Development and Initial Validation of a Momentary Cannabis Craving Scale Within a Homogeneous Sample of U.S. Emerging Adults

Abstract

Given the popularity and ease of single-item craving assessments, we developed a multi-item measure and compared it to common single-item assessments in an ecological momentary assessment (EMA) context. Two weeks of EMA data were collected from 48 emerging adults (56.25% female, 85.42% White) who frequently used cannabis. Eight craving items were administered, and multilevel factor analyses were used to identify the best fitting model. The resulting scale’s factors represented purposefulness/general desire and emotionality/negative affect craving. Convergent validity was examined using measures of craving, cannabis use disorder symptoms, frequency of use, cannabis cue reactivity, cannabis use, negative affect, and impulsivity. The scale factors were associated with cue-reactivity craving, negative affect, impulsivity, and subfactors of existing craving measures. For researchers interested in using a single item to capture craving, one item performed particularly well. However, the new scale may provide a more nuanced assessment of mechanisms underlying craving.

As rates of cannabis use continue to increase within a shifting legal and social landscape (Compton et al., 2019; Hamilton et al., 2019; Hammond et al., 2021), there is a growing need to better assess contributors to use and predictors of the development of cannabis use disorder (CUD; American Psychiatric Association, 2013). Craving, the subjective experience of wanting to use a drug (Tiffany & Wray, 2012), is one such phenomenon that has been widely recognized as an important factor in theoretical models of the development of substance use disorder (Koob & Moal, 1997; Robinson & Berridge, 1993; Solomon, 1980) and of relapse prevention (Witkiewitz & Marlatt, 2004). The relative importance of craving is further emphasized by the fact that craving is 1 of 11 criteria for assessing the presence of a CUD (American Psychiatric Association, 2013). Therefore, craving arguably represents one of the most central constructs in both clinical and research settings for understanding substance use and disorder (Gauld et al., 2023).

Despite its prominence in virtually all theories of addiction, the measurement of craving has been plagued by a number of concerns, including issues regarding how craving is defined and whether craving is best understood in terms of between- or within-person variability (Drummond, 2001; Tiffany & Wray, 2012). An additional consideration is the number of items needed to assess craving validly and reliably. For example, until craving questionnaires began to be developed in the 1990s (Anton et al., 1995; Tiffany & Drobes, 1991), most studies utilized a single item to measure levels of craving. Within these single-item measures, craving was typically defined narrowly as the urge, desire, want, or need to use a substance (McCusker & Brown, 1990); in other cases, participants were simply asked to rate their level of craving with no definitions provided (Childress et al., 1987; Ludwig, 1974; Voris et al., 1991). As multi-item craving questionnaires gained ground in the literature, some researchers raised new concerns that these longer measures confounded craving with other constructs, such as drug expectancies, and as such, were not pure measures of craving (Kozlowski et al., 1996). An item within these measures might refer to a desire to use a substance to produce a change in emotional state, thus confounding a desire to use with a specific motive or expectancy regarding use. In critiquing the broader measures, some researchers posited that craving, unlike many other constructs, could be accurately and adequately assessed using relatively few items or perhaps even a single item (Kozlowski et al., 1996). This debate about whether measures of craving should be restricted to items assessing desire to use only or whether the inclusion of related constructs, such as expectancies and intentions to use, provides additional utility continues to be an important one, as popular measures of cannabis craving do include items that capture related constructs when assessing one’s desire to use (Heishman et al., 2009).

With the increased popularity of ecological momentary assessment (EMA) studies, researchers can address some of the limitations of between-person measurement of craving by capturing the dynamic, within-person variability of craving in daily life (Shiffman, 2009). Several findings suggest there may be greater clinical utility for in-the-moment measures than for more distal measures of craving (Enkema et al., 2020; Ramirez & Miranda, 2014; Serre et al., 2015). These studies suggest that momentary measures of craving are more strongly associated with use (Enkema et al., 2020; Ramirez & Miranda, 2014) and relapse (Serre et al., 2015) than are more distal measures of craving. Thus, momentary measures of cannabis craving may help better elucidate links between craving and relapse, substance use, and substance use disorder. Yet, EMA studies may be limited, as a single item remains almost exclusively the method of choice for assessing craving within an EMA framework (Buckner, Silgado, & Schmidt, 2011; Buckner, Zvolensky, et al., 2011; Emery et al., 2021; Enkema et al., 2021; Wycoff et al., 2023). Single-item measures are frequently used in EMA, despite their psychometric limitations, due to the intensive assessment and participant burden of EMA studies. The specific item used for assessing craving also varies across studies, which may hamper efforts to replicate effects, as these varying assessment methods likely capture different facets of craving and may have distinct associations with related constructs. Identifying whether associations do indeed differ across single-item craving assessments would better inform researchers’ decisions regarding their measurement of craving. Given the importance and centrality of craving for understanding substance use (Gauld et al., 2023), adequate and reliable measurement is critical.

