Health Psychology 2014, Vol. 33, No. 10, 1232–1240

In the public domain http://dx.doi.org/10.1037/a0035708

Attentional Retraining Administered in the Field Reduces Smokers’ Attentional Bias and Craving William F. Kerst and Andrew J. Waters Uniformed Services University of the Health Sciences Objective: Attentional retraining (AR) is a potential new treatment for addiction. AR trains addicts to attend away from drug-related cues, thereby reducing exposure to drug cues and reducing craving. We examined the utility of delivering AR to smokers on a personal digital assistant (PDA) in the natural environment. Method: Smokers (N ⫽ 60) not seeking to quit were randomly assigned to an AR group or a control group (i.e., a group with no training). They carried a PDA with them for one week. They were prompted to complete four assessments daily, including three attentional retrainings (AR group) or three control trainings (control group), and one evaluation of attentional bias. AR was implemented using a modified visual probe task. Attentional bias was assessed using a standard visual probe task on the PDA. Results: The AR group completed an average of 15.0 attentional retrainings and the control group completed an average of 14.9 control trainings. As hypothesized, attentional bias declined over the week in the AR group, but not in the control group, Group ⫻ Day interaction, F(1, 232) ⫽ 4.77, p ⫽ .03. AR also reduced craving ratings following a briefly presented picture containing smoking and nonsmoking features, group main effect, F(1, 234) ⫽ 3.89, p ⫽ .04. AR did not significantly influence smoking behavior. Conclusion: AR can be administered on a mobile device in the natural environment, and AR can reduce attentional bias and craving. Keywords: attentional retraining, attentional bias, craving, tobacco Supplemental materials: http://dx.doi.org/10.1037/a0035708.supp

centive salience” of the drug cues. Robinson and Berridge (1993) argued that the latter response is critical and that it mediates attention to drug cues, craving, and approach behavior. Franken’s (2003) model expanded on the work of Robinson and Berridge (1993) and proposed that attentional bias can cause craving. An individual with high attentional bias to drug cues will be more likely to attend to drug cues, which provoke craving, than an individual with lower attentional bias. Attentional bias therefore elevates cue-provoked craving, rather than abstinence-induced craving, which is more a function of state of deprivation (Shiffman, 2009). Franken’s model also proposed that craving can cause attentional bias, and that craving can cause drug use or relapse. Although the relevance of laboratory-assessed cue-provoked craving to smoking/relapse has been questioned (Perkins, 2009), an association between smoking cues and smoking/relapse has been found in naturalistic settings (Shiffman, 2009). In addition, smoking cues may provoke smoking and relapse via pathways that do not involve cue-provoked craving. For example, smoking cues may elicit cognitive processes that deplete attentional resources, undermining attempts to reduce smoking or to remain abstinent (Franken, 2003), or the cues may elicit stimulus–response “habits” (Tiffany, 1990). In sum, interventions that reduce attentional bias should reduce exposure to smoking cues and therefore reduce both craving and smoking. Empirical data have also suggested that attentional bias may be an appropriate cognitive target for intervention. A meta-analysis revealed that attentional bias is associated with craving (Field, Duka et al., 2009). However, this association is of modest magnitude, particularly for smoking-related attentional bias, and the

Recent research has examined cognitive processes underlying drug addiction with the hope of identifying cognitive targets for intervention (Wiers & Stacy, 2006). Beginning with Tiffany (1990), many addiction researchers have highlighted the role of automatic (or implicit) processes in drug addiction (see Wiers & Stacy, 2006). Automatic processes are fast, parallel, effortless, and may not engage conscious awareness (Wiers & Stacy, 2006). Measures of automatic/implicit cognition are associated with substance use (Rooke, Hine, & Thorsteinsson, 2008). One widely studied automatic process is automatic attention capture by drug cues, or “attentional bias” to drug cues (Field & Cox, 2008). Theory and data suggest that attentional bias may be an appropriate target for intervention. Attentional bias develops through classical conditioning (Field & Cox, 2008). Historically, drug cues were thought to elicit either drug-like or drug-opposite responses. In the incentive-sensitization theory (Robinson & Berridge, 1993), it was proposed that the conditioned response also involves a conditioned “wanting” response that reflects the “in-

This article was published Online First May 12, 2014. William F. Kerst and Andrew J. Waters, Department of Medical and Clinical Psychology, Uniformed Services University of the Health Sciences. William F. Kerst is now based at Joint Base Elmendorf-Richardson, Anchorage, Alaska. Correspondence concerning this article should be addressed to Andrew J. Waters, Department of Medical and Clinical Psychology, Uniformed Services University of the Health Sciences, 4301, Jones Bridge Road, Bethesda MD, 20184. E-mail: [email protected] 1232

ATTENTIONAL RETRAINING ADMINISTERED IN THE FIELD

nature of the causal relationship is unclear. Other studies have reported that attentional bias prospectively predicts outcomes in addiction, including tobacco addiction (e.g., Janes et al., 2010; Powell, Dawkins, West, & Pickering, 2010; Waters et al., 2003). Interventions that directly target automatic cognitive processes such as attentional bias have been examined in recent research. Cognitive tasks such as the visual probe (VP) task (e.g., Field & Cox, 2008) are modified so that the tasks change (rather than measure) automatic cognitive processes. Attentional retraining (AR) refers to the use of modified tasks to change attentional bias, and is most often conducted using a modified VP task. AR has shown promise in the treatment of anxiety-related conditions (see Hakamata et al., 2010; Hallion & Ruscio, 2011). Hakamata and colleagues (2010) reported in a meta-analysis that AR administered using a modified dot-probe task produced significantly greater reductions in anxiety than control training. Theory and data such as that described above have motivated researchers to apply AR to addiction. Field and Eastwood (2005) trained one group of heavy social drinkers to attend away from alcohol stimuli and toward neutral stimuli (“attend-neutral”) and a second group to attend toward alcohol stimuli (“attend-alcohol”). The attend-neutral group exhibited lower (more negative) attentional bias, reported lower craving, and consumed less alcohol on a taste test (see also Fadardi & Cox, 2009). In a randomized trial of AR in alcohol-dependent patients, Schoenmakers et al. (2010) reported that AR, delivered in five sessions over 21 days, reduced attentional bias, and that AR also reduced the time to discharge from inpatient treatment. In smokers, one study documented that an attend-neutral training group exhibited significantly lower (more negative) posttraining attentional bias than an attendsmoking training group. They also reported less craving in response to a smoking cue, although the effect was found only in men (Attwood, O’Sullivan, Leonards, Mackintosh, & Munafo, 2008). However, other laboratory studies reported that AR did not affect craving in smokers (Field, Munafo et al., 2009; McHugh et al., 2010), and one study reported that AR did not reduce attentional bias (McHugh et al., 2010). AR in the aforementioned studies of smokers was limited to a single session. The effect of multisession AR in smokers is not known. AR in addiction studies has hitherto been delivered in a laboratory setting. Ecological momentary assessment (EMA) can be used to extend the study of AR. EMA involves assessing phenomena at the moment they occur in a person’s natural environment. Cognitive tasks can be administered successfully on a personal digital assistant (PDA; e.g., Waters & Li, 2008). As such, we used EMA methods for both cognitive intervention (AR) and assessment. EMA allows for easy administration of multiple trainings per day. More “doses” of AR may lead to greater reductions, or more sustained reductions, in attentional bias (Hakamata et al., 2010). One theorized cognitive mechanism underlying AR is the development of an implicit production rule, which specifies that if there are both salient and neutral stimuli present in the environment, then attention should shift to the neutral stimulus (Mathews & MacLeod, 2002). In AR for anxiety reduction, participants are trained to automatically attend away from stressful features of provocative stimuli to more benign features. An effect of AR on anxiety is more likely to be observed when participants are assessed with a task that includes both stressful and benign features,

