the craving experience questionnaire

The Craving Experience Questionnaire: a brief,theory-based measure of consummatory desire andcraving

Jon May1, Jackie Andrade1, David J. Kavanagh2, Gerald F. X. Feeney3, Mathew J. Gullo4,5,Dixie J. Statham6, Jessica Skorka-Brown1, Jennifer M. Connolly2, Mandy Cassimatis2,Ross McD. Young3,7 & Jason P. Connor3,4,5

School of Psychology, Cognition Institute, Plymouth University, Plymouth, UK,1 Institute of Health and Biomedical Innovation and School of Psychology andCounselling,Queensland University ofTechnology, Kelvin Grove,Queensland, Australia,2 Alcohol and Drug Assessment Unit, Princess Alexandra Hospital, Brisbane,Queensland, Australia,3 Centre for Youth Substance Abuse Research, The University of Queensland and Queensland University of Technology, Brisbane,Queensland, Australia,4 Discipline of Psychiatry, School of Medicine, The University of Queensland, Brisbane, Queensland, Australia,5 School of Social Sciences,University of the Sunshine Coast, Maroochydore, Queensland, Australia6 and Faculty of Health Sciences, Queensland University of Technology, Brisbane,Queensland, Australia7

ABSTRACT

Background and Aims Research into craving is hampered by lack of theoretical specification and a plethora ofsubstance-specific measures. This study aimed to develop a generic measure of craving based on elaborated intrusion(EI) theory. Confirmatory factor analysis (CFA) examined whether a generic measure replicated the three-factor struc-ture of the Alcohol Craving Experience (ACE) scale over different consummatory targets and time-frames.Design Twelve studies were pooled for CFA. Targets included alcohol, cigarettes, chocolate and food. Focal periodsvaried from the present moment to the previous week. Separate analyses were conducted for strength and frequencyforms. Setting Nine studies included university students, with single studies drawn from an internet survey, acommunity sample of smokers and alcohol-dependent out-patients. Participants A heterogeneous sample of 1230participants. Measurements Adaptations of the ACE questionnaire. Findings Both craving strength [comparativefit indices (CFI = 0.974; root mean square error of approximation (RMSEA) = 0.039, 95% confidence interval(CI) = 0.0350.044] and frequency (CFI = 0.971, RMSEA = 0.049, 95% CI = 0.0440.055) gave an acceptable three-factor solution across desired targets that mapped onto the structure of the original ACE (intensity, imagery, intrusive-ness), after removing an item, re-allocating another and taking intercorrelated error terms into account. Similarstructures were obtained across time-frames and targets. Preliminary validity data on the resulting 10-item CravingExperience Questionnaire (CEQ) for cigarettes and alcohol were strong. Conclusions The Craving ExperienceQuestionnaire (CEQ) is a brief, conceptually grounded and psychometrically soundmeasure of desires. It demonstratesa consistent factor structure across a range of consummatory targets in both laboratory and clinical contexts.

Keywords Alcohol, assessment, chocolate, craving, food, measure, tobacco.

Correspondence to: Jon May, School of Psychology, University of Plymouth, Plymouth PL4 8AA, UK. E-mail: [email protected] 23 May 2013; initial review completed 13 August 2013; final version accepted 19 December 2013

INTRODUCTION

Intense desire or craving is recognized increasingly as acore feature of addiction, as shown by its inclusion inDSM-5 substance use disorder criteria [1] and in ICD-10[2]. While substance use does not require that intensedesires precede it [3], craving presents a cue for addictivebehaviour and a challenge for attempts at control [4].Kavanagh, Andrade & May [5] proposed Elaborated

Intrusion (EI) theory as a description of processes under-pinning desires, bringing together multiple strands ofresearch. EI theory proposes that craving to consumedrugs, food or drink and desire for other activities is acognitiveaffective phenomenon that involves an initial,apparently spontaneous intrusive thought (triggered bycues from the environment, mind and body), followedby controlled processes of elaboration, which tend toinclude construction of multi-sensory imagery [6]. These

METHODS AND TECHNIQUES

bs_bs_banner

doi:10.1111/add.12472

2014 Society for the Study of Addiction Addiction, 109, 728735

embodied cognitions are initially experienced as pleasur-able, but if acquisition of the target is prevented ordelayed the sensory imagery becomes aversive, because itmakes physiological deficits more salient.

