Context Effects as Foundation for Unified Theory

Is unified theory possible? Unification has been lost to sight amid the increasing fragmentation of the psychological field. Unification may be possible, nevertheless, by primary focus on two problems that are fundamental in every area. These two problems are context-stimulus interaction and stimulus-stimulus integration. In principle, both problems may be solved jointly with algebraic models of stimulus integration. In practice, this integrationist approach has been reasonably successful across diverse fields of psychology. A key empirical result is that valuation and integration are distinct modules. Valuation modules allow for context-stimulus interaction, which is ubiquitous. Integration modules provide true measurement of context-stimulus interaction. Most important, these results imply that context effects provide a foundation for developing unified theory. Is unified theory of psychology possible? This goal, advocated long ago by Fechner, has disappeared in the maelstrom of fragmentation that characterizes current psychology. This fragmentation has come to be accepted as normal. Mini-theories are plentiful, but mini-theories do not lead to unified theory. FOUNDATION FOR UNIFIED THEORY To attain unified theory requires conceptual reorientation, based on these two axioms: Axiom of Purposiveness. Axiom of Stimulus Integration. Axiom of Purposiveness. The axiom of purposiveness is manifest in our sensory systems, most of which are affective, as with taste, temperature, and sex. These sensory systems embody a pleasure–pain axis that subserves reactions of goal-approach and goal-avoidance. Purposiveness thus confers a priceless simplification — by reducing complex reality to one-dimensional values of approach and avoidance. Although this one-dimensional representation omits much, it captures a central characteristic of perception, thought, and action. Axiom of Stimulus Integration. The axiom of stimulus integration defines a key issue for unified theory. This axiom reflects the evident fact that all perception, thought, and action depend on integrated action of multiple stimuli. The "taste" of a food, to take an everyday example, depends on odor, temperature, texture, and visual appearance, as well as sweet-sour-salt-bitter. Quest for unification must thus seek general laws of stimulus integration  a focus on Fechner's inner psychophysics. Measurement: A Critical Difficulty. A critical difficulty confronts the search for laws of stimulus integration. Linear (equal interval) scales of the stimulus variables, and of the response, are needed. To illustrate, consider the hypothesis that the size−weight illusion obeys an addition law. The psychological sensation of heaviness is then the sum of the psychological sensation values of the two stimulus variables of weight and size: ρ Heaviness = ψ Gram Weight + ψ Size . To verify this equation, we must be able to measure the three terms on true subjective scales. Without true psychological measurement, we cannot establish even this simple addition law. But if we cannot deal with this simple case, we can hardly hope to deal with more complex integration. This problem of psychological measurement has been controversial ever since Fechner's proposition that just noticeable differences are equal psychologically and can be used as additive units to determine psychological scales. Functional Measurement Theory. The problem of psychological measurement has been solved in many situations with functional measurement theory. The logic of functional measurement is to use algebraic laws of stimulus integration as the base and frame for psychological measurement. The simplest form of functional measurement is the parallelism theorem for addition laws. Parallelism Theorem. Let SAj and SBk be stimulus levels for two factors, manipulated in a two-way design. Denote their psychological values by ψ Aj and ψ Bk. Let ρjk and Rjk be the implicit and observed responses, respectively, to the stimulus combination, {SAj, SBk}. Two premises are employed: ρjk = ψAj + ψBk ; (Premise 1: addition) Rjk = c0 + c1 ρjk (Premise 2: linear response scale) The linearity premise says the observable response, Rjk, is a linear (equal interval) function of the implicit response, ρjk (here c0 and c1 are zero and unit constants that may be set at 0 and 1 for simplicity). Two conclusions follow: Conclusion 1: The factorial graph will be parallel. Conclusion 2: The row means of the factorial data table will be a linear (equal interval) scale of ψ Aj ; the column means will be a linear (equal interval) scale of the ψ Bk . Note the simplicity of this parallelism analysis: Just graph the subject's responses and look. Three-Fold Benefit. Observed parallelism provides support for three conjoint benefits. 1. True measurement of the psychological values of the stimuli. Prior stimulus measurement is not needed. The functional stimulus values derive from Conclusion 2. Note especially that the psychophysical law follows directly from Conclusion 2. It is just the function that relates the sensation values provided by the parallelism theorem to the physical values of the stimuli. Success of the parallelism property can resolve the long controversy over the psychophysical law. 2.Exact structure of the integration law. Observed parallelism supports Premise 1 of additivity. This is strong support although not absolute proof. Logically, it is possible that nonadditivity in the integration is exactly cancelled by nonlinearity in the response measure. This logical possibility no longer seems serious (Anderson, 1996, pp. 45f, 94-98). 3. True measurement of the response. Observed parallelism supports response linearity (Premise 2). Response linearity is not an assumption; it is tested by the parallelism analysis. Response linearity has fundamental importance, as noted later under Response Generality. Algebraic Psychology. Whether functional measurement will be useful depends on whether algebraic laws have empirical reality. The parallelism theorem has minor value unless Nature has endowed organisms with adding-type rules. Happily, extensive empirical work has revealed algebraic laws—addition, multiplication, and especially averaging—in almost every area of psychology. To establish algebraic psychology faced serious obstacles. One obstacle was that by far the most common algebraic law is averaging, not addition. But averaging yields parallelism only in the special case of equal weighting. If two levels of a factor carry unequal amounts of information, the data will be nonparallel. Observed nonparallelism was not infrequent in the initial work, therefore, but it was ambiguous, for it could merely reflect nonlinear response. This seemed the more likely because we were using a rating method, which was universally condemned because of well-known nonlinear biases. The initial work was thus perplexing. Fortunately, this obstacle was overcome, in part by developing simple procedures to linearize the rating method. Indeed, some of the strongest evidence for functional measurement came from experimental manipulations of the weights and verifying predicted deviations from parallelism. Conceptual Implications. Conceptual implications of algebraic laws are more important than measurement per se (Anderson 1996, p. 467). One conceptual implication is value invariance—the value of one stimulus does not depend on the other stimuli that are being integrated. This invariance property is implicit in Premise 1. If instead ψAj depended on SBk, the data would generally violate parallelism. Note that value invariance does not deny context effects. On the contrary. The axiom of purposiveness implies that the value of any stimulus is generally very sensitive to context. Context-stimulus interaction is thus ubiquitous. But stimulus-stimulus interaction during integration is relatively rare. A related conceptual implication is that valuation and integration of stimuli are distinct operations, or modules, to use a current term. This independence of valuation and integration is a wonderful blessing of Nature; it provides a rock of stability in the sea of context-dependent values. This valuation–integration independence is a key to unified theory because it has a cutting edge with algebraic laws. Context effects, in particular, can provide new tools to study inner psychophysics.