Speaking Rate and Fundamental Frequency as Speech Cues to Perceived Age

Summary

This study aimed to specify a set of acoustic cues fundamental to vocal aging and to establish their perceptual relevance, using acoustic analysis and perceptual testing. Three experiments were conducted to identify the perceptual correlates of the aging voice. The first experiment analyzed important voice parameters that signal a person's age for 16 older males and 14 younger males. In the second and third experiments, these acoustic patterns were systematically shifted through resynthesis to see if perceived age would be significantly influenced. In the second experiment, the older and younger male voices were resynthesized by manipulating speaking rate and fundamental frequency to shift the perceived age of the groups toward each other. In the third experiment, older and middle-aged male voices were resynthesized in a similar manner. In both perceptual studies, an age estimation task with naive listeners was used. The results of the first experiment showed that, in older speakers, sentence, word, and diphthong durations were all significantly longer and mean fundamental frequency was significantly higher than for the younger group. In the second experiment, only the manipulation of speaking rate resulted in a significant shift in perceived age, and it did so only for the older subjects. In the third experiment, a significant shift in age estimates was observed for the middle-aged, but not the older, voices when speaking rate was manipulated. The results of both perception tests suggest that speaking rate, but possibly not fundamental frequency, is a perceptually relevant cue to age in voice.

Introduction

While the total population of elders can be shown to be increasing, the effect of the normal aging process on speech production, and how these changes are perceived by listeners, is only marginally understood. A better understanding of how aging affects speech production should contribute to general models of the anatomical and physiological consequences of this process. Speaker age is also likely to be an important “indexical” (nonlinguistic) aspect of the speech signal. Other indexical properties of speech, such as token characteristics, speaker identity, and gender, have been demonstrated, in prior research, to be stored in an integrated fashion with linguistic information (eg, phonemes, syllables, words) in long-term memory.¹ These indexical properties have been shown to play an important role in speech perception and verbal learning.² If speech perception and verbal learning are to be modeled, the role of all relevant indexical properties must be understood, and speaker age is likely to be relevant for several reasons. Important among them is the finding that listeners are relatively accurate in estimating age from speech samples.3, 4, 5, 6 Finally, research on speech and aging is vital for clinical purposes to evaluate speech disorders in the elderly population. Without a model of these characteristics for normal, healthy older individuals, it becomes difficult to isolate the consequences of different speech and/or voice pathologies in that population.

As stated, prior work has shown that listeners can accurately gauge speaker age, with correlations between chronologic and perceived age ranging from r = 0.68 to r = 0.90.3, 4, 5, 6 The acoustic cues that are presumed to signal speaker age are likely the product of age-related physiological changes to the vocal tract. Some of the physiological changes that have been identified include the lengthening of the vocal tract or oral cavity,7, 8 a reduction in pulmonary function,⁹ laryngeal cartilage ossification,¹⁰ an increased stiffening of the vocal folds,10, 11, 12, 13 and a reduction in vocal fold closure.14, 15, 16 These physiological changes have been used to predict acoustic correlates of aging in voices. These predictions can be summarized as follows:

• Fundamental frequency. Mean fundamental frequency (f₀) increases in older males while remaining constant or decreasing in older females.15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 Based on these studies, it does not appear that mean f₀ changes linearly across most of the adult life span, but rather shows a shift after late middle age. f₀ also has been shown to increase in variability with age for both sexes, indicating reduced laryngeal control.9, 19, 20, 32, 33, 34, 35 It must be noted, however, that a subset of these studies failed to find changes in f₀ variability and/or in mean f₀ as a function of age for both men and women.18, 20, 36, 37 Nonetheless, the most consistent pattern that emerges across them is one in which male and female voices coalesce with respect to mean f₀ from middle to old age—a pattern that modifies the shift observed at puberty.38, 39, 40

