The purpose of this study was to investigate some of the acoustic properties of Mandarin tones in citation form and in spontaneous running speech. The experiments described in Chapters 2-5 are designed to investigate the primary acoustic correlates for tones in Mandarin, as well as the interaction between tones and intonation.

The production data can be summarized as follows. Mandarin syllables produced in citation forms have best demonstrated the fundamental frequency patterns of tones hypothesized in phonological studies. These distinct tonal patterns, however, are not always present in spontaneous running speech. When produced in citation forms, the intrinsic duration of the vowels varies according to different tonal patterns, yet such distinctions in duration were not maintained in spontaneous running speech. This result suggested that duration may be a concomitant acoustic correlate for tones. Our study of the production of spontaneous running speech demonstrated a low correlation between the theoretical prediction of lexical tones and the actual acoustic information. This result suggested that in the production of spontaneous running speech in verbal communication, the major acoustic correlate needed for producing tones, i.e., the fundamental frequency patterns, was only partially present. Speakers’ integrated knowledge of the linguistic system enables them to produce only partial acoustic information of lexical tones in a sentence in spontaneous running speech. The perceived phonetic output of sentences is thus the ultimate consequence derived from a complicated interaction among the different levels of linguistic/extra-linguistic knowledge involved. Phonology and phonetics alone are not sufficient in predicting the actual output of a sentence. In establishing the context, there appears to exist some trade-off effect among all the interacting linguistic/extra-linguistic levels of information. Nevertheless, well-formed, read sentences would seem much less complicated since a good deal of the possible interacting factors could be eliminated. In other words, the more the constraints are, the less likely the output appears to vary. Production data were also utilized in testing some different theories of intonation, i.e., the breath-group theory (Lieberman, 1967) and the declination theories (Maeda, 1967; Breckenridge and Liberman, 1977, 1979; Sorensen and Cooper, 1980). The study has shown that the declination theories failed to capture the overall intonation patterns of simple declarative sentences in Mandarin Chinese. The claim that the declination effect is a near-universal property is thus seriously questioned. Declination in intonation might be more of a discourse effect rather than the baseform for declarative sentences (Umeda and Coker, 1980).

The perception data can be summarized as follows. In the perception of Mandarin tones produced in citation forms, native speakers were able to identify these tones without the support of context, due to the fact that the fundamental frequency patterns of these tones were sufficient acoustically. That is, in order for the subjects to perceive such tones, the fundamental frequency contours of the tokens possess the specified tonal patterns. In the perception of syllable tones edited from spontaneous running speech, which often carried insufficient acoustic information specified for tones, subjects’ behaved rather poorly. That is, without the support of contextual information, native speakers were unable to identify syllable tones on the basis of insufficient acoustic information alone. This result indicates that in the perception of tones in spontaneous speech, native speakers may very well “reconstruct” or “recover” the lexical tones that are not physically present from the context. Speech perception is thus a complicated decoding process based not on the acoustic information only.