Locus equations derived from compensatory articulation

Mapping typical and hypokinetic dysarthric speech production network using a connected speech paradigm in functional MRI

We developed a task paradigm whereby subjects spoke aloud while minimizing head motion during functional MRI (fMRI) in order to better understand the neural circuitry involved in motor speech disorders due to dysfunction of the central nervous system. To validate our overt continuous speech paradigm, we mapped the speech production network (SPN) in typical speakers (n = 19, 10 females) and speakers with hypokinetic dysarthria as a manifestation of Parkinson disease (HKD; n = 21, 8 females) in fMRI. We then compared it with the SPN derived during overt speech production by ¹⁵O-water PET in the same group of typical speakers and another HKD cohort (n = 10, 2 females). The fMRI overt connected speech paradigm did not result in excessive motion artifacts and successfully identified the same brain areas demonstrated in the PET studies in the two cohorts. The SPN derived in fMRI demonstrated significant spatial overlap with the corresponding PET derived maps (typical speakers: r = 0.52; speakers with HKD: r = 0.43) and identified the components of the neural circuit of speech production belonging to the feedforward and feedback subsystems. The fMRI study in speakers with HKD identified significantly decreased activity in critical feedforward (bilateral dorsal premotor and motor cortices) and feedback (auditory and somatosensory areas) subsystems replicating previous PET study findings in this cohort. These results demonstrate that the overt connected speech paradigm is feasible during fMRI and can accurately localize the neural substrates of typical and disordered speech production. Our fMRI paradigm should prove useful for study of motor speech and voice disorders, including stuttering, apraxia of speech, dysarthria, and spasmodic dysphonia.

The effects of measurement error and vowel selection on the locus equation measure of coarticulation

Effects on the slope of introducing error in the F2 Hz values in locus equations (LEs) and of using fewer than ten vowels were investigated. For each of the initial consonants /b, d, g/, 2000 simulated sets were generated using Monte Carlo techniques. The sets were altered with 50, 100, or 200 Hz error being randomly applied to each F2 value within a set. Selected vowels were then removed from the sets and the effects on the slopes were measured. Results suggest that the LE slopes are generally resistant to error and reduced number of vowels. Effects of adding 50 Hz of random error to the F2 values in sets using eight or ten vowels were minimal, yielding a mean absolute change in slope less than 0.07.

The integrity of anticipatory coarticulation in fluent and non-fluent tokens of adults who stutter

This article analysed the acoustic structure of voiced stop + vowel sequences in a group of persons who stutter (PWS). This phonetic unit was chosen because successful production is highly dependent on the differential tweaking of right-to-left anticipatory coarticulation as a function of stop place. Thus, essential elements of both speech motor planning and execution can be parsimoniously assessed. Five adult PWS read three passages 3 times in a randomised order. These passages contained an overabundance of words beginning with initial [bV], [dV] and [gV] sequences. Digital audio and visual recordings were analysed to first identify fluent and stuttered target words, which were then spectrally analysed to yield locus equation (LE) regression plots. The slope of the LE regression function directly indexes the coarticulatory extent of the vowel's influence on the preceding stop consonant. The PWS revealed LE parameters falling within the normal ranges based on previously documented data obtained from fluent speakers. Theoretical considerations of possible underlying factors responsible for stuttering disfluencies are discussed relevant to these findings.

Perception of Place of Articulation for Plosives and Fricatives in Noise

This study aims at uncovering perceptually-relevant acoustic cues for the labial versus alveolar place of articulation distinction in syllable-initial plosives {/b/,/d/,/p/,/t/} and fricatives {/f/,/s/,/v/,/z/} in noise. Speech materials consisted of naturally-spoken consonant-vowel (CV) syllables from four talkers where the vowel was one of {/a/,/i/,/u/}. Acoustic analyses using logistic regression show that formant frequency measurements, relative spectral amplitude measurements, and burst/noise durations are generally reliable cues for labial/alveolar classification. In a subsequent perceptual experiment, each pair of syllables with the labial/alveolar distinction (e.g., /ba,da/) was presented to listeners in various levels of signal-to-noise-ratio (SNR) in a 2-AFC task. A threshold SNR was obtained for each syllable pair using sigmoid fitting of the percent correct scores. Results show that the perception of the labial/alveolar distinction in noise depends on the manner of articulation, the vowel context, and interaction between voicing and manner of articulation. Correlation analyses of the acoustic measurements and threshold SNRs show that formant frequency measurements (such as F1 and F2 onset frequencies and F2 and F3 frequency changes) become increasingly important for the perception of labial/alveolar distinctions as the SNR degrades.

