Humans can be instructed verbally to perform computationally complex cognitive tasks; their performance then improves relatively slowly over the course of practice. Many skills underlie these abilities; in this paper, we focus on the particular question of a uniform architecture for the instantiation of habitual performance and the storage, recall, and execution of simple rules. Our account builds on models of gated working memory, and involves a bilinear architecture for representing conditional input-output maps and for matching rules to the state of the input and working memory. We demonstrate the performance of our model on two paradigmatic tasks used to investigate prefrontal and basal ganglia function.

M. Botvinick and D. C. Plaut (2006) argued that many of the criticisms of their earlier simple recurrent network (SRN) model of routine sequential action raised by R. P. Cooper and T. Shallice (2006) were criticisms of the specific implementation rather than criticisms of the underlying theory. Cooper and Shallice (2006) reject this assessment and raise concerns with several implementational adjustments that Botvinick and Plaut made to address their criticisms of the SRN account. Moreover, Botvinick and Plaut are questioned for not addressing potential interactions between their suggested implementational changes. Cooper and Shallice also reconsider the implications of the role of the training set in shaping the SRN model’s normal and error-prone behavior, the role of goals in their original interactive activation network model and routine behavior more generally, and the relation between the putative routine and nonroutine action control systems within the 2 models.

Abstract not found

We propose a model of routine sequence actions based on the Memory for Goals model. The model presents a novel process description for both perseveration and anticipation errors, as well as matching error data from a previously collected dataset. Finally, we compare the current model to previous models of routine sequential action.

Abstract not found

Humans can be instructed verbally to perform computationally complex cognitive tasks; their performance then improves relatively slowly over the course of practice. Many skills underlie these abilities; in this paper, we focus on the particular question of a uniform architecture for the instantiation of habitual performance and the storage, recall, and execution of simple rules. Our account builds on models of gated working memory, and involves a bilinear architecture for representing conditional input-output maps and for matching rules to the state of the input and working memory. We demonstrate the performance of our model on two paradigmatic tasks used to investigate prefrontal and basal ganglia function.