Retraining Automatic Action Tendencies in Addiction: Underlying Mechanisms and Potential Moderator Variables

Retraining Automatic Action Tendencies in Addiction:

Underlying Mechanisms and Potential Moderator Variables

Wenyu Nie

Literature Thesis Research Master Brain and Cognitive Sciences

Amsterdam Brain and Cognition Center, University of Amsterdam

Supervisor: Dr. R.W. Wiers

Second assessor: Dr. R.H. Phaf

July, 2017 – September, 2017

Abstract

Recent research on addiction has suggested that cognitive biases embedded in the automatic psychological processes are important contributors to addictive behavior. Approach bias is one of these cognitive biases which is shown to be relevant. Approach bias modification is a novel intervention in which approach biases for addiction-related cues are modified. The present review aims to analyze the

underlying mechanisms of the formation and change of approach bias in the context of substance addiction, from the perspective of embodiment, conditioning, and then neural basis. In addition, it examines the moderator variables which are potentially relevant to the effectiveness of the approach bias modification. In conclusion, the heterogeneity of the components involved in the approach bias and the moderating role of the control resources were highlighted. Suggestions are made about the ideal settings for lab experiments and clinical trials. Further studies can extend the

investigation on how different components in the approach bias are susceptible to modification, and how the allocation of control resources can influence the

Introduction

Addiction is a neural and psychological dysfunction characterized by compulsive engagement in rewarding stimuli, despite adverse consequences. Different types of addiction can be formed by different substances including drugs, alcohol, tobacco or even unhealthy food. Addiction to some of these substances can lead to acute diseases and even death, and dependence on the others can contribute to the development of a number of chronic health conditions, including cancer, cardiovascular diseases, and obesity. Besides the detrimental effects to health, these addictive conditions can also bring out severe damage to the finance and social function of the addicted (WHO, 2017). It is very urgent to develop effective interventions to modify the addictive behavior.

Traditional interventions for addictive behavior are largely based on providing information about the detrimental effects of addiction and encouraging individuals to reflect and monitor their behavioral tendencies (Kakoschke, Kemps, & Tiggemann, 2017). Many studies have found a negative correlation between executive function and the development of addiction, which indicates the involvement of intention and control in disinhibiting addictive behavior (Lancaster & Stead, 2005). The traditional approach to fighting addiction takes effect by making people aware of the unhealthy consequences, so that they can form a conscious intention to stop the addictive behavior. However, this approach fails to produce a lasting change in the addictive behavior (Marteau, Hollands, & Fletcher, 2012).

One explanation of this failure refers to the dual-processes of the human mind (Kahneman & Frederick, 2002). The dual-process theory attempts to distinguish two different types of thinking processes in human psychology. One type is the

implicit/unconscious, automatic/impulsive processes and the other is the

explicit/conscious, controlled/reflective processes. These two different streams of thinking operate simultaneously, interacting and competing with each other, and ultimately affect observable behavior (Strack & Deutsch, 2004).

This theory also explains the development of addictive behavior. When the reflective processes persuade the individual to quit the addictive behavior and to form a conscious intention to stop using the substance, the impulsive processes still work to draw the individual towards rewards. When the impulsive processes overcomes the reflective processes, the addictive behavior continues (Stacy & Wiers, 2010). A negative correlation was usually found between executive function and addictive behavior. This reveals how vulnerable the reflective processes of an addicted individual are. The impulsive processes can be more dominant in forming the addictive behavior (Deutsch, Gawronski, & Hofmann, 2016, Chapter 11).

The investigation to find specific mechanisms behind the impulsive processes sheds light on how these processes work to influence a person to seek the rewards of addictive behavior. The formation of an addictive tendency in the implicit processes can be extrapolated from evidence at the neural level, according to the

incentive-sensitization theory. Robinson and Berridge (2003) argue that the activation of the mesolimbic dopamine system and the increase in the level of dopamine

increase the salience of the addiction-related stimuli. At the behavioral level, the person associates the addictive substance with the rewarding effect. At the cognitive level, the phenomenon is summarized by a variety of cognitive biases, such as attentional bias, approach bias, and implicit associations (Larsen, Kong, Becker, Cousijn, Boendermaker, Cavallo, & Wiers, 2014). With the development of the IAT (Greenwald, McGhee, & Schwartz, 1998), the implicit positive memory associations with addictive substance are widely explored. Later research has also looked into how attention is automatically allocated to the addiction-related cues. These findings converge to conclude that, the implicit biases can predict the addiction-related tendency and behavior (Larsen et al, 2014).

Approach Bias and Approach/Avoidance Training

The approach bias is one of these biases which is found to be relevant. Individual attitudes towards a group of people or a category of objects can be implicitly shown through their action tendencies of approaching or avoiding the attitude target (Solarz, 1960). The approach bias in addiction refers to the automatic approach tendency towards addiction-related cues in the addicted individuals (Palfai & Ostafin, 2003). Palfai and Ostafin (2003) modified the IAT to measure approach-avoidance

tendencies and found that people with higher alcohol-approach tendencies

experienced stronger subjective craving, more intense affective responses towards alcohol and more frequent binge drinking episodes. The approach bias was later also measured by Stimulus Response Compatibility (SRC) Task and Approach Avoidance Task (AAT). In the classic SRC task, the participants are explicitly instructed to

categorize the relevant features of the stimuli (i.e. addiction-related or neutral) usually with the symbolic approach and avoidance movements of a manikin. In different blocks, the two action tendencies are associated with addiction-related cues or control cues. The approach bias is quantified by contrasting the reaction time between

compatible blocks (approach associated with addiction-related cues) and incompatible blocks (avoidance associated with addiction-related cues). In the AAT, by contrast, the responses to addiction-related cues are measured in an implicit manner by instructing participants to respond to irrelevant features, such as the orientation (e.g. Cousijn, Goudriaan, & Wiers, 2011) or the format (portrait v.s. landscape, e.g. assessments in Wiers, Eberl, Rinck, Becker, & Lindenmeyer, 2011) of the cue pictures. In this manner, the trials do not need to be organized block-wise and the approach bias is measured by contrasting the compatible trials (approach when the cue is

addiction-related) and incompatible trials (avoid when the cue is addiction-related) all the way through the task. Usually in the AAT, the approach and avoidance movements are implemented with a joystick triggering visual feedback. Pushing the joystick makes cues zoom out, whereas pulling the joystick makes cues zoom in. The approach bias, like the implicit associations and the attentional bias, is found to predict craving, existence of addiction and relapse after treatment in a wide range of addiction substances, including alcohol (e.g. Wiers, Rinck, Dictus, & Van den Wildenberg, 2009; Wiers, Rinck, Kordts, Houben, & Strack, 2010), tobacco (e.g. Watson, de Wit, Cousijn, Hommel, & Wiers, 2013), and unhealthy food (e.g.

Jansen, 2010).

