Master Thesis
The Sneaky Snack Game:
Developing a New Paradigm to Investigate the Development of Habits by Overtraining
April 2nd, 2016
By Johannes Wimberg Student No. 10004448
Abstract
Dual-process theories (Dickinson, 1985) state that initially goal-directed behavior goes over to habitual control by overtraining, i.e. initially outcome-controlled responses become controlled by context stimuli. Evidence comes from devaluation studies, which have shown that instru-mental responses for a reward persist despite devaluation of that reward, e.g. by inducing nau-sea. The present study aims to investigate in human subjects whether overtraining renders goal-directed behavior habitual. In training, two stimuli are trained with the same outcome in sepa-rate blocks. The outcome is valued in one block but not in the other. Spacebar pressing is trained with valued stimulus-outcome pairings. When mixing the stimuli in a test session, response to one stimulus has to be changed. It is hypothesized that after extensive training, participants will struggle to alter their response once the outcome value pertaining to a stimulus is reversed, but not so much after short training. It is furthermore explored how mindfulness relates to reliance on habitual behavior control, while controlling for impulsivity. The results show habitual be-havior by a devaluation effect, but this effect was evident after both short and long training. Neither mindfulness nor impulsivity had a relation to reliance on habits. The failure to find a difference in the devaluation effect after short and long training is proposed to be influenced by sufficient exposure to the stimuli, propositional intentions skipping response-outcome associ-ations, and lacking ecological validity of the task. Missing relations of impulsivity and mindful-ness might corroborate prior research, but need to be bolstered by other measures.
Introduction
Various psychopathological conditions are maintained and perpetuated by habits, e.g. addiction (Dickinson, Wood, & Smith, 2002), obsessive-compulsive disorder (Gillan et al., 2014), but also health concerns like obesity (Volkow & Wise, 2005). It is hence important to understand the development of habits in order to develop and refine adequate interventions. Goal-directed behavior is distinguished from habits by two characteristics. First, goal-directed behavior is controlled by the knowledge between an instrumental response (R) and an outcome (O). That is, an agent knows that a response brings an outcome. This is to be distin-guished from stimulus-response (S–R) associations as proposed by Thorndike's law of effect (1911). Evidence comes from studies reinstating an extinguished response by non-contingent delivery of the outcome (Ostlund & Balleine, 2007). For example, rats are trained an instrumen-tal response like lever pressing in order to receive for food pellets, which is then extinguished by not delivering the pellets. When the outcome is delivered freely without being required to lever press, that response is reinstated. Thus, when rats receive some food pellets after extinc-tion, they resume lever pressing when having the opportunity to do. Such reinstatement of a response cannot be explained in terms of S–R associations. The fact that non-contingent presentation of the outcome is able to reinstate the instrumental response suggests that a rep-resentation of the outcome is associated with the response. Thus, a hallmark of goal-directed behavior is the association between a response and an outcome, rather than a stimulus-response association.
A second characteristic of goal-directed behavior is that the response is performed only if the outcome is desired. Thus, performance of the response depends on the degree to which the agent desires the outcome, such that the response is reduced when the outcome is not de-sired. Studies employing outcome devaluation procedures have provided evidence for the role of desire in goal-directed behavior. These procedures are usually conducted in animals, in which e.g. rats start with instrumental training of usually two responses, for example lever pressing and chain pulling, which are to be performed in order to receive one or two outcomes, for example food pellets or sucrose solution. Following completion of an instrumental training phase, training is continued in a different context, after which one of the outcomes is being devalued by inducing nausea (Adams & Dickinson, 1981) or satiety (Dickinson & Balleine (1998). Then, in a so-called extinction test, the performance of the responses is tested – in ex-tinction, without the outcomes being delivered (in order to dissociate the effect of the devalued outcome from learning experiences during test session). To the degree that the performance of the response is controlled by the outcome value, one would assume that the response for the devalued outcome is reduced compared to the outcome that has not been devalued. And that
goal-directed ability has found considerable support (Colwill & Rescorla, 1985, Colwill & Rescorla, 1988, de Wit, Corlett, Aitken, Dickinson, & Fletcher, 2009; see also Daw, Niv, & Da-yan, 2005, de Wit & Dickinson, 2009). Hence, rats are capable of goal-directed behavior that is controlled by the degree to which an agent values or desires an outcome.
Research has found that overtraining of a response renders it impervious to outcome devaluation (Adams, 1982). In Adams’ (1982) study, the response for a devalued outcome oc-curred as frequently as the response for a non-devalued outcome when instrumental training was given in ten daily sessions (instead of three sessions). Several studies have replicated the effect of extended training to make a response impervious to devaluation (Dickinson, Nicholas, & Adams, 1983, Dickinson, Balleine, Watt, Gonzalez, & Boakes, 1995, Dickinson, Squire, Varga, & Smith, 1998). This finding suggests that performance of the response is not controlled by the outcome value anymore, it is thus habitual rather than goal-directed. It has been argued that R– O and S–R learning processes occur simultaneously (Dickinson, 1985, Daw et al., 2005). Ini-tially, goal-directed R–O associations control behavior, but habitual S–R associations take over behavioral control in the long run.
Research on the resistance to devaluation after overtraining has largely been conducted in animals. In humans, it was successfully demonstrated by Tricomi, Balleine, and O'Doherty (2009). In their study, fractal images are linked to certain buttons. When a fractal appears on the screen, participants are free to press the linked button at their own discretion. Button presses are rewarded on an interval schedule with either an image of M&M's for half of the fractal-button-pairs, or an image of Fritos for the other half. And for half of the participants, the train-ing consisted of short traintrain-ing with one session, whereas for the other half, traintrain-ing consisted of extended training with three daily sessions. After training, one of the outcomes was devalued by satiety. Participants are instructed to eat as much as they can. Subsequently, participants do the same task except that it is conducted in extinction, i.e., the participants will not be rewarded with pictures of the rewards. The study has shown that responding for the devalued outcome was reduced after short training but not after extended training, suggesting that the responses were goal-directed in the former condition. The fact that responding was unaffected by deval-uation in the extended training condition confirms that the behavior has become habitual. Thus, extended training leads to habitual behavior in humans as well.