Speech characteristics of monozygotic twins and a same-sex sibling: an acoustic case study of coarticulation patterns in read speech

This case study reports on an acoustic investigation of the motor speech characteristics of a set of young adult male monozygotic (MZ) twins and compares them to those of an age- and sex-matched sibling who participated in the study 2 years later to match for demographic factors. Coarticulation patterns were investigated from read samples of consonant-vowel sequences in monosyllabic words containing a variety of consonants and vowels. This was done by examining F(2) vowel onsets and F(2) vowel targets, plotted as F(2) locus equations. Data were processed for between-sibling differences using a number of statistical tests. Results indicated that the MZ twins displayed F(2) parameters and coarticulation patterns which were more similar than those of their age- and sex-matched sibling. The results of this case study therefore suggest that acoustic phonetic parameters used to index coarticulation patterns have the potential to profile some of the similarities and differences in the speech characteristics of genetically related individuals.

Evidence of motor programming deficits in children diagnosed with DAS

In this study the hypothesis of motor programming involvement in developmental apraxia of speech (DAS) was investigated by studying articulatory compensation. Five children with DAS and 5 normally speaking children (age 5;0 [years;months] to 6;10), and 6 adult women produced utterances in a normal speaking condition and in a bite-block condition in which the mandible was kept in a fixed position. Throughout the utterances, the course of the second formant was used to determine articulatory compensation and the effect of the bite block on anticipatory coarticulation. Results showed that the bite-block condition in normally speaking children, like in adult women, did not affect the extent of anticipatory coarticulation. In the speech of children with DAS, the bite block had large effects on coarticulatory patterns and on vowel quality, which, contrary to expectations, had improved. These results are interpreted as a clear demonstration of deficient motor programming in DAS.

Second formant transitions in fluent speech of persistent and recovered preschool children who stutter

This study investigated frequency change and duration of the second formant (F2) transitions in perceptually fluent speech samples recorded close to stuttering onset in preschool age children. Comparisons were made among 10 children known to eventually persist in stuttering, 10 who eventually recovered from stuttering, and 10 normally fluent controls. All were enrolled in the longitudinal Stuttering Research Project at the University of Illinois. Subjects fluently repeated standard experimental sentences. The same 36 perceptually fluent target segments (syllables embedded in words) from each subject's repeated sentences were analyzed. The syllables were divided into three phonetic categories based on their initial consonant: bilabial, alveolar, and velar placement. The frequency change and duration of F2 transitions were analyzed for each of the target CV segments. F2 transition onset and offset frequencies and their interval (duration) were measured for each utterance. Data indicate that near stuttering onset, children whose stuttering eventually persisted demonstrated significantly smaller frequency change than that of the recovered group. It is suggested that the F2 transitions should continue to be investigated as a possible predictor of stuttering pathways.

LEARNING OUTCOMES

(1) Readers will learn about studies regarding second formant transition related to stuttering. (2) Readers will learn about differences between children who persist in stuttering and those who recover from stuttering. (3) Readers will learn about research concerned with early identification of risk criteria in persistent stuttering.

Preservation of place and manner cues during simultaneous communication: a spectral moments perspective

Spectral moments, which describe the distribution of frequencies in a spectrum, were used to investigate the preservation of acoustic cues to intelligibility of speech produced during simultaneous communication (SC) in relation to acoustic cues produced when speaking alone. The spectral moment data obtained from speech alone (SA) were comparable to those spectral moment data reported by Jongman, Wayland, and Wong (2000) and Nittrouer (1995). The spectral moments obtained from speech produced during SC were statistically indistinguishable from those obtained during SA, indicating no measurable degradation of obstruent spectral acoustic cues during SC.

EDUCATIONAL OBJECTIVES

As a result of this activity, the participant will be able to (1) describe SC; (2) explain the role of SC in communication with children who are deaf; (3) describe the first, third, and fourth spectral moments of obstruent consonants; and (4) identify spectral moment patterns in speech produced during SC.

