Ambivalence is widely assumed to prolong grief. To examine this hypothesis,
the authors developed a measure of ambivalence based on an algorithmic com
bination of separate positive and negative evaluations of one's spouse. Pre
liminary construct validity was evidenced in relation to emotional difficul
ties and to facial expressions of emotion. Bereaved participants, relative
to a nonbereaved comparison sample, recollected their relationships as bett
er adjusted but were more ambivalent. Ambivalence about spouses was general
ly associated with increased distress and poorer perceived health but did n
ot predict long-term grief outcome once initial outcome was controlled, In
contrast, initial grief and distress predicted increased ambivalence and de
creased Dyadic Adjustment Scale scores at 14 months postloss, regardless of
initial scores on these measures. Limitations and implications of the find
ings are discussed.