Assessing Craving for Alcohol

David J. Drobes, Ph.D., and Suzanne E. Thomas, Ph.D.

Craving for alcohol is common among alcohol-dependent people. Accordingly, measures toassess craving can play important roles in alcohol research as well as in alcoholism treatment.When developing or employing craving-assessment instruments, researchers and cliniciansmust consider numerous factors, such as the specific characteristics of craving to be evaluated,the instrument’s psychometric properties, and the timeframe over which craving is assessed.The measures most commonly used for assessing craving in clinical settings are single-itemquestionnaires, although several multi-item questionnaires also have been developed.Behavioral measures (e.g., amount of alcohol consumption or performance on cognitive tests)and psychophysiological measures (e.g., changes in salivation, respiration, or heart rate) arebeing used primarily in research settings. The assessment of craving can have numerous clinicalbenefits, such as helping the clinician to evaluate the severity of a patient’s alcohol dependence,to select appropriate treatment approaches, and to monitor changes throughout a patient’streatment. The role of craving assessment in predicting treatment outcome, however, remainscontroversial. KEY WORDS: AOD (alcohol and other drug) craving; assessment of variables andmethods; specific AODU (alcohol and other drug use) measurement and test; AOD use behavior;psychophysiology; alcohol cue; temporal context; patient assessment; patient-treatmentmatching; disease severity; patient monitoring; self report; questionnaire; literature review

Vol. 23, No. 3, 1999 179

The phenomenon of craving hasreceived increasing attention inrecent years from researchers and

clinicians working with alcohol-relateddisorders. Scientists have developed severaltheoretical models of the development,neurobiology, and phenomenology ofcraving, some of which are described inother articles in this journal issue (e.g.,see the articles in this issue by Anton,pp. 165–173, and by Tiffany, pp.215–224). Despite increasing interestin this topic, little agreement exists onhow best to conceptualize or measurecraving. In addition, fundamental gapsin knowledge remain concerning therelationship between craving and actualdrinking behavior.

As researchers and clinicians knowfrom experience, craving is a commonoccurrence among alcohol-dependentpeople and a frequent topic of discussionin most alcoholism treatment and research

settings. Accordingly, measures that accu-rately assess craving can play an importantrole in alcoholism research and treatment.This article reviews several critical issuesassociated with craving assessment andprovides an overview of currently availablemethods for measuring alcohol craving.The article also provides some suggestionsfor assessing craving in routine clinicalpractice. Finally, it briefly discusses recentadvances that may enhance the under-standing and measurement of craving.

Factors Affecting Craving Assessment

Variables Assessed

The accuracy with which various instru-ments, or indices, measure craving foralcohol and other drugs (AODs) dependsto a considerable extent on the type of

variables the instruments evaluate. Earlyperspectives on alcohol craving focusedon the subjective nature of craving. Thatis, craving was viewed as an experiencethat could only be assessed through theverbal report of the alcoholic. Con-sequently, the accuracy of most cravingindices was limited by the ability andwillingness of the individual alcoholicto accurately report his or her personalexperience.

More recent conceptualizations ofcraving have fostered a broader per-spective on the nature of craving and,consequently, on sources of data that

DAVID J. DROBES, PH.D., is an assistantprofessor and SUZANNE E. THOMAS,PH.D., is a research instructor at theCenter for Drug and Alcohol Programs,Medical University of South Carolina,Charleston, South Carolina.

could provide important informationon craving. For instance, most researchersassume that craving among alcoholics isinexorably linked to alcohol consump-tion and to relapse to drinking in absti-nent alcoholics. This perspective considersbehaviors related to seeking or consum-ing alcohol as direct manifestations ofcraving. Some investigators, however,have challenged the assumption thatalcohol-related behaviors directly resultfrom craving on both theoretical andempirical grounds (see Tiffany 1990).Therefore, researchers must furtherdetermine for whom, when, and underwhat circumstances meaningful rela-tionships may exist between cravingand alcohol consumption.

Instruments that measure autonomicphysiological activity1 (e.g., changes inheart rate, blood pressure, or sweatgland activity) in response to alcohol-related cues, such as the sight and smellof alcohol, also have received increasedattention in studies of alcohol craving.Instruments that assess such autonomicresponses to alcohol-related cues areparticularly relevant to theories of cravingthat postulate a role for classical condi-tioning2 (e.g., Drobes et al. in press).

Psychometric Issues

With the recent surge in interest in thetheoretical and clinical importance ofcraving, researchers have attempted todevelop ways to measure craving as wellas ways to assess the utility of existingcraving instruments. The utility of anycraving assessment instrument dependson its psychometric properties; specifi-cally, on whether the instrument reli-ably and validly measures craving.

The term “reliability” refers to thedegree to which items on an instrumentyield consistent results. Several types of reliability exist, not all of which are

applicable to every type of assessmentinstrument. For example, test-retestreliability refers to the consistency ofresults when a subject is tested severaltimes. Because craving is considered arelatively transient state that is expectedto differ from one occasion to another,a high degree of test-retest reliability is

not necessarily desirable for cravinginstruments. Other indices of reliabilityare more important to craving instru-ments—for example, whether consis-tent results are obtained with an instru-ment when several testers evaluate thesame person (i.e., inter-rater reliability)and whether similar items on a test donot yield inconsistent or contradictoryresponses (i.e., internal consistency).