Voiceless consonants and locus equations: a comparison with electropalatographic data on coarticulation

The appropriateness of using locus equations in a study of voiceless consonants is examined. The consonants investigated are the stops /t k/ and the fricatives /theta s integral/. The slope value of the locus equation, indicating degree of coarticulation in the CV syllable, is compared with electropalatographic (EPG) data on coarticulation. It is shown that, overall, there is a very poor correlation between locus equation and EPG data as regards coarticulation. It is also shown that more accurate locus equation results - in terms of their correlation with EPG data - are obtained for stop consonants when F2 onset is sampled at stop release, rather than at the onset of voicing for the vowel. Finally, results for the voiceless consonants are compared with results for their homorganic voiced counterparts. Whilst there is no significant difference between voiced and voiceless consonants in the EPG data, there is a significant difference between voiced and voiceless consonants in the locus equation data. These results suggest that locus equations cannot provide invariant cues for stop and fricative place of articulation across the voiced-voiceless distinction.

Representation of phonological categories: a functional role for auditory columns

A speculative model for representing phonological categories for stop place of articulation is offered. The model is analogous to neural columns in both inferotemporal cortex in the macaque for achieving constancy in object recognition and in the inferior colliculus of the barn owl for resolving ambiguous phase disparities for sound localization. Hypothetical auditory columns for encoding the vowel context-induced variability of F2 transitions are described using acoustic data from locus equation scatterplots of human speech. Such analyses serve to absorb the allophonic variability inherent in coarticulated speech utterances.

Effects of facial paralysis and audiovisual information on stop place identification

This study investigated how listeners' perceptions of bilabial and lingua-alveolar voiced stops in auditory (A) and audiovisual (AV) presentation modes were influenced by articulatory function in a girl with bilateral facial paralysis (BFP) and a girl with normal facial movement (NFM). The Fuzzy Logic Model of Perception (FLMP) was used to make predictions about listeners' identifications of stop place based on assumptions about the nature (clear, ambiguous, or conflicting) of the A or AV cues produced by each child during /b/ and /d/ CV syllables. As predicted, (a) listeners' identification scores for NFM were very high and reliable, regardless of presentation mode or stop place, (b) listeners' identification scores for BFP were high for lingua-alveolar place, regardless of presentation mode, but more variable and less reliable than for NFM; significantly lower (overall at a chance level) for bilabial place in the A mode; and lowest for bilabial place in the AV mode. Conflicting visual cues for stop place for BFP's productions of /bV/ syllables influenced listeners' perceptions, resulting in most of her bilabial syllables being misidentified in the AV mode. F2 locus equations for each child's /bV/ and /dV/ syllables showed patterns similar to those reported by previous investigators, but with less differentiation between stop place for BFP than NFM. These acoustic results corresponded to the perceptual results obtained. (That is, when presented with only auditory information, on average, listeners perceived BFP's target /b/ syllables to be near the boundary between /b/ and /d/.)

Temporal and spectral aspects of coarticulation in ataxic dysarthria: an acoustic analysis

In order to analyze the impact of cerebellar disorders on temporal and spectral aspects of coarticulation, 9 individuals with cerebellar dysfunction and 9 controls were asked to produce test sentences comprising a target vowel (V = [a], [i], or [u]) within a schwa-t-V-t-schwa environment. The control speakers were investigated both at their habitual speech tempo and under a slow speaking condition. The squared distances between averaged FFT spectra served as a quantitative estimate of target-induced coarticulation, a method that can be applied to consonants as well as vowels, and which avoids the shortcomings of formant analysis. In order to test the significance of coarticulation effects at the level of individual speakers and to obtain F values as a further measure of the strength of coarticulation, multivariate tests of target effects were performed, with the first 6 principal components derived from the spectra of each speaker. First, inconsistent patterns of anticipatory vowel-to-vowel (W) interactions emerged across individuals, and neither significant group differences nor any effects of speech rate could be detected. The underlying control mechanisms thus seem to be segment-linked subject-specific specifications of the pretarget vowel, uncompromised by cerebellar dysfunction. Second, all participants exhibited highly significant anticipatory [t(h)]-to-vowel (CV) coarticulation. This effect was slightly smaller in the cerebellar group than in the control group, which can be at least partially explained by reduced spectral distances among the 3 target vowels. Speech rate did not influence the CV effects of the control group. As concerns temporal aspects of coarticulation, no significant group differences emerged in terms of length adjustments of the pretarget consonant to the intrinsic duration of the target vowel. Third, ataxic speakers showed a tendency toward enlarged perseverative vowel-to-[t(h)] (VC) and W effects if their slow speech rate was taken into account. Retentive coarticulation turned out to be similar in slow ataxic speakers and in fast-speaking controls. However, significant attenuation of these effects emerged in the latter group under the condition of decreased speech tempo. In summary, these results corroborate the suggestion of different mechanisms of gestural overlap in the temporal domain: Whereas perseverative coarticulation, presumably, reflects biomechanical or motor constraints, anticipation seems to represent higher level phonetic processing.