Until now, Tricomi et al.’s study is the only known study to demonstrate the effect of overtraining on the development of habitual behavior in humans, and therefore deserves atten-tion to be corroborated. The paradigm developed for the present study aims to increase sensi-tivity to distinction of habits from goal-directed behavior. Tricomi et al. (2009) employed a variable interval schedule to reward responses. However, variable interval schedules have been
shown to be more insensitive to outcome devaluation compared to ratio schedules (Dickinson et al., 1983, Hilario, Clouse, Yin, & Costa, 2007). Dickinson (1985) has argued that response-outcome associations are less readily built up in interval schedules since there is no clear corre-lation between responses and outcome delivery. It seems that interval schedules favor habitual S–R-learning processes. The present experiment uses a continuous reinforcement schedule to improve sensitivity to dissociate goal-directed from habitual behavior. Another modification in the present study is the devaluation procedure, which exploits response conflict introduced during a test session. The experimental procedure begins with training of two discriminative stimuli (cues) trained in separate blocks with the same outcome, where one cue precedes the outcome when valued, while the other cue precedes the outcome when it is not valued. For example, participants learn to press a spacebar when a red circle precedes some outcome, e.g. an ice cream. In a separate block, a blue square precedes the same ice cream, but participants learn to withhold response since the ice cream is not valued in that block (and participants are to respond to different outcomes). Having completed training, the originally separate cues are mixed up in a test session, which creates a response conflict with one cue when a particular outcome is now devalued, since response to that one cue has to be changed. Thus, when the ice cream is not valued in test, they are required to stop responding to the red circle. Similar exper-imental procedures exploiting response conflicts have successfully been used in so-called slips of action tasks (de Wit, Ridderinkhof, Fletcher, & Dickinson, 2013). The advantage of such pro-cedures is the lucidity of devaluation compared to devaluation through satiety (as employed by Tricomi et al., 2009). Whereas it is not always clear when someone is sated on an outcome, the employed methodology in the present study in unambiguous with regards to devaluation, and therefore increases reliability of the experimental manipulation.
In addition to investigate whether overtraining renders goal-directed behavior habitual, the present study aims to explore how mindfulness relates to the transition from goal-directed to habitual behavior. Hogarth, Chase, and Baess, 2012 have shown that individual differences in motor impulsivity relate to behavioral sensitivity to devaluation. After instrumental training for either water or chocolate (by pressing one of two outcome-specific buttons), one of the out-comes was devalued by satiety. As expected, devaluation led to reduction of the response for the devalued outcome. In addition, Hogarth et al. have demonstrated that the devaluation effect was significantly weaker in a high impulsivity group compared to the devaluation effect in a low impulsivity group. This particular result suggests that impulsivity relates to relatively stronger reliance on habitual learning processes rather than goal-directed learning. Interestingly, re-search has shown that dispositional mindfulness is negatively correlated with impulsivity (Witt-mann, Peter, Gutina, Otten, Kohls, & Meissner, 2014). Given a negative correlation between
mindfulness and impulsivity, and a positive correlation between mindfulness between impul-sivity and reliance on habits, it is hypothesized that mindfulness will negatively correlate with a propensity to rely on habitual behavior. According to Hölzel et al. (2011), mindfulness has pos-itive effects on four domains: 1. attention, more specifically conflict monitoring and orienting as defined in Posner and Pedersen’s (1990) model of attention, 2. body awareness, 3. emotion regulation, and 4. a decentered perspective on the self. The attentional benefits brought about by mindfulness practice have been linked with a thicker and more active anterior cingulate cor-tex (Grant, Courtemanche, Duerden, Duncan, & Rainville, 2010), which helps monitoring stim-uli in the environment to regulate behavior to pursue goals (van Veen & Carter, 2002). In de-valuation studies, it may be this aspect that helps regulating responding in such a way that im-pulsive stimuli-triggered responses are withheld and outcome-driven responses are facilitated. Interestingly, while mindfulness is associated with increased rostral ACC volume (Grant et al., 2010), increasing impulsivity is associated with a decreased rostral ACC volume (Matsuo, Ni-coletti, Peluso, Hatch, Nemoto, Watanabe, … & Soares, 2009). Thus, mindfulness might tap the same construct and not better explain the reliance on habits than impulsivity itself. Therefore, while exploring the relation between mindfulness and reliance on habits, it will be controlled for impulsivity.
The aim of the present study is thus to replicate the transition from goal-directed to habitual behavior by means of overtraining in humans. A novel paradigm has been developed for that aim. In a training over seven daily sessions, participants are to press the spacebar in response to certain cues which precede a reward. These rewards bring points, which are con-verted to a money reward after training and is given to the participants. In such a way, partici-pants learn stimulus-outcome associations, which should trigger the responses. Simultaneously, there is the opportunity to develop S–R associations. Training consists of two separate blocks where a particular cue precedes an outcome when valued in one block, while another cue pre-cedes the same outcome in the other block, which is then not valued however (i.e., it will not yield points). After training in a test session, cues are mixed together in blocks. Devaluation is implemented by asking participants to collect certain rewards in a test block but not for others. As a consequence, participants have to suppress responding to one cue when an outcome is not valued during test session. Similarly, participants have to start responding to a certain cue when an outcome is valued during test although it was not valued during training. Such trials where the outcome value associated with a particular cue is changed during test session are referred to as incongruent trials. Trials where the associated outcome value of a particular cue is unchanged during test session are referred to as congruent trials.
After short training, responding is assumed to be controlled by outcome-response as-sociations rather than stimulus(cue)-response asas-sociations. Therefore, the participants should respond accurately in both congruent and incongruent trials, that is, they respond accurately to both the valued stimulus and the devalued stimulus that an outcome is associated with, regard-less of whether the participants are forced to alter their response.
However, after extended training, responding is assumed to be controlled by stimulus-response associations. Thus, the cues will trigger the stimulus-response that the participants have learned, even when responding for the signaled outcome will currently lead to loss of points. Thus, par-ticipants will respond adequately during congruent trials but not so much during incongruent trials. While participants will accurately respond to cues where outcome value is unchanged during test, they will respond less accurately to those cues where outcome value is changed during test, that is, either they will tend to withhold their response as they learned in the training sessions although the associated outcome is now valued, or they tend to persist responding to stimuli although the pertaining outcome is not desired anymore. There may be a slight drop in accuracy effect in the short training condition, but it is assumed that the drop is greater in the long training condition.Aside from impaired response accuracy during incongruent trials after extended training, it is hypothesized that extended training will lead to superior accuracy dur-ing congruent trials, compared to accuracy in the short traindur-ing condition.
It is furthermore hypothesized that mindfulness relates to relatively weak reliance on habits. In other words, reliance on goal-directed R–O-learning processes increases as a function of dispositional mindfulness. As a consequence, the effect of devaluation increases as a function of mindfulness, and there will be more accurate responding during incongruent trials after short training, and maybe even extended training. Also, it is expected that mindfulness relates to relatively enhanced responding in congruent trials.