More importantly, researchers started investigating whether the approach bias can be influenced and whether it can play a causal role to intervene the addictive behavior. For this purpose, the approach avoidance task was adapted into the approach

avoidance training (AAT), a form of cognitive bias modification (CBM). In the training task, participants respond to an addiction-irrelevant feature of stimuli (e.g., the shape of the picture frame). The association with addiction becomes a

task-irrelevant feature (this is true for most cases, for special cases, see the moderator variable instruction). By repeating this training, the avoidance tendency towards the addictive substances should be implicitly trained. This was found in most intervention studies of AAT on alcohol abuse (Wiers, et al, 2010, 2011; Eberl et al, 2013, 2014) and eating behavior (Schumacher, Kemps, & Tiggemann, 2016; Brockmeyer, Hahn, Reetz, Schmidt, & Friederich, 2015). However, null results were reported in the other studies (Lindgren et al., 2015). The contradictory results from different studies indicate the effectiveness of the AAT is likely to be moderated by task settings, participant related features or choice of outcome measurement. This intends to

delineate the underlying mechanisms for the AAT and sort out the potential moderator variables to better understand the modification of approach bias in addiction.

Underlying Mechanisms

processes in supporting and forming the abstract, complicated psychological processes. In the case of approach and avoidance, the psychological processes

involving positive emotions or attitudes are embodied through the approach behavior or tendency, whereas those involving negative emotions or attitudes are embodied through avoidance behavior. At the behavioral level, positive stimuli facilitate the approach action tendencies, whereas negative stimuli facilitate the avoidance action tendencies (Phaf, Mohr, Rotteveel, & Wicherts, 2014). In the seminal study by Solarz (1960), participants were explicitly instructed to categorize words as pleasant or unpleasant by moving word cards towards or away from themselves on a movable platform. The pleasantness of the cards and the instruction of moving towards or away made up four possible contingencies. By contrasting the contingencies, it is shown that participants were faster to pull positive words towards themselves and push negative words away from themselves.

There is an evolutionary basis for the association of approach and avoidance tendencies with appetitive cues and aversive cues. In the primitive era of human history, it was important to use the four limbs to explore the uncertain world, grabbing the objects beneficial for individual survival and pushing away the objects threatening life (Firdland & Wiers, 2017). Evolutionary computer simulations have also shown the automatic development of the approach tendency to appetitive stimuli and the avoidance tendency to aversive stimuli after generations of evolution (Heerebout & Phaf, 2010).

Some researchers believed that affective processing is embodied directly through action tendencies for arm flexion and extension (Chen & Bargh, 1999). However, this view was challenged by later studies (Rotteveel et al., 2015, non-replication). A study by Rotteveel and Phaf (2004) managed to dissociate the actual arm movements from the change of distance by replacing the usually used joystick task with a vertical stand with three buttons. Pressing the upper button or the lower button in the vertical stand corresponds to arm flexion or extension, but there would be no actual or symbol change of distance in this setting. However, they found the compatibility effect only when the stimuli were evaluated consciously. When they reduced the conscious evaluation of the affective feature, no action tendencies were observed. It should be noted that, most training studies of addiction have used implicit instructions,

corresponding to the reduced consciousness condition in the Rotteveel and Phaf study. This indicates the modified approach and avoidance tendencies in the AAT should be embodied through mechanisms other than actual physical muscle movements.

Symbolic change of distance

The approach and avoidance tendencies might be embodied through the

representation of self and the distance of the target to self. In a study putting the names of participants on a reference block on the screen, it has been found that the compatibility effect is contingent on the approach and avoidance to the name tag, regardless of the actual physical actions of pulling or pushing (Markman & Brendl, 2005). A later study also eliminates the explanation of self-categorization for this phenomenon (Van Dantzig, Zeelenberg, & Pecher, 2009). This indicates the approach

and avoidance tendencies are associated with the symbolic situational distance between self and target stimuli. Indeed, the manikin task, as an alternative task to the joystick-based approach avoidance task, in which participants would control a fictional manikin man to approach or avoid addiction-related cues, measures the approach bias just as well (Krieglmeyer & Deutsch, 2010). Furthermore, in an assessment study of addictive behavior with implicit instructions, the approach and avoidance were implemented purely through the zooming feedback without actual physical movements, and the researchers were still able to find the association

between addictive behavior and the approach bias measured in this manner (Peeters et al., 2012).

It is safe to conclude that symbolic change of distance is involved in the

assessment of the approach bias, whereas the involvement of physical arm movements is in doubt. In the training studies, the usually used joystick-based versions of the AAT with zooming feedback embraces both two components. However, the usually used implicit instructions in these studies indicate the change of approach bias are probably embodied through the symbolic distance change, rather than the actual arm movements (cf. Krieglmeyer, Houwer, & Deutsch, 2013).

Behavioral Psychology: Conditioning Approach Bias in Addiction

Approach tendency to addictive substances

As is discussed in the section on embodiment, the approach tendency towards appetitive cues has an evolutionary basis. For example, in the case of high-calorie food, due to its role as necessities for survival, there might be a strong, inherited,

unconditional approach tendency (Drewnowski & Almiron-Roig, 2009). This inherited tendency is best illustrated by an interesting experiment of ‘looking glass world’ (Hershberger, 1986). In this experiment, the researchers attempted to overwrite the unconditional approach tendency of the chicks towards food stimuli with the instrumental response of leaving the food stimuli. The movable food container in the experiment was set to move in the same direction but twice as fast as the moving chicks. It could be expected, if the approach tendency towards the food was conditioned during the development of the chicks rather than inherited, the conditioned approach behavior should have soon faded away because of the

extinction of rewarding outcomes. However, the result of the experiment has shown that most of the chicks could not switch the approach tendency, indicating the unconditional nature of the approach tendency in food consumption.

For some other addictive substances, the approach bias involves instrumental conditioning of goal-directed behavior. Actually, many addictive substances, like alcohol and drug, can be first associated with negative feelings before addiction. Many would associate alcohol, smoking or drug use with danger and threat to health and life safety (e.g. in smoking, Glock, Kovacs, & Unz, 2014). The approach

tendency towards these addictive substances does not need to be inherited. In the initial period, the addicted individuals are pleased by the positive effects and herefore, form a positive expectation towards the addictive substances (Robinson & Berridge, 2003). The crucial element for the instrumental conditioning is exactly this kind of contingency between an action and a motivationally relevant outcome (Skinner, 1938).

For example, approaching the alcohol facilitates the consumption of alcohol, and the consumption of alcohol makes people feel more sociable and less inhibited (Watson, de Wit, Hommel, & Wiers, 2012). These positive feelings drive potential addicted individuals to approach alcohol. Approaching these substances is the goal-directed behavior conditioned by the rewarding outcome.

Approach bias measured with addiction-related cues

The approach bias towards the addictive substances converts into the approach bias towards the addiction-related cues presented in the approach avoidance tasks. The addiction-related cues serve as the stimulus (S) in the scenario and the approach tendency elicited is the response (R). The S-R association depends on the predictive relationship between the addiction-related cues and the emergence of the addictive substances, which shapes the Pavlovian conditioning between addiction-related cues and approach responses (Pavlov, 1927). Let us take the case of food as an example. In the schema of a Pavlovian conditioning, the food itself serves as the unconditional stimulus which triggers a response of approach behavior towards the food stimulus, whereas the environment cues, for example, the atmosphere of people holding and enjoying the food, or the presence of the container of the food, are the conditional stimulus in this case. The co-occurrence of the food and these environmental cues makes these cues predictive of the presence of the food and therefore, they elicit the conditioned approach tendency as well.