The term “validity” refers, in general,to the degree to which an instrumentmeasures what it purports to measure.Because the empirical study of cravingis a relatively new endeavor, and inves-tigators consequently are still trying tounderstand exactly what craving is (andis not), assessing the validity of an instru-ment that proposes to measure cravingis a difficult (and possibly premature)task. After more data have been gener-ated and some agreement has beenestablished on what constitutes craving,researchers can more confidently assessthe validity of craving instruments.

As with reliability, several types ofvalidity exist. Some of the most commonmeasures of validity include constructvalidity, external validity, discriminantvalidity, and criterion-oriented validity.Construct validity refers to the degree towhich results from an instrument reflectthe underlying quality (e.g., craving)that the instrument is trying to measure.

External validity describes the degree to which the results obtained with theinstrument agree with results obtainedwith a different instrument with estab-lished validity. Discriminant validityindicates an instrument’s ability toaccurately discriminate between popu-lations with and without the quality ofinterest. For example, for alcohol craving,discriminant validity might be tested bywhether the instrument discriminatesalcoholics from nonalcoholics. Finally,criterion-oriented validity refers to howwell scores on an instrument correlatewith behaviors that are supposedly rele-vant to the quality being measured. Forexample, good criterion-oriented validitymight be suggested if a craving instru-ment yields high scores for people whoare willing to work for a drink in an experi-mental setting and low scores for peoplewho are not willing to work for alcohol.

Most clinical and research assessmentsof craving have relied on brief measure-ment scales with unknown psychomet-ric properties. In fact, the majority ofinvestigators and clinicians have usedinstruments consisting of only a singlequestion (i.e., single-item tests) to assesscraving. Typically, these instrumentsask questions such as, “How strong isyour craving for alcohol?” or “Howmuch do you crave an alcoholic bever-age when you have gone without a drinkfor 1 to 2 days?” Because these indicesconsist of only one question, it isimpossible to obtain an estimate oftheir internal consistency. Nonetheless,single items or brief scales that captureimportant aspects of the craving constructmay demonstrate satisfactory inter-rater or test-retest reliability. Therefore,brief scales may provide especially use-ful tools where multiple craving assess-ments are desired and lengthier instru-ments would be impractical.

Alcohol cravings are generally thoughtto arise either from the desire to experi-ence alcohol’s positive effects (i.e., posi-tive reinforcement) or from the desire toavoid the negative effects of withhold-ing alcohol, such as withdrawal symptoms(i.e., negative reinforcement). More recentmodels have suggested other importantdimensions of craving, such as the desireand intention to consume alcohol, lackof control over alcohol use, and preoccu-

180 Alcohol Research & Health

1Autonomic physiological responses are those reac-tions that are not normally controlled consciously,such as heart rate and blood pressure.

2Classical conditioning is the process by which anoriginally neutral stimulus (e.g., sight of a bar) caninduce a predictable response (e.g., an increase inheart rate) if it is repeatedly paired with anotherstimulus which elicits that response (e.g., alcoholconsumption).

The majority of investigators andclinicians have used

instruments consisting of only a single question

to assess craving.

pation with drinking-related thoughtsand/or behavior. It is possible that cer-tain aspects of craving characterize theexperience of some alcoholics better thanthat of other alcoholics. Similarly, patients’craving descriptions may change through-out the treatment process, necessitatingmeasures that are sensitive to the dimen-sions of craving experienced at differenttime points or under differing circum-stances. The development of indices thatare sensitive to such differences may behelpful in the development of treatmentplans based on the relative importanceof various features of craving acrossindividual patients and over time.

Unfortunately, the brevity of mostavailable scales has not permittedresearchers and clinicians to disentanglethe multiple aspects of craving. Severalnewer multi-item instruments that takeinto account various dimensions of crav-ing are discussed in the section “Methodsof Assessing Alcohol Craving,” below.One should not assume, however, thateach of these scales measures the samedimensions of craving or even that thesame dimensional structure would beproduced by the same instrumentwhen administered to different samplesof alcoholics (see Kranzler 1999).

Timeframe of Assessment

Another critical aspect to consider whenmeasuring craving is the timeframe ofthe assessment. Craving measures fallinto two main categories with respectto their timeframe: (1) state measures,which focus on the patient’s current(“here-and-now”) craving status, and(2) global measures, which ask thepatient to describe his or her generalexperience of craving over the course of 1 day, 1 week, 1 month, or an evenlonger time period.

The specific timeframe of the assess-ment instrument used should dependon the particular goals of the researcheror clinician administering the assessment.State measures are useful for assessingpatients’ craving experiences at specifictime points (e.g., during treatment ses-sions that expose the patients to alco-hol or use other cognitive-behavioralapproaches) and for determining changesin craving over time. Such immediate

assessments of craving are particularlyimportant in experimental settings,because they help researchers to betterunderstand the neurological, biochemi-cal, psychophysiological, cognitive,subjective, and emotional mechanismsinvolved in craving.

