ArticlePDF Available

Implicit measures of “wanting” and “liking” in humans



IST makes the clear prediction that (1) there should be a positive correlation between indices of “wanting” (e.g., drug consumption) and implicit “wanting” scores. Similarly, there should be a positive correlation between indices of “liking” (e.g., various expressions of subjective pleasure) and implicit “liking” scores; (2) there should be higher “wanting” scores in substance abusers or frequent substance users compared to non-users or infrequent users, and there should be no differences in “liking” between these groups (or even less “liking” in frequent substance users); (3) manipulations of “wanting” should affect implicit “wanting” scores whereas manipulations of “liking” should affect implicit “liking” scores. However, studies that tested these hypotheses did not produce equivocal results. To shed light on these discrepancies, we first discuss the different definitions of “wanting” and “liking” and the different tests that have been used to assess these processes. Then, we discuss whether it is reasonable to assume that these tests are valid measures of “wanting” and “liking” and we review correlational, quasi-experimental, and experimental studies that inform us about this issue. Finally, we discuss the future potential of implicit measures in research on IST and make several recommendations to improve both theory and methodology.
Implicit measures of “wanting” and “liking” in humans
Helen Tibboel1
Jan De Houwer1
Bram Van Bockstaele2,3
1 Department of Experimental-Clinical and Health Psychology, Ghent University, Belgium
2 Department of Developmental Psychology, University of Amsterdam, the Netherlands
3 Department of Child Development and Education, University of Amsterdam,
the Netherlands
Author for correspondence:
Department of Psychology, Ghent University, Henri Dunantlaan 2, B-9000 Ghent, Belgium
Phone: 0032 9 264 9143
Fax : 0032 9 264 64 89
Incentive Sensitization Theory (IST; e.g., Robinson & Berridge, 1993; 2003) suggests
that a common dopamine system that deals with incentive salience attribution is affected by
different types of drugs. Repeated drug use will sensitize this neural system, which means that
drugs increasingly trigger the experience of incentive salience or “wanting”. Importantly,
Robinson and Berridge stress that there is a dissociation between drug “wanting” (the
unconscious attribution of incentive salience) and drug “liking” (the unconscious hedonic
experience when one consumes drugs). Whereas the former plays an essential role in the
development and maintenance of drug addiction, the latter does not. Although this model was
based mainly on research with non-human animals, more recently the dissociation between
“wanting” and “liking” has been examined in humans as well. A widely used and promising
means of studying these processes are behavioral implicit measures such as the Implicit
Association Test (IAT), the Approach-Avoidance Task (AAT), different types of Stimulus-
Response Compatibility (SRC) tasks, and Affective Simon Tasks (AST).
IST makes the clear prediction that (1) there should be a positive correlation between
indices of “wanting” (e.g., drug consumption) and implicit “wanting” scores. Similarly, there
should be a positive correlation between indices of “liking” (e.g., various expressions of
subjective pleasure) and implicit “liking” scores; (2) there should be higher “wanting” scores
in substance abusers or frequent substance users compared to non-users or infrequent users,
and there should be no differences in “liking” between these groups (or even less “liking” in
frequent substance users); (3) manipulations of “wanting” should affect implicit “wanting”
scores whereas manipulations of “liking” should affect implicit “liking” scores. However,
studies that tested these hypotheses did not produce equivocal results. To shed light on these
discrepancies, we first discuss the different definitions of “wanting” and “liking” and the
different tests that have been used to assess these processes. Then, we discuss whether it is
reasonable to assume that these tests are valid measures of “wanting” and “liking” and we
review correlational, quasi-experimental, and experimental studies that inform us about this
issue. Finally, we discuss the future potential of implicit measures in research on IST and
make several recommendations to improve both theory and methodology.
1. Introduction
Many researchers have been baffled by the so-called addiction paradox. Why do
addicts continue to pursue and use drugs even though they are fully aware of the harmful
consequences of their behavior? One of the most influential theories that attempted to tackle
this question is Robinson and Berridge’s Incentive Sensitization Theory (IST; e.g., 1993;
2003; 2008). Robinson and Berridge (2000) note that earlier theories on addiction focussed
mainly on negative reinforcement (i.e., addicts want to avoid unpleasant withdrawal
symptoms and therefore continue to use drugs) and positive reinforcement (i.e., addicts
continue to use drugs because they enjoy the pleasurable experience; e.g., Solomon & Corbit,
1973). Such theories could not explain several important questions, such as why addicts often
relapse even though they no longer experience withdrawal symptoms, and why they continue
to use drugs even when the pleasurable effects have become small or non-existent.
IST suggests that in vulnerable individuals, intermittent drug use sensitizes specific
drug effects, whereas other effects habituate, and still other effects remain stable. According
to the model, two processes become sensitized: first, there is the sensitization of psychomotor
effects (i.e., habitual drug users become more active, aroused, and explorative, and they show
stronger general approach tendencies). Second, and most importantly, brain systems that are
involved in attributing a rewarding value (i.e., incentive salience) to stimuli are sensitized
also. This hypersensitivity means that conditioned stimuli that are associated with drug
administration become increasingly sought after and wanted, that these stimuli capture
attention, and that they become more arousing. This creates a dissociation between drug
“wanting” and drug “liking”. Drugs are no longer pursued because of their euphoric effects
(“liking”) but because of the incentive-motivational properties of drugs and drug-related
stimuli (“wanting”). IST suggests that incentive sensitization has lasting effects. It can thus
explain why so many addicts relapse long after their withdrawal symptoms have disappeared:
drug cues remain desired and highly salient even when the physical need to use drugs is
extinguished (e.g., Berridge & Robinson, 2003; Robinson & Berridge, 1993, 2001, 2003,
Even though usually “wanting” mirrors “liking” (i.e., we “want” and pursue things we
“like” and we avoid things we do not “like”), these processes do not need to overlap.
Winkielman and Berridge (2003) suggest that “wanting” may have evolved separately from
and earlier than “liking”. They distinguish two types of “wanting” that might have evolved at
different stages. First, they state that “wanting” was originally (i.e., before conscious
experience) an elementary and unconditioned means to make decisions about the pursuit of
innate rewards such as food, water, and mates. This type of “wanting” was unrelated to
“liking”. Second, “wanting” evolved to become the conditioned motivational process to
pursue incentives that were previously “liked”. This means that different brain mechanisms
are at play for “liking” and “wanting” and that consequently (and crucially) “wanting” can
occur without “liking”. Specific dopamine systems are assumed to play a role solely in the
attribution of incentive salience (“wanting”), whereas they do not play a role in hedonic
aspects of substance use (“liking”) (e.g., Robinson & Berridge, 2008). Within the nucleus
accumbens (NAcc), for instance, a “hedonic hotspot” has been identified that is assumed to be
involved in “liking” but not “wanting”, whereas other areas within the NAcc are assumed to
be involved in “wanting” but not (necessarily) “liking”. It should thus be possible to find
dissociations between the two processes (e.g., to observe “wanting” in the absence of
Behavioral evidence for “wanting” in animal studies is typically assessed by
measuring the extent to which an animal will choose, consume, or work for a specific
substance (e.g., Berridge & Robinson, 1998). Animal studies show that sensitized rats work
harder to obtain a drug reward, are quicker to learn conditioned place preference for locations
paired with drug cues, are more prone to relapse after receiving a priming dose of the
substance, and attach more value to other rewards as well (e.g., Robinson & Berridge, 2003).
Measures of “liking” include an assessment of affective facial expressions. For instance,
tongue protrusions are considered to be indices of hedonic pleasure, whereas gaping is
thought to indicate disliking.
Dissociations between “wanting” and “liking” in animals can be shown, for instance,
in studies on Pavlovian-Instrumental transfer (PIT). Wyvell and Berridge (2000) trained rats
to press one of two levers to obtain a reward (in this case, sucrose pellets). They also taught
the rats that a specific Pavlovian cue (Conditioned Stimulus; CS+) co-occurred with the
presentation of sucrose pallets (Unconditioned Stimulus; US). During a crucial test phase,
rewards were no longer presented, and lever-pressing was measured during the presence and
absence of the CS+. Results showed that rats pressed the lever more often (i.e., they worked
harder to obtain the substance, implying they “wanted” the substance more) when the CS+
was present, showing that this cue increased incentive motivation. This effect increased when
rats had received micro-injections of amphetamine, a dopamine agonist that is known to
increase instrumental performance. Importantly, Wyvell and Berridge showed that
amphetamine injections did not increase hedonic facial reactions, which implies that “liking”
was not affected by this manipulation, therefore providing evidence that “wanting” and
“liking” can indeed become dissociated.
Note that from a strict functional perspective (see Catania, 2013, for a review), there is no contradiction in
saying that a disliked stimulus can be a reinforcer. A stimulus qualifies as a reinforcer of a behavior whenever
the probability of the behavior increases as the result of the behavior-stimulus relation. Hence, the concept
“reinforcer” is defined merely in terms of a stimulus’ function (i.e., it increases the likelihood of behaviors to
which it is related). The concept “reward”, on the other hand, differs from the concept “reinforcer” in that it
points at a stimulus property that is supposed to explain why a stimulus functions as a reinforcer. Whereas it is
typically assumed that rewards have a reinforcing value because of their hedonic properties, Robinson and
It is important to note that apart from the enormous impact that Incentive Sensitization
Theory has had on addiction research, it has affected the study of other types of (health)
behaviors and psychopathology as well. For instance, dissociations between “wanting” and
“liking” have been assumed to play a role in food consumption, (e.g., Finlayson, King, &
Blundell, 2007; but see also Havermans, 2011), impulsive economic choice (e.g., Lades,
2014), depression (e.g., Bushmann, Moeller, Konrath, & Crocker, 2012), schizophrenia
(Heerey & Gold, 2007), compulsive sexual behavior (Voon et al., 2014), and autism (Kohls,
Chevallier, Troiani, & Schultz, 2012). Even though IST has provided valuable new insights in
drug addiction and a striking influence on the study of a broad range of behaviors, there is
uncertainty regarding the operation of incentive sensitization in humans. There is some
evidence for the role of sensitized activations in the mesolimbic dopamine system in addiction
(Leyton, 2007), but the question remains how to measure incentive processes behaviorally in
humans (e.g., Robinson & Berridge, 2000). An increasingly popular line of research focuses
on using so-called implicit measures to examine “wanting” and “liking” in humans (e.g.,
Wiers & Stacy, 2006). The aim of this paper is to examine the validity of behavioral implicit
measures of “wanting” and “liking”. Because a measure can be valid only to the extent that
there is clarity about the to-be-measured concept (Borsboom, Mellenbergh, & van Heerden,
2004; De Houwer, Teige-Mocigemba, Spruyt, & Moors, 2009), we first analyse the
definitions of “wanting” and “liking” as they are described in the IST literature. Next, we
discuss different behavioral measures that are assumed to assess “wanting” and “liking” in the
context of addiction. Subsequently, we examine the validity of implicit measures of
“wanting” and “liking”. Finally, we formulate suggestions for future research and theory
Berridge (2000) argue that, at least in some contexts, the reinforcing function of a stimulus is mediated by a non-
hedonic property which they refer to as “incentive salience”.
Our evaluation of the validity of implicit measures of “wanting” and “liking” was
inspired by the criteria proposed by Borsboom et al. (2004) and De Houwer et al. (2009), who
stress the importance of causality. From this perspective, measures of “wanting” and “liking”
can be considered as valid if measurement scores are causally determined by the degree of
“wanting” and “liking”. This type of validity can be assessed on the basis of studies using
correlational, quasi-experimental, and experimental designs. Although experimental designs
can provide the strongest evidence for causality, correlational and quasi-experimental designs
are not without merit for establishing whether the preconditions for causality are met (see De
Houwer et al., 2009; Van Bockstaele et al., 2014). Hence, we discuss only studies that
examine (one or more of) the hypotheses of IST: (1) there should be a positive correlation
between indices of “wanting” (e.g., drug consumption, drug craving; e.g., Wiers et al., 2002)
and “wanting” scores on a particular implicit measure (i.e., correlational designs). Similarly,
there should be a positive correlation between indices of “liking” (e.g., various expressions of
subjective pleasure) and “liking” scores on a particular implicit measure; (2) higher “wanting”
(but not necessarily “liking”)
scores in substance abusers or frequent substance users
compared to non-users or infrequent users (i.e., quasi-experimental designs), and no
differences in “liking” or even less “liking” in frequent substance users; (3) manipulations of
“wanting” should affect “wanting” scores (and not “liking” scores) whereas manipulations of
“liking” should affect “liking” scores (and not “wanting” scores; i.e., experimental designs).
This means that we will not discuss papers in which modified implicit behavioral tasks are
used to manipulate “wanting” or “liking” effects (e.g., Eberl et al., 2013), unless they include
(pre- and post-) assessments using the implicit measures of “wanting” and “liking” that are
It is important to note that according to IST, “wanting” and “liking” are commonly related (e.g., Berridge &
Robinson, 2003; Winkielman & Berridge, 2003). Only in extreme circumstances (i.e., when casual drug use has
evolved into an addiction) are “wanting” and “liking” dissociated. Thus, we would only expect dissociations
between “wanting” and “liking” measures in samples of drug-abusers (i.e., not in casual drug users).
central in this paper (e.g., Wiers, Eberl, Rinck, Becker, & Lindenmeyer, 2011). This also
means that we will not discuss papers in which other factors than “wanting” or “liking” are
manipulated (e.g., mindfulness, negative affect, cognitive load,; e.g., Ostafin, Bauer, &
Myxter, 2012; Ostafin & Brooks, 2011; Sharbanee, Stritzke, Jamalludin, & Wiers, 2014) and
their effect on “wanting” and “liking” measures is examined. The reason for this exclusion
criterion is that IST makes no clear predictions about how these manipulations should affect
“wanting” and “liking”. Thus, such studies do not inform us about the validity of implicit
measures of “wanting” and “liking”.
2. Definitions of “wanting” and “liking”
Berridge (1996) notes that it is not easy and maybe even uncalled for to demand a
complete and strict definition of reward. He therefore argues for a minimal definition of
reward that will grow as the amount of research on this topic increases. There is a vast
literature on the concepts of “wanting” and “liking” which reveals that the definition of these
concepts has changed considerably over the years. Based on an analysis of the literature, we
were able to distinguish between four possible definitions of the terms “wanting” and
“liking”. In the following section, we discuss each of these definitions. A summary can be
found in Table 1.
2.1. Mental process definition: “wanting” and “liking” as (preconscious) affective and
cognitive processes
Most commonly, “wanting” and “liking” are defined as mental processes. Hence, we
will refer to this definition as a “mental process” definition. Initially, Robinson and Berridge
(1993) used the terms “wanting” and “liking” in accordance with the common definitions of
these processes. “Liking” was simply defined as pleasure, whereas “wanting” was a strong
feeling of craving that resulted in drug administration (p. 249). “Wanting” was considered to
be the result of incentive salience attribution (i.e., associating a rewarding value to drug-
related cues; p. 261). Importantly, the authors raised the possibility that these processes were
more implicit rather than explicit (p. 267) and later stress that these implicit processes do not
necessarily need to be consciously experienced (Berridge & Robinson, 2003).
