Elsevier

Journal of Phonetics

Volume 44, May 2014, Pages 8-24
Journal of Phonetics

Laryngeal–oral coordination in mixed-voicing clusters

https://doi.org/10.1016/j.wocn.2014.02.004Get rights and content

Highlights

  • We study laryngeal–oral coordination in German mixed-voicing obstruent+sonorant onset clusters.

  • Fiberoptic transillumination of the glottis was used.

  • Depending on the sonorant the devoicing gesture may be strengthened compared to singletons.

  • The devoicing gesture is planned to meet the demands of the syllable onset as a whole.

Abstract

Laryngeal–oral coordination was studied in German clusters of voiceless fricative or plosive plus /l/ or /r/ by means of videofiberendoscopy and transillumination. In all cases voice onset time (i.e. the time from release of C1 to onset of voicing) was longer in the clusters compared to the single fricative or plosive controls. However, the coordination patterns leading to this consistent acoustic effect were quite varied, ranging from a passive effect of aerodynamic conditions at release of C1, via shortening of C1 with constant glottal gesture, to enhancement of the glottal gesture. Active reorganization was particularly clear in the rhotic clusters. For the single consonants the duration of the glottal gesture was quite constant over place of articulation but occlusion duration varied systematically. Accordingly, for both clusters and singletons peak glottal opening did not keep a constant timing relationship to landmarks in the oral occlusion of C1. The above findings were robustly present over a range of prosodic conditions. Prosodic strengthening itself had a particularly clear influence on the magnitude of the devoicing gesture.

Introduction

This paper will be concerned with laryngeal–oral coordination in syllable-initial clusters of German consisting of a voiceless element (plosive or fricative), followed by a sonorant (/l/ or /r/). Our basic contention is that such sound sequences can be potentially very useful for highlighting the gaps in our knowledge about the principles underlying interarticulatory coordination (just as much as the more frequently investigated purely voiceless clusters; e.g. Löfqvist and Yoshioka, 1980, Ridouane et al., 2006, Yoshioka et al., 1981). Let us assume as the most basic hypothesis that in both a singleton onset such as /p/ as well as in a cluster onset such as /pl/ the devoicing gesture is organized with respect to the underlyingly voiceless segment /p/. The addition of /l/ to the onset should not then have any effect on the timing of landmarks in the glottal gesture relative to those in the oral gesture of the /p/. Such a scenario would account perfectly well for the well-known substantial devoicing of, for example, /l/ in /pl/ and /r/ in /pr/ in languages such as English and German as a simple coarticulatory process, given that the laryngeal–oral timing pattern for syllable-initial /p/ involves peak glottal opening located close to the release of the oral occlusion.1 As we will see below, this basic scenario is actually not well supported by currently available findings. There is even some support for radically different patterns, for example that addition of an underlyingly voiced sonorant to an otherwise voiceless syllable onset can lead to an increase in the magnitude of the glottal gesture. Thus, the fundamental motivation for our own investigation is that we are currently ignorant about how the laryngeal–oral coordination relations should be formulated, and that mixed-voicing syllable onsets have interesting implications for the level of the syllabic hierarchy at which the devoicing gesture is organized.

In the following paragraphs we will review earlier work, aiming to identify potential patterns of laryngeal–oral coordination, summarizing them schematically in Fig. 1 to provide a framework for further discussion.

A convenient point of departure for consideration of previous findings is Docherty's (1992) acoustic investigation of English. Based on his own results, and other relevant results available at that time, two fairly pervasive generalizations can be identified:

  • 1.

    VOT (i.e. the period of voicelessness following release of the stop or fricative) is longer in /Cl/ and /Cr/ sequences than in simple CV sequences.

  • 2.

    It is well documented that stops and fricatives generally have a shorter occlusion duration when they occur in clusters (e.g. Haggard, 1973, Hawkins, 1979, Klatt, 1975, Klatt, 1973), though in fact we will also be encountering some cases where this effect is quite weak.