Sensory imagery is common to the phenomenology ofcraving in addiction and for food and drink [79].Because craving imagery loads limited capacity workingmemory processes, craving can be reduced by cognitivetasks that compete for those same working memory pro-cesses. This competition is greater in the imagerysmodality; therefore, imagery of neutral visual or olfac-tory stimuli reduces craving strength for cigarettes[6,10], coffee [11] and food [12], relative to auditory orverbal tasks.

A barrier to progress in the understanding of desireshas been a lack of conceptually sound and agreedapproaches to their measurement. Kavanagh et al. [13]reviewed existing measures of alcohol craving, demon-strating that many can be confounded with correlatedbut separate phenomena (drinking behaviour, intentionsto drink, outcome expectancies about perceived benefitsor consequences of drinking and perceptions of behav-ioural control or self-efficacy). Similar issues apply tocraving measures in other domains: for instance, theQuestionnaire on Smoking Urges (QSU) [14] and stateversion of the Food Craving Questionnaire (FCQ-S) [15]include items on expectancy and intention (QSU) orcontrol (FCQ-S). Kavanagh et al. proposed that cravingmeasures should avoid such separate phenomena, butinclude inherent craving attributes such as the frequencyand intensity of episodes and the salience of cravingthoughts. The measure should be sufficiently brief toallow repeated use during experiments or in high-risksituations for addictive behaviours, and should capturekey features of the subjective experience.

The Alcohol Craving Experience (ACE) questionnaire[16] was developed from EI theory. It assesses the inten-sity and intrusiveness of alcohol craving, along with viv-idness of imagery during the craving episode. The ACE isan 11-item scale with two forms: strength (ACE-S), whererespondents rate the strength of craving experiencesduring a specific episode, and frequency (ACE-F), wherethey rate how often they experienced strong urges, intru-sive desires and vivid imagery of alcohol use. Both formsinclude an explicit time-scale. Each form has the samefactor structure, with items divided between correlatedsubscales measuring craving intensity, imagery vividnessand cognitive intrusiveness. Frequency and intensityare assessed separately to allow detection of both infre-quent, intense craving and frequent but weaker episodes.The ACE discriminated between clinical and non-clinicalpopulations, and within the non-clinical sample it dis-criminated those who screened positive for higher risk ofalcohol dependence from those who did not.

The similar processes by which craving is triggeredand maintained may allow development of a measurewith broad application across consummatory targets.Such a measure would contribute to research oncommonalities across addictions (e.g. food addiction[17,18]). A version of the ACE may fulfil this role. Likeother craving scales, the original ACE focused on a singletarget (alcohol). However, because it was derived froma general theory of desire, it should be possible todecontextualize items to produce a measure with widerapplication. As a first step, we used variations of the ACEtranslated for cigarettes, food or chocolate, analogous toadaptations of the QSU to the Questionnaire of AlcoholUrges [19] and Alcohol Craving Questionnaire [20].

This paper pools data from12 studies using variants ofthe ACE, to test whether the measure has the same facto-rial structure, independent of consummatory target andtime-scale. We used confirmatory factor analysis (CFA) tocompare a three-factor and single-factor model, andexamined modification indices to evaluate other possiblestructures.

METHOD

Twelve studies conducted between 2008 and 2012 inPlymouth, UK and Brisbane, Australia are summarized inTable 1. Each provided data on the 11 ACE items identi-fied by Statham et al. [16]: 10 are inTable 3; the 11thwasbody [how vividly/often did you imagine how your bodywould feel if you had (the target)?]. All studies gaveresponses to strength items, and six also contributed dataon frequency. Only studies 13 have already been pub-lished [7,21]. The primary purpose of the studies was tocompare the effects of various interventions uponcraving. Wording of the ACE items was modified foreach study, but all had the original 11-point visualanalogue scales (anchored 0 = not at all; 10 = extremely/constantly). While sample sizes of each study wereappropriate for their primary aims, none were sufficientlylarge to permit factor analyses. Samples are thereforepooled in the current paper to provide a large, albeit het-erogeneous, sample for analysis.

Participants and procedures

Studies 19 were conducted at Plymouth University (UK)and recruited undergraduate psychology students,rewarding them with points to recruit other students fortheir own research. Participants were recruited throughan online system advertising research studies to psychol-ogy undergraduates. Overall, 496 students took part(330 females, 109 males, 57 unknown), aged 1850years (mean = 23). Studies 14 used an incidentalcraving induction method, where participants chose a

CEQ: Craving Experience Questionnaire 729


chocolate to eat later, and then rated craving for choco-late now (studies 1 and 2) or in the last 10 minutes(studies 3 and 4). In studies 3 and 4, ratings wereobtained after participants completed an experimentaltask to test hypotheses that visual tasks would reducecraving, but verbal tasks would have little effect.

