Publications

Abstract

One of the challenges that infants have to solve when learn- ing their native language is to identify the words in a con- tinuous speech stream. Some of the experiments in Artificial Grammar Learning (Saffran, Newport, and Aslin (1996); Saf- fran, Aslin, and Newport (1996); Aslin, Saffran, and Newport (1998) and many more) investigate this ability. In these ex- periments, subjects are exposed to an artificial speech stream that contains certain regularities. Adult participants are typ- ically tested with 2-alternative Forced Choice Tests (2AFC) in which they have to choose between a word and another sequence (typically a partword, a sequence resulting from misplacing boundaries).

Abstract

How do different languages encode location and position meanings? In conjunction with the BowPed picture series and Caused Positions task, this elicitation tool is designed to help researchers (i) identify a language’s resources for encoding topological relations; (ii) delimit the pragmatics of use of such resources; and (iii) determine the semantics of select spatial terms. The task focuses on the exploration of the predicative component of topological expressions (e.g., ‘the cassavas are lying in the basket’), especially the contrastive elicitation of positional verbs. The materials consist of a set of photographs of objects (e.g., bottles, cloths, sticks) in specific configurations with various ground items (e.g., basket, table, tree).

Abstract

We introduce a biomechanical model of oropharyngeal structures that adds the soft-palate, pharynx, and larynx to our previous models of jaw, skull, hyoid, tongue, and face in a unified model. The model includes a comprehensive description of the upper airway musculature, using point-to-point muscles that may either be embedded within the deformable structures or operate exter- nally. The airway is described by an air-tight mesh that fits and deforms with the surrounding articulators, which enables dynamic coupling to our articulatory speech synthesizer. We demonstrate that the biomechanics, in conjunction with the skinning, supports a range from physically realistic to simplified vocal tract geometries to investigate different approaches to aeroacoustic modeling of vocal tract. Furthermore, our model supports inverse modeling to support investigation of plausible muscle activation patterns to generate speech.

Abstract

Psycholinguistic studies of metaphor processing must control their stimuli not just for word frequency but also for the frequency with which a term is used metaphorically. Thus, we consider the task of metaphor frequency estimation, which predicts how often target words will be used metaphorically. We develop metaphor classifiers which represent metaphorical domains through Latent Dirichlet Allocation, and apply these classifiers to the target words, aggregating their decisions to estimate the metaphorical frequencies. Training on only 400 sentences, our models are able to achieve 61.3 % accuracy on metaphor classification and 77.8 % accuracy on HIGH vs. LOW metaphorical frequency estimation.

Abstract

How do we decide where events begin and end? Like the ECOM clips, this questionnaire is designed to investigate how a language divides and/or integrates complex scenarios into sub-events and macro-events. The questionnaire focuses on events of motion, caused state change (e.g., breaking), and transfer (e.g., giving). It provides a checklist of scenarios that give insight into where a language “draws the line” in event integration, based on known cross-linguistic differences.

Abstract

How do we decide where events begin and end? In some languages it makes sense to say something like Dan broke the plate, but in other languages it is necessary to treat this action as a complex scenario composed of separate stages (Dan dropped the plate and then the plate broke). This document introduces issues concerning the linguistic and cognitive representations of event complexity and integration, and provides an overview of tasks that are relevant to this topic, including the ECOM clips, the Questionnaire on Event integration, and the Questionnaire on motion lexicalisation and motion description.

Abstract

How do we decide where events begin and end? In some languages it makes sense to say something like Dan broke the plate, but in other languages it is necessary to treat this action as a complex scenario composed of separate stages (Dan dropped the plate and then the plate broke). The “Event Complexity” (ECOM) clips are designed to explore how languages differ in dividing and/or integrating complex scenarios into sub-events and macro-events. The stimuli consist of animated clips of geometric shapes that participate in different scenarios (e.g., a circle “hits” a triangle and “breaks” it). Consultants are asked to describe the scenes, and then to comment on possible alternative descriptions.

Abstract

Disfluencies, such as uh and uhm, are known to help the listener in speech comprehension. For instance, disfluencies may elicit prediction of less accessible referents and may trigger listeners’ attention to the following word. However, recent work suggests differential processing of disfluencies in native and non-native speech. The current study investigated whether the beneficial effects of disfluencies on listeners’ attention are modulated by the (non-)native identity of the speaker. Using the Change Detection Paradigm, we investigated listeners’ recall accuracy for words presented in disfluent and fluent contexts, in native and non-native speech. We observed beneficial effects of both native and non-native disfluencies on listeners’ recall accuracy, suggesting that native and non-native disfluencies trigger listeners’ attention in a similar fashion.