However, “wanting” was sometimes also described as drug administration itself
(Robinson & Berridge, 2000, p. 94). Whereas this might be a useful operational definition in
animal research, drug consumption in humans is dependent on more factors besides
“wanting”, such as social desirability and the availability of cognitive resources to inhibit
drug use (e.g., Wiers, Bartholow, et al., 2007). In later papers, both “wanting” and “liking”
were defined solely in terms of pre-conscious processes (e.g., Berridge, 1996; Berridge &
Robinson, 1995). Incentive salience attribution was assumed to be a pre-conscious process,
whereas the result of this process was thought to be conscious. Similarly, “liking” was not
described as a conscious process either, but it could lead to a conscious feeling of pleasure
(e.g., Berridge & Robinson, 1995, p. 72; Berridge & Robinson, 2003, p. 508). Wanting and
liking without quotation marks were suggested to refer to wanting and liking in their common
sense, whereas “wanting” and “liking” with quotation marks refer to their meaning in the IST
context (e.g., Berridge, 2010).
Subsequently, to clarify this distinction between wanting and liking in their common
sense and “wanting” and “liking” in the IST framework, Berridge and Robinson (2003, p.
508) explained that both affect and motivation consist of two subcomponents. Affect consists
of conscious pleasure (liking) and core hedonic impact (“liking”). Within motivation, they
distinguished cognitive incentives (wanting) from incentive salience (“wanting”). Cognitive
incentives refer to cognitive goals and explicit desires, whereas incentive salience refers to a
motivational process that translates affect (“liking”) into action (pursuit of the reward). It is
important to note that whereas “liking” is an affective process, “wanting” is not. Interestingly,
earlier papers thus considered “wanting” to be the result of incentive salience (e.g., Robinson
& Berridge, 1993), but this view shifted: the two were considered to be synonymous in later
papers (Berridge & Robinson, 2003).
Berridge (2004) stressed that both “wanting” and “liking” play an essential role in
reward, and that reward is not complete if only one of these two components is present.
Importantly, he states that “liking” in itself is nothing more than an affective state, a brain
reaction that is triggered by the experience of something pleasant. This means that in “liking”,
there is no object of desire, no incentive target (p. 195). “Wanting” is the process that links a
specific stimulus, experience, or action to this hedonic experience, thus transforming this
stimulus, experience, or action into an object of desire.
2.2. Effect definition: “wanting” as characteristics of drug-related CSs, and “liking” as
hedonic impact
Berridge, Robinson, and Aldridge (2009) elaborately defined “wanting” in terms of
the effects that specific drug-related CSs have on animal behavior. We therefore will refer to
this definition as an effect” definition.
They suggest that when an individual attributes
incentive salience to a CS, this CS gains three different “wanting” properties (p. 68).
First, the rewarding CS becomes a “motivational magnet” that triggers strong
approach tendencies that can become compulsive. An example of this “wanting” feature is so-
called “auto-shaping” or sign-tracking. When a CS is consistently paired with the delivery of
a reward, an individual might begin to approach or even try to consume the CS (Uslaner,
Acerbo, Jones, & Robinson, 2006).
Second, a rewarding CS automatically triggers “wanting” for the US with which it was
associated. An example is the PIT phenomenon as discussed above in which the presentation
It must be noted that, surprisingly, there is no “effect definition” of “liking”. In this context, “liking” is
regarded as a mental process. Thus, some of the relevant literature is focused mainly on “wanting” rather than
“liking”. This means that there are fewer definitions of “liking” than of “wanting”. However, the aim of this
paper is to focus on both processes, and more importantly, on dissociations between these two processes.
of a CS increases behavioral indicators of “wanting” (i.e., rats more often pressed a lever that
previously delivered a reward) even when rewards were no longer delivered. Another aspect
of this “wanting” feature is that cue-triggered US “wanting” can also spill over to other USs
(i.e., drug abusers can also exhibit compulsive approach toward other types of rewards, such
as gambling or sex).
Finally, the CS can become a conditioned reinforcer. This means that an individual
will work to gain the CS, even when a US is not presented. The CS thus becomes rewarding
in itself. An example is instrumental conditioned reinforcement, which is the process by
which a CS supports instrumental behavior. This was examined by, for instance, Parkinson,
Roberts, Everitt, and Di Ciano (2005). Their study started with a Pavlovian acquisition phase
in which rats were presented with two levers. Above each lever, a light was located. When
one light (the CS+) was illuminated, the lever below it would deliver a sucrose solution (US).
When the other light (the CS-) was illuminated, nothing happened. Subsequently, the US was
devalued. In a test phase rats were again placed in an operant chamber with two levers, with
lights located above each lever. Lever presses never resulted in sucrose delivery, but only in
the illumination of the CS+ or the CS-. Results showed that rats were inclined to approach the
lever that caused the illumination of the CS+ more than the other lever. This study thus
showed that the CS+ in itself became rewarding because rats worked to obtain the CS+ in the
absence of the US. Importantly, this effect persisted even when the US was devalued (i.e., in
the absence of “liking”).
Note that in this definition, a clear distinction is made between the different behavioral
“wanting” measures in animals. The “motivational magnet” feature refers more to approach
behavior, whereas the latter two refer to the extent to which an animal will work to obtain the
US (the cue-triggered US “wanting” feature) or the CS (the conditioned reinforcer feature).
2.3. Utility definition: “wanting” as decision utility, “liking” as experienced utility
“Wanting” and “liking” have also been defined in terms of utility. We will refer to this
definition as a “utility” definition because dissociations between “wanting” and “liking” are
considered to depend on a difference in the level of different types of utility. Berridge and
Aldridge (2008) examined how the utility of specific hedonic goals is represented in the brain
and how it can drive behavior. They distinguished four different types of utility. Predicted
utility refers to the expectation of the hedonic impact of a future reward; decision utility is the
essence of the decision at the moment it is made; experienced utility is the hedonic experience
produced by the reward after it is obtained; remembered utility is the memory of how pleasant
the reward was. Berridge and Aldridge suggested that “wanting” can be understood as
“decision utility” (e.g., p. 627), or the means to make decisions concerning which types of
rewards should be pursued, and that “liking” can be understood as “experienced utility” (see
also Berridge, 2010; Berridge & O’Doherty, 2014).
Berridge and Aldridge (2008) proposed that decisions are not optimal when there is a
mismatch between the different types of utility. “Miswanting” occurs when there is a
dissociation between an individual’s decision utility and predicted utility on the one hand, and
experienced utility on the other hand, meaning that the individual decided to pursue a reward
because they expected a large hedonic impact. However, when they obtained the reward, they
were disappointed because the reward could not live up to their high expectations. The
individual thus made an error in judgement, but they did make a rational decision because
they pursued a reward that they assumed they would like. Note that this dissociation cannot be
understood as a dissociation between “wanting” and “liking” as initially hypothesized by IST
(e.g., Robinson & Berridge, 2003) if we assume that (a) decision utility is “wanting”,
experienced utility is “liking”, and (b) predicted utility is a cognitive process regarding
explicit expectations and desires. According to IST, “wanting” (decision utility) and predicted
utility are independent: one decides to pursue drugs regardless of whether one expects
positive or negative outcomes. It is possible that decision utility and predictive utility are both
positive (i.e., they both promote drug use), but this is not necessary (e.g., Winkielman &
Berridge, 2003).
“Irrational miswanting” occurs when there is a dissociation between decision utility
(i.e., “wanting”) and predicted utility (i.e., expected “liking”). An individual does not expect
that the reward is desirable, but decides to pursue this reward anyway. The outcome is wanted
even though the individual correctly judges that the outcome will not be pleasurable (i.e.,
there is no experienced “liking”). The individual thus made a correct estimation of the
likeability of the outcome (i.e., they know they will not like it), but the choice was irrational
because he or she chose an unliked outcome. Again, if we use the definitions discussed above,
this does not imply that “irrational miswanting” can be understood as a dissociation between
“wanting” and “liking” in the way IST would predict. It does, however, reflect a dissociation
between “wanting” as an unconscious desire (decision utility) and wanting as a conscious
declarative process that involves explicit goals and expectations (predicted utility; e.g.,
Berridge & O’Doherty, 2014). Finally, it is interesting to note that the emphasis on the
unconscious nature of “wanting” and “liking” disappeared in this literature, and it is suggested
that “wanting” may be only a facet of decision utility (Berridge & Aldridge, 2008).
2.4. Neurological definition: “wanting” as activity in the mesocorticolimbic dopamine
system and “liking” as activity in subcortical hedonic hotspots
A final definition regards “wanting” and “liking” as neurological processes. We will
therefore refer to this definition as the neurological definition. According to this view,
“wanting” is (the result) of activity in the mesolimbic dopamine system, which involves the
midbrain, the NAcc, parts of the striatum, the amygdala, and the prefrontal neocortex. The
most prominent neurotransmitter within this system is dopamine, but opioids, glutamate, and
GABA also play a role (Berridge & Robinson, 2003; Berridge, 2009).
“Liking”, in contrast, takes place in a more limited system consisting of small parts
within the NAcc and the ventral pallidum, so-called “hedonic hotspots”. Whereas “wanting”
is triggered by activation in a single “wanting” hotspot, “liking” is triggered only when the
multiple hedonic hotspots work in synchrony (Berridge & Robinson, 2003; Berridge &
Kringelbach, 2008). Importantly, in contrast to the “wanting” system, the “liking” system
does not involve dopamine but only opioid stimulation. Thus, boosting dopamine
transmission in these networks will cause increases in “wanting”, but not in “liking”.
2.5. Summary
In short, there are four types of definitions of “wanting” and “liking”. It is possible to
view each type of definition as only a sub-component of these complex concepts: First, the
mental process definition in which “wanting” was a combination of wanting in its common
sense and the result of the preconscious process of incentive salience attribution. It was
referred to as a need or desire, but also as drug use itself. “Liking” was referred to as an
experience of pleasure. A later, but closely related view is that “wanting” refers solely to
(preconscious) incentive salience attribution and “liking” refers only to (preconscious)
hedonic impact. This view also holds that “wanting” and wanting are the two subcomponents
of motivation, whereas “liking” and liking together form affect. Second, the effect definition
suggests that “wanting” can be seen as an effect of a set of characteristics of the drug-related
CS: the “motivational magnet” feature, the cue-triggered US “wanting” feature, and the
conditioned reinforcer feature. Third, “wanting” and “liking” have been defined as different
types of utilities: decision utility (i.e., the essence of the decision at the moment it is made)
and experienced utility (the hedonic experience after obtaining an outcome), respectively.
Fourth, “wanting” and “liking” have been defined in terms of neural activity in subcortical
brain systems.
Thus, whereas the definition of “liking” is quite consistent, the definition of “wanting”
varies across and even within papers (e.g., the description of “wanting” as both a
preconscious process that motivates drug consumption and as drug consumption in itself;
Robinson & Berridge, 2000, p. 94). We want to emphasize that regarding complex concepts
from different perspectives is commendable because it enriches our understanding of these
concepts. However, the occasional inconsistencies within the different definitions can hinder
our understanding of the theory and our ability to articulate hypotheses regarding “wanting”
and “liking” in addiction.
3. Dissociating “wanting” and “liking” in humans
Even though IST was based mainly on research with non-human animals, many
researchers have tried to apply IST to humans. Some efforts have been made to examine
“wanting” and “liking” in light of the effect definition and the neurological definition. For
instance, several studies have examined PIT in humans (cf. the automatic cue-triggered US
wanting feature) in combination with fMRI (e.g., Talmi, Seymour, Dayan, & Dolan, 2008),
and even though neurological studies are still scarce, there is increasing evidence that the
mesolimbic dopamine system plays an important role in drug addiction and relapse in humans
(e.g., Leyton, 2007), as IST suggests (e.g., Robinson & Berridge, 2003).
However, the current focus is on research on the mental process definition of
“wanting” and “liking”. This research has become increasingly popular over the last few
decades, even though it faces an important challenge: According to IST (e.g., Berridge &
Robinson, 1995), “wanting” and “liking” are not directly subjectively experienced. As a
result, one cannot simply resort to direct measurement procedures in which participants self-
assess their level of “wanting” and “liking”. For this reason, researchers have turned to the
use of so-called implicit measurement procedures. Before we discuss different implicit
measures of “wanting” and “liking”, we first discuss the rationale behind the use of implicit
measures. Next, we review the use of such measures within the context of IST.
Implicit measures are assumed to have certain advantages over traditional explicit
measures. Most importantly, it is commonly assumed that compared to explicit measures,
implicit measures are (a) less susceptible to extraneous factors such as social desirability
concerns, deception, and other conscious control strategies and (b) might capture processes
that are not introspectively accessible or cannot be easily controlled. There is less agreement
about what exactly distinguishes implicit measures from explicit measures.
De Houwer et al. (2009) suggested that an implicit measure can be defined as the
outcome of a measurement procedure that is caused by the to-be-measured-attribute
automatically. The process that underlies a measure is automatic when it meets one or more of
the following criteria: it is uncontrolled, unintentional, independent of goals, stimulus driven,
unconscious, efficient, and fast (e.g., Moors & De Houwer, 2006). Most implicit measures are
automatic according to some of these criteria, but not all. Some researchers have focussed
only on the question whether implicit measures can capture processes or attitudes of which
the participant is unaware. However, there is very little evidence that implicit measures are
automatic in this sense (e.g., Gawronski, Hofmann, & Wilbut, 2006). Nevertheless, there is
reason to argue that most implicit measures are automatic in the sense that they are based on
processes that are less intentional and less easy to control than the processes that underlie
most traditional, explicit measures. On the basis of this knowledge, researchers have
suggested that implicit measures are possibly able to capture automatic processes such as
“wanting” and “liking” (e.g., Wiers, Van Woerden, Smulders, & De Jong, 2002).
The Implicit Association Test (IAT; Greenwald, McGhee, & Schwarz, 1998) is
beyond any doubt the most well-known implicit measure. The IAT was designed to capture
associations between target and attribute concepts. In a first block, stimuli are classified as
belonging to one of two target concepts (e.g., as “flower” or “insect”) by pressing one of two
keys. In a second block, different stimuli are classified as referring to one of two attribute
concepts (e.g., as “positive” or “negative”) by using the same keys as in the first task. The
critical blocks are those in which both target and attribute stimuli are presented. In some
blocks, target and attribute categories are assigned to responses in an association-compatible
manner (e.g., press a first key for “flowers” and “positive” and a second key for “insects” and
“negative”) whereas in other blocks, the category-response assignments are association-
incompatible (e.g., press a first key for “insects” and “positive” and a second key for
“flowers” and “negative”). In the association-compatible blocks, response latencies are
usually shorter than in association-incompatible blocks. The difference in response latencies
between both types of blocks is thought to reflect the relative strength of the associations. The
IAT is assumed to be an implicit measure in the sense that it is difficult to control (e.g.,
Greenwald et al., 1998). By changing the labels and stimuli, the IAT can measure a wide
variety of implicit attitudes.
