In this talk, I contrast two fundamental theories within whose context experiments are designed, and the resulting data interpreted. By linear theory, which dominates research in psychology, a dependent variable is assumed to be the sum of a collection of main-effect and interaction terms. By dimensional theory, the independent variables in an experiment are assumed to yield values on some number of internal dimensions which, in turn, determine observed performance. I frame my discourse within an investigation of a well-known phenomenon in face-processing research, the face-inversion effect, which reflects the greater processing disadvantage of inverting a visual stimulus for faces compared to non-faces.
I describe two experiments in which faces or non-faces are shown upright or inverted in the study phase, but not in the test phase, of an old-new recognition procedure. In this talk (Vol I) the experiments all involve simulated data (in Kill SILL Vol II, which will occur during Winter quarter, I will describe additional experiments involving real data). Using the simulated data, I demonstrate two things. First, certain data patterns produce an inescapable theoretical conflict in that the conclusion about whether there is or is not a face-inversion effect depends entirely on whether the data are interpreted within the context of linear theory or of dimensional theory. Second, dimensional theory is more effective and consistent than linear theory in identifying the underlying psychological structures that generated the data: This happens because dimensional theory is intrinsically more flexible and realistic as a description of psychological processes than is linear theory.