Professional Documents
Culture Documents
Cognitive Concepts of Craving: Stephen T. Tiffany, PH.D
Cognitive Concepts of Craving: Stephen T. Tiffany, PH.D
Traditional models of craving have been based primarily on the concept of classical
conditioning. In recent years, however, researchers increasingly have introduced cognitive
concepts, such as memory, expectancies, interpretation, and automatic behavior, into their
conceptualizations of craving. These efforts have culminated in the development of four
cognitive models of craving: cognitive labeling, outcome expectancy, dual-affect, and
cognitive processing. The cognitive processing model posits that although many alcohol use
behaviors have become automatized processes in the course of an alcoholic’s drinking
career, craving is a nonautomatic process that requires mental effort and is limited by a
person’s cognitive capacity. This model also implies that alcohol use and alcohol-seeking
behavior can occur in the absence of craving. In addition to introducing various new
concepts and models into craving research, the cognitive sciences also offer well-established
methodologies for testing these models and analyzing craving processes. KEY WORDS: AOD
(alcohol and other drug) craving; scientific model; cognition; research method and evaluation;
conditioning; expectancy; emotion; memory; AOD use behavior; literature review
I
magine that you are an alcoholic races and your hands sweat; you realize mechanisms (for a review of the classi-
trying to quit drinking. You have that this is craving at its worst. You are cal conditioning model, see sidebar, p.
not had a drink in a month, but drawn inexorably into the bar. There is 216, and Tiffany 1995a).
during the past several days, you have no way you can fight it any longer; you Numerous components of the sce-
thought about alcohol constantly. These must have a drink. nario described previously, however,
thoughts occupy your mind, making it Although fictional, this situation is go beyond simple conditioning pro-
nearly impossible to concentrate on not farfetched. In fact, many alcoholics cesses. For example, the fictional scene
anything else. Everything around you will describe in vivid detail similar sto- includes descriptions of alcohol-specific
seems to invoke memories of how pleas- ries about craving and relapse (Ludwig memories, positive expectancies about
ant and satisfying drinking can be. You 1988). The conventional explanation alcohol use, difficulties in concentration,
have wrestled with the idea of having a for this scenario, based on classical con- decisions about drinking, attention
drink, but you have decided to wait at ditioning models, is relatively straight- focused on alcohol cues, interpretations
least a little longer. Today, however, forward: Over a long history of drinking, of physiological reactions, and auto-
after leaving work, you find yourself stimuli and events routinely paired with matic behavior (i.e., automaticity),
somewhat mindlessly driving by your alcohol consumption (e.g., the sight of all of which are cognitive concepts.
favorite bar. You cannot help but notice a bar) become conditioned stimuli—
the front door of the bar propped open, that is, they induce the same responses
seeming to beckon you inside. You pull that are produced by alcohol itself. These STEPHEN T. TIFFANY, PH.D., is a professor
over to the curb, park your car, and conditioned stimuli activate conditioned in the Department of Psychological Sciences,
find yourself standing at the door. As motivational states1 that produce crav- Purdue University, and a member of the
you look through the doorway, it is all ing experiences, physiological reactions, Purdue University Neurosciences Program,
so familiar: The bar stools, the televi- and alcohol-seeking behaviors. Thus, West Lafayette, Indiana.
sion flickering in the corner, and even all the events described in the opening
the smell of stale cigarette smoke are paragraph could be viewed as the con- 1
For a definition of this and other technical terms
comfortable and inviting. Your heart sequences of classical conditioning used in this article, see glossary, p. 223.
cited cognitive processes as central to cognitive science paradigm, which focuses craving exist (see table, p. 218). These
the development of craving, they did on information processing, cognitive include two cognitive-behavioral mod-
not elaborate on how those cognitive architectures, memory, and decision- els—the cognitive labeling model and
processes might operate. making, is more clearly representative of the outcome expectancy model—and
Other more recent models describe contemporary cognitive psychology. Until two cognitive science models—the dual-
the role of cognition in craving somewhat recently, however, this paradigm did not affect model and the cognitive process-
more specifically. This article reviews substantially influence craving research. ing model. Each model offers distinct
selected cognitive models of craving and Both the cognitive-behavioral and examples of how various cognitive con-
discusses their implications for craving information-processing approaches to cepts may help explain craving and
research and assessment. Although this craving differ from the conventional may therefore complement each other.
