e.g. geology or dental health). However, it is by no means established to what extent
attrition is indeed the outcome of such a 'war of attrition' between the two competing
languages (and thus for example dependent on the amount of exposure that a speaker retains
with his or her L1, see below).
When asked about the extent to which they themselves are affected by attrition, most
speakers will immediately latch on to problems of lexical access, and this is also often pointed
out in attrition research as the aspect of linguistic knowledge that is most vulnerable to
attrition effects (de Bot, 1996; Hulsen, 2000; Köpke and Nespoulous, 2001; Köpke &
Schmid, 2004; Montrul, 2008; Opitz, 2011, to name but a few). However, most studies which
have put this assumption to the test have failed to find any truly dramatic lexical loss,
problems of lexical access (e.g. Schmid & Jarvis, submitted), nor substantially reduced speed
of lexical retrieval (Yılmaz & Schmid, forthc.). Proficient bilinguals may, on occasion, indeed
1
The term language attrition pertains to the process of language change experienced among speakers for whom the
language under observation had stabilized prior to the onset of attrition effects, ie. migrants who left the country in which
their L1 was spoken and in which they had acquired it after puberty. This phenomenon should be distinguished from
incomplete acquisition, that is, the restructuring of linguistic knowledge by the acquisition of another language earlier in
life (see e.g. Montrul, 2008). In the present contribution, I will reserve the term attrition exclusively for the former case
and not talk about pre-puberty migrants.
experience some crosslinguistic transfer in their lexicon. Schmid (2011), based on the
taxonomy offered by Pavlenko (2004), attempts to provide a classification of such
phenomena, comprising the following categories:
1. Borrowing: the use of an item from the L2 in the L1, often in such a way that it is
integrated phonologically and/or morphologically (e.g. English tends to assign initial
stress to bisyllabic French loanwords which are stressed on the second syllable in
the donor language, and German forms the past participle of borrowed verbs by
means of the circumfix ge-verb-t, leading to bizarre forms such as 'gedownloadet').
2. Restructuring: the meaning of an existing L1 word is extended to include the
meaning of its L2 translation equivalent. This process often involves verbs which
have undergone semantic bleaching and been turned into quasi-auxiliaries, such as
the English verb to run which occurs in the collocations 'to run short of something'
or 'to run for office'. Attriters often transfer such fixed expressions from the L2 to
the L1.
3. Convergence: where both languages offer lexical items that are similar in form but
different in meaning (faux amis), attriters may sometimes be mislead to use the L1
term with the L2 meaning. For example, Sharwood Smith (1983) quotes the example
of an L1 English-L2 Dutch bilingual, who used the verb to overdrive to mean to
exaggerate, based on the Dutch homophonic item overdrijven.
4. Shift: unlike the processes listed under 1-3 above, shift is not limited to individual
lexical items but concerns entire lexical fields, as Pavlenko (2002, 2004) has
demonstrated with respect to emotion terms. Shift usually affects areas of the
linguistic repertoire which are highly culture-specific, such as politeness sytems.
What these types of crosslinguistic influence have in common is that they are, to a greater or
lesser degree, visible on the surface of an utterance; speakers, listeners and researchers
observe and notice them immediately. This is probably why they are often (and in my opinion
mistakenly) interpreted as evidence for L1 attrition. There is, however, no reason to conclude
that, just because a speaker has borrowed an item or used it in an inappropriate collocation,
s/he is at that point unable to retrieve or use the original and target form of the item. It simply
means that, on this particular occasion, the resources provided by the L2 underpinned lexical
choice.
More interesting from the point of view of the loss of linguistic proficiency, but also
more difficult to establish in practice, is the question of whether the actual volume of
vocabulary (either productive or receptive) is truly reduced in the process of language
attrition, and to what extent its accessibility can become affected. Investigations of L1
attrition using controlled tasks which allow the speaker to focus fully on the retrieval of
lexical items (placing no demands on other components of language production), such as
Picture Naming or Verbal Fluency tasks, typically find access problems to be extremely
limited. Attriters may be slightly slower to name items (in particular low-frequency ones, see
Yılmaz & Schmid, forthc.) or be slightly less productive in the number of items of a particular
category that they can name within a certain time span (for an overview, see Schmid & Jarvis,
submitted), but these findings may be ascribed to the fact that bilinguals have to manage a
substantially larger number of lexical items than monolinguals.
Investigations of the way in which attriters make use of their productive vocabulary in
(elicited) free speech - that is, in situations where all aspects of language production come
into play and it is not possible for the speaker to focus attention solely on lexical retrieval -
typically tend to find a higher number of disfluency markers (Schmid & Beers Fägersten,
2009), sometimes accompanied by a slight reduction in lexical diversity measures such as
VOCD (for an overview see Schmid & Jarvis, submittted). Again, it is uncertain to what
extent such findings may be ascribed to an actual reduction in lexical accessibility of the L1
items, or merely to the demands of managing two linguistic systems and inhibiting the (more
highly activated) L2. The fact that bilinguals become slower to name objects in their first
language very shortly after the onset of bilingualism (and so probably not due to attrition) has
been demonstrated e.g. by Mägiste (1979). On the whole, the differences between attriters and
monolingual controls tend to be robust but hardly dramatic, contradicting the general
assumption that the lexicon is that part of the linguistic repertoire that is extremely vulnerable
to L1 attrition.
Where grammar is concerned there is a substantial body of research investigating to what
extent particular features of a language may be restructured under conditions of L1 attrition.
Many of these studies are situated within the Chomskyan framework, e.g. the Principles and
Parameters approach (e.g. Gürel, 2004, 2007; Gürel & Yılmaz, 2011; Kim, Montrul & Yoon,
2010) or the Minimalist Program (e.g. Tsimpli, Sorace, Heycock and Filiaci, 2004; Tsimpli,
2007). In the latter context, recent grammatical investigations of language attrition have often
focussed on Sorace's Interface Hypothesis (e.g. Sorace, 2011) which predicts that "structures
involving an interface between syntax and other cognitive domains are less likely to be
acquired completely than structures that do not involve this interface" (Sorace, 2011: 1) and