State measures may be less useful,however, in analyses assessing the rela-tionship between craving and generalalcohol use behavior. Studies typicallyhave found only insignificant or weakrelationships between the strength ofsubjective craving at a given time pointand concurrent or subsequent alcoholuse behavior. This lack of a correlationmay result from the presence of multi-ple factors that may differentially influ-ence subjective ratings and behavior.Examples of such factors include thepatient’s internal state (e.g., mood orwithdrawal symptoms), the presence of environmental cues related to pastdrinking, the perceived availability ofalcohol, and the patient’s current moti-vation to consume alcohol.

In contrast to state measures of crav-ing, global craving assessments tend toshow a stronger relationship with actualalcohol use. One potential explanation forthis stronger correlation is that alcohol-dependent people may be able to resistdrinking during brief instances of crav-ing, but cravings that occur frequentlymay have a cumulative impact ondrinking. Therefore, instruments thatask patients to integrate their cravingsover longer periods of time may offer a more reliable assessment of the extentto which cravings have caused distressor interference for the patient. Thismay, in turn, bear a closer connectionto whether the person will engage indrinking behavior.

Methods of AssessingAlcohol Craving

Assessment tools for craving generallyfall into two broad categories: (1) self-report instruments and (2) behavioraland psychophysiological measures. Self-report instruments typically are ques-tionnaires that can be either filled outby the patients themselves or adminis-tered by clinicians. These instruments

are frequently used in both researchand clinical settings. Behavioral andpsychophysiological measures are pri-marily used in experimental settings.

Self-Report Instruments

Self-report is the most frequently usedmethod to obtain information aboutcraving. In everyday clinical practice,therapists usually administer single-item instruments on which the patientreports his or her level of subjectivecraving. These instruments includequestions such as “How strong is yourcraving for alcohol?” or “How strong is your urge to drink?” The therapistalso provides anchor statements, suchas “not at all” and “the most I’ve everfelt.” The patient then rates his or hercraving either by selecting the mostappropriate choice on a 7- or 10-pointscale (i.e., a Likert-type scale) or byindicating a vertical mark along a linethat connects the two anchor statements(i.e., a visual analog scale [VAS]). WithVAS scales, the distance between the“no-craving” end of the line and thepatient’s mark serves as the index of crav-ing. For both types of assessment, thepatient may be asked to rate his or hercurrent level of craving or average levelof craving over a longer time period.

Brief Likert-type and VAS scalesprovide a straightforward and time-effective approach to assessing a patient’slevel of subjective craving. However, asnoted earlier, these assessments are limitedin their ability to provide informationabout the multiple elements that candefine the craving experience (Tiffany1992). Furthermore, researchers cannotdetermine the internal consistency ofthe instrument used. Because of theselimitations, brief Likert-type or VASratings should be supplanted withmulti-item instruments with desirablepsychometric properties whenever pos-sible. Several such instruments aredescribed in the following paragraphs(also see table on p. 182).

The Yale-Brown Obsessive CompulsiveScale for Heavy Drinking (Y–BOCS–hd). This scale, which was adapted fromthe Yale-Brown Obsessive CompulsiveScale (Y–BOCS) (Goodman et al.

Vol. 23, No. 3, 1999 181

Assessing Craving for Alcohol

182 Alcohol Research & Health

Summary of Questionnaires for Assessing Alcohol Craving

Source Reference PsychometricInstrument (Supporting References) Description Administration Properties Reported

Yale-Brown Modell et al. 1992a Ten items in a Administered by Good psychometric Obsessive (Goodman et al. 1989a) structured interview; a trained clinician; properties reportedCompulsive subscale scores for 15- to 30-min interview on parent instrument Scale for heavy alcohol obsessions (Y–BOCS); discrimi-drinkers and compulsive nant validity of (Y–BOCS–hd) behaviors related Y–BOCS–hd sup-

to alcohol can ported by high sensi-be obtained tivity and specificity

in correctly classify-ing alcoholics andnormal controls

Obsessive Anton et al. 1995 Fourteen items; Self-administered High internal consis-Compulsive (Kranzler et al. 1999; measure thoughts about questionnaire; requires tency (0.86); moderateDrinking Scale Roberts et al. 1999) alcohol during 5 to 10 min to complete to high internal con-(OCDS) nondrinking period; can sistency for individual

be scored for multiple subscales and factorssubscales and factors (0.71–0.85); concur-

rent validity evidencedby strong correlationswith measures of alcohol dependence severity and Y–BOCS–hd; good predictive validity shown by relationshipbetween subscale scores and midtreat-ment drinking

Alcohol Urge Bohn et al. 1995 Unidimensional Likert- Self-administered High internal consis-Questionnaire (AUQ) type scale with eight questionnaire; tency (0.91); high test-

items to assess requires 1 min to retest reliability (1-dayacute craving complete interval, 0.82); constuct

validity evidenced by strong correlationswith measures of alcohol dependence severity and the OCDS

