Distortion of ERP averages due to overlap from temporally adjacent ERPs: analysis and correction

In studies of event-related potentials (ERPs), short interstimulus intervals (ISIs) are often employed to investigate certain neural or psychological phenomena. At short ISIs, however, the ERP responses to successive stimuli may overlap, thereby distorting the ERP averages. This paper describes a signal processing approach for analyzing the distortion of ERP averages due to such overlap. In general, the distortion is modeled in terms of mathematical convolutions of the ERP waveform elicited by each type of adjacent stimulus with the corresponding distribution in time of those stimuli relative to the averaging epoch. Using this framework, a number of implications of ERP overlap for experimental design and interpretation are examined, with special emphasis given to selective attention paradigms. It is shown that the possibility of confound due to ERP overlap is widespread in short-ISI experiments, and even the widely used procedure of stimulus randomization does not necessarily control for differential distortion of the ERPs to attended versus unattended stimuli. Problems due to ERP overlap can be particularly serious in short-ISI studies that examine how ERPs (and associated perceptual processes) are influenced by the nature of the preceding stimulus (i.e., stimulus sequence effects). A set of algorithms is presented for estimating and removing the residual distortion due to response overlap from recorded ERP averages. The use of these algorithms, collectively termed the Adjacent Response (Adjar) Technique, can alleviate many of the overlap-related problems that arise when short ISIs are used, thereby enhancing the power of the ERP technique.