Quantity in Infant-directed Speech

This paper reports on an investigation of how the phonological rule of the complementary relationship between vowel and consonant duration is manifested in Swedish infant-directed speech (IDS), as compared to adult-directed speech (ADS). Cross-linguistic investigations indicate that vowels tend to be prolonged in speech directed to infants. In contrast, there are signs that consonants may be shortened in this speech style. How then, would these tendencies affect the phonological rule of V:C and VC: sequences in IDS? Analysis of vowel and consonant durations in disyllabic content words from six Swedish mothers’ IDS and ADS revealed that the VC proportionality is preserved in the IDS. The results suggest that important information of the linguistic structure of the ambient language is maintained. The results are discussed from an infant developmental perspective in a model of IDS transforms. INTRODUCTION In the phonological system of Swedish quantity distinction plays an important role. Minimal word pairs like ‘kapa’ and ‘kappa’ or ‘lät’ and ‘lätt’ are contrasted by different vowel and consonant lengths (quantity). The relevant domain for this quantity distinction is limited to lexically stressed VC syllables, as in other Germanic languages [1, 2]. The proportionality of the segment durations in such VC sequences is complementary, i.e. a long vowel is followed by a short consonant (V:C) and a short vowel is followed by a long consonant (VC:). Often, the quantity contrast is enhanced by correlated phonetic markers like vowel quality and diphthongization [3]. However, the observations of this quantity contrast are based on speech directed to adults. In the interaction with infants a quite different speech style is typically adopted. Cross-linguistic investigations show systematic differences between infant-directed speech (IDS) and adult-directed speech (ADS) in e.g. extensive prosodic modulations in IDS [4, 5]. More importantly, with respect to the quantity contrast, acoustic studies at the segmental level show overspecification in terms of increased vowel durations in IDS [6, 7, 8]. In contrast, a recent investigation reveals decreased consonant duration as measured in Voice Onset Time (VOT) [9]. Since the quantity contrast in Swedish involves duration characteristics in VC syllables the question asked is whether the IDS transformations are characterized by prolonged vowels and shortened consonants, thereby causing a degradation of the quantity distinction. Or do the mothers adopt a selective strategy by keeping the complementary VC proportions and in this way preserve the quantity contrast? To elucidate this question speech samples from six Swedish mothers’ IDS and ADS are analyzed. 1.1. Method The IDS samples were recorded when the mothers, native speakers of central standard Swedish, interacted with their three-month-olds in a sound-treated studio. The mothers were requested to interact with their infants as they would normally do at home, while using a selection of toys. The ADS samples were collected immediately after the IDS session when the investigator entered the studio and talked informally with the mother attempting to direct the conversation to the topics of the preceding mother-infant session. Both sessions were videotaped. 1.1.1. Speech material To obtain comparable word lengths while avoiding potential prosodic junctures due to morphological structure or compression effects in consonant clusters, content words meeting the following criteria were selected for analysis: • Disyllabic words carrying primary stress on the first syllable, i.e. trochaic stress pattern. • Only simple forms • Words with consonant clusters of maximally two adjacent consonants in word medial positions The criteria for syllable boundary division between lexically stressed and unstressed syllables followed those proposed in Bruce [1]. 1.1.2. Analysis The speech samples were digitized at 16 kHz and acoustically analyzed using ESPS/WAVES, 5.0. With this program a waveform display, a spectrogram and a labeling window are simultaneously available on the screen. The vowel and consonant lengths were measured in the selected syllables and the ratio between the durations (V/C) was calculated. Word length was also measured. RESULTS The distribution of the number of VC sequences as a function of speech condition and target vowel length is displayed in table 1.