Currently, there is no empirically validated scale of momentary cannabis craving for use in EMA studies. In addition, to our knowledge, no research has examined the practical gains to be achieved by multi-item craving questionnaires in comparison to single-item measures within EMA designs, which is particularly relevant for EMA studies where participant burden is of concern. Given that there is historical debate about the necessity of broader questionnaire measures of craving (Kozlowski et al., 1996), research is needed to evaluate a momentary measure of cannabis craving alongside commonly used single items.

Current Study

The current study represents the first effort to develop a brief measure of within- and between-person variability in cannabis craving for future use in EMA and daily diary studies. We made use of multilevel factor analysis (MFA) to model both within- and between-person variability in cannabis craving within an EMA protocol conducted among emerging adults. First, we examined the relative extent of within- and between-person variance in each cannabis craving item and then conducted a series of exploratory MFAs. These analyses estimated the factor structure for within- and between-person levels of cannabis craving; model fit was evaluated using several indicators. Next, we evaluated convergent validity of the factors by examining their between-person associations with a baseline measure of cannabis craving (Marijuana Craving Questionnaire-Short Form [MCQ-SF]; Heishman et al., 2009), the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association, 2013) CUD craving criterion, CUD symptom counts, and past 30-day frequency of cannabis use. At the within-person level, we examined convergent validity in association with momentary cannabis cue-reactivity craving, cannabis use, negative affect/stress, and impulsivity. We evaluated associations with past month and momentary cannabis use because some (but not all; Emery et al., 2021) EMA studies have shown that cannabis craving is related to cannabis use (Serre et al., 2015). Negative affect was evaluated because of (a) its proposed role in craving of alcohol and other substances (Baker et al., 1986, 2004; Stasiewicz & Maisto, 1993) and (b) popular craving measures often include subscales assessing negative affect-related (or relief) craving (Heishman et al., 2001, 2009). Finally, we examined relations to impulsivity because research suggests craving may mediate associations between impulsivity and substance use (Coates et al., 2020; Flaudias et al., 2019), and individuals who score higher on measures of impulsivity also tend to score higher on measures of craving (Joos et al., 2013). However, within-person associations and the role of impulsivity in cannabis use and craving are less studied (Rinehart & Spencer, 2021).

Given the ease and popularity of using a single item to assess craving within EMA studies, any longer measure of cannabis craving may need to demonstrate stronger convergent validity to receive widespread use. Therefore, we sought to informally compare the convergent validity of the momentary cannabis craving scale to that of each of the eight single-cannabis craving items. Standardized values are reported to permit comparison of effect magnitude across items and scales. This comparative analysis of each item also allowed us to examine whether associations differed depending on the single item chosen, which would have implications for researchers’ selection of a single-item cannabis craving measure.

Methods

Participants

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. Study participants were 48 emerging adults aged 18–21 years (M = 19.58, SD = 1.03; 56.25% female; 85.42% White) from the Southeastern United States who reported using cannabis at least three times per week over the past 30 days. Data were collected from participants between 2019 and 2021. To determine that cannabis use eligibility criteria were met, participants completed an oral fluid cannabinoid test (Δ⁹-tetrahy drocannabinol [THC]; Forensic Fluids Laboratories) or provided urine sample. Study exclusion criteria were (a) planning to quit or reduce cannabis use within the next month, (b) the presence of a severe alcohol or opioid use disorder, (c) current enrollment in treatment for substance use, (d) being pregnant or lactating, or (e) having any severe condition that could interfere with the ability to complete study procedures. Given that several cannabis craving items were added after study enrollment began, only participants who enrolled after all craving items were included were used in the analyses, resulting in the aforementioned sample size of 48 individuals.

Procedure

Participants were recruited for the study using community and clinic advertisements, including flyers at local medical clinics, restaurants, gyms, and coffee shops, as well as internet and social media posts. Individuals who were interested in participating completed a phone screen, and those who were deemed potentially eligible were invited to schedule an initial visit to complete baseline measures (see Figure 1 for study procedure). Informed consent was obtained from all study participants, and study procedures were approved by the institutional review board. To confirm recent cannabis use for inclusion purposes, participants either submitted a urine sample in-person or returned an oral fluid sample by mail (if participating fully remotely due to the COVID-19 pandemic). For participants returning oral fluid samples by mail, sample collection was video observed to ensure specimens belonged to the participant. Relevant to the current study, the baseline assessment included measures of demographics, typical cannabis use methods (e.g., leaf/bud, concentrates/oils, or edibles/drinks/tinctures), past 30-day cannabis use, CUD symptoms, and baseline levels of cannabis craving (via the MCQ-SF; Heishman et al., 2009). At the baseline session, participants were provided with an orientation to the EMA protocol and selected four 2-hour time blocks during which they would be able to answer EMA prompts each day. Weekday and weekend time blocks could vary. EMA prompts were completed across 14 days using an iOS app (see Wray et al., 2015). EMA sessions assessed current levels of cannabis craving using eight items. Craving was assessed before and after presentation of cannabis cues using a Cue Reactivity Ecological Momentary Assessment (CREMA) methodology (Warthen & Tiffany, 2009). Craving assessed prior to cue exposure was used for measure development to ensure that our measure captured naturalistic craving rather than cue-elicited craving. Cue-elicited craving was used as one validation metric to evaluate whether craving scores increased following exposure to cannabis cues. Additional study procedures have been described elsewhere (Gex et al., 2022). This study was not preregistered.