Alcohol Craving Singleton et al. 1994 General form contains Self-administered Moderate to strongQuestionnaire (ACQ) (Love et al. 1998) 47 items and short form questionnaire; general factor (0.75–0.97)

contains 12 items; form requires 10 min and subscale both forms provide and short form less than (0.77–0.86) reliabilitymeasure of acute 5 min to completecraving and can be scored for multiplesubscales or factors

Min = Minute(s)NOTE: Discriminant validity indicates an instrument’s ability to accurately discriminate between populations with and without the quality of interest (e.g., alcoholics andnonalcoholics). Construct validity refers to the degree to which results from an instrument reflect the underlying quality (e.g., craving) that the instrument is trying to measure.Concurrent and predictive validity are types of criterion-oriented validity; they both describe the degree to which the results obtained with an instrument agree with otherdevices that purport to measure the same characteristic. Concurrent validity examines the relationship between measures taken around the same time, and predictive validityevaluates the relationship between the instrument and another measure taken at a later time. The reliability of an instrument indicates the degree to which items on an instru-ment yield consistent results. Test-retest reliability refers to the consistency of results when a subject is tested on multiple occasions. Internal consistency reliability indicatesthat similar items on a test do not yield inconsistent or contradictory responses. Reliability estimates can range from 0 to 1.0, with 1.0 indicating the highest degree of reliability.

1989a,b), was the first multi-iteminstrument developed and validated forthe specific purpose of measuring crav-ing for alcohol (Modell et al. 1992a).The Y–BOCS–hd conceptualizes alco-hol craving as obsessions and compul-sions relating to alcohol consumption.The term “obsession” refers here to the frequency and intrusive nature ofthoughts about drinking, especially after1 or 2 days of abstinence, whereas theterm “compulsion” refers to the loss ofcontrol over drinking.

The Y–BOCS–hd consists of 10questions, of which 5 comprise an obses-sionality subscale and the remaining 5comprise a compulsivity subscale. Theinstrument is administered in a struc-tured clinical interview lasting approxi-mately 15 to 30 minutes.

In a study assessing the validity andreliability of the Y–BOCS–hd, Modelland colleagues (1992b) administeredthe questionnaire to 62 alcoholics. Theresults were then compared with ratingsfrom a single-item subjective cravingmeasure (i.e., “How much do you cravean alcoholic beverage when you’ve gonewithout a drink for 1 to 2 days?”). Thestudy found that for both the total scoreand the two subscale scores, a statisti-cally significant, albeit moderate, corre-lation existed between the ratingsobtained with the two instruments.

The Obsessive Compulsive DrinkingScale (OCDS). Anton and colleagues(1995) modified the Y–BOCS–hd to derive the Obsessive CompulsiveDrinking Scale (OCDS), a 14-itemquestionnaire that the patient can self-administer and complete in about 5minutes. Because the instrument is self-administered, each respondent is likelyto interpret the questions similarly eachtime that he or she completes the ques-tionnaire, thereby improving test-retestreliability and eliminating or reducinginterviewer bias and differences in inter-pretation between interviewer andrespondent. The OCDS is a globalmeasure in which patients are asked torate their craving over a period of 1 or2 weeks (but no less than 1 day).

Like the Y–BOCS–hd, the OCDScontains an obsessive subscale, whichconsists of eight items, and a compulsive

subscale, which consists of six items.However, the questions in the obsessivesubscale on the OCDS relate to theoccurrence of intrusive thoughts aboutalcohol at any time when the respondentis not drinking and thus encompass amore general timeframe than that spec-ified in the Y–BOCS–hd. Patientsrespond to each item by selecting oneof five statements that range from min-imal to maximum endorsement of theitem. The OCDS has demonstratedgood test-retest reliability and highinternal consistency; furthermore, itsscores are strongly and significantlycorrelated with ratings obtained withthe Y–BOCS–hd (Anton et al. 1995).The OCDS also has been shown to bea valuable tool for outcome assessmentand for monitoring a patient’s progressduring treatment (Anton et al. 1996).

The Alcohol Craving Questionnaire(ACQ). The self-administered ACQ(Singleton et al. 1995) contains 47 items,each of which is scored on a 7-pointLikert-type scale ranging from “stronglydisagree” to “strongly agree.” Each itemis related to one of five domains thatare considered relevant to alcohol craving:(1) desire to drink alcohol, (2) intentionto drink alcohol, (3) lack of controlover the use of alcohol, (4) anticipationof positive effects from drinking (i.e.,positive outcome expectancy), and (5)expectancy of relief from withdrawal oralcohol’s negative effects. The ACQ is astate measure providing an index of acutecraving, because the questions relate tothe degree to which the respondent iscurrently experiencing these urges.Initial validation work with the ACQrevealed four dimensions with moder-ate to high internal consistency. Thesefactors were labeled emotionality, pur-posefulness, compulsivity, and expectancy(Singleton and Gorelick 1998). TheACQ also has been modified into ashort form, the ACQ–SF, that containsthe 12 items most strongly correlatedwith the total ACQ score.

