INTRODUCTION
Despite documentation of the near-universal linguistic and prosodic features contained in infant-directed speech (IDS) (Fernald & Simon, Reference Fernald and Simon1984; Snow, Reference Snow, Gallaway and Richards1994), there is little data on how adult speech directed toward children is modified as children develop. IDS has been characterized as communication that is listener-oriented, and communication that attracts infants' attention; it is also hypothesized to assist language development (Cooper, Abraham, Berman & Staska, Reference Cooper, Abraham, Berman and Staska1997; Fernald, Reference Fernald1991; Grieser & Kuhl, Reference Grieser and Kuhl1988). The language abilities of preschool children are still developing; age-related changes in maternal speech addressing children could therefore be examined if a longitudinal design was used to record mothers talking to the same child at different ages during childhood.
Studies have demonstrated a general tendency for adults to adjust the linguistic complexity of IDS, especially the semantic and pragmatic features, in accord with an infant's age and stage of language development (Brousseau, Malcuit, Pomerleau & Feider, Reference Brousseau, Malcuit, Pomerleau and Feider1996; Cross, Reference Cross, Snow and Ferguson1977; Murray, Johnson & Peters, Reference Murray, Johnson and Peters1990; Snow, Reference Snow, Gallaway and Richards1994). However, some studies show no age-related adjustments in the syntactic features of IDS (e.g. Kavanaugh & Jirkovsky, Reference Kavanaugh and Jirkovsky1982; Newport, Gleitman & Gleitman, Reference Newport, Gleitman, Gleitman, Snow and Ferguson1977). Whether maternal speech input is generally fine-tuned to a child's age or language level is therefore still debatable.
The prosodic features of IDS also exhibit age-related changes (e.g. Garnica, Reference Garnica, Snow and Ferguson1977; Kitamura & Burnham, Reference Kitamura and Burnham2003; Stern, Spieker, Barnett & Mackain, Reference Stern, Spieker, Barnett and Mackain1983), but the association may not be straightforward. In a longitudinal study (Stern et al., Reference Stern, Spieker, Barnett and Mackain1983), the pitch range in IDS was more exaggerated for infants at specific ages (0 ; 4) than for either younger (newborn) or older children (1 ; 0 & 2 ; 0). An explanation offered for this age-specific prosodic modification is the child's linguistic development: a lower level of IDS stimulation is sufficient to keep a newborn alert, whereas greater diversity in pitch is needed to encourage four-month-old infants to vocalize, and IDS is not the primary cue to alertness or vocalization in older infants (Stern et al., Reference Stern, Spieker, Barnett and Mackain1983). Another study (Kitamura & Burnham, Reference Kitamura and Burnham2003) reported that mothers adjust their pitch level and pitch range to infants at birth, 0 ; 3, 0 ; 6, 0 ; 9 and 1 ; 0 to indicate the mother's communicative intent. For example, mothers tend to use higher mean F0 to convey a positive affect in their phrases addressed to infants at 0 ; 6 and 1 ; 0, and use more directive utterances with greater F0 range and lower mean F0 in speech directed towards infants at 0 ; 9. These studies support the idea that maternal speech patterns complement an infant's changing language abilities during the first year of life.
Studies of age-related changes in CDS for children beyond two years of age have also produced mixed results. Garnica (Reference Garnica, Snow and Ferguson1977) found that mothers used higher pitch, wider pitch range, more final-rising in imperative sentences and more whispers when interacting with two-year-olds than with five-year-olds. In contrast, Warren-Leubecker & Bohannon (Reference Warren-Leubecker and Bohannon1984) reported no age-related changes; mothers modified their prosody equally when speaking to two- and to five-year-olds. In a small group of longitudinal subjects, Japanese parents tended to use higher pitch in IDS than ADS when addressing their 0 ; 0–1 ; 6 children. While parents' pitch height decreased with infant age from 0 ; 0 to 1 ; 6, parents did not elevate pitch significantly when addressing their 1 ; 7–5 ; 0 children (Amano, Nakatani & Kondo, Reference Amano, Nakatani and Kondo2006).