Study 5 recruited people who reported smoking 10cigarettes/day, and asked them either to rate cigarettecravings now or over the last week.

Studies 69 tested participants who had not eaten for25 hours (except study 8) after an indirect cravinginduction asking about their plans to eat (except in study7). In studies 68, cravings now were rated (in study 8after a visual task), while in study 9 cravings in the last10 minutes were rated (after experimental tasks).

Study 10 (n = 399 aged 1960 years, mean = 28; 66males) was an online survey advertised to staff and stu-dents at Plymouth University. No incentivewas offered forcompleting the survey. Respondents identifying them-selves as current or former smokers were asked aboutcigarette cravings, while the remainder rated their food or

chocolate cravings. Current smokers all reportedsmoking more than six cigarettes/day.

Study 11, at Queensland University of Technology,used an opportunity community sample of currentsmokers who wanted to quit smoking (n = 59 aged 1860 years, mean = 31, 27 females). Participants wererecruited via e-mails or flyers at Brisbane universities andwork-places, and approaching smokers individually. Par-ticipants reported smoking 15 cigarettes/day and werenot currently using nicotine replacement or receivingsmoking treatment. They were asked to abstain fromsmoking for 5 hours prior to testing, and then completedthe QSU, the Revised FagerstrmTolerance Questionnaire(RTQ; [22]) and the ACE.

Study 12, at an alcohol and drug out-patient servicein Brisbane, drew from 542 participants (178 female,aged 1870 years, mean = 40) meeting DSM-IV-TR cri-teria for current alcohol dependence. Participants werenot dependent on other substances (except nicotine), andwere not taking anti-cravingmedication. They completedthe ACE at their first clinical session of a 12-week

Table 1 Summaries of the studies.

Target Study Abstention Induction Context Time-frame nCEQ-SM SD n

CEQ-FM SD

Chocolate 1 No Yes Now 35 3.92 1.65 Chocolate 2 No Yes Now 45 5.22 1.74 Chocolate 3 Yes Yes Verbal (digits) Last 10 minutes 19 4.74 2.38 19 4.80 2.72Chocolate 3 Yes Yes Visual (matrix) Last 10 minutes 17 4.22 2.27 17 4.06 2.36Chocolate 4 No Yes Verbal (counting) Last 10 minutes 45 4.12 2.42 45 3.70 2.26Chocolate 4 No Yes Visual (DVN) Last 10 minutes 45 3.34 2.51 45 2.68 2.26Chocolate 4 No Yes Visual (clay) Last 10 minutes 48 3.44 2.36 48 2.77 2.31Chocolate 10 No No Last time 76 5.50 1.88 76 4.36 2.02Chocolate overall 330 4.40 2.28 250 3.62 2.37Cigarette 5 No No Now 20 5.24 1.87 Cigarette 5 No No Last week 20 5.97 1.96 Cigarette 10 No No Smoker Last time 89 4.53 2.14 89 3.89 2.08Cigarette 10 No No Ex-smoker Last time 124 4.20 2.24 124 1.17 1.86Cigarette 11 Yes No Last 5 minutes 59 4.57 2.35 59 4.74 2.55Cigarette overall 312 4.54 2.23 272 2.83 2.61Food 6 Yes Yes Now 56 3.78 2.22 Food 7 Yes No Now 30 4.42 1.91 Food 8 No Yes Visual (clay) Now 56 3.79 2.05 Food 9 Yes Yes Verbal (counting) Last 10 minutes 20 4.41 2.38 20 4.05 2.34Food 9 Yes Yes Mind wandering Last 10 minutes 20 5.03 2.20 20 4.32 1.80Food 9 Yes Yes Visual (clay) Last 10 minutes 20 4.99 1.87 20 4.64 2.16Food 10 No No Last time 110 5.61 1.57 110 4.02 2.08Food overall 312 4.68 2.07 170 4.11 2.09Alcohol 12 No No Last week 276 5.22 2.22 276 4.04 2.33All Substances 1230 4.69 2.22 968 3.60 2.43

Target indicates the substance for which cravings were being reported; Study identifies the study described in the Method text; abstention and inductionindicate whether participants had abstained from the target substance before completing the questionnaire or if they had been subjected to a cravinginduction procedure; context describes any experimental task that had been experienced before completing the questionnaire; time-frame indicates theperiod over which cravings were being reported; for both the CEQ-S and CEQ-F formsM indicates themean scale total over all 10 items (010) and SD thestandard deviation of these scale totals.Ns are given for both forms to allow the pooled sample sizes to be shown. Bold face rows showoverall values pooledover studies for the relevant substance. CEQ-S = Craving Experience Questionnaire-strength; CEQ-F = Craving Experience Questionnaire-frequency.