1233

than when assessed at unprovoked occasions (Hallion & Ruscio, 2011; Mathews & MacLeod, 2002). In addiction, AR may also be mediated by an implicit production rule (an automatic process). However, Schoenmakers and colleagues suggested that AR may also be mediated through more controlled processes that involve awareness of training contingencies (Schoenmakers et al., 2010). Either way, an effect of AR on craving may be most likely observed after presenting participants with a stimulus containing both smoking and neutral (nonsmoking) features. AR-trained smokers should automatically (or, perhaps, deliberately) attend to the neutral features. Compared with untrained smokers, they should experience less exposure to the smoking content and therefore experience less craving. In sum, AR may be useful in the treatment of addiction. In previous studies of AR in tobacco addiction, AR was administered in a single laboratory session. Here we examined the utility of administering multiple sessions of AR over a 1-week period in a naturalistic setting. Our two primary hypotheses were that: (a) AR would reduce attentional bias, and (b) AR would reduce craving in response to stimuli containing both smoking and nonsmoking features. In exploratory analyses, we examined the effect of AR on smoking behavior. Finally, we explored the effect of awareness of AR training contingencies on outcomes.

Method Participants Participants were 60 adult community-based smokers from the metropolitan area of Washington DC (see Figure 1). To qualify, participants had to be current smokers, smoke 10 or more cigarettes per day for the past 2 years, and be 18 – 65 years of age. The exclusion criterion was an expired breath carbon monoxide (CO) level lower than 10 ppm. Participants were recruited by advertising targeted at 18 – 65-year-old smokers in local newspapers, craigslist.com, and through the use of flyers. The study was approved by the institutional review board of the pertinent university.

Laboratory Procedures Study procedures and measures are summarized in Figure 2. After a telephone screening, eligible participants were invited to attend a baseline visit. Research assistants described the study, answered questions, confirmed eligibility (by obtaining CO levels), and obtained written informed consent. The informed-consent form stated that there were two treatment conditions, but it did not specify how the two conditions differed. Participants were then randomly assigned to the AR or control condition, stratified by gender. Participants and researchers were both blinded to condition assignment. Participants then completed the VP assessment task (attentional bias assessment) on the PDA, provided a saliva sample for analysis of smoke exposure, and completed a number of questionnaires, described later. Participants were given a smoking diary (see the Measures section). Participants were told that they could smoke as much or as little as they liked during the week. They were trained on the use of the PDA, and they returned the PDA (and smoking diary) 1 week later at Visit 2. At Visit 2, participants again completed the VP task, provided breath samples for CO, and provided a second saliva sample. Awareness of

KERST AND WATERS

1234

Figure 1. Study flowchart. Participants were randomly assigned to treatment conditions, stratified by gender. Due to researcher error, 16 men and 14 women completed the control condition. Key: 1 in the field, AR participants completed 434 attentional retrainings and 146 VP assessments; 2 in the field, control participants completed 448 control trainings and 145 VP assessments. For participants who completed at least one field assessment, AR participants (n ⫽ 28) completed a mean of 5.1 (SD ⫽ 1.9) assessments, and control participants (n ⫽ 29) completed a mean of 5.3 (SD ⫽ 1.7) assessments, F(1, 58) ⫽ 0.00, p ⫽ .97.

training contingencies was assessed (see Measures), and participants were debriefed about the purpose of the study.

RAs by 5 min (up to three times per RA). RAs scheduled during a suspend were not replaced, nor were RAs lost due to change in bed time, PDA error, or participant error (e.g., failure to

Field Procedures The PDA was programmed to audibly prompt the participants at random four times per day (random assessments; RAs). As described below, RAs were used to deliver the interventions (trainings, three per day) and for assessment (one per day)1. Participants could prevent the PDA from presenting RAs for up to 2 hr (using the “suspend” function), and could also delay

1 Using the selected wake-up time and bed time on a given day, the program divided the day into four equal “periods.” An RA was scheduled at a random time during each period. In the AR and control groups, respectively, trainings occurred as follows: Earlier than 11:00 a.m., 17.5% and 13.1%; 11:00 a.m.–3:00 p.m., 30.3% and 29.5%; 3:00 p.m.–7:00 p.m., 34.7% and 38.5%; later than 7:00 p.m., 17.5% and 18.9%.

ATTENTIONAL RETRAINING ADMINISTERED IN THE FIELD

1235

Figure 2. Study procedures and measures. AR ⫽ attentional retraining; CON ⫽ control training; FTND ⫽ Fagerström Test for Nicotine Dependence; smoked so far that day ⫽ whether or not participant had smoked so far that day; QSU ⫽ Questionnaire for Smoking Urges; cig ⫽ cigarette; CO ⫽ carbon monoxide; PDA ⫽ personal digital assistant; computer ⫽ assessments administered by desktop computer. Base ⫽ baseline. Several items (not shown in Figure 2 and not reported) were interposed between the two craving items; these items included assessment of negative affect. Cued craving pictures contained both smoking and nonsmoking content, for example, one picture consisted of a group of individuals in which two individuals were holding a cigarette and several individuals were not. Compensation ⫽ participants were paid $20 for each of two laboratory visits; $5 for each day they provided data; $2 for each completed training and field assessment. Key: 1 attentional bias assessment; 2 craving assessments; 3 smoking assessments; 4 used as covariates; 5 used for secondary subgroup analyses.

charge the PDA). A pilot study with nine participants revealed that this protocol was feasible; the pilot participants responded to 76.9% of presented RAs.