The Alcohol Urge Questionnaire(AUQ). The AUQ (Bohn et al. 1995)is an eight-item, self-administered statemeasure that assesses the patient’s urgefor an alcoholic drink at the time the

questionnaire is completed, therebyproviding an index of acute craving. Aswith the ACQ, the items are scored alonga seven-point Likert-type scale. TheAUQ contains four items pertaining tothe desire for a drink, two items regard-ing expectations of positive effects fromdrinking and two items relating to theinability to avoid drinking if alcoholwere present. The AUQ has demon-strated good test-retest reliability, bothafter a 1-day and 1-week interval.Furthermore, the instrument has shownsignificant (but moderate) positive cor-relations with the patient’s alcohol-dependence severity and with scores onthe OCDS, as well as a negative correla-tion with the length of time the patienthas been abstinent (Bohn et al. 1995).

Psychophysiological and Behavioral Measures

Researchers frequently use psychophys-iological and behavioral indices in con-junction with self-report measures toelucidate the full range of responsesthat may relate to craving. A majoradvantage of measuring these additionalresponse domains is that they providemore objective data than do self-reportsand thus may be influenced less by var-ious sources of bias. In addition, manycraving theories make specific predictionsregarding the physiological responsesthat should accompany craving reportsand the relationships that should beobserved across various indices of crav-ing. The assessment of behavioral andpsychophysiological variables can helpsupport or disprove these predictions.

In laboratory settings, psychophysi-ological and behavioral responses areoften measured within “cue reactivity”studies, which assume that the subject’sresponses to alcohol-related stimuli canreflect craving (for more informationon cue-reactivity studies, see the articlein this issue by Litt and Cooney, pp.174–178). Cues that have been used instudies of alcohol craving include thesight, smell, and taste of alcohol; picturesor videos of alcohol or alcohol-relatedscenarios, such as a barroom; the studyparticipant’s belief that he or she willconsume alcohol; and mental imageryof alcohol-related situations. Negative

Vol. 23, No. 3, 1999 183

Assessing Craving for Alcohol

mood manipulations, which typicallyinvolve the participant imagining beingin an unpleasant situation or being pre-sented with a stressor that will invokesuch a mood, also have been used as acue for alcohol craving in some studies.Unfortunately, these studies have rarelydemonstrated close relationshipsamong behavioral, psychophysiological,and subjective measures of craving.Alcoholics’ responses to craving-inductionprocedures appear to vary considerablyas a result of numerous individual dif-ferences and situational factors. None-theless, cue-reactivity measures haveprovided valuable information aboutthe correlates of craving. Significantly,some studies have shown that autonomicreactions to alcohol cue presentations,such as changes in skin conductance,3can predict later relapse to drinking(e.g., Drummond and Glautier 1994).

Psychophysiological Measures. Severalpsychophysiological variables have beenassessed in response to alcohol cues,including salivation, skin conductance,skin temperature, respiration, heart rate,and blood pressure. Carter and Tiffany(1999) recently analyzed the results ofnumerous studies (i.e., conducted a meta-analysis) of cue-reactivity effects on psy-chophysiological variables across a rangeof AODs. Consistent with the assump-tion that alcoholics exhibit increasedurges and cravings for alcohol in responseto alcohol cues, this analysis revealedthat alcoholics across studies demonstratesignificant increases in heart rate and skinconductance in response to alcohol cues.That is, most studies showed that alco-holics responded with greater autonomicresponses to alcohol cues than to controlcues. However, substantial variabilityexisted in the sizes of these effects amongthe studies included in this analysis, and itis important to note that the effects weregenerally larger for self-reports of crav-ing than for the psychophysiologicalsignals. The fact that most of these stud-ies included single-item ratings suggeststhat as alluded to earlier, these itemsmay indeed reliably assess craving inthe context of cue-reactivity studies.

One physiological measure that wasnot included in the meta-analysis byCarter and Tiffany (1999) was salivation.

Previous reviews have suggested thatsalivation increases reliably in the presenceof alcohol cues (Niaura et al. 1988;Rohsenow et al. 1990–1991). This find-ing was confirmed in another recent meta-analysis, which also indicated a moderateincrease in salivation among alcoholicsin response to alcohol cues relative tocontrol cues (Tiffany et al. in press).

Behavioral Measures. The behavioralmeasure most closely related to the con-cept of alcohol craving is alcohol con-sumption itself. Some researchers consideralcohol seeking and drinking during orafter treatment to reflect, at least in part,craving processes. Various direct mea-sures of drinking behavior can serve asoutcome variables during or after alco-holism treatment, including frequencyof drinking, number of drinks perdrinking occasion (i.e., drinking inten-sity), and latency to drink.4 More fine-grained behavioral analyses of drinkingpatterns also have been conducted.

Although researchers generally arereluctant to conduct studies involvingalcohol consumption with alcohol-depen-dent subjects, craving studies performedin laboratory rather than treatment set-tings occasionally include direct measuresof alcohol consumption (Drobes andAnton in press). For example, Davidsonand colleagues (1999) recently demon-strated that treatment with the medica-tion naltrexone5 for 1 week diminishedoverall alcohol consumption and drinkingspeed in a laboratory session amongyoung (i.e., with a mean age of 22years) heavy drinkers.