730 Jon May et al.


cognitive behavioural programme for alcohol depend-ence, along with the ObsessiveCompulsive DrinkingScaleObsessions subscale (OCDS-Obsessions) [23]. Assome had contributed to development of the ACE [16],the current confirmatory factor analyses presented belowused only the 276 who had not been in the earlierstudy (90 female, aged 1969 years, mean = 40). Thefull sample was used for preliminary tests of convergentvalidity.

Craving time-frame

For studies where the reported time-frame was now,only the strength form (ACE-S) was administered. Studiesreferring to a time-frame also allowed the frequency form,ACE-F, to be administered. The time-frame of 10 minutesin some studies reflects the duration of the experimentaltasks being compared for their impact on craving cogni-tions. Studies 1012 were correlational in design. Study10 asked people to rate the last time they experienced acraving for ACE-S, but the last week for ACE-F. Study 11asked participants to report the strength and frequency ofcravings over the previous 5minutes. The clinical samplein study 12 rated the previous week. The differing time-frames reflect the varied purposes of the original studies,and provide an opportunity to examine stability of factorstructures over short (510 minutes) and longer (lasttime/last week) periods.

Preliminary validity data

Examples of relationships with other variables were pro-vided by studies 11 (on current smokers) and 12 (onalcohol-dependent out-patients). Study 11 provided datafrom factor 1 (desire to smoke) of the QSU and nicotinedependence on the RTQ. Study 12 provided correlationswith the OCDS-Obsessions subscale and with a single-item rating of the extent to which respondents currentlyfelt like a drink [24].

Confirmatory factor analyses

Combining the 12 samples provided 1230 participantson the ACE-S (330 chocolate, 312 food, 276 alcohol,312 cigarettes); 968 of these also completed the ACE-F(250 chocolate, 170 food, 276 alcohol, 272 cigarettes).Maximum likelihood confirmatory factor analyses wereconducted using SPSS AMOS version 20 (IBM Corp,Armonk, NY, USA), with multi-group analyses on thefour craved substances. We compared models usingchange in 2 values, 2/d.f. ratios (values >3 indicatingpoor fit), comparative fit indices (CFI, values 0.05 indicating poor fit), with pClose beingused to test whether RMSEA was greater than the 0.05threshold.

Tests of multivariate normality indicated that datawere non-normal, with most items having non-normalunivariate skewness or kurtosis. As reported by Gao,Mokhtarian & Johnston [25], removing a large number ofcases or transforming the data can create more funda-mental problems inmodel interpretation. Henly [26] pro-vides a rule of thumb that samples showing multivariatenon-normality should be 600 to obtain unbiasedparameter estimates. As our sample size exceeded thisthreshold, we chose to proceed with analyses on thecomplete, untransformed data.

RESULTS

We began by applying the three-factor model reported byStatham et al. [16], with items want, need, urge andthink loading on intensity; picture, taste, smell,mouth and body on imagery; and not think andintrusive on intrusiveness (see Table 3 for full items).

We first included all items from ACE-S and ACE-Fwithin a single analysis to determine if it was appropriateto treat the forms separately (Table 2). A six-factor modelwith separate intensity, imagery and intrusivenessfactors for each form fitted the data more closely than athree-factor model combining subscales over the twoforms (change 2(48) = 2879, P < 0.001). We thereforeproceeded to analyse the forms separately.

We next compared the ACE three-factor model withone where all 11 items were linked to a single cravingfactor. For both forms, the three-factor model produceda significantly better fit than the single factor (ACE-S:change 2(12) = 2032, P < 0.001; ACE-F: change 2(12) =1315, P < 0.001). We then added correlations betweenpairs of error components within a factor that showeda high and consistent pattern of modification indicesover the substance groups. This further improved fitfor both forms (ACE-S: change 2(16) = 201, P < 0.001;ACE-F: change 2(20) = 287, P < 0.001), with ACE-S nowmeeting CFI and RMSEA fit criteria, although still exceed-ing 3 on CMIN/d.f. ACE-F also met the CFI criterion, butnot RMSEA or CMIN/d.f. criteria.

