Even with more successful smoking cessation treatments, relapse continues to be a major issue, and the most common outcome of a smoking cessation attempt (Hartmann-Boyce et al. 2018). A widely used model of the maintenance and relapse process considers that attempts to maintain abstinence are punctuated by periodic temptations in which intense craving is experienced, and the likelihood of lapsing is substantially increased (Businelle et al. 2016; Ferguson and Shiffman 2014; Marlatt and Gordon 1985; O'Connell and Martin 1987; Shiffman et al. 1996b). A robust literature, going back decades, documents the circumstances surrounding temptation episodes and lapses in daily smokers (DS; O'Connell et al. 2008; Shiffman 1982; Shiffman et al. 1996a). This research finds that these temptations are often triggered by situational cues (Shiffman 1982; O'Connell and Martin 1987; Shiffman et al. 1996b). This might be expected to be particularly true of non-daily, intermittent smokers (ITSs), since such cues are especially associated with their ad libitum smoking (Shiffman et al. 2014a).

Although some cues may be idiosyncratic, studies of lapses in DS have identified some cues that generally promote temptation and lapsing. Two prominent and powerful contexts associated with temptations and lapses are consuming alcohol (Bold et al. 2016; Businelle et al. 2016; Lam et al. 2014; Shiffman 1982; Shiffman et al. 1996b; Van Zundert et al. 2012) and seeing others smoke (Bolman et al. 2018; Borland 1990; Businelle et al. 2016; Lam et al. 2014; O'Connell and Martin 1987; Shiffman 1982; Shiffman et al. 1996b). Both of these may be expected to be strongly associated with ITS’ temptations, since both have particularly strong relationships with smoking among ITS. In a direct comparison between ITS and DS smoking, alcohol and seeing smoking doubled to tripled the degree of linkage to smoking among ITS versus DS (Shiffman et al. 2014a). Other factors, such as being at work, or being engaged in work, had similar smoking-suppressing effects on ITS and DS.

In contrast, negative affect was not associated with ad libitum smoking in either ITS or DS (Shiffman et al. 2014a), yet is very strongly associated with temptations and lapses in DS (Lam et al. 2014; Marlatt and Gordon 1985; Minami et al. 2014; O'Connell and Martin 1987; Shiffman 1982; Shiffman et al. 2007; Shiffman et al. 1996b; Shiffman and Waters 2004). Some accounts of relapse (Edwards and Kendler 2011) interpret this to indicate that nicotine withdrawal, which is marked by negative affect and by affective reactivity (Hughes 2007; Hughes and Hatsukami 1986), plays a major role in triggering lapses. Thus, a simple model of dependence based on nicotine regulation and withdrawal avoidance may explain the role of negative affect in DS’ temptations as due to withdrawal. However, this explanation would not apply to ITS, who do not experience withdrawal symptoms such as negative affect or craving when abstaining (Shiffman et al. 2015a). There were no increases in negative affect or craving when ITS abstained, and this cannot be explained by positing that ITS are perpetually in withdrawal due to their inability to maintain nicotine levels, because ITS’ negative affect and difficulty concentrating even when not smoking are similar to those of DS smoking ad libitum, and their craving is lower (Shiffman et al. 2014b).

Thus, given that ITS do not experience withdrawal, a simple nicotine regulation withdrawal avoidance model of dependence would predict no role for negative affect in ITS’ temptations. However, we have previously proposed that such a model is incomplete, as it does not take into account factors such as automaticity (Tiffany 1990) or the important role of cues and stimulus control in maintaining smoking. We have proposed a two-factor model of dependence (Shiffman et al. 2015b) in which stimulus control of smoking by cues plays an important role. We suggested that stimulus control can itself lead to a kind of behavioral dependence that affects ITS, and helps explain their difficulty quitting smoking. On this model, the role of stimulus control in maintaining smoking is not limited to ITS, but also plays a part in daily and dependent smokers’ smoking, but is diluted and masked by nicotine regulation during ad libitum smoking. This is consistent with analyses contrasting ITS’ and DS’ self-reported motives for smoking using the WISDM (Shiffman et al. 2012a. ITS gave greater relative weight to Secondary Dependence motives such as cues, but DS also scored high on this dimension, giving it almost equal weight with craving and tolerance, and greater weight than automaticity. Thus, stimulus control is important to DS as well as to ITS, but is, we posit, diluted during ad libitum smoking. We have suggested that stimulus control emerges as dominant in abstinence, helping to explain the strong role played by smoking-relevant cues (Ferguson and Shiffman 2014), even when nicotine is completely replaced, and even in the absence of withdrawal (Ferguson and Shiffman 2009). Thus, it is possible that the negative affect typically associated with DS lapses may not be due to withdrawal, but may arise from everyday sources of affective distress, and act as a smoking cue. Negative affect may also affect temptations and lapses by draining the mental resources (Ferguson and Shiffman 2014; Muraven and Baumeister 2000; Drobes et al. 1994) that are thought to be needed for effective coping and cognitive self-control (Muraven et al. 1998). These non-withdrawal explanations would be made more plausible if a link between negative affect and temptation and lapsing were also observed among ITS. Accordingly, we sought to assess the role of negative affect in ITS temptations and lapses.

