Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Measuring approach–avoidance tendencies towards food with touchscreen-based arm movements

Measuring approach–avoidance tendencies towards food with touchscreen-based arm movements Most tasks for measuring automatic approach–avoidance tendencies do not resemble naturalistic approach–avoidance behav- iors. Therefore, we developed a paradigm for the assessment of approach–avoidance tendencies towards palatable food, which is based on arm and hand movements on a touchscreen, thereby mimicking real-life grasping or warding movements. In Study 1 (n = 85), an approach bias towards chocolate-containing foods was found when participants reached towards the stimuli, but not when these stimuli had to be moved on the touchscreen. This approach bias towards food observed in grab movements was replicated in Study 2 (n = 60) and Study 3 (n = 94). Adding task features to disambiguate distance change through either corresponding image zooming (Study 2) or emphasized self-reference (Study 3) did not moderate this effect. Associations between approach bias scores and trait and state chocolate craving were inconsistent across studies. Future studies need to examine whether touchscreen-based approach–avoidance tasks reveal biases towards other stimuli in the appetitive or aversive valence domain and relate to relevant interindividual difference variables. Introduction examples include manikin tasks, in which a symbol of a person needs to be moved towards or away from the tar- Humans typically approach appetitive stimuli and try to get stimulus (de Houwer, Crombez, Baeyens, & Hermans, avoid aversive ones. To bypass limitations of self-report 2001), and joystick-based tasks, in which the target stimulus measures of such partially automatic approach–avoidance needs to be moved towards or away from oneself by pulling tendencies towards environmental stimuli, several behavioral or pushing a joystick (Rinck & Becker, 2007). reaction time tasks have been developed. Two prominent While such tasks represent established and relatively well-validated tools for measuring approach–avoidance ten- dencies, studies that aimed at demonstrating an approach Electronic supplementary material The online version of this bias towards (high-calorie) food produced mixed findings. article (https ://doi.org/10.1007/s0042 6-019-01195 -1) contains Specifically, the majority of studies that used joystick- supplementary material, which is available to authorized users. based tasks did not demonstrate an approach bias towards (high-calorie) food relative to reactions to control stimuli or * Adrian Meule adrian.meule@sbg.ac.at found such a bias only in certain subgroups of participants (Brockmeyer, Hahn, Reetz, Schmidt, & Friederich, 2015; Department of Psychology, University of Salzburg, Kakoschke, Kemps, & Tiggemann, 2015; Maas, Keijsers, Hellbrunner Straße 34, 5020 Salzburg, Austria et al., 2017; Maas, Keijsers, Rinck, Tanis, & Becker, 2015; Center for Cognitive Neuroscience, University of Salzburg, Maas, Woud, et al., 2017; Machulska, Zlomuzica, Adolph, Salzburg, Austria Rinck, & Margraf, 2015; Paslakis, Kühn, Grunert, & Erim, Schoen Clinic Roseneck, Prien am Chiemsee, Germany 2017; Paslakis et al., 2016). We previously demonstrated an Department of MultiMedia Technology, Salzburg University approach bias towards chocolate-containing food in a pre- of Applied Sciences, Puch, Austria dominantly young, female sample, but this bias was only Institute of Psychology, University of Goettingen, Göttingen, found when stimulus categories (food vs. objects) were Germany explicitly associated with approach–avoidance instructions Behavioural Science Institute, Radboud University Nijmegen, (Lender, Meule, Rinck, Brockmeyer, & Blechert, 2018). Nijmegen, The Netherlands Vol.:(0123456789) 1 3 Psychological Research One reason for these inconsistencies may be that (dragging time). Across three studies, we used pictures approaching or avoiding real items such as food involves of chocolate-containing foods and pictures of non-edible reaching, grasping, and moving these objects, which are objects, which we also employed in our previous studies movements that are not executed during the above-men- (Lender et al., 2018; Meule et al., 2019). We expected that tioned computerized tasks. Recent research has started to participants in the present studies would show an approach examine how to measure approach–avoidance tendencies bias towards food (e.g., shorter reaction times in pull food based on more naturalistic movements. Eerland, Guadalupe, than pull objects trials) in our touchscreen-adapted task. This Franken, and Zwaan (2012), for example, used posture on a bias may be reflected in grab movements (which would be balance board as an index for approach–avoidance behaviors. in line with the findings by Schroeder et al., 2016), in drag Recently, Schroeder, Lohmann, Butz, and Plewnia (2016) movements (which would correspond to manikin or joystick demonstrated an approach bias towards food based on grasp- tasks, where no grab movements are required), or both. ing hand movements in a virtual reality setting. Shen, Zhang, In Study 1, participants were instructed to respond to and Krishna (2016) have also demonstrated the relevance either pictures of food or objects by reaching towards them of directly, naturalistically interacting with food stimuli in and moving them to the opposite side of the screen. That a food choice task. In their studies, participants performed is, when target stimuli were displayed near the top (distal) the same task either on a touchscreen device or on a desktop edge of a horizontally positioned touchscreen, they had to be computer. The authors found what they termed a “direct- pulled (approached) and when target stimuli were displayed touch effect” such