Method Sampling
Participants are (largely) university students who participate for course credit or for compensation of EUR 50.00, plus a variable bonus depending on the participant’s score of max.
EUR5.00. For technical reasons, only participants owning a device running Microsoft Windows could participate.
Materials
Sneaky Snack Game. The Sneaky Snack Game is specifically developed for the purpose to induce and assess R–O-learning, S–O-learning, and S–R-learning. Cues serve as stimuli that
precede some non-valued outcomes and valued outcomes, that participants are to respond for by pressing the spacebar in order to gain points – which are worth money. Those cues are sym-bols which are shown on ice cream vans running across the screen, for example the red outline of a crown. The Sneaky Snack Game is programmed with the stimulus presentation program Presentation.
Training. During training, participants learn that certain stimuli (cues) are linked to
particular outcomes, and are to respond to those stimuli when outcomes are valued in the cur-rent block. When an outcome is valued, correct responding will yield points. Such training oc-curs over seven daily sessions. That way, the game sets up S–O, R–O, and S–R associations.
Initiation. When the Game starts, participants will be welcomed with an initial screen that introduces to the task. The following text is displayed in a white type, centered on a black background (a slash indicates a double return): You’re about to play the Sneaky Snack game for about 30 minutes. / This time you’re collecting ice creams: Magnums, Cornettos, rocket pops, and soft serve. / At the beginning of each trial, you’ll be informed which ice creams you’re supposed to collect. / Remember: you gain 1 point when you press the space bar for the correct ice cream, but you lose 1 point when pressing for the wrong ice creams. All the points you collect across the study will be converted to a bonus, which you’ll receive on top of the regular compensation. You’re start-ing with 50 points. Are you ready? Then press the space bar.
Target display. Following the initiation and prior to each trial block, participants are shown a screen displaying which two of four outcomes are valued in the upcoming training block. These outcomes are ice creams: Cornettos, soft serve, Magnums, and rocket popsicles. The target display is shown in Figure 1a. The non-valued outcomes are shown on the left half of the screen, framed by a red rectangle. The valued outcomes are shown on the other half in a green frame. On the bottom line of the frames, a minus sign is shown for the non-valued out-comes, and a plus sign is shown for the valued outcomes. The target display is shown for four
(a) (b)
Figure 1. (a) Target display screen showing the valued and non-valued outcomes, (b) target check screen following
seconds. In the center of the screen between the two frames, a countdown indicates the time remaining until the subsequent screen appears.
Target check. Following the target display, participants indicate which of the outcomes they were asked to collect by clicking on pictures of these particular outcomes, which are dis-played as shown in Figure 1b. If the participants indicate the wrong targets, the target display is shown again, followed by the target check. The target check is implemented to ensure partici-pants’ awareness of the targets (and thus, proper responding). When the response is correct, the game starts with the first training trial.
Training trial. Each trial starts with the default screen, as shown in Figure 2a, that is, an empty street, the score centered on top, and a skater boy right to the score. This screen will also be shown during intertrial intervals (ITI’s), of which the duration varies between 500-1500ms.
Figure 2. Sequence of a training trial. (a) Default screen during ITI’s. (b) The cue is on the vehicle’s trunk after
ITI, response frame is opened during first half of the vehicle’s run across the screen (600 ms). (c) Halfway the screen (at 600 ms), the outcome is shown on top of the screen. Response frame is closed. (d/e) During the last half of the vehicle’s run across the screen, the skater moves downward to pick up the ice, converging with the van at the end of the run. (f) When the vehicle leaves the screen, feedback is shown for responses.
Following the ITI, an ice cream van or scooter appears on the left side of the screen (figure 2b). The vehicle moves along the street to the right side of the screen in 1200ms. Halfway (at 600ms), one of the four outcomes will be shown on top of the van as shown in figure 2c.
One of eight symbols is shown on the vehicle’s trunk which cue the type of ice cream the vehicle provides. The initial 600ms of the vehicle’s drive form the response frame during which participants are to press the spacebar in reaction to the cues. Once the outcome is dis-played above the vehicle, response frame is closed.
At that point, the skater starts moving by a thirty-degree angle to the point where the van leaves the screen, converging with the van at the end, see figure 2d and 2e. If participants press the space bar within the response frame for valued outcomes, a ‘+ 1’ string in green color will be shown above the skater, as shown in figure 2f. This point will be added to the score shown at the top of the screen. If participants respond for non-valued outcomes within the response frame, a ‘– 1’ in red color will be shown and one point is subtracted from the score. At the end of the trial, the default screen is shown (with the adjusted score). The following trial begins after the ITI.
Table 1
Example of a participant’s stimulus-outcome set.
Set No.
Stimulus
No. Example
Outcome Valued?
Magnum Cornetto Soft serve Rocket pop
1 1 Yes * 2 Yes 3 No * 4 No 2 5 No 6 No * 7 Yes 8 Yes *
Note. Stimuli are randomly paired with one outcome per participant. Each outcome is always a valued target in
one block, but a non-valued outcome in the other block – separate cues are associated with outcome value. One of the two outcome-stimulus pairings is trained under short training in one set, the other under extended train-ing in the other set. An asterisk indicates that the assigned stimulus trained under extended traintrain-ing. Yes/No indicates whether the pertaining outcome is valued or not.
Training blocks. The game consists of ten blocks, where two blocks are constantly alter-nated. In one block, a fixed pair of two of the four outcomes are valued (e.g. Magnums and rocket pops), while the other two outcomes are valued in the other block (e.g. soft serve and Cornetto’s). Each training block has unique cue-outcome pairings that are randomly assigned per participant. In total, there are eight distinct symbols with distinct colors. Half these symbols are shown in one block, and the other half in the other block. In each block, one of the desired (and non-desired) outcomes will be trained three times more often than the other. This is to implement extended (as opposed to short) training. Also, each block uses different vehicles. While the cues are shown on an ice cream van’s trunk in one block, they are shown on a scooter’s trunk in the other block. Table 1 illustrates the cue-outcome-training pairings. One training block consists of 32 trials of a total duration of 1.71 to 2.81 minutes (depending on total duration of all ITI’s). The entire session with ten training blocks thus takes approximately 17-28 minutes.
Stimulus-outcome review. After each training block, participants are asked to indicate which cues were paired with which outcome. On four successive screens, participants click on one of the outcomes (aligned vertically on the right side) that they think was paired with the vehicle shown on the top right side of the screen, as shown in figure 3a. Once the participants click on an outcome, a bar filled with a gradient going from red (very uncertain) to green (very certain) appears below the vehicle’s location, as shown in figure 3b. Participants indicate their certitude about the pairing by clicking on the corresponding location. Stimulus-outcome pair-ings are reviewed in order to assure S–O-learning.
