SECOND LANGUAGE EFFECTS ON A FIRST LANGUAGE

Although the phonetics and phonology of a first language (L1) are less likely to be affected by a second language (L2) than, for example, the lexicon, a number of experimental investigations have detected changes in either perception or production in both cases, showing bilingual speakers, to be differentiated from the monolingual norm. These studies imply that knowledge and use of a second language may, under some circumstances, affect a bilingual speaker's first language (see Cook, 2003; Schmid, 2002).

One thread in experimental studies of bilingual perception and production has focused on voicing as specified by voice onset time (VOT). One of the first was conducted by Caramazza, Yeni-Komshian, Zurif, and Carbone (1973). They found that their subjects' first language, French, was affected by their second language, English, in the perception of voicing but not in production. The bilingual participants were all native speakers of French and had begun to learn English as children, by no later than their seventh birthday. Bilinguals did not switch from one perceptual set to another in order to match the perceptual functions of the monolingual speakers. Rather, they appear to have based their perceptual decisions on the same criteria in both French and English. In production, the participants' use of VOT in French was not affected by their second language. Williams (1976, 1977, 1978) and Elman, Diehl, and Buchwald (1977) replicated the Caramazza et al. study with Spanish–English bilinguals, and found—just as Caramazza et al. had—that the bilingual speakers preserved the phonetic characteristics of each language in production but not in perception. On the other hand, Mack (1984, 1989) found no significant differences between the performance of bilingual and monolingual talkers in either perception or production of voicing.

Flege and Hillenbrand (1984) and Major (1992) have suggested that laboratory studies may underestimate the effects of a strong second language on a first language. Major (1992) investigated the English pronunciation of five native speakers of American English who had been living permanently in Brazil from 12 to 35 years. All were middle-class college educated speakers who used English on a daily basis in connection with their professions as English teachers or administrators in an English language institute. In spite of both practice in English and, presumably, a desire to maintain their first language, all of the speakers showed some changes in their pronunciation of voiceless stops, as indicated by VOT measurements, in either casual or formal settings, or in both. All of the changes were in the direction of Portuguese voice onset time values even for speakers whose Portuguese pronunciation differed markedly from that of native speakers. Flege and Hillenbrand (1984) examined the production of native French speakers who were living in an English-speaking environment and using English on a daily basis. None of these talkers resembled monolinguals. Rather, all of them employed VOT values substantially in excess of the values observed in the speech of French monolinguals, showing the effects of extensive use of a second language. These shifts in perception and production can take place relatively rapidly, over months rather than years (Sancier & Fowler, 1997). All these studies agree, however, in their finding that some aspect of the phonetics of L1 shifts so that bilingual talkers employ phonetic values in their L1 that are more similar to the phonetic values of L2 than the values employed by monolingual talkers.

Although much of the research has examined voicing, L2 effects on the production and perception of vowels in L1 have also been reported. The characteristics of L2 influence are not always easy to specify. For example, Mack (1984, 1989) reported that English–French bilinguals treated vowel quality for the /i/-/I/ contrast differently from monolinguals. She concluded that bilingual phonetic transfer or systematic restructuring of the phonetic component can occur in a dominant language used regularly and productively since early childhood. A case study of contact between French and English, Bullock and Gerfen (2004) showed a pattern that might be related, that is, the merger of mid front rounded vowels as schwar (

). Baker and Trofimovich (2005) reported that the acoustic structure of Korean vowels was “colored” by the English vowel system for some Korean–English bilinguals.

