Validity and Reliability of the Cigarette Purchase Task when Participant Cigarette Consumption is Unconstrained

Abstract

Hypothetical purchase tasks offer effective and efficient methods to assess the reinforcing value of various substances, including cigarettes. The purpose of the present study is to examine the validity and reliability of the Cigarette Purchase Task (CPT) in an experimental arrangement in which participants were receiving free cigarettes. Critical to the validity of the CPT is that those who smoke can accurately estimate how much they would smoke under varying economic constraints. Participants (N = 9) were provided free study cigarettes for eight weeks. Participants completed the CPT once weekly. To examine the validity of the five CPT demand indices (i.e., demand intensity, Pmax, Omax, breakpoint, and alpha), we used a simple linear regression stratified by session number to model which of the five CPT demand indices were associated with the number of cigarettes smoked per day during Week 1 of the experiment. Significant associations in the hypothesized direction were noted across the five CPT indices with the evidence for validity greatest for intensity, followed by Omax, Pmax, breakpoint, and alpha. To examine CPT test-retest reliability, we estimated interclass correlation coefficients (ICC) between Sessions 1 and 4 and Sessions 5 and 8. All but one ICC supported “good” or “excellent” reliability with the only exception seen with the alpha index between Sessions 1 and 4, which was moderate reliability. Collectively, these results provide evidence supporting the construct validity and temporal stability/reliability of the CPT demand indices under conditions of limited economic constraint.

Behavioral economics integrates the disciplines of economics and psychology to help understand why people deviate from decisions that might be in their best interest (Reed et al., 2013). Behavioral economics has been used widely to understand substance misuse and dependence, including nicotine dependence. Dependence can be conceptualized as overvaluing of a substance to the exclusion of other important reinforcers, such as food, work, or family (Higgins et al., 2004). One such behavioral economic tool is the Cigarette Purchase Task (CPT) that assesses hypothetical cigarette consumption under increasing economic constraint. Participants are typically asked how many hypothetical cigarettes they would smoke over a range of price points (e.g., $0.25/cigarette, $1.00/cigarette) in a 24 hr period. Hypothetical purchase tasks are quick and easy to administer, which contrasts with laboratory measures of demand that require participants to respond repeatedly for a substance, long sessions, and repeated visits (e.g., Shahan et al., 1999). The value of the CPT was recognized by the Consortium on Methods Evaluating Tobacco, which recommended that it be incorporated into research that informs the Food and Drug Administration’s regulation of tobacco products (Berman et al., 2018).

Five demand indices can be generated in the CPT: (a) Demand Intensity is the demand for cigarettes when they are free or consumption is unconstrained. (b) Omax is the maximum amount of money someone is willing to spend on cigarettes. (c) Pmax is the point at which demand for cigarettes becomes elastic (i.e., small price increases result in corresponding reductions in consumption). (d) Breakpoint is the cost of cigarettes at which consumption falls to zero. (e) Alpha (α) describes the rate at which consumption decreases as a function of price increases.

The correspondence between CPT indices and cigarette smoking (and their correlates) is a measure of the validity of the CPT. Zvorsky et al. (2019) reviewed 34 studies that implemented the CPT and examined its relation to smoking and correlates (e.g., e-cigarette use, negative affect, cotinine levels, nicotine dependence). These studies varied in methodology and included cross-sectional, longitudinal, and experimental analyses. Demand intensity, followed by Omax, had the highest number of significant associations with other cigarette smoking outcomes and the largest effect sizes (Zvorsky et al., 2019). A similar pattern was observed alcohol and other substance use.

Researchers have tested whether respondent self-reported consumption of cigarettes on the CPT is associated with actual consumption of cigarettes. If the CPT indices correlate with or predict actual consumption the measure can be considered valid. Wilson et al. (2016) evaluated the association between the CPT and purchases of research cigarettes at four values ranging from $0.12 to $1.00 per cigarette. Participants purchased research cigarettes for one week at each of these five values. Self-reported consumption on the CPT differed significantly from actual consumption, although most of the values were correlated (rs ranged = 0.49–0.74). Smith et al. (2016), in a large randomized clinical trial examining effects of cigarettes that varied in nicotine content, collected CPT data at baseline and cigarettes smoked per day during Week 6 of exposure to free study cigarettes. Demand intensity and actual cigarettes smoked per day were highly correlated (r = .68), but the raw differences were not reported. Nighbor et al. (2020) compared self-reported baseline demand intensity to consumption of free cigarettes. Hypothetical demand was correlated (rs ranged from = 0.67–0.70) with actual consumption although the former was on average 4.4 cigarettes per day greater than actual consumption. Although demand intensity overestimated actual consumption, it was nevertheless sensitive to differences in other well-established correlates of nicotine dependence and cigarette smoking, including consumption between those with lower versus higher educational attainment, and those with versus without opioid use disorder (Nighbor et al.).

