John W. Black

Study of Twenty‐Six Intervocalic Consonants as Spoken and Recognized by Four Language Groups

Twenty‐six intervocalic consonants were recorded by three speakers of each of four languages—Hindi, English, Arabic, and Japanese—and heard by 24 speakers of each of them. The data were treated in two ways. (1) An analysis of variance indicated that listening groups differed and that consonants were unequal in their intelligibility and showed statistically significant interactions between speakers and consonants, between listeners and consonants, and among consonants, listeners, and speakers. All speakers spoke better and all listeners listened better when saying and hearing sounds of their native language. (2) A quantitative procedure employed by Miller and Nicely was adapted to ascertain which features were retained by the listeners in their error responses and whether or not these were similar from one language group to another. All consonants were classified in a binary manner in relation to each of seven channels into which the voice communication network was subdivided. The unusual outcome lay in th...

Relationships among Fundamental Frequency, Vocal Sound Pressure, and Rate of Speaking:

Twenty males who could control their vocal effort to reach specified soft and loud vocal levels, spanning 30 db., practised and recorded three vowels and three phrases at four levels, ranging from soft to loud. Increments in vocal effort were accompanied by increase in fundamental frequency, the latter shifting upward increasingly with successive steps in sound pressure. The vocal changes that occurred from one level of speaking to another were somewhat specific to the material that was spoken. Phrases that were spoken with different amounts of vocal effort, soft to loud, were spoken at the slowest rate when said softly.

A Study of Pauses in Oral Reading of One's Native Language and English

Seventy-two students, representing equally Hindi, Spamsh, and Japanese speakers, served as experimental subjects. Each group was equally subdivided between students who were more and ones who were less proficient in aural comprehension of English. All participants recorded a philosophical essay in both their native language and in English. The statistical treatment was based upon the median duration of each speakers distribution of pauses in reading a two-minute portion of the passage. The groups of higher and lower levels of proficiency in the aural reception of English differed significantly both in the median length of pauses and the accompanying semi-interquartile range. Language groups differed in median duration of pauses, but not in semi-interquartile range. There was no difference in the length or distribution of pauses that the students used in reading in their native language and in English.

Reply to “Speaker identification by speech spectrograms: some further observations” [R.H. Bolt, F.S. Cooper, E.E. David, P.B. Denes, J.M. Pickett, and K.N. Stevens, J. Acoust. Soc. Am. 54, 531–534 (1973)]

In their letter to the editor, “Speaker identification by speech spectrograms: some further observations,” Bolt et al. [J. Acoust. Soc. Am. 54, 531–534 (1973)] express an opinion on voice identification as produced in real‐life cases by specially trained professional examiners using a combined aural and visual examination of speaker samples. The opinion of these authors is not based on personal experience or even on a direct observation of these examinations. In addition, they disregard crucial facts that strongly interact with the reliability of those positive decisions produced by professional, full‐time examiners, such as their special training and responsibility, the five possible decisions they are entitled to produce after each examination, the number of samples, and the length of time used to perform each examination. It is our contention that opinions based on feelings other than in actual experience are of little value, irrespective of the scientific authority of those who produce such an opinion.

Some effects of sentence structure on speechreading

This study compares the difficulty in speechreading various transformations of a kernel sentence. Different transformations were found to vary in their intelligibility under these conditions.

Cross-lingual measurements of interconsonantal differences

Native speakers recorded pairs of consonant-vowel (CV) and vowel (V) syllables in Japanese, American English, Serbo-Croatian, and Italian. All initial consonants of each language were paired with each other. For example, a speaker would read a-a, a-pa, a-ma, etc., through all of the consonants. The next series would commence pi-i, pi-pi, pi-bi, pi-mi, etc., through the consonants. Five vowels were used. Following typical instructions of the method of magnitude estimation, native panels of 20 to 26 listeners, all university students, individually assessed the sameness or difference of a pair of syllables. The measures were normalized and averaged for each series of pairs of syllables.Sixteen of the prevocalic sounds, including the absence of a consonant, were deemed phonetically similar from language to language. First, intercorrelations were computed among the languages with respect to the 16 judgments made of the differences between each consonant and the consonants of that language, itself included. Second, a factor analysis was made of each matrix of interconsonantal distances. Third, a cluster analysis was made of the 64 (=16×4) sounds of the four languages.

Effects of delayed sidetone, noise, and syntactic structure on the level and duration of speech

The effect of eight syntactic structures in conjunction with four levels of sidetone and three levels of noise, on the duration and loudness of sentences was measured. Three different populations of readers were used. In terms of both measures, delayed sidetone and different sentence-types significantly affected oral reading. In all circumstances a kernel sentence was affected least by the speech-retarding variables. Negative and query sentences, which uniformly contain more rather than fewer transformational nodes (in the tree-diagram system) were more adversely affected than kernel sentences. There was a striking similarity between the outcome of sidetone and noise with respect to the effects of sentence structure.

The Reading of Messages of Different Types and Numbers of Syllables Under Conditions of Delayed Side-Tone

Eight sets of phrases were constructed for oral reading. Two sets were comprised of five syllables, in one instance consonant-vowel syllables and in the other, vowel-consonant. Three sets were made to include only syllables of two, three, or four sounds and a total of either 15 or 16 sounds. Three additional sets were the same as the preceding ones except that the total number of sounds was either 20 or 21. Each group of twenty-four young male adults read a set of the phrases under 12 conditions of delayed side-tone ranging from zero delay to 0.30 sec. delay. The duration of the oral phrase was measured. The reading of phrases of vowel-consonant syllables was more adversely affected by delayed side-tone than was the reading of consonant-vowel syllables, and the disparity between the two increased as the amount of delay of side-tone was increased to 0.21 sec. Otherwise syllables and phrases of different lengths were responded to “alike,” there being no interaction between the amount of delay of side-tone and the structure of the phrase.

Predictability as related to style

Randomly chosen brief samples of language from the manuscripts of speeches by Rufus Choate, his contemporaries, and our contemporaries are compared. The principal tool is a generously modified version of Shannons letter‐prediction procedure. The modification becomes a language‐prediction procedure in which university students, singly, guess successive words and are corrected with successive letters. The generating process is treated as tracking a sequence of prose and likened to a listeners role in “communication in process.” Choates phrases were less predictable than the others, suggesting a greater surprise‐element for the listener. This quality, in turn, has been lauded by rhetoricians as a desirable property of style.

Verbal Materials for Adult Aphasics in a Letter-Prediction Task

Summary The principal assumption under test was that redundant spaces, as determined by unanimity of correct responses by 10 normal adults, are equally simple for aphasic persons to predict. This is not the case. Paragraphs may be predicted in the same manner and be unequal in difficulty as measured by correct responses. This difference is not revealed by pooling the relative uncertainty of equally spaced data points in different paragraphs and using uncertainty or diversity among the 20 responses per data point as the criterion. This is not inconsistent with the earlier finding. If the spaces are unequal in ease of prediction, then they would probably be unequal in the diversity of the results, space by space. As these inequalities are pooled the unsystematic differences would tend to cancel each other. The basic fact is that a redundant space, based on 10 normal predictors, is not a tenable generalization, at least when extended to aphasic predictors. This does not alter the usefulness of the material for repeated measures and for comparisons with counter balanced stimuli.