that participants who performed the task at the bottom (proximal) edge of the screen, they had to be on a touchscreen made more hedonic food choices than those pushed (avoided). In Study 2, we examined whether intro- who performed the task with a non-touch interface. Finally, ducing a zooming effect would facilitate approach–avoidance Zech and colleagues implemented an approach–avoidance inclinations. As in joystick tasks (Rinck & Becker, 2007), task (AAT) on a smartphone that could be moved closer or the picture would enlarge in pull trials and would shrink in further away to simulate naturalistic grasping or rejection push trials. In Study 3, we examined whether manipulating movements (Cring, 2017; Zech, 2015). self-reference would modulate approach–avoidance move- Due to the popularity and wide availability of touch- ments. Specifically, we tested whether presenting a mani- screen-based devices such as smartphones and tablet com- kin (representing the participant) at the bottom half of the puters, we have recently examined an implementation of an screen would facilitate approach–avoidance inclinations and AAT on a touchscreen monitor (Meule, Lender, Richard, whether presenting a manikin at the top half of the screen Dinic, & Blechert, 2019). Here, participants had to pull would reverse response patterns. As approach bias towards or push pictures of chocolate-containing foods or objects food was related to higher trait or state food craving in pre- towards or away from themselves by dragging the pictures vious studies (Brockmeyer et al., 2015; Lender et al., 2018; to the top or bottom of a screen, which was horizontally Meule et al., 2019), we explored whether trait chocolate positioned in front of them. An approach bias towards these craving as well as state chocolate craving before and after foods, however, was only found in individuals who reported the task were associated with approach–avoidance tenden- that they frequently crave chocolate. Moreover, while this cies across all three studies. Although relationships between paradigm represents an AAT with more naturalistic arm craving and approach biases have not been consistently movements in terms of pushing and pulling the pictures, it found in the literature, positive relationships may provide still deviates from real-life approach and avoidance behav- an indication of convergent validity of our new paradigm. iors. For example, approaching food may involve reaching towards that food first and then pulling it closer and avoid- ing food may involve first grabbing it and then moving it Study 1 away from oneself. Therefore, we modified our previous touchscreen-based paradigm such that both grabbing and Methods dragging movements would be required. In the current studies, we thus tested a paradigm in which Participants Eighty-five individuals (82.4% female, participants had to lean forward and reach out to distal n = 70) participated in this study. Mean age was 22.1 years stimuli and then drag these stimuli towards themselves (i.e., (SD = 2.63) and mean body mass index was 22.5  kg/m approach) and to grasp proximal stimuli and then drag these (SD = 3.23). Mean hunger ratings on a scale from 0 = not stimuli away from themselves (i.e., avoidance) on a touch- hungry at all to 10 = very hungry were 4.66 (SD = 2.69). screen monitor. Through this setup, we were able to die ff ren- Thirty participants (56.7% German, n = 17; 40.0% Austrian, tiate between the time participants needed to reach the target n = 12; 3.30% other citizenship, n = 1) were tested at the Uni- stimulus (grabbing time) and the time participants needed versity of Salzburg, Austria. Fifty-five participants (96.4% to move the stimulus towards or away from themselves German, n = 53; 3.60% other citizenship, n = 2) were tested 1 3 Psychological Research at the Radboud University Nijmegen, The Netherlands. Par- When participants reached to the wrong picture, this picture ticipants’ sex (χ = 0.03, p = 0.861), hunger ratings, and could not be moved (i.e., a trial could only be completed by (1) age (both ts < 1.26, ps > 0.211) did not differ between the grabbing the correct picture and moving it to the other side two study centers. of the screen). The task consisted of two blocks and partici- AAT The AAT included 16 pictures displaying choco- pants were instructed to respond to the food pictures in one late-containing foods and 16 pictures displaying non-edible block and to the objects pictures in the other block. Block objects, which were obtained from the food-pics database order was counterbalanced across participants. Within each (Blechert, Meule, Busch, & Ohla, 2014). Picture numbers block, each picture was presented four times at the top and in the food-pics database are 004, 079, 107, 111, 137, 140, four times at the bottom (in randomized order), totaling 128 162, 163, 165, 168, 189, 286, 289, 465, 500, 510 (chocolate trials in one block. Thus, the task consisted of 256 trials pictures), and 1004, 1015, 1045, 1056, 1059, 1095, 1146, in total, and participants had to pull food, push food, pull 1188, 1212, 1226, 1227, 1260, 1265, 1279, 1283, 1293 (neu- objects, and push objects in 64 trials each. tral pictures). Food and objects pictures did not differ in Questionnaires The German, chocolate-adapted ver- color, size, brightness, contrast, complexity, recognizability, sion of the Food Cravings Questionnaire–Trait–reduced or familiarity (all ts < 1.27, ps > 0.214). Each picture had a (FCQ–T–r; Meule & Hormes, 2015) was used to measure resolution of 96 dpi (600 × 450 pixels). The pictures have the frequency and intensity of chocolate cravings in general. been previously used in a joystick-based AAT (Lender et al., The scale has 15 items which are scored from 1 = never to 2018). 