Measures Attentional bias. The VP task was used to assess attentional bias (in lab and field), which is widely used in addiction research (Field & Cox, 2008). Participants were instructed that they would perform a task assessing RTs. They were informed that a dot would be presented on the left or right side of the PDA screen. They were required to indicate the dot’s position as quickly as possible by pressing a “Left” or “Right” button on the PDA screen using their thumbs. The standard VP task (assessment of attentional bias) consisted of 80 trials, presented in a new random order for each assessment. At the start of each trial, a fixation cross was displayed in the center of the screen for 500 ms. A picture pair, containing one smoking picture and one neutral picture, was then presented for 500 ms, one picture on each side of the central position. This presentation duration (500 ms) was chosen because AR in addiction has been successful in reducing attentional bias with this duration (but not with 200 ms) (Schoenmakers et al., 2010), and this duration of AR has most commonly been used for anxiety (Hakamata et al., 2010). The dot probe was displayed immediately after the offset of the pictures. It remained on the screen until the participant made a response. After the response, the fixation cross for the subsequent trial was presented after 500 ms. On the standard VP task, for each picture pair, on half of the trials the probe replaced the smoking picture, and on the other half, the probe replaced the neutral picture. On half of the trials, the smoking picture appeared on the right, and on the other half the

smoking picture appeared on the left. After the final (80th) trial, an item assessed how many times the participant reported being interrupted during the task (response options: no times, 1 time, 2 times, 3 times, 4⫹ times). This item was administered so that data could be analyzed while excluding assessments with a large number of interruptions. Craving. The 10-item Questionnaire for Smoking Urges (QSU) is a self-report measure of craving (Cox, Tiffany, & Christen, 2001). We reported the total score (Cronbach’s alpha ⫽ .95 at baseline and .92 at Visit 2) for the current study. Each item assessed craving in response to a “compound” smoking picture (cued craving item). A picture with both smoking and nonsmoking (neutral) features was presented on the PDA in landscape mode for 1 s. Participants subsequently reported their craving for cigarettes on the PDA on a 1–7 scale (1 ⫽ strongly disagree, 7 ⫽ strongly agree). Studies have shown that assessment of cue-provoked craving using photographic images on PDAs during EMA is feasible and produces valid data (e.g., Wray et al., 2011). A second (noncued) craving item was also administered. Smoking. The smoking diary was a paper form containing boxes for each study day; participants were instructed to make an entry into the appropriate box each day before they went to bed indicating how many cigarettes they had smoked on that day. Two biological measures assessed heaviness of smoking: cotinine levels in saliva (cotinine is the major metabolite of nicotine), and CO levels in exhaled breath (using a CO monitor, Vitalograph, Lexena, KS; SRNT, 2002). On the PDA, one item assessed “smoking so far that day” (Response Options: Had not smoked that day; Had smoked that day). A second item assessed time since last cigarette (response options: 1 ⫽ no cigarettes smoked for ⬎ 2 hr; 2 ⫽ last

1236

KERST AND WATERS

cigarettes smoked between 30 min and 2 hr; 3 ⫽ last cigarettes smoked between 30 min and 5 min; 4 ⫽ just smoked/smoking now). Contingency awareness. The interview question, “Did you notice any pattern in how the dots replaced certain pictures?” was used to assess contingency awareness. Responses were transcribed and coded as aware or unaware. An AR participant was coded as aware if he or she was able to describe the relationship between the probes and the picture types (i.e., dots were more likely to follow neutral pictures than smoking pictures). Demographics and smoking measures. Author-constructed demographic and smoking history questionnaires were administered. Participants were asked to indicate how many cigarettes they smoked on average per day, the number of lifetime quit attempts lasting at least 24 hours they had made, and their intent to quit. The Fagerström Test for Nicotine Dependence (FTND), a widely used self-report measure of nicotine dependence that yields scores from 0 –10 (higher scores reflecting greater dependence), was also administered (Heatherton et al., 1991).

Control Training Condition

Stimulus Materials

Data Reduction and Analysis

Pictures for trainings and VP assessments were taken from the International Smoking Image Series (ISIS; Gilbert et al., 2007), previous studies (e.g., Swanson, Swanson, & Greenwald, 2001), and author-constructed pictures. An initial pool of more than 300 pictures was categorized on smoking content (smoking vs. nonsmoking) and the presence/absence of human features (human features vs. objects) and rated by three research assistants for appropriateness and clarity (global judgment). The highest rated 20 pictures from each category were selected for the final set of 80 pictures. Smoking “human” pictures (e.g., a hand with a cigarette) were randomly paired with nonsmoking “human” pictures (e.g., a hand with a key). The same was true for smoking “object” pictures (e.g., a cigarette) and nonsmoking “object” (e.g., a key) pictures. From these picture pairs, eight different picture sets were created. Order of presentation of the eight picture sets was randomized across participants. Participants were scheduled to be exposed to each picture set for no more than one “cycle” during the study. (A cycle was defined as three trainings followed by one assessment). On each cycle, participants received three AR or control trainings using an assigned picture set (e.g., Set A), and attentional bias was also assessed using Set A. On the subsequent cycle, they were trained using a different picture set (e.g., Set B) and assessed using Set B. Therefore, attentional bias was always assessed using materials on which participants had received training. At the second laboratory visit, participants were assessed on a set of (randomly selected) pictures from the eight sets. For the cued craving item, stimuli were eight pictures in landscape orientation that contained both smoking and nonsmoking content. At each assessment, a picture was selected randomly.