What might explain the increased VOT in sequences such as /pl/ compared to singleton /p/? One possible scenario is that it is a simple link between the above two generalizations: /p/ and /pl/ may have a glottal gesture of similar duration, but a shorter occlusion duration in cluster /pl/ (see scenario “Short C1” in Fig. 1). In Hoole and Bombien (2010) we have suggested that an organizational principle of this kind may be in operation for, in effect, the mirror-image case in Icelandic: pre-aspirated plosives, on the one hand, and voiceless nasal plus plosive sequences, on the other hand, have a glottal gesture of similar duration but shorter plosive occlusions in the latter case. In fact, Docherty's data did not give much support for this scenario, since there were cases where VOT increased more (in the clusters) than would be predicted by the reduction in occlusion duration (i.e. for some clusters the reduction was negligible). In effect, for some clusters, the total duration of voicelessness was longer than in the singletons, indicating that the duration of the glottal devoicing gesture might be longer in the clusters (scenario “long glottal gesture” in Fig. 1).2

This is a striking result, because probably no coarticulation theory would predict that adding an underlyingly voiced sound to a syllable onset could result in an increase in devoicing. There may well be one simple passive explanation for the unexpected fact that the stop-continuant cluster has a longer period of devoicing than the simple stop, namely that the aerodynamic conditions in the continuant are not conducive to initiation of phonation, due to the fact that the oral tract is still partially occluded ( see Hanson & Stevens, 2002, especially p. 1175 ff). Recent work of our own using articulatory synthesis (Hoole, Pouplier, Beňuš, & Bombien, 2013) confirms the potential relevance of partial constrictions for aerodynamic conditions: In simulations of /br/ the re-initiation of voicing after devoiced /b/ was delayed if strong dorsal constriction for /r/ was already present at release of /b/. A schematic illustration of how the presence of an aerodynamic effect would be reflected in our measurements is given by scenario “aerodynamics (+short C1)” in Fig. 1. This simply uses the oral and glottal (abduction and adduction) timing patterns from the “short C1” scenario and shifts the right edge of the VOT phase (i.e. the location of voice onset) to a later point in time. Comparison of “short C1” and “aerodynamics (+short C1)” also illustrates the important point that depending on the strength of the influence of aerodynamic conditions there can be a mismatch between the acoustically measured duration of voicelessness and the articulatorily measured duration of the laryngeal abduction and adduction movement. Since in the present study we measure the laryngeal movements themselves, it is possible to weigh up the contributions to any increase in duration of voicelessness more directly than would be possible in a purely acoustic study. Specifically, for the baseline singleton case in Fig. 1 (an aspirated plosive) we assume that at the time when glottal adduction is approaching completion no aerodynamically relevant constriction of the vocal tract remains. Accordingly, the precise time point of voice onset is directly determined by the glottal configuration itself: in Fig. 1 we show a situation that is typical in our recordings, namely that the acoustically determined voice onset occurs slightly before glottal adduction is complete (as measured from the transillumination signal). In the cluster scenario “short C1” the position of voice onset relative to the end of glottal adduction is the same as in the singleton case. If such a pattern occurs in the empirical data below we would take this as evidence that any constriction related to C2 has no aerodynamically relevant effects. In the “aerodynamics (+short C1)” scenario the voice onset shifts to a later time relative to completion of glottal adduction, leading to the opposite conclusion.3

Based on the results of previous investigations we assume that it is simply an open issue whether the longer period of voicelessness in clusters compared to singletons reflects a longer glottal abduction–adduction cycle in clusters or, rather, reflects aerodynamically less favorable conditions for voicing in clusters (or some combination of these two mechanisms; further discussion in Docherty (1992, pp.148–149)). But given that lengthening of the glottal gesture itself is a realistic possibility (confirmed by instrumental results we discuss below), it is interesting to consider what could drive speakers to adopt this pattern of movement.

An output oriented style of explanation for lengthening of the glottal gesture might be that it is perceptually important to have a substantial amount of devoicing on the second element in a cluster when the first element is underlyingly voiceless (e.g. to separate “played”, “blade”, “prayed”, “braid”). In other words, languages such as English and German exhibit a large number of contrasting combinations of C1 and C2 in syllable-onset position in which the listener needs to be able to recover the voicing status and place of articulation of C1 as well as the manner of articulation of C2. Since a low onset frequency of voiced F1 is known to support voicing percepts it may be advantageous to delay voice onset in voiceless clusters to reduce the salience of low F1 due to the constriction of C2. Perhaps also it is easier for listeners to distinguish voiceless (devoiced) sonorants if their duration is long.