6 = always. Internal reliability was α = 0.961 in the current The task was programmed in unity (https://unity 3d.com ) study. The German, chocolate-adapted version of the Food and displayed on a 23-inch iiyama ProLite T2336MSC-B2 Cravings Questionnaire–State (FCQ–S; Meule & Hormes, touchscreen monitor with a resolution of 1920 × 1080 pix- 2015) was used to measure the intensity of current choco- els. Participants were seated in front of a table on which the late craving and hunger before and after the AAT. The scale touchscreen monitor was positioned in portrait orientation has 15 items (12 items for the chocolate craving subscale with an angle of approximately 15° relative to the horizontal and 3 items for the hunger subscale) which are scored from table top (Fig. 1). Each trial started with presentation of a 1 = strongly disagree to 5 = strongly agree. Internal reli- hand symbol in the center of the screen. When participants abilities of the craving subscale were α = 0.939 before and placed five fingers on the hand symbol, two pictures simul- α = 0.950 after the task. Internal reliabilities of the hunger taneously appeared on the screen, which were either a food subscale were α = 0.902 before and α = 0.929 after the task. picture on the top (distal part of the screen) and an object Procedure The study was approved by the institutional picture on the bottom (proximal part of the screen) or vice review board of the University of Salzburg. Participants versa (Fig. 2a). Participants were instructed to reach for the were recruited and tested at the University of Salzburg and target picture. Instructions then read “When the [target] pic- at Radboud University. A few days prior to the laboratory ture is in the lower half [of the screen], push it away from testing session, participants completed an online survey, yourself. When the [target] picture is in the upper half [of which included the FCQ–T–r. In the laboratory testing ses- the screen], pull it towards you.”. When participants reached sion, participants signed informed consent, provided the the target picture, the other picture disappeared so that the sociodemographic information, and completed the FCQ–S. target picture could be moved to the other side of the screen. They then performed the AAT and, subsequently, completed the FCQ–S again. Participation was reimbursed with course credits. Data analyses Trials in which participants lifted their hand too early or reached to the wrong picture were excluded from analyses (4.25% of all trials). We differ- entiated two types of reaction times: the time between picture onset until participants reached the target stim- ulus (grabbing time) and the time participants needed to move the target stimulus to the border of the screen (dragging time). Bootstrapped split-half reliability esti- mates for each condition (pull food, push food, pull objects, push objects) were obtained using the average function of the R package splithalf version 0.5.2 (Par- sons, 2018) performing 5000 random splits. Reliability Fig. 1 Experimental setup in all three studies. Participants sat in front estimates were r = 0 . 9 0 – 0 . 9 1 ( S p e a r m a n – B row n - c o r- of a table on which a touchscreen monitor was positioned in portrait rected r = 0.95–0.96) for grabbing time and r = 0.98 orientation with an angle of approximately 15° sb 1 3 Psychological Research Fig. 2 Representative pull trials in a food block in Study 1 (a), Study 2 (b), and Study 3 (c). Each trial began with the display of a hand symbol in the center of the screen. When participants touched this symbol with five fingers, two pictures appeared at the top and bottom of the screen. Participants were instructed to either respond to pictures with food or to pictures with non-edible objects and to move pictures at the top towards themselves (to the bottom of the screen) and to move pictures at the bottom away from them- selves (to the top of the screen). The picture disappeared and the next trial started when the picture reached the opposite border of the screen. In Study 1, all participants performed the same task (a). In Study 2, one group of participants performed the task with a zoom feature and one group of participants performed the task as in Study 1 (b). In Study 3, one group of participants performed the task with a manikin displayed at the bottom, one group of partici- pants performed the task with a manikin displayed at the top, and one group of participants performed the task as in Study 1 (c). Note that the arrows were not used in the task but are pre- sented here for illustration 1 3 Psychological Research (Spear man–Brown-cor rected r = 0.99) for dr agging Results sb time. In line with joystick-based AAT studies (Rinck & Becker, 2007), median reaction times of all trials as a Grabbing time A significant main effect of stimulus function of condition were calculated for each partici- [F = 56.7, p < 0.001, η = 0.403] indicated that partici- (1,84) p pant. Analyses of variance for repeated measures with pants reacted faster to food (M = 779  ms, SD = 96.8) than trial type (pull vs. push) and stimulus (food vs. objects) as to objects (M = 827  ms, SD = 95.9). This effect, however, within-subjects factors were run separately for grabbing was qualified by a significant interaction trial type × stimu- times and dragging times. To examine correlates of AAT lus [F = 5.45, p = 0.022, η = 0.061]. Following up this (1,84) p performance, approach bias scores were calculated sepa- interaction effect with paired t tests was inconclusive, as rately for grabbing time and dragging time (approach bias grabbing times in pull versus push trials were not signifi- score = [reaction time for pushing food − reaction time for cantly different for either stimulus category (both t s < 1.87, pulling food] − [reaction time for pushing objects − reac- ps > 0.065) and were faster for food versus