review focuses on alcohol craving, nei- conditioning perspectives in several ways. The four models differ considerably in
ther cognition-based nor conditioning- Most important, traditional condition- the extent to which they utilize con-
based contemporary craving models ing models conceptualize craving as a temporary cognitive science theories.
are unique to alcoholism research. somewhat biologically primal, homoge- Indeed, one approach—the cognitive
Consequently, this review also draws neous state that directly represents the labeling model—uses concepts that are
from craving research on other drugs. fundamental motive for AOD use no longer widely accepted in the cogni-
in the addict. Conversely, cognitive tive sciences. Nevertheless, this model is
approaches consider craving the prod- included here because it still is invoked
Cognition and Craving uct of higher order mental functions. occasionally when researchers speculate
Research Thus, from the cognitive perspective, about the cognitive features of craving.
craving is not a primitive motivational
The increasing emphasis on cognition state but a complex, multidimensional
in craving research reflects the aftermath process that reflects how AOD-relevant
The Cognitive Labeling Model
of the “cognitive revolution” that swept information controls an addict’s behavior. One problem associated with conven-
the social sciences, particularly psychol- Furthermore, cognitive models contend tional conditioning models is explain-
ogy, in the 1970s. Since then, cognitive that craving processes can be measured ing how conditioned responses are
perspectives have dominated many of in ways not envisioned by traditional translated into craving experiences. For
the major fields of psychological research. conceptualizations of craving. (For example, assume that an alcoholic pre-
Even basic models of classical condi- more information on cognition-based sented with the sight of his or her
tioning, which are frequently invoked approaches to measuring craving, see favorite bar responds to the situation
in craving research, have increasingly the section “Craving and Cognitive with a racing heart and sweaty palms.
incorporated cognitive perspectives (e.g., Methodology,” p. 222.) How can these supposed conditioned
Rescorla 1988). For example, many mod- These two perspectives on craving autonomic responses generate craving?
ern theories of classical conditioning (i.e., cognitive-behavioral and informa- Some craving theorists have suggested
have adopted an information-process- tion processing) also have somewhat that conditioned responses are con-
ing perspective. This approach posits different implications for the development verted, in some unspecified way, into
that the learning of an association of interventions to prevent or treat craving. experiences of craving (e.g., Poulos et al.
between stimuli depends on the extent For example, if, as suggested by the 1981). Other researchers have hypothe-
to which the occurrence of one stimu- conventional view, craving is a biologi- sized that craving experiences arise
lus provides information about the cally primal state, it makes sense to seek from some form of cognitive interpre-
occurrence of the other stimulus. biological treatments (e.g., pharmaco- tation of conditioned reactions (e.g.,
When considering the impact of logical therapy) that directly target this Ludwig and Wikler 1974; Melchior
cognitive concepts on craving research, motivational state. Conversely, if crav- and Tabakoff 1984). This latter
it is useful to distinguish between two ing arises from the operation of dynamic, hypothesis has resulted in the formula-
types of cognitive theories: cognitive- multidimensional information-processing tion of the cognitive labeling model.
behavioral models and the cognitive systems, attempts to reduce craving might The cognitive labeling model of
science paradigm. Cognitive-behavioral be most successful if they target the craving is a variant of Schachter and
models (also called social learning cognitive processes regulating craving. Singer’s (1962) cognition-arousal theory
models) emphasize such constructs as of emotion. According to that theory,
expectancies, attributions, imitation, an emotional experience results from
and self-efficacy in the control of drink- Cognitive Models the interaction between physiological
ing behavior (e.g., Marlatt and Gordon of Craving arousal and a cognitive interpretation
1985). These models, which draw heav- of the arousal. The interpretation pro-
ily from the social and personality psy- Many researchers have speculated that vides an emotional label that deter-
chology of the 1960s and 1970s, are cognition plays an important role in mines the quality (e.g., pleasant or
relatively well represented in studies of generating craving. To date, however, unpleasant) of the emotional state. The
alcohol dependence. Conversely, the only four detailed cognitive models of intensity of that emotion is determined
Cognitive labeling model Craving is an emotion generated after exposure to alcohol-related cues that lead to both physical
arousal and a cognitive response identifying the arousal as craving; the extent of craving depends
on the extent of the arousal.