The role of DS’ coping responses in resistance to temptation—i.e., keeping a temptation from progressing to a smoking lapse—has been shown in several studies (O'Connell et al. 1998; Shiffman 1984; Shiffman et al. 1996b). The research has been mixed on the relative efficacy of cognitive and behavioral coping responses, respectively, and also as to whether engaging in both kinds of coping is superior to just doing one. The previous studies have been based on heavy DS; coping effects among ITS, whose temptations may be milder and easier to overcome, are not known.

It is important to better understand smoking cessation and relapse process among ITS, as they now constitute 25–33% of all adult smokers in the USA (Reyes-Guzman et al. 2017). Moreover, despite their lack of significant dependence (Shiffman et al. 2012b) and the absence of withdrawal when they abstain (Shiffman et al. 2015a), they have very low success rates in smoking cessation, and do make use of smoking cessation treatment (Tindle and Shiffman 2011).

In this paper, we examine the situational factors associated with the onset of temptations, and those associated with the outcome of the temptation—whether it progresses to a smoking lapse or is resolved without smoking (a “resolved temptation”), in a sample of ITS attempting to maintain abstinence in a quit attempt. The data were collected in the context of a randomized trial of nicotine gum to support quitting in an ITS sample (Shiffman et al. 2019). Successful maintenance was uncommon, with 6-month continuous abstinence rates below 10%. Across multiple end points, the study found no effect of active gum vs. placebo. In the present analyses, to avoid confounding influences of pharmacological effects, we examine data from the placebo group to assess the situational factors associated with temptations and with their outcome, including analyses of coping as an influence on the outcome of temptations. The analyses draw from Ecological Momentary Assessment (EMA; Shiffman et al. 2008; Stone and Shiffman 1994) data in which participants reported their craving, mood, location, activity, and social setting (including exposure to others smoking) whenever they experienced a resolved temptation or a smoking lapse, with comparison to the circumstances reported when they were prompted at random times for assessment of their “background” states.

Among our key research questions was the role of negative affect in temptations and lapses. We hypothesized that negative affect would play little role in these episodes, as ITS were not expected to experience withdrawal. Given the strong role of situational stimuli in DS’ temptations and lapses (Shiffman et al. 1996b), and their particularly strong role in ITS’ ad libitum smoking (Shiffman et al. 2014a), we expected to observe a strong role for smoking cues, including alcohol consumption, in temptations and lapses. Such cues were expected to characterize temptation episodes, and differentiate them from randomly time-sampled background contexts, while also adding to the risk of lapsing in temptation situations. However, based on past research on DS lapse processes (Shiffman 1982; Shiffman et al. 1996b), we hypothesized that the performance of coping would be the biggest factor differentiating resolved temptations from those that progressed to lapses to smoking.

Methods

Subjects

Participants were adult ITS (≥ 18 years old) recruited from the Pittsburgh area via a variety of media, who reported smoking non-daily (4–27 days per month, regardless of the number of cigarettes smoked) for at least 1 year, and smoking at any rate for at least 3 years. Eligible participants expressed interest in making an attempt to quit smoking and willingness to try nicotine replacement gum. (See Shiffman et al. 2019 for inclusion criteria.) As shown in Supplemental Table 1, participants in this analysis averaged 44.0 years of age (SD = 14.9), were 43.9% male and 53.9% non-Caucasian, and smoked an average of 3.7 (SD = 2.7) cigarettes per day on smoking days (mean = 3.5; SD = 1.3 days per week). Notably, 68.5% had a Fagerstrom Test of Nicotine Dependence (Heatherton et al. 1991) score of 0, indicating no dependence.