The speech recording conditions provide one possible explanation for this inconsistency among studies. Garnica (Reference Garnica, Snow and Ferguson1977) collected maternal speech samples in the laboratory as mothers helped their five-year olds perform a highly structured task, and this more information-oriented activity may have elicited speech that did not emphasize IDS characteristics. In contrast, Warren-Leubecker & Bohannon (Reference Warren-Leubecker and Bohannon1984) recorded speech samples during free play at the child's home, in which a higher proportion of social regulatory speech was elicited. Mothers generally use social regulatory speech more when talking to younger children than to older children, and also use a greater pitch range in regulating children's social behaviors than in informational speech (Fernald, Reference Fernald1991). It is possible therefore that both the pragmatic context of recordings and children's ages affected the pitch features of maternal speech.
In addition to prosodic modifications used when addressing infants, recent studies show that the acoustic-phonetic features of IDS are exaggerated when compared to ADS. The acoustic vowel space enclosed by the three point vowels (/i/, /a/ and /u/) is expanded in IDS vs. ADS across many languages (e.g. English, Swedish, Russian, Mandarin) (Burnham, Kitamura & Vollmer-Conna, Reference Burnham, Kitamura and Vollmer-Conna2002; Kuhl et al., Reference Kuhl, Andruski, Chistovich, Chistovich, Kozhevnikova, Ryskina, Stolyarova, Sundberg and Lacerda1997; Liu, Kuhl & Tsao, Reference Liu, Kuhl and Tsao2003). Liu et al. (Reference Liu, Kuhl and Tsao2003) show that the vowel expansion in IDS does not differ with infant's age during the first year of life; that is, the vowel space area of IDS is similar in Mandarin IDS in infants aged 0 ; 6–0 ; 8 and 0 ; 10–1 ; 0. No studies have examined vowel expansion for older children. Moreover, it has been shown that the degree of vowel space exaggeration in IDS is positively associated with infants' speech discrimination performance for native phonetic contrasts in both age groups, suggesting that the more clearly articulated IDS may facilitate infants' early phonetic learning (Liu et al., Reference Liu, Kuhl and Tsao2003). If phonetic exaggeration facilitates language learning at all levels (e.g. articulation, semantics and pragmatics), vowel space expansion observed in IDS should extend beyond the first year of life because preschool children's language skills are still developing. Alternatively, the extent of vowel acoustic expansion in maternal speech may be reduced as children age because the expansion may especially support phonetic learning and five-year-old children's phonetic development is quite advanced.
The exaggeration seen in IDS for vowels, and observed across languages, is not limited to vowels. Additional phonetic features in different language systems have also shown this exaggeration. For example, lexical tones, essential phonetic units signaling syllable lexical meaning in tonal languages, are also exaggerated in IDS. Maternal speech in Mandarin and Thai, both tonal languages, shows an elevated pitch and expansion of the pitch contours over syllables, phrases and sentences (e.g. Grieser & Kuhl, Reference Grieser and Kuhl1988; Kitamura, Thanavishuth, Burnham & Luksaneeyanawin, Reference Kitamura, Thanavishuth, Burnham and Luksaneeyanawin2002; Liu, Tsao & Kuhl, Reference Liu, Tsao and Kuhl2007). However, the prosodic modifications made by mothers to emphasize the distinctions that are critical phonetically did not distort the F0 contour shape that is crucial to lexical meaning at the syllable level. For example, the ‘turning point’, the point in time at which the F0 direction changes from falling to rising in the vocalic segment, a feature critical to lexical tone, is not altered in IDS (Liu et al., Reference Liu, Tsao and Kuhl2007). Thus the modification of IDS selectively emphasizes phonetic differences for young children.
Age-related changes in IDS pitch height and range that have been observed in tone languages show interesting differences that merit further study. Pitch modifications in suprasegmental units (i.e. words and phrases) in Thai IDS were shown to change with age during the second half of the first year of life (e.g. Kitamura & Burnham, Reference Kitamura and Burnham2003), but infant age during the first year did not affect pitch modifications for lexical tones at the syllable level in Mandarin IDS (Liu et al., Reference Liu, Tsao and Kuhl2007). No studies have explored both segmental (e.g. vowel) and suprasegmental (e.g. lexical tone) modifications beyond the age of 1 ; 0.