A smaller number of studies have assessed the reliability and temporal stability of the CPT. Few et al. (2012) recruited people who smoked (N = 11) an average of 22.3 cigarettes per day from the community and had them complete the CPT two times, 1-week apart. Temporal reliability was excellent across all indices (rs ranged = 0.76–0.99). Wilson et al. (2016) assessed reliability across six administrations of the CPT and reported excellent reliability (median Cronbach’s alpha = 0.99) among people who smoked (N = 19). Additional data on the reliability of the CPT may be warranted in light of the increasing use of this method. Numerous studies have tested the validity of the CPT (see Zvorsky et al., 2019 for a review), but additional data is needed in the context of an experiment in which study cigarettes are of particular interest, as in Donny et al. (2017). Questions remain regarding whether participants can accurately predict the number of cigarettes they would smoke with cigarettes they have less experience smoking, and whether these reports correlate with other measures of nicotine dependence. Moreover, study cigarettes are typically provided free of charge so participants don’t have experience purchasing these cigarettes as they would with their usual brand cigarettes.

Key to the validity of the CPT is that those who smoke can accurately estimate how much they would consume under varying economic constraints. As such, studies like the present one that provide participants with free cigarettes provide an opportunity to examine associations between self-reported consumption on the CPT and actual consumption of cigarettes when available at no or minimal economic constraint. The purpose of the present study is to assess the validity and reliability of the CPT under such conditions.

Method

Participants.

Eleven adult participants met eligibility requirements and consented to participate, although two dropped out after completing a single experimental session (N = 9). Participants that dropped out from the study did not have enough data to be analyzed because data from each weekly session was needed. Participants were recruited for a study to compare differences in nicotine dependence based upon varying cigarette nicotine yields (nicotine yield refers to the amount of nicotine that a user inhales rather than the nicotine content of the cigarette). It is typically machine measured, though people who smoke can vary their smoking topography to acquire more nicotine [US Department of Health and Human Services, 2001]. The protocol was approved by the University Human Subjects Committee (#17160). Data are available upon reasonable request.

To be included in the study, participants had to report smoking high nicotine yield cigarettes, no plans to quit smoking within the next 3 months, report smoking ≥ 5 cigarettes per day for the past year, provide a breath carbon monoxide (CO) sample ≥ 5 parts per million (PPM), report no current serious mental illness, be sufficiently literate to complete research tasks, not pregnant or nursing, and report no use of other tobacco products within the past month. Exclusion criteria included exclusive use of roll-your-own cigarettes, recent history of suicidal plans, use of mentholated cigarettes, or positive self-report of illicit drug use in the past 30 days (not including marijuana), an expired breath sample positive for recent alcohol use, breath CO ≥ 80 PPM, or participation in another smoking study in the past 30 days. Smoking mentholated cigarettes was an exclusion because it was difficult to match nicotine yield in comparable mentholated cigarettes, and some of the critical nicotine yield data was missing (Agnew-Heard et al., 2016).

We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study, and we follow JARS (Applebaum et al., 2018). All data, analysis code, and research materials are available upon reasonable request. Data were analyzed using R, version 4.3.3 (R Core Team, 2020) and the package “beezdemand”, version 0.1.2 (Kaplan et al., 2019). This study’s design and its analysis were not pre-registered.

General Procedure.

Sessions were conducted in a small clinic, equipped with desks, chairs, and a Windows-based computer that was used to complete assessments. Participants completed 10 similarly structured experimental sessions, with approximately one week between sessions. In chronological order, participants completed an intake session, baseline session, and eight experimental sessions (during which time they smoked free study cigarettes), followed by a final follow-up session. In each session participants reported daily smoking consumption as well as any other tobacco product use on the Timeline Follow Back (TLFB; Sobel & Sobel, 1992), completed a range of questionnaires that measured nicotine dependence, craving, and other psychological distress (e.g., Fagerstrom Test of Nicotine Dependence [Heatherton et al., 1991], Wisconsin Inventory of Smoking Dependence Motives-Brief scale [Piper et al., 2008], Beck Depression Inventory [Beck et al., 1961], The Minnesota Nicotine Withdrawal Scale [Hughes & Hatsukami, 1986], and researchers collected carbon monoxide and breath alcohol levels. At the eight experimental sessions participants were provided a one-week supply of study cigarettes. The final follow-up session was similarly structured to other sessions, but followed a week in which participants did not receive free study cigarettes.