In research on “wanting” and “liking” in addiction
, researchers have explored not
only the potential of the IAT but also other implicit measures, some of which are variants of
the IAT. In the remainder of this section, we provide a brief overview of the various measures
that have been used in this research (see Gawronski & De Houwer, 2014; and Nosek,
Hawkins, & Frazier, 2011, for more complete reviews of implicit measures in general). A
summary can be found in Table 2. The personalized IAT is a variant of the IAT that was
developed to remove the influence of extra personal associations (e.g., societal stereotypes;
Olson & Fazio, 2004). In this task, labels refer to the participants themselves (e.g., “I like”
Even though researchers have performed many studies examining the role of “wanting” and “liking” in the
context of over-eating as well, we will not discuss this literature. The main reason for this is that there is no
theoretical basis for the assumption that incentive sensitization plays a role in eating behavior (e.g., Havermans,
instead of “positive” and “I dislike” instead of “negative”), to avoid confusion between
societal norms and participants’ own attitudes. Other IATs have been developed that reduce
the relativity of the measure. Karpinski and Steineman (2006) introduced the Single Category
IAT, a task that aims to measure attitudes towards a single target concept (e.g., flowers)
without the need for a complementary category (e.g., insects). Another version of the IAT is
the unipolar IAT (e.g., Jajodia & Earleywine, 2003), in which only one attribute category is
needed (e.g., either “positive” or “negative” but not both). This allows researchers to take into
account that participants might have an ambiguous attitude (i.e., both positive and negative)
toward specific targets.
Importantly, researchers have assumed that by varying the labels of the IAT, different
versions of the IAT can be used to capture different concepts. In line with this idea, “wanting”
versions of the IAT have been created that are assumed to reflect “wanting”, whereas effects
on “liking” versions of the IAT are assumed to reflect “liking”. First, Wiers et al. (2002)
developed both a valence IAT, in which the target categories are “alcohol” and “soda”, and
the attribute categories are “positive” and “negative”, and an arousal IAT, with the same
target categories, but with “passive” and “active” as the attribute categories. Because
“wanting” is assumed to cause drug cues to become more arousing, the arousal IAT was
assumed to be a valid measure of “wanting”. Second, Tibboel et al. (2011) designed a liking
IAT, in which the attribute labels were “I like” and “I do not like” and a wanting IAT, in
which the attribute labels were “I want” and “I do not want”. They reasoned that these labels
better reflected the processes proposed by Robinson and Berridge. Finally, in the approach-
avoid IAT, the labels “approach” and “avoid” are used. It was designed as a measure of
“wanting”, as it is assumed it can capture automatic approach tendencies (e.g., Palfai &
Ostafin, 2003).
Besides these variations of the IAT, several other implicit measurement tasks have
been developed to assess the motivational component of “wanting”. In these tasks,
participants are required to make approach responses to one category of items, and avoidance
responses to another category of items. In another block, the opposite responses must be given
to the same categories of stimuli. These tasks are based on the assumption that it is easier to
make an approach-movement towards stimuli that are rewarding than to avoid them.
In the (Affective) Simon Task (AST) and the Approach-Avoidance Task (AAT),
stimuli need to be judged on a stimulus dimension that differs from the to-be-measured
dimension (e.g., whether a picture has a portrait or landscape format; e.g., Field, Caren,
Fernie, & De Houwer, 2011; Wiers, Rinck, Dictus, & Van den Wildenberg, 2009). In
contrast, in the “relevant Stimulus Response Compatibility Task” (rSRC; Mogg, Bradley,
Field, & De Houwer, 2003), participants approach or avoid stimuli on the basis of the
stimulus dimension that is being assessed (e.g., whether a picture depicts an alcoholic drink or
a soft drink). Importantly, this measure can still be considered implicit or indirect in the sense
that participants are not asked to explicitly judge the attitude objects. Instead, their approach-
or avoidance tendencies are inferred on the basis of the relative speed with which they make
approach- or avoidance movements toward the objects. Responses can be given either by
moving a manikin towards or away from the stimulus (in the rSRC, the AST) or by moving a
joystick towards or away from the stimulus (in the AAT).
5. Are implicit measures of “wanting” and “liking” valid?
Our assessment of the validity of implicit measures of “wanting” and “liking” is
guided by the perspective on validity that was put forward by Borsboom et al. (2004). They
suggested that a test is valid if a) the to-be-measured attribute exists and if b) there is a causal
relationship between variations in the attribute and variations in the measurement outcomes.
Although questions can be raised about whether “wanting” and “liking” do exist as
psychological attributes, it is difficult to arrive at definite answers to this type of ontological
issues. Hence, we sidestep the ontological debate and focus on the causality aspect of validity.
We first investigate whether, on an a priori theoretical basis, there are reasons to assume that
“wanting” and “liking” as defined in the literature can causally influence performance on
implicit measurement tasks that are assumed to capture “wanting” and “liking”. Next, we
review whether there is empirical support for the hypotheses that “wanting” and “liking”
actually determine implicit measures that are assumed to capture these attributes. In doing so,
we distinguish between three types of empirical studies. First, there are quasi-experimental
studies, in which implicit measures of “wanting” and “liking” are examined in groups that are
assumed to differ on these constructs (e.g., comparing heavy and light drinkers). Second,
there is correlational research, in which correlations between scores on implicit “wanting”
and/or “liking” measures and other indices of “wanting” and “liking” are examined. Third,
there are experimental studies, in which researchers manipulate “wanting” and/or “liking” and
examine how this affects outcomes of implicit “wanting” and “liking” measurement
procedures. Borsboom et al. (2004) argue that experimental studies are preferred, because
both measurement and validity are causal concepts. Thus, manipulations of “wanting” and
“liking” should causally affect “wanting” and “liking” respectively.
5.1. Are there a priori theoretical reasons to assume that implicit “wanting” and “liking”
measures are valid?
Ideally, valid measures of “wanting” and “liking” should be able to capture all facets
of “wanting” and “liking” described in the mental process definition. Thus, a “wanting” test
should assess an individual’s desire or need to use drugs, instrumental drug-seeking, drug-
taking behavior, and the (preconscious) process of incentive salience attribution. A valid
measure of “liking”, on the other hand, should capture the (preconscious) experience of
Before we discuss the validity of each type of implicit measure separately, one first
argument that makes us question the validity concerns the assumption that they can capture
preconscious processes. According to several IST papers that refer to “wanting” and “liking”
as psychological processes, “wanting” and “liking” are assumed to be preconscious (e.g.,
Berridge, 1996). Bargh (1994) notes that an unconscious attitude can be “unconscious” in
three different ways. First, an individual may not be aware of the source of the attitude (i.e.,
the experiences that gave rise to the attitude), they may not be aware of the content (i.e.,
whether it is positive or negative), or they may not be aware of the impact of the attitude on
behavior. Importantly, there is no evidence that implicit measures capture unconscious or
preconscious processes (Gawronski, Hofmann, & Wilbur, 2006; Hahn & Gawronski, 2014).
For instance, Hahn, Judd, Hirsch and Blair (2014) recently showed that participants were very
good at predicting their own IAT scores, suggesting that the IAT does not measure
unconscious processes. Interestingly, even though prediction accuracy was high, there was
still low correspondence between the IAT and explicit attitude measures. This implies that
implicit measures do tap into different, more automatic processes than explicit measures.
Thus, whereas it is doubtful that implicit measures can provide us with a valid
estimation of “wanting” and “liking”, implicit measures do seem to be more promising in
capturing automatic processes than explicit measures. Furthermore, it is important to note,
most researchers who examine IST using implicit measures interpret “wanting” and “liking”
not as preconscious processes but as implicit or automatic processes (e.g., Wiers et al., 2002),
a definition that has been used by Berridge and Robinson (2003; page 267). In this case,
implicit measures might be considered as valid and valuable ways to examine “wanting” and
“liking”, because, as we mentioned above, one property of automaticity is indeed that it is
unconscious (e.g., Moors & De Houwer, 2006), but a process can still be considered as
automatic to some extent even when it does not (fully) meet this criterion.
Another consideration is that the different definitions of “liking” all stress that it
concerns a temporary experience of pleasure. According to the mental process definition, it is
a hedonic response that is experienced at the moment of drug consumption and that might not
involve a specific “object of desire” (e.g., Berridge, 2004). It seems unlikely that IATs can
access such responses. Because IATs are designed to assess memory representations, they are
more likely to capture memory-based pre-consumption appetitive responses. Unlike the
“liking” responses that are triggered during consumption, pre-consumption responses are
assumed to reflect learned cue-affect associations in memory and might be more akin to
5.1.1. Wanting, Liking, and Approach-Avoidance IATs
Even though the definition of “liking” seems more straightforward than the definition
of “wanting”, it is not clear whether the valence IAT (with labels “positive” and “negative”)
of Wiers et al. (2002) is a good “liking” test. First, the attribute labels “positive” and
“negative” are not synonymous for “liking” and “disliking”. Even though individuals usually
experience “liking” for positive things, the definition of “liking” is limited to merely the
hedonic experience of pleasure, whereas the label “positive” is a much broader concept.
Second, stimulus items such as “sociable”, “tedious” (Wiers et al., 2002), “sincere”, and
“vindictive” (De Houwer et al., 2004) which are typically used in the valence IAT are
generically positive and negative words that are not even mentioned in any of the definitions
by Robinson and Berridge (e.g., 1993, 2000, 2003). Because the nature of the (attribute) items
can determine the conceptualization of the (attribute) labels (i.e., “positive” and “negative”;
see Govan & Williams, 2004), it is unlikely that the attribute labels in the valence IAT are
conceptualized by participants in line with the conceptualization of “liking” in IST. Similar
arguments can be made to question the validity of the liking IAT of Tibboel et al. (2011).
Even though they used target labels that were more closely related to IST (i.e., “I like” and “I
do not like”), they used generic positive and negative words (e.g., “holiday”, “pain”) as well,
making it quite plausible that participants recoded these labels to mean nothing more that
“positive” and “negative” instead of “I like” and “I do not like”. In sum, it can be argued that
none of the available valence or liking IAT effects reflect the hedonic impact as put forward
in IST.
Wiers and colleagues (2002) assumed that their arousal IAT measures “wanting”,
which they interpreted as “sensitized arousal”. One problem with this interpretation is that
even though the attribution of incentive salience can coincide with feelings of arousal,
research has shown that arousal and incentive salience cannot be equated (e.g., Berridge,
Venier, & Robinson, 1989). Moreover, IST does not assign a major role for arousal to the
development and maintenance of addiction (e.g., Robinson & Berridge, 1993). As such, an
important theoretical problem with the arousal IAT is that it lacks content validity. Hence, at
best it is causally determined only by a non-essential and trivial facet of “wanting”.
One can also doubt whether the “wanting” IAT used by Tibboel and colleagues (2011)
actually captures “wanting”. Because they used the labels “I want” and “I do not want”, it
could be that they captured the “layman’s” definition of wanting. However, it is less clear
whether this IAT can measure “wanting” in the IST sense. First, there is no reason to assume
that participants coded these labels to mean something along the lines of the IST definitions
mentioned above. Second, the design of this IAT faces the same problem as the design of
Wiers et al. (2002) in the sense that the stimuli were quite general positive and negative
words. Hence, there is no reason to assume that these words captured the theoretical meaning
of “wanting”. Because all “I want” items were positive and all “I do not want” items were
negative, it is even possible that participants recoded both the “wanting” and the “liking”
labels in this study as merely meaning “positive” and “negative”.
The question whether the approach-avoid IAT can capture all facets of “wanting” can
also not be answered in an unequivocally positive manner. The target labels in the task of
Palfai and Ostafin (2003) were “alcohol” and “electricity” whereas the attribute labels were
“approach” and “avoid”. Even if it is safe to assume that an IAT with these labels indeed
captures “alcohol-behavioral disposition associations”, as the authors suggest (p. 1152), we
cannot conclude that this version of the IAT is causally influenced by “wanting” in the sense
of (the result of) the preconscious process of incentive salience attribution, a need or desire to
use drugs, or drug use itself.
5.1.2. Measures of automatic motivational tendencies
The AAT is thought to be a measure of appetitive approach tendencies, which is
assumed to be a property of “wanting”. The AAT seems to assess the extent to which the
presentation of a drug-related stimulus automatically triggers approach behavior. Eder and
Rothermund (2008), however, argued that the activation of approach and avoid responses is
driven merely by the valence of the presented stimuli. More specifically, they suggest that
positive stimuli automatically activate positive responses, including the approach responses in
an AAT. Thus, AAT effects could be driven by the mere valence of stimuli instead of their
motivational salience. Furthermore, Krieglmeyer, De Houwer, and Deutsch (2013) suggested
that such AAT-like effects occur mainly when participants have the conscious goal to
evaluate stimuli. This implies that it would not be possible to interpret such effects as
reflections of “wanting”, as this process should not be affected by conscious goals (e.g.,
Berridge, 1996). In addition to these objections, it is also unclear how AAT effects relate to
the other facets of “wanting”. There is little reason to believe that they capture drug use,
desire, or incentive salience attribution. The same holds for the modified versions of the
Simon task and the rSRC task (Mogg et al., 2003).
5.2. Is there empirical evidence for the validity of “wanting” and “liking” measures?
5.2.1. Quasi-experimental research. Wanting, Liking, and Approach-Avoid IATs
In a seminal study on implicit measures of “wanting” and “liking”, Wiers et al. (2002)
introduced the valence and the arousal alcohol IAT in a sample of heavy drinkers and a
control group of light drinkers. Interestingly, their results suggested that both heavy and light
drinkers had negative alcohol associations. This was assumed to imply that neither of the
groups experienced “liking” for alcohol.
However, the heavy drinkers, but not the light
drinkers, did reveal an arousal IAT effect that was indicative of associations between alcohol
and arousal. This led the researchers to conclude that heavy but not light drinkers experienced
“wanting” for alcohol. In line with these results, De Houwer, Crombez, Koster, and De Beul
(2004) found that, just like heavy social drinkers, alcohol dependent patients showed negative
alcohol associations and positive alcohol arousal associations. However, Dickson, Gately and
Field (2013) used two unipolar IATs: one to measure positive alcohol associations, and one to
measure negative alcohol associations. Their data showed that alcohol dependent patients’
positive alcohol associations did not differ from controls, but their negative alcohol
associations were less strong than those of controls. This suggests that valence might indeed
play a role in addiction, although it is nuanced (i.e., a lack of negative associations instead of
strong positive associations).
Other studies however, yielded contrasting results. First, Wiers, Houben, and De
Kraker (2007) showed that cocaine users have stronger associations between cocaine and
sedation than between cocaine and arousal. Furthermore, valence IATs showed that they had
It is important to note that one cannot interpret positive and negative IAT scores to reflect positive and negative
attitudes respectively. Blanton and Jaccard (2006), for instance, suggest that the zero-point of an IAT is
meaningless and should be interpreted with caution. We want to emphasize that relative measures such as the
implicit tasks that we describe in this manuscript should be interpreted only in a relative way (e.g., comparing
scores between groups; examining pre- and post-measurements; or in correlations)..
stronger positive than negative associations with cocaine. Second, several other studies using
standard, personalized, and unipolar versions of the valence IAT did yield higher liking scores
for substance-related stimuli in substance (ab)users compared to controls (e.g., De Houwer,
Custers, & De Clercq, 2006; De Houwer & De Bruycker, 2007; Jajodia & Earlywine, 2003;
McCarthy & Thompsen, 2006; Robinson, Meier, Zetocha, & McCaul, 2005), even though IST
would predict negative “liking”. Third, several experiments by Swanson and colleagues
(2001) revealed no differences in implicit “liking” between smokers and non-smokers.
Fourth, Tibboel and colleagues (2011) used wanting and liking IATs in a group of smokers
and a control group of non-smokers. Results showed that smokers had more positive scores
on both IATs compared to non-smokers, suggesting that smokers both “like” and “want”
smoking more than non-smokers. In contrast, IST would predict larger differences between
the two groups on their wanting scores compared to their liking scores.