The instrumentally conditioned goal-directed approach behavior can also contribute to the approach bias towards addiction-related cues. The Pavlovian to

instrumental transfer (PIT) is the bridge between the R-O (reward-outcome)

relationship in the instrumental learning and the S-R relationship in the measurement of the approach tendency. In a PIT scenario, the R-O relationship is extended with a preceding conditional stimulus serving as the context. For example, in the case of training rats to press levers in the operant box, if the food pellet reward is only associated with the lever-pressing behavior when the green light is on, the rat will learn to discriminate different contexts and will press the lever only in the

green-light-on context (Hogarth, 2012). The rewarding outcome in a case of PIT, therefore, will specifically reinforce the S-R binding. In the cases of addiction, there is also a context for this instrumental learning, for example, the environment of other people holding the glass and enjoying their beer or the image of a wine bottle with a nice wine brand name on the tag. These addiction-related cues are constantly

reinforced with the approach behavior by the reward of being disinhibited and sociable. Consequently, the goal-directed approach gradually becomes a habit that is stimulus-bound to the addiction-related environmental cues. Moreover, in the process, the actual rewarding value of the outcome becomes less relevant. In animal studies, it is found that after the S-R binding is solidly formed, outcome devaluation procedure (e.g. through satiation of the relevant desire beforehand) or extinction procedure (e.g. taking away the rewarding outcome directly) would not reduce the instrumental response, indicating the S-R habits become independent of the rewarding outcome after time.

exclusively determined by the automatic conditioning mechanisms. The approach bias score does not only represent the facilitation of the approach tendency towards

addiction-related cues, but also the control over conflicting avoiding trials. The latter requires executive function processes beyond the scope of conditioning. In a study dissociating the two distinctive sources of contribution to the approach bias, the researchers separated the ‘pull alcohol’ trials and ‘push alcohol’ trials (Sharbanee, Stritzke, Wiers, Young, Rinck, & MacLeod, 2013). It was found that in the ‘push alcohol’ trials, where the instructed response was incongruent with the potential approach bias, the reaction time was correlated with the working memory capacity, a common index for executive function. In the congruent ‘pull alcohol’, however, there was no such phenomenon. This clearly illustrates the two distinct components in the measurement of the approach bias to the addiction-related cues.

Modification of the approach bias

Modification of the approach bias with the AAT training will therefore involve the breaking down and modification of the aforementioned conditioned associations. In the training program, the participants approach the neutral cues/avoid the

addiction-related cues and receive feedback about their performance. In the PIT scenario, the feedback can be seen as a mild version of reward and punishment. Associating addiction-related cues with the approach behavior will bring out the incorrect answer, triggering the negative feedback. This should gradually diminish the approach behavior in the context of addiction-related cues. In addition, a new

established through the reinforcement of the positive feedback.

Neural Mechanism of Training Automatic Tendencies in Addiction

Dual-process model at the neural level

In line with the dual-process model at the cognitive level, similar delineation can be made at the neural level (Wiers et al., 2017). The bottom-up impulsive processes mostly involve the fronto-limbic areas including striatum, amygdala, insula, anterior cingulate cortex, and other prefrontal areas. These areas together play an important role in the formation of impulsive action tendencies. Specifically, two separate mechanisms were proposed to contribute to the formation of automatic behavior and they correspond to the functioning of the different brain areas. The first mechanism involves amygdala, nucleus accumbens, and ventral medial prefrontal cortex, associating the relevant stimuli with subjective value and motivational salience. The second mechanism concerns the habit formation. As discussed in the behavioral

psychology section, some conditioned S-R associations, after repetitive reinforcement, turn out to be habitual behavior, less susceptible to the extinction procedure that

changes the motivational salience of the stimuli. This behavioral pattern corresponds to the shift from activation of ventral striatum to the activation of the dorsal striatum

(Everitt & Robbins, 2013).

The top-down reflective processes, on the other hand, involve imagination, planning and intentional calculation of the future consequences. Conflict monitoring and control is one component of the reflective processes. The human brain has the function of monitoring and solving the conflict between the action tendency produced

in the reflective processes and that produced in the automatic processes. These high-level control functions involve the posterior area of the ventromedial prefrontal cortex and the anterior cingulate cortex and some basal ganglia regions (Alvarez & Emory, 2006).

Reflective and impulsive processes in addiction

The bottom-up and top-down processes are crucial for the formation of the addictive behavior. The bottom-up impulse towards the addictive substances can be related to sensitized incentives, habitual behavior, and negative reinforcement. According to the incentive sensitization theory (Robinson & Berridge, 2003), the addiction to a specific addictive substance relies on the motivational salience of the addictive substance registered in the neural system. The association between motivational salience and the addiction-related cues occurs in the amygdala, which is known for its role in

evaluative Pavlovian conditioning and emotional memories (Wiers et al., 2014). Moreover, the medial prefrontal cortex is involved in tagging subjective values to the addictive stimuli. The motivational information converges in the ventral striatum, connecting the rewarding value of the addictive substances with the impulsive

behavior of approaching them. Negative reinforcement involves the same process, but in the opposite direction for the avoidance of addictive substance. The avoidance of addictive substances brings about strong withdrawal symptoms. At the neural level, the negative affective state was correlated with strong activation of the insula. The information of the negative emotion streams to the ventral striatum and the impulse against avoidance behavior is formed. With the repetitive reinforcement, the approach

bias further develops to be S-R habits regulated by the dorsal striatum.

The impulsive behavioral tendency for the addictive substances is also regulated by the upstream prefrontal cortices involved in the reflective processes. When the impulsive behavior towards the addictive substances is in conflict with the long term health goals formed in the reflective processes, the ventral prefrontal cortex will take charge and exert inhibitive control. The mechanism which makes addiction very stubborn against eradication is the vicious cycles in the positive loop of impulse and the negative loop of executive function. The impulse is constantly amplified, firstly because of the incentive sensitization mechanism. It is found that the motivational salience of the addictive stimuli is not be desensitized but rather, becomes more sensitive over time (Koob & Volkow, 2016). The repetition of the impulsive

satisfaction also reinforces the S-R route of habituation, producing ‘wanting’ without ‘liking’ (Watson et al., 2012). In addition, the constant abuse of some addictive substances is found to be detrimental to the development of prefrontal cortex and extremely harmful to the executive function (Fernandez-Serrano, Perez-Garcia, & Verdejo-Garcia, 2011). The lack of the executive function capacity in return makes the individuals more subject to addictive behavior, which further undermines their

capability of willpower and intentional action.

