From real-time measures to real world differences: New [and old] statistical approaches to individual differences in real-time language processing.

In the last 30 years, the Visual World Paradigm has rapidly become a dominant experimental paradigm for understanding real-time language processing. Part of this derives from the rich visualizations of the millisecond-by-millisecond timecourse of language processing that it offers. While the field has converged on strong statistical approaches for this in experimental paradigms, a great deal of recent work has sought to apply the Visual World Paradigm in individual differences paradigms to examine factors such as development and aging, multilingualism and various types of communicative and cognitive disorders. However, these models are less useful when confronting issues that are at the core of individual differences work: of collinearity and co-morbidity among predictor variables. We review a new approach that may complement existing approaches by capitalizing on regression-based techniques that were developed to handle these issues. We argue that analysis should start by developing a quantitative "profile" of processing that can be captured in a small number of index variables that capture meaningful dimensions such as the degree of competition or the speed of settling on a target interpretation. These can then be used in more sophisticated regressions like hierarchical regression, commonality analysis and mediation analysis which can unpack shared and unique variance, and test multiple causal pathways. We illustrate this with tutorials based on our own work on hearing loss, development and language disorders that illustrate how these approaches can provide greater insight into individual differences.