Speaking-rate-induced variability in F2 trajectories

This study examined speaking-rate-induced spectral and temporal variability of F2 formant trajectories for target words produced in a carrier phrase at speaking rates ranging from fast to slow. F2 onset frequency measured at the first glottal pulse following the stop consonant release in target words was used to quantify the extent to which adjacent consonantal and vocalic gestures overlapped; F2 target frequency was operationally defined as the first occurrence of a frequency minimum or maximum following F2 onset frequency. Regression analyses indicated 70% of functions relating F2 onset and vowel duration were statistically significant. The strength of the effect was variable, however, and the direction of significant functions often differed from that predicted by a simple model of overlapping, sliding gestures. Results of a partial correlation analysis examining interrelationships among F2 onset, F2 target frequency, and vowel duration across the speaking rate range indicated that covariation of F2 target with vowel duration may obscure the relationship between F2 onset and vowel duration across rate. The results further suggested that a sliding based model of acoustic variability associated with speaking rate change only partially accounts for the present data, and that such a view accounts for some speakers' data better than others.

The effect of speaking style on a locus equation characterization of stop place of articulation

Locus equations were employed to assess the phonetic stability and distinctiveness of stop place categories in reduced speech. Twenty-two speakers produced stop consonant + vowel utterances in citation and spontaneous speech. Coarticulatory increases in hypoarticulated speech were documented only for /dV/ and [gV] productions in front vowel contexts. Coarticulatory extents for /bV/ and [gV] in back vowel contexts remained stable across style changes. Discriminant analyses showed equivalent levels of correct classification across speaking styles. CV reduction was quantified by use of Euclidean distances separating stop place categories. Despite sensitivity of locus equation parameters to articulatory differences encountered in informal speech, stop place categories still maintained a clear separability when plotted in a higher-order slope x y-intercept acoustic space.

The role of perception in the sound change of velar palatalization

Voiceless velar stops may become palatoalveolar affricates before front vowels. This sound change is not only one of the most common types of palatalization, but is a very common sound change in the world's languages. Nevertheless, we do not have an adequate understanding of how this sound change takes place. Three experiments reported here test the hypothesis that velar palatalization is the result of listeners' on-line perceptual reanalysis of fast rate speech. It is shown that velars before front vowels are both acoustically and perceptually similar to palatoalveolars. This supports the proposal that velar palatalization is perceptually conditioned.

On the relevance of locus equations for production and perception of stop consonants

We examined the possible relevance of locus equations to human production and perception of stop consonants. The orderly output constraint (OOC) of Sussman, Fruchter, and Cable (1995) claims that humans have evolved to produce speech such that F2 at consonant release and F2 at vowel midpoint are linearly related for consonants so that developing perceptual systems can form representations in an F2ons-by-F2vowel space. The theory claims that this relationship described by locus equations can distinguish consonants, and that the linearity of locus equations is captured in neural representations and is thus perceptually relevant. We investigated these claims by testing how closely locus equations reflect the production and perception of stop consonants. In Experiment 1, we induced speakers to change their locus equation slope and intercept parameters systematically, but found that consonants remained distinctive in slope-by-intercept space. In Experiment 2, we presented stop-consonant syllables with their bursts removed to listeners, and compared their classification error matrices with the predictions of a model using locus equation prototypes and with those of an exemplar-based model that uses F2ons and F2vowel, but not locus equations. Both models failed to account for a large proportion of the variance in listeners' responses; the locus equation model was no better in its predictions than the exemplar model. These findings are discussed in the context of the OOC.

The perceptual relevance of locus equations

Identification curves were estimated for the English consonants /b,d,g/ using five-formant CV synthetic stimuli comprehensively sampling the F2 onset-F2 vowel acoustic space in the vicinity of /b,d,g/ locus equations [H. Sussman et al., J. Acoust. Soc. Am. 90, 1309-1325 (1991)]. The stimuli included 10 English monophthongal vowel contexts, 11 levels of F2 onset per vowel, and 3 levels of F3 onset orthogonally varied with the F2 variables (10 vowels x 11 F2 onsets x 3 F3 onsets = 330 stimuli). After brief training, each of six subjects, three male and three female, was presented eight trials of each of the stimuli, one or two trials per day over a period of several days. Systems of identification curves were visualized as identification surfaces situated in locus equation acoustic space and were overlaid with acoustic data from five male speakers in order to judge the degree of correspondence between perception and acoustic data. A chi square analysis was also performed in order to quantify the correspondence between the observed perception data and expected frequencies derived from the acoustic data. The results, when interpreted in terms of a dominance hierarchy hypothesis, strongly indicate F2 onset and F2 vowel, in combination, serve as important cues for stop consonant place of articulation.

