F2 Transition Research Articles

The role of rate of transition in the perception of place of articulation

Listeners can reliably identify the place of articulation of bilabial and alveolar stops in severely truncated CV syllables from just the initial 20 ms, but they need 40 ms or more to identify a velar. The need for a longer interval may reflect listeners' expectations that rate of transition should be slower for velars because of the slower opening gesture at this place of articulation [Kewley‐Port et al., J. Acoust. Soc. Am. 73, 1779–1783 (1983)]. This paper investigates the perceptual effects of differences in rate of transition in two experiments. First, synthetic CV syllables, where the direction and magnitude of the F1, F2, and F3 transitions matched natural [ba,da,ga] syllables, but where the transition durations were varied in 15‐ms steps between 15 and 120 ms were presented to listeners for two‐interval, forced‐choice discrimination and for identification along a stop—glide dimension. If listeners expect longer transitions with velars, then they should accept stimuli with longer transitions as stops when the formants indicate a velar than when they indicate a bilabial or alveolar. Second, variation in F2 and F3 transitions between values appropriate for [da] and values appropriate to [ga] were combined with variation in the rate of transition; rate of transition was varied in 5‐ms steps between 20 and 60 ms. These stimuli were also presented to listeners in two‐interval, forced‐choice discrimination and identification of place of articulation tasks. It is expected that faster rates of transition will shift listeners' response to stimuli with F2 and F3 transitions near the [d]/[g] boundary toward [d] and slower transitions will shift them toward [g].

Analysis of formant transitions for English fricatives spoken by five male and five female talkers

Nonsense words of the form C′VC′VC′VC were recorded for the four voiceless fricatives of English [f,th,s,sh] in ten vowel environments by each of five male and five female talkers. The database was digitized at 16 000 samples/s and digital spectrograms were produced. Formant frequency values, F2 and F3, were measured by hand from the spectrograms at voicing onset and vowel midpoint for each CV transition. These data were examined in order to seek regularities of several possible forms: (1) invariant onset frequencies independent of speaker and vowel environment, (2) existence of locus frequencies that differ as a function of place of fricative articulation, but are similar across speakers, and (3) similar transition behavior (e.g., rising or falling F2 and/or F3) across speakers for a given vowel with no attempt to seek regularities across vowels. In all cases, distributions of transition-related measures across speakers for each place of articulation show disturbing amounts of overlap, even when male and female data are plotted separately. While it has often been said that [f] and [th] are distinguished perceptually by F2 and F3 transition behavior (because frication spectra are often quite similar), the data suggest that it will be difficult for a speech recognition device to make this distinction reliably from observed formant transition behavior in the general population. Some reassessment of the appropriate perceptual cues seems to be required. [Work supported in part by an NIH grant.]

Frequency discrimination and consonant recognition in the hearing impaired

The frequency difference limen (DL) for short‐duration second‐formant transitions (F2) presented alone and in speechlike environments was examined in normal hearing and hearing‐impaired subjects. Four stimulus conditions were included: F2 transition alone; F2 transition in the presence of all formants (full formant); full‐formant stimuli preceded by a burst and followed by a vowel; and full‐formant stimuli preceded and followed by a vowel. For the simplest stimulus condition (F2 alone), all hearing‐impaired subjects had DLs within the 95% confidence interval around the mean of the normals. For all other conditions, however, the hearing impaired demonstrated extreme intersubject variation. Some hearing‐impaired subjects could not detect a formant transition of even 800 Hz, whereas others continued to perform like normal hearers. Consonant identification performance was also examined in CV and VCV environments. Consonants whose identification is believed to involve formant transitions were selected for study. In spite of modifications to the frequency discrimination task and use of relevant syllables, relatively weak correlations continue to be observed between F2 DL and consonant recognition ability, as well as between F2 DL and degree of hearing loss. [Work supported by NIH.]

Acoustic study of Russian vowel allophones

The five Russian vowels vary in auditory quality depending on their consonant contexts; one traditional reference gives several major and minor impressionistic allophones for each stressed vowel. The consonants are reported to be relatively unaffected by their vowel contexts, and to block vowel‐to‐vowel interactions. This paper presents a study of Russian vowels in CV, VC, and CVC syllables for three speakers. Preliminary results for vowels /i/ and /a/ across a wide variety of contexts suggest that the allophonic variation is of three kinds. First, there is target variation: Each of these vowels has one target in some contexts, and a second target in other contexts. Second, some additional vowel variation is due to undershoot of those targets after extensive F2 and F3 transitions. Finally, however, in many contexts the formants do reach their targets, despite the extensive transitions. In these cases, the perceived vowel quality variations seem to arise from differences in the transitions alone. [Work supported by NSF.]