Finally, Tibboel et al. (2015) recently aimed to assess “wanting” and “liking” in a
group of light social drinkers, a group of heavy social drinkers, and a group of alcohol-
dependent patients. On the basis of IST, they expected that alcohol-dependent patients would
experience more “wanting” and less “liking” compared to the light drinkers. They suggested
that heavy social drinkers would have scores that would fall in between those of the light
drinkers and the patients. To test this hypothesis they used two personalized single target
IATs with labels “I want” and “I do not want” and labels “I like” and “I do not like”.
Surprisingly, results showed that there were no differences in “wanting” and “liking” scores.
Furthermore, there were no differences between light drinkers and alcohol-dependent patients.
Surprisingly, both groups had significantly more negative scores on both IATs compared to
heavy social drinkers. These data suggest that IATs in fact do not tap into “wanting” and
“liking”. Instead, the IAT “liking” measure that Tibboel et al. used might reflect personal
alcohol-associations that are affected by personal experiences. For light drinkers, a negative
“liking” score might reflect a lack of positive experiences or a disinterest towards alcohol,
whereas for alcohol-dependent patients, the same score might reflect their lifelong struggle
with the negative consequences of their addiction. In contrast, heavy social drinkers might
have had negative experiences (e.g., hangovers) which might not yet outweigh the positive
experiences. Furthermore, Tibboel et al. suggested that their “wanting” measure might not
reflect incentive sensitization but instead might reflect more declarative goals. In this sense,
alcohol is not “wanted” by alcohol-dependent patients, because it stands in the way of their
goal to be healthy and productive.
Quasi-experimental research with the approach-avoid IAT is quite rare. De Houwer et
al. (2006) found that both smokers and non-smokers had negative valence IAT scores but that
smokers had positive approach associations with smoking, whereas this score was negative
for non-smokers. This is in line with the findings of Wiers et al. (2002) and, more
importantly, it is in line with the predicted dissociation between “wanting” and “liking” in the
To summarize, quasi-experimental studies do not yield consistent evidence for the
hypothesis that substance (ab)users have higher scores on arousal, wanting, or approach-avoid
IATs and equal (or lower) scores on valence IATs compared to non-users or less frequent
users. Measures of automatic motivational tendencies.
Research with the rSRC task showed that smokers have a stronger approach bias
toward smoking-related pictures compared to non-smokers (Mogg et al., 2003). Similarly,
using an rSRC task, Field, Kiernan, Eastwood, and Child (2008) showed that heavy social
drinkers had an approach bias toward alcohol whereas light drinkers did not. Field, Caren, et
al. (2011) asked social drinkers to perform both an rSRC task and an alcohol Simon task.
Their rSRC task revealed that heavy drinkers had a stronger approach bias toward alcohol,
whereas the Simon task did not. Thus, it seems that the rSRC is better able to capture
motivational processes in addiction than modified versions of the Simon Task. However,
some studies yielded contrasting results. Spruyt and colleagues (2012) used an rSRC task to
examine alcohol-approach tendencies in alcohol-dependent patients. In contrast to their
hypothesis, they found that patients had an avoidance bias, whereas a control group of non-
problematic social drinkers had an approach bias toward alcohol. Another discrepant result
was found by Barkby, Dickson, Roper, and Field, (2012), who showed no difference between
alcohol-dependent patients and controls on their rSRC task.
In several quasi-experimental studies the AAT has been used to examine group
differences in automatic approach tendencies. Wiers et al. (2009) found that heavy drinkers
with at least one g-allele in a mu-opioid receptor gene (i.e., a gene that is associated with
rewarding effects of alcohol) showed a stronger approach bias towards alcohol and other
appetitive stimuli than heavy drinkers who did not carry this allele, suggesting that AAT
scores are higher in at least a subgroup of heavy drinkers. Similarly, Cousijn, Goudriaan, and
Wiers (2011) used a cannabis-AAT to show that heavy cannabis users had a stronger
approach bias towards cannabis than controls.
Some results are more difficult to interpret: Sharbanee, Stritzke, Wiers, et al. (2013)
did an AAT study with heavy and light drinkers. They found that heavy drinkers with low
working memory capacity had more difficulties to avoid alcohol than light drinkers. However,
heavy drinkers with high working memory capacity were quicker to avoid alcohol than light
drinkers. It is not clear whether we can assume that these AAT findings support IST, because
IST does not make specific predictions regarding the influence of one’s working memory
capacity on “wanting”.
Other findings are clearly not in line with IST’s predictions. First, Cousijn, Snoek, and
Wiers (2013) asked two groups of heavy cannabis users to perform a similar AAT. One group
was tested while they still had the intention to smoke but before they actually smoked
cannabis; the other group was tested after they had smoked cannabis. Even though cannabis
users who had the intention to smoke had higher levels of subjective craving and lower levels
of satiation than the group who had already smoked cannabis, the pattern of AAT scores was
surprising: the “intention” group showed no implicit approach bias, whereas the group who
had just smoked cannabis did have positive AAT scores. Clearly, this goes against the idea
that satiation reduces “wanting” and should thus reduce approach scores on the AAT. Second,
Wiers and colleagues (2013) asked a group of smokers and a group of ex-smokers to perform
a smoking AAT. They found that only current smokers and not ex-smokers had positive
approach scores. Whereas IST would predict positive AAT scores in current smokers, it
would predict also that ex-smokers would have positive AAT scores, as incentive
sensitization has very long lasting effects (e.g., Robinson & Berridge, 2003). In sum, quasi-
experimental research is either scarce (for the modified Simon task) or produced mixed
results (for the rSRC task and the AAT).
5.2.2. Correlational research.
Because IST does not specify how incentive processes in humans manifest themselves
in behavior (e.g., Robinson & Berridge, 2000), it is difficult to decide what kind of variables
different “wanting” and “liking” test scores should correlate with in order to be valid.
Nevertheless, as mentioned above, several criteria that seem in line with IST have been
suggested and used by several researchers. For instance, it has been argued that “wanting”
should correlate with drug use and questionnaires regarding drug craving or drug problems
and that measures of “liking” should correlate with explicit liking ratings and questionnaires
as well (e.g., Wiers et al., 2002).
In this section, we review studies in which implicit
measures of “wanting” and “liking” were related to this type of criterion variables.
Besides the fact that it is difficult to judge what kind of variables “wanting” and
“liking” test scores should correlate with, it is also difficult to decide what kind of variables
they should not correlate with. At first sight, IST clearly predicts that “wanting” and “liking”
are dissociated, and should thus not correlate with each other. This implies that we would
expect weak or non-existing correlations between “wanting” and “liking” test scores, between
“liking” test scores and “wanting” indices, and between “wanting” test scores and “liking”
indices. However, it is important to note that, as we mentioned above, “wanting” and “liking”
are commonly closely related (e.g., Winkielman & Berridge, 2003) and only become
dissociated after casual drug use has developed into addiction. The samples in most of the
studies we report here consist of casual drug-users rather than drug abusers, which means that
hypotheses concerning correlations are not clear-cut. In casual drug-users “wanting” and
“liking” test scores and other behavioral manifestations of “wanting” and “liking” can all be
expected to be intercorrelated to some extent, because “wanting” and “liking” are not (yet)
dissociated. This sheds doubt on the assumption that correlational designs with casual drug-
users can be used to draw any conclusions about the respective validity of different “wanting”
and “liking” measures and suggests that results should be interpreted with caution. Instead, as
suggested by an anonymous reviewer, researchers could focus on quasi-experimental designs
that compare the predictive validity of wanting and liking measures in addicts and non-
It is important to note, however, that IST assumes that “wanting” and “liking” are independent of their
conscious counterparts (e.g., Berridge, 2010), implying that calculating correlations between implicit and explicit
measures are not meaningful in this context. However, we will discuss these correlations because it is common
practice to calculate and report such correlations and to draw conclusions on the basis of these correlations (e.g.,
Houben & Wiers, 2007a; 2007b; 2008).
addicts. For instance, wanting measures might have unique value in predicting drug
consumption in addicts but not in non-addicts. Wanting, Liking, and Approach-Avoid IATs.
Wiers et al. (2002) found a positive correlation between scores on their arousal IAT
and alcohol (ab)use (i.e., participants with stronger alcohol-arousal associations used more
alcohol), a finding that was replicated with a unipolar IAT (with labels “active” and “neutral”
instead of “active” and “passive”) as well (e.g., Houben & Wiers, 2006). However, no
correlations were found between arousal IATs and explicit arousal ratings. To address the
issue of predictive validity, Wiers and colleagues (2002) also measured prospective alcohol
use, by asking participants to record their alcohol consumption during one month after
participating in the experiment. Results showed that the combined IATs could explain some
of the variance in alcohol use. Research with another version of the arousal IAT showed that
positive alcohol-arousal associations could predict alcohol use in heavy drinkers (Houben,
Rothermund, & Wiers, 2009). In contrast, Van den Wildenberg et al. (2006) found a positive
correlation between arousal IAT scores and a decrease in alcohol-induced heart rate (i.e.,
higher arousal IAT scores were associated with more relaxation instead of arousal after an
alcohol induction).
Using the valence IAT, Wiers and colleagues (2002) found a negative correlation with
alcohol use (i.e., participants with more positive implicit attitudes towards alcohol consumed
less alcohol). Similar results have been found, for instance, by Houben and Wiers (2007a) for
both the traditional valence IAT and a personalized valence IAT. However, they found no
correlations between implicit and explicit valence measures, and no such correlations were
found in other studies using similar valence IATs (e.g., Houben & Wiers, 2006). To
complicate things even further, other research has shown that valence IATs were positively
correlated with prospective alcohol consumption (e.g., Houben & Wiers, 2007a,, 2008). For
the wanting IAT, Tibboel et al. (2011) calculated correlations with explicit measures for
deprived and sated smokers. When smokers were deprived, they found positive correlations
between the wanting IAT scores and ratings of arousal, valence, and wanting, general
smoking urges, and the urge to smoke a cigarette after the session. When smokers were sated,
they found correlations with valence ratings, smoking urges, craving, and the urge to smoke
after the session. Similar correlations were found for the liking IAT. As the smokers in this
study were relatively light smokers, one could argue that for them, “wanting” and “liking”
were not yet dissociated, explaining the similarities between the correlations. However, partial
correlations, controlling for liking IAT scores, revealed a significant correlation between the
wanting IAT and arousal ratings in the deprivation session, and with the urge to smoke a
cigarette at the end of the session in the satiation session.
A number of correlational studies have employed the approach/avoid IAT. Lindgren et
al. (2015) reported positive correlations between scores several scores reflecting alcohol
problems and habitualness of drinking. Palfai and Ostafin (2003) also found positive
correlations between scores on the approach/avoid IAT and binge drinking, difficulty to
control alcohol consumption, and general sensitivity to positive incentives. Furthermore,
participants were asked to rate their urge to drink, affect, and positive expectancies twice:
Once during a baseline phase, and once during a cue exposure test. During this test,
participants were presented with a glass of beer which they were told they would drink after
they filled out the ratings a second time. The IAT score predicted the urge to drink, subjective
arousal, and pleasantness ratings during cue exposure. However, after controlling for baseline
responding (i.e., responding when they were not yet exposed to the alcohol-cue), the
correlation with pleasantness no longer reached significance. The approach IAT score did not
correlate with several temptation-related variables, nor with behavioral inhibition. Ostafin,
Marlatt, and Greenwald (2008) observed no correlation between the approach IAT and
explicit motivation to drink alcohol, but they did find a positive correlation with alcohol
consumption on a taste test. Ostafin, Palfai, and Wechsler (2003) tried to disentangle
approach and avoidance by calculating separate approach and avoidance scores. They used an
approach-avoid IAT in a sample of social drinkers, who varied in their alcohol consumption.
Their results showed that whereas indices of approach associations did not reliably predict
binge drinking and alcohol problems, indices of avoidance associations did: Lower avoidance
scores were associated with more binge drinking and alcohol problems. Thus, it seems that
the avoidance but not the approach component of the approach-avoid IAT is a more important
predictor of alcohol-related behavior. This observation suggests, however, that the approach-
avoid IAT might not be the best measure to assess “wanting” because IST suggests that
“wanting” is driven by approach-related motivational processes rather than by (a lack of)
Cohn et al. (2012) performed regression analyses to examine how much variance on
the approach/avoid IAT was predicted by temptation to drink alcohol and by alcohol
restriction. They found only a significant effect of restriction, but not temptation. The IAT
was not correlated with the urge to drink alcohol. Finally, other papers in which the use of the
approach/avoid IAT was discussed did not find any correlations with “wanting” measures
(e.g., Lindgren et al., 2013; Van den Wildenberg et al., 2006).
To summarize, the arousal IAT has been found to correlate with substance use or
substance-related problems (e.g., Wiers et al., 2002), but also with indices of sedation (Van
den Wildenberg et al., 2006). For the valence IAT, both positive, negative and no correlations
with relevant variables were found (e.g., Houben & Wiers, 2006, 2007a; Wiers et al., 2002).
Furthermore, research by Tibboel et al. (2011) provides no clear support for the conclusion
that wanting and liking IATs specifically capture “wanting” or “liking”, respectively. Finally,
approach-avoid IAT scores tend to correlate with some addiction-related variables, but not
with others. Measures of automatic motivational tendencies
Several studies have reported positive correlations between alcohol rSRC effects and
alcohol consumption (Christiansen, Cole, Goudie, & Field, 2011; Field, Kiernan, Eastwood,
Child et al, 2008). Furthermore, Mogg et al. (2003), also found correlations between their
smoking rSRC task and the time participants looked at smoking pictures (as measured with an
eye tracker), attentional bias, and pleasantness ratings, but did not find correlations with other
criteria (e.g., the number of cigarettes participants smoked on an average day, time since their
last cigarette, how long they smoked, etc.).
Clinical studies examining approach bias in actual alcohol-dependent patients revealed
contrasting results. Whereas Barkby et al. (2012) report a positive correlation between rSRC
scores and alcohol consumption prior to treatment (i.e., supporting IST principles), Spruyt et
al. (2012), in contrast, found that alcohol-dependent patients with a stronger alcohol
avoidance bias were more likely to relapse three months after they were discharged from the
hospital. Thus, rSRC scores can correlate positively, negatively, or not at all, with drug-
related behaviors. Research with the modified Simon Task is scarce, but so far no significant
correlations with drug-related behavior have been found (Field et al., 2011; Van Hemel-Ruiter
et al., 2011).
Finally, Wiers et al. (2009) did not find any significant correlations between the AAT
and subjective craving (other correlations are not reported). In another study by Wiers and
colleagues (2010), regression analyses showed that whether or not participants underwent an
approach or avoidance training could explain a significant part of the variance in alcohol
consumption during a taste test: participants who had learnt to avoid alcohol drank less than
participants who were trained to approach alcohol. However, this was the case only for those
participants for whom approach (or avoidance) training resulted in faster alcohol-approach (or
avoid) reaction times. Van Hemel-Ruiter et al. (2011) reported a negative correlation between
the AAT score for alcohol-related pictures and alcohol use, but a positive correlation with
subjective valence of alcohol (i.e., the stronger participants’ approach bias, the less alcohol
they used, and the more positively they judged alcohol). Furthermore, in their study
concerning cannabis use, Cousijn and colleagues (2011) report that AAT scores correlated
negatively with subjective craving (i.e., a stronger approach bias was associated with weaker
subjective craving) but they did not find significant correlations with other criteria (e.g.,
baseline cannabis use, cannabis-related problems, smoking). However, their study did
demonstrate predictive validity in the sense that they found a positive relation with cannabis
use after six months (i.e., a stronger approach bias was associated with more cannabis
consumption). Thus, to summarize, just like rSRC scores, correlations between AAT scores
and drug-related behavior can be negative, positive, or non-significant .
