Separate and joint evaluations differ in the manner in which attribute information is processed. We propose that negative attributes weigh more in separate compared to joint ratings. To assess the relative weights of positive and negative attributes, stimuli were employed containing mixed attribute valence, with good and poor attributes. In three experiments it was demonstrated empirically that, using such mixed valence options, the negative attributes indeed weighed more than the positive ones in separate compared to joint ratings. Direct evidence for the increased weight of negative attributes in separate ratings was provided by showing preference shifts between joint and separate ratings for options with mixed attribute valence.