Alveolar Tap Research Articles

Speech rate influences categorical variation of English flaps and taps during normal speech

This research demonstrates that North American English speakers with faster syllable iterance rates are more likely to produce reduced /t/ and /d/ as taps, which have two directions of motion, than flaps, which have one. Using B/M mode ultrasound to capture tongue motion direction, coupled acoustic recordings to capture syllable speed, we identified categorical variants of flaps and taps produced by 18 speakers and compared the average likelihood of each variant in relation to the rate at which speakers can repeat `ta'. Faster iterators produce more alveolar taps in single flap phrases (e.g. `autumn), and double flap phrases (e.g. `edit a'). Alveolar taps require more changes in direction of tongue motion, and are often produced to avoid articulatory conflicts with neighbouring vowels. The results support an argument that individual variation in flap/tap production is partly due to a motor skill constraint limiting cycles of tongue motion, weighed against avoiding articulatory conflict. Currently a second...

Articulatory Variation of the Alveolar Tap and Implications for Sound Change

UC Berkeley Phonology Lab Annual Report (2012) Articulatory variation of the alveolar tap and implications for sound change ∗ Chundra Cathcart Introduction As a natural class, the label “rhotic” comprises an extremely heterogeneous set of sounds which often show little in common phonetically (Ladefoged and Maddieson, 1996). In the same way, individual rhotic phonemes in the world’s languages have been shown to exhibit highly variable articulatory behavior. Several phonetic studies have revealed much subphonemic variation in the realization of individual rhotic phonemes, with different articulatory configurations corresponding to one categorical acoustic signature. An old aphorism says that the seeds of sound change can be found in the details of synchronic variation. If this insight is correct, then it follows that among variants of a phone or phoneme, possible seeds of of attested changes should be identifiable. The aim of this paper is to make a tentative connection between rhotic articulatory variability and a particular sound change seen in unrelated language families. Observing lingual ultrasound data, I analyze features of the alveolar tap that vary within and across speakers. I then propose a provisional link between some articulatory patterns seen in the data and the development of progressive rhotic postalveolarization or retroflexion, a sound change that has come about independently in some dialects of Swedish and Norwegian, in some English dialects, in Sanskrit, in some East Iranian languages, and elsewhere. In an overwhelming number of situations, progressive rhotic retroflexion appears to have been triggered by an alveolar tap, a “non-retroflex” sound; such a development has been deemed “unnatural” in the literature, due to an overly abstract conception of the articulatory and acoustic properties of rhotics such as the alveolar tap. The variation uncovered in this study sheds light on a possible mechanism underlying the development of rhotic ∗ The author is greatly indebted to Grant McGuire and Emily Saltzman of the University of California, Santa Cruz, without whose help recruitment of subjects and collection of data would not have been possible.

Residual /s/ and /r/ distortions: The perspective of the speaker

The purpose of this study was to investigate the impact of residual distortions from the perspective of the speaker. Eighty adult speakers of Brazilian Portuguese with a distortion of the alveolar fricative /s/, the alveolar tap , or both, completed a questionnaire that documented their therapy history, awareness of the speech distortion, and perceived impact of the disorder. Only a small number of the participants reported that they received negative reactions to their speech and only a minority were of the opinion that their speech interfered with work or social life. Yet a considerable number were of the opinion that their speech was different from that of other people and felt insecure in speech situations. The results suggest that there is a high tolerance toward minor speech errors in present day society that may be related to the fact that such disorders are not at all rare.

Subphonemic planning across syllable and word boundaries.

Previous research demonstrated that there are no fixed motor programs or tasks in speech, and there is evidence for subphonemic planning in speech within a word across up to two phoneme boundaries [Derrick and Gick (Submitted)]. Because this evidence is word-internal, it could be suggested that speakers simply memorize many motor programs for each word and draw on them as needed. We demonstrate that speech planning extends across three phonemes, two syllables and a morpheme boundary. Participants produce more up-flaps the first flap of words such as “editor” and “auditor”, which often end with the tongue tip up, versus alveolar taps in “edify” and “audify,” which end with the tongue tip down. We also demonstrate planning across word boundaries. Participants also produce more up-flaps in “edit” and “audit” if these words are followed by “a,” which has a double flap sequence, than if the words are followed by “the,” which does not. Planning across morpheme and word boundaries would ultimately require memorization of an infinite number of motor programs or tasks.