The present study had two goals. First, we examined whether the acoustic-phonetic features of vowels and lexical tones are modified in Mandarin CDS addressed to five-year-old children relative to ADS. There are inconsistent findings about whether CDS to five-year-olds exhibits any prosodic modifications (e.g. Amano et al., Reference Amano, Nakatani and Kondo2006; Garnica, Reference Garnica, Snow and Ferguson1977). Second, we examined whether the extent of acoustic-phonetic modification in maternal speech addressing children varies with age using a longitudinal design at two ages: preverbal infants vs. five-year-olds.
Regarding the first goal, we hypothesized that CDS would differ from ADS, reflecting the fact that at age 5 ; 0, children are still developing their language abilities. Regarding the second goal, we hypothesized that there are age-related differences in the acoustic-phonetic features of maternal speech addressed to children at the preverbal versus five-year-old stage. Specifically, the extent of speech exaggeration, measured both in vowels (i.e. vowel space and duration) and in lexical tone at the syllable level (i.e. pitch height and range), were hypothesized to be reduced as a function of age as children develop from 0 ; 7–1 ; 0 to 5 ; 0. A strength of the design used to test the hypothesis in the present experiment is the use of both acoustic parameters of vowel expansion and lexical tone expansion in Mandarin CDS as dependent variables and a longitudinal design which reduces speaker variation in the acoustic measures collected at different ages. In our previous cross-sectional study using a large sample (n=32) of mothers of infants aged 0 ; 6–0 ; 8 and 0 ; 10–1 ; 0, and the same recording materials and measures, the IDS acoustic measures did not significantly change with infant age (Liu et al., Reference Liu, Tsao and Kuhl2007; Liu et al., Reference Liu, Kuhl and Tsao2003). Therefore, IDS from mothers of infants aged 0 ; 7–1 ; 0 were combined to examine the age-related differences between preverbal IDS and five-year-old CDS in this longitudinal study.
METHODS
Participants
Seventeen mother–child dyads participated in the speech recordings at two times, when the children were aged 0 ; 7–1 ; 0 (mean=0 ; 9.21; range=0 ; 6.24 to 1 ; 0.0; boys=10, girls=7) and 5 ; 0 (mean=5 ; 3.0; range=5 ; 0.0 to 5 ; 9.0). All participants were recruited from the list of the House Registry Offices of Kaoshuing, a metropolitan area in Taiwan. A language background questionnaire was administrated before the speech recording to verify that mother–child dyads used Mandarin Chinese as the dominant language in their homes. Infants had no known physical, sensory or mental disabilities.
Infants were recruited from middle-class families. Mothers typically were high school and four-year-college graduates (M=13·2 years, SD=1·91) and fathers had similar educational backgrounds (M=13·51 years, SD=2·01). The average annual household income for the participants placed them close to Taiwan's average household income.
Speech stimuli
Speech samples were collected from all mother–child dyads and mother–experimenter dyads, using twelve Mandarin Chinese preselected bisyllabic target words. These target words were constructed as (C)VCV(V) and contained the three corner vowels, /i/, /a/, /u/, with four lexical tone patterns equally distributed in the first syllable, such as /i2ma1/ ‘aunt’, /ma3t
h
1 / ‘wagon’, /u1kuε1/ ‘tortoise’ (the tone numeral was attached to the end of each syllable). The tonal contexts were matched for target syllables – the tone pattern of the second syllable in the bisyllabic words was always the high flat tone, Tone 1, to minimize the effects of neighboring syllables on the tone of the target syllable (i.e. the first syllable). That is, the tones of the bisyllabic words that contain the vowel /i/ in the first syllable were [11], [21], [31] and [41], and the same situation was applied to target words with /a/ and /u/ in the first syllable. All selected target words were nouns in Mandarin Chinese, including people's names, objects and animal names.