We therefore examined the modification indices foritem loadings. These suggested removing the body item,and reassigning think from the intensity to the intru-siveness factor. The modified three-factor model resultedin improved fit for both forms on all criteria, with theACE-S producing a CMIN/d.f. ratio below 3, and ACE-Fnow meeting the criterion for RMSEA (Table 2).

Following this re-organization and item reduction,and to reflect the generic nature of the questionnaire, we



refer to the two forms of the 10-item scale as the CravingExperience Questionnaire (CEQ).

Invariance over substances

The analyses reported so far treated the four groups ofparticipants who had reported cravings to alcohol, ciga-rettes, chocolate and food as separate groups. Constrain-ing measurement weights and structural covariances toforce factor loadings between items and the three factorsto be identical over the four substance groups worsenedmodel fit significantly (CEQ-S constrained: 2 change(39) = 131, P < 0.001; CEQ-F, constrained: 2 change(39) = 172, P < 0.001). Therefore, while the overallfactor structure remained invariant over different con-summatory targets, specific loadings changed.

Invariance over time-frames

We then examined invariance over the different time-scales referred to in variants of the CEQ, pooling oversubstances. Of those who had completed both CEQ-S andCEQ-F, 695 rated their experiences over a long time-span(the last week, the last time or since the last time),whereas 293 rated a short time-scale (the last 5 or 10

minutes). On the CEQ-S, an additional 242 rated theircraving now (or at this time). We thus conductedmulti-group analyses using two time-frames for theCEQ-F and three time-frames for the CEQ-S (Table 2).

For both forms, these analyses produced acceptablefit. Constraining measurement weights and structuralcovariances made fit worse (CEQ-S: 2 change (26) = 73,P < 0.001; CEQ-F: 2 change (13) = 44, P < 0.001). Inboth cases, CMIN/d.f. and RMSEA improved in con-strained models, but CFI and AIC worsened slightly. Thissuggests that different time-frames resulted in similarpatterns of response.

Invariance over studies

To confirm that it was appropriate to pool the 12 studiesinto a single data set, despite their heterogeneity, we con-ducted multi-group analyses by study. All 12 studies usedthe CEQ-S and six used the CEQ-F. Constraining measure-ment weights and structural covariances had a smalleffect upon fit (CEQ-S: 2 change (39) = 51, P = 0.016;CEQ-F: 2 change (13) = 49, P < 0.001), with indiceschanging negligibly or improving. This indicates thatstudy heterogeneity is not a problem.

Table 2 Comparison of models over different substances and time-frames.

Model d.f. 2 CMIN/d.f. CFI RMSEA (95%CI) pClose AIC

Combined forms, 3 factors 824 6730 8.17 0.721 0.086 (0.084 to 0.088)

Final version of the CEQ

The CEQ, with modified wording to avoid reference to aspecific substance, is shown inTable 3. For CEQ-S, partici-pants either rate current craving or the strongest cravingover a time-frame; CEQ-F requires a time-frame to bespecified. Total scores summarize strength and frequencyof craving, respectively. Coefficient alphas are 0.91 forCEQ-S (corrected item/total correlations: 0.430.75) and0.94 for CEQ-F (0.570.84). The lowest item-total corre-lations were for the not think item, which is primarilyrelevant when people are trying to suppress desirethoughts. The CEQ-S and CEQ-F totals correlated 0.72,and within each form the three subscale intercorrela-tions ranged from 0.52 to 0.61 for CEQ-S and 0.63 to0.73 for CEQ-F. Between the CEQ-S and the CEQ-F, the two

imagery subscales correlated 0.72; intensity correlated0.67 and intrusiveness 0.71.

Preliminary validity data

The total CEQ-S and CEQ-F over the previous 5 minuteswere correlated strongly with the QSU desire to smokefactor (the subscale according most closely with our defi-nition of cigarette craving) and with current nicotinedependence on the RTQ (Table 4, study 11). Relation-ships were particularly strong with intensity andimagery subscales.

In study 12, the total CEQ-S and CEQ-F had thetime-frame of the previous week. Both forms had robustcorrelations with OCDS-Obsessions, which does not spe-cifically specify a time-frame (e.g. howmuch of your time

Table 3 Composition of the 10-item Craving Experience Questionnaire (CEQ).