Neural signatures in measurement and modification of approach bias

The approach bias measured in the context of the approach avoidance task involves the neural mechanisms of addiction discussed above. The fronto-limbic system plays the crucial role to initiate the automatic approach tendency towards the

addiction-related cues. Neuroimaging studies of the approach and avoidance task have found that the BOLD responses in the nucleus accumbens and medial prefrontal cortex were stronger for the alcohol-dependent patients when they were presented with alcohol-related cues, and the amygdala activity was positively associated with the subjective craving for alcohol-dependent patients, when they were approaching addiction-related cues (Wiers et al., 2014). This finding converges with the

functionality of these brain areas, confirming the three-way connection of the fronto-limbic areas, increased motivational salience of the stimuli, and the observed approach bias in the behavioral tasks.

The brain areas involved in executive function also take part in the measurement of the approach bias, by regulating the interference of avoiding the addiction-related cues. However, there is less evidence for the involvement of the frontal areas related to executive function in the neuroimaging studies of the approach bias. For example, there is no evidence showing that the activation of dlPFC, the brain area involved in the conflict control, was different between alcohol-dependent patients and control group (Wiers et al., 2014). The non-significance can be explained by the

heterogeneity of the control group. People not addicted to alcohol fall into two groups. Some people never develop the approach tendency towards alcohol, and therefore, the avoidance-alcohol contingency will not trigger response to incongruency in the dlPFC area at all. On the other hand, there are some people who stay clean from the addictive substances by exerting strong intentional control. For the latter group of people, the avoidance-alcohol contingency does indeed trigger response to incongruency, and the

responses are even stronger than those of the addicted individuals, as a result of their stronger executive function capacity (Barbey, Koenigs, & Grafman, 2013). Potentially, the neural activity of these two groups in dlPFC goes opposite ways, and might have cancelled out with each other.

Studies on the neural effects of the training have so far been limited to inpatients addicted to alcohol and to the training with the implicit joystick-based AAT. Wiers and colleagues (Wiers, Stelzel, et al., 2015) used the cue-reactivity task as the measurement method and found that the training reduced the activation of the

amygdala, which mediated the reduction of alcohol craving. Reduction of cue-induced reactivity in the amygdala may therefore be an important mechanism of the

therapeutic effectiveness of the AAT. Another study investigated the neural signatures of automatic approach tendencies after the modification and found the reduction of activation in mPFC after CBM training compared to placebo. It is also found that the reduction of mPFC activation was correlated with a reduction in the approach bias scores only in the training group (Wiers, Ludwig, et al., 2015). The results in the two studies suggest that training influences the functioning of mPFC and amygdala, brain areas involved in the formation of the automatic alcohol approach bias, which shed light on the neural basis underlying the clinical effectiveness of the AAT training.

Moderating Variables

Previous research has shed some light on the effectiveness of the approach bias training program. However, the training effect was not always replicated in different

experimental settings, when the measured outcome indices range from the approach bias itself to clinical indices (e.g. Becker, Jostmann, Wiers, & Holland, 2015; Lindgren et al., 2015). A meta-analysis even claimed to find no evidence for the effectiveness of the AAT on translating the change of bias into effects on clinical outcomes (Cristea, Kok, & Cuijpers, 2016). The contradictory findings from different studies might be a result of potential moderating variables, ranging from the source of participants to the selection of outcome indices. A comprehensive examination of the potential variables will help to define the optimal settings and ideal targets for the training program.

Task and Treatment Related Moderators

Joystick task v.s. manikin task

The joystick task is commonly used in the modification studies. In the original joystick task, the approach and avoidance behavior is operationalized as the physical pulling and pushing of a vertically positioned joystick forward or backward

horizontally. In the manikin task, however, the participants are instructed to move a manikin towards or away from the task-relevant stimuli on the computer screen. In this scenario, the manikin is taken as the reference point, and the approach and avoidance behavior is operationalized by the distance change of the manikin to the relevant stimuli.

The potential advantage of the manikin task should not be ignored. As is

discussed in the embodiment section about the underlying mechanisms, the symbolic change of distance is more likely to contribute to the approach and avoidance

tendencies. Krieglmeyer and Deutsch (2010) tried to establish the criteria for comparing the joystick task and the manikin task. They argued that the ease of

recategorization and the means of distance change should be taken as the validity and sensitivity evaluations of the task quality. There were previous studies indicating that the meaning of the physical actions in the joystick task is easy to change. Participants could easily recategorize the extension of the arm as ‘reach towards’ rather than ‘push away’. In addition, the manikin task might represent better means of distance change, since the automatic response to danger is withdrawal rather than pushing away (Lang, Bradley, & Cuthbert, 1990; Markman & Brendl, 2005; Seibt, Neumann, Nussinson, & Strack, 2008). In the study contrasting the two tasks, they indeed find the manikin task to have better validity and sensitivity for measuring the approach and avoidance tendencies in different contexts.

The feedback joystick task is an adaption of the joystick task. In the adapted version, the participants will see the stimuli zooming in after pulling the joystick and vice versa after pushing the joystick. This new setting is supposed to prevent the misinterpretation of the arm movements. The feedback joystick task was indeed more sensitive than the original joystick task, but still, it is shown to be less sensitive than the manikin task, especially in its predictive power to the clinical measurements in the context of anxiety disorder (Krieglmeyer & Deutsch, 2010). The juxtaposition of the three tasks indicated the superiority of manikin task over joystick tasks. It should be noted though, the two tasks were compared in the setting of an explicit instruction in the comparison study. However, considering that the embodiment of actual arm

movements is likely to be inhibited in the setting of implicit instructions (Rotteveel & Phaf, 2004), it would only put the joystick feedback task in more disadvantage. In line with this, it is worthwhile to examine the sensitivity and validity of the manikin task in the training setting. In addition, with this rationale, the physical arm movements should not be necessary, at least in the setting of implicit instructions. Indeed, some studies managed to measure approach bias with only the zooming visual effects and without the actual arm movements (Peeters et al., 2012). Researchers are also trying to extend this finding to the training task in order to see whether the symbolic change of distance is sufficient for the effectiveness of training. This might open the door for using the keyboard as the alternative response instrument, enabling online and long distance participation in the training studies where the joystick equipment is

inaccessible.

Implicit instruction v.s. explicit instruction

Although in the assessment studies, both the SRC task with explicit instructions and the AAT task with implicit instructions were used, the latter was more usually adapted into the task in training studies. In a seminal training study, the researchers

distinguished the explicit condition and the implicit condition in the joystick AAT training. In the implicit condition, the participants were instructed to respond to the shape of the picture format cues (whether it is landscape or square), whereas in the explicit condition, the participants were directly instructed to respond to whether the picture stimuli are addiction-related or not (Wiers et al., 2011). The result of the study did not indicate any significant differences in training outcomes between the two

conditions. In a study of training attentional bias in anxiety disorder, however, it was reported that in the explicit condition there was better acceptability of the training whereas only in the implicit condition, the training effect can be generalized to the anxiety against subsequent stressors (Grafton, Mackintosh, Vujic, & MacLeod, 2014).