Study design and procedures

In this clinical intervention trial, approved by the University of Pittsburgh Institutional Review Board, after a 2-week baseline period, participants were randomized 1:1 to receive either active 2 mg nicotine gum or an inactive placebo gum for 6 weeks. All participants received behavioral counseling focused on identifying, avoiding, and coping with smoking cues. There were two counseling visits before the designated quit day, and visits on the quit day and 1, 2, 4, and 6 weeks afterwards. Data were collected between June 2015 and September 2018.

Participants were provided with an Android smartphone (BLU© Dash 4.0 D270A) custom-programmed for EMA data collection (HBART, University of Tasmania Behavioural and Situational Research Group, https://www.utas.edu.au/health/research/groups/tasmanian-school-of-medicine/behavioural-and-situational-research-group-bsrg/hbart). Starting at the quit date, participants were to record all episodes of temptation (spikes of craving or coming close to smoking) and lapses (episodes of smoking), and to respond to randomly scheduled prompts (approximately 4 per day), for up to 6 weeks after the target quit day (regardless of actual quitting). During each report, participants were presented with a series of questions assessing situational factors (see Supplemental Table 2) at that moment (random assessments) or at the start of the episode (temptations, including lapses). These items pertained to participants’ craving (craving and need to smoke, averaged; 0–100 point scale, not subject-mean-centered), mood state (e.g., bored, contented, sad; complete list in Supplemental Table 2), location, activities, and social setting, smoking restrictions, and exposure to smoking and smoking cues. Mood items were factor-analyzed to yield 4 factors: negative affect, positive affect, arousal, and difficulty concentrating. These yielded factor scores, which were expressed as T-scores (mean 50, SD 10). In lapses and temptations, participants also identified what triggered the episode and what the main trigger was. They were also asked about their use of behavioral and/or cognitive coping to try to avoid smoking (none, behavioral only, cognitive only, or both; behavioral and cognitive coping were analyzed as variables). As part of their training for EMA, participants were oriented on what constituted behavioral coping (doing something) and cognitive coping (thinking or telling oneself something). Since the effect of the placebo gum would be primarily behavioral (like confectionery gum that quitters are sometimes advised to use when tempted), use of study gum in the hour preceding the report of a lapse or resolved temptation was counted as behavioral coping for the primary analyses; sensitivity analyses examined coping without counting use of gum.

Analysis

Data processing

Data for three participants (2% of the sample) were removed due to consistent non-compliance (weekly random prompt response < 50%). We removed one additional participant, an outlier who had entered > 400 temptation episodes, because those data prevented the analysis from converging. A total of 2.5% of observations were removed due to being in weeks of poor compliance with random prompts (1.8% of observations) or to abuse of the facility for suspending prompting (0.7% of observations). A total of 0.4% of additional days were lost due to hardware or software failures, battery exhaustion, or loss of the devices.

Because ITS smoke so intermittently even when smoking ad libitum, it can be difficult to distinguish the process associated with quitting from “normal” smoking. To ensure a focus on cessation-related processes, analyses were limited to subjects who achieved initial abstinence (regardless of whether they subsequently lapsed or relapsed), defined as 7 consecutive days without smoking (Shiffman et al. 2019), with data starting on the first day of that abstinence period. (Thus, all participants had to have more than 7 days of data after the target quit date.) This yielded a sample of 130 individuals, with EMA data for an average of 32.6 days, with an average of 80% response to the prompts for randomly scheduled assessments. The dataset consisted of 976 temptations episodes (829 resolved and 147 resulting in a smoking lapse) and 11,446 randomly scheduled background assessments.

