In his target article Lockhead calls attention to numerous complications that prevent a valid straightforward or Fechnerian interpretation of psychophysical data. Here I describe three additional sources of bias, all involving the influence of stimuli presented before the start of an investigation.

[Comment, pp 759-780] Lockhead (1992) [Target Article] is undoubtedly right to attack so-called intensity scaling or the estimation of subjective magnitudes as an invalid perversion of tasks requiring quantitative judgments of aspects of objects, stuffs, and situations. He goes too far, however, in claiming that feature scales do not exist... ... A perceived physical pattern (sensory feature or channel) and the cognitive process that integrates it with its context are characterized by determining to which particular combination of specified stimulus patterns (...) the response is most sensitive. This is the principle of causal system analysis that is at the basis of all theory-building science. (shrink)

In terms of physics, mass is the fixed attribute of an object while weight varies with the accelerative force. Neither weight nor mass are simple sensory stimuli as both involve the integration of sensory and motor information with higher cognitive processes. Studies of apparent heaviness yield only vague information about sensorimotor mechanisms.

In contrast to Lockhead's view it is argued that psychology as a genuine science must not be based on other sciences and that psychological measurements have to be validated inside psychology. It is pointed out that psychological scalings, unaffected by judgment contexts, can be obtained if the experimental setting is compatible with everyday situations.

Psychophysical scales can be constructed under suitable restrictions from appropriate data, but they still do not justify privileged internal sensations.

Given that psychophysical responses are not a function of a single property but vary with a variety of stimulus and context variables, Lockhead has little hope for laws relating behavior to the environment. However, progress can be made with tasks that manipulate multiple sources of information to test formal information-processing models.