Proceedings, FONETIK 2005, Department of Linguistics, Göteborg University Tonal word accents produced by Swedish 18- and 24month-olds Germund Kadin and Olle Engstrand Department of Linguistics, Stockholm University accent contrast and that, in consequence, Swedish word accent acquisition typically takes place during the 18-24 months age interval. In the present study, this assumption was tested using a group of 18-month-olds. To verify the language-specificity of observed accent patterns, comparable groups of American English 18- and 24-month-olds were used as controls. Abstract F0 measurements were made of disyllabic words produced by several Swedish and American English 18- and 24-month-olds. The Swedish 24- and 18-monthers produced accent contours that were similar in shape and timing to those found in adult speech. The Swedish 18monthers, however, produced very few words with the acute accent. It is concluded that most Swedish children have acquired a productive command of the word accent contrast by 24 years of age and that, at 18 months, most children display clear tonal ambient-language effects. The influence of the ambient language is evident in view of the F0 contours produced by the American English children whose timing of F0 events tended to be intermediate between the Swedish grave and acute contours. The relative consistency with which grave accent contours were produced by the Swedish 18-monthers suggest that some children are influenced by the ambient language well before that age. Methods Subjects were drawn from a larger group of approximately 60 Swedish and 60 American English children at the ages of 6, 12, 18, 24 and 30 months. Audio and video recordings were made as described in Engstrand et al. (2003). The present study is based on recordings of • 11 Swedish 24-month-olds (6 girls, 5 boys) • 13 American 24-month-olds (6 girls, 7 boys) • 11 Swedish 18-month-olds (6 girls, 5 boys) • 16 American 18-month-olds (9 girls, 7 boys), i.e., a total of 51 children, including the 24monthers used in Engstrand and Kadin (2004). All disyllabic words with stress on the first syllable were analyzed according to criteria described in Engstrand and Kadin (2004). F0 was measured at five points in time: at 1) the acoustic onset of V1, 2) the F0 turning-point in V1 (if the F0 contour was monotonic throughout the vowel, the turning-point was assigned the value of the onset), 3) the acoustic offset of V1, 4) the acoustic onset of V2, and 5) maximum F0 in V2 (if F0 declined throughout the vowel, maximum F0 was assigned the value of the onset). A Fall parameter was defined as the F0 difference between V1 turning-point and offset, and a Rise parameter was defined as the F0 difference between V2 maximum and V1 offset. All measurements were made using the Wavesurfer program package. Introduction Swedish has a contrast between a ‘grave’ and an ‘acute’ tonal word accent. The acute accent is associated with a simple, one-peaked F0 contour. The grave accent typically has a twopeaked F0 contour with a fall on the primary stress syllable and a subsequent rise towards a later syllable in the word (Bruce 1977, Engstrand 1995, 1997). A preliminary report on Swedish children’s acquisition of the word accents was presented in Engstrand and Kadin (2004). The results, which were based on 6 Swedish and 6 American English 24-month-olds, suggested that, at that age, Swedish children are well on the way to establishing a productive command of the accent contrast. The present study was carried out to test this preliminary conclusion using additional subjects. In addition, previous studies (Engstrand et al. 1991, Engstrand et al. 2003) have suggested that most 17-18-montholds display a much less consistent use of the Results Auditory judgments first suggested that a majority of the words produced by the Swedish children (both 18- and 24-month-olds) had a grave-like tonal contour and that, in general, 67 Proceedings, FONETIK 2005, Department of Linguistics, Göteborg University grave and acute accents were assigned to words in accordance with the adult norm. In contrast, none of the American English word productions sounded convincingly grave. this fall was, on the whole, less marked than for the older group (the grand means were 76 and 50 Hz for the 24- and 18-monthers, respectively). The Rise parameter, too, was positively evaluated for most 18-monthers but, again, less consistently than for the 24monthers (grand means 111 and 55 Hz, respectively). Two of the 18-monthers (SW18F3 and SW18M5) did not evidence any rise into the second vowel. Measurement results are summarized in table 1-5 (a full statistical treatment will be reported elsewhere). The tables present means and standard deviations for the Fall and Rise parameters. The bottom line of each table shows grand means and standard deviations. In the left column, SW and AM represent