Second language learners and the variable speech signal.

Abstract

In the article “Age of acquisition andallophony in Spanish-English bilinguals”Barlow (2014) presents production dataof /l/ from two groups of Spanish-Englishbilinguals, who differ on age of acquisi-tion of English (before 5 years or after 6years of age). Barlow’s contribution is awelcome addition to the relatively under-studied field of allophone acquisition bysecond language learners. In what followsI expand upon issues touched upon byBarlow in her article and comment moregenerally on why the issue of variabilityin the speech stream (of which allophonesin complementary distribution is but onetype) must be addressed differently for L2learners than for infants acquiring a firstlanguage. I restrict the discussion to per-ception and will primarily address issueskey to adult (i.e., individuals who beganto acquire their second language after thesound system of their first language is inplace) second language acquisition.Research on how adults perceive non-native sounds has received considerableattention over the past thirty years (seework by Flege and Best for the most influ-ential models of L2 perception) and thevast majority of this work has lookedat the way in which non-native soundsassimilate into native-language sound cat-egories, independent of the context inwhich they occur (for an exception, seeLevyandStrange,2008).Inacertainsense,it can be said that much of this researchabstracts away from speech perception asit unfolds in real time (McMurray andJongman,2011).Partofthechallengereal-time speech perception represents for L2learners involves dealing with the way co-occurring sounds (or abstract contextssuch as stress, see Shea and Curtin, 2011)affect each other or lead to variability,whether predictable or indexical innature.Thestudyofallophoneacquisitionrep-resents an effort to break away fromthis tradition and can be included inthe broader research program that exam-ines how learners deal with the variabil-ity found in the input. Indeed, variabilityitself is “highly variable” and can be dueto individual speaker differences, dialectdifferences, speech rate, and formality.These kinds of variability are often dis-tinguishedfromallophonicvariabilitythatis the result of phonetic or phonologi-cal factors and tend to occur in a moreacross-the-board fashion in speech.In terms of L1 acquisition, part oflearning a language’s sound system nec-essarily involves learning which soundscontrast and which do not. Research sug-gests that distributional knowledge andphonetic similarity play a key role inguiding infants toward identifying non-phonemic sounds in their language (seeSeidl and Cristia, 2012 foranexcellentoverview; see Yeung and Werker, 2009,for work showing that a lack of lexi-cal contrast can be used by infants toacquire allophones in non-contrastive dis-tributionsaswell).Forexample,inarecentstudy, Seidl et al. (2009) examined therole of phonemic vs. allophonic contrastsin infant speech perception. They famil-iarizedFrench-learning11-month-oldandEnglish-learning 11- and 4-month-oldinfants to syllables in which the finalconsonants conditioned the nasality ofthe previous vowel. In French, nasality isphonemicwhileinEnglishitisallophonic.The results showed that French-learning11-month-olds and English-learning 4-month- olds had a reliable pattern of pref-erence while English 11-month-olds wereinsensitive to the patterning, orientingequally to syllables following and violat-ing the familiarized patterns. The authorsconclude that language-specific sensitiv-ity to context-driven allophonic contrastsemerges as early as 11 months of age.In contrast, adult native listenersdistinguish allophonic contrasts at a pho-netic level less accurately than phonemiccontrasts. For example, Pegg and Werker(1997), using an AX discrimination task,showed that native English-speaker adults’performance on the allophonic contrastbetweenvoiced[d]andthevoicelessunaspirated [t] was better than chance,but nonetheless worse than that on aphonemic contrast (for similar results seeWhalen et al., 1997).In addition to perceiving the dif-ference between two different phones,there is another important component toallophonic acquisition: its context-drivennature. Specifically, allophonic perceptioncannot be truly categorized as such unlessthe sounds occur in the context in whichthey are expected (or not, see Shea andCurtin, 2011 for details; Key, 2014). Forexample, Peperkamp et al. (2001),usingthe French [χ]-[] alternations showedthat French listeners could discriminatebetween allophonic segments in CV sylla-bles but as soon as the CV syllables wereput into their allophonic contexts, suchdiscrimination disappeared. Thus, to trulyspeak of “allophone perception” listenersmust be aware of the contrast but also thecontext in which it occurs.The mechanism by which infants buildtheirsoundcategoriesisbasedupontrack-ing distributional frequencies across thespeech stream (Maye et al., 2002). A num-ber of laboratory studies reveal that suchlearning is possible in both infants and