Factor/Item Strength form, CEQ-Sa Frequency form, CEQ-Fa

Intensity Right now . . ./At that time . . . Over the last [time-frame] how often . . .Want . . . how much do/did you want it? . . . did you want it?Need . . . how much do/did you need it? . . . did you need it?Urge . . . how strong is/was the urge to have it? . . . did you have a strong urge for it?Imagery Right now/At that time, how vividly do/did you Over the last [time-frame] how often did you . . .Picture . . . picture it . . . picture itTaste . . . imagine its taste? . . . imagine its taste?Smell . . . imagine its smell? . . . imagine its smell?Mouth . . . imagine what it would feel like in your mouth or throat? . . . imagine what it would feel like in your mouth

or throat?Intrusiveness Right now . . ./At that time . . . Over the last [time-frame] how often . . .Not think . . . how hard are/were you trying not to think about it? . . . were you trying not to think about it?Intrusive . . . how intrusive are/were the thoughts? . . . were the thoughts intrusive?Think . . . how hard is/was it to think about anything else? . . . was it hard to think about anything else?

F = frequency; S = strength. aAlso tested in the analyses, but not included in the final CEQ, was body (how vividly/often did you imagine how yourbody would feel if you had a drink/cigarette etc.?).

Table 4 Correlations between the Craving Experience Questionnaire (CEQ) and other measures.

CEQ form and subscale

Study 11: current smokersa Study 12: alcohol-dependent out-patientsa

QSUDesire tosmokeb

RTQc

Nicotinedependence

OCDS-obsessionsd

Borg scale attreatmentsession 1e

CEQ-S total 0.75 0.67 0.59 0.34CEQ-S intensity 0.74 0.64 0.55 0.33CEQ-S imagery 0.65 0.60 0.50 0.31CEQ-S intrusiveness 0.48 0.43 0.53 0.27CEQ-F total 0.76 0.61 0.62 0.38CEQ-F intensity 0.77 0.61 0.59 0.40CEQ-F imagery 0.67 0.55 0.53 0.33CEQ-F intrusiveness 0.52 0.41 0.56 0.30

F = frequency; S = strength. All P 0.001. aIn study 11, the CEQ referred to the previous 5 minutes; in study 12, it referred to the previous week. Inboth cases, respondents rated strength items in relation to the time when they most wanted a cigarette/alcohol drink in that period. bQuestionnaire onsmoking urges, factor 1 (desire to smoke). cRevised Fagerstrm Tolerance Questionnaire (RTQ). dObsessivecompulsive drinking scale (OCS-S)obsessions subscale (items 16, 13). eHow much respondents currently felt like a drink (single item, 05).



when youre not drinking is occupied by ideas, thoughts,impulses or images related to drinking?). Unsurprisingly,it was more moderately associated with a cross-sectionalmeasure of the extent to which session 1 attendeescurrently felt like a drink.

DISCUSSION

The strength and frequency forms of the 10-item CEQshow a robust three-factor structure which, with theexception of one re-allocated and one omitted item,replicates the structure of the alcohol-specific ACE.Together, the CEQ-S and CEQ-F measure the core desirecomponents of frequency, intensity and the salience ordismissability of related intrusive thoughts, as identifiedin the recent review by Kavanagh et al. [13]. They alsotrack a key cognitive feature of cravingdesire-relatedimagerythe theoretical and empirical importance ofwhich is demonstrated in other work [5,27], and by theclose association of the imagery subscale and QSUdesire to smoke.

Specific structural weights varied across substancesand time-frames, but the factor structure was invariant,giving confidence that CEQ subscales apply across con-texts. The stability over time-frames occurred despitetheir different demands upon memory. Frequency andstrength forms were statistically distinct, showing thatrespondents were not conflating intensitywith frequency,although there is inevitably an association between thesetwo facets of craving that leads to substantial correla-tions between the two forms, and between CEQ-F and theQSU, which is a measure of craving strength. Futureresearch could assess the accuracy of the CEQ-F, to deter-mine if it is actually measuring craving frequency or if itis simply a re-phrased measure of craving strength. Thestability of the overall factor structure provides furtherevidence that normal, everyday desires lie on the samecontinuum as dysfunctional ones, and are driven bysimilar cognitive mechanisms.