Analyses used generalized linear mixed models (SAS PROC GLIMMIX) with random intercepts to test differences between the EMA occasion types in each of the EMA assessment items. Logistic regression was used for categorical variables, and linear regression for continuous variables (these are designated in Table 1 by showing means and SEs). We addressed our research questions in two comparisons. The first addressed the antecedents of temptations by contrasting situations associated with temptations (regardless of their outcome) with those reported at randomly selected times. The second addressed factors associated with the outcome of temptations, by contrasting those that resulted in smoking (“lapses”) to those that were resolved without smoking (“resolved temptations”). The dichotomous variables for behavioral and cognitive coping were combined to structure a series of contrasts shown in Table 2.

Table 1 Descriptive data and statistical comparisons for temptations (compared to background) and for lapses compared to resolved temptations
Full size table
Table 2 Relationships between coping responses and lapsing during temptation episodes
Full size table

Results

Factors associated with experiencing temptation

As expected, craving was much higher in temptations (Table 1), rising from a very low value (17 on a 0–100 scale) at randomly sampled times to a mean above the scale midpoint (57). Each 10-point increase in craving was associated with a 90% increase in the odds of temptation (compared to the random background level).

Mood was significantly worse in temptation episodes: Negative affect was significantly higher and positive affect significantly lower, but modestly so. (To supplement the analysis of factor scores, we also analyzed two more easily interpreted items: for each 10-point increase in anger [negative affect], the odds of temptation increased 12%, and for each 10-point increase in happiness [positive affect], the odds of temptation declined by 4%.) To assess whether these effects held among ITS with no signs of dependence, we re-tested the effects of negative and positive affect among participants with FTND = 0. The effects held and, indeed, were somewhat accentuated, with the odds of a temptation nearly doubled for every 1-SD increase in negative affect (OR = 1.97, 1.71–2.27) and reduced by nearly a quarter for every 1-SD increase in positive affect (OR = 0.78, 0.67–0.91).

Additionally, overall, subjects reported having slightly more difficulty concentrating in temptations (a 9% increase in the odds of temptation for a 10-point increase in rated difficulty concentrating) and very slightly higher arousal (a 10-point decrease in tiredness was associated with a 0.4% increase in the odds of temptation).

The likelihood of temptation varied substantially by location. Being in a bar was associated with a more than 10-fold increase in the odds of temptation (more on this below), and being in others’ homes or outside also was associated with increased risk (all compared to being home). Participants’ activity was also influential. Working, particularly at one’s job (but also at chores), was associated with lower risk of temptation, whereas social interaction, especially socializing, was associated with increased risk. Arguing was rarely recorded, but when it was, it was strongly associated with temptation. Engaging in media consumption was associated with lower risk.

Drinking alcohol had a particularly strong association with temptation, increasing the odds of a temptation more than 4-fold. However, among those who were drinking, neither the number of drinks nor self-reported drunkenness was associated with temptation. Drinking caffeinated beverages was also associated with elevated risk of temptation but non-caffeinated (non-alcoholic) beverages or eating food were not.

The social setting also affected the risk of temptations. Being with friends and/or acquaintances was associated with increased risk; being alone or with co-workers was associated with less risk. The smoking behavior of those present was very important: whether people who were part of the participant’s group (e.g., friends at dinner) were smoking or others were smoking in the participant’s view, the odds of temptation increased more than 4-fold.

Relatedly, exposure to smoking-specific cues was important, increasing the odds of temptation 5-fold. Seeing cigarettes, matches, and so on increased the odds of temptation, and smelling smoke increased the odds 5-fold. Perhaps summatively, the majority of temptation situations occurred in situations where participants indicated they used to smoke (Table 1), with an odds ratio of 7.

Regulations regarding smoking were also relevant to the risk of temptation, with temptations less likely when smoking was either discouraged or forbidden, particularly when forbidden by law. Availability of cigarettes was also strongly associated with temptation, increasing the odds of temptation approximately 3-fold.

Multivariable models

The above analyses identified many situational factors that were associated with temptations to smoke, in a univariate manner. The data suggest that the smoking environment—that is, factors such as the regulations regarding smoking and the presence of other smokers—is a particularly important influence, as this tends to correlate with other risk factors, such as being in a bar or at work. To control for these potential correlates or confounds, we performed focused multivariable analyses in which these variables were included as covariates.

