Abstract
Inter-individual variability in global/local processing is often quantified through performance on the well-known hierarchical letters task (Navon, 1977). Here, a global letter stimulus is composed of local letters and participants perform a task (e.g., identification) on the global or local letter(s) in global and local attention conditions, respectively. Congruency between levels (e.g., same or different letter) is manipulated. Global precedence refers to faster average response times in the global versus local attention condition, whereas global/local interference refers to slower response times in local/global attention conditions for incongruent versus congruent stimuli. The size of these reaction time differences is used as a quantification of the ease with which individuals can identify the global letter (precedence) or are hampered by incongruency at the other level (interference). However, a quantification of the statistical evidence for or against individual differences in either measure rarely occurs. In this study, we applied order-constrained modelling (Haaf & Rouder, 2017) to three different data sets (N = 275, N = 116, and N = 40) from three different labs. This technique allows researchers to quantify the evidence for inter-individual differences in precedence and interference and, if present, whether these effects vary are in the same direction for everyone. In our two large-N data sets (with only 50–80 trials per condition), we observed that everyone shows precedence, with unconstrained variability (local and global precedence). Furthermore, everyone shows global and local interference, without inter-individual variability. This implies that precedence is not global for everyone and that interference does not vary between individuals, calling into question its usefulness as an individual difference characteristic. However, the data set with 40 observers and 320 trials per condition showed variability for global interference, and zero local interference. A proper quantification of variability in interference thus seems to require many more trials per observer than standard practice.
Acknowledgement: Research Foundation Flanders