On the standard VP task (field, 292 assessments), reaction times (RTs) ⬍ 100 ms were discarded (0.09% of trials). Median RTs (correct responses) were used to reduce the influence of outliers. An attentional bias score for each assessment was computed as median RT for trials on which the probe replaced the neutral picture minus the median RT for trials on which the probe replaced the smoking picture (i.e., difference score). Faster RTs on probes that replaced smoking pictures (i.e., positive attentional bias score) reflected attentional bias toward the smoking picture. Faster RTs on the probes that replace neutral pictures (negative attentional bias score) reflected attentional bias away from the smoking pictures toward the neutral pictures, or “avoidance.” One extreme score (10.5 SD from mean) was removed, leaving 291 scores for analyses. The same methods were used for the laboratory assessments (i.e., the standard VP task). ANOVA for continuous variables or Pearson’s chi-square test for categorical variables were used to examine between-groups differences at baseline. ANCOVA, with one between-groups variable (group) was used to analyze lab data. ANCOVA is preferred to the analysis of pre–post change scores when participants are randomly assigned to groups and it has greater power (Fitzmaurice, 2001). The dependent variables were attentional bias, craving (cued and noncued craving), and smoking behavior (CO in breath, cotinine in saliva), assessed at Visit 2. Linear mixed-models (LMM) analyses (PROC MIXED command in SAS, Cary, NC) were used to analyze continuous outcome variables assessed in the field. LMMs allow for the fact that subjects differ in the number of observations available for analysis, and take into account the clustering of data within subjects. We used a random (subjectspecific) intercept and an autoregressive model of Order 1 for the residuals within subjects (all dependent variables). Day in study was entered as a continuous variable, along with Group and the Group ⫻ Day interaction term. If the interaction term was not significant, it was dropped from the model. To analyze “smoked so far that day,” we used multilevel logistic regression (PROC GLIMMIX in SAS, using maximum likelihood with adaptive quadrature estimation). For analysis of lab and field data, baseline (preintervention) measures of the dependent variables were included as covariates in the respective analyses, and each dependent

Attentional Retraining Condition AR participants were scheduled to complete three modified VP tasks (attentional retrainings, 160 trials each) followed by one standard VP task (attentional bias assessment, 80 trials) per cycle on the PDA. On ARs, the dot always replaced the neutral picture. There was a perfect correlation between picture type and dot location. In all other respects, the procedures were the same as those for the standard VP task.

Control participants were scheduled to complete three control trainings (160 trials each) followed by one standard VP task (80 trials) per cycle on the PDA. On control trainings, the dot was equally likely to replace the smoking picture and the neutral picture. Therefore, there was zero correlation between picture type and dot location. This control condition ensured that (a) the duration of each training should not differ between groups; (b) AR and control participants received equal practice on motoric aspects of the VP tasks; and (c) the same smoking and neutral pictures were presented with equal frequency to AR and control participants. In each cycle (both AR and control conditions), the assessment (the standard VP task) could occur on the same or following day as the previous training (AR or control), depending on the time of day the cycle was initiated. The first cycle was initiated by the first training after Visit 1. The first training (of a set of three) occurred at different times of day across participants.

ATTENTIONAL RETRAINING ADMINISTERED IN THE FIELD

variable was tested in a separate model. For primary analyses, parameter estimates (PE; 95% CI) from ANCOVAs and LMMs were reported as unstandardized measures of effect size (Wilkinson, 1999). By aggregating over participants, Cohen’s d was also reported for the two primary outcome variables, attentional bias and cued craving. (Measures of smoking behavior were secondary outcomes). For all tests, ␣ was set to .05, and all tests were two-tailed.2 The sample size was selected based on estimated population parameters derived from the pilot study and from the meta-analysis of Hakamata et al. (2010), which revealed an effect size of d ⫽ 0.61 for the effect of AR on anxiety. In our pilot, after Day 5, attentional bias in the AR group (⫺29.4 ms) was 46.4 ms lower (more negative) than in controls (⫹17.0 ms; SD ⫽ 70.7 ms). If the intraclass correlation coefficient was no greater than 0.3, then we had at least 90% power to detect an effect of Group.

Results AR (n ⫽ 30) and control (n ⫽ 30) participants did not differ in age, M ⫽ 41.8, SD ⫽ 10.2 vs. M ⫽ 43.6, SD ⫽ 14.0, respectively, F(1, 57) ⫽ 0.32, p ⫽ .58; gender, 50.0% female vs. 46.7% female, ␹2(1, 60) ⫽ 0.07, p ⫽ .80; race, 20.0% White, 56.7% Black, 23.3% other vs. 16.7% White, 66.7% Black, 16.7% other, ␹2(2, 60) ⫽ 0.67, p ⫽ .72; cigarettes smoked per day, M ⫽ 15.3, SD ⫽ 8.5 vs. M ⫽ 16.2, SD ⫽ 7.4, F(1, 58) ⫽ 0.21, p ⫽ .65; FTND scores, M ⫽ 5.4, SD ⫽ 2.3 vs. M ⫽ 5.2, SD ⫽ 2.2, F(1, 58) ⫽ 0.12, p ⫽ .73; number of lifetime quit attempts, M ⫽ 5.3, SD ⫽ 3.1 vs. M ⫽ 5.0, SD ⫽ 3.0; F(1, 58) ⫽ 0.13, p ⫽ .72; and intent to quit, 20% in next 30 days, 67% in next 6 months, and 13% not at all vs. 27% in next 30 days, 60% in next 6 months, and 13% not at all, respectively, ␹2(2, N ⫽ 60) ⫽ 0.39, p ⫽ .82. AR participants responded to (completed) 79.1% of 735 presented RAs and control participants responded to 80.3% of 738 presented RAs. The mean time of day for training was 3:34 p.m. (SD ⫽ 3.54 hr; AR) and 3:04 p.m. (SD ⫽ 3.79 hr; control), F(1, 827) ⫽ 1.34, p ⫽ .25, and for assessment was 3:49 p.m. (SD ⫽ 5.04 hr; AR group) and 4:26 p.m. (SD ⫽ 5.02 hr; control), F(1, 235) ⫽ 0.20, p ⫽ .66. The mean duration of assessments (n ⫽ 291) was 6.1 min (SD ⫽ 1.4). The mean duration of attentional retrainings and control trainings was 8.4 min (SD ⫽ 2.3) and 8.4 min (SD ⫽ 1.3), respectively, F(1, 826) ⫽ 0.07, p ⫽ .79. AR participants (n ⫽ 29) completed a mean of 15.0 (SD ⫽ 3.6) attentional retrainings and control participants (n ⫽ 30) completed a mean of 14.9 (SD ⫽ 3.4) control trainings, F(1, 56) ⫽ 0.35, p ⫽ .56. At the assessments, the mean duration from the preceding AR or control training was 9.7 hr (SD ⫽ 8.9) and 7.9 hr (SD ⫽ 6.7) for the AR and control groups, respectively, F(1, 230) ⫽ 1.84, p ⫽ .18. AR and control participants completed a mean of 2.74 (SD ⫽ 1.29) attentional retrainings and 2.70 (SD ⫽ 1.23) control trainings respectively since the previous assessment, F(1, 234) ⫽ 0.12, p ⫽ .73. The mean duration from the preceding AR or control training to Visit-2 (lab) VP assessment was 18.7 hr (SD ⫽ 17.5) and 15.5 hr (SD ⫽ 12.6) for the AR and control groups respectively, F(1, 56) ⫽ 0.47, p ⫽ .50.