Temptations were associated with drinking alcohol, being in a bar, and being where smoking is not forbidden and where others are smoking. Supplemental Table 3 shows the results of a multivariable model including all of these variables. It shows that when these variables are taken jointly into account, drinking alcohol was no longer significantly associated with temptation, nor were smoking regulations. Being in a bar, where smoking was allowed, and where others were smoking, continued to be associated with temptation, even after adjusting for the other variables.

Similarly, a multivariable model assessed whether reporting exposure to smoking cues increased the risk of temptation, independent of seeing smokers, being where smoking was allowed, and having cigarettes available. The effect of reported exposure to cues was attenuated, but not eliminated, by controlling for these other factors.

Factors associated with lapsing in temptations

The analyses also evaluated the relationship between situational factors and the likelihood of lapsing once participants had experienced temptation (that is, they contrasted lapse episodes vs resolved temptation episodes). Contrary to expectations, ITS did not report more intense craving in situations that led to smoking (lapses) than in situations resolved without smoking (resolved temptations). The lack of association with this core dependence-related experience led us to assess whether the likelihood of lapsing was higher for ITS with at least some indication of dependence (i.e., FTND > 0). It was not (OR = 1.18; 0.50–2.79, p = 0.71).

ITS reported higher NA and lower PA in lapses; the differences were in the same range as those that differentiated temptations from background. These associations were not significant in the subset of ITS with FTND = 0, though the effects in this subset were in the same direction (NA: OR = 1.12, p = 0.47; PA: OR = 0.78, p = 0.13). Higher arousal was reported in lapses, but slightly less difficulty concentrating.

The likelihood of lapsing varied substantially by location, in ways that paralleled the differences between temptations and the background. Being in a bar, in others’ homes, or outside also increased risk (compared to being home). Participants’ activity was also influential, again in ways that paralleled correlates of temptation. Working at one’s job was associated with lower risk of lapsing, whereas social interaction, especially socializing and arguing, was associated with increased risk. Engaging in media consumption was associated with lower risk. Drinking alcohol was associated with a 7-fold increase in the odds of smoking, but the number of drinks and self-rated drunkenness was not associated with smoking. Neither drinking non-alcoholic beverages nor eating was associated with smoking.

Just as with temptations, being with friends and/or acquaintances was associated with increased risk of smoking, whereas being alone or with co-workers was associated with less risk. Whether those present were smoking was very important: when people were smoking in view, the odds of smoking increased almost 3-fold, and increased 8-fold if those smoking were part of the participant’s group and not just strangers in view.

Exposure to smoking cues was also important, increasing the risk of smoking more than 7-fold, with especially strong associations with seeing matches or lighters. Smelling smoke increased the odds of lapsing more than 5-fold. Regulations regarding smoking are also relevant to the risk of smoking, which was considerably reduced when smoking was either discouraged or forbidden by the smoker’s own rules, but not by others’ rules or by law. Not surprisingly, since cigarettes had to be available for smoking to occur, availability of cigarettes (as perceived by the participant) was also strongly associated with smoking, increasing the odds of smoking 7-fold (when cigarettes were seen as available, but with difficulty) to 46-fold (when easily available). Perhaps summatively, lapse situations were more often characterized by participants as ones where they used to smoke, doubling the odds of a lapse.

Triggers

Participants were more likely to identify multiple triggers for lapses than for resolved temptations (OR = 2.46, 1.41–4.28). However, individual particular triggers did not significantly distinguish lapses from resolved temptations, with the notable exception that good mood and smoking cues were more often cited as a trigger of lapses (Table 1). When participants had to designate a single primary trigger, only good mood predicted smoking.

To explore the association between good mood as a trigger and lapsing, a multivariable analysis examined designating this as the trigger for lapsing was associated with other, potentially confounding factors. A prior analysis of temptations (Shiffman 1986) suggested that positive affect marked situations associated with social situations such as parties. When we re-assessed the association between good mood as a trigger and lapsing while controlling for situational factors of drinking alcohol, being with friends, and being around others smoking, good mood as the trigger was no longer associated with lapsing (Supplemental Table 3).

Multivariable analyses

As with predictors of temptations, multivariate models considered the unique contributions of correlated situational factors such as the smoking environment (Supplemental Table 3). Unlike the findings for temptations, modeling indicated that drinking alcohol was independently associated with lapsing, even after accounting for smoking regulations, others smoking, and being in bars (which was no longer significant). When a collection of variables related to cues, smoking regulations, and others smoking were considered jointly in a multivariate model, regulations, seeing others smoking, and reporting exposure to smoking cues were no longer related to lapse risk, but having others in one’s group smoking and having cigarettes available (especially easily) were strongly related to lapsing.

