Methods for modular modelling: Additive factors and cognitive neuropsychology

Abstract

Theorizing about how people perform any cognitive information-processing task typically takes the form of proposing a modular model of the cognitive system that people use to accomplish that task. Some of these models are stage models; many are not. In particular, models in which the passage of information from one module to another is cascaded rather than discrete are currently very popular, but these by definition are not stage models. The additive factor method as described by Sternberg (2011) is designed specifically for working with stage models. How useful is it for theorists whose models are not stage models? The goal of the additive factor method is the verification or discovery of the parts (modules) of cognitive systems. That is also the goal of the method of cognitive neuropsychology. I concur with Sternberg's view that these are complementary methods that can inform each other.

Keywords:

• Modularity
• Additive factors
• Computational modelling

Acknowledgments

I thank Saul Sternberg, Derek Besner, Claudio Mulatti, and Veronica Cembrani for much useful discussion.

Notes

¹ Sternberg refers to this kind of practice—I think with approval—as a “fishing expedition”. I share his approval of scientific fishing expeditions. Not all journal-article reviewers do.

² Here the output from a module is thresholded. As pointed out to me by Claudio Mulatti, one can also imagine input thresholding: A module does not respond to input until the strength of that input exceeds some threshold value.

³ Such additivities are seen only when at least some of the items in reading-aloud experiments are nonwords, a point to which I will return.

⁴ Alternatively, it might be possible to capture the additive effects by making some other module of the DRC model—the letter identification level, for example—a stage, as Besner et al. (2010) have suggested.