Effect of AR on Attentional Bias The mean error rate (field assessments) was 0.4% (SD ⫽ 1.4). Summary statistics are reported in Table 1. There was significant

1237

attentional bias at baseline, N ⫽ 60, M ⫽ 12.2 ms, SD ⫽ 37.5, p ⫽ .01, and no between-groups difference in attentional bias at baseline, F(1, 58) ⫽ 0.29, p ⫽ .68. LMM (field data) revealed a Group ⫻ Day interaction, PE ⫽ 8.54, 95% CI [0.81, 16.3], p ⫽ .03 (see Figure 3)3, which persisted when controlling for time since last cigarette, PE ⫽ 8.61, 95% CI [0.88, 16.3], p ⫽ .03, and FTND scores, PE ⫽ 8.51, 95% CI [0.80, 16.2], p ⫽ .03. Attentional bias decreased over days in the AR group, PE ⫽ ⫺7.58, 95% CI [⫺14.4, ⫺0.72], p ⫽ .03, but not in the control group, PE ⫽ 0.92, CI [⫺3.67, 5.51], p ⫽ .69. Across all days, there was significant attentional bias in the control group, PE ⫽ 10.9, 95% CI [1.49, 20.3], p ⫽ .02, but not in the AR group, PE ⫽ ⫺3.26, 95% CI [⫺18.7, 12.2], p ⫽ .67. The two groups exhibited significantly different attentional bias after Day 5, PE ⫽ 31.4, 95% CI [5.89, 56.8], p ⫽ .01 (d ⫽ 0.69). However, the AR group did not exhibit statistically significant attentional avoidance after Day 5, PE ⫽ ⫺17.6, 95% CI [⫺38.1, 2.87], p ⫽ .09. Using ANCOVA, there was no effect of Group on attentional bias at Visit 2, PE ⫽ ⫺0.26, 95% CI [⫺20.8, 20.2], p ⫽ .98.

Effect of AR on Craving The mean cued craving rating at baseline (N ⫽ 60) was 4.38 (SD ⫽ 2.11). There was no between-groups difference in cued craving at baseline, F(1, 58) ⫽ 0.30, p ⫽ .59. LMM conducted on cued craving did not reveal a Group ⫻ Day interaction, PE ⫽ ⫺0.004, 95% CI [⫺0.21, 0.20], p ⫽ .96 (Figure 3, lower graph). In the reduced model (which excluded the interaction term) there was a significant main effect of group, PE ⫽ 0.77, 95% CI [0.00, 1.55], p ⫽ .04; cued craving ratings were significantly lower in the AR group (between-groups difference in cued craving, d ⫽ 0.56; see Figure 3). In contrast, there was no effect of group, PE ⫽ 0.41, 95% CI [⫺0.38, 1.20], p ⫽ .31, nor a Group ⫻ Day interaction, PE ⫽ ⫺0.025, 95% CI [⫺0.23, 0.18], p ⫽ .81, on the noncued craving item. It is noteworthy that the effect of group on cued craving persisted when controlling for noncued craving, PE ⫽ 0.42, 95% CI [0.09, 0.74], p ⫽ .01, and continued to persist when controlling for time since last cigarette, PE ⫽ 0.41, 95% CI [0.07, 0.74], p ⫽ .01, and FTND scores, PE ⫽ 0.43, 95% CI [0.10, 0.75], p ⫽ .01. Using ANCOVA, there was no effect of group on cued craving, PE ⫽ ⫺0.25, 95% CI [⫺1.30, 0.81], p ⫽ .64, noncued craving, PE ⫽ ⫺0.34, 95% CI [⫺1.32, 0.65], p ⫽ .50, or the QSU total score, PE ⫽ 0.22, 95% CI [⫺0.78, 1.22], p ⫽ .66 at Visit 2. 2 We used data from the 60 protocol completers for the primary analyses (see Figure 1). We also conducted intent-to-treat analyses using all available data. For lab data, this included 65 participants, including one participant who did not receive trainings. For participants with missing data at Visit 2, Visit 1 data were used. We analyzed the field data of the 60 participants with available data (including the three protocol violators). Results did not differ from analyses of protocol completers. 3 Number of reported interruptions during the VP task (AR vs. control) were as follows and did not differ by group (p ⫽ .96): No times ⫽ 58.2% versus 56.7%; 1 time ⫽ 19.2% versus 20.0%; 2 times ⫽ 14.4% versus 11.7%; 3 times ⫽ 2.7% versus 6.9%; 4⫹ times ⫽ 2.8% versus 4.8%. When the 15 assessments with 4⫹ interruptions were excluded, the p value for the Group ⫻ Day interaction became smaller, PE ⫽ 9.85, 95% CI [2.69, 17.0], p ⫽ .007.

KERST AND WATERS

1238

Table 1 Summary Statistics on Dependent Variables by Training Group and Day Day Condition

Variable

AR

Attentional bias (ms) Cued craving (1–7) Noncued craving (1–7) Smoked so far that day No. cigs smoked (diary) QSU total ratings (0–10) CO (ppm) Cotinine (ng/ml) Control Attentional bias (ms) Cued craving (1–7) Noncued craving (1–7) Smoked so far that day No. cigs smoked (diary) QSU total ratings (0–10) CO (ppm) Cotinine (ng/ml)

Base

1

10.1 (36.8) 16.1 (33.7) 4.53 (2.01) 4.25 (1.67) 4.83 (1.86) 3.50 (1.93) 96.7 (18.3) 100.0 (0.0) 12.6 (7.2) 5.05 (2.76) 15.9 (5.35) 394 (181) 14.4 (38.7) ⫺14.2 (52.5) 4.23 (2.22) 4.47 (1.85) 4.57 (2.18) 3.80 (2.14) 100.0 (0.0) 100.0 (0.0) 14.1 (6.1) 4.31 (2.86) 15.6 (4.68) 420 (253)

2

3

4

5

6

11.2 (116.9) 3.37 (2.03) 3.68 (2.08) 89.5 (31.5) 13.8 (7.9)