Perception of Persian uvular and velar stop consonants by speakers of American English

Recent cross‐language studies by Werker and Tees [J. Acoust. Soc. Am. 75, 1866–1878 (1984)] describe difficulties in English speakers' perception of a Thompson glottalized uvular‐velar stop contrast. However, the acoustic information that differentiated the contrasting phones was difficult to specify. The present study extends the examination of non‐English place contrasts by investigating native English speakers' ability to perceive a nonglottalized uvular‐velar contrast in Persian (Farsi). Multiple instances of velar and uvular stop consonants produced in several VCV contexts by a native speaker of Farsi were presented. Correct identification of the medial consonant varied systematically as a function of vowel context from very few errors to near change performance. Acoustic analysis revealed systematic F2 and F3 formant transition differences between uvular and velar consonants produced in the same vowel context. Variations in voicing offset (murmur) were also observed in several vowel contexts. These results further our understanding of the role of linguistic experience in the perception of phonetic categories. [Research supported by NINCDS.]

A truncation model for the F2 trajectory in syllables with complex vocalic components

Phonetic orthodoxy treats the acoustic realization of a CVn, syllable (N.B. V = vowel or glide) as the concatenation of consecutive transitions from one element to the following one in the syllable. This study proposes an alternative view of the acoustic structure of this type of syllable, based on LPC trackings of F2 in sets of selected Chinese and English CVn syllables. The basic hypothesis is that the underlying F2 transitions between any two phonologically adjacent elements in the same syllable (e.g., the C → V1, V1 → V2,…) all originate at the same temporal position—the syllable initiation. Each transition has a specified rate. The acoustic realization will be a programmed “truncation” process; that is, the phonologically preceding transition truncates the phonologically following transition (e.g., the C → V1 transition truncates the V1 → V2 transition; the truncated V1 → V2 in turn truncates the V2 → V3 transition). This model provides elegant accounts and quantitative predictions of the undershooti...

Rate‐dependent variability in English and Japanese complex vowel F2 transitions

Are the effects of changes in speech rate on F2 transition slopes in languages with and without unitary diphthongs explicable in terms of general phonetic and physiological principles? Five American English diphthongs and four Japanese vowel sequences were recorded, spoken at three different tempos by eight and seven speakers, respectively. LPC spectra were computed at 10‐ms intervals, and the resulting F2 contours were analyzed by a computer program designed to objectively define and measure steady‐state and transitional portions of the vowel. Contrary to the claims of Gay [J. Acoust. Soc. Am. 44, 1570–1573 (1968)], speech rate was found to significantly affect two measures of transition slope in English, with slope decreasing as speech rate increased. Individual diphthong populations in both languages displayed high correlations between a transition, slope and its F2 range, indicating that rate adjustment is not simply a matter of target undershoot at faster speeds. Japanese showed less rate‐dependent variability, suggesting that temporal reorganization for different speech rates is affected by language‐specific structures. [Work supported by NIH.]

Millimeter-wave optical double resonance spectra of NO2: How good a quantum number is N?

Optical transitions to the 2B2 electronic state region of NO2 have been observed in double resonance with three millimeter-wave ground state rotational transitions. The probed ground state transitions were the N=2 (F1), N=4 (F1), and N=10 (F2), K=0→1 Q branch transitions near 250 GHz. The optical range scanned was 16 800 to 16 950 cm−1. The 100,10–101,9 F2 spectrum was combined with previous double resonance results with the ground state 91,9–100,10 F1 transition. This comparison shows 70 out of 160 upper states with J=19/2 to have indefinite or mixed N values. It establishes violations of the ΔN=ΔJ selection rule in this spectral region which contains bands with a wide range of spin-rotation constants. Lack of intensity correlation between the two spectra suggests two classes of upper states are present: a group of strongly N-mixed upper states, and a group essentially unmixed in N. Heller’s F parameter, which measures the fraction of all possible symmetry-restricted basis states contributing to a typical eigenstate, is similar for the double-resonance spectra of the three N values, but is greater for K=1 than for K=0. By finding upper states in common, R lines from N=2 and P lines from N=4 have been identified.