(a) (b)
Figure 3. Stimulus-outcome review. Participants are asked to indicate with which outcome the shown symbol is
associated with. (a) One of the four screens shown per symbol, (b). When having selected an outcome, a bar ap-pears on which participants indicate their certainty about the cue-outcome pairing.
Postgame questionnaire. Following each training session, participants are asked to indi-cate on a bar a) how well they’ve slept, b) how much they were distracted during the session, and c) how motivated participants were to do the task.
Test. Four test blocks are given, in which participants collect all possible combinations
of two outcomes (i.e., block 1: A+B, block 2: B+C, block 3: A+C, block 4: A+D).
Test trials are designed almost identically to the training trials except for two modifica-tions. First, no feedback is given. The outcomes are not shown and are masked instead, and the score will not be shown as well. Second, cues of both originally separated blocks are now mixed. Mixing up the cues has important consequences. Now that test trials also contain both cues that of an outcome, thus the cue that preceded the outcome when valued, as well as the cue that preceded the outcome when non-valued, participants will have to change their responding to one of the cues, depending on whether it is valued or non-valued during test block. Response types are summarized in table 2. If responding is driven by pure S-R-habits, participants will less accurately respond in incongruent trials compared to congruent trials. If responding is driven by S-O-R-associations, that is, controlled by outcome value, participants should be able to respond as accurately during incongruent trials as during congruent trials.
Five Facet Mindfulness Questionnaire (FFMQ, Baer et al., 2006). The FFMQ will be taken as a measure of trait mindfulness. The FFMQ is a 39-item questionnaire that consists of statements like e.g. “When I’m walking, I deliberately notice the sensations of my body moving.” The FFMQ consists of five subscales, however, only the total score will be of interest. Partici-pants rate on a 5-point Likert scale to which degree a statement is true according to the partic-ipants’ own opinion. Research has shown appropriate correlations with total duration of mind-fulness practice (Baer et al., 2006), and appropriate convergent or discriminant validity with
Table 2
Response types in test session, depending on outcome value in training and during test.
S–O Desired in Training
O Desired in Test
Yes No
Yes Congruent:
Keep responding as learned.
Incongruent:
Withhold learned response.
No Incongruent:
Start responding to stimulus.
Congruent:
No responding as learned.
Note. Trials where responding to a cue differs from the original response during training
are referred to as incongruent trials. Trials where responding to cues is unchanged are referred to as congruent trials.
several psychopathological constructs and constructs of positive psychology (Baer et al., 2006, Baer et al., 2008, Taylor & Millear, 2016). The Dutch version of the questionnaire has shown similar validity (de Bruin et al., 2012, Veehof et al., 2011).
Barrat Impulsivity Scale 11 (BIS-11, Patton et al., 1995). The BIS-11 is a self-report questionnaire that measures impulsivity. It consists of thirty statements to which participants indicate how often they are true for the participants on a four-point Likert-scale: 1. rarely/never, 2. occasionally, 3. often, 4. almost always/always. The BIS-11 contains six subscales: attention (e.g. “I don’t ‘pay attention’.”), cognitive instability (e.g. “I have ‘racing’ thoughts.”), motor im-pulsivity (e.g. “I do things without thinking.”), perseverance (e.g. “I change jobs.”), self-control (e.g. “I plan tasks carefully.”), and cognitive complexity (e.g. “I save regularly.”). The total score will be used as an indicator of impulsivity. The BIS-11 has shown good internal consistency (Patton et al., 1995).
Debriefing Questionnaire. After the final testing session, participants fill in a question-naire where are asked whether they took the task serious and whether they cheated at some point during training and if so, asked to specify how (e.g. occasionally having another person perform the task). These data will be used to determine whether participants will be excluded from analysis if their test data are to be considered outliers.
Procedure
Participants make an appointment to come to the lab of the University of Amsterdam. There, the stimulus presentation software ‘Presentation’ will be installed on their laptop. They will perform a demo version of the Sneaky Snack Game afterwards. The demo consists of a distinct cue set and distinct outcomes (pizzas), but keeps the same vehicles. During the demo, experimenters are present in order to explain the game and to assure proper understanding of the game’s functioning. The demo will be followed by the actual Sneaky Snack Game as de-scribed, during which the experimenter leaves the room. For the following six days, participants are to do the Sneaky Snack Game at home. Participants are asked to perform it at roughly the similar time as much as possible and to send the data files each day (to assure that participants commit to the task). They are also told that they receive a small reward on top of the regular compensation depending on the number of points they collect during training. No require-ments were set with regards to the context in which training occurs. One week after the first appointment, directly following the training, participants come to the lab for a final run of the game and the final test session. The game is performed at the participants’ own laptops in order to minimize potential effect of context learning. Finally, participants complete the FFMQ, the BIS and the debriefing questionnaire.
It must be noted that control for impulsivity has been opted for at a later stage of the study. While the BIS was implemented into the experimental procedure for the remaining par-ticipants, the participants who already completed the study were invited to fill out the BIS online. The data of 25 participants could eventually be collected, which will be used for the regression analysis.
Results Participants
53 participants entered the study. 17 participants were excluded from analysis due to technical problems running the game (three participants), self-reported failure to understand the game during the test session (one participant), self-reported cheating during home training (one participant), indication of the wrong targets to collect after a training block (three partic-ipants). Another nine participants were excluded from analysis because their response accuracy during the test session was lower than 0.50 for one or more of the response types shown in table 2, suggesting that the participants may not have understood the changed rules during the test session. Six participants missed one training session but were still included in analysis. 36 par-ticipants remain in analysis. Mean age was 22.31 years (SD = 4.66). Nine parpar-ticipants were male (25%).