Recording procedure
Each mother was audio-recorded when talking to a native Mandarin-speaking adult and to her child in a quiet room in the speech laboratory. A high-quality digital audio tape-recorder (SONY TCD-D100), with 16-bit resolution, a sampling rate of 44·1 KHz and a microphone (SONY ECM-MS907) were used for recording. In this semi-structured situation, preselected toys and pictures corresponding to individual target words were provided during the speech recording to help mothers easily use the target words in mother–child interaction. Multiple speech samples of each target word were collected during recording for each speech style. The equipment, materials and procedure were held constant in the two speech recording settings at the two ages (infancy and childhood). During the IDS recording, mothers were told that the goal of the study was mother–child play and were encouraged to talk as freely and naturally as they did at home.
During ADS recordings, mothers talked with the experimenter about their child's interest in the same set of preselected toys and pictures while their infant (child) was playing with an adult in the adjacent room. The ADS samples were recorded at both child ages to examine the reliability of acoustic analysis at the two time-points. Acoustic measures of ADS were not predicted to show any age-related differences. The recording sequence of IDS (or CDS) and ADS samples was counterbalanced across all subjects.
Acoustic analysis
The KAY Elemetrics Computerized Speech Laboratory (CSL) software was used to conduct the acoustic analysis. The vowel and tone in the first syllable was the target unit for acoustic analysis. The acoustic-phonetic measures included: vowel duration, vowel formant frequencies (first two formants), and mean F0 and F0 range for pitch contours in Mandarin lexical tones at the syllable level. To measure the vowel features, narrow-band spectrograms, FFT spectra and autocorrelation LPC spectra were used to judge the locations of formants. The onset of each vocalic segment was marked when both formants (F1 and F2) were visible on the spectrogram. Vowel offset was marked at the point where F2 and/or F1 were no longer visible. Each vowel's duration was measured from vocalic onset to offset. Vowel formant frequencies were measured at the cursor that marked the onset, central (i.e. vowel steady state) and offset positions of the vocalic segment. Averaged (across the three measuring points) F1 and F2 values, representing the dynamic pattern of vowel-formant change, were used to calculate acoustic vowel space areas for individual mothers in each speech condition. The F1 and F2 of vowels were viewed as Descartes' coordinates on the x–y plane; the area of vowel space compassing /i/, /a/ and /u/ was equal to the triangular area constructed from the three (F1, F2) pairs of each point vowel in the x–y plane. The vowel space area was calculated using the following equation:
where ‘ABS’ is absolute value, ‘F1i’ symbolizes the F1 value of vowel /i/, ‘F2a’ symbolizes the F2 value of vowel /a/, and so on.
For the lexical tone analysis, the F0 value in the initial, middle and final positions of the vocalic segment in each syllable, along with the highest and lowest (valley) points of the F0 contour of the vowel, were measured to track the overall pitch contour of individual lexical tones. The ‘turning point’, an essential temporal cue for tone distinctions (Liu et al., Reference Liu, Tsao and Kuhl2007), was calculated as the relative timing of the point where the pitch contour changes direction from falling to rising, [(the duration of onset to valley÷duration of vowel)∗100%].
A second well-trained phonetician analyzed 5 percent of the speech samples, randomly selected to validate the inter-rater reliability of acoustic analysis. The inter-rater reliability for the acoustic analysis procedure was high (r=0·92).
RESULTS
Did mothers modify the acoustic-phonetic features of maternal speech when addressing their five-year-old children, compared with ADS? Did speech modifications in language addressed to children vary with the child's age? To address these questions, the acoustic-phonetic features of maternal speech of seventeen mother–child dyads were analyzed twice, first when infants were 0 ; 7–1 ; 0 and again when they reached the age of 5 ; 0. Table 1 lists the vowel space, vowel duration, pitch height and pitch range of ADS, IDS and CDS.
Means of acoustic features of Mandarin ADS and IDS measured at two ages (standard deviation in parentheses)
Vowel characteristics
Vowel space area
The vowel space area containing /i/, /a/ and /u/ in the two speech conditions were compared at both ages. Compared to ADS, maternal speech addressing both preverbal infants and preschool children exhibited an exaggeration of vowel space. A one-way repeated-measure ANOVA (IDS, CDS, ADS-infancy, ADS-childhood) on vowel space areas showed a significant speaking style effect (F(3, 48)=14·83, p<0·001, ηp2=0·481). The Bonferroni post-hoc test (p=0·01) shows the order of vowel space: IDS≈CDS>ADS-infancy≈ADS-childhood. Thus, the results show that, in infancy, the vowel space of IDS is significantly larger than that of ADS. The expanded vowel space is also seen at age 5 ; 0; the vowel space of CDS is significant larger than that of ADS. The vowel space of maternal speech directed to preverbal infants is similar to the vowel space of CDS, suggesting that vowel area in maternal speech addressing children is not greatly reduced with children's ages.
