The benefits of accurately reading a partners’ emotions may depend on whether those emotions threaten the relationship. Married couples (N=140) were videotaped during a 15-minute conflict conversation. Later, individuals viewed the video and continuously rated their own affect and (separately) their partner’s affect. Marital satisfaction was the average score across the Marital Adjustment Test and the Marital Relationship Inventory. Average subjective affect was the mean rating dial position for self-rated affect. Average perceived affect was computed similarly for partner-rated affect. Illusory positivity was inferred by controlling for partner’s average subjective affect when using average perceived affect as a predictor. Empathic accuracy was the cross-correlation of the two time series of self-rated and partner-rated affect (3-sec lag). Our hypothesis was that when illusory positivity is high, low empathic accuracy will not undermine marital satisfaction. We used Actor-Partner-Interdependence Model (APIM) to estimate the predicted model with good model fit. Beyond husbands’ average subjective affect, wives’ empathic accuracy interacted with wives’ average perceived affect to predict both wives’ (b=-11.614, p=0.037) and husbands’ marital satisfaction (b=-14.812, p=0.007). No parallel effects emerged for husbands. This pattern suggests that wives’ (not husbands’) positive illusions compensate for low empathic accuracy in predicting both partners’ marital satisfaction.