Figure 4. Response accuracy shown per day. Separate lines training schedule (long/short), and for outcome value
(valued/non-valued) associated with a stimulus-outcome pairing. 0,7 0,8 0,9 1 1 2 3 4 5 6 7 8 Ac cu ra cy
Training Session Number
Home training
Figure 4 shows response accuracy for all response types in figure 4 for all participants, including the data of participants who missed one training session. To check whether training was successful, response accuracy scores on the first and last training session were compared by a repeated measures ANOVA. The factors day (first/last), outcome value (valued/non-valued), and training length (short/long) were entered into the analysis. Significant main effects were superseded by two interactions. An interaction of value and training length was found, F (1, 34) = 14.182, MSE = 0.005, p = 0.001, 𝜂"# = 0.294. Figure 5a plots the accuracy for long and short
training schedules, with separate bars for outcome value. Accuracy is generally lower after short training than after long training, F(1,34) = 99.410, MSE = 0.003, p < 0.001, 𝜂"# = 0.745, and
accuracy generally higher for valued outcomes, F(1,34) = 13.627, MSE = 0.004, p = 0.001, 𝜂"# =
0.286. The difference in accuracy between valued and non-valued outcomes is smaller after long training than after short training, M = 0.066, SE = 0.018, t(35) = 3.766, p < 0.001. In other words, participants perform better after long training and for valued outcomes, and the performance gap between valued and non-valued outcomes diminishes after long training. Also, a significant interaction of day and training length was found, F(1, 45) = 35.959, MSE = 0.004, p < 0.001, 𝜂"#
= 0.514. Figure 5b shows the accuracy scores on the first and last day with separate bars for long and short training. Performance is generally better on the last day than on the first day, F(1, 34) = 88.455, MSE = 0.004, p < 0.001, 𝜂"# = 0.722, suggesting that training was successful. And as to
be expected, accuracy is lower after short training than after long training, but the difference in accuracy between the long and short schedules is significantly smaller on the last day, M =
(a) (b)
Figure 5. a) Response accuracy for long and short training, with separate bars for valued and non-valued outcomes.
b) Response accuracy on the first and last training day, with separate bars for valued and non-valued outcomes. Error bars indicate standard error values.
0,5 0,6 0,7 0,8 0,9 1
Short Training Long Training
Ac cu ra cy Non-Valued Valued 0,5 0,6 0,7 0,8 0,9 1
First Training Day Last Training Day
Ac
cu
ra
cy
0.0923, SE = 0.0154, t(35) = 5.997, p < .001. Thus, manipulation of short and long training was successful.
Performance during Test Phase
It was hypothesized that response accuracy would be reduced in incongruent trials compared to congruent trials, i.e., participants would respond less accurate to stimuli that are associated with outcomes of which the value was changed in the test session. It was furthermore hypothe-sized that this difference would be of lesser magnitude after short training. A repeated measures ANOVA was performed in order to investigate the influence of value change and training length on response accuracy during the test session. A significant main effect of outcome value during training was found, F(1, 35) = 12.13, MSE = 0.05, p < 0.001, 𝜂"# = 0.257. During the test
session, response accuracy was higher for outcomes that were not desired during training (M = 0.824, SE = 0.012) than for outcomes which were desired during training (M = 0.795, SE = 0.012), although not significantly, F(1, 34)= 0.101, MSE = 0.014, p > 0.05. Furthermore, a sig-nificant interaction of outcome value during training and outcome value during test was found, F(1, 35) = 20.612, MSE = 0.015, p < 0.001, 𝜂"# = 0.371. Figure 6 shows the response accuracy for
each response type. While accuracy during congruent trials was highest, accuracy dropped sub-stantially during incongruent trials. More specifically, the detrimental impact on accuracy was stronger when an outcome was devalued (M = 0.753, SE = 0.017) than when an outcome was made valuable during test session (M = 0.800, SE = 0.017). Contrary to the hypotheses, there
Figure 6. Response accuracy during test session. Bars with dashed borders represent
incon-gruent trials. Error bars indicate standard error values.
0,50 0,60 0,70 0,80 0,90 1,00
Training: Valued Training: Non-valued
Ac
cu
ra
cy
was no significant difference of training length on response accuracy. Both long and short train-ing led to less accurate performance in incongruent trials.
Six participants missed one training session. It might change the results and therefore, the same analysis was performed under exclusion of these participants. The results reflect those of the prior analysis. Aside from a significant main effect for the outcome value during training, F(1, 27) = 14.679, MSE = 0.005, p = 0.001, 𝜂"# = 0.352, a significant of outcome value during
training and outcome value during test was found, F(1, 27) = 12.339, MSE = 0.011, p < 0.05, 𝜂"#
= 0.314. Thus, exclusion of the participants that missed one session does not significantly change the results.
Relating Task Performance to Mindfulness
To investigate the relationship between mindfulness and reliance on habits, an accuracy score was calculated for congruent and for incongruent trials. Training schedules and outcome value during training were collapsed. In a further step, the difference between these two accu-racy scores was calculated. The difference score gives a measure of the reliance on habits. An increasing score indicates a stronger effect of devaluation, and thus more reliance on habitual behavior control. The relation between mindfulness and this difference score was investigated by a regression. Please note that only the data of 25 participants are used for this analysis, as explained in the procedure section. Both the FFMQ and the BIS score were entered as predictors for the regression analysis, with the difference score as dependent variable. The coefficients of the regression are shown in table 3. Mindfulness and impulsivity together account for 1.3% of the variance. Mindfulness is not a significant predictor of reliance on habits, t(19) = –0.493, p = 0.63, as is reflected by a weak correlation, r = – 0.098. Impulsivity as well is no significant pre-dictor of reliance on habits t(19) = – 0.357, p = 0.73, r = – 0.057.
In a study cited in the introduction, Hogarth et al. (2012) found an inverse correlation between impulsivity and reliance on habits only for the motor impulsivity subscale. It was thus checked whether the BSI subscales relate to reliance on habits by entering all subscales into a
Table 3
Parameter estimates for the regression analysis with FFMQ and BSI scores as predictors and accuracy difference score as outcome variable.
Beta t Pearson r Partial r
Constant 0.343
FFMQ –0.090 –0.423 –0.091 –0.090
BSI 0.066 0.312 0.068 0.066
regression. Only a trend was found for the self-control subscale, t(19) = 1.904, p = 0.072, r = 0.265, partial r = 0.400. No significant relation was found for the other factors. Thus, neither mindfulness nor impulsivity are related to reliance on habits.
Conclusions & Discussion
The present study investigated in human subjects whether overtraining leads to habitual behavior. Habitual behavior was conceived as insensitivity to revaluation of an outcome, i.e., initially learned goal-directed behavior perseveres although a formerly valued outcome is not valued anymore, or when a formerly non-valued outcome has assumed a positive value. The present study has clearly shown that training leads to habitual behavior. Contrary to the hy-potheses, both short and long training led to habitual behavior. As an exploratory matter, it was investigated how mindfulness relates to reliance on habits. It was hypothesized that increased trait mindfulness would be associated with a smaller revaluation effect, i.e., more goal-directed behavioral control. No relation of mindfulness with reliance on habits was found. It was also found that impulsivity does not relate to reliance on habits.