It seems that the implicit version of the AAT training should be preferred in consideration of the results of these two studies, and that all the studies afterward indeed followed the implicit version of the task. However, it should be noted that the statistic power of the seminal study was limited in contrasting the two different conditions. Moreover, the conclusion in the Grafton study should not be extended to the study of approach bias studies so easily. For example, the task-irrelevant word cues and the task-relevant dot probes were presented sequentially in the Grafton study. The participants might consciously or unconsciously establish the strategy to

discriminate the emotional valences of the word cues for the facilitation of their judgment on the probe cues. However, in the training studies of the approach bias, the task-irrelevant addiction-related cues and the task-relevant picture format cues were presented simultaneously and the participants might adopt a strategy to focus on the format cues without any registration of the addiction-related cues in mind. In fact, a meta-analysis on the affectively primed approach and avoidance tendencies found that the instruction type was a significant moderator, with the explicit instruction

producing a stronger effect than the implicit instruction (Phaf et al., 2014). A measurement study contrasting the two versions of instruction in the context of alcohol addiction also found that the explicit version was more sensitive and had

higher reliability in distinguishing heavy and light drinkers, indicating the explicit encoding of the addiction-related cues might be necessary for the elicitation of the approach tendency (Field, Caren, Fernie, & De Houwer, 2011).

Time scheduling of the treatment

According to the limited studies across different types of addiction, it is easier to find generalized effect of the training program in multiple-session studies, compared to single-session studies (e.g. Manning et al, 2016). The number of trials in a session seems to make less difference (Kakoschke et al., 2017). There seems to be a dose-effect relationship between the number of training sessions and training effectiveness. In a study focusing on the time schedules of the training program, the researchers (Eberl et al., 2014) found the mean number of sessions needed to reach the strongest training effect was six. However, it should be noted, there were strong individual differences in the time course of the bias modification. Most of the patients had yet to reach their plateau after six sessions, and the mean was dragged down towards six by a few individuals reaching their strongest effect within the first two sessions. There were also individuals not reaching the optimal bias change until the very late in the last session. One explanation of the individual differences in the time course is the awareness of the training contingencies in the task. As it was discussed in the previous session, the awareness of the task purpose might influence the effectiveness of the training. The awareness can be a deliberate manipulation of the task settings, or alternatively, by the insight of the individual participants. In addition, the optimal number of sessions for the strongest change of bias does not necessarily

equal to the optimal number of sessions for the strongest clinical outcome. There can be a bottleneck for the impact of approach bias on clinical outcomes, but there has been little research on this issue.

Participant Related Moderators

Executive function

As it is mentioned in the discussion of the dual-process model at cognitive and neural levels, the formation and change of approach bias is regulated by cognitive control. As it is proven in a number of studies using implicit measures to predict addiction, individual differences in the executive function capacity moderate the relative balance between implicit cognition and explicit cognition, with the implicit cognition having a stronger influence on behavior in individuals with relatively weak executive control. Specifically, for implicit approach tendencies, Peeters and colleagues (2013) found that they predicted escalation of alcohol use in a sample of impulsive high-risk youth with weak cognitive control, whereas, in a study with a sample of normal young participants (adolescents who reported drinking more than once), the approach bias was less predictive of alcohol use (Pieters et al., 2014).

Taking this insight into the training, the potential moderating role of executive function in the outcome measurements should be considered. With this rationale, in individuals with weak cognitive control capacity, the decrease of impulsive approach tendencies will be shown in behavioral and clinical outcomes, whereas in individuals with strong cognitive control capacity, it will not cause a change in behavioral and clinical outcomes, in that the highly functioning reflective processes will inhibit the

impulses anyway. If this hypothesis is true, there will be a dissociation of change in bias and change in behavior. Moderating effect of cognitive control would exist only on the change of addictive behavior. For individuals with high cognitive control capacity, the training would take effect on the approach bias, but not on the change of behavior.

However, Sharbanee and colleagues (2014) provided another theory about the moderating role of executive function. In this theory, the moderation of executive function occurs at the training stage. In the implicit version of the joystick task, the addiction-related cues in the pictures are task-irrelevant. The participants are supposed to inhibit the addiction-related cues and respond to the format feature of the pictures. The better inhibitory control capability over the task-irrelevant cues might forestall the implicit registration of the addiction-related cues, making the congruency effect impossible to be actualized. If this hypothesis is true, the moderating effect of executive function would occur on the change of bias. The non-replication in the studies with samples of normal people, therefore, might be a result of this lack of actual training on their automatic processes (e.g. in Lindgren et al., 2015).

Motivation to control

Although direct evidence from studies of the approach bias is limited, studies of other implicit cognitive biases have shown that the motivation to control modulates the influence of the implicit cognitive biases on behavior. In a study using urge and stimulant effects to the sight and smell of alcohol as impulse measures, the researchers (Tahaney, Kantner, & Palfai, 2014) distributed the participants to

high-restraint and low-restraint groups according to their motivation to abstinence. Interestingly, it was found that only in people with high restraints, the interaction between the executive function capacity capacity and the impulse measures was predictive of the alcohol consumption. This serves as a typical example that, whether impulses will dominate behavior depends on how much you allocate your control resources to fight against it.

In a recent study of working memory and alcohol use, the researchers (Van Deursen et al., 2015) found an interplay among appetitive memory associations, working memory capacity and motivation to change the alcohol consumption behavior. The motivation to change moderated the interaction between the working memory capacity and the implicit memory associations. Only in high motivation groups, there is an interaction between the working memory capacity and the implicit associations on alcohol consumption. This again reveals that to inhibit impulse, individuals should be motivated to allocate the resources for the current goal. In a real-world environment with addiction-related cues, trained individuals still need the control resources to prevent them from approaching the addictive substances. If the individuals choose not to allocate any control resources for inhibition, even mild approach tendencies are enough to cause approach behavior. This might be true in those cases of social drinkers or normal student samples, in which the participants do not wish to change at all (e.g. in Lindgren et al., 2015).

Nature of the approach bias

formed by various types of conditioning, ranging from inherited unconditioned responses to habitual S-R responses. These components of different natures might differ in their susceptibility to change. The experiment of ‘looking glass world’ (Hershberger, 1986) has shown how difficult it is to break down the inherited approach tendencies towards food and establish new conditioning contingencies against it. Evolutionary studies indicate a genetically hard-wired preference for the high-calorie food (Drewnowski & Almiron-Roig, 2009). In addition to motivation to change and time scheduling of the treatment, this might also help to explain why the effectiveness is difficult to reach in some training studies of eating disorders (Becker et al., 2015).

The ambivalence of the approach bias might also become an issue. In some of the training studies, a clear approach bias was not found in the patients against control before the training, but still, the training program took effect in reducing approach bias and improving clinical outcomes (Wiers et al., 2011, Eberl et al., 2013). In these cases, the patients might have internalized the long-term goal of avoiding addictive substances, and therefore, hold both approach and avoidance tendencies (McEvoy, Stritzke, French, Lang, & Ketterman, 2004; Wiers et al., 2006). The training program in this scenario could reduce the ambivalence by strengthening the avoidance

tendencies and weakening the approach tendencies. In contrast with the change of approach bias in the addiction-related cues, in most cases, there is no change of the approach tendencies to the control cues, even though the training on the control cues is as frequent as the training on the addiction-related cues (Eberl et al., 2013). The

lack of effect on the nonclinical samples of training might be for the same reason. For non-addicted, the addiction-related cues are just as control cues for the clinical

patients, entailing no ambivalence in action tendencies.