Speech perception as pattern recognition

This work provides theoretical and empirical arguments in favor of an approach to phonetics that is called double-weak. It is so called because it assumes relatively weak constraints both on the articulatory gestures and on the auditory patterns that map phonological elements. This approach views speech production and perception as distinct but cooperative systems. Like the motor theory of speech perception, double-weak theory accepts that phonological units are modified by context in ways that are important to perception. It further agrees that many aspects of such context dependency have their origin in natural articulatory processes. However, double-weak theory sides with proponents of auditory theories of phonetics by accepting that the real-time objects of perception are well-defined auditory patterns. Because speakers find ways to "orderly" output conditions" (Sussman et al., 1995), listeners are able to successfully decode speech using relatively simple pattern-recognition mechanisms. It is suggested that this situation has arisen through a stylization of gestural patterns to accommodate real-time limits of the perceptual system. Results from a new perceptual experiment, involving a four-dimensional stimulus continuum and a 10-category/hVC/response set, are shown to be largely compatible with this framework.

An investigation of stop place of articulation as a function of syllable position: a locus equation perspective

Locus equations were employed to phonetically describe stop place categories as a function of syllable-initial, -medial, and -final position. Ten speakers, five male and five female, produced a total of 2700 CVC and 4500 VCV utterances that were acoustically analyzed to obtain F2 onset, F2 vowel, and F2 offset frequencies for locus equation regression analyses. In general, degree of coarticulation, as indexed by locus equation slope, was reduced for post-vocalic (VC) stops relative to pre-vocalic stops (pooled data from initial and medial positions), but significant differences were observed as a function of stop consonant. All stops showed significantly reduced R2 values and increased standard errors of estimate for VC relative to CV productions. Separability of stop place categories in a higher-order slope X y-intercept acoustic space also diminished for VC vs CV stop productions. The degradation of classic locus equation form (high correlation and linearity) for VC relative to CV productions was attributed to greater articulatory precision in the production of pre-vocalic compared to post-vocalic stops. This greater articulatory precision was interpreted as reflecting a greater need to normalize vowel context-induced variability of the F2 transition for syllable onset relative to final stops. The decline in acoustic lawfulness of syllable-final stops is discussed in terms of coarticulatory interactions and expected perceptual correlates.

Locus equations as phonetic descriptors of consonantal place of articulation

This experiment tested whether locus equation coefficients, slope, and y-intercept could serve as indices of place of articulation for obstruents sharing the same place of articulation across different manner classes. Locus equations for 22 speakers were derived from CV/t/ words with initial voiced stop /d/, voiceless aspirated stop /t/, nasal /n/, voiced fricative /z/, and voiceless fricative /s/ preceding 10 vowel contexts. Post hoc tests revealed /d/ = /z/ = /n/ for slope means. Voiced /d/ and voiceless /t/ were also equivalent when F2 transition onset measurement points were equated. Scatterplots of locus equation coefficients revealed three nonoverlapping and distinct clusters when the diverse coronal group was compared with labials and velars. A discriminant analysis using slope and y-intercept as predictors successfully categorized all five coronals into one alveolar group with 87.1% accuracy. The collective results support the contention that locus equations can serve as effective phonetic descriptors of consonant place of articulation across manner classes.

Consonant-vowel interdependencies in babbling and early words: preliminary examination of a locus equation approach

Consonant-vowel productions at two distinct stages of language development were studied in a single female child. At 12 months canonical babbling syllables (N = 144) identified by a panel of listeners as comprising [bV], [dV], and [gv] tokens were acoustically analyzed by measuring F2 transition onset and F2 midvowel frequencies and plotting their relationship as locus equations for each stop category. A regression analysis performed on these scatterplots revealed differential slopes and y-intercepts as a function of stop place. The same analysis was performed 9 months later on CV utterances (N = 243) produced as syllable-initial segments of real words by the same child. Whereas labial and velar locus equation parameters moved toward more adult-like values, alveolar slope and y-intercept moved away from adult values and more in the direction of decreased coarticulation between vowel and consonant. There was greater scatter of data points around the regression line for production of words compared to babbling. These results are compared to locus equations obtained from 3-5-year-olds and adults. Locus equations appear to be useful as an empirical developmental probe to document how CV productions gradually approach adult categorical standards.