Usefulness of the Mean in Psychological Scaling of Cleft Palate Speech

Abstract

Psychological scaling techniques have frequently been used for research purposes of obtaining measures of various aspects of speech of individuals with cleft palate (3). These techniques all require judgments by observers, or listeners, for each of an array of stimuli, that is, in this case, an array of speech samples. Techniques which have been used to obtain measures for cleft palate speech samples include the method of pair comparisons, the method of direct magnitude—estimation, and the method of equal-appearing intervals. Of these three methods the most frequently used has been the method of equal—appearing intervals (5'). Application of the method of equal-appearing intervals (2, 7) includes rating by observers, or listeners, of each of an array of speech samples on a scale under the assumption that the interval between any two adja— cent points on the scale is equal to the interval between any other two adjacent points on the same scale. Often used is a seven-point scale extending from one, for least, to seven, for most, of the particular aspect of the stimuli to be rated. The present purpose is to evaluate the usefulness of one of two possible methods of computing equal—appearing-intervals scale values for samples of cleft palate speech for each of the following: nasality, articulation, language development, and general defectiveness. Scale values obtained by the method of equal-appearing intervals have customarily been the medians of the sets of ratings assigned by the observers to the stimuli (2). Use of the medians rather than the means avoids undesirable influence upon the scale values from possible skewing of distributions of ratings. That is, extreme ratings toward one end of the scale which are far from close agreement with the other ratings would have less effect upon the median than upon the mean of the set of ratings. The extremes of the scale, that is, one and seven on a seven-point scale, or one and nine on a nine—point scale, are identified by the extremes in the particular array of stimuli for which scale values are obtained. The points on a scale for one array of stimuli thus cannot be assumed to have the same value as the same points on a scale for another array of stimuli.