16.7 (26.0) 4.48 (1.94) 4.48 (2.16) 90.5 (30.1) 13.0 (7.9)

9.3 (61.4) ⫺23.1 (101.9) 2.90 (1.89) 3.88 (2.45) 3.29 (2.12) 4.00 (2.35) 90.5 (30.1) 68.0 (47.6) 14.1 (7.0) 13.7 (9.2)

16.0 (48.6) 4.55 (2.28) 4.25 (2.27) 85.0 (36.7) 14.4 (7.0)

18.8 (39.3) 4.32 (2.01) 4.18 (2.13) 90.1 (29.4) 15.1 (7.1)

8.6 (31.1) 4.48 (2.13) 4.39 (2.15) 87.0 (34.4) 14.8 (6.4)

29.5 (62.6) 4.00 (2.21) 4.25 (2.29) 91.7 (28.2) 13.8 (7.9)

7

Visit 2

⫺2.8 (57.6) ⫺22.6 (92.6) ⫺0.0 (46.1) 3.35 (2.58) 3.91 (2.36) 3.83 (2.22) 3.60 (2.52) 3.94 (2.42) 3.90 (2.02) 90.0 (30.8) 87.5 (33.6) 89.7 (31.0) 13.9 (7.5) 13.5 (8.0) 3.84 (2.33) 15.5 (7.70) 410 (211) 10.4 (30.1) 0.1 (72.0) ⫺0.3 (30.4) 3.85 (2.23) 4.24 (2.14) 3.43 (2.24) 3.35 (2.16) 4.19 (2.16) 3.43 (2.21) 95.0 (22.4) 76.2 (43.6) 90.0 (30.5) 13.3 (6.6) 13.3 (7.8) 3.66 (2.45) 14.2 (5.81) 408 (268)

Note. Mean (SD) for bias scores (ms), craving, and other study measures. Base ⫽ baseline; data for baseline and Visit 2 are from laboratory sessions. Data for smoked so far that day data are % yes responses. Data from Days 1–7 derive from ecological momentary (EMA) assessments. Day 7 column includes EMA assessments occurring on Day 8 (Visit 2 day) prior to the Visit 2. At baseline, ns were 30 for AR and control; ns for Days 1–7 for the control group were 15, 20, 22, 23, 24, 20, 21. For the AR group, ns were 8, 19, 21, 21, 25, 20, 32 for Days 1–7, respectively.

Effect of AR on Smoking

Training Data

AR and control participants reported smoking 13.5 (SD ⫽ 7.3) and 13.7 (SD ⫽ 6.7) cigarettes per day, respectively (smoking diary), and reported having “smoked so far that day” on 87.3% and 87.4% of assessments, respectively (field data). At baseline, there was no between-groups difference in CO, F(1, 58) ⫽ 0.03, p ⫽ .86, or cotinine, F(1, 58) ⫽ 0.21, p ⫽ .65. Using ANCOVA, there was no effect of group on CO levels, PE ⫽ ⫺1.07, 95% CI [⫺4.13, 1.99], p ⫽ .49, or salivary cotinine levels at Visit 2, PE ⫽ ⫺23.4, 95% CI [⫺105.4, 58.6], p ⫽ .57. Using LMM, there was no effect of group on cigarettes smoked per day, PE ⫽ 0.02, 95% CI [⫺1.84, 1.90], p ⫽ .98, nor a Group ⫻ Day interaction, PE ⫽ ⫺0.30, 95% CI [⫺0.62, 0.02], p ⫽ .074. Using mixed-model logistic regression, there was no effect of group on smoking so far that day, PE ⫽ ⫺0.18, 95% CI [⫺1.25, 0.90], p ⫽ .75, nor a Group ⫻ Day interaction, PE ⫽ 0.04, 95% CI [⫺0.35, 0.43], p ⫽ .85.

In exploratory analyses, training RTs for AR participants (i.e., on probes replacing neutral pictures, M ⫽ 756.8 ms, SD ⫽ 239.2) did not get faster over days, PE ⫽ ⫺1.19, SE ⫽ 3.70, p ⫽ .75. However, the effect of day on training RTs (AR participants) was moderated by awareness, PE ⫽ ⫺25.4, SE ⫽ 7.13, p ⫽ .0004. RTs of aware participants became slower over days, PE ⫽ 12.2, SE ⫽ 4.62, p ⫽ .009, whereas RTs of unaware participants became faster over days, PE ⫽ ⫺12.5, SE ⫽ 4.42, p ⫽ .005.

Contingency Awareness Fourteen (46.8%) AR participants were aware (i.e., they were able to describe the relationship between the probes and the picture types) and 16 (53.3%) were unaware. On field data (AR participants), in exploratory analyses, awareness did not moderate the effect of day on attentional bias (Awareness ⫻ Day interaction, PE ⫽ ⫺5.61, SE ⫽ 5.79, p ⫽ .33)5. However, aware participants exhibited generally smaller (more negative) attentional bias; main effect of awareness, PE ⫽ 37.4, SE ⫽ 13.5, p ⫽ .007. There was no effect of awareness of cued craving (field), PE ⫽ ⫺0.25, SE ⫽ 0.63, p ⫽ .69, noncued craving (field), PE ⫽ 0.10, SE ⫽ 0.58, p ⫽ .87, attentional bias (lab), PE ⫽ ⫺3.03, SE ⫽ 17.9, p ⫽ .87, or any other measure (all ps ⬎ .17).

Discussion The main findings were as follows. First, as hypothesized, AR delivered using a PDA in the natural environment reduced attentional bias to smoking cues over days. Second, AR did not reduce noncued craving, but, as hypothesized, AR did reduce craving following a stimulus that included smoking and nonsmoking content. Third, there was no evidence that AR reduced smoking in this sample of nontreatment-seeking smokers. As hypothesized, attentional bias declined over days in the AR group. Notably, the decline took a number of days to develop (see Figure 3). More attentional retraining sessions, or attentional retraining sessions with more training trials, may be necessary for more rapid and larger reductions in attentional bias. In addition, AR in the current study occurred mainly when participants had been smoking. 4 Although the Group ⫻ Day interaction was not significant, we calculated the effect of day for each training group. The effect of day was not significant for AR participants, PE ⫽ 0.12, SE ⫽ 0.13, p ⫽ .32, or for control participants PE ⫽ ⫺0.18, SE ⫽ 0.10, p ⫽ .07. 5 Although the Awareness ⫻ Day interaction was not significant, we calculated the effect of day for each awareness group. The effect of day was not significant for the aware, PE ⫽ ⫺6.16, SE ⫽ 3.86, p ⫽ .12, but was significant for the unaware, PE ⫽ ⫺10.6, SE ⫽ 4.01, p ⫽ .01.