Coping

Compared to doing no coping, both behavioral and cognitive coping were associated with averting a lapse (Table 2). However, the effect was much bigger for behavioral coping, which was associated with higher odds of averting a lapse 130-fold, compared to 9-fold for cognitive coping. Individuals who reported only behavioral coping had 18 times better odds or averting a lapse than those who reported only cognitive coping. Behavioral coping also had incremental effects even when individuals had also performed cognitive coping: in cases where cognitive coping was performed, behavioral coping increased the odds of averting a lapse 7-fold. In contrast, when behavioral coping was performed, those who also performed cognitive coping did not see any advantage, with the data actually leaning towards worsening outcomes (OR = 1.70, ns). The lowest probability of lapsing was seen when participants reported only behavioral coping, and behavioral coping was associated with a steep increase in the odds of surviving a temptation without smoking. That said, some coping was reported in 60% of lapses, so coping more often than not failed to prevent lapses.

As a sensitivity test, we conducted tests of coping effects that did not count the reported use of study gum as a behavioral coping response. The pattern of results (Supplemental Table 4) was similar: both behavioral and cognitive coping, considered individually, were significantly and very strongly associated with warding off lapsing, compared to no coping, with much larger effects for behavioral coping (OR = 0.04) than cognitive coping (OR = 0.21). However, with the removal of gum use from behavioral coping, the contrasts between behavioral and cognitive coping, and between combinations of both types compared to either alone, were no longer significant, though it was notable that the combination of cognitive and behavioral coping again tended to be associated with increased risk of lapsing (OR = 2.42, ns), compared to behavioral coping alone.

To further test the effects of coping in preventing lapses, we re-examined the coping analyses while excluding episodes where cigarettes were unavailable, reasoning that this presented a better test of coping, because lapses were unlikely in any case when cigarettes were unavailable. Notably, the main change when this constraint was applied was to increase the risk of lapsing when no coping was reported. Nevertheless, the analyses produced similar results, but with much stronger effects of coping (Table 2). The odds of avoiding a lapse were dramatically decreased when behavioral coping was enacted (a 1000-fold decrease in lapse risk for behavioral coping compared to no coping). Cognitive coping also very substantially decreased the risk of lapsing (14-fold), but to a lesser degree.

Mediation of alcohol and mood effects through coping

The effects of alcohol consumption and affective distress on lapse risk could be due to the effect of these variables on suppressing coping. To assess this hypothesis, analyses explored the relationship between these situational variables and coping, and, in turn, their association with lapse risk once coping had been accounted for. These analyses focused on the situations where cigarettes were available, as this was where the effects of coping were greatest. Bivariate analyses showed that coping was not related to whether participants had been drinking, but the likelihood of any coping was indeed significantly diminished when negative affect was elevated, and increased when positive affect was elevated (Supplemental Table 5). The effects were only evident for the aggregate variable of any coping, and not for either behavioral or cognitive coping, when considered individually (though effects on cognitive coping approached significance). Consistent with a mediational model, negative affect and positive affect were no longer associated with lapse risk once coping had been accounted for in a multivariable model.

Discussion

These data provide, for the first time, an account of the circumstances that lead ITS who are trying to quit to experience temptations to smoke, and the circumstances that are associated with progressing to actual smoking lapses when temptation is experienced. In many respects, the circumstances surrounding temptations and lapses among ITS resemble those that have been documented for daily and heavy smokers (Marlatt and Gordon 1985; O'Connell and Martin 1987; Shiffman 1982; Shiffman et al. 1996b). This suggests a similarity in process, which may be surprising, given that ITS show little to no dependence (Shiffman et al. 2012b), and that relapse is often attributed to dependence and even regarded as a hallmark of dependence (Sussman and Sussman 2011; Zou et al. 2017).