Figure 1.

Study Protocol and Procedures.

Note. Protocol differed slightly for those participants who completed the study entirely remotely. For these participants, a sample of saliva was provided at screening, and as a result, the screening and orientation sessions took place on different days to allow time for the saliva sample to be returned and tested. In addition, remote participants did not provide a bioassay on Day 15.

Measures

Cannabis Craving Items¹.

Momentary cannabis craving was assessed within EMA sessions prior to cannabis cue exposure using eight items (see Table 1 for questions). These items were selected from related measures or previous EMA studies assessing momentary craving. Five of the items were selected from the MCQ-SF, with three items coming from the emotionality subfactor and two items from the purposefulness subfactor (Heishman et al., 2009). MCQ-SF items that reflected momentary or transient states of craving/desire to use cannabis were prioritized with an effort to capture potentially lower-intensity craving (e.g., “It would be great to smoke cannabis right now.”), as well as emotionally-valenced craving. Three additional items were added from other sources to capture higher-intensity craving. One item (“I am craving cannabis right now.”) was from a previous study of within-person cannabis craving (Buckner, Silgado, & Schmidt, 2011), which found that it was highly correlated with scores on the MCQ (Heishman et al., 2001). Another item (“I have a desire to use cannabis.”) was adapted from an item used in a prior cue reactivity study and was found to be associated with cannabis cue reactivity (Gray et al., 2008). Finally, one item (“Nothing would be better than using cannabis right now.”) was modified from the Questionnaire on Smoking Urges-Brief used for assessing tobacco craving (Cox et al., 2001). All cannabis craving items were assessed on a seven-point Likert-type scale ranging from “Strongly disagree” to “Strongly agree.”

Table 1.

Sample Characteristics.

Variable	% (N) or M (SD)
Age	19.58 (1.03)
Gendera
Cisgender woman	56.25% (27)
Cisgender man	43.75% (21)
Race
White	85.42% (41)
Asian/Asian American	8.33% (4)
Black/African American	6.25% (3)
Ethnicity
Non-Hispanic/Latinx	89.58% (43)
Hispanic/Latinx	10.42% (5)
University student	83.34% (40)
Typical mode of cannabis use
Bongs	27.08% (13)
Blunts	20.83% (10)
Other (e.g., gravity bongs, oil cartridges)	18.75% (9)
Wax	12.50% (6)
Bowl/pipe	10.42% (5)
Joints	6.25% (3)
Vaporizers	4.17% (2)
Past-year cannabis use disorder	89.36% (42)
Age of first cannabis use	15.79 (1.57)
Past 30-day cannabis use (# days)	25.56 (6.69)
No. of cannabis uses on typical day	2.95 (2.00)
Average monthly spending on cannabis ($)	164.42 (149.03)
Past 30-day cigarette use (# days)	2.61 (5.93)
Past 30-day e-cigarette use (# days)	20.29 (12.67)
Past 30-day alcohol use (# days)	8.41 (5.62)

^aBoth sex at birth and gender identity were assessed, and for all participants, gender identity (man/woman) aligned with sex at birth (male/female).

Marijuana Craving Questionnaire-Short Form.

At baseline, participants completed the MCQ-SF, which is a 12-item measure of cannabis craving with four factors corresponding to compulsivity, emotionality, expectancy, and purposefulness craving. The MCQ-SF has been empirically validated among adult cannabis users (Heishman et al., 2009). Response scales for each item ranged from 1 (strongly disagree) to 7 (strongly agree).

Past 30-Day Cannabis Use.

Past 30-day cannabis use was assessed using a timeline follow back (TLFB) interview (Hjorthøj et al., 2012; Sobell & Sobell, 1992). For each day, participants were asked whether they had used cannabis. If a participant reported use on a given day, they were asked additional details, including their method (e.g., leaf/bud, concentrates/oils, or edibles/drinks/tinctures) and estimated amount of use (i.e., in grams) on that day.

Cannabis Use Methods.

Within the TLFB interview for cannabis use, participants reported their cannabis use methods in the past 30 days. In the baseline assessment, they also reported their most typical method of use during the past 30 days, during their period of heaviest cannabis use, and during their lifetime. For the purposes of the current study, we utilized the measure of typical cannabis use methods in the past 30 days, as this was deemed to be most relevant for evaluating the performance of the EMA craving items. Response options were joint, bowl/pipe, blunt (i.e., cannabis rolled in tobacco paper or wraps), bong, vaporizer, in food/edibles, in beverages, hookah, spliff (i.e., a joint rolled with a mixture of cannabis and tobacco), wax, other, or a combination of methods.