Study Cigarettes.

Participants received a free weekly supply of study cigarettes prior to each experimental session. Participants received 150% of the self-reported number of cigarettes smoked per day from the 7-day baseline report of cigarette smoking. Participants received Marlboro Reds for four weeks and Marlboro Silvers for four weeks. Marlboro Red cigarettes are considered high-nicotine yield and Marlboro Silvers are considered low-nicotine yield. Order of study cigarette distribution was counterbalanced across participants.

Cigarette Purchase Task.

The CPT assessed hypothetical consumption of cigarettes at increasing amounts of money for a 24-hour period. Verbal instructions from the experimenter included the following: (a) The available cigarettes are the study cigarettes; (b) You have the same income/savings that you have now and NO ACCESS to any cigarettes or nicotine products other than those offered at these prices; (c) You can smoke without restrictions and without factoring in what might occur in the next 24 hours related to your participation in the study; (d) You would smoke the cigarettes that you request at this time, not save or stockpile cigarettes for a later date. The following twenty cigarette prices were assessed, and prompts included the corresponding price per pack to help participants assess spending: $0.00, $0.02, $0.05, $0.10, $0.20, $0.30, $0.40, $0.50, $0.60, $0.70, $0.80, $0.90, $1, $2, $3, $4, $5, $10, $20, $40. The CPT implementation followed best-practice guidelines outlined by Reed et al. (2020).

CPT data were analyzed for systematic responding according to Stein et al. (2015). Data from Participant 110 (week 2) was flagged for insensitivity to price increases (i.e., 20 cigarettes were endorsed at each price point). In this case we retained the participant’s data in the analyses because all other CPT data from this participant were deemed systematic. We analyzed empirical (observed) CPT indices for Intensity, Omax, Pmax, and Breakpoint. Demand intensity was coded as the number of cigarettes purchased at $0.00. Omax was coded as the largest amount of money a participant was willing to spend on cigarettes, determined by multiplying consumption by price at each price point. Pmax was coded as the cigarette value at which Omax occurred. Breakpoint was coded as the last individual cigarette price with any consumption. If participants endorsed consumption at the final CPT price point (i.e., no breakpoint) then their breakpoint was coded as $40 (Higgins et al., 2017). Alpha (α) was derived according to the exponentiated demand equation (Koffarnus et al., 2015) using Graphpad Prism 7.05 (Graphpad Software, La Jolla, California). We used Derek Reed’s freely available Graphpad Prism template to calculate alpha for each participant and timepoint (http://www.behavioraleconlab.com/resources---tools.html).

Timeline Follow-back.

The TLFB (Sobel & Sobel, 1992) was used to assess self-reported number of cigarettes smoked per day (CPD). The TLFB was completed in the clinic as an interview at each experimental session. To assess how well participants accurately predicted their use of free cigarettes on the CPT we used the TLFB as the gold-standard. Specifically, the number of cigarettes participants reported smoking at $0.00 (free) was compared to the TLFB number of cigarettes smoked the following day. This comparison was made once for each week while participants were receiving free cigarettes.

Analytic Strategy.

To examine the validity of CPT demand indices we used simple linear regression models, stratified by session number, in which the five CPT demand indices were regressed on CPD during the first day of the first week participants received free study cigarettes. We calculated a bootstrapped coefficient and bootstrapped 95% confidence interval for each model to ascertain a likely population parameter. Means and standard deviations of bootstrapped coefficients are reported. We hypothesized that intensity, Pmax, Omax, and breakpoint would be positively associated with CPD, while alpha would be inversely associated with CPD. To establish test-retest reliability on demand indices, we calculated an intraclass correlation coefficient (ICC) for times 1 and 4, and additionally on times 5 and 8 for each demand index (cigarette assignments of low- or high- nicotine yield were consistent within these time periods).

