A Diffusion Model Decomposition of the Practice Effect

Article excerpt

When people repeatedly perform the same cognitive task, their mean response times (RTs) invariably decrease. The mathematical function that best describes this decrease has been the subject of intense debate. Here, we seek a deeper understanding of the practice effect by simultaneously taking into account the changes in accuracy and in RT distributions with practice, both for correct and error responses. To this end, we used the Ratcliff diffusion model, a successful model of two-choice RTs that decomposes the effect of practice into its constituent psychological processes. Analyses of data from a 10,000-trial lexical decision task demonstrate that practice not only affects the speed of information processing, but also response caution, response bias, and peripheral processing time. We conclude that the practice effect consists of multiple subcomponents, and that it may be hazardous to abstract the interactive combination of these subcomponents in terms of a single output measure such as mean RT for correct responses. Supplemental materials may be downloaded from http://pbr.psychonomic-journals.org/content/supplemental.

When people repeatedly perform the same task, their performance becomes fast, accurate, and relatively effortless. For example, you are able to read this text quickly, virtually without errors, and, hopefully, without investing too much effort. The difference between your performance now and when you first learned to read is staggering; from a slow, error-prone, and effortful endeavor, your reading has matured into automatized skill.

Traditionally, researchers in the field of skill acquisition have quantified the effect of practice primarily in terms of a reduction in the time to execute a given task (i.e., response time or RT; Logan, 1992; Newell & Rosenbloom, 1981; Woodworth & Schlosberg, 1954). Almost every study has shown that the RT benefits due to practice are greatest at the start of training and then slowly diminish over time.

This ubiquitous result, many researchers have argued, is best captured by a power function that relates the mean RT to practice via the equation

MRT = a + bN^sup -c^, (1)

where MRT is the mean RT for correct responses, a quantifies asymptotic performance, b quantifies the difference between initial and asymptotic performance, N represents the amount of practice, and c is the rate parameter that determines the shape of the power law. Empirical support for the power function relation between RT and practice has been reported across a range of tasks-for instance, in cigar rolling and maze solving (Crossman, 1959), fact retrieval (Pirolli & Anderson, 1985), and a variety of standard psychological tasks (Logan, 1992). Support for the power function relation has appeared so strong that the relation is often referred to as a law (e.g., "the ubiquitous law of practice," Newell & Rosenbloom, 1981, p. 3).

Nevertheless, some researchers have questioned whether the speedup with practice is really governed by a power function. In particular, Heathcote, Brown, and Mewhort (2000) argued that the power law is an artifact of averaging practice functions over participants (see also R. B. Anderson & Tweney, 1997; Myung, Kim, & Pitt, 2000). Heathcote et al. showed that for the data of many experiments on skill acquisition, individual learning curves were better described by an exponential function that relates mean RT to practice via the equation

MRT = a + b exp(2cN ), (2)

where the interpretation of the parameters is the same as in Equation 1.

Regardless of the specific shape of the function that relates the amount of practice to mean RT, the previous discussion illustrates that most empirical studies on the practice effect have focused on the decrease in mean RTs for correct responses. By doing so, the field has largely neglected two other important sources of information- namely, accuracy (i.e., proportion of correct responses) and the distribution of RTs (e. …