Outcome measurement

The theoretical background of the approach avoidance training relies on the construct of the approach bias, and therefore, it should be expected that any short term or long term change produced by the training in reality should be mediated by the change in the approach bias scores. Studies of other cognitive biases have shown that clinical effects almost always appear in parallel with the change in the relevant cognitive biases (e.g. Macleod & Clarke, 2015, the attentional bias modification). A series of studies have also shown that the clinical effects were not mediated by alternative mechanisms (e,g. not by the change in the attention bias, Sharbanee, Stritzke, Wiers, & MacLeod, 2013; Sharbanee et al., 2014). Effectiveness can therefore only be discussed after the guarantee of the change in approach bias itself. In some of the studies, the mediation of change in approach bias was not reached or reported (e.g. Wittekind, Feist, Schneider, Moritz, & Fritzsche, 2015). These studies should not be considered as evidence against the effectiveness of the training program. Rather, the methods and experiment designs of the studies should be carefully examined.

In addition, from the theoretical perspective, it is crucial to answer the causal relationship question. It should be noted that one theoretical mission of the training program is to verify the causal role of the approach bias in improving clinical outcomes of addiction. Cases failing to change the approach bias actually neither

support nor falsify the causational hypothesis.

A variety of outcome measurement has been applied in the training studies. Their potential differences are noteworthy. In general, it seems the short term effect with implicit measures was stronger than the long term effect on explicit measures. For example, training groups are more likely to choose healthier snacks (Fishbach & Shah, 2006), consume less chocolate (Schumacher et al., 2016) or alcohol (e.g. Wiers et al., 2011) right after the training. It should be noted though, these short term effects were found only in participants who successfully changed their approach bias, which indicates that the reaction of the participants to the training was heterogeneous. On the other hand, in the self-report results of craving or consumption, the effects were small or nonsignificant, especially in the self-report of long term relapses. The divergence between the implicit measures and explicit measures can be a result of different mechanisms involved in these processes. The long term effect of the training program remains to be determined.

Conclusion

The discussion of the underlying mechanisms of approach bias illustrated the potentially different components in the automatic approach tendencies towards the addictive substances. From the perspective of embodiment, the symbolic change of distance can independently contribute to the embodiment of approach (Krieglmeyer & Deutsch, 2010; Peeters et al., 2012). However, there is less evidence showing the

contribution of actual arm movements (Rotteveel & Phaf, 2004; Rotterveel et al., 2015).

From the perspective of conditioning, the approach bias can be a combination of inherited, classically conditioned, instrumentally conditioned, and habitual approach tendencies towards the addictive substances (Watson et al., 2012). These different components of the automatic approach tendencies might also have a different neural basis. The nucleus accumbens in the ventral striatum plays a role in the evaluation of the rewarding values of the stimuli and the subsequent formation of reward-directed behavior, whereas the dorsal striatum is involved in the habitual responses to stimuli regardless of its rewarding values (Balleine, Delgado, & Hikosaka, 2007). It is also insightful to distinguish the ambivalence of the automatic action tendency from the absence of the automatic action tendency. The two conditions represent different tensions between the approach and avoidance tendencies. This again adds more complexity to the dimensions of the automatic approach tendencies, potentially offering theoretical explanations for a lot of contradictory findings in different experimental settings.

Involvement of the Control Resources

The involvement of the control resources adds even more complexity. Although the study of the approach bias towards addictive substance originates from the

dual-process theory and focuses on the impulsive and automatic nature of the addictive behavior, it should be noted that it is not a process pure phenomenon. The investigation of the behavioral and neural mechanisms of the approach bias

assessment and modification task reveals a strong involvement of the cognitive control processes, and its influence on the formation and modification of approach bias is illustrated by the moderating role of executive function and motivation to control. The executive function capacity and the motivation to control can be considered as different aspects of the involvement of the control resources. The executive function capacity represents the maximum possible resources that can be allocated to the control of the automatic tendencies, whereas the motivation to control represents the actual amount of the control resources allocated to the regulation of the automatic tendencies.

Besides the dissociation of the capability and the motivation to control, the different timing of the involvement of the control processes should also be distinguished. At the stage of the approach bias measurement, there is a conflict control component in the trials of avoiding addictive substances and therefore, the final score of the approach bias does not purely represent the automatic tendencies generated from the automatic processes. There is a chance that the individuals who are measured to have low scores actually have strong approaching impulses, which are watered down by the strong conflict control. At the stage of outcome measure, the successful control of the substance consumption can be a result of the reduced

automatic approach tendencies, but it would not work alone without the allocation of control resources for inhibiting consumption, in which the motivation to control begins to take part and interact with the capability to control. More interestingly, how does the allocation of control resources influences the training stage? A possibility is,

it is precisely the failure to allocate control resources in the training that makes the automatic tendencies exposed and subject to change.

Settings of the Training Task

Previous research on the modification of the approach bias can be put into two categories. There are lab experiments on the causal mechanisms of approach bias and the clinical trials for the improvement of the addictive behavior. Most of the lab experiments had one session, targeting non-clinical participants, such as

undergraduate students. Short term effects of substance consumption or self-report craving were measured right after the completion of the sessions. Most of these studies found significant results in the change of approach bias and in the short term effect. But non-significant results were found in some cases (Becker et al., 2015). There may be many reasons for this non-significance. The addiction relevant in this case is eating disorder, involving the prevention of high-calorie food consumption. The inherited nature of the approach tendencies towards high-calorie food might render it more immune from modification, and therefore more sessions might be needed for a sizable change of the approach bias and behavior. Another reason might be the features of the participants in these studies. All the three studies of Becker and colleagues (2015), for example, recruited female students of normal weight, a

population whose motivation to control their food consumption behavior is in doubt. In consideration of the aforementioned moderators, there is a more ideal manner to implement the experiments of the approach bias modification. Firstly, participants should be motivated to quit the addiction. The participants should be patients

motivated to change their addictive behavior. At least, the executive function capacity and the motivation to control should be measured explicitly to eliminate

misinterpretation. Secondly, the addiction to smoking or alcohol will serve as a more ideal context for the lab experiments on causal mechanisms compared to the domain of eating. Thirdly, the test on consumption behavior should be included as the main measurement of the short-term effect, in that it has more ecological validity and it is more directly related to the change of automatic tendencies rather than the self-report of subjective craving.

In clinical trials, the training program is usually supplementary to the

psychosocial therapy (e.g. Eberl et al., 2013), and the participants were patients in treatment or individuals clearly identified as having the addictions. Apart from the self-report of craving, some of the studies also reported the relapse rate as the outcome measure. The settings of the clinical trials should follow the similar rules with the settings of the lab experiments. Moreover, it will be even more insightful if experience sampling methods can be applied and the change of daily craving and consumption can be tracked.