5.2.3. Experimental research.
Even though experimental research is the most powerful approach to make claims
about validity, studies employing such designs are scarce within the literature on implicit
measures of “wanting” and “liking”. Wanting, and Approach-Avoid IATs.
First, in addition to examining differences between smokers and non-smokers, Tibboel
et al. (2011) performed an experimental manipulation of “wanting”. The authors tested a
group of smokers in two sessions, once when they were deprived of smoking for 12 hours,
and once after they had just smoked. They reasoned that wanting scores should increase when
smokers are deprived (e.g., Field, Mogg, & Bradley, 2004). However, contrary to their
hypotheses, they found that depriving smokers did not affect scores on the wanting IAT, but
that deprivation caused an increase in scores on the liking IAT. These results are the exact
opposite of what IST would propose.
Using an approach/avoid IAT, Farris and Ostafin (2008) showed that at-risk drinkers
had an increased approach bias after consuming alcohol compared to a baseline. However,
participants were free to decide how much alcohol they consumed, which implies that some
participants might have been sated whereas others were merely primed. This is problematic
because priming should increase craving whereas satiation should not (e.g., Christiansen,
Rose, Cole, & Field, 2013). Moreover, there was no control group and no placebo condition.
Both elements limit the extent to which we can draw solid conclusions. Thus, for both the
wanting IAT and the approach/avoid IAT, experimental evidence regarding their validity is
limited and unclear. Measures of automatic motivational tendencies
Christiansen et al. (2013) did an elaborate study in which social drinkers were asked to
perform an rSRC task in three different sessions. In each session, they were first asked to
consume a beverage. In one session, this beverage contained a priming dose of alcohol, in
another session the drink was a placebo (i.e., participants were informed that the drink
contained alcohol, but in fact it did not), and in another session the beverage was a non-
alcoholic control drink (and participants were explicitly told that it did not contain alcohol).
Results showed that participants had positive rSRC scores both after consuming alcohol and
after consuming the placebo, but not after consuming the control drink. A similar study by
Schoenmakers, Wiers, and Field (2008) revealed no difference in rSRC scores after alcohol
consumption and placebo consumption either. Thus, it seems that not only the manipulation
of “wanting” (i.e., consuming a priming dose of alcohol) but also the manipulation of wanting
(i.e., installing the conscious belief that alcohol is consumed) affected rSRC scores,
suggesting that the rSRC is not (only) a measure of “wanting” but also of more conscious
Watson, De Wit, Cousijn, Hommel, and Wiers (2013) have used an AAT to assess
“wanting” for nicotine in two groups of smokers. One group was tested just before and after
smoking. A control group of smokers performed the AAT twice in a deprived state. Their
results showed that AAT scores were positively correlated with craving when smokers were
deprived. However, smoking a cigarette led to reduced explicit craving but an increase in
AAT scores. For smokers who did not get a chance to smoke, AAT scores remained stable.
This is clearly in contrast with the predictions of IST. If AAT scores are an index of
“wanting”, these scores should be reduced when smokers have just smoked a cigarette.
In their clinical study, Wiers et al. (2011) examined differences in both approach-
avoid IAT scores and AAT scores in alcohol-dependent patients before and after treatment.
Both the IAT and AAT scores were positive initially, and became negative after treatment.
However, the change in IAT and AAT scores (pre and post treatment) could not predict
relapse. If one assumes that treatment decreases “wanting”, these results partly support the
validity of these measures: a decrease in “wanting” causes a decrease in wanting scores.
However, if one assumes that relapse is a key indicator of “wanting” (e.g., Robinson &
Berridge, 2001), these results are less convincing.
Even though the data are limited, if anything, they seem unable to provide (clear)
evidence regarding the validity of these implicit measures.
6. Discussion
Overall, both our theoretical analysis and our review of the empirical evidence
provides little support for the validity of different versions of the valence IAT as measures of
“liking” (defined as implicit hedonic impact), and of different versions of the arousal IAT, the
approach/avoid IAT, the wanting IAT, the AAT, and the rSRC as measures of “wanting”
(defined as implicit incentive salience). Although these doubts stem partly from discrepancies
between different definitions and conflicting results (most notable in the IAT studies), they
also stem from the lack of studies that allow us to draw clear conclusions regarding the
validity of these measures. Given this state of affairs, we conclude that more research is
needed that is directed at rectifying a number of shortcomings of the existing literature.
6.1. Experimental designs and longitudinal studies
First, in line with the recommendations of (Borsboom et al., 2004), we suggest that
future studies using different implicit measures of “wanting” and “liking” use experimental
designs that should aid us in testing a causal relationship between changes in the to-be-
measured construct and changes in implicit measures. “Wanting” can be manipulated by, for
instance, depriving participants from a substance for a sufficient amount of time (e.g., Tibboel
et al., 2011) or by a craving induction (e.g., Dewit & Chutuape, 1993). “Liking”, on the other
hand, can be manipulated by flavor evaluative conditioning (e.g., Wardle, Mitchell, &
Lovibond, 2007). Ideally, researchers should include both implicit measures of “wanting” and
implicit measures of “liking” to examine whether these measures are differentially affected by
these manipulations and to examine the extent to which each of these measures explains
variance on different criterion variables (e.g., drug use, relapse).
On a related note, it is important to examine the development of substance (ab)use
over time and the influence of interventions and therapies on substance (ab)use. In this
respect, longitudinal studies examining the development of substance (ab)use over time are
invaluable. A serious shortcoming of the existing literature is the paucity of longitudinal
studies that examine the development of addiction and the changes in scores on implicit
measures of “wanting” and “liking”. Researchers could, for instance, repeatedly test youths
who are at risk for developing addiction on both “wanting” and “liking” measures. If these
measures are valid, “liking” scores should decrease, whereas “wanting” scores should
increase when casual drug use develops into an addiction. One recommendable study was
performed by Peeters et al. (2013). They tested a large group of adolescents who were at risk
for developing substance abuse problems at two points in time, with an interval of six months.
One of their most important findings was that in adolescents with weak inhibition skills,
higher AAT scores at T1 were related to more self-reported alcohol use at T2. Similarly, more
intervention studies are needed, in which participants perform “wanting” and “liking” tests
both before and after treatment (see Wiers et al., 2011, for an important exception). It might,
however, be less straightforward to formulate hypotheses for these intervention studies
because IST postulates that sensitized “wanting” persists for a long time even after
withdrawal symptoms have disappeared.
6.2. Sample selection and context effects
This brings us to a second issue that is often overlooked in the existing literature on
implicit measures of “wanting” and “liking”, and that may explain some of the discrepant
findings: IST assumes that there are differences between drug users and drug addicts.
According to the theory, both “liking” and “wanting” initially play an important role in drug
pursuit. The “wanting” system becomes sensitized and the role of “liking” diminishes only
after frequent drug use (e.g., Robinson & Berridge, 1993). Researchers should take this into
account when they choose their sample. For instance, the drinking behavior of a heavy social
drinker may be driven by both “liking” and “wanting”, whereas only “wanting” might play a
role in clinical levels of alcoholism. Moreover, a sample of participants who classify as heavy
social drinkers might contain both drinkers who are in the earlier developmental stage of
addiction (i.e., the stadium that involves both “wanting” and “liking”) as well as drinkers who
are in later stages (i.e., stages in which “liking” no longer influences their behavior).
Contrasting findings and null-results could thus be due to both hidden across- and within-
study variability in the selection of participants.
On a related note, the contrasting results we reviewed in this paper might be due to
another sample-related question. Whereas IST suggests that “wanting” plays a crucial role in
the development and maintenance of all kinds of substance abuse, research has shown that
different types of reinforcement play different roles in different types of substance abuse. For
instance, intoxication and bingeing play an important role in psychostimulant addiction,
whereas they play a considerably less crucial role in nicotine addiction (in which negative
affect and withdrawal are the most important motivators for continuing drug use; e.g., Koob
& Volkow, 2010). Furthermore, it is possible that the weight of these different processes
changes with the development of an addiction. For instance, whereas alcohol abuse might
start with an emphasis on the motivation to become intoxicated, this might change gradually
in a motivation to avoid the negative consequences of withdrawal. Note that this can differ
between individuals as well (e.g., Koob & Le Moal, 2008). In other words, it is possible that
some substances automatically trigger memories or expectations regarding the appetitive
characteristics of drugs, whereas others substances are associated more with their negative
effects. Importantly, implicit measures of “wanting” and “liking” might be sensitive to these
different drug-effects. It is possible that our “wanting” and “liking” measures are not valid
measures of “wanting” and “liking” but instead reflect the automatically activated
expectations regarding specific drug effects.
Similarly, substance abusers might differ in their reactions depending on the context
they are tested in. Research has shown that generally, implicit measures are malleable (see
Han, Czellar, Olson & Fazio, 2010, for a review and some interesting considerations). Hence,
the results of studies on addiction might be influenced by the circumstances under which
participants were tested. For instance, a treatment context might trigger only negative affect
and avoidance tendencies, whereas another (e.g., social) context might trigger more positive
affect and approach tendencies (e.g., Barkby et al., 2012; Dickson et al., 2013; Spruyt et al.,
2012). Future research should examine the influence of these factors more thoroughly. Note,
however, that if effects indeed differ according to the context, this adds further evidence for
our hypothesis that the implicit measures we address in this paper are not valid “wanting” and
“liking” measures: “wanting” and “liking” are assumed to be more hard-wired and stable (see
Field, Mogg, Mann, Bennett & Bradley, 2013, for similar arguments regarding the role of
attentional bias in alcohol abuse) and should not be affected by one’s surroundings.
6.3. Behavioral manifestations of “wanting” and “liking”
Third, we also need clearer hypotheses regarding the behavioral manifestations of
“wanting” and “liking”. Without such hypotheses, it is impossible to determine the criterion
variables that implicit measures of “wanting” and “liking” should relate to. For instance, it
would be helpful to know whether and when one should expect correlations between implicit
and explicit measures (e.g., subjective pleasantness or craving ratings). According to IST
(e.g., Berridge, 2010) “wanting” and “liking” are independent of conscious wanting and
liking, suggesting that implicit measures and explicit measures should not correlate. However,
many researchers do calculate these correlations and draw conclusions on the basis of these
correlations (e.g., Houben & Wiers, 2007a; 2007b; 2008). Furthermore, clarity is obscured by
the tendency of addiction researchers to include in their studies a large range of explicit
ratings and questionnaires. For instance, studies on alcohol consumption often incorporate
questionnaires to assess alcohol problems such as the Rutgers Alcohol Problem Index (White
& Labouvie, 1989) and the Alcohol Use Disorders Identification Test (Babor, Higgings-
Biddle, Saunders, & Monteiro, 2006), alcohol craving questionnaires such as the Desire for
Alcohol Questionnaire (Love, James, & Willner, 1998), alcohol consumption questionnaires
such as the Alcohol Timeline Follow Back (Sobell, & Sobell, 1995), Likert scales concerning
alcohol expectancies (e.g., Houben et al., 2009), as well as other measures of explicit alcohol
attitudes (e.g., Houben & Wiers, 2006). Commonly, correlations are found for only a (small)
subset of criterion measures, but which correlations are significant varies between studies.
Hence, when correlations are observed, they might well be spurious, especially because it is
uncommon to correct for multiple statistical tests. If these correlations are not spurious, the
question remains why results are inconsistent across studies. In order to advance, researchers
need to make clear a priori predictions regarding correlations between implicit measures and
criterion measures and subsequently report the results concerning all tested hypotheses.
Exploratory analyses remain useful but should be acknowledged as such.
In this context, it is also important to note that the reasons behind the difficulty to
specify criterion variables for implicit measures of “wanting” and “liking” might be related to
the manner in which IST has been developed. Berridge (1996) stressed the benefits of the
volatile definitions of “wanting” and “liking” because there should be room for these
definitions to grow with the increasing amount of research findings. However, to be able to
test a theory, strict definitions and clear hypotheses are needed, because these are the only
kinds of hypotheses that are falsifiable (e.g., Kuhn, 1962). From this perspective, it is
troublesome that IST does not clearly commit itself to the possibility that “wanting” and
“liking” (as mental processes) could be measured behaviorally in humans. This raises the
question whether adopting other definitions allows for more straightforward hypothesis-
testing. Whereas the effect definition might present a too simplified image of drug use for
human subjects (e.g., Wiers et al., 2007), and the utility definition is currently restrained to
theoretical work (e.g., Berridge & Aldridge, 2008), the neurological definition might offer
some perspective. For instance, in some papers (e.g., Berridge, 2004) “liking” is reduced to a
brain reaction that leads to an experience of pleasure that is not linked to an object of desire.
This means that we cannot measure “liking” for a substance, as “liking” is per definition not
linked to any object at all. If this assumption is correct, this leaves only possibilities for
neuropsychological research to shed more light on the IST hypotheses: it suggests that the
only possible way to assess “liking” is to examine brain activity in hedonic hotspots exactly
when a participant consumes a substance. Although such a theoretical conceptualization
might have merit, it does seriously constrain the possibility of constructing implicit or other
behavioral measures of “liking”.
6.4. Neural manifestations of “wanting” and “liking”
This leads us to a final point that neural manifestations of “wanting” and “liking”
might eventually provide the best criterion variable to assess the validity of implicit
(behavioral) measures of “wanting” and “liking”. Whereas IST is not perfectly clear
concerning the behavioral manifestations of “wanting” and “liking”, it does make more
specific hypotheses concerning the neural manifestation of these processes (e.g., Berridge,
2010; Berridge & Kringelbach, 2008). Cognitive neuroscience could provide us with the tools
to test the hypotheses of IST and lend a helping hand in examining the validity of implicit
measures of “wanting” and “liking”. Two examples of studies that could further our
understanding of neuropsychological mechanisms that underlie drug abuse and their relation
to implicit measures were performed by Cousijn et al. (2012) and Korucuoglu, Gladwin, and
Wiers (2014). Cousijn and colleagues asked a group of heavy cannabis users and a group of
controls to perform an rSRC task in the fMRI scanner. Their results did not reveal differences
in rSRC scored between groups, but they did show that in heavy cannabis users, approach
bias activations in several fronto-limbic areas were associated with cannabis use, and that
approach bias activations in the dorsolateral prefrontal cortex and anterior cingulate cortex
could predict cannabis problem severity six months after testing. In a recent EEG study,
Korucuoglu et al. (2014) examined the influence of a priming dose of alcohol or a placebo on
performance and its influence on event-related desynchronization in the beta band (beta
band ERD) which is assumed to provide an index of advanced response preparation. Whereas
their behavioral data only showed a (marginal) AAT effect, and no effect of priming dose,
their EEG data showed an interesting pattern: Beta band ERD was higher in the congruent
block (i.e., approach alcohol; avoid soft drink) in the alcohol condition compared to the
placebo condition. Moreover, within the alcohol condition, beta band ERD was higher for
approach alcohol trials compared to avoid alcohol trials, and higher for avoid soft drink trials
compared to approach soft drink trials. Both studies show that implicit measures alone might
not be sensitive enough to pick up on differences between groups or conditions, but they
might still be related to neuropsychological processes that play a role in incentive salience
attribution. These studies provide us with new and exciting ways to gain more insight into the
questions raised in this paper.
