Regression analyses presented in the IVIV and VIVI tables tested the possibility that confidence variables represented an additional unequivocal variance in the transcription agreement, which goes beyond canonicality. The two analyses also assessed the relative contributions of canonity and the confidence variables modeled in two ways. In fact, both analyses showed an increase in the R square, which was taken into account when the confidence variables of the canonical variables were added in Model 2. Although the two confidence measures (canonicity confidence and confidence in transcription) showed a zero-order correlation greater than canonicality alone in analysis 21, the role of canonity in the regression equation was more important in the analysis of statements 21, where canonity played a statistically significant independent role in the influence of transcription chord. While the analysis of 30 vocals showed a change in square R when confidence variables were added to the model, which were more than 1.5 times the R square that was solely responsible for canonity, the R-square change in language analysis 21, although still significant, was not as dramatic. In particular, the overall gap, attributable to the complete regression equation, was significantly higher in external analysis 21 (622) than in the 30-outside analysis (.385). The absolute increase in square R variation, which is due to confidence variables, was comparable in the two analyses (243 and 243 respectively . 230). The results here show that the relationship is reliable. In analyses 30 and 21, correlations between confidence and canonity were statistically significant, and variance shares were 0.45 and 0.27, respectively.
The results indicate that confidence in transcription systematically predicts canonity. As with the relationship between canonicality and transcription agreement, if we could achieve a more reliable degree of canonity, we might expect an even greater correlation between canonity and confidence in transcription. In addition, the results of the additional sample provided additional support for a strong relationship between coders` trust and transcription agreement. Coders have statistically reliably shown greater confidence in canonical statements than non-canonical statements. In addition, the magnitude of the impact of canonity in the analyses available on the primary sample may have been influenced by variability in the assessment of canonity itself. Coders varied greatly how often they reported segments as non-canonical. The coherence between donors on canonity itself could be greatly improved through more comprehensive and edified training that was not at the centre of this study. Our approach here, for the primary sample, was to define canonity and illustrate the definition during trainings conducted in a single week. Given the variability between coders in the selection of a canonical assessment criterion, it is reasonable to speculate that the correlations we achieved between canonity and transcription agreement were lower than they would have been if the canonical variable had been controlled by additional formations. In addition, a fully instrumental acoustic approach to determining canonity could in the future improve judgment on canonity and lay the groundwork for a more accurate assessment of the correlation of the role of canonity in transcription agreement.