Lastly, demand intensity for study cigarettes (consumption at $0) was compared to self-reported smoking rates on the TLFB on the following day for each week in the study (the use of the first day of cigarette smoking following the CPT is consistent with Nighbor et al., 2020). The dependent variable for this research question was the difference between demand intensity and self-reported CPD on the day after completing the CPT. We estimated a linear mixed model in which difference scores were regressed on the fixed effect of “time” (i.e., the experimental session number), while the random intercept was each participant’s ID number. The beta coefficient for the time variable was assessed for significance at an alpha level of 0.05. To assess the generality of the findings a t-test was conducted between the differences in consumption of usual brand cigarettes and predicted consumption of usual brand cigarettes at $6.00 per pack at baseline and follow-up (i.e., did participants become more accurate at predicting consumption of usual brand cigarettes?). Sample sizes were calculated when the study was conducted on the primary, but not secondary outcomes. Data and analysis code are available upon reasonable request.

Results

Participants.

Participants were primarily male (56%), White (78%), and divorced (44%), with an average age of (M ± SD) 43 ± 14 years. Fifty-six percent of the sample obtained a 2-year college degree and the average reported household income was $22,156 ± 16,630. Twenty-two percent of the sample reported being currently employed. Participants smoked 16.9 ± 5.5 cigarettes per day and had an FTND score of 6.1 ± 1.3.

Validity of the CPT.

Results of the validity analysis for the CPT are presented in Table 1. Regarding intensity, the beta coefficients at each time point for the CPD variable were positive consistent with demand intensity and CPD being positively correlated. Similar positive associations were found for CPD with Omax and Pmax.

Table 1.

Association between cigarettes smoked per day the first day after receiving free study cigarettes during week 1 (CPD) and cigarette purchase task demand indices at eight time points

	Intensity	Omax	Pmax	Breakpoint	Alpha
Variable	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)
CPD T1	0.83 (0.55, 1.96)	0.42 (0.08, 1.86)	0.32 (0.18, 0.56)	0.28 (0.05, 1.05)	−0.02a (−0.04, −0.01)
CPD T2	0.27 (0.02, 0.83)	0.52 (0.23, 0.70)	0.62 (0.15, 1.64)	0.61 (0.15, 1.13)	−0.02a (−0.04, −0.01)
CPD T3	0.72 (0.56, 0.84)	0.79 (0.25, 3.84)	0.51 (0.01, 1.83)	0.49 (0.04, 2.20)	−0.03a (−0.04, −0.01)
CPD T4	0.66 (0.21, 0.80)	0.64 (0.11, 1.02)	0.50 (0.01, 1.43)	0.46 (0.44, 1.22)	−0.05a (−0.07, −0.02)
CPD T5	0.55 (0.46, 0.86)	0.07 (0.05, 0.40)	0.03a (0.01, 0.73)	−0.01 (−0.04, 0.07)	−0.03a (−0.09, −0.02)
CPD T6	0.72 (0.55, 0.76)	0.03a (0.01, 0.07)	0.23 (0.16, 0.62)	0.24 (0.03, 0.91)	−0.03a (−0.08, −0.01)
CPD T7	1.33 (1.08, 1.55)	0.95 (0.24, 2.75)	0.78 (0.02, 1.47)	0.72 (0.01, 3.01)	−0.07a (−0.18, −0.02)
CPD T8	0.90 (0.51, 1.34)	0.61 (0.29, 1.04)	0.43 (0.34, 0.47)	0.45 (0.06, 3.05)	−0.04a (−0.13, −0.03)

Note. b = bootstrapped beta coefficient; 95% CI = bootstrapped 95% confidence interval;

^a= dependent variable log transformed; M = mean of the bootstrapped beta coefficients; T = time

Overall, the strength of relationship – or the evidence for validity – was greatest for intensity (M ± SD)(M = 0.75 ± 0.30), followed by Omax (M = 0.50 ± 0.32), Pmax (M = 0.42 ± 0.23), breakpoint (M = 0.41 ± 0.23), and then alpha (M = −0.03 ± 0.02). Regarding breakpoint, all beta coefficients for CPD were positive, except for the model for Session 5. Regarding alpha, all coefficients were negative (i.e., in the expected direction) for CPD, indicating that as the number of free cigarettes smoked during the first week of exposure to free study cigarettes increased, demand was increasingly inelastic.

Reliability of CPT.