Abstract

Speech perception involves a number of processes that deal with variation in the speech signal. One such process is normalization for speechrate: local temporal cues are perceived relative to the rate in the surrounding context. It is as yet unclear whether and how this perceptual effect interacts with higher level impressions of rate, such as a speaker’s nonnative identity. Nonnative speakers typically speak more slowly than natives, an experience that listeners take into account when explicitly judging the rate of nonnative speech. The present study investigated whether this is also reflected in implicit rate normalization. Results indicate that nonnative speech is implicitly perceived as faster than temporally-matched native speech, suggesting that the additional cognitive load of listening to an accent speeds up rate perception. Therefore, rate perception in speech is not dependent on syllable durations alone but also on the ease of processing of the temporal signal.

Abstract

The notion of relational grammar is extented to ternary relations and illustrated by a fragment of English. Some of Peirce's terms for ternary relations are shown to be incorrect and corrected.

Abstract

This paper introduces the EU-FP7 project CLARIN, a joint effort of over 150 institutions in Europe, aimed at the creation of a sustainable language resources and technology infrastructure for the humanities and social sciences research community. The paper briefly introduces the vision behind the project and how it relates to speech research with a focus on the contributions that CLARIN can and will make to research in spoken language processing.

Abstract

This task is designed to elicit expressions of spatial relations. It was originally designed by Melissa Bowerman for use with young children, but was then developed further by Bowerman in collaboration with Pederson for crosslinguistic comparison. It has been used in fieldsites all over the world and is commonly known as “BowPed” or “TPRS”. Older incarnations did not always come with instructions. This entry includes a one-page instruction sheet and high quality versions of the original pictures.

Abstract

This entry suggests ways to elicit descriptions of caused motion involving topological relations (the domain of English put IN/ON/TOGETHER, take OUT/OFF/APART, etc.). There is a large amount of cross-linguistic variation in this domain. The tasks outlined here address matters such as the division of labor between the various elements of spatial semantics in the sentence. For example, is most of the work of expressing PATH done in a locative marker, or in the verb, or both?

Abstract

The precursor to the Bowped stimuli, this entry suggests various spatial configurations to explore using real objects, rather than the line drawings used in Bowped.

Abstract

This study investigated pronunciation variants of word-final obstruent-liquid-schwa (OLS) clusters in casual French and the variables predicting the absence of the phonemes in these clusters. In a dataset of 291 noun tokens extracted from a corpus of casual conversations, we observed that in 80.7% of the tokens, at least one phoneme was absent and that in no less than 15.5% the whole cluster was absent (e.g., /mis/ for ministre). Importantly, the probability of a phoneme being absent was higher if the following phoneme was absent as well. These data show that reduction can affect several phonemes at once and is not restricted to just a handful of (function) words. Moreover, our results demonstrate that the absence of each single phoneme is affected by the speaker's tendency to increase ease of articulation and to adapt a word's pronunciation variant to the time available.

Abstract

The aim of the present study was to investigate whether speech-in-speech recognition is affected by variation in the target-background timing relationship. Specifically, we examined whether within trial synchronous or asynchronous onset and offset of the target and background speech influenced speech-in-speech recognition. Native English listeners were presented with English target sentences in the presence of English or Dutch background speech. Importantly, only the short-term temporal context –in terms of onset and offset synchrony or asynchrony of the target and background speech– varied across conditions. Participants’ task was to repeat back the English target sentences. The results showed an effect of synchronicity for English-in-English but not for English-in-Dutch recognition, indicating that familiarity with the English background lead in the asynchronous English-in-English condition might have attracted attention towards the English background. Overall, this study demonstrated that speech-in-speech recognition is sensitive to the target-background timing relationship, revealing an important role for variation in the local context of the target-background relationship as it extends beyond the limits of the time-frame of the to-be-recognized target sentence.

