Carol A. Chapelle - The Concise Encyclopedia of Applied Linguistics

Language interaction in bilinguals is manifest in multiple phenomena. Among them are accented speech and accented comprehension, that is, the production of particular linguistic elements that differ from the way monolingual speakers would produce them, and differences in language comprehension processes as compared with monolinguals listening to or reading analogous language input. The word “accent” is used in a broad sense here, referring to differences in both language production and language comprehension in bilinguals as compared with monolinguals, and to differences in all linguistic subdomains, not just phonology.

One possible source of accents is parallel activation of representations of linguistic elements in bilinguals' two language subsystems, even when they have selected one of their languages (the “target” language) for current use. Because the language system of monolinguals only stores linguistic units belonging to a single language, such parallel activation does not occur when monolinguals process language. According to this view, the representation units themselves do not need to differ between monolinguals and bilinguals. For instance, the representation of the English phoneme /t/ and the stored meaning for English cat in an English–French bilingual are identical to the representation of English /t/ and the stored meaning of English cat in a monolingual English speaker. The second possible source of bilingual speech accents is that bilinguals may have developed memory representations of specific linguistic units that differ from the representations of the corresponding units in monolingual memory. For instance, bilinguals may have developed representations that merge a pair of corresponding representations in monolingual speakers of their two languages. The former source of accents may be regarded a difference in processing or “performance”; the latter a difference in knowledge or “competence.”

Flege and his collaborators have shown a phonological accent, in both L1 and L2, when bilinguals produce speech sounds (e.g., Flege, 2002). In some studies they measured the “voice onset time” (VOT) of consonants spoken by bilinguals and monolinguals in the context of a larger language fragment. The VOT is the time between the release of the air and the moment the vocal cords start to vibrate when a speaker produces a consonant. The VOT for one and the same consonant may differ between languages. For instance, the consonant /t/ is spoken with a longer VOT in English than in French and Spanish. This fact gives rise to the question of how bilingual speakers of two languages that exploit different VOT values in producing one and the same consonant utter this consonant. Flege and his colleagues have shown that the VOT values of such consonants differ between monolinguals and bilinguals. Specifically, when spoken by bilinguals these consonants take on VOT values that are intermediate between those of the same consonants spoken by monolingual speakers of the two languages concerned. For instance, if English–French bilinguals and English monolinguals are asked to pronounce the speech fragment two little dogs , the VOT of the /t/ sound in two is shorter for the English–French bilinguals than for the English monolinguals. Conversely, if English–French bilinguals and French monolinguals are asked to pronounce the speech fragment tous les chiens , ‘all dogs’, the VOT of the /t/ sound in tous is longer for the English–French bilinguals than for the French monolinguals.

Flege (2002) attributed these phonological accents to two L2 speech learning processes. One of these, “phonetic category assimilation,” is thought to lead to representations that merge closely similar L1 and L2 sounds into a single phonetic category in memory. The second, “phonetic category dissimilation,” is thought to operate when an L2 sound is very different from all L1 sounds stored in memory. A separate representation for the new L2 sound is then formed in memory, but the position it takes up in phonetic space differs from the position occupied by this sound in monolingual speakers of the language concerned. Furthermore, while inserting a phonetic category for this new sound into the phonetic space, it pushes away one or more of the categories that represent L1 sounds from their original positions (causing an accent).

Though category assimilation and dissimilation provide a plausible explanation of the phonological accents in bilingual speech production, an account in terms of parallel activation of two analogous L1/L2 phonetic categories (e.g., a French‐like /t/ and an English‐like /t/) appears equally plausible, at least for both early bilinguals and late proficient bilinguals. Early bilinguals can already perceive the difference between certain pairs of closely similar L1 and L2 phonetic categories from 10 to 11 months onward (e.g., Sundara, Polka, & Molnar, 2008) and late proficient bilinguals can also do this (Flege, 2007). This discrimination ability clearly points toward the existence of separate phonetic representations for similar L1/L2 sounds because it is hard to see how a difference between two such speech sounds can be perceived at all if they share one and the same representation. The very existence of such pairs of representations for speech sounds that are similar in L1 and L2 renders an interpretation of accented speech sounds in terms of their parallel activation plausible (see De Groot, 2014).

A grammatical accent in bilinguals can, for instance, be witnessed by looking at the way they parse sentences that are (temporarily) structurally ambiguous in one of their languages but not in the other, or sentences that are structurally ambiguous in both languages but for which the preferred solution differs between these languages. An example of the first type of ambiguity is the English sentence The leader defeated in the election resigned one day later (Rah & Adone, 2010), where defeated can either be the simple past of the transitive main verb or the passive participle of a reduced relative clause, the nonreduced form being who was defeated . In other languages, such as German and Dutch, the relative clause construction always takes a nonreduced form so no temporary ambiguity exists. An example of the second type of ambiguity is the English sentence Someone shot the son of the actress who was on the balcony , where either the head of the complex noun phrase ( the son ) or the second noun in this phrase ( the actress ) can be the subject of the relative clause ( who was on the balcony ). Though both structural solutions occur in English, English favors a “low attachment” analysis of this type of sentences, where the second noun in the complex noun phrase ( the actress ) is most often the subject of the relative clause. In other languages, such as Spanish, this ambiguous structure also exists ( Alguien disparó contra el hijo de la actriz que estaba en al balcón ), but the “high attachment” solution is more often correct and preferred. That is, el hijo is most often (initially) assigned the role of relative‐clause subject, forcing a reanalysis of the sentence if later on this solution turns out to be the wrong one (when estaba is encountered). The central question in this line of research is whether bilinguals parse such ambiguous constructions differently from monolingual speakers of the two languages in question, thus evidencing a grammatical accent.

Dussias and her colleagues examined how Spanish–English bilinguals parse sentences of the second type. The results suggested an influence of the other language on the way bilinguals analyze them and that the context of testing may modulate this effect: When testing took place in a predominantly English‐speaking environment in the USA, both Spanish L1/English L2 and English L1/Spanish L2 bilinguals generally favored low attachment over high attachment irrespective of the language of the presented sentences, English or Spanish (Dussias, 2003). In other words, the Spanish sentences were analyzed according to the English‐like parsing strategy, demonstrating an accent in Spanish. In another study (Dussias & Sagarra, 2007), Spanish–English bilinguals immersed in L1 Spanish and presented with Spanish sentences behaved like the monolingual Spanish control subjects, favoring high attachment. In contrast, Spanish–English bilinguals presented with Spanish sentences but immersed in L2 English preferred the low attachment solution that is most common in English, thus showing a grammatical accent in L1. In short, bilinguals appear to prefer the parsing procedure that is most common in the language they are currently exposed to most. This in turn suggests that the two grammatical‐knowledge structures that enable the two different parses are activated to different degrees across different language contexts.