Thumbnail Image

Examining the Tools Used to Infer Models of Lexical Activation: Eye-tracking, Mouse-tracking, and Reaction Time

Most models of auditory word recognition describe the activation of lexical items in a continuous and graded manner. Much evidence in favor of these models comes from the visual-world paradigm, using either eye fixations or computer cursor trajectories as dependent measures. In particular, Spivey, Grosjean and Knoblich (2005) relied on their observation of unimodality in the distribution of cursor trajectories to argue in favor of a single cognitive process consistent with a continuous model of lexical activation. The present study addresses two questions: (1) whether the logic of inferring the number of cognitive processes from distributional analyses can be extended to a different dependent variable – reaction times, and (2) how robust the distribution of cursor trajectories is to changes in cursor speed (mouse gain). In Experiment 1, eye movements and reaction times were recorded in a visual-world paradigm and reaction times were modeled using ex-Gaussian curve-fitting. Participants responded slower to trials with a phonological competitor presented alongside the target than to trials with a control image presented alongside the target. Crucially, this difference was manifested as a shifting of the distribution rather than as a skewing of the distribution and lends additional support for a continuous model of lexical activation. Experiment 2 measured eye and mouse movements concurrently in a similar visual-world task to investigate the relationship between these two dependent measures at the level of the individual trial. In addition, Experiment 2 manipulated the speed of the cursor (mouse gain) between subjects. The low mouse gain served to reduce the effect of phonological competition. Moreover, the shape of the distribution of cursor trajectories across phonological competitor and control conditions was indistinct with low mouse gain, while the shape of the distributions across the two conditions differed with high mouse gain. This effect of mouse gain shows that the distribution of cursor trajectories is not robust to changes in mouse gain. Moreover, it raises questions about the strength of the linking hypothesis necessary to interpret the distribution of cursor trajectories.