Some recent cue-reactivity studieshave incorporated indirect behavioralmeasures in lieu of actual drinking

behavior. For example, Tiffany (1990)proposed that craving involves mentalprocesses requiring conscious cognitiveefforts (i.e., nonautomatic cognitiveprocessing) and therefore should inter-fere with simultaneous activities thatalso require conscious cognitive effort(for more information on this model,see the article in this issue by Tiffany,pp. 215–224). To evaluate this hypoth-esis, Sayette and colleagues (1994) ran-domly assigned 24 alcoholic menadmitted for inpatient alcoholism treat-ment to hold either a glass of water ortheir preferred alcoholic beverage for 3 minutes and to sniff the contents atspecified intervals. Simultaneously, themen were instructed to respond bypressing a button whenever they hearda computer-generated tone. The reac-tion time on this button-pressing taskwas considered an indirect marker ofthe cognitive efforts associated withcraving. The study found that duringthe period when the subjects were askedto sniff the beverages, the men holdingthe alcoholic beverage exhibited signifi-cantly longer reaction times than themen holding a glass of water, suggest-ing that the men in the alcoholic bever-age group experienced greater craving.

Another indirect behavioral measurethat may potentially serve as an indicatorof alcohol craving is choice cue-viewingtime—that is, the length of time a per-son chooses to look at a stimulus. Withinstudies on motivation and emotion,choice cue-viewing time has been relatedto various subjective and physiologicalindicators of arousal (e.g., Lang et al.1993). Several other novel behavioralmeasurement approaches are currentlybeing developed in the alcohol-cravingresearch arena, some of which borrowmethodologies from research on otherdrugs and from the emotion and moti-vation research field.

Clinical Utility of Craving Measurements

Most alcoholics report that cravings area frequent and troublesome aspect oftheir addiction. Consequently, cliniciansshould monitor their patients’ cravings.Assessment and discussion of craving

184 Alcohol Research & Health

3The term “skin conductance” refers to the abilityof the skin to conduct weak electrical currents. Thisability varies depending on the amount of moistureon the skin and can therefore be used as a measureof sweat gland activity.

4The term “latency to drink” in this context refersto the elapsed time from when alcohol is first madeavailable to the point at which consumption begins.

5Naltrexone is a medication that has been approvedby the U.S. Food and Drug Administration for thetreatment of alcoholism when used in conjunctionwith psychosocial therapy. Although the exactmechanism of action of this agent is unknown, onehypothesis is that naltrexone may reduce craving.

may improve rapport between clini-cians and patients, thereby increasingthe relevance of therapeutic interactions.In everyday clinical practice, self-reportinstruments will likely continue to bethe most frequently employed type ofcraving measure, because behavioraland psychophysiological measures,despite their theoretical relevance,clearly extend beyond the resourcecapacities of most treatment settings.

Incorporating craving measure-ments into routine clinical practice canproduce several potential benefits. First,craving-related measures appear to pro-vide a good index of a patient’s overalllevel of alcohol dependence. Anton andDrobes (1998) reported that alcoholicsscored increasingly higher on the OCDSas a function of the severity of theirdependence and, consequently, level of treatment involvement. That is,OCDS scores increased in order fromnonalcoholics, to non-treatment-seek-ing alcoholics (i.e., patients with a low level of dependence severity), toalcoholics receiving outpatient therapy,to alcoholics in inpatient treatment(i.e., patients with the greatest depen-dence severity).

Second, craving assessment can pro-vide valuable information concerning apatient’s capacity to recognize and mon-itor internal states that may be relatedto his or her drinking. If a patient canbe taught to recognize his or her spe-cific level of craving, then that personmay be more likely to use learned cop-ing strategies at the appropriate time orto return for additional treatment sessionsas craving-inducing situations ariseduring recovery.

Third, clinicians may find it usefulto assess craving in order to recommendappropriate treatment alternatives. Forinstance, studies have suggested thatalcoholics who report higher levels ofcraving benefit the most from the med-ication naltrexone as an adjunct to psy-chosocial treatment components (Jaffeet al. 1996).

Finally, because craving measureshave been shown to predict alcoholconsumption during active treatment(Anton et al. 1995), regular monitoringof craving throughout treatment mayaid clinicians and patients in decisions

regarding treatment intensity and dura-tion. For example, patients who con-tinue to experience high levels of crav-ing (either constantly or in certain situ-ations) despite compliance with theircurrent treatment may require contin-ued support, more intensive treatment,or possibly changes in their environ-ment (e.g., social network and activities).

The relationship between cravingand drinking outcome is still somewhatcontroversial in the alcohol field. Somestudies have demonstrated that self-reported craving can predict treatmentoutcome or relapse, whereas otherstudies report that acute craving oftendoes not lead to alcohol consumptionin either active or relapsing alcoholics.One explanation for this discrepancymay involve the types of measures usedto assess craving. It is possible that globalmeasures of craving are more stronglyrelated with drinking behavior than arestate-oriented measures, because globalmeasures involve the cognitive integra-tion of craving-related experiences overa longer time period. As a result, suchmeasures may more reliably incorpo-rate aspects of craving that cannot beobtained from state measures of crav-ing. In particular, global measures maybe better suited to assess the overallanguish and distress experienced by analcoholic, which may be related to thelikelihood of relapse. Conversely, state-oriented measures only give a snapshotof an alcoholic’s immediate alcohol-related struggle. Whether drinking occursat the time of the assessment, however,likely depends on several other factors.