To assess CPT test-retest reliability, we estimated ICCs between Sessions 1 and 4, as well as between 5 and 8. According to Koo and Li (2016) ICCs calculated with 95% CIs should be interpreted as follows: those below 0.5 are “poor”, 0.50–0.75 are “moderate”, 0.75–0.90 are “good”, and 0.90 and above are “excellent”. ICCs for Demand Intensity between Sessions 1 and 4, and between 5 and 8 were 0.77 and 0.92, respectively. ICCs for Omax between Sessions 1 and 4, and between 5 and 8 were 0.97 and 0.78, respectively. ICCs for Pmax between Sessions 1 and 4, and between 5 and 8 were 0.99 and 0.77, respectively. ICCs for breakpoint between Sessions 1 and 4, and between 5 and 8 were 0.99 and 0.76, respectively. ICCs for Alpha (log transformed) between Sessions 1 and 4, and between 5 and 8 were 0.68 and 0.91, respectively. All but one ICC was indicative of good or excellent reliability. The exception is alpha between Sessions 1 and 4, which had moderate reliability. Collectively, these results support the temporal stability/reliability of the CPT demand indices.

Demand Intensity.

Demand intensity for free study cigarettes was (M ± SD) M = 26.7 ± 10.3 prior to Week 1 (Figure 1). On the first day of Week 1 participants smoked M = 21.6 ± 10.8 cigarettes, which was M = 5.1 ± 5.3 cigarettes lower than demand intensity for study cigarettes. Demand intensity for study cigarettes prior to Week 8 of the study was M = 25.6 ± 9.5, and participants smoked 23.4 ± 8.6 cigarettes on the first day of week 8. Demand intensity was M = 2.1 ± 5.5 cigarettes higher than what was consumed. This difference was smaller than what was observed in Week 1. The difference between Demand Intensity and CPD was fairly consistent from Week 1–6 and decreased sharply in Week 7–8. Differences in CPD and Demand Intensity scores at time points 1 vs 6, 1 vs 8, and 6 vs 8 were tested in post-hoc paired t-tests. Results for 1 vs 6 showed that differences were not significant (t = 1.54, p = 0.16). Results for 1 vs 8 showed that differences were significant (t = −2.59, p = 0.03). Results for 6 vs 8 showed that differences were not significant (t = 0.79, p = 0.45). A paired t-test on the difference scores from baseline to follow-up was not significant (t = 2.14, p = 0.07). The mean difference of the difference scores was 5.22 (95% CI = −0.41, 10.85).

Figure 1.

Relationship between demand intensity and study cigarettes smoked the following day across ten experimental sessions.

Note. Demand data for usual brand cigarettes was shown at $6 per pack which is the approximate cost of cigarettes per pack for participants where the study was conducted.

The downward trend depicted in Figure 1 was substantiated by the results of our linear mixed model. Specifically, results showed that time was inversely associated with difference scores (b = –0.52, SE = 0.18, p = 0.01). The random effect variance was 5.26 (SD = 2.30). Considering all possible comparisons between demand intensity and number of cigarettes smoked the following day, demand intensity was higher in 43 of 72 opportunities (59.7%), identical in 22 of 72 (30.6%) opportunities, and lower in 7 of 72 opportunities (9.7%).

Discussion

Data from the present study support the validity of the CPT in the context of an experiment in which participants consumption of actual cigarettes was largely unconstrained. All five demand indices were related to consumption of study cigarettes. Construct validity refers to the notion that the CPT is measuring what it is purported to measure (reinforcing value of smoking, or perhaps an index of dependence severity). Similar to previous observations, demand intensity and Omax had the highest validity scores (Zvorsky et al., 2019). These two demand indices appear to be the most sensitive and are most consistently related to demand for cigarettes as well as use of other substances. Other indices have also been correlated with substance use outcomes, but not as frequently (alpha = 72.1%, breakpoint = 62.1%, and Pmax = 48.8%; Zvorsky et al.). Therefore, it can be difficult to interpret negative findings related to those indices. One tact researchers use is to combine indices into a small number of meaningful conceptual units, was done through factor analysis (e.g., O’Connor et al., 2016). These factors are amplitude (demand intensity) and persistence (breakpoint, alpha, Pmax, and Omax), which were both associated with nicotine dependence and other smoking outcomes and correlates.

We assessed test-retest reliability over two distinct four-week periods and found that each of the CPT indices had moderate-to-excellent reliability. The present findings are consistent with previous reports of the reliability of the CPT (Few et al., 2012; Wilson et al., 2016). According to previous research using the alcohol purchase task, demand intensity and Omax had the highest reliability scores among demand indices (Murphy et al., 2011). ICCs for all demand indices were fairly high in the present study and no demand index appeared to be more reliable than the others. The present study provides additional evidence on the temporal stability of the CPT and extends previous research to situations in which participants complete the CPT related to study cigarettes.

