Individual differences during reading via process-based eye-movement modeling

Advances in psychological science draw on experimental effects and associated individual differences. Increasingly, advances depend on multivariate statistics and computational modeling of these effects. In this case study, we examined correlation parameters of such effects that were estimated in linear mixed models fitted to eye-movement data from 275 readers as well as to 300 simulations from the SWIFT (2012) model of eye-movement control, implementing a narrowing of the attentional span and global inhibition when difficult (i.e., low-frequency) words are fixated. As hypothesized, the model dynamics reproduced a pattern of negative correlations between the fixated word frequency and its two neighbors’ frequencies as well as a positive correlation between the two neighbors found for human readers. Moreover, in simulations with model “lesions”, disabling the zoom lens overpredicted the correlation pattern whereas disabling global inhibition led to a qualitative incompatibility. Integrating experimental, differential, and computational psychology yielded theoretical insights.