SummaryOver the past 15 years, researchers have shown an increasing interest in using event-related potentials (ERPs) to study depression. These studies generally fall into four classes: 1), ERPs as a means of detecting depression; 2), ERPs as a tool for distinguishing subtypes of depression; 3), ERPs as a measure of pharmacological effectiveness; 4), ERPs as indicators of defective cognitive operations in depressed subjects. Results from these heterogeneous approaches are often inconsistent and disappointing. Although some ERP components often show increased latencies and diminished amplitudes, these changes seem to reflect principally a variety of non-specific disorders affecting a wide range of cognitive functions rather than a precise and consistent deficit of a particular function. These disappointing results seem to be attributable to methodological problems (heterogeneous patient populations, disproportionate use of the odd-ball paradigm), and do not necessarily call into question the value of studying the ERPs. Furthermore, recent advances in ERP methodology have opened up new perspectives for ERP use in psychopathology.