A further possibility is that given the aerodynamic conditions in the vocal tract, completion of glottal adduction early in C2 might not lead to reliable re-initiation of voicing anyway, so speakers reorganize the gesture to be completed closer to the time at which aerodynamic conditions are favorable, potentially reinforcing the total duration of voicelessness (we have argued elsewhere (Hoole, Pouplier, & Kühnert, 2012) that speech movements are often organized to exploit the physical conditions in the vocal tract: to ‘go with the flow’).

As already emphasized, empirical verification is required as to whether lengthening of the glottal gesture actually occurs. But previous results do, however, make it seem unlikely that duration of the glottal gesture is shorter in clusters. This is in itself a significant finding since, given a tendency to shorter occlusion durations, a shorter glottal gesture could well be expected under the plausible assumption that the component gestures of an aspirated plosive become modified in parallel (this is what we referred to as the “basic hypothesis” in the first paragraph of Section 1). For example, working within the framework of the Task-Dynamics model, Saltzman and Munhall (1989) point to evidence from perturbation experiments that the laryngeal gesture is modified when the bilabial closure for /p/ is interfered with experimentally. They cite this as evidence for a level of intergestural cohesion that must exist (otherwise speakers would hardly be able to systematically produce VOTs in the range appropriate for their language). We include the scenario in which oral and glottal gestures shorten in parallel as the pattern labeled “constant relative timing” in Fig. 1. The key feature distinguishing this scenario from the other ones shown in the figure is that peak glottal opening (at the boundary between abduction and adduction) stays at the same location relative to the release of the C1 occlusion as it does for the baseline singleton case. Even though previous results have provided little support for this scenario (note also that in the form shown in Fig. 1 it incorrectly predicts a shortening of VOT) it is conceptually important. Indeed, based on the above discussion of aerodynamic effects, a combined scenario can be envisaged in which the glottal and oral kinematics show constant relative timing but voice onset is delayed beyond the end of glottal adduction if constriction conditions in C2 are very unfavorable for voicing (in fact, a delay in voice onset may be more likely in the “constant relative timing” case than, for example, in the “short C1” case because in the former case glottal adduction is completed at a time when the C2 constriction is still fully formed, whereas in the latter case C2 may already have weakened).

Before turning to discussion of currently available instrumental findings one further important element in setting the scene is given by the gestural approach of Browman and Goldstein (1986). Based on their analysis of voiceless clusters Browman and Goldstein (1986) come to the conclusion that it can be stated as a regularity of English that a word (syllable) can only begin with one devoicing gesture (and we assume initially, because of the similar status of aspiration, that German is likely to pattern in a similar way).

  • (1)

    If a fricative gesture is present, coordinate the peak glottal opening with the midpoint of the fricative.

  • (2)

    Otherwise, coordinate the peak glottal opening with the release of the stop gesture.

These two simple rules are very attractive because they already account for important regularities in English such as short VOT in a plosive following a fricative, and devoicing of sonorants in clusters of concern here like /pl/ (see also Iverson & Salmons, 1995; Kehrein & Golston, 20044).

The rules thus form a useful framework in the present context. But the scenarios we have presented above indicate that departures from these coordination patterns may be observed. Thus one aim of the present work is to determine whether these rules need to be supplemented by statements taking the segmental context into account (we come back to the idea of context-sensitive implementation of the devoicing gesture again below, e.g. in discussion of the effect of place of articulation in single consonants).

Turning to investigations that have looked directly at laryngeal behavior in the relevant clusters (i.e. as in our own investigation by means of laryngeal fiberscopy and transillumination), Jessen (1999) considers explicitly the question of whether duration of voicelessness in the sonorant is directly related to shortening of the plosive occlusion. The material came from one speaker of German and consisted of /p, t, k/ combined (to the extent possible in German) with /l, r, n/. The results were somewhat mixed: The clusters with initial /k/ (/kl/, /kn/) were overall reasonably consistent with an account relating VOT duration to the shortening of the plosive occlusion duration, while the p-initial cluster /pl/ showed the expected increase in VOT but with only negligible shortening of the occlusion duration. A rather striking finding in this investigation involved stop-/r/ clusters. All the combinations (/pr, tr, kr/) showed a substantial increase in duration of the glottal gesture. In addition, Jessen (1997) reported on three-element clusters: Those with /r/ as the last element (e.g. /ʃpr/) were the only ones to show a double-peaked abductory movement in the transillumination signal. In the discussion section we will consider in more detail what might contribute to the emergence of the particularly extensive glottal abduction in /r/-final clusters. For the moment we simply note that the net result is that /r/ is realized as a voiceless uvular fricative, and voiceless fricatives are, of course, known to require strong glottal abduction. In any case it is interesting that a contrast such as /pr/ vs. /br/, that underlyingly only appears to involve the first element in the onset, is actually planned in terms of a pattern of activity that involves the whole onset.