Abstract

Speakers of the Mayan language Tzeltal use two frames of reference for spatial reckoning: an absolute system (based on the south/north axis abstracted from the overall slope of the land) and an intrinsic system utilizing spatial axes of the reference object to establish body parts. This paper examines the use of absolute, intrinsic, and landmark cues in descriptions of spatial relations by 22 pairs of Tzeltal children aged between 5 and 17. The data are drawn from interactive space games, where a Director describes a spatial layout in a photo and the Matcher reproduces it with toys. The paper distinguishes use of ad hoc landmarks ('Red Cliffs', 'the electricity post') from genuine absolute reference points ('uphill'/'downhill'/’across’), and shows that adults in this task use absolute ('cow uphill of horse'), intrinsic ('at the tree's side') and landmark ('cow facing Red Cliffs') descriptions to communicate the spatial relations depicted. The youngest children, however, do not use landmark cues at all but rely instead on deictics and on the absolute 'uphill/downhill' terms; landmark terms are still rare at age 8-10. Despite arguments that landmarks are a simpler, more natural, basis for spatial reckoning than absolute terms, there is no evidence for a developmental progression from landmark-based to absolute-based strategies. We relate these observations to Slobin’s ‘thinking for speaking’ argument.

Abstract

How do children isolate the semantic package contained in verb roots in the Mayan language Tzeltal? One might imagine that the canonical CVC shape of roots characteristic of Mayan languages would make the job simple, but the root is normally preceded and followed by affixes which mask its identity. Pye (1983) demonstrated that, in Kiche' Mayan, prosodic salience overrides semantic salience, and children's first words in Kiche' are often composed of only the final (stressed) syllable constituted by the final consonant of the CVC root and a 'meaningless' termination suffix. Intonation thus plays a crucial role in early Kiche' morphological development. Tzeltal presents a rather different picture: The first words of children around the age of 1;6 are bare roots, children strip off all prefixes and suffixes which are obligatory in adult speech. They gradually add them, starting with the suffixes (which receive the main stress), but person prefixes are omitted in some contexts past a child's third birthday, and one obligatory aspectual prefix (x-) is systematically omitted by the four children in my longitudinal study even after they are four years old. Tzeltal children's first verbs generally show faultless isolation of the root. An account in terms of intonation or stress cannot explain this ability (the prefixes are not all syllables; the roots are not always stressed). This paper suggests that probable clues include the fact that the CVC root stays constant across contexts (with some exceptions) whereas the affixes vary, that there are some linguistic contexts where the root occurs without any prefixes (relatively frequent in the input), and that the Tzeltal discourse convention of responding by repeating with appropriate deictic alternation (e.g., "I see it." "Oh, you see it.") highlights the root.

Abstract

This is a German translation of Brown 1980, How and why are women more polite: Some evidence from a Mayan community.

Abstract

In this paper I take up the idea that human thinking is systematically biased in the direction of interactive thinking (E. Goody's anticipatory interactive planning), that is, that humans are peculiarly good at, and inordinately prone to, attributing intentions and goals to one other (as well as to non-humans), and that they routinely orient to presumptions about each other's intentions in what they say and do. I explore the implications of that idea for an understanding of politeness in interaction, taking as a starting point the Brown and Levinson (1987) model of politeness, which assumes interactive thinking, a notion implicit in the formulation of politeness as strategic orientation to face. Drawing on an analysis of the phenomenon of conventionalized ‘irony’ in Tzeltal, I emphasize that politeness does not inhere in linguistic form per se but is a matter of conveying a polite intention, and argue that Tzeltal irony provides a prime example of one way in which humans' highly-developed intellectual machinery for inferring alter's intentions is put to the service of social relationships.

Abstract

This article is a reprint of chapter 1, the introduction to Brown and Levinson, 1987, Politeness: Some universals in language usage (Cambridge University Press).

Abstract

Under noise or speech reductions, young adult listeners flexibly adjust the parameters of lexical activation and competition to allow for speech signal unreliability. Consequently, mismatches in the input are treated more leniently such that lexical candidates are not immediately deactivated. Using eyetracking, we assessed whether this modulation of recognition dynamics also occurs for older listeners. Dutch participants (aged 60+) heard Dutch sentences containing a critical word while viewing displays of four line drawings. The name of one picture shared either onset or rhyme with the critical word (i.e., was a phonological competitor). Sentences were either clear and noise-free, or had several phonemes replaced by bursts of noise. A larger preference for onset competitors than for rhyme competitors was observed in both clear and noise conditions; performance did not alter across condition. This suggests that dynamic adjustment of spoken-word recognition parameters in response to noise is less available to older listeners.