objects in both tion time for pulling objects]). Thus, positive values indi- pull and push trials (both ts > 5.26, ps < 0.001). The pattern cate an approach bias towards chocolate-containing food of the means, however, suggests that grabbing objects in pull and negative values indicate an avoidance bias from choc- trials was slightly slower than in push trials, potentially due olate-containing foods, relative to non-edible objects. For to basic motor movement characteristics (i.e., more shoulder this approach bias score, reliability estimates using the muscle activity necessary when reaching towards the distal difference-of-difference function of splithalf were r = 0.50 side of the touchscreen). Taking this object-related move- (Spearman–Brown-corrected r = 0.66) for grabbing time ment pattern as a reference, this slowing was not observed sb and r = 0.43 (Spear man–Brown-cor rected r = 0.59) for for grabbing food in pull versus push trials, which might sb dragging time. hint at a facilitation of the grab movement in pull trials due to food approach (Fig. 3a). The main effect of trial type was not significant [F = 1.26, p = 0.265, η = 0.015]. (1,84) p Fig. 3 Grabbing times as a function of trial type (push vs. pull) and time was not included in grabbing times in Study 2 and Study 3. stimulus (food vs. objects) in Study 1 (a), Study 2 (b), and Study 3 Therefore, grabbing times are longer in Study 1 than in Study 2 and (c). Note that grabbing times in Study 1 include the time participants Study 3 and include a main effect of stimulus (i.e., that participants needed to recognize the pictures and decide to which picture they had were faster for food than objects across trial types), which was simi- to reach (i.e., the time between picture onset and the moment when larly found in Study 2 and Study 3 when decision time was analyzed participants lifted their hand off the starting position). This decision separately. Error bars represent standard errors of the mean 1 3 Psychological Research Dragging time A significant main effect of trial type For this, we used a between-subjects design where one group [F = 19.4, p < 0.001, η = 0.187] indicated that partici- of participants performed the same task as in Study 1 and (1,84) p pants were faster in push (M = 408  ms, SD = 108) than in another group of participants performed the task with a pull trials (M = 423 ms, SD = 108). The main effect of stimu- zoom feature. lus [F = 1.57, p = 0.214, η = 0.018] and the interaction An additional change compared to Study 1 concerns the (1,84) p trial type × stimulus [F = 0.80, p = 0.373, η = 0.009] calculation of reaction times. In Study 1, we calculated grab- (1,84) p were not significant. bing time as the time between picture onset and reaching the Correlates of approach bias scores Grabbing time target stimulus. However, this conflated the time participants approach bias scores did not correlate with trait choco- needed to recognize the pictures and decide whether they late craving, current chocolate craving or hunger before or have to reach to the picture at the top or bottom and the after the task (all rs < − 0.125, ps > 0.257). Dragging time time participants needed to reach to the target stimulus. To approach bias scores did not correlate with current hunger remedy this, we differentiated between three reaction times before or after the task (both rs < −  0.106, ps > 0.337) but in Study 2: the time between picture onset and the moment correlated positively with trait chocolate craving (r = 0.250, when participants lifted their hand off the screen (decision p = 0.021) and current chocolate craving before (r = 0.239, time), the time between the moment when participants lifted p = 0.028) and after the task (r = 0.217, p = 0.046). their hand off the screen and when they reached the target stimulus (grabbing time), and—as in Study 1—the time par- Conclusion ticipants needed to move the target stimulus to the border of the screen (dragging time). This allowed us to conduct a Study 1 revealed an approach bias towards food as indicated more fine-grained analysis of action preparation and motor by the trial type × stimulus interaction. This effect was not movement execution effects, in line with previous research found for drag movements and is, therefore, in line with that indicated that approach–avoidance biases might emerge the findings by Schroeder et al. (2016) who examined grasp prior to the execution of the actual motor movement (Rot- movements towards food in a virtual reality setting. When teveel & Phaf, 2004). examining correlates of approach bias scores, however, higher approach bias scores for drag but not grab movements Methods were related to higher trait and state chocolate craving, in line with findings previous studies (Brockmeyer et al., 2015; Participants Sixty women participated in this study at Lender et al., 2018; Meule et al., 2019). In sum, although the the University of Goettingen, Germany. Mean age was data seemed promising for a first touchscreen-based imple- 23.5  years (SD = 2.88) and mean body mass index was mentation of an AAT, there was a need for replication and 21.3 kg/m (SD = 2.44). Most participants had German citi- clarification, which motivated Study 2. zenship (93.3% German, n = 56; 6.70% other citizenship, n = 4). Mean food deprivation (i.e., time since participants’ last meal) was 3.08 h (SD = 3.26). Study 2 AAT The t ask and apparatus were equal to Study 1, except that half of participants performed the task that included Study 1 revealed an approach bias towards food in grab a zooming effect when dragging the stimuli on the screen movements, but post hoc tests when comparing the sin- (Fig. 2b). In pull trials, picture size increased by 20% dur- gle conditions were not clear. No approach bias was found ing the drag movement and—when the