How can we evaluate interference in attentional tests? A study based on bi-variate non-parametric tolerance limits.

In attentional tasks a basic performance is often contrasted with the same task administered with some additional load, defined here as "interference". However, it is questionable how interference should be quantified. The raw difference between the interference-loaded ("complex") task and the basic task is marred by measurement artefacts. There are alternative ways, but the choice of which solution to employ appears arbitrary. Bivariate non-parametric tolerance limits were introduced as they retain the information linked to the bivariate nature of the observation, and in this study we compared the different approaches for evaluating interference using real data from 209 normal participants who performed Visual Reaction Times (basic and go / no-go) and the Stroop Test. Our analysis indicated that, among the univariate indices, the use of the complex score covaried for the basic score yields the most satisfactory evaluation of interference; however, its use can be decided only after data inspection. Bivariate non-parametric tolerance limits offer advantages in terms of generality of use.