Elsevier

Cognition

Volume 106, Issue 2, February 2008, Pages 682-706
Cognition

Persistent stress ‘deafness’: The case of French learners of Spanish

https://doi.org/10.1016/j.cognition.2007.04.001Get rights and content

Abstract

Previous research by Dupoux et al. [Dupoux, E., Pallier, C., Sebastián, N., & Mehler, J. (1997). A destressing “deafness” in French? Journal of Memory Language 36, 406–421; Dupoux, E., Peperkamp, S., & Sebastián-Gallés (2001). A robust method to study stress’ deafness. Journal of the Acoustical Society of America 110, 1608–1618.] found that French speakers, as opposed to Spanish ones, are impaired in discrimination tasks with stimuli that vary only in the position of stress. However, what was called stress ‘deafness’ was only found in tasks that used high phonetic variability and memory load, not in cognitively less demanding tasks such as single token AX discrimination. This raised the possibility that instead of a perceptual problem, monolingual French speakers might simply lack a metalinguistic representation of contrastive stress, which would impair them in memory tasks. We examined a sample of 39 native speakers of French who underwent formal teaching of Spanish after age 10, and varied in degree of practice in this language. Using a sequence recall task, we observed in all our groups of late learners of Spanish the same impairment in short-term memory encoding of stress contrasts that was previously found in French monolinguals. Furthermore, using a speeded lexical decision task with word–nonword minimal pairs that differ only in the position of stress, we found that all late learners had much difficulty in the use of stress to access the lexicon. Our results show that stress ‘deafness’ is better interpreted as a lasting processing problem resulting from the impossibility for French speakers to encode contrastive stress in their phonological representations. This affects their memory encoding as well as their lexical access in on-line tasks. The generality of such a persistent suprasegmental ‘deafness’ is discussed in relation to current findings and models on the perception of non-native phonological contrasts.

Introduction

After spending years in a foreign country, adults continue to experience important difficulties in dealing with several aspects of the foreign language. These difficulties are especially salient in the area of phonology, both in production (Flege et al., 1997, Piske et al., 2002) and in perception (MacKay, Meador, & Flege, 2001). Concerning difficulties in perception, they have been linked to the interference from the native-language phonology. Early work by Polivanov, 1931, Trubetzkoy, 1939/1969 led to the conclusion that learners of a second language incorrectly apply their native set of phonological contrasts to parse those of the second language. More recently, a number of such interference effects have been documented in the psycholinguistic literature. Perhaps most well-known is the difficulty that native Japanese listeners, who use only a single liquid phoneme, experience in perceiving the English /r-l/ contrast (Goto, 1971, Miyawaki et al., 1981). These difficulties are persistent, as they are found in listeners with several years of exposure to English (Takagi & Mann, 1995). Moreover, Japanese monolinguals can improve their performance on /r-l/ discrimination over several weeks of intensive computerized training, but their performance remains significantly below that of English monolinguals (Lively, Logan, & Pisoni, 1993), and they continue to use different acoustic cues than native listeners (Takagi, 2002). Similar problems have been reported for the perception of vocalic contrasts, like the Catalan /e-ε/ contrast, which is difficult for native speakers of Spanish, (Pallier et al., 1997, Sebastián-Gallés and Soto-Faraco, 1999, Sebastián-Gallés et al., 2005), or the English /a-ʌ/ and /a-æ/ contrasts, which are difficult for native speakers of Italian (Flege and Mackay, 2004, Flege et al., 1999). It is noteworthy, however, that not all non-native contrasts are equally difficult. Best, McRoberts, and Sithole (1988) documented that an acoustically minimal contrast in Zulu clicks can be discriminated very well by English listeners despite the fact that English uses no such contrasts. For Japanese listeners, /b/ vs. /v/ is initially as difficult as /r/ vs. /l/; however, practice yields a large increase in performance in the former, not the latter (Guion, Flege, Akahane-Yamada, & Pruitt, 2000).