Cannabis Use Disorder Symptoms.

At baseline, each individual criterion for CUD was assessed (American Psychiatric Association, 2013) via the Mini International Neuropsychiatric Interview (MINI; Sheehan et al., 1998) adapted for DSM-5. Symptom counts were calculated by summing the number of symptoms endorsed for each participant. Participants who endorsed two or more of the 11 symptoms within the past year were identified as having a current CUD.

Cannabis Cue Reactivity.

Cannabis cue reactivity was assessed as the total change in craving after viewing a cannabis cue image within EMA sessions. To calculate this, pre-cue exposure craving ratings were subtracted from post-cue craving ratings and were then summed across items to result in total cue reactivity. Cannabis cue images were from a previously validated cannabis cue reactivity paradigm (Karoly et al., 2019) and involved images of cannabis flower, paraphernalia such as bongs and pipes, and joints/blunts; some images showed cannabis being smoked, while others were passive in nature. Cannabis cue reactivity was only used in analyses evaluating within-person convergent validity.

Momentary Cannabis Use.

Between random EMA prompts, participants were asked to initiate a report if they used cannabis. In addition, at each EMA prompt, participants were asked whether they had forgotten to log any cannabis use since the last report. If participants logged any cannabis use between EMA assessments and reported unlogged use, momentary cannabis use was coded ‘1’; otherwise, momentary use was coded “0.” Craving scores at one timepoint (e.g., T1) were assessed in relation to momentary use reported after T1 but before T2 (i.e., via participant-initiated logs) or at the next assessment (i.e., T2 reports of unlogged use since T1).

Momentary Impulsivity.

Momentary impulsivity was measured in EMA sessions via the Momentary Impulsivity Scale (Tomko et al., 2014), which is a four-item scale that is correlated with common trait impulsivity measures. Participants were asked to consider their behavior since the last prompt and indicate the extent to which they had (a) said things without thinking, (b) spent more money than they meant to, (c) felt impatient, and (d) made a “spur of the moment” decision.

Momentary Negative Affect/Stress.

A sum score was created from four items that were used to assess momentary negative affect/stress at each random prompt (referred to throughout as “negative affect”). Participants were asked to indicate how much they were feeling each of the following right now: (a) stressed, (b) frustrated, (c) relaxed (reverse-coded), and (d) sad.

Data Analysis

As an initial step, we calculated the proportion of between- and within-person variability for each item (see Table 2). Between-person variability reflects differences across participants in levels of cannabis craving, while within-person variability reflects state-related changes in cannabis craving. Next, exploratory MFA was conducted within Mplus version 7.4 (Muthén & Muthén, 2015). MFA can effectively model both within- and between-person variability in cannabis craving, a difference which is largely ignored in traditional single-level factor analysis approaches (Muthen, 1991). Utilizing the multilevel structure of EMA data, MFA allows for determination of the optimal number of factors at the within- and between-person levels. Similarly, factor scores are calculated at both levels of the data. Model fit was compared using the root mean square error of approximation (RMSEA, with good fit indicated by values <0.05; Steiger, 1990), standardized root mean squared residual (SRMR, with good fit indicated by values <0.08; Maydeu-Olivares, 2017), comparative fit index (CFI, with good fit indicated by values >0.95; Bentler, 1990), and Tucker-Lewis index (TLI, which good fit indicated by values >0.95; Tucker & Lewis, 1973). In this process of evaluating models, we also planned to remove any uninformative items that had weak loadings (i.e., ⩽ 0.40) or that generalized across factors (i.e., cross-loading ⩾ 0.40 on multiple factors; Matsunaga, 2010).

Table 2.

Item-Level Descriptive Statistics and Intraclass Correlation Coefficients.

Itema	M (SD)	ICC	Skewness	Kurtosis
1. “It would be great to smoke cannabis right now.”	4.01 (2.01)	0.57	0.03	−1.16
2. “I would feel more in control of things right now if I could smoke cannabis.”	3.03 (l.98)	0.64	0.62	−0.84
3. “Smoking cannabis would be pleasant right now.”	4.16 (2.05)	0.56	−0.13	−1.20
4. “I have a desire to use cannabis right now.”	3.82 (2.05)	0.51	0.11	−1.26
5. “If I smoked cannabis right now, I would feel less tense.”	3.40 (2.05)	0.65	0.36	−1.16
6. “I am craving cannabis right now.”	3.53 (2.03)	0.58	0.30	−1.14
7. “I would feel less anxious if I smoked cannabis right now.”	3.34 (2.03)	0.65	0.41	−1.09
8. “Nothing would be better than using cannabis right now.”	3.15 (2.02)	0.60	0.58	− 0.94

Note. Mean refers to the grand mean across all administrations and participants. SD = standard deviation, ICC = intraclass correlation coefficient, representing the proportion of between-person variance in items.

