Intrinsic Pitch in Opening and Closing Diphthongs of German

Perception experiments using rising and falling F0 slopes in 4 German word-final opening and closing diphthongs show that closing/opening diphthongs support the perception of falling/ rising pitch movements. The latter effect is suggested to be enhanced when the initial close vowel quality is retracted and rounded. Hence, the current knowledge of the relationship between intrinsic pitch and vowel quality seems transferable to diphthongs. The estimated intrinsic pitch values of +1.5% are better explained by a psychoacoustic pitch-shift than by a process of speech perception compensating for intrinsic F0. 1. Research Questions This paper addresses two questions. The first one is, if and in what way intrinsic pitch (as opposed to intrinsic F0) occurs in opening and closing diphthongs of German. If intrinsic pitch was found experimentally, then the second question must be, whether this phenomenon is due to a controlled compensation process of speech perception or the result of a pitch-shift characteristic of the perception of any kind of complex tone. To the first question, a number of studies from different language backgrounds agree that the pitch of a vowel is not exclusively due to its fundamental frequency, e.g. [1,2,3]. It is also influenced by the location of the vowel in the three dimensional vowel space. The aperture dimension is of major importance. The more open a vowel the higher it is perceived when compared with a close vowel at the same fundamental frequency. The pitch differences are larger when the close vowel is more rounded and/or retracted than the open one. Inasmuch as this contribution to the pitch of a vowel is inherent in the vowel itself, it is referred to as intrinsic pitch. So far, our knowledge of the existence and the amount of intrinsic pitch is restricted to monophthongs, i.e. subjects had to judge different vowel qualities as part of separate stimuli, interrupted by pauses and/or other segments. It cannot be assumed that our knowledge gained from such a task is simply transferable to diphthongs, where different vowel qualities are to be judged as part of a continuous articulatory and acoustic transition in a holistic unit. Moreover, an acoustic analysis of German closing diphthongs by [4] shows that, in contrast to the onset, stable phases at the offset (if there are any) need not exist for F1 and F2. Further, onset and offset have unequal durations of stable phases and target values in F1 and F2 that differ from those found in monophthongs. Provided our knowledge of intrinsic pitch in monophhongs is applicable to diphthongs, the following hypotheses are tested, using closing and opening diphthongs of German: Since a diphthong mainly consists of a continuous transition from one vowel quality to another, closing diphthongs lead to a gradual decrease of intrinsic pitch. Thus, closing diphthongs support the perception of falling pitch contours. This effect is stronger when the close vowel quality at the offset is retracted and rounded. On the other hand, opening diphthongs lead to a gradual increase of intrinsic pitch and support the perception of rising pitch contours. Analogously, this effect is more pronounced for a back rounded vowel at the onset. As regards the second question, a high negative correlation was found (cp. [1,3]) between intrinsic pitch and intrinsic F0 values for corresponding monophthongs measured in various languages, e.g. [5,6,7]. Higher/lower intrinsic pitch is connected with lower/higher intrinsic F0. Although the reason for this intrinsic F0 is still controversial, a mechanical linkage between the structures of the larynx and other components of the vocal apparatus involved in the production of vowels (esp. the tongue) is the most probable explanation (cp. [2]). Following this physiological explanation, intrinsic F0 is a highly uncontrollable but consistent phenomenon. Combined with the negative correlation, this led to the hypothesis put forward by [2] that intrinsic pitch is caused by a perceptual process parsing F0 into its intentional and intrinsic component. Since the intrinsic component is used as a cue for segment identification, it is compensated for at the prosodic level. However, this hypothesis is not supported by the majority of studies dealing with intrinsic pitch. The values found for intrinsic pitch in these studies are much smaller (at most 3% of the F0 value, see [8] for an overview) than those measured for intrinsic F0 under comparable conditions. These intrinsic pitch values are more adequately explained by the general psychoacoustic mechanism of the virtual pitch model [9] valid for any kind of complex tones of which vowels are a subset. In this model, the pitch of a vowel is calculated with reference to its F0 and spectral envelope. Different spectral envelopes, especially F1 but also F2, cause small but perceptible pitchshifts which are quite similar to the empirical values found for intrinsic pitch. Nevertheless, some studies, e.g. [10], found values for intrinsic pitch, too large to be explained by pitch-shift alone, thus supporting the hypothesis of compensation. The empirical variations in intrinsic pitch values may be due to the strong variation between subjects reported in every study. Alternatively, the larger intrinsic pitch values may also result from judging (nonsense) words containing the relevant vowels instead of comparing the pitch of synthesized vowels in isolation. The former experiments have a closer relationship to speech and might therefore activate more or different perceptual processes than the latter, more psychoacoustic experiments. In order to test this hypothesis, resynthesized German words were used in the experiments of this study. Whether the results point to a process of compensation or pitch-shift is a matter of the amount of intrinsic pitch found. However, it is not the primary aim of this study to quantify intrinsic pitch in diphthongs. Speech Prosody 2004