Tsuchida and Cohn (2000) examined clusters with /l/ for one speaker of American English. For plosives, they found the expected slight shortening of the oral occlusion (for cluster vs. singleton), but the glottal gesture was actually somewhat longer in the clusters. By contrast, in the comparison of single fricatives and fricative-initial clusters, the clusters showed a reduction in duration of the glottal gesture that roughly paralleled the reduction of the fricative oral occlusion duration. Thus, in terms of the scenarios introduced above we have a first indication that plosive-initial and fricative-initial clusters need not necessarily behave in the same way: fricative-initial clusters following “constant relative timing”, plosive-initials showing a combination of “short C1” and “long glottal gesture”.

Hoole (2006, some preliminary results of this study also in Hoole, Fuchs, and Dahlmeier, 2003) examined clusters with /l/ for three speakers of German. The only plosive-initial cluster recorded was /pl/. On the one hand there was a consistent finding of the expected longer VOT in the cluster compared to the singleton /p/, and the alignment of the glottal gesture with the oral gesture for /p/ was different in the cluster compared to the singleton. However, it was not possible to identify a consistent change in the coordination pattern over the three subjects leading to the longer VOT. This was tentatively seen as an indication that lengthened VOT in the cluster condition might indeed be the key output constraint guiding the speakers' articulatory behavior; this constraint could be fulfilled by quite a variety of coordination strategies: reducing the occlusion duration or lengthening the glottal gesture are perhaps just the two most obvious ones; there was even an indication that keeping glottal gesture duration constant but delaying its onset might be a further possibility (illustrated as the “late glottal gesture” scenario in Fig. 1). In this study two fricative+/l/ clusters were recorded (/fl/ and /ʃl/). There was once again a consistent finding that the alignment of the glottal gesture with the oral gesture of C1 changed when going from the singletons to the clusters. Reduction of the duration of C1 occlusion seemed to make a more substantial contribution to the shift in coordination than was the case for the /pl/ cluster.

The small number of subjects in the studies just discussed would itself be sufficient motivation for the present study. But this is all the more so because on the basis of previous results not even the simplest statements about laryngeal–oral coordination in these still relatively straightforward sound sequences can be made with any confidence: in other words, glottal gesture durations in the cluster that are shorter, longer or the same as in the corresponding singleton are all realistic possibilities. For the particular case of /pl/ in our earlier study we advanced an output-oriented account that might rationalize apparently heterogeneous articulatory strategies, but three subjects is clearly insufficient to be confident that this heterogeneity really exists; conceivably, with more subjects a clearly preferred strategy might emerge. A further specific area where more subjects would be highly desirable relates to the striking findings of Jessen for clusters with /r/: this raises the possibility that superficially rather similar syllable onsets such as /pr/, /pl/ might require differing laryngeal specifications. This is particularly intriguing because both would be regarded on paper as ‘good’ (unproblematic) onsets in terms of preferred sonority profiles, while an active realization of /r/ as a voiceless fricative would result in a much less preferred sonority profile for the syllable onset.

Although the main focus of the present investigation is on laryngeal–oral coordination in clusters we will also lead in to the cluster results by looking at coordination in single consonants. The motivation for this is that in particular the relationship between place of articulation of plosives and VOT duration potentially involves similar scenarios to those introduced above for clusters. Specifically, it has been a common finding that VOT is shorter in /p/ than /k/, while occlusion duration is typically shorter in the latter. Thus, comparably to the “short C1” scenario shown for singleton vs. cluster in Fig. 1 it is conceivable that glottal gesture duration remains the same, with the increased VOT being a direct consequence of the shorter occlusion. All the studies quoted above present data relevant to this issue (see in particular Jessen (1999), for a very clear exposition of the possible scenarios; see also Cooper, 1991; Hoole, Pompino-Marschall, & Dames, 1984; Hutters, 1984). Broadly speaking, these studies indicate that shortening of the occlusion duration is certainly a significant element in the lengthening of VOT, but on the other hand the duration of the glottal gesture does not necessarily remain completely constant over all places of articulation. In their wide-ranging cross-language study of VOT, Cho and Ladefoged (1999) suggest that, just as the specification of target values of laryngeal–oral timing for a category such as voiceless aspirated must be part of the language's grammar (because across languages different choices are made from the range of possible values), so, too, it may also be necessary to include context-specific rules referring to place of articulation.

