Influence of explicit standards on judgments with unbalanced response scales.

Abstract

In previous research (Weiss & Hodgson, 1963) , skewed distributions of response categories were shown to affect the absolute judgments of line-drawn squares by ptoducing a shift in the assignment of stimuli toward the segment of the scale which was overrepresented in the response system. The unbalanced scales had three categories on one side and only one on the other side of the midpoint interval. This effect was removed when a stimulus that was designated as an example of an average-sized square was shown from time to time, even though absolute judgments of the other stimuli were still required. In order to obtain a more sensitive and rigorous test of the potential influence of the unbalanced scale when a standard for judgment is available, it seemed desirable to carry out research in which Ss would serve in all conditions and, hence, intra-subject rather than inter-subject comparisons would be the basis of the analyses and in which the method of constant stimuli coupled with the rating-scale feature of the method of single stimuli would be used. Also, since the effect of the skewed scales when a middle standard is present might be evidenced only when discrimination is relatively poor, a more suitable test would require small stimulus differences between the standard and the immediately neighboring, series stimuli. Smdents (N = 27) at Hunter College judged the lengths of comparison lines thac were exposed one at a time in the window of a memory drum. A standard line, which represented the middle line o f the series, was shown for 2 sec. followed by a blank interval of 2 sec. and then a comparison line for 2 sec. From a complete set of 15 lines, forming an arithmetic progression from 13/16 in. to 3 3 /16 in., with 2/16 in. the constant addend, three partially overlapping subsets were formed; these comprised the lines designated 1-9, 4-12, and 7-15, in order of increasing length. Three response scales were used: a 3category balanced one, containing the categories + (larger), 0 (equal) , - (small e r ) ; and two 5-category unbalanced ones comprising +, 0, -1, -2, -3 or +3, +2, f 1 , 0, -. Nine different Graeco-Latin squares were formed. The cells represented the combinations of response scales with stimulus subsets; the columns, the order of the three combinations for any S; and the rows, Ss. Hence, each S was exposed to each scale and to each subset. Each set of lines with its assigned response scale was judged six times before the next combination was presented. Based on analyses of variance, there was no evidence rhat the nature of the scale had any differenrial effect on the number of stimuli judged smaller than, larger than, or equal to rhe standard, o r o n the size of stimulus iudged equal to the standard. This result was obtained whether the early trials, the later trials, o r all trials for a subset were examined, and whether only the responses to the three middle stimuli or to all stimuli were analyzed. Hence, the data confirm the outcomes of the research by Weiss and Hodgson (1963) and support their inference thac the effects of unbalanced scales are limited to situations where the average category of judgment is not held by some standard to a specified portion of the stimulus continuum.