Outcome expectancy model Craving is generated after exposure to environmental alcohol-related cues that trigger positive
expectations about alcohol’s effects.
Dual-affect model Craving can be generated by both negative emotional systems (e.g., negative emotional states,
aversive events, and withdrawal) and positive emotional systems (e.g., positive emotional states
and consumption of small alcohol doses). Positive- and negative-affect craving are mutually
exclusive; the extent of craving depends on the extent to which positive- or negative-affect systems
are activated.
Cognitive processing model Craving represents a nonautomatic cognitive process that is activated when the execution of
automatized drinking behavior is voluntarily or involuntarily blocked. Craving-inducing situations
require cognitive processing and mental efforts and may thereby interfere with other cognitively
demanding tasks. Craving is not required for either alcohol seeking or alcohol use.
sufficient for causing alcohol consump- both alcoholics and nonalcoholics to thought to be mutually inhibitory—
tion, because a person can desire alco- drinking cues and subsequently mea- that is, stimulation of one system
hol without intending to drink. sured the respondents’ changes in out- would suppress activation of the other
Since Marlatt’s (1985) development come expectancies and craving. The system. Consequently, the model
of the original outcome expectancy study found some evidence that cue assumes that a person cannot experi-
model, researchers have further elabo- exposure enhanced craving and increased ence both positive- and negative-affect
rated on the cognitive aspects of this expectancies of positive outcomes in craving simultaneously.
theory (e.g., Goldman and Rather 1993). both subject groups. However, the According to the dual-affect model,
For example, these researchers use sev- expectancy effects were not found con- craving systems are structured at a cog-
eral contemporary concepts of memory sistently across all measures, and some nitive level into propositional net-
and information processing to describe aspects of the findings could even be works—memory networks that store
the mental processes that control alco- interpreted as being inconsistent with information on the stimuli that trigger
hol expectancies. These reformulations the expectancy model (see Tiffany the craving system, AOD-related
have not, however, specifically addressed 1995b). Nevertheless, to date, too little responses, and the meaning or interpre-
the craving aspects of Marlatt’s original research exists on expectancies and tation of stimuli and responses. A given
theory.3 Thus, expectancy research gen- craving for AODs to allow meaningful network is activated when environmental
erally has focused on the beliefs that conclusions about the validity of the stimuli match the stimulus information
people have about alcohol’s effects and outcome expectancy model. stored in that network.
on the associations of those beliefs with The extent of the activation (and,
drinking behavior. Few studies have consequently, craving) depends on the
examined the relationship between
The Dual-Affect Model extent to which multiple cues match
craving and situation-specific outcome Baker and colleagues (1987) proposed the triggering stimuli stored in the
expectancies. that craving is controlled by complex propositional network. For example,
In Marlatt’s theory, positive outcome emotion-processing systems that influ- an alcoholic in a positive, celebratory
expectancies about alcohol should be ence physiological responses, self-reports mood may experience some alcohol
more readily triggered in some settings of craving and emotion, and drug-seek- craving. The same alcoholic also may
than in others. Thus, certain situations ing behavior. The researchers posited experience craving when presented
that are routinely associated with alco- that craving can reflect the operation of with the opportunity to consume an
hol consumption, such as being in a both negative and positive emotion, or alcoholic beverage. The intensity of the
bar setting, would be much more likely affect, systems. craving, however, may differ in each
to activate positive outcome expectan- Negative-affect craving can be trig- situation. Furthermore, the person’s
cies about the effects of alcohol than, gered by a negative emotional response craving will be considerably stronger
for example, being in church. Further- or feeling (e.g., depression or anger), if he or she is in a positive mood and
more, researchers have not investigated aversive events (e.g., having to give a is given the opportunity to drink.