Duration effects on labeling of tonal analogs to F1‐cutback continua

F1 cutback has been demonstrated to be at least a contributing cue, and very possibly a major cue, for the perception of voicing contrasts. In an attempt to evaluate the feasibility of there existing an acoustic basis for the role of F1‐cutback cues in the perception of voicing stimuli, our recent research has focused on comparing stimulus cue dependencies for voicing boundaries with similar stimulus changes in analogous tonal stimuli varying in complexity. An earlier paper [Pastore et al., J. Acoust. Soc. Am. 75, S65 (1984)], described parallel labeling boundaries as a function of tonal analogs to F1 frequency and F2 transitions, but no stimulus duration. We now have identified conditions under which our F1‐cutback analog stimuli exhibit both labeling boundary locations, and changes as a function of duration, which are characteristic of voicing boundaries for speech stimuli. [Supported in part by NSF.]

Trading relations in the perception of /r/‐/l/ by articulation‐delayed children

Articulation‐delayed children who did not produce /r/ correctly and who showed slow progress in therapy, were tested on identification of two synthetic series contrasting prevocalic /r/ and /l/ in a trading relations paradigm. The primary spectral cue (F3 and F2 onset and transition) was varied in ten steps in both series. For one series, the secondary temporal cue (F 1 steady‐state and transition duration) was /r/‐like; in the other, it was /l/‐like. Result showed that phoneme boundaries along the spectral dimensions shifted appropriately as a function of the temporal cue value (i.e., a trading relation was demonstrated). However, the shift was somewhat smaller than for adults, and both boundaries were displaced toward the /l/ end of the spectral dimension, relative to the adults. These results suggest that such children have particular difficulty with the perceptual differentiation of spectral information, as might be predicted from their articulatory substitution patterns. [Supported by NIMH, NINCDS.]

Effect of formant transition rate on the differentiation of synthesized child and adult (w) and (r) sounds.

The purpose of this research was to assess the perceptual effects of a range of second (F2) and third (F3) formant transition rates that occur naturally in the production of /w/ and /r/ by children and adults. Synthesized CV continua that varied in the second (F2) and third (F3) formant onset frequencies between values appropriate for /w/ and /r/ were used as stimuli. Subjects participated in four experimental conditions that involved changing the rate of transition of either F2 or F3 by varying the duration of the transition between the glide onset and an /eI/ vowel nucleus for child and adult stimuli. In each condition, the transition rates of the /w/-endpoint stimulus, the /r/-endpoint stimulus, and at least one of the midpoint stimuli were varied across values appropriate for /w/ and /r/. Mean ratings of the stimuli were compared to test the predictions that /w/ perception increases with increased F2 transition rate and increases with decreased F3 transition rate. The results were as follows: (a) one of the 10 comparisons for the child F2 stimuli was significant, but it involved a change opposite to the predicted direction; (b) four of the six comparisons for the child F3 stimuli were significant, but they involved changes opposite to the predicted direction; (c) 14 comparisons of adult F2 stimuli were significant, but 12 of these 14 comparisons involved changes opposite to the predicted direction; and (d) four of the six comparisons for the adult F3 stimuli were significant, but none of them involved changes in the predicted direction. Only one of all the comparisons involved a significant change in rating between the /w/ and /r/ categories.

F1 transition as a cue to [± voice] in a following utterance‐final stop

Previous studies have shown that the cues of [± voice] in an utterance‐final stop produced without release or vocal vibration during closure include duration of the previous vowel and presence or absence of a falling F1 transition [T. Walsh and F. Parker, J. Phonet. 12, 207–218 (1984)]. Raphael [J. Acoust. Soc. Am. 51, 1296–1303 (1972)] employed synthetic speech to measure the perceptual effect of varying vowel length while holding F1 slope constant. The present study is intended as a complement to Raphael. That is, our study measures the perceptual effect of varying F1 slope at given vowel durations. The stimuli consist of three‐formant tokens of bat/bad, synthesized without release or closure voicing for the final stop. The vowel /ae/ was chosen because of its high F1, which would allow maximum manipulation of the F1 transition. Each token is synthesized with three durations, corresponding to those in Raphael which cued, respectively, [−voice], mixed, and [+ voice] responses. For each duration tokens are...