ATTENTIONAL RETRAINING ADMINISTERED IN THE FIELD

Figure 3. Effect of training group on attentional bias (upper graph) and cued craving (lower graph). AR ⫽ attentional retraining group; CON ⫽ control group; base ⫽ baseline (preintervention).

AR may be more effective when training occurs in an abstinent state, when attentional bias and craving may be elevated.6 Another possible reason for the relatively slow decline in attentional bias is that the PDA protocol mixed attentional retrainings with attentional bias assessments. Completion of the attentional bias assessments may have slowed the development of attentional retraining because during the VP assessment, there was no correlation between the probe location and the picture location. For an AR participant, completion of a VP assessment was the same as completing a (briefer) control training. Although group differences in attentional bias were observed in the latter part of the field data, there were no group differences in attentional bias (or craving) at the second laboratory visit. Participants never experienced AR in the laboratory. It is possible that an effect of AR is most easily detected when the assessment occurs in the same setting as the training. If so, AR should be administered in as many settings as possible. Alternatively, an effect of AR may have been more difficult to detect at the laboratory visit because there was a longer lag between the previous training and the lab assessment than was the case for field assessments. In addition, field assessments used stimulus materials on which participants had just received training, whereas the lab assessment at Visit 2 used materials from throughout the previous week. The most interesting finding of the study was that AR reduced craving following the compound stimulus. The hypothesis was that smokers undergoing AR would be less likely to attend to the smoking part of the picture, because they would attend to the nonsmoking features. They would therefore experience less “exposure” to the smoking component, and therefore experience less craving. Overall, this hypothesis received support. However, it is surprising that an

1239

effect of AR on cued craving did not become larger over days (see Figure 3). Although counterintuitive, the pattern of data resembles findings from AR studies using a modified dot-probe task in anxiety. Hakamata et al. (2010) reported that the number of trainings significantly moderates the effect of AR on attentional bias, but not the effect of AR on self-reported anxiety. There were no effects of AR on smoking behavior. The one (nonsignificant) trend indicated that smoking tended to decrease over days in the control group, relative to the AR group. We find it noteworthy that no previous study has documented an effect of AR on smoking behavior (Attwood et al., 2008; Field, Duka, et al., 2009; McHugh et al., 2010). Also, the observation that AR decreased cued craving without reducing smoking suggests that reduction of the former does not necessarily lead to reduction of the latter. These points notwithstanding, the sample was comprised of nontreatment seekers, which may have made it difficult to detect an effect of AR on smoking. Many (53.3%) AR participants were unaware of the training contingencies. Unaware AR participants exhibited reductions in attentional bias over days, and cued craving, comparable to the aware AR participants. AR appears to involve automatic processes for unaware participants, which supports the role of automatic processes in retraining (Mathews & MacLeod, 2002). Aware AR participants, however, exhibited generally smaller (more negative) attentional bias than the unaware, and RTs of aware AR participants got slower over days during training. In some participants, AR may engage controlled processes (Schoenmakers et al., 2010) that, speculatively, can slow down overall responding, but also reduce attentional bias. Alternatively, the slower response times in the AR participants may reflect, in part, a lack of engagement. Research using different presentation durations over longer training periods and manipulating awareness as an independent variable through explicit coaching (Schoenmakers et al., 2010) might clarify the cognitive mechanisms underlying AR in smokers. In particular, it would be important to determine the relative importance of implicit and controlled processes in AR. The study had several limitations. First, it did not investigate whether an effect of AR on attentional bias generalizes to different stimuli or different tasks. However, an effect of AR was observed on craving following the smoking pictures, which were not used in training. Second, the study had a relatively small sample size, with relatively low power to detect small and medium effect sizes for measures taken in the laboratory. Power was greater for EMA measures because of the larger number of assessments. Third, as noted earlier, we recruited nontreatment seekers. Future studies should recruit treatment-seeking smokers. Fourth, the cued craving measure was always presented at the beginning of the assessment and therefore before the noncued craving measure. The absence of counterbalancing precludes making strong comparisons between the two measures. Moreover, the cued craving assessment was developed for this study, and has not been previously validated. Also, we do not know whether the observed effect of AR on cued craving assessed on the PDA generalizes to naturalistic smoking cues, a limitation which is sharpened by the absence of an effect of AR on smoking. Fifth, we did not ask participants to record smoking on the PDA. AR may have exerted a subtle effect on smoking behavior that would have been undetect6 It may be useful to combine the training protocol used here with training administered in a controlled laboratory setting, when participants are verifiably abstinent.

KERST AND WATERS

1240

able with our methods. Sixth, there was no correction for multiple tests. In particular, results of the exploratory analyses should be treated with caution pending replication. Last, further research using these methods is needed to delineate the causal relationships between attentional bias, craving, and smoking behavior. The study also had strengths. In particular, this is the first study to report the use of an AR intervention administered on a mobile device in the natural environment. In summary, multiple sessions of AR administered in the field can reduce attentional bias to smoking cues as well as craving in response to a cue presented on a PDA. Further research should identify the cognitive mechanisms underlying AR and ascertain its clinical relevance in reducing smoking behavior.