picture reached for drag movements. Thus, Study 2 examined whether the border of the screen—picture size increased threefold strengthening approach–avoidance associations with the within 500 ms and disappeared. In push trials, picture size executed arm movements would provide a more clear-cut decreased by 20% during the drag movement and—when pattern of results. Most joystick-based AAT implementa- the picture reached the border of the screen—picture size tions use a zooming feedback to facilitate perceiving pull decreased to zero within 500 ms. and push movements as approach and avoidance behavior Questionnaires As in Study 1, the chocolate version (Laham, Kashima, Dix, & Wheeler, 2015). That is, picture of FCQ–T–r was used to measure trait chocolate craving size increases in pull trials and decreases in push trials. (α = 0.953), and the chocolate version of the FCQ–S was This zooming feature might be crucial for the emergence used to measure current chocolate craving (α = 0.903 before of approach–avoidance biases (Phaf, Mohr, Rotteveel, & and α = 0.942 after the task) and hunger (α = 0.850 before Wicherts, 2014). Thus, Study 2 examined whether intro- and α = 0.899 after the task). ducing a zooming effect would produce an approach bias Procedure The study was approved by the institutional towards food when dragging the pictures on the screen and review board of the Institute of Psychology at the Univer- whether the grabbing bias of Study 1 could be strengthened. sity of Goettingen. Participants were recruited and tested at 1 3 Psychological Research the University of Goettingen. In the laboratory testing ses- A significant main effect of group [ F = 5.88, p = 0.018, (1,58) sion, participants signed informed consent and completed η = 0.092] indicated that participants in the group with the the FCQ–S. They were then randomly assigned to the AAT zoom feature (M = 439 ms, SD = 174) were faster than par- either with or without the zooming feature. After the AAT, ticipants in the group without the zoom feature (M = 550 ms, they completed the FCQ–S again as well as the FCQ–T–r SD = 179). No other effects were significant (all F s < 3.29, and other questionnaires that are not reported here. Partici- ps > 0.074). pation was reimbursed with course credits or € 8. Dragging time There were no significant effects (all Data analyses Participants in the group with (n = 30) and Fs < 0.96, ps > 0.331). without (n = 30) the zooming feature did not differ in food Correlates of approach bias scores Decision time and deprivation, age, trait chocolate craving, current chocolate grabbing time approach bias scores did not correlate with craving, or hunger before the task (all ts < 0.81, ps > 0.423). trait chocolate craving, current chocolate craving or hun- Trials in which participants lifted their hand too early or ger before or after the task (all rs between −  0.169 and reached to the wrong picture were excluded from analy- 0.138, ps > 0.195). Dragging time approach bias scores ses (4.11% of all trials). Bootstrapped split-half reliability did not correlate with current chocolate craving or hunger estimates in the four conditions (pull food, push food, pull before or after the task (all rs between − 0.143 and 0.023, objects, push objects) were r = 0.94–0.96 (r = 0.97–0.98) ps > 0.276), but correlated negatively with trait chocolate sb for decision time, r = 0.95–0.96 (r = 0.97–0.98) for grab- craving (r = − 0.361, p = 0.005). sb bing time, and r = 0.98–0.99 (r = 0.99) for dragging time. sb Median reaction times were submitted to analyses of vari- Conclusion ance for repeated measures with group (zoom vs. no zoom) as between-subjects factor, and trial type (pull vs. push) Study 2 replicated the trial type × stimulus interaction and stimulus (food vs. objects) as within-subjects factors. and, thus, the approach bias towards food found in Study Approach bias scores were calculated for each reaction time 1. Importantly, Study 2 provided additional mechanistic as in Study 1. Reliability estimates using the difference- of- insights: the differentiation between decision time and grab- difference function of splithalf were r = − 0.26 (r = − 0.39) bing time indicated that the main effect of stimulus for grab- sb for the decision time approach bias score, r = 0.41 (r = 0.58) bing time found in Study 1 might be attributed to the fact sb for the grabbing time approach bias score, and r = 0.65 that participants were faster to recognize or categorize the (r = 0.78) for the dragging time approach bias score. food pictures than the objects pictures. Therefore, they start sb their motor movements in response to food earlier, regard- Results less of where the stimulus is located. Thus, the approach bias found in Study 1 and Study 2 is restricted to the actual grab Decision time A significant main effect of stimulus movement and—in contrast to the finding by Rotteveel and [F = 11.6, p = 0.001, η = 0.167] indicated that partici- Phaf (2004)—is not reflected in the action preparation stage. (1,58) p pants reacted faster to food (M = 337 ms, SD = 153) than to Adding the zooming feature did not affect reaction times objects (M = 382 ms, SD = 161). A significant main effect as a function of trial type and/or stimulus category. Instead, of group [F = 6.11, p = 0.016, η = 0.095] indicated main effects of the zooming feature emerged for decision (1,58) p that participants in the group without the zoom feature and grabbing times that are hard to interpret (as these trial (M = 314 ms, SD = 153) reacted faster than participants in phases should not be affected by zooming). It may be spec- the group with the zoom feature (M = 405  ms, SD = 132). ulated that these effects could be due to a slightly longer No other effects were significant (all Fs < 2.22, ps > 0.142). inter-trial interval in the zoom group (because of the picture Grabbing time As in Study 1, the interaction trial fade-out after reaching the border of the screen). Finally, type × stimulus was significant [F = 6.16, p = 0.016, in contrast to Study 1, dragging time approach bias scores (1,58) η = 0.096]. Yet again, following up this interaction effect were not correlated with state chocolate craving, and even with paired t tests was inconclusive as grabbing times in pull negatively related to trait chocolate craving. versus push trials were not significantly different for either stimulus category (both ts < 1.85, ps > 0.069) and did not dif- fer for food versus objects in either trial type (both ts < 1.52, Study 3 ps > 0.133). Similar to Study 1, however, grabbing objects in pull trials was slightly slower than grabbing objects in Study 2 replicated the approach bias towards food as push trials and this direction reversed for foods: they were reflected in grab movements and no approach bias as grabbed faster in pull than in push trials at a descriptive level reflected in drag movements. Yet, it might still be that the (Fig. 3b). Thus, this crossed interaction again points to an lack of finding an approach bias towards food when dragging approach bias towards food as reflected in grab movements. the pictures on the screen may be because participants did 1 3 Psychological Research not associate the executed arm movements with approach- as in Study 1 and as the no zoom group in Study 2; Fig. 2c). ing and avoiding the stimuli. Therefore, we aimed to ensure Participants in the two manikin groups were instructed that that participants actually perceive the required movements the person symbol represented themselves (“A person sym- as approach and avoidance behavior in Study 3. In the litera- bol will be displayed at the top/bottom of the screen. This is ture, this has been achieved by explicitly labeling responses you!”). That is, in the group with the manikin at the bottom, as towards and away from oneself (Eder & Rothermund, the instruction to move the target stimulus towards or away 2008). In fact, it has been found that response patterns can from themselves corresponded to the actual position of the even be reversed by changing response labels. For example, participant in front of the touchscreen monitor. In the group Seibt, Neumann, Nussinson, and Strack (2008) found oppo- with the manikin at the top, however, this was reversed: the site compatibility effects when using inverse instructions instruction to move the target stimulus towards themselves regarding self- versus object-related reference points. There- now corresponded to dragging the stimuli to the top of the fore, to exclude the possibility that the lack of an approach screen and the instruction to move the stimulus away from bias in dragging time might be because participants did themselves now corresponded to dragging the stimuli to the not perceive the downward dragging as moving the stimuli bottom of the screen. towards them and the upward dragging as moving the stimuli Questionnaires As in Study 1 and Study 2, the chocolate away from them, we manipulated self-reference in Study 3. version of FCQ–T–r was used to measure trait chocolate For this, we used a between-subjects design where one craving (α = 0.929), and the chocolate version of the FCQ–S group of participants performed the task as in Study 1 but, was used to measure current chocolate craving (α = 0.897 additionally, a manikin representing the participant and the before and α = 0.923 after the task) and hunger (α = 0.883 participant’s first name were displayed at the bottom of the before and α = 0.910 after the task). screen. As this clearly labeled downward dragging as mov- Procedure The study was approved by the institutional ing the stimuli towards oneself and upward dragging as mov- review board of the University of Salzburg. Participants ing the stimuli away from oneself, this manipulation was were recruited and tested at the University of Salzburg. In expected to facilitate an approach bias towards food ree fl cted the laboratory testing session, participants signed informed in dragging time. Another group of participants performed consent and completed the FCQ–S. They were then ran- the task with the manikin and name displayed at the top of domly assigned to the AAT either with the manikin at the the screen, which was expected to reverse response patterns bottom, with the manikin at the top, or without the manikin. (Seibt et al., 2008). This group allowed investigating which After the AAT, they completed the FCQ–S again as well as type of distance cue would dominate approach bias: if the the FCQ–T–r and other questionnaires that are not reported physical location of a target stimulus at the distal side of here. Participation was reimbursed with course credits. the screen speeded responses to food in pull trials despite Data analyses Participants in the group with the manikin moving the food away from the symbolic self, a dominance at the bottom (n = 31), the group with the manikin at the of physical over symbolic cues can be inferred. Finally, a top (n = 32), and the group without the manikin (n = 31) did control group of participants performed the task as in Study not differ in sex (χ = 1.17, p = 0.585), food deprivation, (2) 1, that is, without the manikin and name displayed on the age, trait chocolate craving, or current chocolate craving screen. or hunger before the task (all Fs < 1.73, ps > 0.182). Trials in which participants lifted their hand too early or reached Methods to the wrong picture were excluded from analyses (4.37% of all trials). Bootstrapped split-half reliability estimates Participants Ninety-four individuals (74.5% female, n = 70) in the four conditions (pull food, push food, pull objects, participated in this study at the University of