specifically the proposed distinction presentation to a group of strangers), The activation level of a network
between craving and urges (Tiffany AOD withdrawal, cues paired with also determines the consistency of the
1995b). previous withdrawal episodes, and various responses generated by that
If the outcome expectancy model is information that the drug is not avail- network. That is, as a network becomes
correct in assuming that expectancies able. When activated, the negative- more fully engaged, stronger correla-
control alcohol consumption, then a affect system induces craving experi- tions should exist between the extent of
correlation should exist between the ences, drug-seeking behavior, negative self-reported craving and the extent of
extent of the expectancies and the extent affect, and physiological reactions that drug-seeking behavior. Finally, the
of alcohol intake. Indeed, several stud- mimic withdrawal. dual-affect model predicts that partial
ies have demonstrated that expectancies Conversely, positive-affect craving is activation of a propositional network
about alcohol’s effects tend to correlate associated with positive emotions and should lower the threshold for further
significantly with general and situation- with pleasurable or positively reinforc- activation of that network. For exam-
specific alcohol consumption levels ing AOD effects. This craving system is ple, an alcoholic in a celebratory mood
(Goldman and Rather 1993). activated by positive emotional states, a (i.e., whose craving system is partially
Cooney and colleagues (1987) con- small AOD dose, cues paired with AOD activated) should respond more
ducted a study in which they exposed use, and information that the drug is strongly to the presentation of alcohol
available. Activation of this system cues than would an alcoholic in a neu-
3
Indeed, in their elaboration of expectancy theory, induces craving, drug-seeking behavior, tral mood (i.e., whose craving system is
Goldman and Rather (1993) have adopted the positive affect, and physiological responses not activated).
position suggested by Tiffany (1990) that the moti- that mimic the stimulating AOD effects. Unlike other cognitive models of
vational control of alcohol use in the highly practiced
alcoholic may be automatized and not reflected in An important feature of the dual- craving, the dual-affect model has been
reports of craving (also see section “The Cognitive affect model is that both positive- and developed sufficiently to allow
Processing Model,” p. 220). negative-affect craving systems are researchers to generate and test fairly
“Others essentially think instinctively, they are faced with cues and situations have strongly supported the core assump-
short circuiting both imagery and cog- that trigger portions of their automatic tion of this model—that craving is not
nitions, and are inclined to act without drinking behaviors; consequently, they necessary for drug seeking or AOD use
knowing why. When alcohol becomes must exert considerable mental effort (Tiffany 1990, Tiffany and Carter 1998;
readily available, they drink before they to keep themselves from drinking. The Tiffany and Conklin in press). Neither
think.” constant battles with craving may leave studies of craving and alcohol con-
In contrast to automatic processing, the abstinent alcoholic mentally exhausted sumption in the laboratory nor studies
nonautomatic processing can be char- and unable to cope effectively with the of craving and relapse in the real world
acterized as slow, flexible, intention cognitive demands of everyday life. The have provided strong evidence that
dependent, requiring cognitive effort, cognitive effort associated with absti- craving is directly responsible for alcohol
and restricted by limited cognitive nence may be sustainable as long as no use in alcoholics (Tiffany and Conklin
capacity. Nonautomatic processes occur additional problems arise (e.g., in the in press). For example, the alcohol
under three kinds of circumstances: (1) workplace or in relationships with fam- amounts consumed by alcoholics in
when a person first learns a skill, (2) ily and friends). In stressful, challeng- laboratory studies are only weakly cor-
when a highly automatized sequence is ing situations (e.g., when faced with a related with the alcoholics’ descriptions
activated but some environmental tight deadline at work), however, the of craving. Moreover, studies of relapse
obstacle blocks the completion of that abstinent alcoholic may find it easier to episodes indicate that addicts rarely
sequence, and (3) when a person wants succumb to alcohol craving and take a identify craving as a major direct trig-
to prevent the execution of activated drink rather than struggle with both ger for resuming AOD use (Tiffany
automatized sequences (Shiffrin and craving and work-related pressure. and Carter 1998).