Introduction

Introduction Kimary Shahin and Patricia Shaw Fundamental questions about the relative autonomy of and the nature of the interface between phonetics and phonology continue to stimulate a diverse range of responses. The articles in this special issue on the phonetics-phonology interface illustrate the complexity of the questions that underlie the on-going debates over how distinctive features are defined, whether categorical behaviour is indeed a unique and defining property of phonology as opposed to phonetics, how to account for gradient behaviour within the phonology, what constraints govern gradience and variability, how properties of gestural overlap impact on the notion of phonological unit or segment, where perception fits into the theoretical model in relation to articulatory and acoustic properties of production and phonological properties of mental representations, among others. The research reported on here draws on a diversity of analytical approaches and methodologies. Some approach the issues from the phonetics "upwards", as in the ultrasound studies of Derrick and Gick and the acoustic analyses of Shahin, while others, like Flynn and Kochetov and Alderete, initiate their investigations from the detailed examination of phonological patterns that seem to extend "downwards" into what has traditionally been deemed the domain of phonetics. The analyses are empirically rich, drawing on a wide range of data, some from more familiar languages and families (Germanic, Romance, Japanese, Austronesian) and others fromless well-documented groups (Niger-Congo), including a number of critically endangered languages (Salish, Wakashan, Pomoan, and Athapaskan). This is reflective of how over the past decades documentation of a broader diversity of languages and language families has expanded the empirical database of what is known about what properties languages may share and how languages may differ. A brief overview of the issues addressed by each paper follows. Traditionally, categorical behaviour is deemed to be a hallmark property of phonology. The first article, "Individual variation in English flaps and taps: A case of categorical phonetics" by Derrick and Gick, challenges this claim, based on ultrasound evidence related to the production of the tap/flap variant of /t/ and /d/ in [End Page 303] English. Their data, from cases of articulatory conflict, show that individual speakers' unconditioned phonetic production of the subphonemic variants of the tap/flap is categorical. The subphonemic variants are alveolar tap, postalveolar tap, down-flap, and up-flap. Their finding challenges categoricality as a sure indicator that a given sound property has phonological status. Derrick and Gick expect that any of the subphonemic variants can function contrastively. They see this as emerging where individual speakers' contrastive usage diffuses throughout the language community, and so they urge more experimental focus on individual speakers' phonetic behaviour. Gradience is standardly viewed as a, if not the, diagnostic for phonetic status, and this is the focus of the second article, "Acoustic testing for phonologization" by Shahin. This article first summarizes prevailing perspectives on the nature of phonetics and phonology, the phonetics-phonology interface, and phonologization, suggesting that the interface involves a monitor (tracking degree of gradience in phonetic data) along with the transducer (converting between the physical and cognitive) of Hale and Reiss (2000). Shahin then illustrates a procedure by which phonetic data can be inspected for degree of gradience to determine if a given sound property has phonetic or phonological status. The illustration involves acoustic data but the procedure can be used with any type of phonetic data, and it can reveal cases of language-specific phonetics. In the spirit of Myers (2000), she suggests that the procedure, as it results in clarification of what lies inside and outside phonology, will assist phonologists as they work to provide theoretical explanation for properties and patternings that are phonological. It has long been known that degrees of articulatory difficulty and perceptual salience, known as "phonetic scales", relate to phonology, but it has not always been clear if the scales are part of the grammar and, if they are, whether they are innate or learned. These issues are addressed directly by the third article, "Patterns and scales of expressive palatalization: Experimental evidence from Japanese", by Kochetov and Alderete. The authors present comprehensive crosslinguistic and theoretical evidence, and evidence from speaker intuitions, for a distinction between two types of palatalization...

End-state comfort governs kinematic variation in English flap/tap sequences.

Derrick & Gick [Can. Acoust. 36(3), 162 (2008)] found that North American English flap/tap singletons are produced using up to four phonetically conditioned kinematic variants (up-flaps, down-flaps, postalveolar taps, and alveolar taps), dependent on the position of the tongue tip in neighboring rhotic and nonrhotic vowels. Our pilot work has shown that sequences of multiple flap/tap variants are highly mechanically interdependent (e.g., subjects never produce two up-flaps or two down-flaps in adjacent syllables). The present study tests flap/tap sequences for compliance with “end-state comfort” [Rosenbaum et al., Acta Psychologica 94, 59 (1996)], as evidence for look-ahead planning. Twelve subjects’ tongue positions were measured during production of English words and word pairs with one- and two-flap sequences. Results show that sequences of flaps end with a preferred tongue tip position. For example, for 11 subjects, the /d/ of “edit” is realized as an alveolar tap or down-flap in “edit the,” but an up-flap in “edit a,” both facilitating ending the sequence with a low tongue position. A general avoidance of a sequence-final up-flap was also observed, such that /d/ of “edify” is an alveolar-tap or down-flap, while, for eight subjects, “editor” shows an up-flap, postalevolar tap sequence. [Work supported by NSERC.]

