The acquisition of /s/ - /z/ in a phonemic vs neutralised context: comparing French L1, Italian L1 and Spanish L1 learners of L2 English

The acquisition of /s/ - /z/ in a phonemic vs neutralised context: comparing French_L1, Italian_L1 and Spanish_L1 learners of L2 English

Leonardo Contreas Roa, Paolo Mairano, Caroline Bouzon, Marc Capliez

Université de Lille - UMR 8163 STL

Introduction

English has a high functional load voice contrast between /s/ and /z/, which is active word-initially (sing /s/ - zing /z/), word-medially (fussy /s/ - fuzzy /z/) and word-finally (rice /s/ - rise /z/). However, this contrast is neutralised in the pronunciation of morphemic -s (for plural, 3rd person, genitive, and clitic forms of has and is). In this specific context, it is subject to a progressive voice assimilation rule (/s/ in pets due to /t/ being voiceless, but /z/ in beds due to /d/ being voiced) (cf. ¹).

In this study we investigate the acquisition of /s/ - /z/ in L2 English by comparing contexts in which these sounds have a phonemic value vs contexts in which they are determined by a voice assimilation rule. We observe English_L2 productions by French_L1, Northern Italian_L1 and Southern American Spanish_L1 learners, on the assumption that the three groups will show different patterns depending on the status of [s] and [z] in their L1s. These sounds are phonemic in French (hausse /os/ - ose /oz/), and allophones in varieties of Northern Italian, where [z] appears before voiced Cs and between non-C segments, and [s] in front of voiceless consonants (cf. ²). Spanish only has /s/ (although partial or total voicing can occur in syllable coda due to non-obligatory voice assimilation with the following C in casual speech, cf. [^3]). So, based on SLM predictions (cf. ³), we expect that (i) French_L1 learners will show categorically distinct realizations for /s/ and /z/; (ii) Northern Italian_L1 learners will be able to produce the voice opposition, though potentially to a lesser degree since these sounds are allophones in their L1; (iii) Spanish_L1 learners may not be able to produce any difference for /s/ and /z/; (iv) voice assimilation for morphemic -s will be difficult for all learners because word-final /z/ is universally more marked (cf. ⁴), and all L1s have regressive (rather than progressive) voice assimilation rules.

Data

We analysed productions by 40 instructed learners from the IPCE-IPAC corpus of L2 English. Learners were 15 speakers of Metropolitan French (12 F, 3 M, age = 24, SD = 6.59), 15 speakers of Northern Italian (11 F, 4 M; age = 22.5, SD = 2.38), 10 speakers of Spanish (3 F, 7 M; age = 30.2, SD = 6.98) from Peru (n = 5), Chile (n = 3), Colombia (n = 2). For the present study we only considered recordings of the read-aloud task (506 words), which provides perfectly comparable data. The recordings were transcribed orthographically, phonetized and aligned with WebMAUS, and manually verified. For each occurrence of /s/ and /z/, the proportion of periodic signal was extracted via a custom Praat script, thereby obtaining a value ranging from 0 (no periodicity detected) to 1 (periodicity detected throughout the whole target segment). Additionally, we also extracted the duration of segments as a secondary cue of voicing, but durational data are not discussed in this abstract due to space constraints. The results were then imported to R for visualisation and statistical analysis.

Results

The results for phonemic /s/ and /z/ (Figure 1) reflect the expected pattern: Spanish_L1 learners tend not to produce any difference in periodicity between /s/ and /z/, whereas French_L1 and Italian_L1 participants show distinct realizations for these two sounds. This is confirmed by a linear mixed-effects model predicting the periodicity on the basis of Sound (/s/, /z/), Group (FR, IT, SP), Context (intervocalic, non-intervocalic) and Position (word-media, word-final). Post-hoc tests confirmed that the difference in periodicity between /s/ and /z/ is significant for French_L1 and Italian_L1 learners (p < .001), but not for Spanish_L1 learners (p = .11). Instead, the results for the pronunciation of morphemic -s (figure 2) show that all learner groups tend to reproduce (at least globally) the output of the voice assimilation rule: the target segments are voiceless when following a C[-voice], and partially periodic if following a C[+voiced] or a vowel. While French learners clearly show the highest differentiation among conditions, it is surprising to observe that Spanish_L1 learners seem to be able to reproduce the assimilation pattern, and even more neatly than Italian_L1 learners. This is confirmed by a linear mixed effects model similar to the one above, revealing that differences across conditions are significant for the French_L1 group (all p values < .001) and the Spanish_L1 group (all p values < .03), but for the Italian group the condition C[-voice] does not significantly differ from the condition C[-voice] (p = .27).

Conclusions

SLM predictions were confirmed in phonemic contexts. Instead, we find unexpected results for the voice assimilation rule, whereby SpanishL1 learners manage to produce more voiced realisations than ItalianL1 learners. The existence of a non-obligatory voice assimilation rule in Spanish (incl. morphemic -s, used for plural in Spanish) may promote voice assimilation in syllable-coda in L2 English, although with a change in directionality (regressive to progressive assimilation). Moreover, the behaviour shown by SpanishL1 and ItalianL1 learners in morphemic vs non-morphemic -s may reflect ⁵’sfindings for L1 English that morphemic and non-morphemic -s are not homophonous. If confirmed on more data, such results may have an impact on mainstream models of L2 phonology acquisition.

Figure 1. Average periodic proportion for realizations of /s/ and /z/ in phonemic contexts.

Figure 2. Average periodic proportion for realizations of morphemic -s.