Abstract

Semplates are a new descriptive and theoretical concept in lexical semantics, borne out of recent L&C work in several domains. A semplate can be defined as a configuration consisting of distinct layers of lexemes, each layer drawn from a different form class, mapped onto the same abstract semantic template. Within such a lexical layer, the sense relations between the lexical items are inherited from the underlying template. Thus, the whole set of lexical layers and the underlying template form a cross-categorial configuration in the lexicon. The goal of this task is to find new kinds of macrostructure in the lexicon, with a view to cross-linguistic comparison.

Abstract

Large auditory-visual (AV) speech corpora are the grist of modern research in speech science, but no such corpus exists for Australian English. This is unfortunate, for speech science is the brains behind speech technology and applications such as text-to-speech (TTS) synthesis, automatic speech recognition (ASR), speaker recognition and forensic identification, talking heads, and hearing prostheses. Advances in these research areas in Australia require a large corpus of Australian English. Here the authors describe a blueprint for building the Big Australian Speech Corpus (the Big ASC), a corpus of over 1,100 speakers from urban and rural Australia, including speakers of non-indigenous, indigenous, ethnocultural, and disordered forms of Australian English, each of whom would be sampled on three occasions in a range of speech tasks designed by the researchers who would be using the corpus.

Abstract

According to theories of embodied cognition, understanding a verb like throw involves unconsciously simulating the action throwing, using areas of the brain that support motor planning. If understanding action words involves mentally simulating our own actions, then the neurocognitive representation of word meanings should differ for people with different kinds of bodies, who perform actions in systematically different ways. In a test of the body-specificity hypothesis (Casasanto, 2009), we used fMRI to compare premotor activity correlated with action verb understanding in right- and left-handers. Right-handers preferentially activated left premotor cortex during lexical decision on manual action verbs (compared with non-manual action verbs), whereas left-handers preferentially activated right premotor areas. This finding helps refine theories of embodied semantics, suggesting that implicit mental simulation during language processing is body-specific: Right and left-handers, who perform actions differently, use correspondingly different areas of the brain for representing action verb meanings.

Abstract

What is the relationship between motor action and emotion? Here we investigated whether people associate good things more strongly with the dominant side of their bodies, and bad things with the non-dominant side. To find out, we analyzed spontaneous gestures during speech expressing ideas with positive or negative emotional valence (e.g., freedom, pain, compassion). Samples of speech and gesture were drawn from the 2004 and 2008 US presidential debates, which involved two left-handers (Obama, McCain) and two right-handers (Kerry, Bush). Results showed a strong association between the valence of spoken clauses and the hands used to make spontaneous co-speech gestures. In right-handed candidates, right-hand gestures were more strongly associated with positive-valence clauses, and left-hand gestures with negative-valence clauses. Left-handed candidates showed the opposite pattern. Right- and left-handers implicitly associated positive valence more strongly with their dominant hand: the hand they can use more fluently. These results support the body-specificity hypothesis, (Casasanto, 2009), and suggest a perceptuomotor basis for even our most abstract ideas.

Abstract

What is the relationship between space and time in the human mind? Studies in adults show an asymmetric relationship between mental representations of these basic dimensions of experience: representations of time depend on space more than representations of space depend on time. Here we investigated the relationship between space and time in the developing mind. Native Greek-speaking children (N=99) watched movies of two animals traveling along parallel paths for different distances or durations and judged the spatial and temporal aspects of these events (e.g., Which animal went for a longer time, or a longer distance?) Results showed a reliable cross-dimensional asymmetry: for the same stimuli, spatial information influenced temporal judgments more than temporal information influenced spatial judgments. This pattern was robust to variations in the age of the participants and the type of language used to elicit responses. This finding demonstrates a continuity between space-time representations in children and adults, and informs theories of analog magnitude representation.

Abstract

Speaker transitions in conversation are often brief, with minimal vocal overlap. Signed languages appear to defy this pattern with frequent, long spans of simultaneous signing. But recent evidence suggests that turn boundaries in signed language may only include the content-bearing parts of the turn (from the first stroke to the last), and not all turn-related movement (from first preparation to final retraction). We tested whether signers were able to anticipate “stroke-to-stroke” turn boundaries with only minimal conversational context. We found that, indeed, signers anticipated turn boundaries at the ends of turn-final strokes. Signers often responded early, especially when the turn was long or contained multiple possible end points. Early responses for long turns were especially apparent for interrogatives—long interrogative turns showed much greater anticipation compared to short ones.