Nevertheless, a pattern of high scoreson state-oriented measures of craving,while probably unrelated to drinking at a particular time, may show a similarpositive relationship with eventualdrinking as do more global measures.Although this hypothesis needs to betested empirically, its tenet is that clini-cians need to regularly administer bothstate-oriented and global measures of craving in order to estimate bothcurrent and general levels of alcohol-related distress. Such an approachcould help optimize clinical judgmentsregarding the probability of relapse andthe most appropriate interventions toprevent relapse.

Summary and FutureDirections

The past decade has witnessed thedevelopment and increased availabilityof psychometrically sound and theoret-ically relevant self-report measures ofcraving. The measures described in thisarticle are better able to capture thedynamic and multidimensional natureof craving across and within individualpatients over time compared with tra-ditional single-item assessments. Thesemeasures also can be used in a widerange of research and clinical settingsaccording to the needs of the adminis-trator. Additional work is necessary,however, to further refine and validatethese measures. In particular, the gener-alizability of the available measures tovarious types of drinking populationsmust be evaluated, because differentsubpopulations of alcoholics (e.g.,treatment-seeking, non-treatment-seek-ing, abstinent, and adolescent alco-holics) may require new and uniquecraving instruments. The currentlyavailable measures can serve as usefulguides in the development and valida-tion of such instruments.

Technological advances are likely toimprove the multidimensional mea-surement and understanding of cravingin the coming years. For example,researchers have begun to apply neu-roimaging techniques, such as func-tional magnetic resonance imaging(fMRI) and positron emission tomog-raphy (PET), in studies of craving (formore information on these techniques,see the article in this issue by Hommer,pp. 187–196). These new tools will becritical for identifying the brain regionsactivated concurrent with alcohol crav-ing, particularly when combined withother existing technologies for measur-ing craving. Improvements in existingcue-reactivity approaches also holdpromise for advancing measurementsof craving phenomena. For example,methodological advances within thegeneral study of emotion and motiva-tion could be adapted for alcohol-cuereactivity analyses in order to betterelucidate the role of emotions in alco-hol craving. Such investigations areparticularly important in light of recent

Vol. 23, No. 3, 1999 185

Assessing Craving for Alcohol

studies demonstrating a link betweenemotion, cue reactivity, and treatmentoutcome (Cooney et al. 1997; Litt etal. 1990). ■

References

ANTON, R.F., AND DROBES, D.J. Clinical measure-ment of craving in addiction. Psychiatric Annals28:553–560, 1998.

ANTON, R.F.; MOAK, D.H.; AND LATHAM, P. TheObsessive Compulsive Drinking Scale: A self-ratedinstrument for the quantification of thoughts aboutalcohol and drinking behavior. Alcoholism: Clinicaland Experimental Research 19(1):92–99, 1995.

ANTON, R.F.; MOAK, D.H.; AND LATHAM, P. Theobsessive compulsive drinking scale: A new methodof assessing outcome in alcoholism treatment studies.Archives of General Psychiatry 53(3):225–231, 1996.

BOHN, M.J.; KRAHN, D.D.; AND STAEHLER, B.A.Development and initial validation of a measure ofdrinking urges in abstinent alcoholics. Alcoholism:Clinical and Experimental Research 19(3):600–606,1995.

CARTER, B.L., AND TIFFANY, S.T. Meta-analysis ofcue-reactivity in addiction research. Addiction 94:327–340, 1999.

COONEY, N.L.; LITT, M.D.; MORSE, P.A.; BAUER,L.O.; AND GAUPP, L. Alcohol cue reactivity, negative-mood reactivity, and relapse in treated alcoholicmen. Journal of Abnormal Psychology 106:243–250,1997.

DAVIDSON, D.; PALFAI, T.; BIRD, C.; AND SWIFT,R. Effects of naltrexone on alcohol self-administra-tion in heavy drinkers. Alcoholism: Clinical andExperimental Research 23:195–203, 1999.

DROBES, D.J., AND ANTON, R.F. Drinking in alco-holics following an alcohol challenge research pro-tocol. Journal of Studies on Alcohol, in press.

DROBES, D.J.; SALADIN, M.E.; AND TIFFANY, S.T.Classical conditioning mechanisms in alcohol depen-dence. In: Heather, N.; Peters, T.J.; and Stockwell,T., eds. Handbook of Alcohol Dependence and Alcohol-Related Problems. Chichester, England: John Wiley& Sons, in press.

DRUMMOND, D.C., AND GLAUTIER, S. A con-trolled trial of cue exposure treatment in alcoholdependence. Journal of Consulting and ClinicalPsychology 62:809–817, 1994.