Acquiring novel perceptual categories in a third language: Bengali‐English bilinguals’ perception and learning of Malayalam consonants.

The present study examines the acquisition of novel non‐native speech contrasts by adult bilingual speakers of Bengali and English. One of the underlying issues in this study is the role that phonetic features may play in the development of new phonetic categories. For instance, features utilized in native contrasts may generalize in the learning of novel non‐native contrasts, even if they play a limited or no role in the initial perception of these non‐native contrasts [Polka (1992); Harnsberger (1998)]. To explore this feature generalization hypothesis, a high variability consonant identification training paradigm was used with ten Bengali‐English bilinguals who learned to identify 4 Malayalam place of articulation contrasts (dental versus retroflex nasals, dental versus retroflex lateral approximants, palatoalveolar versus retroflex voiceless fricatives, and alveolar tap versus retroflex approximant). The bilingual Bengali‐English listeners were selected for their extensive experience with the relevant place features (e.g., dental, retroflex, palatoalveolar, and alveolar), though Bengali and English lack any direct correspondents to the Malayalam contrasts. Bilingual performance was analyzed in terms of both patterns of perceptual assimilation as well as rate of acquisition of the novel non‐native contrasts.

Phonetic Typology and Positional Allophones for Alveolar Rhotics in Catalan

The present study reports electropalatographic and acoustic data on the positionaland contextual characteristics of alveolar taps and trills in Majorcan,Valencian and Eastern Catalan. The two consonant classes are invariably opposedby degree of tongue dorsum contact and F2, but only differentiated by place ofarticulation when constriction location for the trill is sufficiently retracted. Trillsare produced with less than three contacts and may exhibit a single contact inutterance-initial position and, less often, in /Cr, VrV/ sequences. Word-final and, toa lesser extent, preconsonantal rhotics are implemented as taps in Majorcan andValencian, and strengthened into trills in Eastern Catalan. Moreover, thereappears to be an inverse relationship between initial strengthening, and intervocalicweakening and the absence of syllable-final strengthening, for Valencianrhotics, which could be indicative of a pattern of intersegmental organization.Shortening and articulatory reduction turned out not to be necessarily related forextremely short Valencian taps, which undergo much undershoot intervocalicallybut are highly constricted in /Cɾ , rC/ sequences. Other research aspects suchas devoicing and intergestural timing for Catalan alveolar rhotics are alsoinvestigated.

A study of /J/ and /r/ in the light of the “DAC” coarticulation model

This is an electropalatographic and acoustic study of the production and coarticulatory characteristics of the alveolar tap /J/ and the alveolar trill /r/ in Catalan VCV sequences. The finding that /r/ involves more closure retraction, more predorsum lowering and less apico-predorsal coupling than /J/ suggests that the two consonants are realized by means of different lingual gestures. Moreover, in comparison with the tap, the trill exerts larger and longer effects on the adjacent vowels and is less sensitive to vocalic effects. This suggests that the tongue body is more constrained for /r/ than for /J/. The finding that the lingual gesture for a consonant (/r/) overrides the lingual gesture for an adjacent vowel (/i/) when the two segments are highly constrained and antagonistic is supportive of a view that consonantal production is more precisely controlled than vowel production. The tap and the trill also differ with respect to the direction of VCV coarticulatory effects: on the one hand, in /VrV/ sequences, anticipatory consonant and vowel coarticulation is predominant; on the other, in /VJV/ sequences, vocalic coarticulation is predominantly carryover, whereas consonantal coarticulation is mainly anticipatory. These patterns of coarticulatory behavior are accounted for in terms of gestural production mechanisms and biomechanical properties of the lingual articulators. Differences in articulation and coarticulation between /r/ and /J/ suggest that the trill is not a geminate correlate of the tap.

An articulatory and perceptual study of Tamil liquids

In Tamil there are 5 liquid consonants. They are the lateral approximation [l], the retroflex lateral approximant [hooked el], the alveolar tap [inverted sm cap jay], the retroflex tap [hooked are], and an obscure segment that has been variously described. This paper presents electropalatography data from one speaker and perceptual data from five listeners exploring the articulatory and acoustic dimensions of the phonetic space for liquids in Tamil. The dimensions which are explored include duration (distinguishing flaps and approximants), dynamic tongue movement (distinguishing retroflex and nonretro- flex), and constriction location/spectral shape (distinguishing [l] and the obscure liquid). Preliminary analyses suggest that the obscure liquid of Tamil is an apicopalatal central approximant. However, unlike other segments in Tamil which require tongue tip backing (the retroflexes) there is no evidence of dynamic tongue movement during the closure and some suggestion of acoustic zeros, perhaps from a lateral opening. Additional perception data designed to clarify this picture will be presented.