Several theoretical approaches have been proposed to account for these patterns of data. In some models, the acoustic perceptual space is ‘warped’ according to experience. This means that the weighting of certain acoustic dimensions or cues (the slope of certain formants, fine-grained timing properties, etc) is enhanced or reduced, depending on their functional values in the language (Francis and Nusbaum, 2002, Iverson et al., 2003, Jusczyk, 1997, Nosofsky, 1986). In these models, non-functional perceptual dimensions will yield poor perceptual discrimination. Other models are based on similar proposals, but rely on abstract phonetic (Flege, 1995) or even phonological features (Lado, 1957, Brown, 1998, Brown, 2000). In such models, when a non-native contrast cannot be parsed in terms of a contrastive feature of the native language, perceptual problems arise. Finally, a third class of models proposes that perceptual difficulties are due to the interference from native phoneme categories or prototypes. Phoneme categories reduce the multidimensional and detailed representations of a stimulus to a discrete linguistic label, plus a goodness rating indicating how far the stimulus is from the center of the category (Grieser and Kuhl, 1989, Best, 1994). Best (1994) proposes that upon exposure to a speech sound, the perceptual system automatically assimilates it to the closest category. If two sounds are assimilated to the same category, they will be very difficult to discriminate, unless they differ greatly in category goodness. Stimuli that are too distinct from any existing category will not be assimilated, and hence, their acoustic details remain available, yielding good discrimination (Flege, 1995).

Considerable effort has been devoted to explicitly distinguishing these models in the realm of vowel and consonant perception, generally supporting models based on perceptual warping and prototype formation (Guion et al., 2000, Kuhl, 2000, Best et al., 2001, Kingston, 2003). Less research has been directed towards the perception of suprasegmentals. Suprasegmentals are interesting, because they involve different types of acoustic cues. Indeed, whereas segments involve fast and fine-grained spectral changes, suprasegmentals involve slower and more global acoustic cues like F0, energy and duration (Lehiste, 1970). These cues are, moreover, also used to carry information regarding prosodic constituent structure, pragmatic content, emotional state, etc., adding to the variability in their distributions. The description in terms of discrete categories with a prototype structure may therefore be less pertinent for suprasegmentals than for segments. Hence, it is not clear that the above perceptual effects for segmental contrasts would also apply to suprasegmental contrasts.

Yet, recent research has pointed to interesting parallels between the perception of segmental and suprasegmental contrasts, in particular ones concerning tones. First, it has been found that native speakers of tone languages perceive their native tonal contrasts in a more categorical fashion than non-native speakers (Hallé et al., 2004, Xu et al., 2006). In addition, they process these contrasts in the left hemisphere, whereas speakers of non-tone languages process them either in the right hemisphere or bilaterally (Gandour et al., 1998, Gandour et al., 2000, Klein et al., 2001). Furthermore, native speakers of English who learn Mandarin have much difficulty in learning the tones of this language, and rely on slightly different acoustic cues than native speakers (Stagray & Downs, 1993). As in the case of segments, not all tonal contrasts are equally difficult. This finding has been linked to the interference from existing suprasegmental contrasts in the native language, like stress for English learners of Mandarin (Kiriloff, 1969, White, 1981, Shen, 1989, Chen, 1997). Finally, as in the case of /r/ and /l/, English adults can be trained to increase their performance on tone perception, with performance reaching a suboptimal plateau (from 66% correct pre-training to 87% post-training, Wang, Spence, Jongman, & Sereno, 1999).

In this paper, we examine the perception of a stress contrast by French listeners. Stress is instantiated by three acoustic cues: F0, duration and energy, none of which is used in French to signal a phonological contrast. Indeed, French has no contrastive tone, pitch accent or stress, and duration is used contrastively in neither vowels nor consonants. Duration is only used allophonically, with vowels being lengthened before certain consonants (Casagrande, 1984). French is described as having a phrasal ‘accent’, which is realized as final syllable lengthening in prosodic groups with no increase in F0 or intensity (Rossi, 1980, Vaissière, 1991); this phrasal accent thus has a demarcative, not a contrastive function. Hence, a contrast between, for instance, Spanish /bébe/ ‘(s/he) drinks’ and /bebé/ ‘baby’ cannot be represented by an existing contrastive phonological or phonetic feature in French.

