We use nine waves of BHPS data to examine interactions between spouses in terms of a behaviour with important health repercussions: cigarette smoking. Correlation between partners' behaviours may be due to correlated effects, as a consequence of matching or information revealed by others' behaviours, or to endogenous effects generated by bargaining within marriage. A simple bivariate probit reveals a positive correlation between own current smoking and partner's past smoking, which is consistent with endogenous effects. However, after controlling for individual effects, we find that own current smoking and partner's past smoking are statistically independent: all of the correlation in smoking status works through the correlation in individual effects. As such the correlation in the raw smoking data is consistent with positive assortative matching in marriage over smoking, rather than bargaining within the couple or social learning.