A simple explanation for the failure to find a difference may lie in sufficient exposure of the stimuli. Even the short training may have exposed the stimuli for a sufficient number that S–R processes have overtaken behavioral control. Each stimulus set is trained with a total of 1,248 trials, of which a short-trained stimulus is presented 156 times across all training sessions. As can be seen in figure 4, response accuracy in training has almost reached ceiling performance after only three training sessions. Performance would only improve marginally after these three sessions. It suggests that this timeframe was sufficient to internalize the task properly and to establish habitual control of behavior. Therefore, even the short training schedule may have exposed the participants sufficiently to form habitual behavior. To investigate whether exposure was sufficient, further research might reduce the number of stimulus presentations in the short training schedule. A between-subjects design with varying number of stimulus presentations might be helpful to trace the zone of transition from goal-directed to habitual behavior.
Another explanation that might account for the failure to find a difference in sensitivity to revaluation after short and long training may lie in propositional rules that form a shortcut between stimuli and outcomes. To efficiently perform the task, participants might think about which stimuli to respond to and to which not to respond to. These propositional rules might contribute to the formation of S–R habits. By forming such rules, participants link a stimulus to a response, omitting a representation of the outcome that mediates the relation between stimulus and response during transition from goal-directed to habitual behavior. This way of responding resembles implementation intentions (Gollwitzer, 1999), which are if-then plans
that specify a certain action in the presence of a specific situation/cue. For example, if someone desires to lose weight, this person might plan to eat an apple when craving for sweets instead of taking the latter. It has been shown that the working mechanism of implementation intentions is the formation of S–R associations, that is, specifically planned actions are made accessible by specifying concrete situations/cues under which the action is to be performed (Webb & Sheeran, 2007, Webb & Sheeran, 2008). In such a way, implementation intentions promote the formation of new habits, and have been shown to promote self-control and a variety of healthy behaviors (Gollwitzer & Sheeran, 2006). This mechanism might explain the formation of habit-ual behavior even after short training. To circumvent the influence of propositional rules, fur-ther research might implement an adjusted task in which the training is performed with sub-liminal or masked presentation of the stimuli. This measure might not only slow down for-mation of habits, but also increase the sensitivity of the task to discriminate between goal-di-rected and habitual behavior.
It could also be argued that the task does not train any R–O associations since there is no actual reward. Even though the participants receive points for correct answers, which are converted into a small compensation, the participants may not be aware of this reward during the game. Rather, the points and the ice creams are only stimuli indicating that their response was right. In associative terms, the participants learned a response to a stimulus, thus could only learn an S–R habit, or a mere motor program. It is difficult to see a goal-directed nature in this task that is generalizable to everyday life. As a consequence, it training length would not make a difference and yield the results found in the present study. To shed light on this matter, sali-ence of the rewards could be manipulated. For example, more concrete rewards could be dis-played in follow-up research, e.g. “+20 cents” for a correct response, or sexual stimuli (pictures) might be presented. Thus, enhancing the salience helps formation of goal-directed R–O associ-ations while preventing bias towards S–R associassoci-ations.
As there was no difference in accuracy after short and long training, it could not be tested whether high mindfulness levels relate and superior accuracy in incongruent trials and also superior accuracy in congruent trials after long training compared to short training. It was opted instead to use the goal revaluation effect as a gauge of reliance on habits. The present study could not establish a link of mindfulness and impulsivity with reliance on habits. It may be due to an insufficiently valid implementation of the task, as discussed above, but could also be due to problems of the instruments measuring mindfulness and impulsivity, as will be dis-cussed below.
Regular mindfulness practice has positive influences on attentional capacities (as con-ceived by Posner & Pedersen, 1990), which are linked with several changes in the anterior cin-gulate cortex (Hölzel et al., 2011). Impulsivity shares a similar neural substrate with mindfulness (Matsuo et al., 2009). Considering that impulsivity does not seem to be related to reliance on habits, it is not surprising that mindfulness is also not related to reliance on habits. The results suggest that mindfulness does not have a significant enough influence on reliance on habits, and that attentional capacities do not significantly influence behavioral control after overtrain-ing, i.e., improved attentional resources do aid in counteracting habitual behavior control. Ex-ecutive functions might play a more significant role in reliance on habits. However, mindfulness practice has been shown to positively influence executive functions (Tang, Yang, Leve, & Har-old, 2012, Teper & Inzlicht, 2012). One would suppose that executive functioning would play a role in behavioral control and learning processes. In addition, the FFMQ as a measure of mind-fulness may lack validity in reflecting cognitive benefits that are brought about by mindmind-fulness. The FFMQ is a questionnaire that has been validated by correlations with meditation experi-ence and a multitude of constructs concerning emotion regulation and well-being (Baer et al., 2006, de Bruin et al., 2012, Veehof et al., 2011). However, at least to the knowledge of the author of this report, there are no studies studies examining convergent validity with cognitive con-structs that are known to be influenced by regular mindfulness practice, especially attention. As a consequence, the findings found in the present study should be embraced with caution. A study comparing reliance on habits in a mindfulness treatment group with a control group may yield different, and certainly more valid results.
Regarding impulsivity, prior research by one study has only found a correlation of a subfacet of impulsivity, i.e. motor impulsivity (Hogarth et al., 2012). High motor impulsivity was associated with a greater devaluation effect and thus more reliance on habits, and was in-terpreted to fit the nature of habits, that is, habitual behavior are cue-triggered actions. Another study by Watson, Wiers, Hommel, and de Wit (2014) could not replicate the relation between impulsivity and habitual behavior. The present study adds evidence that there is no relation between impulsivity and reliance on habits. Still, the conclusions about impulsivity need to be corroborated with other measures. Impulsivity is a construct with differing operationalisations (de Wit, 2009, Dick et al., 2010). For the present study, the BIS-11 was selected because prior research (on which the hypotheses of this study are founded) has also employed the BIS-11, which makes results better comparable. A study by Reynolds, Ortengren, Richards, and de Wit (2006) investigated relations of prevalent measures of impulsivity and found that self-report measures like the BSI-11 are not related to prevalent behavioral measures, whereas the latter fall into two distinct facets: behavioral inhibition as measured by the Stop/Go-paradigm, and delay
discounting tasks. Reynolds et al. (2006) proposed that self-report questionnaires measure an-other subfacet of impulsivity. Aside from these factors, an-other factors have been proposed (de Wit, 2009, Dick et al., 2010). Therefore, it might be useful to use behavioral measures to inves-tigate the relation of impulsivity with reliance on habits. Other measures may reveal significant relations with other facets of impulsivity. Behavioral measures also have the advantage that they overcome the typical concerns of self-report questionnaires, giving a more valid gauge of im-pulsivity.