References Attwood, A. S., O’Sullivan, H., Leonards, U., Mackintosh, B., & Munafo, M. R. (2008). Attentional bias training and cue reactivity in cigarette smokers. Addiction, 103, 1875–1882. doi:10.1111/j.1360-0443.2008 .02335.x Cox, L. S., Tiffany, S. T., & Christen, A. G. (2001). Evaluation of the brief questionnaire of smoking urges (QSU-brief) in laboratory and clinical settings. Nicotine & Tobacco Research, 3, 7–16. doi:10.1080/ 14622200124218 Fadardi, J. S., & Cox, W. M. (2009). Reversing the sequence: Reducing alcohol consumption by overcoming alcohol attentional bias. Drug and Alcohol Dependence, 101, 137–145. doi:10.1016/j.drugalcdep.2008.11.015 Field, M., & Cox, W. M. (2008). Attentional bias in addictive behaviors: A review of its development, causes, and consequences. Drug and Alcohol Dependence, 97, 1–20. doi:10.1016/j.drugalcdep.2008.03.030 Field, M., Duka, T., Tyler, E., & Schoenmakers, T. (2009). Attentional bias modification in tobacco smokers. Nicotine & Tobacco Research, 11, 812– 822. doi:10.1093/ntr/ntp067 Field, M., & Eastwood, B. (2005). Experimental manipulation of attentional bias increases the motivation to drink alcohol. Psychopharmacology, 183, 350 –357. doi:10.1007/s00213-005-0202-5 Field, M., Munafo, M. R., & Franken, I. H. A. (2009). A meta-analytic investigation of the relationship between attentional bias and subjective craving in substance abuse. Psychological Bulletin, 135, 589 – 607. doi: 10.1037/a0015843 Fitzmaurice, G. (2001). A conundrum in the analysis of change. Nutrition, 17, 360 –361. doi:10.1016/S0899-9007(00)00593-1 Franken, I. H. A. (2003). Drug craving and addiction: Integrating psychological and neuropsychopharmacological approaches. Progress in Neuro-Psychopharmacology & Biological Psychiatry, 27, 563–579. doi: 10.1016/S0278-5846(03)00081-2 Gilbert, D. G., Sugai, C., Zuo, Y., Rabinovich, N. E., McClernon, F. J., & Froeliger, B. (2007). Brain indices of nicotine’s effects on attentional bias to smoking and emotional pictures and to task-relevant targets. Nicotine & Tobacco Research, 9, 351–363. doi:10.1080/14622200701188810 Hakamata, Y., Lissek, S., Bar-Haim, Y., Britton, J., Fox, N., Leibenluft, E., & Pine, D. S. (2010). Attention bias modification treatment: A metaanalysis toward the establishment of novel treatment for anxiety. Biological Psychiatry, 68, 982–990. doi:10.1016/j.biopsych.2010.07.021 Hallion, L. S., & Ruscio, A. M. (2011). A meta-analysis of the effect of cognitive bias modification on anxiety and depression. Psychological Bulletin, 137, 940 –958. doi:10.1037/a0024355 Heatherton, T. F., Kozlowski, L. T., Frecker, R. C., & Fagerström, K. O. (1991). The Fagerström Test for Nicotine Dependence: A revision of the Fagerström Tolerance Questionnaire. British Journal of Addiction, 86, 1119 –1127. doi:10.1111/j.1360-0443.1991.tb01879.x

Janes, A. C., Pizzagalli, S. R., Richardt, S., Frederick, B. D., Chuzi, S., Pachas, G., & Kaufman, M. J. (2010). Brain reactivity to smoking cues prior to smoking cessation predicts ability to maintain tobacco abstinence. Biological Psychiatry, 67, 722–729. doi:10.1016/j.biopsych.2009.12.034 Mathews, A., & MacLeod, C. (2002). Induced processing biases have causal effects on anxiety. Cognition and Emotion, 16, 331–354. doi: 10.1080/02699930143000518 McHugh, R. K., Murray, H. W., Hearon, B. A., Calkins, A. W., & Otto, M. W. (2010). Attentional bias and craving in smokers: The impact of a single attentional training session. Nicotine & Tobacco Research, 12, 1261–1264. doi:10.1093/ntr/ntq171 Perkins, K. A. (2009). Does smoking cue-induced craving tell us anything important about nicotine dependence? Addiction, 104, 1610 –1616. doi: 10.1111/j.1360-0443.2009.02550.x Powell, J., Dawkins, L., West, R., & Pickering, A. (2010). Relapse to smoking during unaided cessation: Clinical, cognitive and motivational predictors. Psychopharmacology, 212, 537–549. doi:10.1007/s00213010-1975-8 Robinson, T. E., & Berridge, K. C. (1993). The neural basis of drug craving: An incentive-sensitization theory of addiction. Brain Research Reviews, 18, 247–291. doi:10.1016/0165-0173(93)90013-P Rooke, S. E., Hine, D. W., & Thorsteinsson, E. B. (2008). Implicit cognition and substance use: A meta-analysis. Addictive Behaviors, 33, 1314 –1328. doi:10.1016/j.addbeh.2008.06.009 Schoenmakers, T. M., de Bruin, M., Lux, I. R. A., Goertz, A. G., Van Kerkhof, D. H. A. T., & Wiers, R. W. (2010). Clinical effectiveness of attentional bias modification training in abstinent alcoholic patients. Drug and Alcohol Dependence, 109, 30 –36. doi:10.1016/j.drugalcdep .2009.11.022 Shiffman, S. (2009). Responses to smoking cues are relevant to smoking and relapse. Addiction, 104, 1617–1618. doi:10.1111/j.1360-0443.2009 .02580.x Society for Research on Nicotine and Tobacco Subcommittee on Biochemical Verification. (2002). Biochemical verification of tobacco use and cessation. Nicotine & Tobacco Research, 4, 149 –159. doi:10.1080/ 14622200210123581 Swanson, J., Swanson, E., & Greenwald, A. (2001). Using the Implicit Association Test to investigate attitude– behaviour consistency for stigmatised behaviour. Cognition and Emotion, 15, 207–230. doi:10.1080/ 0269993004200060 Tiffany, S. T. (1990). A cognitive model of drug urges and drug-use behavior: Role of automatic and nonautomatic processes. Psychological Review, 97, 147–168. doi:10.1037/0033-295X.97.2.147 Waters, A. J., & Li, Y. (2008). Evaluating the utility of administering a reaction time task in an ecological momentary assessment study. Psychopharmacology, 197, 25–35. doi:10.1007/s00213-007-1006-6 Waters, A. J., Shiffman, S., Sayette, M. A., Paty, J. A., Gwaltney, C. J., & Balabanis, M. H. (2003). Attentional bias predicts outcome in smoking cessation. Health Psychology, 22, 378 –387. doi:10.1037/0278-6133.22.4 .378 Wiers, R. W., & Stacy, A. W. (Eds.). (2006). Handbook of impliCIt cognition and addiction. Thousand Oaks, CA: Sage. Wilkinson, L. (1999). Statistical methods in psychology journals: Guidelines and explanations. American Psychologist, 54, 594 – 604. doi: 10.1037/0003-066X.54.8.594 Wray, J. M., Godleski, S. A., & Tiffany, S. T. (2011). Cue-reactivity in the natural environment of cigarette smokers: The impact of photographic and in vivo smoking stimuli. Psychology of Addictive Behaviors, 25, 733–737. doi:10.1037/a0023687

Received September 20, 2012 Revision received April 5, 2013 Accepted May 3, 2014 䡲