Graduate Programs in the Life Sciences

Research Focus

All of my research and teaching have been characterized by commitments to (a) rigorous research that relies on the experimental method, (b) the application of mathematical models in research, and (c) the development of predictive theories of cognition. Concerning (a), when it comes to the gain in reliable knowledge that can be derived from research, the experimental method is without parallel. Although important information can be gained from descriptive correlational investigations, especially in the early stages of research, only carefully-counterbalanced studies that are conducted under controlled conditions are able to separate the factors that influence cognition from each other and to establish the cause-effect relations that are essential to theoretical interpretation. For that reason, although I have occasionally published descriptive correlational studies, virtually all of my research has been characterized by hypothesis-driven experimental designs. Concerning (b), a standard conundrum in research on cognitive functions is that the procedures that are used to measure human cognition, from the simplest attentional task to the most complex reasoning task, are not process-pure. That is, performance on such tasks is invariably due to a mixture of contributing cognitive processes. Therefore, by definition, the respective contributions of those processes cannot be separated, let alone quantified, using task performance as the sole source of dependent variables. To accomplish separation and quantification, it is necessary to implement the target processes in a mathematical model that is defined over the task in question and to estimate the model's parameters from task performance. Because it is the parameters that carry information about the target processes, data analysis then centers on estimated values of those parameters, rather than on task performance. This, in turn, allows theoretical hypotheses to be directly evaluated, rather than indirectly evaluated under weak measurement assumptions. An additional strength of model-based research that redounds to the benefit of knowledge gain is that theoretical ideas must be sharpened to a fine edge before they can be embedded in mathematical models. In light of all these advantages, there is no cognitive task that I have studied extensively for which I have not also developed relevant mathematical models. A few examples include Markov models of true recognition and recall, Markov models of logical and scientific reasoning, multinomial models of mathematical reasoning, and multinomial models of false recognition and recall. Last, concerning (c), one of the abiding limitations of cognitive theories is that they emphasize explanation at the expense of prediction. That is, what counts as an acceptable theory is a body of assumptions that can cover the available archive of research findings, hopefully with due regard to parsimony, with little consideration being given to whether new and surprising results fall out of those same assumptions as predictions. However, explaining data is the easy part of theorizing, and experience has shown that is normally possible to formulate multiple competing explanations of any set of data. The more formidable part of theorizing is develop a set of assumptions that will also predict novel findings that would not have been envisioned without those assumptions. That is the type of theorizing that I have stressed in my research and teaching, and fuzzy-trace theory is a prime example. In addition, it is my emphasis on predictive theorizing that is responsible for the fact that my experimental work has so often been concerned with counterintuitive findings--for instance, the cognitive triage effect, the false persistence effect, the mere-testing effect, the memory-independence effect, the memory-interference effect, and the phantom recollection effect.