While pooling of data from different studies provided atotal sample size sufficient for factorial analyses, the het-erogeneity of samples and studies is a potential limitationto the study, although our multi-group analysis indicatedinvariance over studies. Now that we have demonstratedthe feasibility of adapting the ACE to a more genericformat, studies with more balanced sampling areessential to corroborate the CEQs factor structure,and examine the possible value of the body item insubstance-dependent contexts other than alcohol. Recentdata using the imagery items in a large communitysample [28] indicate that the body item consistently pre-dicted alcohol consumption, and that although the morespecific mouth/throat item also correlated with drink-ing, it did not uniquely predict it. These two items would

seem to be tapping the same aspects of imagery, and itmay be that a rephrasing of the more general body itemwill prove appropriate in contexts beyond oral consump-tion. Only one of the studies reported here usedsubstance-dependent participants: further research isneeded to give greater confidence that factor structuresare invariant across degree of dependence.

The sensory foci of the CEQ restrict its current form toconsummatory targets. As the specific senses implicatedmost strongly in craving imagery are likely to change fornon-consummatory targets such as sport or gambling[7,9], different items may be needed for the imagerysubscale in those contexts. However, there is no a priorireason to expect that intensity or intrusiveness factorswill need to change.

The CEQ forms provide sensitivity to two central fea-tures of desires: that they occur with differing frequencyand intensity. The retrospective CEQ-F will be especiallyvaluable in contexts where the number of spontaneousthoughts and their salience are the critical phenomena(e.g. where maintenance of resistance to recurringdesires is being examined).

The CEQ-S represents a distinct departure from mostprevious measures, which ask people to make averageratings of intensity over long or undefined time-periods[13]. Instead, it focuses on the current experience or on ahighly salient episode in the recent past, e.g. strongestcraving in the last day or week. This focus should helprespondents to recall a concrete experience rather thanresponding on the basis of generic knowledge or attempt-ing to derive an average. The flexibility of time-framesallows the repeated use of the full form or the intensitysubscale in self-monitoring or experimental sessions.Together, strength and frequency forms allow researchersto test whether frequency of craving episodes or intensityof the worst episodes determines consumption andrelapse.

In current smokers, versions of the CEQ-S and CEQ-Fthat focused on the previous 5 minutes had strong rela-tionships with current cigarette craving and nicotinedependence. The correlation between CEQ intensityand QSU desire to smoke was especially high. In thealcohol sample, differences in the time-frames usedfor the respective measures probably restricted the sizeof correlations, especially for the single-item measureobtained in the treatment session, when 57% ofthose answering the question had no current craving.However, intercorrelations of the last-week CEQ withOCDS-Obsessions were moderate to strong, and eventhose with current desire in session 1 were highlysignificant.

This study provides robust data on the internal struc-ture of the CEQ and offers preliminary data on itsconvergent validity. This brief, flexible and conceptually

734 Jon May et al.


grounded measure provides a tool for assessing cravingexperience independently of behavioural variables andallows comparisons across desires for a range of consum-matory behaviours.

Declaration of interests

None.

Acknowledgements

Studies 110 were conducted by Cara Goodman, HarrietTaylor, Alexios Tzovaras, Kayleigh Hopkins, LornaStephenson-Rymer, Kayleigh Hughes, Rebecca Holland,Amanda Hynes, Robyn Newport, Lucy Williams, CarlyRudder, Kayleigh Preedy, Hayley Illingworth, CatherineWard, Marco Flatau, AlexWheatley, Claire Blackford andRebecca Chidley. Kerri Gillespie contributed to the datacollection in Study 11. Study 12 was supported by inter-nal funds and its conduct was assisted by staff of theAlcohol and Drug Assessment Unit, Princess AlexandraHospital. Jason P. Connor is supported by an AustralianNational Health and Medical Research Council CareerDevelopment Fellowship (APP1021909).

References

1. American Psychiatric Association. Diagnostic and StatisticalManual of Mental Disorders, Fifth Edition (DSM-5). Arling-ton, VA: American Psychiatric Association; 2013.

2. World Health Organization (WHO). The ICD-10 Classifica-tion of Mental and Behavioural Disorders: Clinical Descriptionsand Diagnostic Guidelines. Geneva: WHO; 1992.

3. Tiffany S. T. A cognitive model of drug urges and drug-usebehaviour: role of automatic and nonautomatic processes.Psychol Rev 1990; 97: 14768.

4. Skinner M. D., Aubin H.-J. Cravings place in addictiontheory: contributions of the major models. NeurosciBiobehav Rev 2010; 34: 60623.