7. Conclusion
In this paper, we evaluated existing efforts to measure the mental processes of
“wanting” and “liking” in implicit ways and uncovered several inconsistencies that need to be
addressed in future research. Of course, these problems do not detract from the value of IST.
This theory has given impetus to groundbreaking research with non-human (e.g., Wyvell &
Berridge, 2000) and human (e.g., Wiers & Stacy, 2006) subjects, on behavioral and
neuropsychological levels (e.g., Evans et al., 2006), and has impacted a wide range of impulse
control disorders (e.g., Noël, Brevers, & Bechara, 2013). It has redefined the role of dopamine
in incentive salience attribution, and it has thereby redefined reward in general (e.g., Berridge
& Robinson, 1998). These developments have led to an increasingly better understanding of
It must be noted that the task used in this study was not a regular AAT, because it required that stimuli were
judged on the relevant stimulus dimension (i.e., alcohol or soft drinks) whereas in most AATs, stimuli are judged
on a different dimension (e.g., portrait or landscape).
the role of reward and affect in the development and maintenance of addiction. However, it is
important to note that even though the way is paved, we have not yet arrived at our
destination, in particular with regard to the definition and (implicit) measurement of the core
concepts of “wanting” and “liking” in humans. We therefore hope that our recommendations
can further improve both empirical and theoretical insights into incentive sensitization and its
role in addiction.
Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (2006). AUDIT: The
Alcohol Use Disorders Identification Test: Guidelines for use in primary care, second
edition, Retrieved June 24, 2006.
Bargh, J. A. (1994). The Four Horsemen of automaticity: Awareness, efficiency, intention,
and control in social cognition. In R. S. Wyer, Jr., & T. K. Srull (Eds.), Handbook of
social cognition (2nd ed., pp. 1-40). Hillsdale, NJ: Erlbaum.
Barkby, H., Dickson, J. M., Roper, L., & Field, M. (2012). To approach or to avoid alcohol?
Automatic and self-reported motivational tendencies in alcohol dependence.
Alcoholism: Clinical and Experimental Research, 36, 361-368.
Berridge, K. C. (1996). Food reward: Brain substrates of wanting and liking. Neuroscience &
Biobehavioral Reviews, 20, 1-25.
Berridge, K. C. (2004). Motivation concepts in behavioral neuroscience. Physiology &
Behavior, 81, 179-209.
Berridge, K. C. (2010). Wanting and liking: Observations from the neuroscience and
psychology laboratory. Inquiry, 52, 1-18.
Berridge, K. C., & Aldridge, J. W. (2008). Decision utility, the brain, and pursuit of hedonic
goals. Social Cognition, 26, 621-646.
Berridge, K. C. & Kringelbach, M. L. (2008). Affective neuroscience of pleasure: Reward in
humans and animals. Psychopharmacology, 199, 457-480.
Berridge, K. C. & O’Doherty, J. P. (2014). From experienced utility to decision utility. In P.
W. Glimcher & E. Fehr (Eds.), Neuroeconomics (Second Edition) (pp. 335-351).
Academic Press, San Diego.
Berridge, K. C., & Robinson, T. E. (1995). The mind of an addicted brain: Neural
sensitization of wanting versus liking. Current Directions In Psychological Science, 4,
71-76. DOI: 10.1111/1467-8721.ep10772316.
Berridge, K. C., & Robinson, T. E. (1998). What is the role of dopamine in reward: Hedonic
impact, reward learning, or incentive salience? Brain Research Reviews, 28, 309-369.
Berridge, K. C., & Robinson, T. E. (2003). Parsing reward. TRENDS in Neurosciences, 26,
Berridge, K. C., Robinson, T. E., & Aldridge, J. W. (2009). Dissecting components of reward:
"liking", "wanting", and learning. Current Opinion in Pharmacology, 9, 65-73.
Berridge, K. C., Venier, I. L., & Robinson, T. E. (1989). Taste reactivity analysis of 6-
hydroxydopamine-induced aphagia: Implications for arousal and anhedonia
hypotheses of dopamine function. Behavioral Neuroscience, 103, 36-45.
Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2004). The concept of validity.
Psychological Review, 111, 1061-1071.
Bushmann, B. J., Moeller, S. J., Konrath, S., & Crocker, J. (2012). Investigating the link
between liking versus wanting self-esteem and depression in a nationally
representative sample of American adults. Journal of Personality, 80, 1453 1469.
Catania, A. C. (2013). A Natural Science of Behavior. Review of General Psychology, 17,
Christiansen, P., Cole, J., Goudie, A. J., & Field, M. (2012). Components of behavioural
impulsivity and automatic cue approach predict unique variance in hazardous
drinking. Psychopharmacology, 219, 501-510.
Christiansen, P., Rose, A. K., Cole, J. C., & Field, M. (2013). A comparison of the anticipated
and pharmacological effects of alcohol on cognitive bias, executive function, craving,
and ad-lib drinking. Journal of Psychopharmacology, 27, 84-92.
Cohn, A. M., Cameron, A. Y., Udo, T., Hagman, B. T., Mitchell, J., Bramm, S., & Ehlke, S.
(2012). Delineating potential mechanisms of implicit alcohol cognitions: Drinking
restraint, negative affect, and their relationship with approach alcohol associations.
Psychology of Addictive Behaviors, 26, 318-324.
Cousijn, J., Goudriaan, A. E., Ridderinkhof, K. R., Van den Brink, W., Veltman, D. J., &
Wiers, R. W. (2012). Approach-Bias Predicts Development of Cannabis Problem
Severity in Heavy Cannabis Users: Results from a Prospective FMRI Study. PLoS
One, 7, e42394.
Cousijn, J., Goudriaan, A. E., & Wiers, R. W. (2011). Reaching out towards cannabis:
approachbias in heavy cannabis users predicts changes in cannabis use. Addiction,
106, 1667-1674.
Cousijn, J., Snoek, R. W. M., & Wiers, R. W. (2013). Cannabis intoxication inhibits
avoidance action tendencies: a field study in the Amsterdam coffee shops.
Psychopharmacology, 229, 167-176.
De Houwer, J., Crombez, G., Koster, E. H. W., & De Beul, N. (2004). Implicit alcohol-related
cognitions in a clinical sample of heavy drinkers. Journal of Behaviour Therapy and
Experimental Psychiatry, 35, 275-286.
De Houwer, J., Custers, R., & De Clercq, A. (2006). Do smokers have a negative implicit
attitude toward smoking? Cognition and Emotion, 20, 1274-1284.
De Houwer, J., & De Bruycker, E. (2007). The identification-EAST as a valid measure of
implicit attitudes toward alcohol-related stimuli. Journal of Behavior Therapy and
Experimental Psychiatry, 38, 133-143.
De Houwer, J., Teige-Mocigemba, S., Spruyt, A., & Moors, A. (2009). Implicit Measures: A
Normative Analysis and Review. Psychological Bulletin, 135, 347-368.
De Wit, H., & Chutuape, M. A. (1993). Increased ethanol choice in social drinkers following
ethanol preload. Behavioral Pharmacology, 4, 29-36.
Dickson, J. M., Gately, C., & Field. (2013). Alcohol dependent patients have weak negative
rather than strong positive implicit alcohol associations. Psychopharmacology, 228,
Eberl, C., Wiers, R. W., Pawelczack, S., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2013).
Approach bias modification in alcohol dependence: Do clinical effects replicate and
for whom does it work best? Developmental Cognitive Neuroscience, 4, 38-51.
Eder, A. B., & Rothermund, K. (2008). When do motor behaviors (mis) match affective
stimuli? An evaluative coding view of approach and avoidance reactions. Journal of
Experimental Psychology: General, 137, 262-281.
Evans, A. H., Pavese, N., Lawrence, A. D., Tai, Y. F., Appel, S., Doder, M., Brooks, D. J.,
Lees, A. J., & Piccini, P. (2006). Compulsive drug use linked to sensitized ventral
striatal dopamine transmission. Annals of Neurology, 59, 852858.
Farris, S.R., & Ostafin, B.D. (2008). Alcohol consumption primes automatic alcohol
motivation. American Journal of Drug and Alcohol Abuse, 34, 703-711.
Field, M., Caren, R., Fernie, G., & De Houwer, J. (2011). Alcohol approach tendencies in
heavy drinkers: Comparison of effects in a Relevant Stimulus-Response Compatibility
Task and an approach / avoidance Simon task. Psychology of Addictive Behaviors, 25,
Field, M., Kiernan, A., Eastwood, B., & Child, R. (2008). Rapid approach responses to
alcohol cues in heavy drinkers. Journal of Behavior Therapy and Experimental
Psychiatry, 39, 209-218.
Field, M., Mogg, K., & Bradley, B. P. (2004). Eye movements to smoking-related cues:
effects of nicotine deprivation. Psychopharmacology, 173, 116-123.
Field, M., Mogg, K., & Bradley, B. P. (2005). Alcohol increases cognitive biases for smoking
cues in smokers. Psychopharmacology, 180, 63-72.
Field, M., Mogg, K., & Bradley, B. P. (2005). Craving and cognitive biases for alcohol cues
in social drinkers. Alcohol & Alcoholism, 40, 504-510.
Field, M., Mogg, K., Mann, B., Bennett, G. A., & Bradley, B. P. (2013). Attentional biases in
abstinent alcoholics and their association with craving. Psychology of Addictive
Behaviors, 27, 71-80.
Finlayson, G., King, N., & Blundell, J. E. (2007). Is it possible to dissociate 'liking' and
'wanting' for foods in humans? A novel experimental procedure. Physiology &
Behaviour, 90, 36-42.
Gawronski, B., & De Houwer, J. (2014). Implicit measures in social and personality
psychology. In H. T. Reis, & C. M. Judd (Eds.), Handbook of research methods in
social and personality psychology (2nd edition, pp. 283-310). New York, NY:
Cambridge University Press.
Gawronski, B., Hofmann, W., & Wilbur, C. J. (2006). Are “implicit” attitudes unconscious?
Consciousness and Cognition, 15, 485-499.
Govan, C., & Williams, K. D. (2004). Changing the affective valence oft he stimulus items
influences the IAT by re-defining the category labels. Journal of Experimental Social
Psychology, 40, 357-365.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual
differences in implicit cognition: The implicit association test. Journal of Personality
and Social Psychology, 74, 1464-1480.
Hahn, A., & Gawronski, B. (2014). Do implicit evaluations reflect unconscious attitudes?
Behavioral and Brian Sciences, 37, 28 29.
Hahn, A., Judd, C. M., Hirsh, H. K., Blair, I. V. (2014). Awareness of implicit attitudes.
Journal of Experimental Psychology: General, 143, 1369 1392.
Han, H.A., Czellar, S., Olson, M. A., & Fazio, R. H. (2010). Malleability of attitudes or
malleability of the IAT? Journal of Experimental Social Psychology, 46, 286-298.
Heerey, E. A., & Gold, J. M. (2007). Patients with Schizophrenia demonstrate dissociation
between affective experience and motivated behavior. Journal of Abnormal
Psychology, 116, 268 278.
Hobbs, M., Remington, B., & Glautier, S. (2005). Dissociation of wanting and liking for
alcohol in humans: a test of the incentive-sensitisation theory. Psychopharmacology,
178, 493-499.
Houben, K., Rothermund, K., & Wiers, R. W. (2009). Predicting alcohol use with a recoding-
free variant of the Implicit Association Test. Addictive Behaviors, 34, 487-489.
Houben, K., & Wiers, R. W. (2006). Assessing Implicit Alcohol Associations with the
Implicit Association Test: Fact or Artifact? Addictive Behaviors, 31, 1346-1362.
Houben, K., & Wiers, R. W. (2007a). Are drinkers implicitly positive about drinking alcohol?
Personalizing the alcohol-IAT to reduce negative extrapersonal contamination.
Alcohol and Alcoholism, 42, 301-307.
Houben, K., & Wiers, R. W. (2007b). Personalizing the alcohol-IAT with individualized
stimuli: Relationship with drinking behavior and drinking-related problems. Addictive
Behaviors, 32, 2852-2864.
Houben, K., & Wiers, R. W. (2008). Implicitly positive about alcohol? Implicit positive
associations predict drinking behavior. Addictive Behaviors, 33, 979-986.
Jajodia, A., & Earlywine, M. (2003). Measuring alcohol expectancies with the implicit
association test. Psychology of Addictive Behaviors, 17, 126-133.
Karpinski, A., & Steinman, R. B. (2006). The single category implicit association test as a
measure of implicit social cognition. Journal of Personality and Social Psychology,
91, 16-32.
Kohls, G., Chevallier, C., Troiani, V., & Schultz, R. T. (2012). Social ‘wanting’ dysfunction
in autism : neurobiological underpinnings and treatment implications. Journal of
Neurodevelopmental Disorders, 4, 1 20.
Koob, G. F., & Le Moal, M. (2008). Addiction and the brain antireward system. Annual
Review of Psychology, 59, 29-53.
Koob, G. F., & Volkow, N. (2010). Neurocircuitry of addiction. Neuropsychopharmacology,
35, 217-238.
Korucuoglu, Ö. , Gladwin, T. E., & Wiers, R. W. (2014). Preparing to approach or avoid
alcohol: EEG correlates, and acute alcohol effects. Neuroscience Letters, 559, 199-
Krieglmeyer, R., De Houwer, J., & Deutsch, R. (2013). On the nature of automatically
triggered approach-avoidance behavior. Emotion Review, 5, 280-284.
Kuhn, T. S. (1962). The Structure of Scientific Revolutions (3rd ed.). Chicago: University of
Chicago Press.
Lades, L. K. (2014). Impulsive consumption and reflexive thought : Nudging ethical
consumer behavior.
Leyton, M. (2007). Conditioned and sensitized responses to stimulant drugs in humans.
Progress in. Neuropsychopharmacology and Biological Psychiatry, 31, 16011613.
Lindgren, K. P., Neighbors, C., Teachman, B. A., Gasser, M. L., Kaysen, D., Norris, J., &
Wiers, R. W. (2015). Habit doesn’t make the predictions stronger: Implicit alcohol
associations and habitualness predict drinking uniquely. Addictive Behaviors, 45, 139
Lindgren, K. P., Neighbors, C., Teachman, B. A., Wiers, R. W., Westgate, E., & Greenwald,
A. G. (2013). I drink therefore I am: Validating alcohol-related implicit association
tests. Psychology of Addictive Behaviors, 27, 1 13.
Love, A., James, D., & Willner, P., 1998. A comparison of two alcohol craving
questionnaires. Addiction 93, 1091-1102.
McCarthy, D. M., & Thompsen, D. M. (2006). Implicit and explicit measures of alcohol and
smoking cognitions. Psychology of Addictive Behaviors, 20, 436-444.
Mogg, K., Bradley, B. P., Field, M., & De Houwer, J. (2003). Eye movements to smoking-
related pictures in smokers: Relationship between attentional biases and implicit and
explicit measures of stimulus valence. Addiction, 98, 825-836.
Moors, A., & De Houwer, J. (2006). Automaticity: A theoretical and conceptual analysis.
Psychological Bulletin, 132, 297-326.
Noël, X., Brevers, D., & Bechara, A. (2013). A neurocognitive approach to understanding the
neurobiology of addiction. Current opinion in neurobiology, 23, 632-638.
Nosek, B. A., Hawkins, C. B., & Frazier, R. S. (2011). Implicit social cognition: From
measures to mechanisms. Trends in Cognitive Sciences, 15, 152159.