Perception of initial glides by impaired‐ or normal‐hearing listeners

An experiment was conducted to study the use of cues to glide consonant identification by hearing‐impaired (n = 22) and normal‐hearing (n = 6) listeners. The syllables /wæk, jæk, bæk, dæk, gæk, æk/ (six different utterances each) were identified for three conditions: (1) unmodified, (2) initial glide murmurs or stop bursts removed, and (3) syllables cut back to the ninth pitch period from the vowel onset. The normal group, listening at 75 or 95 dB SPL, showed 100% glide phoneme perception for the condition of deleted glide murmurs or stop bursts. For this condition performances varied greatly within the hearing‐imparied group (listening at MCL): Near‐normal glide phoneme perception was seen for less than 20% of the impaired listeners (mean 3FA = 72 dB HL), indicating good use of the vowel onset transitions for glide perception by these listeners. Between the conditions: Unmodified versus glide‐murmur/stop‐bursts deleted, the majority of the hearing‐impaired listeners (mean 3FA = 88 dB HL) manifested degraded manner and/or place and/or phoneme perception. For these listeners reduced perception of the duration and/or frequency extent of F1 and/or F2 vowel onset transitions was suggested by the types of consonant confusions evidenced. About 20% of the impaired listeners (mean 3FA = 92 dB HL) showed no differences in performance among the three conditions, indicating inability to distinguish the presence/absence of vowel onset transitions that characterize initial glides.

Trading relations in the perception of /r/‐/l/ by Japanese learners of English

Native English speakers show a trading relation and perceptual integration of a temporal cue (F1 transition duration) and spectral cues (F3 and F2 onset and transitions) differentiating the liquids /r/‐/l/ [L. Polka and W. Strange, J. Acoust. Soc. Am. Suppl. 1 73, S53 (1983)]. To investigate the effects of linguistic experience on these perceptual phenomena, native Japanese speakers learning English completed forced‐choice identification and oddity discrimination tasks, using two synthetic /rak/‐/lak/ series in which the temporal and spectral cues were varied orthogonally. Identification functions of Japanese who could consistently label the /r/‐/l/ series showed a trading relation (a predictable shift in the identification category boundary) and perceptual integration of temporal and spectral cues. Discrimination of pairs differing on both cues was significantly better than spectral‐cue‐alone pairs only when the combination of cue values facilitated phonemic differentiation. Preliminary results with less advanced Japanese English learners showed no advantage in discrimination performance of two‐cue pairs over one‐cue pairs, suggesting that the perceptual integration of cues is a function of language‐specific phonetic processing. [Work supported by NIMH.]

A measurement of the vibrational band strength for the ν3 band of the HO2 radical

The HO2 radicals generated in a discharge-flow system have been observed with tunable diode laser absorption in the P branch of the ν3 vibrational band at 1080 cm−1. The observed line positions agree within (2 to 5)×10−3 cm−1 with those calculated from the molecular constants for the ν3 band obtained from a previous study using laser magnetic resonance spectroscopy. The band strength has been determined by observing line center absorptions when HO2 is produced in the reaction F+H2O2→HO2+HF (k1) with a measured concentration of atomic fluorine and excess hydrogen peroxide. F-atom concentrations are measured by diode laser absorption of the spin-orbit transition at 404 cm−1. The analysis accounts for HO2 losses due to the reactions of HO2+HO2→H2O2+O2 (k3) and F+HO2→HF+O2 (k4). The experimental data are consistent with a value of k3=(1.6±0.3)×10−12 cm3 s−1 and a ratio k4/k1=1.0±0.4. The line strength for the 615←716 F1 transition is 2.9×10−21 cm2 molecule−1 cm−1 which corresponds to a ν3 band strength of 35±9 cm−2 (STP atm)−1. This value is a factor of 1.6 to 6 lower than previous ab initio calculations.

Developmental speech perception of three acoustic cues associated with place of articulation

In this experiment, 30 children and 10 adults participated in three speech perception tasks: discrimination, labeling, and adaptation. Stimuli were four sets of synthetic CV syllables, varying along a bilabial‐to‐alveolar, place‐of‐articulation continuum. The primary acoustic cue in each continuum was the change in slope of the F2–F3 transitions. The continua were constructed so that two had transition lengths of 45 ms, and two of 95 ms. Two continua contained a 5‐ms burst, and two were burstless. The discrimination task was a change‐no change procedure, in which subjects indicated whether a set of four stimuli remained the same or changed. Results indicated a complex developmental pattern. For discrimination, there was a progression in response strategy and sensitivity with increasing age. In contrast, labeling performance was similar for all subjects. Finally, only the adult subjects showed significant adaptation effects. Children's responses were essentially unchanged after adaptation. Further, transit...