Numic[r] Is Not a Spirant

Virtually all scholars working on Numic languages have called [r] a spirant or listed it, without comment, as resulting from spirantization. However, Central Numic shows that [r] results from an early rule of tap formation applying to /t/, with subsequent application of spirantization then affecting other stops. When this analysis is extended to Western and Southern Numic, the result is that in no Numic language has [r] ever belonged to a series of spirants. (Contains 20 references.) (Author) *********************************************************************** Reproductions supplied by EDRS are the best that can be made from the original document. *********************************************************************** PERMISSION TO REPRODUCE THIS MATERIAL BY TO THE EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC).Numic Er] Is Not a Spirant James L. Armagost Kansas State University U.S. DEPARTMENT Of EDUCATION Office 01 Educational Research and Improvement EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) 1211his document has been reproduced as received Irom the person or orgaruzabon origtnattng tt. O Minor changes have been made to improve reproducbon Quality Points of view or opinions stated in this document do not necessarily represent othclal OERI posthon or policy John E. McLaughlin University of California, Santa Barbara Abstract: Virtually all scholars working on Numic languages have called [r] a spirant or listed it, without comment, as resulting from spirantization. However, Central Numic shows that [r] results from an early rile of tap formation applying to /t/, with subsequent application of spirantization then affecting other stops. When we extend this analysis to Western and Southern Numic, the result is that in no Numi.; language has [r] ever 1-elonged to a series of spirants. Virtually all scholars working on Numic languages have called [r] a spirant or listed it, without comment, as resulting from spirantization. However, Central Numic shows that [r] results from an early rile of tap formation applying to /t/, with subsequent application of spirantization then affecting other stops. When we extend this analysis to Western and Southern Numic, the result is that in no Numi.; language has [r] ever 1-elonged to a series of spirants. In this paper we dis=uss languages in the Numic branch of Uto-Aztecan with regard to one facet of the consonant gradations found there.' Specifically, we focus on what is usually referred to as the spirantized series, arguing that one supposed member of this series, the alveolar tap (or flap) (r), is not now and never has been a spirant in any of these languages, and in fact has nothing to do with a process of spirantization. To anyone unfamiliar with Numic languages we may seem to be taking an obvious position, perhaps comparable to arguing that the earth is not flat. Therefore, we first present data from Panamint to show why it has become general usage among Numicists to speak of (r) as resulting from spirantization. We briefly summarize corresponding data in other Central Numic languages and propose an alternative, and we think superior, analysis. We then extend our analysis historically to Western and Southern Numic, and finally to Proto Numic. Our concluding remarks are based on some additional crosslinguistic data. The underlying consonant and vowel system of Panamint is shown in (1).2 We will be concerned with the oral stops, which exhibit gradation patterns when they occur phrase medially. One such pattern is illustrated in (2), where we see a four-way alternation involving the initial stop of the postposition /pa ?an/ 'on'. The first column in (2) gives the citation form Kansas Working Papers in Linguistics, Volume 18, 1993, pp. 131-142

Tongue apex activities during alveolar stops.

Lateral cineradiograms, taken at a rate of 100 frames per second, were used to investigate tongue apex movements associated with alveolar consonants produced by five American English speakers. Consistent with previous research concerning bilabial stops, the results indicated a tendency for faster apical movements towards voiceless stops than toward voiced stops. On the other hand, tongue speeds during voiced alveolar stops and alveolar taps did not appear to differ significantly. Any tendency for faster speeds to be associated with tap movements appeared to be due to corresponding greater distances moved by the tongue tip. Finally, the amount of stress on vowels preceding or following [d] was not found to be an independent factor for determining tongue tip speeds toward the consonant. Once again, any speed differences seemed predictable on the basis of distance.

On the wider application of the term “tap”

It is noted that the length difference claimed between [ꓩ], a voiced alveolar tap and [d], a voiced alveolar stop, exists between [ṇ] and [n] and between [ḷ] and [l], tapped and non-tapped alveolar nasals and laterals respectively in Ghotuọ and some other Ẹdo languages of the Bendel State of Nigeria. It is shown that tapped nasals and tapped laterals are to be found in some other languages of the lower basin of the Niger.However, length or duration is also a primary differentiating feature between lenis and non-lenis sounds (other factors such as “muscular tension” and “force of articulation” being difficult to pin down). Therefore, taps and lenis sounds which share the same phonetic correlates should be described in terms of a multivalued feature [length], following Williamson (1976). Sounds of normal duration would be [0 length] while shorter-than-normal sounds are [–1 length]. Long or so-called fortis sounds would be [1 length].