Abstract

In this paper we explore the feasibility of artificial (formal) grammar recognition (AGR) using spiking neural networks. A biologically inspired minicolumn architecture is designed as the basic computational unit. A network topography is defined based on the minicolumn architecture, here referred to as nodes, connected with excitatory and inhibitory connections. Nodes in the network represent unique internal states of the grammar’s finite state machine (FSM). Future work to improve the performance of the networks is discussed. The modeling framework developed can be used by neurophysiological research to implement network layouts and compare simulated performance characteristics to actual subject performance.

Abstract

Studies on children’s use of intonation in reference are few in number but are diverse in terms of theoretical frameworks and intonational parameters. In the current review, I present a re-analysis of the referents in each study, using a three-dimension approach (i.e. referential givenness-newness, relational givenness-newness, contrast), discuss the use of intonation at two levels (phonetic, phonological), and compare findings from different studies within a single framework. The patterns stemming from these studies may be limited in generalisability but can serve as initial hypotheses for future work. Furthermore, I examine the role of input as available in infant direct speech in the acquisition of intonational encoding of referents. In addition, I discuss how future research can advance our knowledge.

Abstract

Same-different discrimination judgments for pairs of Korean stop consonants, or of Japanese syllables differing in phonetic segment length, were made by adult Korean adoptees in the Netherlands, by matched Dutch controls, and Korean controls. The adoptees did not outdo either control group on either task, although the same individuals had performed significantly better than matched controls on an identification learning task. This suggests that early exposure to multiple phonetic systems does not specifically improve acoustic-phonetic skills; rather, enhanced performance suggests retained language knowledge.

Abstract

When we speak, we spontaneously produce gestures (co-speech gestures). Co-speech gestures and speech production are closely interlinked. However, the exact nature of the link is still under debate. To addressed the question that whether co-speech gestures originate from the speech production system or from a system independent of the speech production, the present study examined the relationship between co-speech and co-thought gestures. Co-thought gestures, produced during silent thinking without speaking, presumably originate from a system independent of the speech production processes. We found a positive correlation between the production frequency of co-thought and co-speech gestures, regardless the communicative function that co-speech gestures might serve. Therefore, we suggest that co-speech gestures and co-thought gestures originate from a common system that is independent of the speech production processes

Abstract

The present study investigated the effect of orthography on the learning and subsequent processing of pronunciation variants in a second language. Dutch learners of French learned reduced pronunciation variants that result from schwa-zero alternation in French (e.g., reduced /ʃnij/ from chenille 'caterpillar'). Half of the participants additionally learnt the words' spellings, which correspond more closely to the full variants with schwa. On the following day, participants performed an auditory lexical decision task, in which they heard half of the words in their reduced variants, and the other half in their full variants. Participants who had exclusively learnt the auditory forms performed significantly worse on full variants than participants who had also learnt the spellings. This shows that learners integrate phonological and orthographic information to process pronunciation variants. There was no difference between both groups in their performances on reduced variants, suggesting that the exposure to spelling does not impede learners' processing of these variants.

Abstract

Odor naming is difficult for people, but recent cross-cultural research suggests this difficulty is culture-specific. Jahai speakers (hunter-gatherers from the Malay Peninsula) name odors as consistently as colors, and much better than English speakers (Majid & Burenhult, 2014). In Jahai the linguistic advantage for smells correlates with a cultural interest in odors. Here we ask whether sub-cultures in the West with odor expertise also show superior odor naming. We tested wine and coffee experts (who have specialized odor training) in an odor naming task. Both wine and coffee experts were no more accurate or consistent than novices when naming odors. Although there were small differences in naming strategies, experts and non-experts alike relied overwhelmingly on source-based descriptions. So the specific language experts speak continues to constrain their ability to express odors. This suggests expertise alone is not sufficient to overcome the limits of language in the domain of smell.