Crucially, ITS were more likely to report temptation when they were experiencing more affective distress—higher negative affect and lower positive affect. This was true even among the ITS with no signs of dependence at all (FTND = 0). Further, the ITS sample as a whole was more likely to actually lapse when the temptations occurred under greater affective distress. These differences in affect were relatively small, but were in the range of effects observed for DS (Baer et al. 1989; O'Connell and Martin 1987; Shiffman et al. 1996b), and were reinforced by the fact that temptations and lapses were more likely when participants were arguing. If these were dependent smokers, one might attribute these experiences to nicotine withdrawal. However, as ITS do not seem to show withdrawal, and the effect was evident for those with an FTND score of zero, other explanations are needed. It is plausible that ITS may have experienced some affective relief as a result of smoking (Baker et al. 2004; Kassel et al. 2003), resulting in emotional distress provoking craving and temptation, without any involvement of withdrawal.

It could also be that, in individuals trying to exert self-control effort to avoid smoking, affective distress disrupts such efforts (Marlatt and Gordon 1985; Muraven and Baumeister 2000; Shiffman and Waters 2004), leading to temptation and to actual smoking. Indeed, mediational analyses showed that effects on performance of coping completely accounted for the effects of affective distress (higher negative affect, lower positive affect) on the risk of lapsing in the face of temptation. This is consistent with laboratory findings showing that experimentally induced negative affect can inhibit coping, and that this was related to subsequent success in a quit attempt (Drobes et al. 1994). These findings from ITS suggest that the affective distress observed among DS in connection with temptations may similarly not be products of withdrawal, but rather part of normal variation in affect, arousal, and concentration. This may also help explain why nicotine replacement therapy is not completely protective against temptation or lapsing (Hartmann-Boyce et al. 2018), even when it effectively reduces or even eliminates withdrawal symptoms (Ferguson and Shiffman 2014). Research in DS is needed to better elucidate the role of negative affect in temptations and lapses. Other characteristics of temptations and of lapse episodes suggested the operation of conditioning, as temptations and lapses tended to happen in the situations where, according to previous research (Shiffman et al. 2014a), and the participants of the study themselves, ITS typically smoked. In this respect, ITS lapses very much resembled those of daily and even heavy and dependent smokers, and suggest the importance of conditioned stimuli, and stimulus control of smoking in relapse. We had previously speculated (Shiffman et al. 2013a; Shiffman et al. 2013b) that such processes play important parts in both daily and non-daily smokers’ smoking and relapse, and suggested that the operation of such processes produces a kind of behavioral dependence that makes smoking persistent and relapse-prone even without tolerance and withdrawal, which are often designated as the core of dependence. These data are consistent with that notion.

As expected, reported craving was much more intense in temptation episodes; this is almost their defining feature. The data confirm findings from ad libitum smoking (Shiffman et al. 2014a) indicating that ITS do experience craving, particularly when provoked by cues. Surprisingly, craving was no higher in situations that led to smoking than in successfully resolved temptations. Also, ITS with some signs of dependence (i.e., FTND > 0) were no more likely to lapse during temptations than were ITS with no sign of dependence (FTND = 0). This is consistent with the idea that once a temptation occurs, other factors, particularly coping, determine its outcome. Of course, dependence-related process themselves may affect coping, for example by making it hard to access cognitions about coping or about the ill effects of smoking (Sayette and Hufford 1997), or invoking smoking as an automatic response (Tiffany 1990).

As previously reported (Shiffman 1982; Shiffman et al. 1996b) for DS, the performance of coping was a key factor in averting smoking when facing temptation, very steeply reduced the odds of lapsing (especially for behavioral coping, compared to no coping). Analyses of temptations and lapses among DS (Shiffman 1982; Shiffman et al. 1996b) have also shown substantial effects of coping, but the effect sizes were much smaller. It may be that coping is particularly effective for ITS, because their drive to smoke is not as great as DS’. The magnitude of coping effects was reduced when use of the placebo gum was not counted as behavioral coping, but the patterns were very similar and the effects remained quite large.

