Sensory scaling: Unanswered questions

Despite the range of literature he refers to, Krueger fails adequately to consider alternative interpretations and this limits the weight that can be given to his conclusions. His argument relies on underlying assumptions that can be questioned and have not been justified. In the space available I shall attempt a brief discussion of some of these difficulties. (1) What Is It that Is measured? Krueger refers to "subjective magnitude (conscious sensation)" and proposes that "sensations are phenomenally experienced (i.e., introspectable via direct magnitude estimation and category ratings)." In normal scientific practice, the claim that a procedure delivers a measure of a variable requires that we define the variable and offer some validation of the proposed measurement procedure. Thus, "motivation" might be defined as the determinant of certain aspects of behavior, and a presumed measure of motivation would be validated by examining correlations with behavior. But we have no clear definition of "subjective magnitude" and we are not offered any procedure for validating proposed measures of it. Instead, discussion in this area usually rests on intuitive convictions. To illustrate, suppose that a loud noise is presented to a subject and he responds by emitting the number "ten"; a low noise is presented and his response is "five." It may seem intuitively evident that each noise produced a unique subjective experience; the sensory experience produced by the loud noise is different from that produced by the low noise; and "ten" and "five" capture the difference between the former and the latter. But this subjectivist view (SV) is at best an intuitive correlation, and we know that correlation does not guarantee a causal relation. The leading proponent of this view was Stevens (1957b; 1960); Krueger's arguments follow the same line. But other views are possible. If, to avoid relying on the untestable basis of intuition, we follow the example of attempts to define the concept of "intelligence" and define the sensory scale as that which sensory-scaling procedures measure, we arrive at an information-processing view (IPV) of sensory scaling (Treisman 1962; 1964a; 1964b). According to this view, the response "ten" is determined by underlying neural processes consequent on the physical stimulus processes that at best parallel in some intuitive sense the corresponding "subjective experiences"; any scale we derive from such responses can only refer to the underlying physical processes. In a procedure such as magnitude estimation, the data are in the form of a table of correspondences between stimulus intensities, /, and corresponding mean responses R. An empirical relationship is found between the two, such as R = alp. The IPV rests on the observation that any such relationship can be expanded into a pair of equations that include an intermediate measure S. Then S = f{I), R = g(S), where/and g are any functions (out of an infinite set of possible pairs) such that g[/(/)] = O/P. (More than two stages are of course possible.)