Abstract

The lexicon, considered as a component of the process of recognizing speech, is a device that accepts a sound image as input and outputs meaning. Lexical access is the process of formulating an appropriate input and mapping it onto an entry in the lexicon's store of sound images matched with their meanings. This chapter addresses the problems of auditory lexical access from continuous speech. The central argument to be proposed is that utterance prosody plays a crucial role in the access process. Continuous listening faces problems that are not present in visual recognition (reading) or in noncontinuous recognition (understanding isolated words). Aspects of utterance prosody offer a solution to these particular problems.

Abstract

A distinction between strong and weak vowels can be drawn on the basis of vowel quality, of stress, or of both factors. An experiment was conducted in which sets of contextually matched word-intial vowels ranging from clearly strong to clearly weak were cross-spliced, and the naturalness of the resulting words was rated by listeners. The ratings showed that in general cross-spliced words were only significantly less acceptable than unspliced words when schwa was not involved; this supports a categorical distinction based on vowel quality.

Abstract

English and French listeners performed two tasks - click location and speeded click detection - with both English and French sentences, closely matched for syntactic and phonological structure. Clicks were located more accurately in open- than in closed-class words in both English and French; they were detected more rapidly in open- than in closed-class words in English, but not in French. The two listener groups produced the same pattern of responses, suggesting that higher-level linguistic processing was not involved in these tasks.

Abstract

One of a listener's major tasks in understanding continuous speech is segmenting the speech signal into separate words. When listening conditions are difficult, speakers can help listeners by deliberately speaking more clearly. In three experiments, we examined how word boundaries are produced in deliberately clear speech. We found that speakers do indeed attempt to mark word boundaries; moreover, they differentiate between word boundaries in a way which suggests they are sensitive to listener needs. Application of heuristic segmentation strategies makes word boundaries before strong syllables easiest for listeners to perceive; but under difficult listening conditions speakers pay more attention to marking word boundaries before weak syllables, i.e. they mark those boundaries which are otherwise particularly hard to perceive.

Abstract

Two phoneme goodness rating experiments addressed the role of orthographic knowledge in the evaluation of speech sounds. Ratings for the best tokens of /s/ were higher in words spelled with S (e.g., bless) than in words where /s/ was spelled with C (e.g., voice). This difference did not appear for analogous nonwords for which every lexical neighbour had either S or C spelling (pless, floice). Models of phonemic processing incorporating obligatory influence of lexical information in phonemic processing cannot explain this dissociation; the data are consistent with models in which phonemic decisions are not subject to necessary top-down lexical influence.

Abstract

Two lexical decision experiments in Cantonese are described in which the recognition of spoken target words as a function of phonological similarity to a preceding prime is investigated. Phonological similaritv in first syllables produced inhibition, while similarity in second syllables led to facilitation. Differences between syllables in tonal and segmental structure had generally similar effects.

Abstract

Speakers and listeners have a shared goal: to communicate. The processes of speech perception and of speech production interact in many ways under the constraints of this communicative goal; such interaction is as characteristic of prosodic processing as of the processing of other aspects of linguistic structure. Two of the major uses of prosodic information in situations of communication are to encode salience and segmentation, and these themes unite the contributions to the symposium introduced by the present review.

Abstract

This chapter highlights that most language behavior consists of speaking and listening. The chapter also reveals differences and similarities between speaking and listening. The laboratory study of word production raises formidable problems; ensuring that a particular word is produced may subvert the spontaneous production process. Word production is investigated via slips and tip-of-the-tongue (TOT), primarily via instances of processing failure and via the technique of via the picture-naming task. The methodology of word production is explained in the chapter. The chapter also explains the phenomenon of interaction between various stages of word production and the process of speech recognition. In this context, it explores the difference between sound and meaning and examines whether or not the comparisons are appropriate between the processes of recognition and production of spoken words. It also describes the similarities and differences in the structure of the recognition and production systems. Finally, the chapter highlights the common issues in recognition and production research, which include the nuances of frequency of occurrence, morphological structure, and phonological structure.

Abstract

The performance of automatic speech recognition systems is usually assessed in terms of error rate. Human speech recognition produces few errors, but relative difficulty of processing can be assessed via response time techniques. We report the construction of a measure analogous to response time in a machine recognition system. This measure may be compared directly with human response times. We conducted a trial comparison of this type at the phoneme level, including both tense and lax vowels and a variety of consonant classes. The results suggested similarities between human and machine processing in the case of consonants, but differences in the case of vowels.