In any case, by incorporating single consonants into our discussion we extend the range of material on which to discuss the extent to which speakers try to get as much mileage as possible out of a rather constant glottal gesture – even if the duration is not fixed in a hard and fast sense – with oral gestures showing a shifting pattern relative to this laryngeal substrate. Here it is worth recalling Shipp's (1982) suggestion that the highly preprogrammed nature of the abductory–adductory cycle may make the larynx ‘one of the basic metronomes of the speech production process’ (p. 111; see also Weismer, 1980).

For several reasons it was considered interesting to investigate the organization of the laryngeal devoicing gesture over a range of prosodic conditions in which the target sounds occurred in contexts differing with respect to prosodic strength (achieved, for example, by manipulating phrasal accent position; further details below). Firstly, prosodic manipulation can be regarded as a probe to test the robustness of any differences between singletons and clusters, or between different clusters. For example, if it did indeed turn out that clusters showed enhancement of the glottal gesture compared to singletons then we wanted to be able to test whether the effect only emerged in prosodically strong contexts. Secondly, we were interested in prosodic strengthening as a phenomenon in its own right. There has been considerable interest in recent years in investigating how prosodic structure is signaled not just in contrasting intonation patterns but also in segmental changes. A fair amount of knowledge has now also accumulated on the articulatory substrate of such segmental changes (e.g. using electropalatography or electromagnetic articulography; Bombien et al., 2010, Bombien et al., 2013, Byrd and Choi, 2010, Keating et al., 2003). Of particular interest in the present case is the fact that, at least for English, VOT is often found to be longer in prosodically strong contexts (see e.g. Cho & McQueen, 2005, for discussion). In articulatory terms this suggests that the laryngeal gesture is enhanced in such cases, but detailed information is currently lacking. There has been a fair amount of analysis of word-stress effects using transillumination (e.g. Cooper, 1991, Fuchs, 2005, Hoole et al., 1984, Löfqvist and McGarr, 1987), giving a consistent picture of more glottal abduction in more strongly stressed syllables, but to our knowledge nothing is known about whether such findings extend to further sources of prosodic strengthening. We will focus, in particular, on the influence of contrastive focus, since this was the condition that was easiest to embed in a corpus suitable for transillumination. But we also included a condition designed to provide information on domain-initial strengthening because of the equally well-documented relevance of prosodic boundaries for articulatory strengthening processes (e.g. Cho and McQueen, 2005, Keating et al., 2003). Even if the expectation is indeed that the English findings will extend to German, and that word-stress findings will extend to phrasal stress and to boundary-induced strengthening, there still remains much scope for clarification, i.e. does the glottal gesture simply lengthen with longer VOT, or does it also increase in magnitude? Is it perhaps simply timed later, rather than lengthened? Does gestural enhancement apply equally to fricatives and plosives, even though enhancement of VOT is not relevant in the former case? The third and final reason for including prosodic variation in the experimental design was motivated by the fact that the present work forms part of a larger study investigating laryngeal articulation in languages differing in voicing typology (German, Dutch, French); we expect that the effect of prosodic strengthening should reflect different underlying organization of the voicing contrast in these languages. This will be the subject of a future report.

Section snippets

Experimental setup

Five native speakers of German (2 female, age 26 and 28; 3 male, age 35, 35, and 26), one of them the second author, were recorded by means of photoglottographic transillumination (PGG). For this purpose a flexible endoscope (Olympus ENF-P3) connected to a Rehder rpCam-X light source and video camera was inserted through the nose into the pharyngeal cavity by a physician such that the tip of the endoscope was positioned above the glottis. After positioning, the endoscope was mounted on an

Results (1) – single stops

We will start the presentation of the results by looking briefly at timing patterns with respect to place of articulation in stops. It turns out that one of the coordination scenarios we outlined above as potentially applicable to both place of articulation in stops, as well as to the singleton-cluster comparison can indeed be identified in the place of articulation analysis: specifically, differences in VOT linked to differences in occlusion duration while glottal gesture duration varies very

