In this issue, Fidler and colleagues compare the ability of two standard measures of dependence and a measure of urges to smoke during a normal smoking day, for their ability to predict successful cessation 1. Perhaps surprisingly, the authors report that the single-item measure of urge was the strongest predictor of short-term abstinence. With this finding, Fidler and colleagues have added to the list of indicators of relapse risk; a list that includes measures of nicotine dependence 2 and responsiveness to smoking-related cues (i.e. cue–reactivity 3), among others. Such measures are potentially clinically relevant: if health-care professionals are able to identify interested quitters who are going to have difficulty maintaining abstinence then they can successfully triage the smokers whom they see, focusing their efforts on the smokers who are most likely to have the greatest difficulty quitting. In health-care systems with limited resources—and we know of no other kind—the ability to focus resources on smokers in the most need is likely to be beneficial. The obvious next question, however, remains: once identified, how do we help these particularly vulnerable smokers to maximize their chances of success? Sadly, despite years of research into cessation treatments, we have made precious little progress on this front—relapse rates for first-line treatments remain unacceptability high 4. Without substantial improvements in the available treatment options there is limited practical use in our ability to identify those smokers who are at greater risk of failure. The finding that craving intensity experienced during ad libitum smoking predicts relapse risk concurs with previous findings that craving intensity experienced during a quit attempt predicts relapse risk 5-7. It is puzzling, however—in terms of our understanding of the underlying drivers of relapse—that treatments successful at attenuating craving are not more successful at helping smokers to quit. For example, we have reported that nicotine patches suppress background craving (along with many commonly reported withdrawal symptoms) experienced during a quit attempt, but that this effect does not account for the efficacy of patches 5. This discrepancy suggests that urge itself may not be a direct measure of relapse risk. It is possible, however, that self-reported urge is a marker for more proximal factors that relate to relapse risk. Fidler et al.'s measure of craving asked smokers to summarize craving retrospectively over a 24-hour period. While, on the face of it, this would seem to refer to ‘average’ craving over that time, it is likely that such recall in fact gives much greater weight to moments of intensive or ‘peak’ craving 8. This would probably reflect smokers' response to smoking cues, or susceptibility to cue-induced cravings, which we know are important in relapse 9, 10, and are not treated adequately with current regimes 9, 11. Gaining a better understanding of the nature of craving and response to smoking cues may suggest new, and hopefully more effective, treatment strategies. In conclusion, Fidler and colleagues' work provides evidence that a simple one-item measure of urge intensity can be used to successfully predict relapse risk, and suggests the need to explore further the dynamics of craving during ad libitum smoking. The development of effective treatments is necessary so that we can better assist interested quitters who are at a high risk of failure. Through their work at Pinney Associates, Drs Ferguson and Shiffman serve as consultants to GlaxoSmithKline Consumer Healthcare on an exclusive basis on matters relating to smoking cessation. Dr Shiffman also has an interest in a venture to develop a new nicotine replacement medication. Dr Bruno has nothing to disclose.