Revisiting stop‐consonant perception for two‐formant stimuli

The aim of this study was to re‐examine the factors leading to stop consonant perception for consonant‐vowel stimuli with just two formants over a range of vowels, using stimuli and test procedures somewhat different from those used in earlier studies. Five two‐formant CV stimulus continua were synthesized, each covering a range of second‐formant (F2) starting frequencies, for vowels corresponding roughly to [iiæɑu]. Listeners identified the beginning of the syllable under two different instructions: a forced‐choice response of b, d, or g, and an open response set. The stimuli with the most unanimous responses were those with substantial F2 transitions, i.e., [bi bɪ du]. There were few consistent g responses. Under open response instructions, stimuli with straight or small transition ranges were identified as beginning with no consonant. The results are compared with those of other studies, particularly with regard to differences in stimulus characteristics, and are discussed in relation to theories of invariance. [Supported in part by grants from NIH.]

Perception of sine-wave analogs of voice onset time stimuli.

It has been argued that perception of stop consonant voicing contrasts is based on auditory mechanisms responsible for the resolution of temporal order. As one source of evidence, category boundaries for nonspeech stimuli whose components vary in relative onset time are reasonably close to the labeling boundary for a labial stop voiced-voiceless continuum. However, voicing boundaries change considerably when the onset frequency of the first formant (F1) is varied--either directly or as a side effect of a change in F1 transition duration. Stimuli consisted of a midfrequency sinusoid that was initiated 0-50 ms prior to the onset of a low-frequency sinusoid. Results showed that the labeling boundary for relative onset time increased for longer durations of a low-frequency tone sweep. This effect is analogous to the F1 transition duration effect with synthetic speech. Further, the discrimination of differences in relative onset time was poorer for stimuli with longer frequency sweeps. However, unlike synthetic speech, there were no systematic effects when the frequency of a transitionless lower sinusoid was varied. These findings are discussed in relation to the potential contributions of auditory mechanisms and speech-specific processes in the perception of the voicing contrast.

Propensity rules in rotationally inelastic collisions of diatomic molecules in 3Σ electronic states

The formalism for the treatment of rotationally inelastic collisions of molecules in 3Σ electronic states is exposed, for the general case of intermediate coupling. The reduced matrix elements of the T operator are expressed in a Hund’s case (a) basis. Within the infinite-order-sudden (IOS) limit, the symmetry of the reduced T-matrix elements and the large-J limit of certain vector coupling coefficients can be used to derive several propensity rules bearing on changes in the Fi(i=1–3) label. In particular, for large J only the Fi conserving transitions will have large cross sections. From a physical viewpoint this propensity rule implies that the scattering, at least in the IOS limit at large J, is independent of Σ, the spin projection quantum number. The experimental study by Caughey and Crosley [J. Chem. Phys. 71, 736 (1979)] of rotational relaxation in the B 3Σu− state of S2 confirms our theoretical propensity rules, although the experimental rate constants for the F1 → F2, F3 transitions are considerably larger than would be predicted theoretically.

The role of "chirp" identification in duplex perception.

Duplex perception occurs when the phonetically distinguishing transitions of a syllable are presented to one ear and the rest of the syllable (the “base”) is simultaneously presented to the other ear. Subjects report hearing both a nonspeech “chirp” and a speech syllable correctly cued by the transitions. In two experiments, we compared phonetic identification of intact syllables, duplex percepts, isolated transitions, and bases. In both experiments, subjects were able to identify the phonetic information encoded into isolated transitions in the absence of an appropriate syllabic context. Also, there was no significant difference in phonetic identification of isolated transitions and duplex percepts. Finally, in the second experiment, the category boundaries from identification of isolated transitions and duplex percepts were not significantly different from each other. However, both boundaries were statistically different from the category boundary for intact syllables. Taken together, these results suggest that listeners do not need to perceptually integrate F2 transitions or F2 and F3 transition pairs with the base in duplex perception. Rather, it appears that listeners identify the chirps as speech without reference to the base.