Although L2 influence on L1 vowel production or perception has been documented, the exact specification of the influence appears to be quite complex, reflecting age, similarity between the phonological systems of languages, patterns of use of the two languages, and so forth. In particular, although it is possible for the categories of the two languages to become more similar, the reverse pattern of phonetic influence in bilingualism has also been reported. In this pattern, the phonetic values of L1 differ from a monolingual norm by exaggerating differences. This development has been found for both voicing and vowels. Flege and Eefting (1987a) examined the voicing contrast in the speech of three groups of English–Spanish bilinguals. In this study, the second language, English, affected the pronunciation of Spanish, but, instead of using a compromise value for both languages, the bilingual talkers produced the Spanish voiceless stops with significantly shorter VOT values than monolingual Spanish speakers. That is, their Spanish productions were less, rather than more, English-like. In a similar study examining Dutch–English bilinguals, Flege and Eefting (1987b) found that proficient Dutch speakers of English produced Dutch /t/ with shorter VOT values than less proficient speakers, emphasizing the contrast between the phonetic realizations of voiceless stops in the two languages. Similarly, Flege, Schirru, and MacKay (2001) found that native speakers of Italian who had started to learn English before adolescence tended to produce English /e/ with more diphthongization than native English speakers, suggesting that they were dissimilating the category of the English vowel from the Italian monophthongal /e/. Finally, Guion (2003) found that Quichua vowels produced by Spanish–Quichua bilinguals were raised in comparison with the vowels produced by monolingual Quichua speakers. She also argued that this vowel reorganization serves to enhance the distinctiveness of vowels in the two languages.

On the basis of such findings, Flege (1995) has argued that bilinguals create a perceptual representation of phonological contrasts that serve as a compromise between their two languages. If bilingual talkers consider two sounds in their languages to be instances of the same category, then these two sounds will come to be produced and perceived in similar ways; that is, the first language and the second language sounds will be produced with characteristics that are a compromise between the two languages. On the other hand, if bilingual talkers consider a sound in their second language to be “new,” not equivalent to a sound found in L1, they may create a new category for the sound. Because the sounds of the two languages exist in a common phonological space, bilinguals may modify differences among the sounds of L1 in order to maintain phonological contrasts. The effects may take place at the level of phonetic implementation of contrasts, and, apparently, occur without conscious awareness.

Because the production of vowels allows for relatively direct measurement of gradient influences, in this study we examined the pronunciation of Latvian vowels by native Latvian talkers representing various generations since the Russian occupation after World War II. Our objective is to investigate the phonetic structure of Latvian over time, to determine how the pronunciation of the language has been affected by nearly universal bilingualism on the part of Latvians, coupled with exposure to accented Latvian as produced by ethnic Russians.

LATVIAN AND RUSSIAN VOWELS

The vowel systems of the two languages differ in inventory and in the way that prosodic patterns interact with vowel quality. Latvian employs the vowels

. Although there are differences in the distribution of vowels, all can be contrastively long or short in any syllable of a word. Stress is fixed on the first syllable. Unstressed short vowels are sometimes extremely abbreviated or even absent, but they do not appreciably change in quality (Laua, 1984). Latvian does not have any series of consonants that systematically affect vowel quality.

The traditional analysis of the standard Russian vocalic system posits five vowels

, with considerable variability of positional allophones, the most salient being the difference between the so-called hard and soft vowels, that is, vowels following palatalized consonants. In fact, because of these differences, some analyses of Russian posit a sixth vowel, a high, unrounded

. The formant structure of Russian vowels has been extensively described by Fant (1970), Lobanov (1971), and Jones (1971), among many others. Russian employs variable stress, and the vowel system reduces in unstressed syllables: after nonpalatalized consonants, /e/ rises to /i/ and /o/and /a/ merge as either a low central vowel or as

(Padgett, 2005). According to Jones (1971), the most salient differences are found in values of the second formant. Russian does not employ contrastive vowel length; rather vowel length is associated with stress. It is not known whether Russian as spoken in Latvia has developed any characteristics that distinguish it from the standard.

These differences between the two languages suggest that habitual use of Russian has the potential to influence the Latvian pronunciation of vowel quantity and vowel quality. If Latvian vowel production has been directly modified to conform to the Russian pattern, then we might expect the following tendencies:

Weakening of contrast between Latvian

.