• Speaking rate. Speaking rate appears to slow as a function of age for both male and female speakers. This rate reduction has been observed in extemporaneous speech and read speech, including paragraphs, sentences, words, and segments.26, 29, 41, 42, 43, 44, 45 However, it has not necessarily been observed in specific phonetic cues of segments, such as voice onset time, even though an extensive range of segment cues have not been examined to date.46, 47, 48, 49 Overall, speaking rate has been consistently shown to be an acoustic cue differentiating speakers by age; it reflects a more generalized slowing in motor processes in the elderly population.⁵⁰

• Formant frequencies. Vowel formant frequencies have shown modest decreases with age.7, 8, 32, 51 This shift is presumed to be a by-product of the lowering of the vocal folds over the life span, which results in a longer vocal cavity. However, some studies have failed to confirm this pattern.8, 42, 52, 53 In particular, Xue and Hao⁸ did not observe this decrease in formant frequency in all types of vowels measured, as they should have if the predicted glottal lowering is responsible for a consistent, perceptually relevant cue signaling speaker age.

• Fundamental frequency perturbation. Investigators examining cyclic variation in f₀ in sustained vowels have reported increased f₀ perturbation in both older males and females.20, 54 However, older male speakers appear to display more f₀ perturbation than do older females.²⁰ In addition, the values observed have been modest in magnitude and their perceptibility has not been studied directly. For example, the observed values are not comparable with those measured in pathological voices. Moreover, other investigators have not found f₀ perturbation to be a significant aging cue for males³⁷ or for females.15, 32, 26 Thus, to date, studies of f₀ perturbation have not provided useful evidence suggesting that this measure is a robust cue of speaker age.

• Other cues. A few other cues also have been examined by means of acoustic studies of vocal aging. Increased shimmer in older males³⁷ and females²⁰ has been observed by some investigators. Moreover, breathiness has been found to differentiate older females from both middle-aged and young females (as measured by both harmonics-to-noise ratio¹⁵ and long-term averaged spectra¹⁶). Disfluency rate was not found to vary between young, middle-aged, and older males,⁴⁴ although Shuey⁵⁵ did observe a significant increase in speech errors due to age in some vowels and final consonants. Finally, intensity has been suggested to decrease with age because of reduced pulmonary function. This prediction has been supported by some studies,⁴⁸ but others have found no significant effect or even the opposite pattern.⁴¹

In summary, findings for most of the acoustic cues of aging, with the exception of speaking rate and (to a lesser extent) mean f₀, have not been observed to be particularly consistent across studies. Moreover, few if any perceptual experiments have been conducted to validate the relevance of any of these cues. Specifically, prior research involving perceptual tasks has been focused on either listeners' abilities to accurately gauge age3, 56 or impressionistic characteristics of aged voices.⁴ Moreover, acoustic studies of speaker age are difficult to evaluate because the perceptual relevance of a particular acoustic cue that changes (due to age) is unknown. For example, how perceptually relevant is a 100-Hz shift in the second formant frequency of the vowel /i/ in older male voices?⁸ How important is this formant lowering to perceived age relative to, say, a shift in fundamental frequency? Questions such as these cannot be answered without first validating the perceptual relevance of acoustic cues. These observations motivated the design of the studies that followed. They combine an acoustic analysis of potential cues to speaker age with a corresponding perceptual evaluation in which these cues are systematically manipulated.

The purpose of this study was to:

• Verify, if possible, certain effects of vocal aging by examining two acoustic cues cited in earlier work, namely, mean fundamental frequency and speaking rate (experiment 1).

• Determine the perceptual relevance of these cues in perception tests in which the speech materials used in the acoustic analyses are modified by resynthesis (experiments 2 and 3).

Mean fundamental frequency and speaking rate were selected because they have been shown to consistently and significantly vary due to perceived/chronologic age in a relatively large number of previous studies. Moreover, these cues can be systematically manipulated in resynthesis without introducing significant unwanted artifacts in the modified speech signals.