Little is known about when and why observers of harassment incidents choose to intervene. On the basis of a model of observer intervention by Bowes‐Sperry and O'Leary‐Kelly (2005), we examined situational (directness of harassment, relationship to target, knowledge of target orientation, recurrence beliefs) variables as potential predictors of observer decisions to intervene when witnessing sexual orientation harassment. We also examined how costs and benefits analysis and recurrence beliefs relate to the level and immediacy of involvement. Relationship to the target, directness of the harassment, recurrence beliefs, knowledge of target orientation, and costs evidenced some connections to the nature of intervention. We discussed implications for organizations. Copyright � 2011 John Wiley & Sons, Ltd.