Psycholinguistics: Quantitative Methods

Abstract Psycholinguistic research methods like self‐paced reading and eye‐tracking during reading yield precise data in milliseconds that indicate how much time participants spend looking at and reading specific words in a stimulus sentence. Collecting and analyzing such time‐based data entails some special considerations and three fundamental ones are covered in this entry. First, the design of stimulus sentences is an exacting task that must address multiple extraneous variables, such as word length and frequency, which might otherwise affect the data and obscure the effects of interest to the researchers. The easiest and most efficient way to ensure experimental control is to design stimulus sentences that are nearly identical across conditions except for a single word that is related to the manipulated variable. Second, stimulus items are random rather than fixed, meaning that they are similar to participants in that they are a sample drawn from a larger population and that the goal is to generalize results to a broader population. Statistical analysis should therefore capture random variation between stimulus items. Current practice accomplishes this with mixed‐effects models that include crossed random effects for subject and item with unaveraged data. Finally, recent attention to effect sizes in applied linguistics research has generated interest in establishing specific effect size guidelines that are appropriate for research with different methods. For self‐paced reading and eye‐tracking, it appears that effect sizes tend to be small, probably because reading is a largely automatic and unconscious process and because random variation in the data is relatively high.