There is one important limit of the Sneaky Snack Game that has to be discussed, i.e., its lack of experimental control, which raises three issues. First, there is no control of the training context. Participants may perform the task in different contexts, e.g. at work, in the train, at home in bed, or at different times, which form different temporal contexts. Thraikill and Bou-ton (2015) have shown that habits do not readily generalize to different contexts. When training is performed in several different contexts, performance during the test session may suffer. This may become problematic when the length of the training schedules is reduced to a point where directed behavior is on the verge of shifting to habitual control. To reliably dissociate goal-directed from habitual behavior control, is would be necessary to omit the possibly mitigating influence of different training contexts. On the other hand, it might be useful to exploit the context sensitivity of habitual behavior. When the test session is performed in an entirely unfa-miliar context, e.g. an MRI tube, possibly with a differently looking task, e.g. replacing the vans on the street by little red riding hood in the forest, participants should show a clear devaluation effect if behavior is habitual, while participants should be able to adapt to the new situation. Adding the context into the measurement of habits might enhance the sensitivity of the task. It requires control of the context during training, however.
A second issue that arises from the lack of experimental control is distraction. When being distracted, participants may not sufficiently internalize the S–O and S–R associations, which eventually compromises performance in the test session and obscures the effect. Distrac-tion may also favor habitual S–R control since resources of effortful processing are occupied by the secondary task, that is, participants are not fully engaged in learning the S–O–R associa-tions. Taken together, lacking control of experimental context and distraction may obscure a potential experimental effect. Conducting the entire study at the lab would be the most ideal solution. It requires a lot of time and logistical effort, however.
The third issue that arises from the lack of experimental control is of practical nature. It was not rare that participants missed one or more training sessions. Or participants may not have performed the task at the same time, although they have been explicitly asked to. Technical issues led to exclusion of some participants. Eventually, it means that a lot of participants have
to be excluded and thus a lot of the experimenters’ time and effort goes lost. When participants do not adhere to instructions, it was opted to remind them only very gently of the instructions without being demanding, since the resulting negative affect might have compromised the mo-tivation to perform the task. Consequently, participants might be more inclined to perform the task by the way, e.g. while watching TV, which might again obscure an experimental effect. And when being too demanding, participants might be pushed to lie when being asked at the end of the task to indicate how much they were distracted. It is preferable to have accurate information since participants could be excluded from analysis if need be. Thus, training at home affords little logistical effort at the expense of experimental control and loss of participants, and requires gentle communication.
To summarize, the present study has found limited evidence that overtraining leads to habitual behavior. It is supposed that short training was sufficient enough to set up habitual behavior, or that the developed paradigm did not properly set up R–O associations, favoring direct S–R associations. In addition, the present study has not found relations of mindfulness nor impulsivity on reliance on habitual behavior. Further research needs to adjust the task and employ alternative measures of impulsivity and mindfulness.
3. References
Adams, C. D., & Dickinson, A. (). Instrumental responding following reinforcer devalua-tion. The Quarterly Journal of Experimental Psychology Section B, (), -.
doi:./
Adams, C. D. (). Variations in the sensitivity of instrumental responding to reinforcer de-valuation. The Quarterly Journal of Experimental Psychology Section B, (), -. doi:./
Baer, R. A., Smith, G. T., Hopkins, J., Krietemeyer J., & Toney, L. (). Using self-report as-sessment methods to explore facets of mindfulness. Asas-sessment, (), -.
Baer, R. A., Smith, G. T., Lykins, E., Button, D., Krietemeyer, J., Sauer, S., … Williams, J. M. G. (). Construct validity of the five facet mindfulness questionnaire in meditating and nonmeditating samples. Assessment, (), -.
Balleine, B. W., & Dickinson, A. (). The role of incentive learning in instrumental out-come revaluation by sensory-specific satiety. Animal Learning & Behavior, (), -. doi:./bf
Colwill, R. M., & Rescorla, R. A. (). Postconditioning devaluation of a reinforcer affects instrumental responding. Journal of Experimental Psychology: Animal Behavior Processes, (), -. doi:.//-...
Colwill, R. M., & Rescorla, R. A. (). Associations between the discriminative stimulus and the reinforcer in instrumental learning. Journal of Experimental Psychology: Animal Be-havior Processes, (), -. doi:.//-...
Colwill, R. M., & Rescorla, R. A. (). Effect of reinforcer devaluation on discriminative control of instrumental behavior. Journal of Experimental Psychology: Animal Behavior Processes, (), -. doi:.//-...
Daw, N. D., Niv, Y., & Dayan, P. (). Uncertainty-based competition between prefrontal and dorsolateral striatal systems for behavioral control. Nature Neuroscience, (), -. doi:./nn
De Bruin, E. I., Topper, M., Muskens, J. G., Bögels, S. M., & Kamphuis, J. H. (). Psycho-metric Properties of the Five Facets Mindfulness Questionnaire (FFMQ) in a Meditating and a Non-Meditating Sample. Assessment, (), -.
De Wit, H. (). Impulsivity as a determinant and consequence of drug use: a review of un-derlying processes. Addiction Biology, (), -. doi:./j.-...x De Wit, S., & Dickinson, A. (). Associative theories of goal-directed behavior: A case for
animal–human translational models. Psychological Research, (), -. doi:./s---
De Wit, S., Corlett, P. R., Aitken, M. R., Dickinson, A., & Fletcher, P. C. (). Differential engagement of the ventromedial prefrontal cortex by goal-directed and habitual behavior toward food pictures in humans. Journal of Neuroscience, (), –.
De Wit, S., Niry, D., Wariyar, R., Aitken, M. R., & Dickinson, A. (). Stimulus-outcome interactions during instrumental discrimination learning by rats and humans. Journal of Experimental Psychology: Animal Behavior Processes, (), -. doi:./-...
De Wit, S., van de Vijver, I., & Ridderinkhof, K. R. (). Impaired acquisition of goal-di-rected action in healthy aging. Cognitive, Affective, & Behavioral Neuroscience, (), -. doi:./s---
De Wit, S., Watson, P., Harsay, H. A., Cohen, M. X., van de Vijver, I., & Ridderinkhof, K. R. (). Corticostriatal Connectivity Underlies Individual Differences in the Balance be-tween Habitual and Goal-Directed Action Control. Journal of Neuroscience, (), -. doi:./jneurosci.-.