^aItems administered to participants used the term marijuana rather than cannabis, but we do not retain this language here due to the term’s racist origins.

Multilevel structural equation modeling (SEM) was then used to evaluate the convergent validity of the factor scores. We examined associations between the factor scores and baseline cannabis craving (as measured by the MCQ-SF), past 30-day frequency of cannabis use, the presence/absence of the CUD craving criterion, and CUD symptom counts. To validate the within-person factor scores, we examined associations with momentary cannabis cue-reactivity craving, cannabis use, negative affect, and impulsivity. For comparison, each individual item was also evaluated in association with each of the validation items. We used standardized effect sizes to compare performance across models, and we evaluated how much variance in items and factors was explained by the associated traits at both the within- and between-person levels. For all analyses in Mplus, full information maximum likelihood was used to handle missing data.

Because our sample size was limited, we also considered the effect sizes of our results in comparison with previous studies that validated cannabis craving measurement (Heishman et al., 2009; Heishman et al., 2001). Regarding current craving, β ranged from 0.10 to 0.45 for MCQ-SF factor scores; for past month craving, β ranged from 0.14 to 0.31 (Heishman et al., 2009). For the long-form MCQ, associations with positive mood were equivalent to β (Peterson & Brown, 2005) ranging from −0.23 to −0.17, and past month cannabis use β ranged from 0.22 to 0.43 (Heishman et al., 2001). Thus, where possible, we compare our findings considering this existing research. Analysis code for this study is provided in the Supplemental Material.

Results

Descriptive Statistics

Slightly more than half (56.25%) of the sample was cis-gender female, and the remaining proportion were cis-gender male (43.75%; see Table 1 for sample characteristics). Participants predominantly identified as White (85.42%), with 8.33% identifying as Asian, and 6.25% identifying as Black. In terms of ethnicity, 10.42% of participants were Hispanic or Latinx. The average age of first cannabis use was around 16 years (M = 15.79, SD = 1.57). On average, participants in the sample reported using cannabis on 25.56 days in the last 30 days (SD = 6.69). Within the past month, the most common cannabis use methods were bongs (27.08%), blunts (20.83%), other (18.75%; follow-up responses indicated that “other” often referred to gravity bongs or oil cartridges), wax (12.50%), and bowl/pipe (10.42%). Most of the sample (89.36%) met criteria for a current CUD. Co-use of alcohol and tobacco was common (see Table 1 for frequency of past 30-day use).

There was a total of 1948 completed EMA reports for the 48 participants, with each participant completing about 40 reports on average (M = 40.58; average compliance with random reports = 72.5%). Compliance rates were higher than that observed in other substance-use disorder EMA studies (i.e., 69.80%; Jones et al., 2019). On average, the first random prompt of the day was completed at 10:51 (range from 3:31 to 16:10), the second random prompt at 13:58 (range from 10:02 to 18:12), the third random prompt at 17:18 (range from 11:47 to 21:22), and the fourth random prompt at 20:21 (range from 14:24 to 23:43). Within-person variance in individual items ranged from 35% to 49%, with the remaining variance due to between-person differences (see Table 2).

Multilevel Factor Analysis

Given that guidelines typically recommend that factors have no fewer than three items (Albano, 2020), an initial set of four multilevel EFA models, ranging from one to two between- and within-person factors, were fit including all eight cannabis craving items. An oblique (i.e., Geomin) rotation was used. Models that specified a single within-person factor fit especially poorly and models with a single between-person factor fared somewhat better (see Table 3 for model fit). However, the best fitting model specified two factors at both the between-person and within-person levels. Although the model fit of a two within-two between-factor model was adequate, χ² (df) = 326.23(26), RMSEA = 0.077, SRMR (within/between) = 0.018/0.031, CFI = 0.976, TLI = 0.948, two items were deemed to be poor indicators. Item 6 (“I am craving cannabis right now.”) loaded onto different factors at the within- and between-person levels, and Item 8 (“Nothing would be better than using cannabis right now.”) cross-loaded onto both factors at the within- and between-person levels. Based on these statistical issues, Items 6 and 8 were dropped from the scale, and a subsequent series of MFAs was conducted including the remaining six items.

Table 3.

Model Fit Statistics From Multilevel Factor Analysis.