In any case, the data reinforce the importance of teaching and emphasizing coping with temptations in treatment. The present findings suggest emphasizing behavioral strategies over cognitive ones, but previous analyses have been mixed on this matter (Shiffman 1982; Shiffman et al. 1996b). The observation that cognitive coping was less effective, at least when gum use was counted as coping, could be due to the potential inclusion of responses such as relying on willpower, or self-punitive self-talk, which have been found to be ineffective, or even counter-productive (Shiffman 1984). The present findings moderate the past findings in suggesting that combining both cognitive and behavioral strategies may improve outcomes. However, some caution is warranted, as coping was self-selected and the self-report coping data may miss some things smokers did to avoid smoking, and, crucially, does not take into account the number, character, quality, vigor, or timeliness of the coping performed. It was notable that most lapses occurred despite reported coping, suggesting that improvements in coping performance are needed. In any case, these data suggest that smokers quitting should be encouraged to implement behavioral coping. While it is likely also useful to implement cognitive coping, smokers should avoid relying solely on cognitive strategies.

The data highlight the role of cigarette availability in temptations and lapses. Temptations were more likely to occur when cigarettes were perceived as available, whether easily or with difficulty, and this held up even when markers of actual availability, such as the presence of others smoking, were controlled for. This could be related to the finding that perceived availability of opportunity to smoke heightens the effect of smoking cue exposure on craving (Sayette et al. 2001; Tiffany and Hakenewerth 1991). Once participants were tempted, what most predicted whether they smoked or not was the easy availability of cigarettes. This likely reflects the practicality that lapsing requires access to cigarettes. But it also suggests that lapses may be impulsive, such that individuals may not engage in extended efforts to overcome barriers to access, but may move quickly or even automatically to smoking if cigarettes are easily available (Tiffany and Hakenewerth 1991). Notably, the availability of cigarettes made lapsing more likely when the smoker failed to make a coping response.

Our analyses shed some light in the important role of alcohol consumption in temptations and lapses. Temptations were more likely when alcohol had been consumed, but it did not matter how much was consumed or how drunk the person felt, suggesting that this may not be a pharmacological effect, but perhaps a product of conditioned association of smoking with drinking (Shiffman et al. 2014a). Moreover, the association with alcohol was no longer significant when other contextual factors, such as the presence of others smoking, were controlled. However, alcohol consumption was strongly associated with lapsing within temptations, even after those factors were controlled, and this effect was not mediated by alcohol effects on coping.

The findings also reinforce the influence of social interactions in ITS’ smoking. Although most cannot be characterized as “social smokers” (Schane et al. 2009; Shiffman et al. 2014a), ITS tend to smoke when socializing with friends and acquaintances (Shiffman et al. 2014a), and these social dimensions of situations were also associated with a greater risk of temptation and lapsing. Although such detailed data are not available for DS, the published data suggest that social settings may play a similar role in DS lapses, where others were present 60% of the time (Ferguson and Shiffman 2014). The role of social interaction and social influence on relapse deserves further study.

In general, lapse situations seemed to be more extreme versions of those associated with resolved temptations: they were characterized by more exposure to smoking cues, more cigarette availability, and more adverse affective states. This suggests that temptations exist on a continuum of severity, with more severe instances being more likely to lead to smoking rather than being resolved without smoking. It was surprising, however, that this was not reflected in more intense craving in lapse situations than in resolved temptations. Processes that are not reflected in consciously perceived craving may drive higher smoking probability in these situations.

The study was subject to some limitations. The sample consisted of self-selected treatment-seekers, who received behavioral treatment, so the sample may not be representative. There was substantial dropout from the study and loss of data due to non-compliance, which could introduce bias. All of the data were self-reported, which could allow for bias, including implicit self-justifications for lapsing or being tempted to smoke (“the cues made me do it”). The study also had significant strengths, notably the use of detailed real-time EMA data to capture details of temptations, and the use of randomly sampled occasions to represent the participants’ experiences outside of temptations. This is also the first study to report such relapse process data for non-daily smokers, who now comprise a quarter to a third of adult smokers in the USA (Reyes-Guzman et al. 2017).

In sum, this study showed that situational factors associated with temptations among non-daily smokers trying to maintain abstinence included exposure to smoking cues, alcohol consumption, and negative affect. The role of negative affect, which is also seen in temptations and lapses among daily and dependent smokers, is particularly important, as it suggests that the negative affect that characterizes temptations and lapses is not withdrawal-related, but due to natural variations of affect in daily life. As in daily smokers, coping played an important, but imperfect, role in preventing smoking in the face of temptation to smoke, suggesting its importance in helping non-daily smokers maintain abstinence.