Olson, M. A., & Fazio, R. H. (2004). Reducing the influence of extrapersonal associations on
the implicit association test: Personalizing the IAT. Journal of Personality and Social
Psychology, 86, 653-667.
Ostafin, B. D., Bauer, C., & Myxter, P. (2012). Mindfulness decouples the relation between
automatic alcohol motivation and heavy drinking. Journal of Social and Clinical
Psychology, 31, 729-745.
Ostafin, B. D., & Brooks, J. J. (2011). Drinking for relief: Negative affect increases automatic
alcohol motivation in coping-motivated drinkers. Motivation and Emotion, 35, 285-
Ostafin, B. D., Marlatt, G. A., & Greenwald, A. G. (2008). Drinking without thinking: An
implicit measure of alcohol motivation predicts failure to control alcohol use.
Behaviour Research and Therapy, 46, 1210-1219.
Ostafin, B. D., Palfai, T. P., & Wechsler, C. E. (2003). The accessibility of motivational
tendencies toward alcohol: Approach, avoidance, and disinhibited drinking.
Experimental and Clinical Psychopharmacology, 11, 294-301.
Palfai, T. P., & Ostafin, B. D. (2003). Alcohol-related motivational tendencies in hazardous
drinkers: Assessing implicit response tendencies using the modified-IAT. Behaviour
Research and Therapy, 41, 1149-1162.
Parkinson, J. A., Roberts, A. C., Everitt, B. J., & Di Ciano, P. (2005). Acquisition of
instrumental conditioned reinforcement is resistant to the devaluation of the
unconditioned stimulus. The Quarterly Journal of Experimental Psychology, 58B, 19-
Peeters, M., Monshouwer, K., Van de Schoot, R. A. G. J., Janssen, T., Vollebergh, W. A. M.,
& Wiers, R. (2013). Automatic Processes and the Drinking Behavior in Early
Adolescence: A Prospective Study. Alcoholism Clinical and Experimental Research,
37, 1737-1744.
Robinson, M. D., Meier, B. P., Zetocha, K. J., & McCaul, K. D. (2005). Smoking and the
implicit association test : When the contrast category determines the theoretical
conclusions. Basic and Applied Social Psychology, 27, 201 212.
Robinson, T. E., & Berridge, K. C. (1993). The neural basis of drug craving: An incentive-
sensitization theory of addiction. Brain Research Review, 18, 247-291.
Robinson, T. E., & Berridge, K. C. (2000). The psychology and neurobiology of addiction:
An incentive-sensitization view. Addiction, 95, S91-S117.
Robinson, T. E., & Berridge, K. C. (2001). Mechanisms of action of addictive stimuli:
Incentive-sensitization and addiction. Addiction, 96, 103-114.
Robinson, T. E. & K. C. Berridge (2003). Addiction. Annual Review of Psychology, 54, 1-29.
Robinson, T. E. & K. C. Berridge (2008). The incentive sensitization theory of addiction:
Some current issues. Philosophical Transactions of the Royal Society B-Biological
Sciences, 363, 3137-3146.
Schoenmakers, T., Wiers, R. W., & Field, M. (2008). Effects of a low dose of alcohol on
cognitive biases and craving in heavy drinkers. Psychopharmacology, 197, 169-178.
Sharbanee, J. M., Stritzke, W. G. K., Wiers, R. W., Young, P., Rinck, M., & MacLeod, C.
(2013). The interaction of approach-alcohol tendencies, working memory capacity,
and current task goals predicts the inability to regulate drinking behavior. Psychology
of Addictive Behaviors, 27, 649-661.
Sharbanee, J. M., Stritzke, W. G. K., Jamalludin, M. E., & Wiers, R. W. (2014). Approach-
alcohol action tendencies can be inhibited by cognitive load. Psychopharmacology,
231, 967-975.
Sobell, L. C. & Sobell, M. B. (1995). Alcohol Timeline Followback Users’ Manual. Toronto,
Canada: Addiction Research Foundation.Solomon, R. L. & Corbit, J. D. (1973). An
Opponent-Process Theory of Motivation: II.Cigarette addiction. Journal of Abnormal
Psychology, 81, 158171.
Spruyt, A., De Houwer, J., Tibboel, H., Verschuere, B., Crombez, G., Verbanck, P., Hanak,
C., Brevers, D., & Noël, X. (in press). On the predictive validity of automatically
activated approach/avoidance tendencies in abstaining alcohol-dependent patients.
Drug and Alcohol Dependence.
Swanson, J. E., Rudman, L. A., & Greenwald, A. G. (2001). Using the implicit association
test to investigate attitude-behaviour consistency for stigmatized behavior. Cognition
and Emotion, 15, 207 230.
Talmi, D., Seymour, B., Dayan, P., & Dolan, R. J. (2008). Human pavlovian-instrumental
transfer. Journal of Neuroscience, 28, 360-368.
Tibboel, H., De Houwer, J., Spruyt, A., Brevers, D., Roy, E., & Noël, X. (2015). Heavy
social drinkers score higher on implicit wanting and liking for alcohol than alcohol-
dependent patients and light social drinkers. Journal of Behavior Therapy and
Experimental Psychiatry, 48, 185 191.
Tibboel, H., De Houwer, J., Spruyt, A., Field, M., Kemps, E., & Crombez, G. (2011). Testing
the validity of implicit measures of wanting and liking. Journal of Behavior Therapy
and Experimental Psychiatry, 42, 284-292.
Uslaner, J. M., Acerbo, M. J., Jones, S. A., & Robinson, T. E. (2006). The attribution of
incentive salience to a stimulus that signals an intravenous injection of cocaine.
Behavioural Brain Research, 169, 320324.
Van Bockstaele, B., Verschuere, B., Tibboel, H., De Houwer, J., Crombez, G. & Koster,
E.H.W. (2014). A review of current evidence for the causal impact of attentional bias
on fear and anxiety. Psychological Bulletin, 140, 682-721.
Van den Wildenberg, E., Beckers, M, Van Lambaart, F., Conrod, P. & Wiers, R. W. (2006).
Is the strength of implicit alcohol associations correlated with alcohol-induced heart-
rate acceleration? Alcoholism, Clinical and Experimental Research, 30, 1336-1348.
Van Hemel - Ruiter, M., de Jong, P. J., Wiers, R. W. (2011). Appetitive and Regulatory
Processes in Young Adolescent Drinkers. Addictive Behaviors, 36, 18-26.
Voon, V., Mole, T. B., Banca, P., Porter, L., Morris, L., Mitchell, … Irvine, M. (2014).
Neural correlated of sexual cue reactivity in individuals with and without comopulsive
sexual behaviors. PLOS One, 9, 1 10.
Wardle, S. H., Mitchell, C. J., & Lovibond, P. F. (2007). Flavor evaluative conditioning
and contingency awareness. Learning & Behavior, 35, 233 -241.
Watson, P., De Wit, S., Cousijn, J., Hommel, B., & Wiers, R. W. (2013). Motivational
mechanisms underlying the approach bias to cigarettes. Journal of Experimental
Psychopathology, 4, 250-262.
White, H. R. & Labouvie, E. W. (1989). Towards the assessment of adolescent problem
drinking. Journal of Studies on Alcohol, 50, 30-37.
Wiers, R. W., Bartholow, B. D., Van den Wildenberg, E., Thush, C., Engels, R. C. M. E.,
Sher, K. J., Grenard, J., Ames, S. L., & Stacy, A. W. (2007). Automatic and controlled
processes and the development of addictive behaviors in adolescents: A review and a
model. Pharmacology, Biochemistry, and Behavior, 86, 263 283.
Wiers, C. E., Kühn, S., Javadi, A. H., Korucuoglu, O., Wiers, R. W., Walter, H., Gallinat, J.,
& Bermpohl, F. (2013). Automatic approach bias towards smoking cues is present
in smokers but not in ex-smokers. Psychopharmacology, 229, 187197.
Wiers, R. W., Eberl, C., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2011). Retraining
automatic action tendencies changes alcoholic patients’ approach bias for alcohol and
improves treatment outcome. Psychological Science, 22, 490-497.
Wiers, R. W., Houben, K., & De Kraker, J. (2007). Implicit cocaine associations in active
cocaine users and controls. Addictive Behaviors, 32, 1284-1289.
Wiers, R. W., Rinck, M., Dictus, M., & Van Den Wildenberg, E. (2009). Relatively strong
automatic appetitive action-tendencies in male carriers of the OPRM1 G-allele. Genes,
Brain and Behavior, 8, 101-106.
Wiers, R. W., Rinck, M. Kordts, R., Houben, K., & Strack, F. (2010). Re-training Automatic
Action-Tendencies to Approach Alcohol in Hazardous Drinkers. Addiction, 105, 279-
Wiers, R. W., & Stacy, A. W. (2006). Implicit cognition and addiction. Current Directions in
Psychological Science, 15, 292-296.
Wiers, R. W., Van Woerden, N., Smulders, F. T. Y., & De Jong, P. J. (2002). Implicit and
explicit alcohol-related cognitions in heavy and light drinkers. Journal of Abnormal
Psychology, 111, 648-658.
Winkielman, P., & Berridge, K. (2003). Irrational wanting and subrational liking: How
rudimentary motivational and affective processes shape preferences and choices.
Political Psychology, 24, 657-680.
Winkielman, P., & Berridge, K. C. (2004). Unconscious emotion. Current Directions in
Psychological Science, 13, 120-123.
Wyvell, C. L., & Berridge, K. C. (2000). Intra-accumbens amphetamine increases the
conditioned incentive salience of sucrose reward: Enhancement of reward "wanting"
without enhanced "liking" or response reinforcement. Journal of Neuroscience, 20,
Author Note
Correspondence regarding this article should be addressed to Helen Tibboel, Department of
Experimental-Clinical and Health Psychology, Ghent University, Henri Dunantlaan 2, 9000
Ghent, Belgium. Email: Helen Tibboel is a post-doctoral research
fellow funded by the Special Research Fund of Ghent University (BOF). Bram Van
Bockstaele is a postdoctoral researcher of the Research Priority Area YIELD of the
University of Amsterdam.
Part of this research was supported by grant BOF09/01M00209 of Ghent University to Jan De
Houwer. The authors thank Matt Field for his constructive feedback.
Table 1. Overview of different definitions of “wanting” and “liking”
Definition type
Mental process
(preconscious) craving; (the result of) incentive
salience; drug administration itself; mechanism that
relates the hedonic experience (“liking”) to an object
of desire
(preconscious) pleasure; core
hedonic impact; affective state;
brain reaction without an object of
the “motivational magnet” feature
the cue-triggered US “wanting” feature
the conditioned reinforcer feature
decision utility: the essence of the decision to pursue a
reward at the moment it is made
experienced utility: the hedonic
experience by the reward after it is
activity in mesolimbic dopamine-systems, affected by
dopamine, opioids, and GABA
activity in subcortical hedonic
hotspots, affected only by opioids
Table 2. Overview of different implicit measures of “wanting” and “liking”
Liking measures
To-be-measured attribute
Task summary
Implicit Association Test (IAT)
Categorization task of attributes ("I like" vs. "I do not like") and
targets (drug-related vs. control)
Categorization task of attributes ("pleasant (positive)" vs.
"unpleasant (negative)") and targets (drug-related vs. control)
Wanting measures
Implicit Association Test (IAT)
Categorization task of attributes ("I want" vs. "I do not want") and
targets (drug-related vs. control)
Categorization task of attributes ("Approach" vs. "Avoid") and
targets (drug-related vs. control)
Categorization task of attributes ("Active" vs. "Passive") and
targets (drug-related vs. control)
relevant Stimulus Response
Compatibility Test (rSRC)
Approach or avoid pictures on the basis of relevant (drug-related)
dimension using a mannikin
Affective Simon Task (AST)
Approach or avoid pictures on the basis of irrelevant (not drug-related)
dimension using a mannikin
Approach/avoidance task
Approach or avoid pictures on the basis of irrelevant (not drug-related)
dimension using a joystick
... Wanting refers to the motivation to achieve a rewarding result. Liking indicates the imagined anticipation of the hedonic state of pleasure resulting from achieving the goal [134,[137][138][139]. ...
Full-text available
Life transitions represent moments characterized by changes that can profoundly influence individual life trajectories and subjective well-being. Recently, career coaching has become an important method of helping people expand their self-awareness, facilitate personal development, and increase their performance in the school-to-work transition. Although previous studies have confirmed that one of the most important keys to the success of a coaching program is the quality of the relationship between coach and coachee, there is a lack of knowledge regarding how to objectively measure it. In this pilot study, we adopted a neuroscientific approach to introduce objective measures of the relationship between coach and coachee through the phases of a coaching session. A sample of 14 university students and a professional coach participated in career-coaching sessions while their affective states were measured by recording brain (EEG) and physiological (Skin conductance) activity. Electroencephalographic indicators of valence, arousal, and engagement showed differences between session phases, highlighting the possibility of a neurophysiological measurement of relational dynamics. Our results provide initial evidence that neurophysiological activity can be considered a way to understand differences in the coach-coachee relationship, thereby providing information on the effectiveness of coaching interventions and facilitating a better life transition from school to work.
... While any interpretation of the lack of a relationship should be done with caution, this result supports the possibility that these two tasks tap into distinct aspects of social reward processing. Similar dissociations in different aspects of reward processing-such as wanting and liking, have been reported for other stimuli such as food or drugs, and these have been linked with atypical behaviour [44,45]. Dissociations have also been reported between selfreport and behavioural measures of social reward responsivity in clinical populations [46,47]. ...
Full-text available
Children typically prefer to attend to social stimuli (e.g. faces, smiles) over non-social stimuli (e.g. natural scene, household objects). This preference for social stimuli is believed to be an essential building block for later social skills and healthy social development. Preference for social stimuli are typically measured using either passive viewing or instrumental choice paradigms, but not both. Since these paradigms likely tap into different mechanisms, the current study addresses this gap by administering both of these paradigms on an overlapping sample. In this study, we use a preferential looking task and an instrumental choice task to measure preference for social stimuli in 3–9 year old typically developing children. Children spent longer looking at social stimuli in the preferential looking task but did not show a similar preference for social rewards on the instrumental choice task. Task performance in these two paradigms were not correlated. Social skills were found to be positively related to the preference for social rewards on the choice task. This study points to putatively different mechanisms underlying the preference for social stimuli, and highlights the importance of choice of paradigms in measuring this construct.
... It is possible that in the case of intense use/behavior, the presence of the time range amplified "wanting, " while no such effect was present for low/moderate intensity uses/behaviors, as reflected in low "wanting-after." Another possible explanation is that moderate users are able to satisfy their "wanting" with use, but not intense users, supported by previous studies (12,78) finding that implicit "wanting" scores of smokers were unaffected by nicotine deprivation (12 or 10 h of deprivation vs. immediately after smoke). ...
