This paper addresses two aspects of the model of rational addiction: forward looking behavior and time consistent preferences. It explores smoking by women before, during and after pregnancy using the European Community Household Panel (ECHP).Pregnancy is used as an instrument for a partially predictable future decrease in smoking. Women reduce the average number of cigarettes they smoke and many quit in the period 10 to 15 months before the birth of a child. Our analysis suggests that this effect may be stronger for married than for unmarried women, corresponding to the higher probability that the pregnancies of married women are planned. Pregnancy is also used as an instrument to estimate the parameters of a structural model of addiction. The estimates imply that cigarettes are highly addictive. Finally, we present statistically significant evidence that, even when the expected number of cigarettes smoked one month after the interview is taken into account, expected smoking further in the future has an independent effect on current consumption. This effect remains even when we impose the highest theoretically possible coefficient on expected cigarettes smoked one month after the interview. This means that the null of time consistency is (barely) rejected against the alternative of time inconsistency.