Measuring cognitive information processing


In order for the information contained by linguistic units (and, more generally, psychological stimuli of all kinds) to influence cognition, it is required that this information is processed by the mind. This is to say, it is not sufficient to show that the stimuli contain a certain amount of potentially relevant information and that in some ways this affects processing, but it is also necessary to directly show that this information is actually used by the cognitive system.

An intensely researched topic in language processing is how different corpus-derived measures of the frequency of linguistic constituents (ranging from the basic morphemes up to larger discourse units) influence their comprehension and production. An aspect of my research, stemming from my dissertation work, has been to clarify the effects of these measures on behavior. Starting from a characterization of the different factors of word forms and word meanings that affect the recognition of a word, I developed an information-theoretical account of morphological complexity measures that influence word recognition (Moscoso del Prado, Kostić, & Baayen, 2004). In further developments I have extended this work to show how behaviorally relevant measures of informativity can be fully derived from plain collections of texts (Moscoso del Prado, 2007) and how it can be used to characterize larger scale regularities of the morphological system (Milin, Filipović-Ɖurđević, & Moscoso del Prado, 2009).

A great part of my recent work is concerned with developing a method to determine the amount of information that has actually been processed by the cognitive system. This has resulted in a qualitatively different interpretation of reaction time data than that traditionally used (Moscoso del Prado, 2011; see also the highlight of this work that appeared in MIT Technology Review). This new approach uses insights from physics to provide a fundamentally new perspective on the study of cognition. This development has unearthed important properties of the cognitive system and the dynamical mechanisms that it employs for information processing. Among other things, I find evidence for an adaptive system that systematically adjusts online its rate of information processing to adapt to the requirements of the information being processed. Furthermore, a direct link can be established between measures of informativity and actual behavioral mechanisms (Moscoso del Prado, submitted).

Integrating the measure of complexity of cognitive processing, with an account of the complexity of words and texts is one my principal goals in the immediate future. This development would enable the study of linguistic, behavioral, and neurophysiological complexity, within the same mathematical framework.