The cognitive neuropsychology of everyday action and planning

Abstract

Today's cognitive neuropsychology of everyday action had its inception in studies of ADS—action disorganisation syndrome—that were inspired by the Norman-Shallice theory of controlled and automatic action selection. It is now known that errors in everyday action and planning are commonplace with many types of brain damage, and that effects associated with gradations in severity, and with the presence of distractor objects, are surprisingly uniform across clinically diverse patient samples. The Norman-Shallice model of automatic action selection, having been implemented for two everyday tasks, has had some success in explaining these facts. This paper reviews the patient and modelling studies and discusses some unanswered questions and methodological challenges that confront future research in this area.

Acknowledgments

Preparation of this paper was supported by a grant from the NIDRR Research Field Initiated Program CFDA 84.133G (Laurel Buxbaum, Principal Investigator).

Notes

The vulnerability to errors of action in CHI and LCVA was expected, based on the literature on ideational and frontal apraxia. The high error rates in RCVA, on the other hand, were completely unexpected. Goldenberg (personal communication, January 2004) has recently completed a study of naturalistic action production in LCVA and RCVA which confirms that the two groups are equally vulnerable to errors of action.

To create “simulated patients,” the model was run many times at settings of a parameter that ranged from small deviations from default (corresponding to mild impairment) to large deviations (corresponding to severe impairment); these runs were then sampled in proportion to the severity distribution of the actual patient sample. The “clinical sample” used the largest possible sample from the simulation data to create another realistically biased sample, which was used to test empirical predictions. Readers are encouraged to consult the original manuscript for further details (Cooper et al., in press).

The quantitative fit is not perfect. For example, relative to the actual patients, the simulated patients have a tendency to make too many omissions. This shows up with particular clarity in the mild severity range, where most actual patients make no omissions but most simulated patients make some. As discussed in Cooper et al. (in press), omissions happen in several ways in the model, one of which is failure to complete a task subsequent to an error. Omissions of this type are commonplace in more severe patients (e.g., Schwartz et al., 1998), but my impression is that they rarely occur in those with mild deficits. Adding further error-correction mechanisms to the model might bring it more in line with the actual rate of omissions throughout the severity range.