GOODMAN, W.K.; PRICE, L.H.; RASMUSSEN, S.A.;MAZURE, C.; FLEISCHMANN, R.L.; HILL, C.L.;HENINGER, G.R.; AND CHARNEY, D.S. The Yale-Brown Obsessive Compulsive Scale: I. Development,use, and reliability. Archives of General Psychiatry46:1006–1011, 1989a.

GOODMAN, W.K.; PRICE, L.H.; RASMUSSEN, S.A.;MAZURE, C.; DELGADO, P.; HENINGER, G.R.; AND

CHARNEY, D.S. The Yale-Brown ObsessiveCompulsive Scale: II. Validity. Archives of GeneralPsychiatry 46:1012–1016, 1989b.

JAFFE, A.J.; ROUNSAVILLE, B.; CHANG, G.;SCHOTTENFELD, R.S.; AND MEYER, R.E. Naltrexone,relapse prevention, and supportive therapy withalcoholics: An analysis of patient treatment match-ing. Journal of Consulting and Clinical Psychology64:1044–1053, 1996.

KRANZLER, H.; MULGREW, C.; MODESTO-LOWE,V.; AND BURLESON, J.A. Validity of the ObsessiveCompulsive Drinking Scale (OCDS): Does cravingpredict drinking behavior? Alcoholism: Clinical andExperimental Research 23:108–114, 1999.

LANG, P.J.; GREENWALD, M.K.; BRADLEY, M.M.;AND HAMM, A.O. Looking at pictures: Affective,facial, visceral, and behavioral reactions. Psychophysiology30(3):261–273, 1993.

LITT, M.D.; COONEY, N.L.; KADDEN, R.M.; AND

GAUPP, L. Reactivity to alcohol cues and inducedmoods in alcoholics. Addictive Behaviors 15(2):137–146, 1990.

LONG, A.; JAMES, D.; AND WILLNER, P. A compari-son of two alcohol craving questionnaires. Addiction93:1091–1102, 1998.

MODELL, J.; GLASER, F.; MOUNTZ, J.; SCHMALTZ,S.; AND CYR, L. Obsessive and compulsive charac-teristics of alcohol abuse and dependence:Quantification by a newly developed questionnaire.Alcoholism: Clinical and Experimental Research16(2):266–271, 1992a.

MODELL, J.G.; GLASER, F.B.; CYR, L.; AND MOUNTZ,J.M. Obsessive and compulsive characteristics ofcraving for alcohol in alcohol abuse and dependence.Alcoholism: Clinical and Experimental Research16(2):272–274, 1992b.

NIAURA, R.S.; ROHSENOW, D.J.; BINKOFF, J.A.;MONTI, P.M.; PEDRAZA, M.; AND ABRAMS, D.B.Relevance of cue reactivity to understanding alco-hol and smoking relapse. Journal of AbnormalPsychology 97:133–152, 1988.

ROBERTS, J.S.; ANTON, R.F.; LATHAM, P.K.; AND

MOAK, D.H. Factor structure and predictive valid-ity of the Obsessive Compulsive Drinking Scale.Alcoholism: Clinical and Experimental Research23(9):1484–1491, 1999.

ROHSENOW, D.J.; CHILDRESS, A.R.; MONTI, P.M.;NIAURA, R.S.; AND ABRAMS, D.B. Cue reactivity inaddictive behaviors: Theoretical and treatmentimplications. International Journal of the Addictions25(7A–8A):957–993, 1990–1991.

SAYETTE, M.A.; MONTI, P.M.; ROHSENOW, D.J.;GULLIVER, S.B.; COLBY, S.M.; SIROTA, A.D.;NIAURA, R.; AND ABRAMS, D.B. The effects of cueexposure on reaction time in male alcoholics.Journal of Studies on Alcohol 55:629–633, 1994.

SINGLETON, E.G., AND GORELICK, D.A. Mechanismsof alcohol craving and their clinical implications.In: Galanter, M., ed. Recent Developments in Alcohol-ism: Volume 14. The Consequences of Alcoholism.New York: Plenum Press, 1998. pp. 177–195.

SINGLETON, E.G.; TIFFANY, S.T.; AND HENNINGFIELD,J.E. Development and validation of a new ques-tionnaire to assess craving for alcohol. In: Harris,L.S., ed. Problems of Drug Dependence, 1994:Proceedings of the 56th Annual Scientific Meeting,The College on Problems of Drug Dependence, Inc.National Institute on Drug Abuse ResearchMonograph Series. Vol. 153. Rockville, MD: theInstitute, 1995. p. 289.

TIFFANY, S.T. A cognitive model of drug urges anddrug use behavior: Role of automatic and nonauto-matic processes. Psychological Review 97:147–168,1990.

TIFFANY, S.T. Critique of contemporary urge andcraving research: Methodological, psychometric, andtheoretical issues. Advances in Behaviour Researchand Therapy 14:123–139, 1992.

TIFFANY, S.T.; CARTER, B.L.; AND SINGLETON,E.G. Challenges in the manipulation, assessmentand interpretation of craving relevant variables.Addiction, in press.

186 Alcohol Research & Health