Flawed foundations of associationism? Comments on Machado and Silva (2007).

A. Machado and F. J. Silva have spotted an important conceptual problem in scalar expectancy theory's account of the 2-standard-interval time-left experiment. C. R. Gallistel and J. Gibbon (2000) were aware of it but did not discuss it for historical and sociological reasons, owned up to in this article. A problem of broader significance for psychology, cognitive science, neuroscience, and the philosophy of mind concerns the closely related concepts of a trial and of temporal pairing, which are foundational in associative theories of learning and memory. Association formation is assumed to depend on the temporal pairing of the to-be-associated events. In modeling it, theorists have assumed continuous time to be decomposable into trials. But life is not composed of trials, and attempts to specify the conditions under which two events may be regarded as temporally paired have never succeeded. Thus, associative theories of learning and memory are built on conceptual sand. Undeterred, neuroscientists have defined the neurobiology-of-memory problem as the problem of determining the cellular and molecular mechanism of association formation, and connectionist modelers have made it a cornerstone of their efforts. More conceptual analysis is indeed needed.