8-Item scale models
Model (factors)	χ2(df)	RMSEA	SRMR (within/between)	CFI	TLI
1 Within - 1 Between	1521.82 (40)	0.138	0.058/0.096	0.882	0.834
2 Within - 1 Between	482.06 (33)	0.084	0.018/0.087	0.964	0.939
1 Within - 2 Between	1359.02 (33)	0.144	0.057/0.032	0.894	0.820
2 Within - 2 Between	326.23 (26)	0.077	0.018/0.031	0.976	0.948
6-Item scale models
Model (factors)	χ2 (df)	RMSEA	SRMR (within/between)	CFI	TLI
1 Within - 1 Between	1290.09 (18)	0.19	0.073/0.091	0.848	0.747
2 Within - 1 Between	217.33 (13)	0.090	0.007/0.081	0.976	0.944
1 Within - 2 Between	1091.43 (13)	0.206	0.073/0.008	0.871	0.703
2 Within - 2 Between	23.862 (8)	0.032	0.007/0.007	0.998	0.993

Note. Bold indicates optimal model.

In MFAs including the remaining six items, the two within-two between-factor model was the best fitting one across all fit indices, see Table 3; χ² (df) = 23.862(8), RMSEA = 0.032, SRMR (within/between) = 0.007/0.007, CFI = 0.998, TLI = 0.993. Model fit values indicated that this was an excellent fitting model for the data. Within this model, Items 1, 3, and 4 loaded highly onto one factor (purposefulness/general desire to use), and Items 2, 5, and 7 loaded highly onto a second factor (negative affect/emotionality) at both the within- and between-person levels (see Table 4). The two factors were highly correlated at both the within-person (r = 0.744, SE = 0.017) and between-person (r = 0.788, SE = 0.057) levels.

Table 4.

Standardized Factor Loadings (Standard Errors) of the Optimal Model.

	Within-person		Between-person
Itema	Factor 1	Factor 2	Factor 1	Factor 2
1. “It would be great to smoke cannabis right now.”	0.851 (0.009)	0.007 (0.002)	1.004 (0.031)	−0.006 (0.035)
2. “I would feel more in control of things right now if I could smoke cannabis.”	0.104 (0.030)	0.655 (0.028)	0.086 (0.128)	0.782 (0.112)
3. “Smoking cannabis would be pleasant right now.”	0.934 (0.028)	−0.068 (0.033)	0.977 (0.025)	0.003 (0.027)
4. “I have a desire to use cannabis right now.”	0.777 (0.026)	0.117 (0.030)	0.712 (0.066)	0.300 (0.072)
5. “If I smoked cannabis right now, I would feel less tense.”	−0.001 (0.017)	0.895 (0.020)	0.002 (0.011)	1.006 (0.011)
7. “I would feel less anxious if I smoked cannabis right now.”	−0.002 (0.019)	0.877 (0.021)	− 0.010 (0.026)	0.996 (0.023)

Note. Items 6 (“I am craving cannabis right now”) and 8 (“Nothing would be better than using cannabis right now”) were removed due to cross-loading onto both factors. Bold indicates that the item was retained on that factor. Factor 1 refers to purposefulness/general craving, and Factor 2 refers to negative affect/emotionality craving.

^aItems administered to participants used the term marijuana rather than cannabis, but we do not retain this language here due to the term’s racist origins.

Within-Person Convergent Validity

Momentary cannabis use was not associated with either factor scores or any individual item. Higher cannabis cue reactivity (i.e., greater change in craving in response to cannabis cues) was associated with both factor scores and all eight individual items, although the association was strongest for Factor 1 (purposefulness/general craving; β = 0.50 [0.07]). Negative affect was associated with Factor 2 (negative affect/emotionality craving; β = 0.17 [0.07]), as well as Items 5 (β = 0.15 [0.06]) and 7 (β = 0.16 [0.06]), both of which loaded onto Factor 2. Both factors and all items (except for Item 4) were significantly associated with momentary impulsivity, with the association being strongest for both factors and Item 1 (all β = 0.21). The momentary measures accounted for 28.8% of the variance in Factor 1 and 21.9% of the variance in Factor 2 scores. Variance accounted for in items ranged from 16.4% (for Item 2) to 24.6% (for Item 1).

Between-Person Convergent Validity

Baseline scores on the MCQ-SF Purposefulness subscale were associated with Factor 1 (β = 0.44 [0.18]), which reflected purposefulness/general craving. Items 1 (β = 0.47 [0.18]), 3 (b = 0.40 [0.19]), 4 (β = 0.45 [0.17]), 6 (β = 0.46 [0.15]), and 8 (β = 0.34 [0.15]) were also associated with MCQ-SF Purposefulness. Surprisingly, MCQ-SF Compulsivity scores were negatively associated with factor scores and with all items. MCQ-SF Emotionality was associated with Factor 2 scores (β = 0.37 [0.18], reflecting negative affect/emotionality craving). Items 5 (β = 0.44 [0.16]), 6 (β = 0.30 [0.18]), and 7 (β = 0.35 [0.17]) were also significantly associated with MCQ-SF Emotionality. Factor 1 was uniquely associated with MCQ-SF Expectancy (β = 0.39 [0.20]). Items 4 (β = 0.20 [0.10]), 6 (β = 0.34 [0.17]), and 8 (β = 0.39 [0.12]) were significantly associated with endorsement of the CUD craving criterion; though not significant, the magnitude of effect for Factor 2 was similar (β = 0.22 [0.20]). Factors and items were not related to past 30-day frequency of cannabis use or to CUD symptom counts (see Table 5).

