What's New, Pussycat? On Talking to Babies and Animals Denis Burnham, Christine Kitamura, and Uté Vollmer-Conna

Pitch and Phonetic Analyses

The pitch and phonetic analyses centered solely on the target words, sheep, shoe, and shark. All instances of these words, free from background noise spoken during mothers' 10- to 15-minute interactions with each recipient, were excised from the speech stream, and digitized at 10 kHz using Kay Elemetrics CSL (bold entries in Table S1). In a series of macros within CSL, all these words were low-pass filtered at 1KHz, and had impulse (glottal pulse) markers added. The words were then trimmed to just the voiced portion, and mean F0 derived.

The phonetic analyses centered on the values of the first and second formants in the /i/, /u/, and /a/ vowels in infant-, pet-, and adult-directed speech (IDS, PDS, ADS). Analysis procedures were developed after consultation with Jean Andruski. Macros were developed in CSL, as used by Andruski and Kuhl (S1, S2), which allowed comparison between narrowband spectrograms, FFT, and LPC analyses. Unbroken formants could not always be extracted and in such cases formant values could not be derived (one mother accounted for most of the omissions). F1 and F2 values from all suitable samples were extracted, logged and stored for later analysis. A respectable average of approximately 70% of eligible vowels were phonetically analyzed (italicized entries in Table S1). Mothers' mean F1 and F2 values for /i/, /u/, and /a/ were used to calculate vowel triangle areas for each speech register.

Affect Ratings and Analyses

30-second speech samples clear of background noise were randomly selected from each of the 12 mothers' infant-, pet-, and adult-directed speech. The resultant 36 samples were digitized at 10 KHz and low-pass filtered with a cut-off of 400 Hz. Ratings were used to derive affect scores for the three speech registers. 20 undergraduate students rated the 36 samples (with speech registers unidentified and order of presentation randomized) on 5 scales: (i) positive or negative vocal affect -4 (high negative affect) to +4 (high positive affect); and the intention of the speaker to (ii) express affection, (iii) encourage attention, (iv) soothe or comfort and (v) direct behavior from 1 (not at all) to 5 (extremely) (S3). Intra-class correlations were used to determine inter-rater reliability. Average measure intra-class coefficients for the five scales were .78, .84, .82, .74, .42 respectively. All ratings were factor analyzed, and the resulting affect factor weightings were .82, .88, .44, .81, and 0.14. Mean factor scores across raters for each mother were entered into an analysis of variance with speech registers (IDS, PDS, ADS) as a repeat factor, and mothers (N=12) as the random factor.

Supplemental Table 1. Number of words for the intonational (bold entries), and the phonetic analysis (italicized entries)
	/i/	/u/	/a/	Total
IDS	119 (73)	129 (72)	123 (66)	371 (211)
PDS	110 (86)	92 (70)	86 (61)	288 (217)
ADS	74 (66)	68 (71)	74 (59)	216 (196)
Total	303 (225)	289 (213)	283 (186)	875 (624)

References

S1. J. E. Andruski, P. K. Kuhl, P.K. In H. Bunnell & W. Isardi (Eds) Proc. 4th ICSLP, Philadelphia, 3, 1541-1544 (1996).

S2. J. E. Andruski, P. K. Kuhl, A. Hayashi, J. Acoust. Soc. Am. 105, 1095 (1999)

S3. Kitamura, C. Burnham D. Infant Beh & Dev. (in press)

Sound Files

Audio S1: Baby low-pass speech
Audio S2: Pet low-pass speech
Audio S3: Adult low-pass speech