Depending on the theoretical model, the lack of contrastive use of suprasegmentals in French yields two divergent predictions. According to a model where the existing inventory of linguistic features is determinant, French participants should have much difficulty in representing and acquiring contrastive stress. This would parallel the finding that native speakers of Spanish have more difficulties than native speakers of English to learn Swedish vowel duration contrasts (McAllister, Flege, & Piske, 2002); indeed, in Spanish, vowel duration is not used at all, whereas in English, it is a secondary cue to the contrast between tense and lax vowels. In contrast, according to a model where the limiting factor is the similarity to an existing prototype, French listeners should find it very easy to acquire contrastive stress. This would parallel findings that American listeners can distinguish Zulu clicks quite easily (Best et al., 1988), and that French listeners have some strong abilities to distinguish Mandarin tones, even though they behave in a less categorical fashion (Hallé et al., 2004).

So far, the data accumulated on the processing of contrastive stress by French listeners supports both types of models (Dupoux et al., 1997, Dupoux et al., 2001). Indeed, when French participants are tested with a standard AX discrimination task on a stress contrast, they do not differ from Spanish monolinguals. Yet, if memory load and/or talker variability is introduced, French participants display a very strong impairment to perceive stress contrasts (see Table 1).

Prima facie, this pattern of data can be interpreted as showing that French listeners have a true impairment concerning the processing of contrastive stress. According to this interpretation, their poor performance in the high variability experiments shows that they lack a proper phonological representation of stress that they could use to encode stress contrasts in an on-line fashion. It is only in the low variability tasks that they have enough time and resources to discriminate the stress contrast, using an ad hoc acoustic strategy. However, there is an alternative explanation which needs to be addressed. Perhaps French listeners are not impaired in their phonological representation of contrastive stress, but rather have difficulty with the metalinguistic access to this contrast. Note that the short-term memory task requires participants to explicitly associate two stimulus classes to two response keys. In order for this association to be learned efficiently, metalinguistic access to these two classes is required. Yet, monolingual French listeners are likely not to have a metalinguistic representation of Spanish stress, contrary to Spanish, contrastive stress does not exist in French, stress is not marked in the orthography, and it is not taught in school. Would late learners, who surely have metalinguistic knowledge of Spanish stress, show the same limitation as French monolinguals in their processing of contrastive stress?

The aim of this paper is thus to examine whether the performance pattern of French listeners is due to a true processing deficit or to metalinguistic limitations. We test French learners of Spanish, who have learned about contrastive stress as part of their explicit L2 training. If the problem is a true processing limitation, beginning learners of Spanish should not differ from monolinguals, and only more advanced learners should start to show an enhancement in performance. If, in contrast, the problem is metalinguistic, even a limited exposure to Spanish should enable the French participants to use the stress labels and hence enhance their performance Furthermore, in order to evaluate whether stress ‘deafness’ is limited to tasks requiring metalinguistic access to stress categories, we contrast two tasks: the first one is the sequence recall task with high phonetic variability of Dupoux et al. (2001), the second one is a lexical decision task using word–nonword minimal pairs that vary only in stress position. Again, if French listeners have a real processing deficit as far as stress is concerned, then French learners of Spanish should have difficulty rejecting nonwords that differ from a real word with respect to the position of stress only. By contrast, if our previous findings with French monolinguals are due to metalinguistic limitations, then French learners of Spanish should have no difficulty in encoding stress in lexical items, and hence correctly reject such nonwords.

Section snippets

Sample

Thirty-nine native speakers of French who were late learners of Spanish participated in this study. There were 11 men and 28 women, aged between 20 and 57 (mean = 28). Six late learners lived and were tested in Paris; they had been recruited from among university students in a Spanish language department. The remaining 33 had lived in Barcelona for at least 6 months and were tested there; they had been recruited from among undergraduate students who spent a year on an exchange program. Only four

Self-evaluation

All late learners rated their Spanish competence in several domains, i.e. pronunciation, grammar, and vocabulary, on a scale from one to ten, as well as the importance of Spanish in their own lives (Table 3).

These data were subjected to a series of analyses of variance (ANOVA) with the between-participant factor Practice (Beginner vs. Intermediate vs. Advanced) and the dependent variables Pronunciation, Grammar, Vocabulary, and Importance, respectively. For Pronunciation there was a significant

Experiment 1: Stress encoding in short-term memory

In this experiment, we use a modified version of the high phonetic variability sequence recall task of Dupoux et al. (2001). This task is meant to make the acoustic level of representation not accessible and hence to highlight the phonological level. The present version of the task differs from the one used by Dupoux et al. (2001) in three respects. First, it uses sequences of length four only, thus avoiding both floor and ceiling effects. Second, the stimuli are produced by several talkers.