Results (2) – full material

In this section we will first look one by one at the detailed results for each of the parameters outlined in Section 2, and then draw the threads together by summarizing the patterns for each of the main cluster categories (fricative+/r/, fricative+/l/, plosive+/r/, plosive+/l/) relative to the corresponding singletons. As pointed out in Section 2.4, the main test variables in this section are Manner and Type and their interaction. The factor Prosody was already introduced in the previous

Discussion

In this discussion section we will first pull together the results related to prosodic conditions (as presented in Section 3.2 for single stops and just summarized in Section 4.9.5 for the full material including clusters and fricatives) and then move on to the overall more complex pattern of results that emerged from the comparison of different syllable onsets.

At a very basic level, the clearly greater glottal opening in the strong prosodic conditions appears to indicate that this study

Acknowledgments

This work was supported by the German Research Council (Grant HO 3271/3-1 to Philip Hoole). We express our sincere gratitude to Bodo Winter for valuable advice with the statistic analyses in this paper. The second author thanks Louis Goldstein and Khalil Iskarous for kindly providing a working environment in the Linguistics Department at USC, Los Angeles. To René Bombien for being the physician conducting the endoscopy: Mercy buckets! Thanks to Elizabeth Heller for help with segmentation and

References (43)

  • G. Weismer

    Control of the voicing distinction for intervocalic stops and frictives: Some data and theoretical considerations

    Journal of Phonetics

    (1980)
  • C. Browman et al.

    Towards an articulatory phonology

    Phonology Yearbook

    (1986)
  • D. Byrd et al.

    At the juncture of prosody, phonology, and phonetics — The interaction of phrasal and syllable structure in shaping the timing of consonant gestures

  • Cooper, A. M. (1991). Laryngeal and oral gestures in English /p, t, k/. In Proceedings of the XIIth international...
  • G.J. Docherty

    The timing of voicing in British english obstruents

    (1992)
  • S. Fuchs

    Articulatory correlates of the voicing contrast in alveolar obstruent production in German

    ZAS Papers in Linguistics

    (2005)
  • L. Goldstein

    On articulatory binding: Comments on Kingston's paper

  • H.M. Hanson et al.

    A quasiarticulatory approach to controlling acoustic source parameters in a Klatt-type formant synthesizer using HLsyn

    Journal of the Acoustical Society of America

    (2002)
  • H. Hirose et al.

    The relationship between glottal opening and the transglottal pressure differences during consonant production

  • P. Hoole

    Techniques for investigating laryngeal articulation section A: Investigation of the devoicing gesture

  • P. Hoole

    Experimental studies of laryngeal articulation – Part II: Laryngeal–oral coordination in consonant sequences (unpublished habilitation thesis)

    (2006)
  • Cited by (10)

    • Temporal variability in sung productions of adolescents who stutter

      2016, Journal of Communication Disorders
      Citation Excerpt :

      Comparable to VOT reduction, a reduction in stop gap duration from speech to song was found (F (1,14) = 26.65, p < 0.001). In addition, stop gap duration varied with stop category (/p/>/t/>/k/; F (2,28) = 49.59, p < 0.001), a common finding in phonetic analyses of stop gap (Hoole & Bombien, 2014). Stop gap variability (CV) was not different for song and speech.

    • How much is a word? Predicting ease of articulation planning from apraxic speech error patterns

      2015, Cortex
      Citation Excerpt :

      According to our analyses, the probability that all gestures in a cluster are accurate at the same time did not fall short of the product of individual probabilities of accurate gesture production, meaning that the complexity effect in the present data is explainable by the trivial circumstance that clusters offer more opportunities to fail than single consonants. Hence, there was no specific effect of increased timing requirements in consonant clusters, as discussed, for instance, by Hoole and Bombien (2014) for healthy speakers. Another result which may appear counter-intuitive at first sight is that synchronous velar and glottal aperture gestures did not turn out disproportionately taxing, – in the case of the glottal gestures there was even a strong facilitation of aperture gestures preceding or following the closed glottis configuration of the vowel nucleus.

    • Does voicing affect patterns of transfer in nonnative cluster learning?

      2021, Journal of Speech, Language, and Hearing Research
    View all citing articles on Scopus
    View full text