Vowel duration
Mothers elongated vowels when talking with their preverbal infants and their five-year-old children when compared to ADS. A one-way repeated-measures ANOVA (IDS, CDS, ADS-infancy, ADS-childhood) on vowel duration showed a significant style effect (F(3, 48)=34·26, p<0·001, ηp2=0·682). Bonferroni post-hoc analyses (p=0·01) showed the order of vowel duration: IDS >CDS >ADS-childhood≈ADS-infancy. Mothers spoke more slowly with their preverbal infants than with their five-year-old children. The vowel duration of IDS was significantly longer than that of ADS-infancy. The same pattern was shown with the five-year-old children; the vowel duration of CDS was significantly longer than that of ADS-childhood. The results show a significantly reduced durational modification in maternal speech addressing five-year-old children when compared to infants aged 0 ; 7–1 ; 0.
Lexical tone characteristics
Mean F0
Compared to ADS, mothers raised their pitch (i.e. mean F0) when talking with their preverbal children and five-year-old children. A one-way repeated-measure ANOVA (IDS, CDS, ADS-infancy, ADS-childhood) on pitch showed a significant style effect (F(3, 48)=59·65, p<0·001, ηp2=0·788). The Bonferroni post-hoc analyses (p<0·001) showed the pitch height order: IDS>CDS>ADS-infancy≈ADS-childhood. Therefore, there is an age-related F0 change in maternal speech at different ages: mothers used a higher pitch in IDS when compared to CDS. The mean F0 of ADS did not significantly vary with age. Figure 1 illustrates the mean F0 of individual lexical tones.
The mean F0 values of four lexical tones in Mandarin ADS, IDS and CDS to children at two ages (SE in error bars).
Considering the individual tones, the mean F0 was significantly greater in IDS than CDS (one-way ANOVAs all reached a significance level of p=0·05 for each individual tone). The results also demonstrated a consistent order of mean F0 among the four lexical tones across the two ages, i.e. Tone 4>Tone 1>Tone 2>Tone 3. The mean F0 difference between high (i.e. Tones 1 & 4) vs. low tone (i.e. Tone 3) pairs was significantly more exaggerated in IDS when compared to CDS for Tone 1 vs. Tone 3 (t(16)=3·29, p=0·004) and Tone 4 vs. Tone 3 (t(16)=3·69, p=0·002).
F0 range
Compared to ADS, Mandarin-speaking mothers used greater pitch fluctuation (F0 range) at the syllable level in IDS and CDS. A one-way repeated-measure ANOVA (IDS, CDS, ADS-infancy, ADS-childhood) on F0 range showed a significant style effect (F(3, 48)=21·03, p<0·001, ηp2=0·568). The order of F0 range in the Bonferroni post-hoc analysis (p<0·05) is: IDS>CDS>ADS-infancy≈ADS-childhood. Mothers used a more exaggerated F0 range in IDS than in CDS and a larger F0 range in CDS than ADS-childhood, while the F0 range of ADS did not vary with age. The results show age-related changes in F0 range and a reduced F0 range modification in speech addressing older children. For the individual tones, Figure 2 illustrates the F0 range of different speech styles. For the individual tones, the F0 range was significantly greater in IDS than in ADS (measured in infancy), a greater F0 range was observed in CDS than in ADS (measured in childhood) (all one-way ANOVAs reached significance at p<0·001). Moreover, the F0 range of individual tones was significantly greater in IDS than in CDS (all reached the significance level of p=0·05), with the exception of the flat level tone, Tone1 (F(1, 16)=3·63, p=0·075), showing an age-related change of F0 range in maternal speech directed to infants and children. A consistent order of F0 range among the four lexical tones occurred in both IDS and CDS: Tone 4>Tone 3>Tone 2>Tone 1.
The F0 range of four lexical tones in Mandarin ADS, IDS and CDS to children at two ages (SE in error bars).
