4.3 Results
7.3.1 Speech Errors
There were 99 cases where the target syllables themselves were impaired through errors.
Slightly more low-frequency than high-frequency syllables were affected (56 vs. 43). An additional 25 errors concerned the syllables immediately preceding or following the target syllables. Of these, 13 appeared in the context of frequent and 12 in the context of rare syllables. Variations in vowel duration and tenseness were not counted as errors, as it was often difficult to decide whether the speaker intended to produce a long or a short vowel.
Especially quadruples 4, 7, and 8 were ambiguous in this respect, as there are hardly any differences in vowel quality between short and long /a/ in German. A certain amount of variability was also allowed in the vowel of the second syllable of the target words. Some participants tended to pronounce the second syllable vowel [ɐ] as [eːɐ] due to the local regiolect, and the [ə] in the syllable [təl] was often merged with the following [l] or elided completely.
7.3.2 Duration
Mean values for syllable and vowel duration of frequent and rare syllables in each quadruple are given in Table 15. For all quadruples except 5 and 6 there was a slight tendency for frequent syllables to be shorter than rare syllables. A Wilcoxon rank sum test showed a significant effect of syllable frequency across quadruples (W=1 760 773, p<.01). In an investigation of combinations of quadruples and frequency level, results only proved to be significant for quadruple 2 (W=22 399.5, p<.05, corrected for 120 comparisons). Vowel duration differences between frequent and rare syllables were less consistent. Hardly any differences in mean values appeared for quadruples 4 and 5, and for quadruple 3 there was actually a tendency in the opposite direction, with rare syllables tending to have shorter vowels than frequent syllables. Even so, the effect remained significant across quadruples (W=1 799 040, p<.05). No duration effects were visible for syllable onsets and codas or for either of the two context syllables immediately preceding and following the target items.
Investigations of combinations of quadruple identity and frequency level showed a significant frequency effect on vowel duration of quadruple 6 (W=18 601, p<.0001, corrected for 120 comparisons).
Quadruple 1 2 3 4 5 6 7 8 All Frequent
syllables 245.8
66.0 215.7
61.2 254.5
76.0 270.8
123.3 294.0
138.8 233.5
60.3 264.3
72.0 299.3
138.8 259.5 91.8 Rare
syllables 249.4
69.5 227.0
64.6 260.1
71.5 275.9
124.6 294.4
138.5 232.5
68.8 271.8
75.7 306.2
142.3 265.0 94.8
Table 15: Mean duration in milliseconds for syllables (above) and vowels (below) by quadruple and frequency category
Although differences between frequent and rare syllables were very small, they remained visible even in a direct comparison of items. For each error-free group of four syllables produced by the same speaker in the same context, duration values for the two frequent syllables were subtracted from duration values for the two rare syllables.
Figure 18: Duration difference between frequent and rare syllables in milliseconds, grouped by participant, quadruple, and context
In Figure 18, results from a comparison of syllable duration values are presented as a boxplot diagram. Wilcoxon signed rank tests showed significant deviations from zero on the level of the syllable (V=210 810, p<.0001) as well as for vowels (V=205 467.5, p<.0001) and codas (V=170 626, p<.05), but not for onsets. In such direct comparisons, the lengthening effect of rare syllables seemed to extend to the duration of the following context syllable (V=195 667.5, p<.001), while the preceding context remained unaffected.
Both frequent and rare syllables tended to be longer in the first appearance of the target words than in subsequent renditions. An interaction plot between frequency level and order of occurrence (Figure 19) shows that although duration differences decreased as participants became more familiar with the target items, mean values for rare syllables remained above those for frequent syllables. In Wilcoxon rank sum tests for combinations of frequency level and order of appearance, frequency effects for the same order position remained below significance level, as did order effects for the same frequency level (corrected for 28 comparisons). Only the duration difference between the first and last appearance of rare syllables proved to be marginally significant (W=125 886.5, p=.067, mean values: 273.1 ms vs. 260.6 ms).
Figure 19: Mean syllable duration by order of appearance for frequent and rare syllables in milliseconds
To investigate duration variability, absolute duration differences were calculated between segmentally identical syllables spoken by the same participant in different contexts. Mean values for syllable and vowel variability are given in Appendix G. In this examination, vowel duration appeared to be more variable across contexts for rare than for frequent syllables.
Wilcoxon rank sum tests confirmed this effect (W=3 576 768, p<.0001). While no significant tendencies were visible for syllable or onset duration, codas of rare syllables were marginally more variable than codas of frequent syllables (W=3 716 340, p<.097, mean values: 186.8 ms vs. 172.1 ms). An analysis of the context syllables themselves revealed that they tended to
show higher absolute duration differences from one context to the next when surrounding a rare syllable than when appearing together with a frequent syllable (W>3 950 000, p<.05, mean values preceding syllable: 53.5 ms vs. 50.9 ms, mean values following syllable: 48.3 ms vs. 44.7 ms). To confirm the findings, distance values for frequent syllables were subtracted from those for rare syllables for each context comparison. Figure 20 presents differences in vowel variability for individual quadruples. Analyses of differences in variability using Wilcoxon signed rank tests were significant in terms of vowel duration (V=329 791.5, p<.0001), coda duration (V=305 409.5, p<.01), and the duration of the following context syllable (V=325 542, p<.0001). In this analysis, the effect on the preceding context syllable was marginally significant (V=305 453.5, p=.06). When examining vowel variability for combinations of quadruple identity and frequency levels (corrected for 120 comparisons), marginally significant frequency effects appeared for quadruples 4 (W=51 503.5, p=0.079) and 6 (W=49 293, p=0.068). Combinations of quadruple identity and frequency level showed no significant differences in terms of onset, coda, or syllable variability, or the variability of context syllables (corrected for 120 comparisons).
Figure 20: Vowel variability difference between frequent and rare syllables in milliseconds, grouped by participant, quadruple and context comparison