Schneider 1977). The cognitive pro- Overt behavior, self-reports of craving Numerous studies also have sup-
cessing model hypothesizes that craving and emotional distress, and autonomic ported the cognitive processing model’s
represents nonautomatic processes physiological responses are typical reac- prediction that craving activation
that are activated simultaneously with tions that result from the activation of should disrupt cognitive functioning
automatized AOD use sequences. The craving processes. The cognitive pro- (Sayette in press). For example, Sayette
nonautomatic processes are activated cessing model, however, proposes that and colleagues (1994) exposed people
either to overcome obstacles to success- craving may influence a fourth category receiving inpatient alcoholism treat-
fully completing automatized AOD use of responses not typically addressed in ment to both alcohol-related cues (e.g.,
or to prevent the execution of an autom- craving studies—that is, responses to a glass of their favorite alcoholic bever-
atized sequence. These two situations the cognitive demands of craving. age) and neutral cues (e.g., a glass of
generate, respectively, the craving observed According to the model, a craving- water). During cue exposure, the sub-
in alcoholics who are not attempting to inducing situation presents a problem jects were asked to press a button when-
quit drinking and the craving observed in that demands a solution. Furthermore, ever they heard a brief sound, and their
alcoholics who are attempting abstinence. the model posits that solving this prob- reaction times were recorded. The study
The first scenario is exemplified by lem requires nonautomatic cognitive found that reaction times were slower
the alcoholic who goes to his or her processes that function at the expense in the presence of alcohol-related cues
favorite bar and discovers that it is closed of other activities that also necessitate than in the presence of neutral cues,
for remodeling. Resolution of this unex- nonautomatic processing. Because suggesting that the alcoholics’ ability to
pected inconvenience requires nonau- nonautomatic processes involve mental perform the button-pressing task was
tomatic processes, causing changes in effort that exhausts limited cognitive disrupted by the cognitive processing
overt behavior, craving and emotional resources, the alcoholic trying to solve triggered by the alcohol cues. These
distress, and physiological responses. craving-associated problems will have findings support the cognitive process-
For example, the alcoholic will take little mental capacity left for coping ing model’s hypothesis that craving is
action to overcome the obstacle (e.g., with other cognitively demanding situ- cognitively disruptive and therefore
drive to another bar). In addition, the ations. Thus, the nonautomatic cogni- exerts a toll on other cognitively
alcoholic will express desire for alcohol, tive demands of craving processes can demanding tasks.
intentions to find and consume alco- account for the disruptive impact of Beyond their implications for the
hol, and frustration and anger at being craving on daily functioning. cognitive processing model, this and
thwarted in his or her attempt to drink. The cognitive processing model similar studies demonstrate how the
The alcoholic also will likely display allows for several explicit predictions cognitively intrusive and disruptive fea-
physiological responses (e.g., changes in about the cognitive organization of tures of craving—features that addicts so
heart rate or sweat gland activity) that craving and AOD use, the behavioral often complain about—can be studied
reflect either the physical or cognitive manifestations of craving, and the role under controlled laboratory conditions.
demands of the craving situation. of automatic and nonautomatic pro- More generally, the studies illustrate
The second scenario represents the cesses in relapse. Many of these predic- the usefulness of measures and meth-
situation confronting alcoholics who tions have not yet been evaluated sys- ods derived from cognitive sciences for
are trying to quit drinking. Every day tematically; however, various studies investigating craving processes. These
GLOSSARY
Attributions: Inferences on the causes and effects of behavior. Expectancies: Expectations regarding the effects of a behavior
Autonomic responses: Physiological responses that are con- or a drug.
trolled by the autonomic nervous system (e.g., heart rate, Motivational state: A psychological state that initiates or modi-
blood pressure, and sweat gland activity). These responses fies a certain behavior (e.g., seeking and consuming alcohol
are not under conscious control. or other drugs).
Conditioned withdrawal syndrome: The experience of physio- Paradigm: A set of related theories or models.
logical symptoms (e.g., sweating, tremors, and anxiety) of Positively reinforcing effects: The effects of a drug (e.g., alco-
withdrawal from alcohol and other drugs that is not caused hol) that increase the probability that a person will attempt
by the actual withholding of the drug but by stimuli that to consume more of that drug.
have been associated with previous withdrawal episodes.
