- Cited by other publications
- Edinburgh Research Archive
- The Initiation of Sound Change: Perception, production, and social factors - Google книги
The estimates for all predictors in the analysis of the first two principal components of the spectral measures can be found in Table 4. Estimates for all predictors in the analysis of the first two principal components of the spectral measures. Recall that PC1 has the strongest loadings for peak frequency, spectral mean, and skewness; an increase in PC1 corresponds to an increase in peak frequency and spectral mean and a decrease in skewness thus less tilt toward the left. These predictions are indeed borne out.
- Simulation, Spectacle, and the Ironies of Education Reform (Critical Studies in Education & Culture)!
- The initiation of sound change : perception, production, and social factors.
- Sustainable Airport Construction Practices;
- From Innate Immunity to Immunological Memory.
- Java 2 Game Programming!
- Strengthening of Reinforced Concrete Structures: Using Externally-Bonded FRP Composites in Structural and Civil Engineering.
The vocalic effect also varies temporally. Moreover, the sibilant-specific effects of vowel also differ across measurement positions. S EX also interacts significantly with other factors. With respect to the second principal component PC2 , recall that PC2 has strong loadings for standard deviation and kurtosis; an increase in PC2 corresponds to a decrease in standard deviation and an increase in kurtosis i.
These predictions are only partially borne out. In terms of vocalic influence, males appear to exhibit more vocalic influences than females. The perceptual results also indicate that the participants, as a group, exhibit significant effects of perceptual compensation. However, the fact that the inclusion of by-subject random slopes for contextual factors i. The next section explores the nature of this variation in detail. This section considers the nature of individual variability in the perception and production of coarticulated speech by examining the link between them.
The regression models were fitted using the ddply function in the plyr package Wickham, The production measures PC1 and PC2 were modeled using linear regressions. The models included the fixed factors already mentioned above. Table 5 summarizes the predictors whose model estimates were used in the correlation study. Summary of predictors whose model estimates were used in the correlation study. Theoretical predictions: Before diving into the results of the correlation analysis, it is worth noting that, despite the admittedly exploratory nature of the correlation analysis, it is important to consider what correlations might be expected a priori.
As reviewed in the Introduction, the existing literature reported conflicting claims about the perception-production link. Some reported no relationship, while others found indirect or inconsistent evidence. The inconsistency in previous research findings might stem from the fact that the mapping between perception and production is not direct. As shown above, many factors and the interactions between them are responsible for explaining the variances in the perception and production results.
From a purely theoretical perspective, perception models that do not assume non-verdicial encoding of percepts predict a positive relation between perception and production. In its most basic form, one might expect the magnitude of perception compensation to find the analog in the degree of coarticulation reflected in production.
That is, for example, the coefficients for V OWEL from both the perception and production models would be expected to correlate positively. Models that assume non-veridical encodings, such as C-Cure and perceptual models implicitly assumed in listener-misperception models of sound change see more discussion below , would predict the opposite correlations.
That is, individuals with strong coarticulation in production should exhibit the least compensatory response or less context-sensitivity.
Cited by other publications
Results of the correlation analysis: Table 6 summarizes the correlation results. While several correlations show p values below 0. The sex of participant is indicated by the color of the symbol black for females and gray for males. Correlations between coefficients from the perception in columns and production in rows regression models. As the variables being correlated are estimates for interactions between predictors, it would not be feasible to interpret the nature of the correlations without first examining the nature of individual variability concerning a given interaction between predictors within each regression model.
The pattern differs for the individuals in the 1st quartile top panel. While individuals in the 1st quartiles data points toward the left end of Figure 4 exhibited some degree of perceptual compensation, as evidenced by the differences in identification functions across vowel contexts at the crossover point i. The production patterns mirror the perceptual patterns. In the preceding section, we observed significant correlations between individual-level estimates for predictors in the perceptual and production regression models.
These findings are most consistent with perceptual models that assume veridical encoding, such as the auditorist, the gesturalist, and the exemplar-based approaches to speech perception since there is a positive relationship between the perception and production of coarticulation. Our results resonate most strongly with the perception-production loop models such as Pierrehumbert and Sonderegger and Yu Specifically, the context-specific distinctness in the acoustic realization of the sibilants is reflected in shifts in context-specific category boundaries as well as the categoricity of the perceptual responses.