Salzburg, Aus- push objects) were r = 0.98–0.99 (r = 0.99) for decision sb tria. Mean age was 23.4 years (SD = 4.74) and mean body time, r = 0.97–0.98 (r = 0.98–0.99) for grabbing time, sb mass index was 22.8 kg/m (SD = 4.30). Most participants and r = 0.98–0.99 (r = 0.99) for dragging time. Median sb had German (55.3%, n = 52) or Austrian (36.2%, n = 34) citi- reaction times were submitted to analyses of variance for zenship (8.51% other citizenship, n = 8). Mean food depriva- repeated measures with group (manikin at the bottom vs. tion was 3.11 h (SD = 2.94). manikin at the top vs. no manikin) as between-subjects AAT The t ask and apparatus were equal to Study 1, except factor, and trial type (pull vs. push) and stimulus (food vs. that one group of participants performed the AAT with a objects) as within-subjects factors. Approach bias scores manikin and the participant’s first name displayed at the bot- were calculated as in Study 1 and Study 2. Reliability esti- tom of the screen. Another group of participants performed mates using the difference- of-difference function of splithalf the AAT with the manikin and name displayed at the top of were r = 0.03 (r = 0.04) for the decision time approach bias sb the screen. A third group of participants performed the AAT score, r = 0.56 (r = 0.72) for the grabbing time approach sb without the manikin and name (i.e., the same task version 1 3 Psychological Research bias score, and r = 0.59 (r = 0.72) for the dragging time drag movements. Approach bias scores were not related to sb approach bias score. trait or state chocolate craving. Results Additional analyses Decision time A significant main effect of stimulus As the post hoc tests for following up the interaction effects [F = 10.9, p = 0.001, η = 0.107] indicated that partici- (1,91) p for grabbing time in Studies 1–3 were inconclusive, we pants reacted faster to food (M = 312  ms, SD = 168) than explored whether merging data across studies would pro- to objects (M = 347  ms, SD = 186). The three-way inter- vide a more clear-cut picture. For this, we merged grabbing action group × trial type × stimulus was also significant times of Study 2 and Study 3 (as grabbing time in Study 1 [F = 3.91, p = 0.023, η = 0.079]. Following up this inter- (2,91) p included decision time), leading to a combined sample size action by comparing approach bias scores between groups of n = 154. An analysis of variance for repeated measures revealed that approach bias scores were higher in the group again yielded a significant interaction trial type × stimulus with the manikin at the bottom (M = 6.06  ms, SD = 30.8) [F = 11.5, p = 0.001, η = 0.070]. Paired t tests indi- (1,153) p than in the group with the manikin at the top (M = − 14.2 ms, cated that grabbing food (M = 478 ms, SD = 199) was faster SD = 26.3; t = 2.81, p = 0.007). Approach bias scores in (61) than grabbing objects (M = 498 ms, SD = 219) in pull trials the group without a manikin did not significantly differ from [t = 2.09, p = 0.038]. Grabbing times for food and objects (153) the other two groups (both ts < 1.80, ps > 0.076). No other did not differ in push trials [t = 0.67, p = 0.504]. (153) effects were significant (all Fs < 3.69, ps > 0.057). Grabbing time As in Study 1 and Study 2, the interaction trial type × stimulus was significant [F = 5.38, p = 0.023, (1,91) Discussion η = 0.056] and, again, following up this interaction effect with paired t tests was inconclusive as grabbing times in pull The aim of the current studies was to develop a paradigm versus push trials were not significantly different for either for measuring approach–avoidance tendencies towards food stimulus category (both ts < 1.30, ps > 0.199) and did not dif- with arm movements on a touchscreen. Across all three stud- fer for food versus objects in either trial type (both ts < 1.46, ies, an approach tendency towards food (relative to non- ps > 0.149). Similar to Study 2, however, grabbing objects edible objects) was found when participants had to reach in pull trials was slightly slower than grabbing objects in towards the stimuli. Specifically, when stimuli were located push trials and this direction reversed for foods: they were distally—that is, when participants had to reach out to them grabbed faster in pull than in push trials at a descriptive level in preparation to move the stimuli towards them—there was (Fig. 3c). Thus, this crossed interaction again points to an a speeding of grabbing food compared to non-edible objects. approach bias towards food as reflected in grab movements. No such approach bias was found for the speed of dragging No other effects were significant (all Fs < 1.89, ps > 0.157). the stimuli towards or away from oneself. Thus, results differ Dragging time A significant main effect of stimulus from conventional tasks that measure approach or avoid- [F = 4.42, p = 0.038, η = 0.046] indicated that partici- (1,91) p ance biases by requiring participants to move a manikin or pants moved objects (M = 434 ms, SD = 121) faster than food the stimuli on a computer screen (de Houwer et al., 2001; (M = 453 ms, SD = 164). No other effects were significant Rinck & Becker, 2007). However, they are in line with find- (all Fs < 2.04, ps > 0.156). ings from a virtual reality study in which an approach bias Correlates of approach bias scores Decision time, grab- towards food was reflected in grasping movements towards bing time, and dragging time approach bias scores did not stimuli (Schroeder et al., 2016). correlate with trait chocolate craving, current chocolate crav- ing or hunger before or after the task (all rs between − 0.185 Decision time and 0.074, ps > 0.073). Due to the fine-grained measurement of a composite, multi- Conclusion stage, approach–avoidance behavior, our study series gives insights beyond demonstrating an