ficiently to allow for the generation of tions, memory organization, and per- chological characteristics of craving,
formal mathematical or computer formance-based indicators of mental alcohol researchers almost certainly will
models. The cognitive sciences, how- functions likely will elucidate many of need to apply state-of-the-art cognitive
ever, offer a standard of theorizing that the mechanisms underlying craving. processing models and methods. ■
if fully adopted by craving researchers, Although cognitive approaches to crav-
would represent a quantum leap in the ing have considerable potential, they
scientific study of craving processes. still have shortcomings. For example,
References
many of the predictions generated by BAKER, T.B.; MORSE, E.; AND SHERMAN, J.E. The
cognitive models have not been system- motivation to use drugs: A psychobiological analysis
Cognition and the Future atically evaluated. Moreover, some of of urges. In: Rivers, P.C., ed. The Nebraska Symposium
on Motivation: Alcohol Use and Abuse. Lincoln, NE:
of Craving Research the models are insufficiently developed University of Nebraska Press, 1987. pp. 257–323.
or invoke constructs that no longer
The cognitive models discussed in this represent contemporary research and BERRIDGE, K.C., AND ROBINSON, T.E. The mind
of an addicted brain: Neural sensitization of want-
article offer perspectives on craving theory in the cognitive sciences. ing versus liking. Current Directions in Psychological
processes that differ markedly from Despite these problems, the pres- Science 4:71–76, 1995.
conventional, conditioning-based con- ence of cognitive concepts, measures,
CARTER, B.L., AND TIFFANY, S.T. Meta-analysis of
cepts of craving. In some cases, such as and methods in craving research will cue-reactivity in addiction research. Addiction, in press.
the cognitive labeling hypothesis, the almost inevitably increase. This assertion
models merely place a cognitive spin is based at least in part on the observa- CEPEDA-BENITO, A., AND TIFFANY, S.T. The use of
a dual-task procedure for the assessment of cogni-
on basic conditioning approaches. tion that the theories of the behavioral tive effort associated with smoking urges. Psycho-
Conversely, some theories, such as the sciences in general increasingly incor- pharmacology 127:155–163, 1996.
cognitive processing model, depart rad- porate cognitive dimensions. Indeed,
COONEY, N.L.; GILLESPIE, R.A.; BAKER, L.H.; AND
ically from traditional assumptions the cognitive perspective has become KAPLAN, R.F. Cognitive changes after alcohol cue
about the organization and function of the dominant paradigm in most behav- exposure. Journal of Consulting and Clinical Psychology
craving processes. Despite their differ- ioral sciences. Craving research has 55:150–155, 1987.
ences, however, cognitive approaches always tracked—albeit with some DROBES, D.J., AND TIFFANY, S.T. Comparisons of
share common themes. Most important, delay—major trends in the behavioral smoking urges elicited through imagery and in vivo
these approaches posit that craving arises sciences. Beyond this “inevitability of cue exposure. Journal of Abnormal Psychology 106:
from the operation of information-pro- history,” however, cognitive approaches 15–25, 1997.
cessing systems. In other words, these will be welcomed into studies of alcohol GOLDMAN, M.S., AND RATHER, B.C. Substance use
models view craving not merely as a craving because of their potential for disorders: Cognitive models and architecture. In:
primitive drive state but as the product helping researchers disentangle complex Dobson, K.S., and Kendall, P.C., eds. Psycho-
pathology and Cognition. San Diego: Academic Press,
of higher order cognitive functions. craving data. For example, over the next 1993. pp. 245–292.
Cognitive models offer researchers few years, scientists will face entirely
the opportunity to study craving pro- new types of results generated by brain LONDON, H., AND NISBETT, R.E. Elements of
Schachter’s cognitive theory of emotional states. In:
cesses with measures that extend far imaging studies of craving. To under- London, H., and Nisbett, R.E., eds. Thought and
beyond the standard inventory of vari- stand the neurobiological correlates of Feeling. Chicago: Aldine, 1974.
ables suggested by conventional craving craving identified by such studies and
LUDWIG, A.M. Understanding the Alcoholic’s Mind:
theories. For example, studies of attri- to identify the association between The Nature of Craving and How to Control It. New
butions, outcome expectancies, inten- neurobiological features and the psy- York: Oxford University Press, 1988.
To order, write to: National Institute on Alcohol Abuse and Alcoholism, Publications Distribution Center, P.O. Box 10686,
Rockville, MD 20849–0686. Fax: (202) 842–0418. Full text is available on the World Wide Web at http://www.niaaa.nih.gov