Table 5.

Associations With Validation Measures.

Within person	Factor 1	Factor 2	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8
Cannabis use	0.01	−0.01	−0.01	−0.03	0.00	0.01	−0.02	− 0.01	−0.02	0.01
Cue reactivity	0.50	0.34	0.45	0.34	0.46	0.46	0.36	0.44	0.35	0.43
Negative affect	0.00	0.17	−0.02	0.05	−0.03	0.04	0.15	0.05	0.16	0.03
Impulsivity	0.21	0.21	0.21	0.19	0.16	0.09	0.14	0.12	0.16	0.13
Between person
MCQ-SF Purposefulness	0.44	0.22	0.47	0.18	0.40	0.45	0.23	0.46	0.21	0.34
MCQ-SF Compulsivity	−0.27	−0.30	−0.26	−0.19	−0.28	−0.34	−0.33	− 0.25	−0.30	− 0.29
MCQ-SF Emotionality	−0.07	0.37	−0.12	0.32	0.00	0.28	0.44	0.30	0.35	0.26
MCQ-SF Expectancy	0.39	−0.03	0.38	−0.06	0.36	−0.02	−0.10	−0.32	−0.02	− 0.13
Past 30-day cannabis use	−0.09	−0.06	−0.06	−0.05	−0.11	−0.06	−0.06	0.05	−0.04	0.06
CUD craving criterion	0.08	0.22	0.08	0.20	0.10	0.20	0.23	0.34	0.23	0.39
CUD symptom count	0.01	0.11	0.04	0.16	−0.01	0.11	0.13	0.17	0.11	0.22

Note. Bold indicates significance at p < .05. Factor 1 refers to purposefulness/general craving, and Factor 2 refers to negative affect/emotionality craving. MCQ-SF = Marijuana Craving Questionnaire-Short Form, CUD = cannabis use disorder.

Discussion

Within a sample of frequent cannabis-using emerging adults, a two-factor measure of momentary cannabis craving was the best fitting model at both the within- and between-person levels. The resulting two factors each comprised three items and appeared to reflect a purposefulness/general desire to use subscale and an emotionality/negative affect subscale. These subscales are consistent with factors previously identified in existing popular measures of cannabis (Heishman et al., 2001, 2009) and cigarette (Tiffany & Drobes, 1991) craving; furthermore, they align well with models of reward and relief craving (Baker et al., 1986; Wise, 1988). Thus, the new scale fits within existing theoretical frameworks of craving and supports the existence of multiple dimensions underlying momentary craving.

Overall, the momentary cannabis craving measure appeared to have adequate convergent validity. Most associations were in line with expectations based on each factor’s representation and previous measure development research (Heishman et al., 2001, 2009), although neither factor was related to cannabis use. Consistent with the second factor representing emotionality craving, it was uniquely related to momentary negative affect and baseline MCQ-SF Emotionality subscores. Although not significant, emotionality craving also had stronger associations with CUD symptom counts and endorsement of the CUD craving criterion than purposefulness craving. On the other hand, purposefulness craving was more strongly related to cue-reactivity craving and MCQ-SF Purposefulness and Expectancy baseline scores than emotionality craving. These two facets of craving, while highly correlated, also had distinct associations. Therefore, we present initial convergent validity for a novel, momentary measure of cannabis craving for use in EMA and daily diary studies, with items largely derived from an existing craving measure, the MCQ-SF.

For the individual items, associations with validation measures were more varied. Just one item, “I am craving cannabis right now,” (Buckner, Silgado, & Schmidt, 2011) was associated with both MCQ-SF Purposefulness and Emotionality scores, as well as CUD craving endorsement. The item also demonstrated associations with impulsivity and cue-reactivity craving (but not negative affect). For researchers interested in using a single item to reliably capture facets of both purposefulness and emotionality cannabis craving, this item would provide the best fit of those we evaluated. On the other hand, the new six-item scale may provide additional utility for assessing mechanisms underlying craving, including negative affect and cue reactivity (Tiffany & Wray, 2012). For example, of all the items and factors, Factor 1 (purposefulness/general desire to use) had the strongest association with cue reactivity and MCQ-SF Expectancy scores, Factor 2 (emotionality/negative affect) with negative affect, and both factors with impulsivity. When selecting a method for assessing cannabis craving, researchers should reflect on their study’s unique needs and considerations.

Importantly, the varied associations observed across items and factors suggest a need to incorporate more standardized measures of cannabis craving in EMA research, regardless of whether that ultimately means selecting a widely accepted single item or making use of this (or another future) multi-item craving scale. At present, the inconsistency in measurement of cannabis craving across studies may hamper efforts at producing replicable science. Researchers who select a particular item to measure craving may unknowingly be capturing different facets of craving compared with other studies upon which their work builds. Since craving continues to be recognized as a central tenet of substance use disorders and related theories (Gauld et al., 2023; Witkiewitz & Marlatt, 2004), increased attention to its reliable and standardized assessment is warranted.