F0 turning point
The results show that duration is increased significantly in IDS and CDS as opposed to ADS, and this raises an interesting question regarding the ‘turning point’, the point at which the F0 contour direction changes from falling to rising in Tones 2 and 3, in IDS and CDS. In ADS, the turning point distinguishes two tones with similar pitch contour, Tone 2 (rising) and 3 (dipping), and we predicted that it would also do so in IDS and CDS. That is, IDS and CDS would not exaggerate the turning point difference between speech styles because exaggerating this cue could perceptually confuse infants regarding lexical meaning.
The results confirmed our prediction, showing that the turning point was not significantly different for the two speaking styles (one-sample t-test, p>0·1 for each tone pair) in both IDS (Tone 1=49·13%, Tone 2=25·62%, Tone 3=75·33%, Tone 4=93·33%) and CDS (Tone 1=58·97%, Tone 2=29·07%, Tone 3=79·34%, Tone 4=93·30%). This suggests that the essential temporal cues needed to identify lexical tones are preserved in both CDS and IDS, a pattern that replicates an earlier finding showing that mothers use similar turning points in lexical tones when addressing young infants and adults (Liu et al., Reference Liu, Tsao and Kuhl2007).
Examining repeat-recording effects on CDS
In addition to the longitudinal data examining the differences between CDS and ADS, and the age-related changes of acoustic features between IDS and CDS, the present study also analyzed speech samples collected from another group of eleven Mandarin-speaking mothers addressing their five-year-old children to examine the potential effect of repeat recordings on acoustic-phonetic features of CDS. As predicted, the results show no significant difference between the two groups of mothers addressing their five-year-old children (one-way ANOVA, all p>0·1). Thus, the age-related changes of acoustic-phonetic modifications observed between IDS and CDS cannot be attributed to repeat recordings.
DISCUSSION
The results of this study demonstrate that the pattern of acoustic-phonetic exaggeration shown in speech addressed to preverbal infants (Kuhl et al., Reference Kuhl, Andruski, Chistovich, Chistovich, Kozhevnikova, Ryskina, Stolyarova, Sundberg and Lacerda1997; Liu et al., Reference Liu, Kuhl and Tsao2003; Liu et al., Reference Liu, Tsao and Kuhl2007) also exists in speech addressed to five-year-old children. The longitudinal design used in the present study additionally compared the acoustic-phonetic features of speech addressed to the same children as preverbal infants and as five-year-olds. The results show that the extent of the acoustic-phonetic modification in the speech that adults address to children changes as a function of the age of the child – acoustic-phonetic modifications are significantly reduced in CDS when compared to IDS – indicating that a child's age plays a role in maternal speech in infancy and childhood.
The findings that mothers exaggerate vowel space, elongate vowel duration, raise F0 height, expand the F0 range of tone contours and increase the F0 difference between tone pairs when talking with their five-year-old children suggest that speech characteristics may reflect mothers' assessments, though not conscious, of their children's ages and their linguistic abilities. The results are consistent with previous studies showing CDS pitch modifications for five-year-old children (e.g. Garnica, Reference Garnica, Snow and Ferguson1977). Compared to ADS, the exaggerated acoustic features in maternal speech directed towards children at two ages appear to amplify, to different degrees, the essential phonetic features for both preverbal infants and preschool children.
A particular strength of this study is the use of a longitudinal design that reduces the possibility that speaker and parenting experience differences account for the age-related changes. Comparing the acoustic features of speech addressed to preverbal infants and five-year-old children highlights the age-related vowel and lexical tone modifications between IDS and CDS. Further longitudinal studies can now examine the association between CDS modifications and direct assessments of children's language abilities during childhood to test how finely tuned CDS modification is to a specific child's language abilities.
