Artifacts in the Registration and Interpretation of Speech-Process Variables

Abstract

Of the different interacting characteristics of the speech process such as level, intensity and range of voices, on-off-patterns etc., the on-off-patterns are the most readily formalized and objectively recorded; hence the appearance of this feature in particular in some hundred publications. However, the results of different research teams investigating the same subject-matter sometimes vary considerably. In this article the authors report on their duplication of some of these published investigations, using an automatic speech analyser comparable to Cassota, Feldstein and Jaffe's AVTA. Their findings were similar to those of some other research workers, but they found evidence for the hypothesis that these results are based on technical artifacts and that some correlations obtained, for which psychological interpretations have been given, are in fact due to minor, unacknowledged circumstances of recording. The authors comment on these findings in the context of a broad discussion on the problem of artifacts and wrong conclusions arising from weaknesses of technique, interpretation and model construction.