This finding echoes previous findings that show a correlation between how distinct an individual produces a contrast and how well that individual discriminates the contrast Newman et al. Our findings extend these earlier findings, demonstrating the context-specificity of such a contrast-based correlation.
Edinburgh Research Archive
Our findings are not consistent with predictions of perceptual models that assume non-veridical encoding, such as C-Cure Cole et al. Such models predict a negative relationship between the perception of coarticulated speech. That is, assuming that all listeners engage in expectation adjustment somehow, albeit to varying degrees, listeners engaging in weak expectation adjustment i.
A noteworthy aspect of our findings is the fact that the correlation between perception and production of coarticulation is much more nuanced than previous studies have generally assumed. That is, the correlations between the general effects of vocalic contexts on sibilant perception i. This fact is particularly striking since the vocalic context is a significant predictor in both group-level models for the perception and production data. Our findings offer potential insights into why previous studies that focused primarily on the extent of the coarticulatory effects on speech perception and production failed to observe a perception-production link.
The present findings suggest that what is important is not the extent of coarticulation per se , but rather the effects that coarticulation has on the realization of the segments targeted. Such findings are consistent with recent studies demonstrating that listeners have fine-grained sensitivity to acoustic—phonetic cues needed to track their distributions Clayards, Michael K. Further research is needed to ascertain the underlying mechanisms responsible for these individual differences in the perception-production linkage.
Individuals might differ in the type of coarticulated speech input they encounter, which could result in different distributions of context-specific perceptual exemplars cf.
Pierrehumbert, or different understanding of how articulatory gestures are coordinated in different vocalic contexts, to the extent the universalist understanding of gestural knowledge can be relaxed. Variability in coarticulatory experiences need not stem from source variability per se , however. Individual differences in cognitive processing style Yu, might also affect how individuals analyze coarticulated speech even if the input source is the same.
As noted in the introduction, one type of cognitive processing style difference that has been linked to variation in the perception and production of coarticulation is autistic-like traits. While this study does not investigate this factor explicitly, this variable is unlikely to be the type of cognitive processing style difference that mediates the link between the perception and production of coarticulated speech since the effect of AQ is found only within females and not among males at least according to the findings of Yu, Individuals might also vary in terms of oral-motor skills.
From an auditorist perspective, for example, individuals might differ in how well they are at uncovering strategies for producing optimally perceivable acoustic signals. However, existing evidence for individual variation in oral-motor skills remain scant cf. Future research might also elucidate potential lurking variables not tied to knowledge of coarticulation e.
The Initiation of Sound Change: Perception, production, and social factors - Google книги
The fact that the correlation between perception and production is modest deserves some attention as well. The strength of the correlation hovers around 0. Various factors might have contributed to this state of affairs. Also, the fact that the production data but not the perceptual judgments were obtained in lexical contexts might have constrained the strength of the correlation between perception and production.
Finally, as noted earlier, the perception and production tasks targeted different low vowels i. To this end, it is worth noting that Jongman et al.
- capillary gas adsorption chromatography;
- 8 Ways to Great: Peak Performance on the Job and in Your Life.
- ISBN 13: 9789027248411.
- Guide to the Traditions *OP (Mage: The Ascension)!
- The Cybernetic Brain: Sketches of Another Future.
- Most Downloaded Articles.
Given that F2 onset frequency was not a spectral measure included in the acoustic analysis conducted, this difference is unlikely to have affected the results of the perception-production correlation analysis in any significant way. Finally, as summarized succinctly in Beddor et al.
To the extent that listeners are able to adapt efficiently, the production system may not need to be similarly malleable. As such, the perception-production link is unlikely to be a perfect relation. In any event, despite these potential limitations, a significant correlation between the perceptual and production results is nonetheless found, which can be taken as further evidence of the robustness of the correlation observed. To this end, consider the relationship between the perceptual and production results of the three participants highlighted in Figure 6. Each row illustrates the perceptual-production patterns of one participant.