5. Kavanagh D. J., Andrade J., May J. Imaginary relish andexquisite torture: the Elaborated Intrusion theory of desire.Psychol Rev 2005; 112: 44667.

6. May J., Andrade J., Panabokke N., Kavanagh D. J.Visuospatial tasks suppress craving for cigarettes. Behav ResTher 2010; 48: 47685.

7. May J., Andrade J., Panabokke N., Kavanagh D. J. Images ofdesire: cognitive models of craving. Memory 2004; 12:44761.

8. May J., Andrade J., Kavanagh D. J., Penfound L. Imagery andstrength of craving for eating, drinking and playing sport.Cogn Emot 2008; 22: 63350.

9. Kavanagh D. J., May J., Andrade J. Tests of the ElaboratedIntrusion Theory of craving and desire: features of alcoholcraving during treatment for an alcohol disorder. Br J ClinPsychol 2009; 48: 24154.

10. Versland A., Rosenberg H. Effect of brief imagery interven-tions on craving in college student smokers. Addict ResTheory 2007; 15: 17787.

11. Kemps E., Tiggemann M. Competing visual and olfactoryimagery tasks suppress craving for coffee. Exp ClinPsychopharmacol 2009; 17: 4350.

12. Kemps E., Tiggemann M. Modality-specific imagery reducescravings for food: an application of the Elaborated Intrusiontheory of desire to food craving. J Exp Psychol Appl 2007;13: 95104.

13. KavanaghD. J., StathamD. J., FeeneyG. F., YoungR.M.,MayJ., Andrade J. et al. Measurement of alcohol craving. AddictBehav 2013; 38: 157284.

14. Tiffany S. T., Drobes D. J. The development and initial vali-dation of a questionnaire on smoking urges. Br J Addict1991; 86: 146776.

15. Cepeda-Benito A., Gleaves D. H., Williams T. L., Erath S. A.The development and validation of the state and traitfood-cravings questionnaires. Behav Ther 2001; 31: 15173.

16. Statham D. J., Connor J. P., Kavanagh D. J., Feeney G. F.,Young R. M., May J. et al. Measuring alcohol craving: devel-opment of the alcohol craving questionnaire. Addiction2011; 106: 12308.

17. Gearhardt A. N., Corbin W. R., Brownell K. D. Preliminaryvalidation of the Yale Food Addiction Scale. Appetite 2009;52: 4306.

18. Rogers P. J., Smit H. J. Food craving and food addiction:a critical review of the evidence from a biopsychosocialperspective. Pharmacol Biochem Behav 2000; 66: 314.

19. Bohn M. J., Krahn D. D., Staehler B. A. Developmentand initial validation of a measure of drinking urgesin abstinent alcoholics. Alcohol Clin Exp Res 1995; 19:6006.

20. Singleton E. G., Tiffany S. T., Henningfield J. E. The Multidi-mensional Aspects of Craving for Alcohol. Baltimore, MD:National Institute on Drug Abuse, National Institutes ofHealth; 1994.

21. Andrade J., Pears S., May J., Kavanagh D. J. Use of a claymodeling task to reduce chocolate craving. Appetite 2012;58: 95563.

22. Tate J. C., Schmitz J. A proposed revision of the FagerstromTolerance Questionnaire. Addict Behav 1993; 18: 13543.

23. Anton R. F., Moak D. H., Latham P. The Obsessive Compul-sive Drinking Scale: a self-rated instrument for the quantifi-cation of thoughts about alcohol and drinking behavior.Alcohol Clin Exp Res 1995; 19: 929.

24. Borg V. Bromocriptine in the prevention of alcohol abuse.Acta Psychiatr Scand 1983; 68: 10010.

25. Gao S., Mokhtarian P. L., Johnston R. A. Nonnormalityof data in structural equation models. Transport Res Rec2008; 2062: 11624.

26. Henly S. J. Robustness of some estimators for the analysisof covariance structure. Br J Math Stat Psychol 1993; 46:31338.

27. Andrade J., May J., Kavanagh D. J. Sensory imagery incraving: from cognitive psychology to new treatments foraddiction. J Exp Psychopathol 2012; 3: 12745.

28. Connor J. P., Kavanagh D. J., Andrade J., May J., Feeney G. F.X., Gullo M. J. et al. Alcohol consumption in young adults:the role of multisensory imagery. Addict Behav 2013; 39:7214.



the craving experience questionnaire

Documents