Future scenario

The discussion of the underlying mechanisms and potential moderator variables reveals some issues that still remain to be studied. First, the two types of embodiment should be examined more extensively with regards to their roles in the formation and change of the approach bias. Previous studies successfully dissociated the

found the latter to be sufficient for detecting the approach bias (Krieglmeyer & Deutsch, 2010; Peeters et al., 2012). However, the sufficiency of the latter to make a difference does not exclude the possibility of contribution from the former. In fact, in the Rotteveel and Phaf (2004) study, they still found the physical arm movements to embody approach and avoidance tendencies in explicit instruction conditions. The exploration of this question can be operationalized by contrasting the different tasks used for training with both explicit and implicit instructions. As it was mentioned in the discussion of the moderator variable task, the superiority of the joystick feedback task over the manikin task is still unconfirmed. If the physical arm movements indeed make no difference to the effectiveness of the training, the feedback task without joystick movements might also be more convenient and applicable than the joystick feedback task.

The second point remains to be studied is the applicability of the explicit and the implicit versions of the training task. The conclusions from attention bias studies (Grafton et al., 2014) should not be transferred to the context of approach bias, because of the different procedures and therefore potentially different mechanisms underlying the two tasks. Direct investigation of the variable instruction should be conducted in the context of the approach bias to reveal how the registration of stimuli and the embodiment of actual physical arm movements might change in the explicit version of the training task.

The time scheduling of the training sessions should also be further explored in the long term clinical studies. The only study which focused on the time course of the

training effect concentrated only on the curve of bias change (Eberl et al., 2014). It still remains to be discovered what the curve of outcome measures should be like. Is there a bottleneck for the influence of the approach bias change on the change of addictive behavior? Furthermore, there are also large individual differences in the time course of bias change. This suggests the moderation of the participant-related variables on the effectiveness of the training. For example, does the motivation to change affect the shape of individual training curves? Does the training work better in the long run for individuals with higher or lower executive function capacity? Future studies should embrace these potential moderations and explain the individual differences.

The allocation of control resources in the processes of training should also be further studied. Although in a big picture, the interaction of executive function, motivation to control and automatic action tendencies are clear, it is still unclear how the interaction works in the different timing of assessment, training, and behavior in reality. For example, in the studies which argued that the executive function capacity moderates the influence of automatic approach tendency on actual addictive behavior, it is unclear whether the moderation occurs in the process of assessment, diluting the difference between low and high approach bias individuals, or in the process of behavioral test, overshadowing the automatic impulses towards addictive substances. In addition, in the training, considering that the approach avoidance task is not process pure, it is unclear whether the training is mainly on the change in the

the addictive stimuli. Future studies should dissociate the two components in the assessment and the training studies, as was in the Sharbanee (2013) study, in order to demonstrate the specifics of the contribution of these two components. Moreover, even if the effect is indeed on the automatic tendencies, it is still interesting to explore whether the involvement of control resources facilitates the change of automatic tendencies, or actually hinders it (Sharbanee et al., 2014).

In conclusion, the heterogeneity of the components involved in the approach bias and the moderating role of the control resources were discussed in this review. Further studies can extend the investigation on how different components in the approach bias are susceptible to modification, and how the allocation of control resources can influence the effectiveness of the assessment and modification of approach bias in different timing.

References

Alvarez, J. A., & Emory, E. (2006). Executive function and the frontal lobes: a meta-analytic review. Neuropsychology review, 16(1), 17-42.

Balleine, B. W., Delgado, M. R., & Hikosaka, O. (2007). The role of the dorsal striatum in reward and decision-making. Journal of Neuroscience, 27(31), 8161-8165.

Barbey, A. K., Koenigs, M., & Grafman, J. (2013). Dorsolateral prefrontal contributions to human working memory. Cortex, 49(5), 1195-1205.

Becker, D., Jostmann, N. B., Wiers, R. W., & Holland, R. W. (2015). Approach avoidance training in the eating domain: testing the effectiveness across three

single session studies. Appetite, 85, 58-65.

Brockmeyer, T., Hahn, C., Reetz, C., Schmidt, U., & Friederich, H. C. (2015). Approach Bias Modification in Food Craving—A Proof‐of‐Concept Study.

European Eating Disorders Review, 23(5), 352-360.

Cristea, I. A., Kok, R. N., & Cuijpers, P. (2016). The effectiveness of cognitive bias modification interventions for substance addictions: a meta-analysis. PloS one,

11(9), e0162226.

Cousijn, J., Goudriaan, A. E., & Wiers, R. W. (2011). Reaching out towards cannabis: approach‐bias in heavy cannabis users predicts changes in cannabis use.

Addiction, 106(9), 1667-1674.

Deutsch, R., Gawronski, B., & Hofmann, W. (Eds.). (2016). Reflective and impulsive

determinants of human behavior. Psychology Press.

Drewnowski, A., & Almiron-Roig, E. (2009). Human Perceptions and Preferences for Fat-Rich Foods. Fat detection: Taste, texture, and post ingestive effects, 23, 265. 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.

Eberl, C., Wiers, R. W., Pawelczack, S., Rinck, M., Becker, E. S., & Lindenmeyer, J. (2014). Implementation of Approach Bias Re‐Training in Alcoholism—How Many Sessions are Needed?. Alcoholism: Clinical and experimental

Everitt, B. J., & Robbins, T. W. (2013). From the ventral to the dorsal striatum: devolving views of their roles in drug addiction. Neuroscience & Biobehavioral

Reviews, 37(9), 1946-1954.

Fernández-Serrano, M. J., Pérez-García, M., & Verdejo-García, A. (2011). What are the specific vs. generalized effects of drugs of abuse on neuropsychological performance? Neuroscience & Biobehavioral Reviews, 35(3), 377-406.

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(4), 697.

Foorman, B. R., Francis, D. J., Fletcher, J. M., Schatschneider, C., & Mehta, P. (1998). The role of instruction in learning to read: Preventing reading failure in at-risk children. Journal of Educational Psychology, 90(1), 37–55.

Fridland, E., & Wiers, C. E. (2017). Addiction and embodiment. Phenomenology and

the Cognitive Sciences, 1-28.

Glock, S., Kovacs, C., & Unz, D. (2014). Implicit attitudes toward smoking: How the smell of cigarettes influences responses of college-age smokers and nonsmokers. Journal of health psychology, 19(5), 629-641.

Grafton, B., Mackintosh, B., Vujic, T., & MacLeod, C. (2014). When ignorance is bliss: Explicit instruction and the efficacy of CBM-A for anxiety. Cognitive

Therapy and Research, 38(2), 172-188.

differences in implicit cognition: the implicit association test. Journal of

personality and social psychology, 74(6), 1464.

Heerebout, B. T., & Phaf, R. H. (2010). Emergent oscillations in evolutionary simulations: Oscillating networks increase switching efficacy. Journal of

Cognitive Neuroscience, 22(5), 807-823.

Hermans, D., Vansteenwegen, D., Crombez, G., Baeyens, F., & Eelen, P. (2002). Expectancy-learning and evaluative learning in human classical conditioning: Affective priming as an indirect and unobtrusive measure of conditioned stimulus valence. Behaviour research and therapy, 40(3), 217-234.