Full-text available
A few studies have examined the changes in substance-and behavior-related "wanting" and "liking" of human subjects, the key properties of Incentive Sensitization Theory (IST). The aim of this study was to examine the dissociation between "wanting" and "liking" as a function of usage frequency, intensity, and subjective severity in individuals across four substances (alcohol, nicotine, cannabis, and other drugs) and ten behaviors (gambling, overeating, gaming, pornography use, sex, social media use, Internet use, TV-series watching, shopping, and work). Also, the potential roles of impulsivity and reward deficiency were investigated in "wanting," "liking," and wellbeing. The sex differences between "wanting" and "liking" were also examined. Based on our findings using structural equation modeling with 749 participants (503 women, M age = 35.7 years, SD = 11.84), who completed self-report questionnaires, "wanting" increased with the severity, frequency, and intensity of potentially problematic use, while "liking" did not change. Impulsivity positively predicted "wanting," and "wanting" positively predicted problem uses/behaviors. Reward deficiency positively predicted problem uses/behaviors, and both impulsivity and problem uses/behaviors negatively predicted wellbeing. Finally, women showed higher levels of "wanting," compared to men. These findings demonstrate the potential roles of incentive sensitization in both potentially problematic substance uses and behaviors.
... In recent years, the role that cognitive neuroscience technology can play in consumer behavior research has been increasingly recognized (Lin et al., 2018;Plassmann et al., 2015), and the EEG technique has been used to capture rapid changes and implicit attitudes that cannot be easily detected by behavioral experiments or surveys (Carbine et al., 2018). Food is a natural and powerful reward clue, so it is important to combine explicit and implicit measures when examining the individuals' differential responses to different foods (Tibboel, De Houwer, & van Bockstaele, 2015). The results of this study, once again, demonstrate the sensibility of the EEG signals (Andersen et al., 2019;Qiu et al., 2020). ...
People often associate features or stimuli across senses, and these crossmodal correspondences also function as criteria for judging whether a pair of features or stimuli is crossmodally congruent or incongruent. We conducted two studies to examine whether the crossmodal congruency between a drink color and a flavor could also elicit more positive ratings of this drink. In Study 1, the experimental task was to simultaneously view a photo of a colored drink and to taste a beverage whose flavor was either congruent or incongruent with the color. Consequently, the color-flavor congruent drinks received more positive ratings than the color-flavor incongruent ones. In Study 2, we conducted an EEG experiment to investigate both behavioral and oscillatory brain responses to the crossmodal congruency between the successively presented color and flavor. The results of subjective ratings did not reveal any significant difference between the color-flavor congruent and incongruent stimuli. However, the time-frequency analyses on the EEG data revealed that rating the color-flavor incongruent beverages elicited greater alpha band power in the parietal region, compared to the color-flavor congruent stimuli. Collectively, these findings revealed the influence of color-flavor congruency on the subjective ratings of beverages, and demonstrated how this effect was modulated by the temporal synchrony of the color and flavor cues.
Background Repeated action or inaction toward objects changes preferences for those objects. However, it remains unclear whether such training activates approach-avoidance motivation toward the objects, which leads to actual behavior. We conducted a pre-registered online experiment to examine whether approach and avoidance tendencies were affected by the experience of having executed or withheld a button-press response to a stimulus. Methods Participants ( N = 236) performed a Go/NoGo task in which they were asked to repeatedly execute a response to a picture of a mug ( i.e ., Go-primed stimulus) and suppress a response to another picture of a mug ( i.e ., NoGo-primed stimulus). They then received one of two manikin tasks, which were implicit association tests designed to assess approach–avoidance tendencies. One manikin task measured the reaction times of moving a manikin toward or away from the Go-primed stimulus and the other picture of a mug ( i.e ., unprimed stimulus). The other manikin task measured the reaction times of moving a manikin toward or away from the NoGo-primed stimulus and the unprimed stimulus. The participants then rated their preference for the Go-primed, NoGo-primed, and unprimed items. Results The Go-primed item was evaluated as more highly preferable than the unprimed item in the Go condition, while the NoGo-primed item was evaluated as less preferable than the unprimed item in the NoGo condition. In contrast, the mean approach/avoidance reaction times in the manikin task showed no difference between the Go-primed and unprimed stimuli or between the NoGo-primed and unprimed stimuli. Conclusion When participants repeatedly responded or inhibited their responses to an object, their explicit preference for the object increased or decreased, respectively. However, the effect did not occur in approach-avoidance behaviors toward the object.
Background and objective: Theoretical models propose that different cognitive biases are caused by a common underlying mechanism (incentive salience/"wanting") and should, therefore, be interrelated. Additionally, stronger impulsive processes should be related to weaker inhibitory abilities. However, these assumptions have hardly been empirically tested and key psychometric information have hardly been reported in samples of smokers. To extent previous research, the present study aimed (1) to estimate the reliability (split-half) of different cognitive bias measures and (2) to investigate associations between attention, approach and associative biases, response inhibition, and smoking-related variables. Methods: Eighty current, non-deprived smokers completed the following tasks in random order: Approach-Avoidance Task (AAT), Stimulus-Response Compatibility Task (SRCT), Implicit-Association Tests (IAT, approach-avoid, valence), Dotprobe Task, Go-/NoGo Task (GNGT). Additionally, different smoking-related variables were assessed. Split-half reliabilities of the different cognitive (bias) measures and correlations between them were calculated. Results: Split-half reliabilities of the AAT, the SRCT, and the Dotprobe Task were unacceptable whereas both IATs and the GNGT showed good to excellent reliability. Smoking-approach associations were significantly related to nicotine dependence; however, none of the cognitive bias measures correlated with response inhibition or smoking-related variables. Limitations: Pictorial stimuli were the same across paradigms and might not have been relevant to all participants. Conclusions: This is the first study to investigate the association between different cognitive biases, response inhibition, and smoking-related variables. Although findings are at odds with theoretical assumptions, their interpretation is clearly restricted by the low reliability of the cognitive bias measures.
The overvaluation of energy-dense foods is a key contributor to unhealthy eating behaviour, identifying it as a key target for therapeutic interventions. A growing literature has shown that by consistently associating specific food items with the inhibition of a motor response (i.e. stopping), the evaluations of these stimuli can be reduced after training. In this brief review, we focus on measures used to capture food valuation following such training interventions. Evidence for the food devaluation effect has primarily stemmed from studies that employ explicit measures such as ratings of food attractiveness or taste, and implicit measures, such as the implicit association test, which have yielded mixed findings. Although promising, our understanding of the utility of implicit measures in studies of eating behaviour is relatively sparse, and we offer recommendations for the use of explicit and implicit measures in future research.
Full-text available
Changes in food preferences after bariatric surgery may alter its effectiveness as a treatment for obesity. We aimed to compare food reward for a comprehensive variety of food categories between patients who received a sleeve gastrectomy (SG) or a Roux-en-Y gastric bypass (RYGB) and to explore whether food reward differs according to weight loss. In this cross-sectional exploratory study, food reward was assessed using the Leeds Food Preference Questionnaire (LFPQ) in patients at 6, 12, or 24 months after SG or RYGB. We assessed the liking and wanting of 11 food categories. Comparisons were done regarding the type of surgery and total weight loss (TWL; based on tertile distribution). Fifty-six patients (30 SG and 26 RYGB) were included (women: 70%; age: 44.0 (11.1) y). Regarding the type of surgery, scores were not significantly different between SG and RYGB, except for ‘non-dairy products—without color’ explicit liking (p = 0.04). Regarding TWL outcomes, explicit liking, explicit wanting, and implicit wanting, scores were significantly higher for good responders than low responders for ‘No meat—High fat’ (post-hoc corrected p-value: 0.04, 0.03, and 0.04, respectively). Together, our results failed to identify major differences in liking and wanting between the types of surgery and tended to indicate that higher weight loss might be related to a higher reward for high protein-content food. Rather focus only on palatable foods, future studies should also consider a broader range of food items, including protein reward.
Full-text available
Changes in food preferences after bariatric surgery may alter its effectiveness as a treatment for obesity. We aimed to compare food reward for a comprehensive variety of food categories between patients who received a sleeve gastrectomy (SG) or a Roux-en-Y gastric bypass (RYGB) and to explore whether food reward differs according to weight loss. In this cross-sectional exploratory study, food reward was assessed using the Leeds Food Preference Questionnaire (LFPQ). We assessed liking and wanting of eleven food categories. Comparisons were done regarding type of surgery and Total Weight Loss (TWL; based on tercile distribution). Fifty-six patients (30 SG and 26 RYGB) were included (women: 70%; age: 44.0 (11.1) y). Regarding the type of surgery, scores were not significantly different between SG and RYGB, except for 'non-dairy products-without color' explicit liking (p = 0.04). Regarding TWL outcomes, explicit liking, explicit wanting and implicit wanting, scores were significantly higher for Good responders than Low responders for 'No meat-High fat' (post-hoc corrected p-value: 0.04, 0.03 and 0.04, respectively). Together, our results failed to identify major differences in liking and wanting regarding the type of surgery and tended to indicate that higher weight loss might be related to a higher reward for high protein-content food. Rather to focus only on palatable foods, future studies should also consider a broader range of food items, including protein reward.
Aims: To assess whether cognitive biases for drug-related cues are associated with subjective craving and behavioural indices of drug-seeking behaviour, as predicted by incentive models of addiction. Methods: Fifty social drinkers took part in a laboratory study in which their subjective craving and cognitive biases for alcohol cues were assessed, before they completed a progressive ratio operant task for alcohol (beer) reinforcement. Results: Social drinkers with high levels of alcohol craving at the beginning of the experiment had more pronounced attentional, approach, and evaluative biases for alcohol cues, compared with those with low craving. There were also trends for the high craving group to show greater operant responding for beer reinforcement, but the latter findings were inconclusive, and no evidence was found of associations between the operant responding and cognitive bias measures. Conclusions: The finding of a relationship between subjective craving and cognitive biases for alcohol cues is consistent with incentive models of addiction. Methodological factors may have obscured the predicted relationships between cognitive bias and operant performance, such as the use of a specific reinforcer (beer) during the operant task, while a range of alcohol-related cues were used in the cognitive bias tasks.
This chapter discusses how different forms of outcome utility are embedded in brain systems. Experienced utility, the actual pleasure of an outcome when received, is encoded by a neural activations in a network that includes limbic prefrontal cortex as well as deep brain structures below the cortex, but it is possible that causal generation of experienced utility, in the form of intense pleasures, may be more restricted to small hotspots within the deeper structures. Decision utility, manifested in choices to pursue or consume an outcome, is influenced by additional factors, including memories of past experienced utility resulting from outcomes (remembered utility), and predictions or beliefs about how good experienced utilities are likely to be in future (anticipated/predicted utility). Brain circuitry for decision utility can be separated to a degree from circuitry for experienced utility, and the brain mesolimbic dopamine system is one component that is especially important to decision utility as a mechanism for making choices. However, there is some controversy in the field today concerning the precise role of dopamine in decisions. One common view has been that mesolimbic dopamine influences choice by mediating learning and predicted utility (as teaching signal and prediction error), acting as an input to decision utility. An alternative view is that mesolimbic dopamine instead more purely mediates decision utility directly (as incentive salience or 'wanting'), being able to depart from learned or remembered utilities, and not necessary for reward learning or predictions. Berridge and O'Doherty favor different sides in this dopamine controversy, and so the authors here distill those different views into a brief debate. Finally, the relations among brain systems for various utilities opens up interaction possibilities that sometimes lead to irrational choices. These include 'wanting' a particular outcome whether or not that outcome turns out to be actually 'liked'. That phenomenon is particularly vivid in addiction but also may occur to some degree in ordinary life.
The question of addiction concerns the process by which drug-taking behavior, in certain individuals, evolves into compulsive patterns of drug-seeking and drug-taking behavior that take place at the expense of most other activities, and the inability to cease drug-taking, that is, the problem of relapse. In this paper we summarize one view of this process, the “incentive-sensitization” view, which we e rst proposed in 1993. Four major tenets of the incentive-sensitization view are discussed. These are: (1) potentially addictive drugs share the ability to alter brain organization; (2) the brain systems that are altered include those normally involved in the process of incentive motivation and reward; (3) the critical neuroadaptations for addiction render these brain reward systems hypersensitive (“sensitized”) to drugs and drug-associated stimuli; and (4) the brain systems that are sensitized do not mediate the pleasurable or euphoric effects of drugs (drug “liking”), but instead they mediate a subcomponent of reward we have termed incentive salience (drug “wanting”).
What roles do mesolimbic and neostriatal dopamine systems play in reward? Do they mediate the hedonic impact of rewarding stimuli? Do they mediate hedonic reward learning and associative prediction? Our review of the literature, together with results of a new study of residual reward capacity after dopamine depletion, indicates the answer to both questions is 'no'. Rather, dopamine systems may mediate the incentive salience of rewards, modulating their motivational value in a manner separable from hedonia and reward learning. In a study of the consequences of dopamine loss, rats were depleted of dopamine in the nucleus accumbens and neostriatum by up to 99% using 6-hydroxydopamine. In a series of experiments, we applied the 'taste reactivity' measure of affective reactions (gapes, etc.) to assess the capacity of dopamine-depleted rats for: 1) normal affect (hedonic and aversive reactions), 2) modulation of hedonic affect by associative learning (taste aversion conditioning), and 3) hedonic enhancement of affect by non-dopaminergic pharmacological manipulation of palatability (benzodiazepine administration). We found normal hedonic reaction patterns to sucrose vs. quinine, normal learning of new hedonic stimulus values (a change in palatability based on predictive relations), and normal pharmacological hedonic enhancement of palatability. We discuss these results in the context of hypotheses and data concerning the role of dopamine in reward. We review neurochemical, electrophysiological, and other behavioral evidence. We conclude that dopamine systems are not needed either to mediate the hedonic pleasure of reinforcers or to mediate predictive associations involved in hedonic reward learning. We conclude instead that dopamine may be more important to incentive salience attributions to the neural representations of reward-related stimuli. Incentive salience, we suggest, is a distinct component of motivation and reward. In other words, dopamine systems are necessary for 'wanting' incentives, but not for 'liking' them or for learning new 'likes' and 'dislikes'.
This paper presents a biopsychological theory of drug addiction, the 'Incentive-Sensitization Theory'. The theory addresses three fundamental questions. The first is: why do addicts crave drugs? That is, what is the psychological and neurobiological basis of drug craving? The second is: why does drug craving persist even after long periods of abstinence? The third is whether 'wanting' drugs (drug craving) is attributable to 'liking' drugs (to the subjective pleasurable effects of drugs)? The theory posits the following. (1) Addictive drugs share the ability to enhance mesotelencephalic dopamine neurotransmission. (2) One psychological function of this neural system is to attribute 'incentive salience' to the perception and mental representation of events associated with activation of the system. Incentive salience is a psychological process that transforms the perception of stimuli, imbuing them with salience, making them attractive, 'wanted', incentive stimuli. (3) In some individuals the repeated use of addictive drugs produces incremental neuroadaptations in this neural system, rendering it increasingly and perhaps permanently, hypersensitive ('sensitized') to drugs and drug-associated stimuli. The sensitization of dopamine systems is gated by associative learning, which causes excessive incentive salience to be attributed to the act of drug taking and to stimuli associated with drug taking. It is specifically the sensitization of incentive salience, therefore, that transforms ordinary 'wanting' into excessive drug craving. (4) It is further proposed that sensitization of the neural systems responsible for incentive salience ('for wanting') can occur independently of changes in neural systems that mediate the subjective pleasurable effects of drugs (drug 'liking') and of neural systems that mediate withdrawal. Thus, sensitization of incentive salience can produce addictive behavior (compulsive drug seeking and drug taking) even if the expectation of drug pleasure or the aversive properties of withdrawal are diminished and even in the face of strong disincentives, including the loss of reputation, job, home and family. We review evidence for this view of addiction and discuss its implications for understanding the psychology and neurobiology of addiction.