Experiment 2: Stress encoding in the lexicon

In order to explore whether the difficulty with the perception of stress is not limited to short-term memory tasks, but extends to the use of stress for on-line lexical access, we examined the use of stress in lexical representations by means of a lexical decision task. In this experiment, we presented real Spanish words, as well as nonwords obtained from these words by changing the location of stress. For instance, górro is a word (‘hat’), but gorró is not. Participants had to make a speeded

General discussion

The main finding of this study is that French late learners of Spanish have much difficulty in the perception of stress, as assessed in a sequence recall task and a lexical decision task. In the sequence recall experiment, the late learners were undistinguishable from the French monolingual controls and they did not show an effect of practice: those learners who on average had spent three years in a Spanish-speaking country performed as badly as beginners with only a few months of practice with

References (58)

  • N. Sebastián-Gallés et al.

    Online processing of native and non-native phonemic contrasts in early bilinguals

    Cognition

    (1999)
  • C.T. Best

    The emergence of native-language phonological influence in infants: A perceptual assimilation model

  • C.T. Best et al.

    Discrimination of non-native consonant contrasts varying in perceptual assimilation to the listeners’ native phonological system

    Journal of the Acoustical Society of America

    (2001)
  • C.T. Best et al.

    Examination of perceptual reorganization for nonnative speech contrasts: Zulu click discrimination by English-speaking adults and infants

    Journal of Experimental Psychology: Human Perception and Performance

    (1988)
  • C. Brown

    The role of the L1 grammar in the L2 acquisition of segmental structure

    Second Language Research

    (1998)
  • C. Brown

    The interrelation between speech perception and phonological acquisition from infant to adult

  • J. Casagrande

    The sound system of french

    (1984)
  • Q. Chen

    Toward a sequential approach for tonal error analysis

    Journal of the Chinese Language Teachers Association

    (1997)
  • E. Dupoux et al.

    Epenthetic vowels in Japanese: A perceptual illusion?

    Journal of Experimental Psychology: Human Perception and Performance

    (1999)
  • E. Dupoux et al.

    A robust method to study stress ’deafness’

    Journal of the Acoustical Society of America

    (2001)
  • J.E. Flege

    Second-language speech learning: Theory, findings, and problems

  • J.E. Flege et al.

    Perceiving vowels in a second language

    Studies in Second Language Acquisition

    (2004)
  • J.E. Flege et al.

    Native Italian speakers’ perception and production of English vowels

    Journal of the Acoustical Society of America

    (1999)
  • J.E. Flege et al.

    Cross-language phonetic interference: Arabic to English

    Language and Speech

    (1981)
  • A.L. Francis et al.

    Selective attention and the acquisition of new phonetic categories

    Journal of Experiment Psychology: Human Perception and Performance

    (2002)
  • J. Gandour et al.

    A cross-linguistic PET study of tone perception

    Journal of Cognitive Neuroscience

    (2000)
  • J. Gandour et al.

    Pitch processing in the human brain is influenced by language experience

    Neuroreport

    (1998)
  • D. Grieser et al.

    Categorization of speech by infants: Support for speech-sound prototypes

    Developmental Psychology

    (1989)
  • S.G. Guion et al.

    An investigation of current models of second language speech perception: The case of Japanese adults’ perception of English consonants

    Journal of the Acoustical Society of America

    (2000)
  • Cited by (0)

    Research for this paper was funded by grants from the Centre National de la Recherche Scientifique (Aide à Projet Nouveau, and Cognition et Traitement de l’Information, CTI02-15), the Groupement d’Intérêt Scientifique Sciences de la Cognition (99N35/0008), l’Agence Nationale pour la Recherche (ANR-05-BLAN-0065-01), the McDonnell Foundation (Bridging Mind, Brain and Behavior Program, JSMF 20002079), the Spanish Ministerio de Educacion y Ciencia (Grant Contract SEJ2004-07680-C02-01) and the Catalan government (2005SGR-01026). We are grateful to Katherine White for help in running the participants in Paris, and to Anne Christophe and Jacques Mehler for comments and discussion.

    View full text