Ba or pa?
You will hear 13 synthetic syllables that change in small, equal steps from ba to pa. In Part 1 you label each one. In Part 2 you try to tell pairs apart. At the end you will see how your hearing divides a smooth physical change.
Before you start
- Use headphones in a quiet place. The differences are short (a few thousandths of a second).
- On an iPhone or iPad, turn off silent mode, or you will hear nothing.
- Play the two end points and set your volume so both are comfortable.
Which syllable did you hear? Answer by your first impression. Each sound plays once.
You will hear three sounds. The third sound, X, is the same as either the first or the second. Which one does X match? If you cannot tell, guess.
Part 1 done
Your labeling curve is in the results below. Part 2 tests how well you can tell sounds apart.
Your results
Discrimination points sit at the midpoint of each pair (for example, the 15–30 ms pair is plotted at 22.5 ms). For discrimination, 50% is chance.
Show the numbers
What the demonstration shows
Voice onset time (VOT) is the interval between the release of the lips and the start of vocal-fold vibration. In English, syllable-initial /b/ usually has a VOT near 0 to 15 ms, and /p/ has a long lag, usually 40 to 80 ms, filled with breathy aspiration (Lisker & Abramson, 1964). The 13 sounds here differ only in VOT, in equal 5-ms steps.
Categorical perception shows up as two linked patterns (Liberman et al., 1957). First, labeling switches abruptly near one point on the continuum, for English speakers usually around 25 ms for labial stops, rather than changing gradually. Second, discrimination of pairs that differ by the same 15 ms is good when the two sounds fall on opposite sides of that boundary and close to chance (50%) when both fall in the same category (Abramson & Lisker, 1970).
The dashed line is the discrimination you would show if you could tell two sounds apart only by labeling them differently: P(correct) = ½ + ½(p1 − p2)², where p1 and p2 are the proportions of pa responses to the two sounds. Points above the dashed line mean you heard some differences within a category. Listeners usually do somewhat better than this prediction, which is why later work treats perception of stop consonants as strongly but not purely categorical.
The continuum
All 13 stimuli in order. Each spectrogram shows frequency from 0 to 5 kHz over 320 ms, with darker shading for more energy. The tick marks on the top and bottom edges show the release of the lips and the start of voicing. Between them is aspiration noise, with no energy in the lowest formant (F1).
The sounds are generated in your browser by a formant synthesizer modeled on Klatt (1980). Each has a 20-ms silent lead, a brief release burst, aspiration until voicing begins, and a voiced /a/ with formant transitions starting at release (F1 220→750 Hz, F2 950→1220 Hz, F3 2150→2600 Hz). Total duration from release is 300 ms for every token, and pitch falls from 122 to 96 Hz.
References
Abramson, A. S., & Lisker, L. (1970). Discriminability along the voicing continuum: Cross-language tests. In Proceedings of the Sixth International Congress of Phonetic Sciences (pp. 569–573). Academia.
Klatt, D. H. (1980). Software for a cascade/parallel formant synthesizer. Journal of the Acoustical Society of America, 67(3), 971–995.
Liberman, A. M., Harris, K. S., Hoffman, H. S., & Griffith, B. C. (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology, 54(5), 358–368.
Lisker, L., & Abramson, A. S. (1964). A cross-language study of voicing in initial stops: Acoustical measurements. Word, 20(3), 384–422.