Dick, D. M., Smith, G., Olausson, P., Mitchell, S. H., Leeman, R. F., O'malley, S. S., & Sher, K. (). Understanding the construct of impulsivity and its relationship to alcohol use dis-orders. Addiction Biology, (), -. doi:./j.-...x
Dickinson, A. (). Actions and Habits: The Development of Behavioural Autonomy. Phil-osophical Transactions of the Royal Society B: Biological Sciences, (), -. doi:./rstb..
Dickinson, A., & Balleine, B. (). Motivational control of goal-directed behavior. Animal Learning & Behavior, (), -.
Dickinson, A., Balleine, B., Watt, A., Gonzalez, F., & Boakes, R. A. (). Motivational con-trol after extended instrumental training. Animal Learning & Behavior, (), -. doi:./bf
Dickinson, A., Nicholas, D. J., & Adams, C. D. (). The effect of the instrumental training contingency on susceptibility to reinforcer devaluation. The Quarterly Journal of Experi-mental Psychology Section B, (), -. doi:./
Dickinson, A.; Squire, S.; Varga, Z.; Smith, J. W. (). Omission learning after instrumental training. The Quarterly Journal of Experimental Psychology, (), -.
Dickinson, A., Wood, N., & Smith, J. W. (). Alcohol seeking by rats: Action or habit? The Quarterly Journal of Experimental Psychology: Section B, (), -.
doi:./
Gillan, C. M., Morein-Zamir, S., Urcelay, G. P., Sule, A., Voon, V., Apergis-Schoute, A. M., … Robbins, T. W. (). Enhanced Avoidance Habits in Obsessive-Compulsive Disorder. Biological Psychiatry, (), –. http://doi.org/./j.biopsych... Gollwitzer, P. M. (). Implementation intentions: Strong effects of simple plans. American
Psychologist, (), -. doi:.//-x...
Gollwitzer, P. M., & Sheeran, P. (). Implementation Intentions and Goal Achievement: A Meta-analysis of Effects and Processes. Advances in Experimental Social Psychology, , -. doi:./s-()-
Grant, J. A., Courtemanche, J., Duerden, E. G., Duncan, G. H., & Rainville, P. (). Cortical thickness and pain sensitivity in zen meditators. Emotion, (), -.
doi:./a
Hilário, M. R. (). Endocannabinoid signaling is critical for habit formation. Frontiers in Integrative Neuroscience, . doi:./neuro...
Hogarth, L., Chase, H. W., & Baess, K. (). Impaired goal-directed behavioural control in human impulsivity. The Quarterly Journal of Experimental Psychology, (), -. doi:./..
Hölzel, B. K., Lazar, S. W., Gard, T., Schuman-Olivier, Z., Vago, D. R., & Ott, U. (). How Does Mindfulness Meditation Work? Proposing Mechanisms of Action From a Concep-tual and Neural Perspective. Perspectives on Psychological Science, (), -.
doi:./
Matsuo, K., Nicoletti, M. A., Peluso, M. A., Hatch, J. P., Nemoto, K., Watanabe, Y., . . . Soares, J. C. (). Anterior cingulate volumes associated with trait impulsivity in individuals with bipolar disorder. Bipolar Disorders, (), -.
doi:./j.-...x
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (). The Unity and Diversity of Executive Functions and Their Contributions to Com-plex “Frontal Lobe” Tasks: A Latent Variable Analysis. Cognitive Psychology, 41(1), 49-100. doi:10.1006/cogp.1999.0734
Ostlund, S. B., & Balleine, B. W. (2007). Selective reinstatement of instrumental performance depends on the discriminative stimulus properties of the mediating outcome. Animal Learning & Behavior, (), -. doi:./bf
Posner, M. I. & Petersen, S. E. (). The attention system of the human brain. Annual Re-view of Neuroscience, , –.
Reynolds, B., Ortengren, A., Richards, J. B., & Wit, H. D. (). Dimensions of impulsive be-havior: Personality and behavioral measures. Personality and Individual Differences, (), -. doi:./j.paid...
Tang, Y., Yang, L., Leve, L. D., & Harold, G. T. (). Improving Executive Function and Its Neurobiological Mechanisms Through a Mindfulness-Based Intervention: Advances Within the Field of Developmental Neuroscience. Child Development Perspectives. doi:./j.-...x
Taylor, N. Z., & Millear, P. M. (). Validity of the Five Facet Mindfulness Questionnaire in an Australian, meditating, demographically diverse sample. Personality and Individual Differences, , -.
Teper, R., & Inzlicht, M. (). Meditation, mindfulness and executive control: the im-portance of emotional acceptance and brain-based performance monitoring. Social Cog-nitive and Affective Neuroscience, (), -. doi:./scan/nss
Thorndike, E. L. (). Animal intelligence: experimental studies. New York: Macmillan. Thrailkill, E. A., & Bouton, M. E. (). Contextual control of instrumental actions and
hab-its. Journal of Experimental Psychology: Animal Learning and Cognition, (), -. doi:./xan
Tricomi, E., Balleine, B. W., & O’Doherty, J. P. (). A specific role for posterior dorsolat-eral striatum in human habit learning. European Journal of Neuroscience, (), -. doi:./j.-...x
Van Veen, V., & Carter, C. (). The anterior cingulate as a conflict monitor: fMRI and ERP studies. Physiology & Behavior, (-), -. doi:./s-()-
Veehof, M. M., ten Klooster, P. M., Taal, E., Westerhof, G., & Bohlmeijer, E. T. (). Psy-chometric properties of the Dutch five facet mindfulness questionnaire (FFMQ) in pa-tients with fibromyalgia. Clinical Rheumatology, , -.
Volkow, N. D., & Wise, R. A. (). How can drug addiction help us understand obesity? Nature Neuroscience, (), -. doi:./nn
Watson, P., Wiers, R., Hommel, B., & Wit, S. D. (). Working for food you don’t desire. Cues interfere with goal-directed food-seeking. Appetite, , -. doi:./j.ap-pet...
Webb, T. L., & Sheeran, P. (). How do implementation intentions promote goal attain-ment? A test of component processes. Journal of Experimental Social Psychology, (), -. doi:./j.jesp...
Webb, T. L., & Sheeran, P. (). Mechanisms of implementation intention effects: The role of goal intentions, self-efficacy, and accessibility of plan components. British Journal of Social Psychology, (), -. doi:./x
Wittmann, M., Peter, J., Gutina, O., Otten, S., Kohls, N., & Meissner, K. (). Individual dif-ferences in self-attributed mindfulness levels are related to the experience of time and cognitive self-control. Personality and Individual Differences, , -.