Because study cigarettes were provided for free, we closely examined the relation between demand intensity and the number of free cigarettes consumed during the first day of the week after completing the CPT. In Week 1, mean demand intensity was 5.1 (± 5.3) cigarettes higher than smoking rate. In Week 8 demand intensity was 2.1 (± 5.5) cigarettes higher than smoking rate. Consistent with previous research (Nighbor et al. 2020), participants often overestimated demand for free cigarettes. Nighbor et al. found that demand intensity was 4.4 cigarettes per day higher on average compared to the smoking rate of free study cigarettes. Behavioral economists have described various decision-making biases (e.g., Kahneman, 2011). One such bias is the projection bias, in which people forecast future needs based upon current emotions and values (Loewenstein et al., 2003). For example, if someone is ravenous they may say “I can eat a whole pizza” thinking that they will feel the same way they currently do after consuming seven pieces of pizza. In fact, they do not feel the same after eating a few slices and don’t consume the whole pizza. Colloquially, the phrase “your eyes are bigger than your stomach” is used to describe this response pattern. This bias has been observed in a wide range of situations, including situations in which participants predict how they would feel about medical diagnoses and job changes (Loewenstein et al.). When assessing correspondence between demand intensity and cigarette consumption the following day, participants over-estimated on 43 of 72 opportunities (59.7%), demonstrating this response pattern. However, after continued exposure to free cigarettes participants demand intensity more closely matched consumption, although demand was still overestimated. The present study replicates and refines observations made by Nighbor et al. by identifying that participants become more accurate in predicting consumption after experience with free cigarettes.

A particularly gainful use of the CPT is to measure nicotine dependence severity when participants are provided very low nicotine content (VLNC) cigarettes. Studies have found that dependence severity decreases when participants smoke VLNC cigarettes over an extended period of time, implicating the substantial health benefits of a nationwide nicotine reduction policy (Donny et al., 2015). Researchers were concerned that such a policy may not be effective with more vulnerable cigarette users, so they used the CPT to determined whether dependence severity decreased among vulnerable populations that were provided VLNC cigarettes (Higgins et al., 2017; Higgins et al., 2020). Data of such studies indicate that vulnerable cigarette users will benefit from such a policy because both rate of smoking and dependence severity decreased among people with affective disorders, people with low educational attainment, and people diagnosed with opiate use disorder.

A few limitations of the present study merit mention. We did not directly observe participant smoking. It is possible that the data on cigarette smoking reported on the TLFB was not accurate. We suspect the data are accurate, though, because there is abundant evidence that the TLFB is reliable and valid (Brown et al., 1998; Sobell & Sobell, 1992; Sobell et. al., 1988; Weinhardt et al., 1998). In comparison, Nighbor et al. (2020) used a daily telephone use system in which participants called in to report their daily smoking behaviors. A second limitation is the small number of participants that completed the study, which reduced our power to detect effects. We were relying on repeated measures to generate additional datapoints rather than larger numbers of participants. Furthermore, other studies of reliability have had small numbers of participants (e.g., Few et al., 2012).

In conclusion, the accomplishments of the study include verifying the validity of the CPT over an expended period of time, in the context of a study in which participants received free study cigarettes. The CPT continues to be utilized in studies in tobacco regulatory science (in which participants smoke free study cigarettes for an extended period of time) and has been recommended for use in such studies so the validity of this measure should be established in this context (Berman et al., 2018; Donny et al., 2015). Secondly, we added to the small number of studies on the reliability of the CPT, and did so over a longer period of time. Finally, we demonstrated that people become more accurate predicting their cigarette consumption after repeated exposure to free study cigarettes and the CPT. The ability to better predict consumption on the CPT appeared to generalize to usual brand cigarettes, as well. Researchers should be aware that responses to the CPT may be changing with repeated exposure, and so comparisons that rely on repeated measures within participants may be problematic until this effect is better understood. Nonetheless, we established that hypothetical consumption on the CPT corresponded with actual consumption even though these responses were changing over time.

Public Significance.

Results provided evidence of the construct validity and temporal stability of the hypothetical cigarette purchase task while participants were receiving free study cigarettes.