Perkins'  critique of cue–reactivity research is a needed stimulus for the examination of cue–reactivity research, but it is not a basis for abandoning this line of research. Perkins notes that few studies have linked cue–reactivity to relapse; however, the absence of evidence is not evidence of abstinence. It is especially noteworthy that the studies that have not found an effect have been relatively small and under-powered, and thus uninformative rather than definitively negative. Thus, Perkins' critique should be taken as a call for adequately designed and adequately powered studies to carefully examine this association, rather than a call to abandon this area of inquiry.
Perkins notes that a study from our laboratory  did show a relationship between cue–reactivity and relapse, but expresses puzzlement that the effect was seen only in a group treated with nicotine patches. In fact, Perkins' review provides the foundation for understanding this finding. Perkins rightly distinguishes withdrawal-based craving—which we have called ‘background’ craving [3,4]—from cue-based craving (‘breakthrough’ or ‘provoked’ craving), noting that nicotine patches reduce background craving substantially, but do not protect against cue-provoked cravings. Consider that both background and provoked cravings seem to promote relapse; i.e. they are competing causes of relapse. Accordingly, when smokers are not treated with nicotine replacement therapy (NRT), the high levels of background craving alone are sufficient to cause relapse in many, thus diluting and masking the effect of cue–reactivity on relapse. In contrast, when NRT suppresses background craving, the effects of cue-provoked craving in relapse emerge more clearly into relief (i.e. it increases the signal-to-noise ratio). Further studies of smokers treated with NRT are needed to confirm and explore this phenomenon.
Importantly, Perkins' account of cue exposure research omits one of the most important reasons that researchers and clinicians have become interested in cue exposure: most relapse episodes occur in the context of cue exposure and cue-provoked craving [5; see 4]. The vast majority of initial lapses occur in the presence of explicit cigarette cues, alcohol consumption and/or negative affect and are said to be triggered by these cues. Moreover, some of these same cues are also associated with smoking during ad libitum smoking . This is part of what makes the role of such cues seem so important, and reactivity to such cues seems to be a key part of the relapse equation.
None of this is to suggest that the laboratory cue–reactivity paradigm is perfect. Perkins' critique serves the field by stimulating re-examination and improvement of the paradigm. It is striking, as Perkins notes, that smoking itself has been used so rarely as an outcome in cue exposure studies. The existing literature is quite limited in other ways. To begin with, the literature has focused exclusively upon stable individual differences in reactivity, usually assessed after brief abstinence among subjects who are not actually trying to quit. However, it seems plausible that reactivity varies with subjects' state, and that the most relevant reactivity might emerge in more relapse-relevant contexts, i.e. among abstinent subjects who are undertaking a real quit effort. (Incidentally, this is the context in which Waters et al. observed an association with relapse.) It may also be important to study cue–reactivity jointly with background craving, rather than in isolation. Further, the range of cues studied has also been rather limited, most often focusing upon the cigarette itself. This is based upon the idea that these would be the most common, most proximal and most universal cues, but as lapses can be provoked by other cues, and because individuals may have differential sensitivity to various cues, such assessment may underestimate the relationship with relapse. Finally, individual differences in reactivity to cues are unlikely to be the whole story—reactivity becomes most relevant to the extent that the smoker is actually exposed to relevant cues at vulnerable times. No study has examined the interaction of individual cue–reactivity and the frequency or context of actual real-world cue exposures. There are many other ways in which cue–reactivity research can be improved. Let us set about improving the paradigm, and conducting adequate tests of the association with relapse, rather than throwing out the baby with the bathwater.