Hershberger, W. A. (1986). An approach through the looking-glass. Animal Learning

& Behavior, 14(4), 443-451.

Hofmann, W., De Houwer, J., Perugini, M., Baeyens, F., & Crombez, G. (2010). Evaluative conditioning in humans: a meta-analysis. Psychological

bulletin, 136(3), 390.

Hogarth, L. (2012). Goal-directed and transfer-cue-elicited drug-seeking are dissociated by pharmacotherapy: evidence for independent additive controllers. Journal of Experimental Psychology: Animal Behavior

Processes, 38(3), 266.

Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute substitution in intuitive judgment. Heuristics and biases: The psychology of

intuitive judgment, 49, 49-81.

bias for food cues and poor inhibitory control on unhealthy food intake. Appetite,

87, 358–364.

Kakoschke, N., Kemps, E., & Tiggemann, M. (2017). Addictive Behaviors Approach bias modi fi cation training and consumption: A review of the literature.

Addictive Behaviors, 64, 21–28.

Koob, G. F., & Volkow, N. D. (2016). Neurobiology of addiction: a neurocircuitry analysis. The Lancet Psychiatry, 3(8), 760-773.

Krieglmeyer, R., & Deutsch, R. (2010). Comparing measures of approach–avoidance behaviour: The manikin task vs. two versions of the joystick task. Cognition and

emotion, 24(5), 810-828.

Krieglmeyer, R., De Houwer, J., & Deutsch, R. (2013). On the nature of automatically triggered approach–avoidance behavior. Emotion Review, 5(3), 280-284.

Lancaster, T., & Stead, L. F. (2005). Self-help interventions for smoking cessation. Cochrane Database Syst Rev, 3(3).

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion, attention, and the startle reflex. Psychological review, 97(3), 377.

Larsen, H., Kong, G., Becker, D., Cousijn, J., Boendermaker, W., Cavallo, D., ... & Wiers, R. (2014). Implicit motivational processes underlying smoking in American and Dutch adolescents. Frontiers in psychiatry, 5.

Lindgren, K. P.,Wiers, R.W., Teachman, B. A., Gasser, M. L.,Westgate, E. C., Cousijn, J., ... Neighbors, C. (2015). Attempted training of alcohol approach and drinking identity associations in US undergraduate drinkers: Null results from two studies.

PloS One, 10(8), e0134642.

MacLeod, C., & Clarke, P. J. (2015). The attentional bias modification approach to anxiety intervention. Clinical Psychological Science, 3(1), 58-78.

Manning, V., Staiger, P. K., Hall, K., Garfield, J. B., Flaks, G., Leung, D., ... & Verdejo ‐ Garcia, A. (2016). Cognitive bias modification training during inpatient alcohol detoxification reduces early relapse: a randomized controlled trial. Alcoholism: Clinical and Experimental Research, 40(9), 2011-2019.

Markman, A. B., & Brendl, C. M. (2005). Constraining theories of embodied cognition. Psychological Science, 16(1), 6-10.

Marteau, T. M., Hollands, G. J., & Fletcher, P. C. (2012). Changing human behavior to pre- vent disease: The importance of targeting automatic processes. Science,

337(6101), 1492–1495.

Nederkoorn, C., Houben, K., Hofmann,W., Roefs, A., & Jansen, A. (2010). Control yourself or just eat what you like? Weight gain over a year is predicted by an interactive effect of response inhibition and implicit preference for snack foods.

Health Psychology, 29(4), 389–393.

McEvoy, P. M., Stritzke, W. G., French, D. J., Lang, A. R., & Ketterman, R. (2004). Comparison of three models of alcohol craving in young adults: a cross‐ validation. Addiction, 99(4), 482-497.

Pavlov, I. P., & Anrep, G. V. (2003). Conditioned reflexes. Courier Corporation.

Peeters, M.,Wiers, R. W. ,Monshouwer, K.,, van de Schoot, R., Janssen, T., Volle- bergh, W. A. (2012). Automatic processes in at-risk adolescents: the role of

alcohol-approach tendencies and response inhibition in drinking behavior.

Addiction, 107(11), 1939–1946.

Peeters, M., Monshouwer, K., van de Schoot, R. A., Janssen, T., Vollebergh, W. A., & Wiers, R. W. (2013). Automatic processes and the drinking behavior in early adolescence: a prospective study. Alcoholism: Clinical and Experimental

Research, 37(10), 1737–1744.

Phaf, R. H., Mohr, S. E., Rotteveel, M., & Wicherts, J. M. (2014). Approach, avoidance, and affect: a meta-analysis of approach-avoidance tendencies in manual reaction time tasks. Frontiers in psychology, 5.

Pieters, S., Burk, W. J., Van der Vorst, H., Engels, R. C., & Wiers, R. W. (2014). Impulsive and reflective processes related to alcohol use in young adolescents.

Frontiers in psychiatry, 5, 56.

Robinson, T. E., & Berridge, K. C. (2003). Addiction. Annual Review of Psychology,

54(1), 25–53.

Rotteveel, M., & Phaf, R. H. (2004). Automatic affective evaluation does not automatically predispose for arm flexion and extension. Emotion, 4(2), 156. Rotteveel, M., Gierholz, A., Koch, G., van Aalst, C., Pinto, Y., Matzke, D., ... &

Sasiadek, A. (2015). On the automatic link between affect and tendencies to approach and avoid: Chen and Bargh (1999) revisited. Frontiers in psychology,

6.

Seibt, B., Neumann, R., Nussinson, R., & Strack, F. (2008). Movement direction or change in distance? Self-and object-related approach–avoidance

motions. Journal of Experimental Social Psychology, 44(3), 713-720.

Schumacher, S. E., Kemps, E., & Tiggemann, M. (2016). Bias modification training can alter approach bias and chocolate consumption. Appetite, 96,219–224.

Sharbanee, J. M., Stritzke, W. G., Wiers, R. W., & MacLeod, C. (2013). Alcohol‐ related biases in selective attention and action tendency make distinct contributions to dysregulated drinking behaviour. Addiction, 108(10), 1758-1766.

Sharbanee, J. M., Stritzke, W. G., Wiers, R. W., Young, P., Rinck, M., & MacLeod, C. (2013). The interaction of approach-alcohol action tendencies, working memory capacity, and current task goals predicts the inability to regulate drinking behavior. Psychology of addictive behaviors, 27(3), 649.

Sharbanee, J. M., Hu, L., Stritzke, W. G., Wiers, R. W., Rinck, M., & MacLeod, C. (2014). The effect of approach/avoidance training on alcohol consumption is mediated by change in alcohol action tendency. PloS one, 9(1), e85855.

Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. BF Skinner Foundation.

Smeeton, N. J., Williams, A. M., Hodges, N. J., & Ward, P. (2005). The relative effectiveness of various instructional approaches in developing anticipation skill.

Journal of Experimental Psychology, 11(2), 98–110.

Solarz, A. K. (1960). Latency of instrumental responses as a function of compatibility with the meaning of eliciting verbal signs. Journal of Experimental Psychology,