Swedish and American English, respectively, 18 and 24 indicate the respective ages in months, and F and M stand for sex (female or male). The last figure is a reference number that identifies the individual child. Thus, for example, SW24F1 stands for a Swedish 24 months old girl with the reference number 1. Table 2. Means for the Fall and Rise parameters in grave accent words produced by the Swedish 18monthers. Child SW18F1 SW18F2 SW18F3 SW18F4 SW18F5 SW18F6 SW18M1 SW18M2 SW18M3 SW18M4 SW18M5 Grand mean Table 1. Means for the Fall and Rise parameters in grave accent words produced by the Swedish 24monthers. Child SW24F1 SW24F2 SW24F3 SW24F4 SW24F5 SW24F6 SW24M1 SW24M2 SW24M3 SW24M4 SW24M5 Grand mean GRAVE ACCENT Fall (Hz) Rise (Hz) Mean SD Mean SD 69 62 102 121 60 29 69 40 62 45 70 66 113 88 93 129 72 41 99 124 77 38 127 125 102 141 392 286 59 42 70 84 101 122 106 136 70 59 42 78 60 35 50 44 76 64 111 112 N 21 22 19 26 23 12 2 23 23 21 23 GRAVE ACCENT Fall (Hz) Rise (Hz) Mean SD Mean SD 41 44 139 180 59 41 94 67 30 28 1,0 65 64 17 33 24 114 107 52 219 27 18 70 98 43 46 58 85 23 38 45 44 39 33 38 134 35 29 83 34 80 93 -3,0 27 50 45 55 89 N 2 14 4 4 7 2 17 4 6 13 2 75 The Swedish acute productions were relatively few – 37 in all produced by 6 of the 11 24month-olds. Thus, 5 of the Swedish 24monthers lacked acute productions altogether (table 3). Again, the Fall parameter had positive values which were, however, smaller than in the grave productions (and, as we shall see, with a different timing of the F0 peaks). The Rise parameter values differ markedly from those pertaining to the grave words in that they were consistently negative. This means that the acute words displayed a continuous F0 decline from an early peak in the first syllable. In other words, the acute words had a clearly ‘onepeaked’ F0 contour. 215 Swedish grave words produced by the 24month-olds consistently displayed positive values for both the Fall and the Rise parameters (table 1). This means that 1) F0 declined from a turning-point in the primary stress vowel reaching a relatively low value at the end of that vowel, and 2) rose to resume a relatively high position in the second vowel resulting in a ‘two-peaked’ F0 contour. Acute productions by the Swedish 18monthers were too few to provide a basis for reliable generalizations. However, parameter values tended to differ from those for the grave productions and to resemble those pertaining to the 24-monthers. The Swedish 18-month-olds show a similar, but somewhat less consistently two-peaked contour (table 2). The positive values of the Fall parameter indicate that F0 declined from a turning-point in the primary stress vowel, but 68 Proceedings, FONETIK 2005, Department of Linguistics, Göteborg University Table 5. Means for the Fall and Rise parameters in words produced by the American 18-monthers. Table 3. Means for the Fall and Rise parameters in acute accent words produced by the Swedish 24monthers. Child SW24F1 SW24F2 SW24F3 SW24F4 SW24F5 SW24F6 SW24M1 SW24M2 SW24M3 SW24M4 SW24M5 Grand mean ACUTE ACCENT Fall (Hz) Rise (Hz) Mean SD Mean SD 29 11 46 78 96 49 11 2,0 42 9,5 -3,7 25 -79 106 -29 22 -52 1,7 16 2,7 35 31 -73 57 43 15 -37 27 AMERICAN ENGLISH Fall (Hz) Rise (Hz) Mean SD Mean SD N AM18F1 41 68 -47 67 5 AM18F2 19 27 84 171 6 AM18F3 0 -11 1 AM18F4 39 49 -33 190 10 AM18F5 30 52 1 AM18F6 27 24 -48 102 12 AM18F7 28 22 -27 18 4 AM18M1 11 159 1 AM18M2 36 36 24 73 8 AM18M3 36 25 25 83 10 AM18M4 44 78 4,8 79 5 Grand mean 28 41 17 98 63 Child N 0 6 7 3 9 0 0 7 0 5 0 37 The above tables have shown differences between accent types and ambient languages in terms of the Fall and Rise parameters. Thus, timing has so far been disregarded. However, timing of F0 events in relation to segmental structure is crucial as illustrated in figure 1. The figure shows mean data for the Swedish and American English children in both age groups (symbols are explained in the figure legend). Grave and acute productions are shown for the Swedish 24-monthers. F0 values are timealigned to the first measurement point, and the data points are connected by smoothed lines that bear a certain resemblance to authentic F0 contours. The measurement points correspond to acoustic events as described above. F0 values pertaining to the American children tended to resemble those for the Swedish acute productions (see tables 4 and 5; five of the 16 American 18-monthers did not produce any usable utterances and have been left out). The Fall parameter values were moderately positive and the Rise parameters tended to be on the negative side. Thus, the American data suggest a moderate F0 decline in the first vowel which, on average, continued into the second vowel. Table 4. Means for the Fall and Rise parameters in words produced by the American 24-monthers. AMERICAN ENGLISH Child Fall (Hz) Rise (Hz) Mean SD Mean SD N AM24F1 69 82 -53 175 12 AM24F2 58 41 -4,5 45 16 AM24F3 33 23 -2,8 57 13 AM24F4 55 71 -80 81 15 AM24F5 32 31 -28 57 13 AM24F6 51 24 -2,9 76 7 AM24M1 20 27 -6,4 58 16 AM24M2 50 75 5,6 27 16 AM24M3 29 36 -40 82 16 AM24M4 93 149 -0,71 90 7 AM24M5 26 21 31 45 10 AM24M6 27 26 2,6 47 9 AM24M7 34 17 1 Grand mean 45 51 -10 70 151 The main timing effect is found in the first F0 maximum which, for the Swedish children, appears early in the stressed vowel of the grave words and near the following vowel-consonant boundary in the acute words. The productions pertaining to American children are intermediate with an F0 maximum approximately at or after the middle of the stressed syllabe. Also note the differences in the second vowel (mesurement points 4 and 5). The contour describes a steady slope in the Swedish acute productions, whereas the grave contour reaches as new peak in the second vowel. Thus, the acute contour differs from the grave contour both in having a late turning-point in the first vowel and in lacking a secondary peak in the following vowel. Even though the American English contours reflect occasional rises toward the 69 Proceedings, FONETIK 2005, Department of Linguistics, Göteborg University second vowel, these rises are smaller and less systematic than those of the Swedish grave contours. months, many Swedish children begin to produce grave-like F0 contours and to mark the appropriate words with these contours. Engstrand et al. (2003) reached a similar conclusion on the basis of listening tests. Based on those studies as well as preliminary analyses of the present material, Engstrand and Kadin (2004) hypothesized that acquisition of the Swedish tonal word accents typically takes place in the 18-24 months age interval. However, the relative consistency with which grave accent contours were produced by the present 18-monthers would suggest that some children are influenced by the ambient language well before this age. This is in agreement with results of listening tests suggesting occasional grave-like tone contours as early as at 12 months of age (Engstrand et al. 2003). Time (ms) Figure 1. Average F0 contours derived from mean parameter values shown in tables 1 - 5. Symbols: filled diamonds=SW24 grave, unfilled diamonds=SW18 grave, gray diamonds=SW24 acute, filled squares=AM24, unfilled squares=AM18. Acknowledgment Summary and conclusions This work was supported by grant 2003-846014311-29 from the Swedish Research Council (VR) to O. Engstrand. Auditory judgments have shown that disyllabic words produced by Swedish 24- and 18-montholds mainly carry the grave tonal word accent. This is an expected influence of the ambient language, since a majority of disyllabic Swedish words are characterized by that accent. Whereas the Swedish 24-monthers also produced a significant number of acute words, the acute accent occurred very rarely in the younger group. F0 contours were usually shaped and timed according to the adult norm. This was the case in all 24-month-olds and in all but two of the 18-month-olds. This suggests that most Swedish children at 24 months of age have established a productive command of the word accent contrast, and that many 18-montholds are in a fair way to acquiring the grave accent. However, the virtual absence of acute words in the 18-month-olds makes it hard to determine whether a systematic accent contrast has been established at that age. References Bruce G. (1977) Swedish Word Accents in Sentence Perspective. Lund: Gleerup. Engstrand O. (1995) Phonetic interpretation of the word accent contrast in Swedish. Phonetica 52, 171-179. Engstrand O. (1997) Phonetic interpretation of the word accent contrast in Swedish: Evidence from spontaneous speech. Phonetica 54, 61-75. Engstrand O., Williams K. and Strömqvist S. (1991) Acquisition of the tonal word accent contrast, Actes du XIIème Congres International des Science Phonétiques, Aix-enProvence, vol. 1, pp. 324-327. Engstrand O., Williams K. and Lacerda F. (2003) Does babbling sound native? Listener responses to vocalizations produced by Swedish and American 12- and 18-montholds. Phonetica 60, 17-44. Engstrand O. and Kadin G. (2004) F0 contours produced by Swedish and American 24month-olds: implications for the acquisition of tonal word accents. Proceedings of the Swedish Phonetics Conference held in Stockholm 26-28 May 2004, pp. 68-71. The influence of the ambient language is even more evident in view of the F0 contours produced by the American English controls. The timing of those contours tended to be intermediate between the Swedish grave and acute contours. Out of the five Swedish 17-18-month-olds observed in a similar study (Engstrand et al. 1991), three showed a two-peaked, grave-like F0 contour in grave words (even though a rise was consistently present on the second syllable). It was tentatively concluded that, at 17 70
© Copyright 2026 Paperzz