This research provides an experimental test of the self-esteem hypothesis that avoids potential hypothesis guessing and self-presentational concerns associated with previous research by including subtle measures of both social self-esteem and intergroup discrimination. The role of group identification and social self-esteem as determinants of ingroup bias was examined under high and low identity-threatening conditions utilizing an implicit measure of social self-esteem. Participants read a fictitious statement indicating whether their university received a good or bad evaluation relative to a rival university and then made attributions for this situation. High group identifiers had a greater decrease in implicit social self-esteem after a threat than low group identifiers did and they displayed the greatest ingroup favoritism. Greater ingroup-serving bias was associated with a subsequent increase in implicit social self-esteem.