Page 16 - Fall 2006

later become very difficult to acquire. Indeed, it has been argued that when infants acquire intonation languages, they learn to disregard absolute pitches, since individual words here carry the same meaning regardless of the pitch level in
31
that absolute pitch is initially acquired by tone language speakers as a feature of speech, and that its rarity in speakers of intonation languages such as English can be explained by the lack of opportunity to acquire it in infancy. The hypothe- sis that absolute pitch may in principle be acquired very early in life is strengthened by findings that infants can carry out perceptual learning tasks that involve referring to the
on the absolute pitch hypothesis we would not expect to see a reduction in consistency with the passage of time. We gen- erated a list of 12 Mandarin words, with each of the four tones occurring three times in the list.
The subjects were tested in two sessions, which were held on different days, but now they read out the word list twice in each session, with the readings separated by roughly 20 sec- onds. We calculated four difference scores: Between the first readings on Days 1 and 2; between the second readings on Days 1 and 2; between the first and second readings on Day 1, and between the first and second readings on Day 2.
Table 1 shows, for each comparison, the numbers of sub- jects whose averaged signed difference scores fell into each 0.25 semitone bin. It can be seen that remarkable consisten- cies were again obtained. For all comparisons, half of the sub- jects showed averaged pitch differences of less than half a semitone, and one-third of the subjects showed averaged pitch differences of less than a quarter of a semitone. And importantly, there were no significant differences in the degree of pitch consistency in reciting the word list on differ- ent days compared with reciting it twice in immediate suc- cession. In fact, the differences between comparisons across days and under immediate repetition did not begin to approach significance. This leads to the conclusion that although the pitch discrepancies we obtained were very small, they still underestimated the precision of the absolute pitch templates that influenced the subjects’ speech.
How do speakers of intonation language perform on this task? To examine this, we tested a group of 14 native speak- ers of English on a list of 12 English words, using the same procedure as for the speakers of Mandarin. We found that the English speakers showed roughly the same degree of consis- tency as did the Mandarin speakers in reciting the word list twice in immediate succession; however, they were signifi- cantly less consistent than the Mandarin speakers in reciting the list on different days. So the speakers of English differed from those of Mandarin in terms of pitch consistency, both
which they are pronounced.
My colleagues and I have been examining the conjecture
absolute pitches of tones.
32
Absolute pitch and tone language
If tone languages speakers use absolute pitch as a cue to differentiate the meaning of words, then we would expect them to refer to stable and precise absolute pitch templates in reciting the same list of words on different days. In one experiment to examine this conjecture,14 we tested seven speakers of Vietnamese, none of whom had received any sig- nificant musical training. The subjects were handed the same list of ten Vietnamese words to read out in two sessions, which were held on different days. Then for each spoken word, we took pitch (f0) estimates at 5 ms intervals, and aver- aged these pitches along the musical scale (i.e., along a log frequency continuum), so producing an average pitch for each word. Then for each speaker we calculated the differ- ence between the average pitches produced by each word as it was spoken on the different days, and we averaged the signed differences across the words in the list.
Our results showed remarkable consistencies: All speak- ers showed averaged pitch differences of less than 1.1 semi- tone, with two of the seven speakers showing averaged pitch differences of less than a quarter of a semitone. So the speak- ers must have been referring to precise and stable absolute pitch templates in reciting
the list of words.
In our next experi- ment, we tested 15 speakers of Mandarin, who (with one exception) had also received little or no musical training. All these subjects had grown up in China, and spoke Mandarin as their primary language. One purpose here was to test the generality of our findings to a different tone language. A second pur- pose was to explore whether there would be less consistency in reciting the word list on different days, compared to reciting it twice in succession, since
Table 1. Pitch difference scores produced by speakers of Mandarin on reading out the same list of words on different occasions. The table displays, for each comparison, the number of subjects whose pitch difference scores fell into each 0.25 semitone bin.
14 Acoustics Today, October 2006