Limitations

An important limitation of the current study is the use of a predominantly White sample. Compared to the broader community (U.S. Census Bureau, 2022), the current sample had a higher proportion of White individuals (85.42% compared to 74.1% in the community) and a lower proportion of Black/African American individuals (6.25% vs. 19.6% in the community). Although the current sample did include higher rates of Asian Americans (8.33% vs. 1.8%) and Hispanic/Latinx individuals (10.42% vs. 4.2%) than the surrounding community, the relative lack of diversity among participants warrants a need to evaluate the measure’s performance in other racial and ethnic groups. Similarly, the current sample consisted exclusively of emerging adults aged 18–21 years, so the momentary cannabis craving scale should be evaluated in other age groups to support its use within these populations. Given that external data were not available to conduct confirmatory factor analyses, future work will be needed to evaluate the model fit of the scale within different samples. Confirmatory work within larger samples will also be critical to examine the measure’s validity and factor structure further.

Due to the need for brevity in EMA studies, the initial item pool evaluated here was necessarily narrower in scope than many traditional, multi-item craving questionnaires, including the MCQ-SF from which several of the items were derived. Another consideration is that most of the craving items (five of eight) used the term “smoke” when referring to cannabis use, which might limit generalizability to samples where participants primarily employ other means of using cannabis, such as ingestion. Within the current sample, however, all participants endorsed that their typical method of cannabis use was via inhalation/smoking; therefore, we were not able to evaluate whether the measure’s items were differentially endorsed as a function of an individual’s typical consumption method (i.e., smoked/inhaled vs. other methods). This is consistent with national data, where among youth and young adults, smoking remains by far the most prevalent route of administration for cannabis (Wadsworth et al., 2022). Future research should explore whether using more neutral (e.g., “use cannabis”) or personalized language (based on terms preferred by the participant and/or the participant’s typical mode of use) influences endorsement of cannabis craving items. Assessments of cannabis use and craving will need to emphasize continually modernizing language as terms and forms of use are ever evolving.

Across both factors and individual items, there were no significant associations with cannabis use either in the moment or within the past month. It is possible that this lack of an association between the craving measures and cannabis use was because heavy use was so widespread among the sample. On average, participants were using cannabis about three times a day and on 26 days in the past 30 days. In such a sample, it may be difficult to capture enough variability in use behaviors to determine its associations with craving. The measure’s performance among individuals who use cannabis less frequently should be evaluated in the future. In addition, a unidimensional measure of momentary impulsivity was used for validation. Although it is well established that impulsivity is a multidimensional construct, EMA procedures require brief assessments, and the Momentary Impulsivity Scale we used has been found to be moderately correlated with subscale and total scores on the Barratt Impulsiveness Scale-11 (Patton et al., 1995) and the UPPS Impulsive Behavior Scale (Tomko et al., 2014; Whiteside et al., 2005). Thus, this measure appears to capture a broad range of impulsivity facets, but future work examining differential associations between impulsivity facets and scores on the Momentary Cannabis Craving Scale would be informative.

Finally, although “marijuana” was used in all items viewed by participants at the time this study was conducted, we have not used this term throughout the paper when referring to the items. The term marijuana in the United States is rooted in racist and anti-immigration sentiments (MacDonald, 2023), and thus, we recommend that future iterations of the scale use the term cannabis instead. Although the term marijuana remains more popular than cannabis among the public, Google search trends show that the gap between the two has been narrowing (Google Trends, accessed February 1, 2024), with cannabis overtaking marijuana in popularity in some regions. Researchers should consider the familiarity of their participants with both terms when making a suitable choice, but it is increasingly likely that individuals using cannabis will be familiar with this term.

Conclusions

Using MFA, a two-factor momentary cannabis craving scale was derived and validated for use among predominantly White emerging adults within an EMA protocol. This scale comprised a purposefulness/general desire factor and an emotionality/negative affect factor, with both present at the within- and between-person levels. These two factors, although highly correlated, had distinct associations with existing cannabis craving measure subscales and negative affect. Furthermore, the factors aligned well with existing measures of craving for other substances and with reward and relief craving theories. In comparison, associations between validation measures and single items were varied, with one item (“I am craving cannabis right now”) performing relatively well at assessing both purposefulness and emotionality craving. Researchers should consider the relative merits of assessing a broader spectrum of craving dimensions (such as reward [i.e., purposefulness] and relief [i.e., emotionality] craving) alongside the potential drawbacks of increased participant burden when deciding how to assess cannabis craving in EMA studies. Regardless, the varied associations observed across commonly used single items and the multi-item scales highlight a need for more empirical, standardized measurement of craving in future studies to increase replicability and enhance precision of findings.