What is the function of these acoustic exaggerations in maternal speech to infants and preschoolers? Vowels with longer duration are indicative of ‘clear’ versus ‘conversational’ speech among individual speakers (Picheny, Durlach & Braida, Reference Picheny, Durlach and Braida1989). In addition to the temporal cues, a larger vowel space is associated with better speech intelligibility in typical speakers (Bradlow, Torretta & Pisoni, Reference Bradlow, Torretta and Pisoni1996) and atypical speakers with motor speech disorders (Turner, Tjaden & Weismer, Reference Turner, Tjaden and Weismer1995). Therefore, expanding the vowel space and lengthening vowels in maternal speech implies that Mandarin-speaking mothers are speaking more clearly when addressing their infants and children, making vowels perceptually more distinct from one another for children (Kuhl et al., Reference Kuhl, Andruski, Chistovich, Chistovich, Kozhevnikova, Ryskina, Stolyarova, Sundberg and Lacerda1997; Liu et al., Reference Liu, Kuhl and Tsao2003). This may be beneficial for infants because it makes phonetic units more distinct (Kuhl et al., Reference Kuhl, Andruski, Chistovich, Chistovich, Kozhevnikova, Ryskina, Stolyarova, Sundberg and Lacerda1997), which may account for the association between mothers' speech and infants' enhanced performance in speech discrimination tasks (Liu et al., Reference Liu, Kuhl and Tsao2003). For lexical tones, the major acoustic cues for Mandarin-speaking adults' perception of lexical tones are F0 height and F0 range (e.g. Gandour & Harshman, Reference Gandour and Harshman1978). Exaggerating these features in IDS and CDS could make the tones perceptually more distinct for children. In addition, similar patterns for the F0 turning point across speech styles preserves lexical meaning in maternal speech addressing infants and children (Liu et al., Reference Liu, Tsao and Kuhl2007).
Studies on infant speech perception suggest a learning process in which infant perception is shaped by exposure to ambient language (Kuhl, Conboy, Coffey-Corina, Padden, Rivera-Gaxiola & Nelson, Reference Kuhl, Conboy, Coffey-Corina, Padden, Rivera-Gaxiola and Nelson2008). Early skill in native-language speech discrimination can predict the growth of later language development between 1 ; 1 and 2 ; 6 (Tsao, Liu & Kuhl, Reference Tsao, Liu and Kuhl2004; Kuhl, Conboy, Padden, Nelson & Pruitt, Reference Kuhl, Conboy, Padden, Nelson and Pruitt2005), and mothers' clear speech in IDS and CDS may assist this process.
What are the explanations for age-related changes? Phonetic exaggeration was reduced in CDS compared to IDS and this could be interpreted as an adjustment to the child's age or linguistic level (Kitamura & Burnham, Reference Kitamura and Burnham2003; Stern et al., Reference Stern, Spieker, Barnett and Mackain1983). Increased intelligibility in IDS is achieved by exaggerating the acoustic cues to vowels and lexical tones and this could directly enhance the perceptual cues needed for infants to discriminate native-language phonetic differences (Liu et al., Reference Liu, Kuhl and Tsao2003). Given that children's language skills are not completely adult-like at the age of five, the enhancement of speech at the phonetic level may continue to play a role in language learning. However, since children are able to produce lexical tones and vowels in Mandarin Chinese by the age of five (e.g. Hua & Dodd, Reference Hua and Dodd2000), maternal highlighting of the phonetic units (i.e. vowel and lexical tone) during childhood may not be as important. CDS modification could also be associated with the fact that adults emphasize semantic and pragmatic features when addressing older children (Brousseau et al., Reference Brousseau, Malcuit, Pomerleau and Feider1996; Cross, Reference Cross, Snow and Ferguson1977; Snow, Reference Snow, Gallaway and Richards1994). At the age of 5 ; 0, CDS, with its higher pitch and exaggerated lexical tones, longer and more exaggerated vowels, could facilitate children's learning of new words and their development of pragmatic skills.
Further study of the association between acoustic exaggeration at the phonetic level in maternal speech to children and those children's language processing skills in childhood will be necessary to assess the value of adults' modifications of speech, and their effects on children. This study analyzed the acoustic modifications of maternal speech at the phonetic level; other linguistic modifications, such as the semantic, syntactic and pragmatic of IDS (e.g. Snow, Reference Snow, Gallaway and Richards1994) would also be of interest. To fully examine the hypothesis that infant- and child-directed speech facilitates language acquisition, longitudinal studies examining long-term relationships between maternal speech at an early age and children's later language performance would provide very valuable information.