approach bias towards Study 3 again replicated the trial type × stimulus interac- food. First, the present setup allowed for differentiating deci- tion for grabbing time and, thus, the approach bias towards sion time—that is, the time between stimulus onset and start food found in Study 1 and Study 2. Although manipulating of the hand movement (release of the start button)—from the self-reference did affect decision times, it did not change subsequent two movement stages grabbing and dragging. approach bias towards food as reflected in grab movements In contrast to the findings by Rotteveel and Phaf ( 2004), or the absence of approach bias towards food as reflected in those ‘planning times’ (implicated in our decision time) did 1 3 Psychological Research not carry a bias that would point to a facilitated prepara- internal reliability estimates were generally low for grab- tion of affectively compatible movements. As indicated by bing time and dragging time approach bias scores as well, stimulus type main effects, however, decision times were some were in the acceptable range (Parsons, Kruijt, & Fox, faster for foods compared to objects in Study 2 and Study 2018). Similarly, it appears that findings on the relationship 3 (irrespective of trial type). One reason for this might be a between approach biases and craving are rather inconclu- higher degree of attentional capture of appetitive food rela- sive. For example, either trait food craving (Brockmeyer tive to other stimuli (Carbine et al., 2018). Yet, as we did et  al., 2015), state food craving (Lender et al., 2018), or not measure attentional processes (e.g., eye movements) increases in craving during performing an AAT (Dickson, directly, differences in the physical characteristics of food Kavanagh, & MacLeod, 2016) have been reported to cor- and objects pictures—although those were well-matched— relate with an approach bias towards food. When looking cannot be fully excluded. Another reason may be ‘classifi- at the wider literature that include studies using alcohol-, cation speed’: because participants were instructed to react tobacco-, and cannabis-related AATs, relationships between to either food or non-edible objects, they had to categorize approach biases and craving have also been found incon- the pictures to identify the target stimuli and execute the sistently (Cousijn, Goudriaan, & Wiers, 2011; Schoenmak- required movement. Using such instructions typically leads ers, Wiers, & Field, 2008; Wiers et al., 2013, 2014). Thus, to faster response latencies to food versus neutral stimuli in future studies are needed that clarify whether approach bias simple reaction time tasks because the food category is more towards food can be found independent of food craving or specific than the more diverse category of neutral objects whether it rather relates to trait, state, or changes in (i.e., (Loeber, Grosshans, Herpertz, Kiefer, & Herpertz, 2013; cue-induced) food craving. Meule et al., 2014). Future directions Grabbing and dragging time Interpretation of results is limited to the stimuli and partici- When differentiating between decision and grabbing time, pant characteristics in the current study. That is, results may Study 2 and Study 3 converged in showing that approach be different for other stimulus categories (e.g., other appeti- bias towards food only emerged during grabbing motor tive stimuli such as savory foods or alcoholic beverages) and movements and not during dragging. Facilitated grabbing in other samples (e.g., clinical samples such as individuals on the background of comparable decision and dragging with eating disorders or obesity). Moreover, future stud- times suggest that stimulus–response compatibility effects ies need to examine whether approach bias towards food potentially driven by approach biases emerge during motoric as reflected in grab movements in our paradigm relates to control of grab movements. Future research might follow- actual consumption of these foods. up on this in other AAT implementations, for example, in An important next step would be to evaluate whether balance board or virtual reality studies (Eerland et al., 2012; modifying the current paradigm to a training (e.g., by con- Schroeder et al., 2016) or other setups (e.g., computer gam- sistently presenting food stimuli at the bottom and control ing, 3D navigation). Dragging was not modulated by stimu- stimuli at the top) results in decreased approach bias and lus or trial type, suggesting that ‘securing’ food from a distal intake of the avoided foods. Previous findings on modify - to a proximal position is not particularly biased, at least in ing behavior using approach–avoidance trainings have been our setup. Yet, the very high internal reliabilities suggest that ambiguous. For example, while joystick-based trainings there was not much variation in dragging times across trials. (e.g., repeatedly avoiding appetitive stimuli in terms of Likely, participants operated at maximal speed in all trials, push movements) have been found to reduce approach ten- which may have concealed stimulus and trial type effects. dencies towards appetitive stimuli, effects on actual intake are less consistent (Becker, Jostmann, & Holland, 2018; Correlates of approach bias scores Kakoschke, Kemps, & Tiggemann, 2017). That is, sev- eral studies did not find that an avoidance training reduced When examining correlates of approach biases, however, actual consumption of alcoholic beverages (Leeman et al., no consistent associations were obtained. Grabbing time 2018), soft drinks (Krishna & Eder, 2018), or high-calorie approach bias did not correlate with trait and state chocolate foods (Becker, Jostmann, Wiers, & Holland, 2015; Dickson craving and although dragging time approach bias did cor- et al., 2016; Ferentzi et al., 2018; Warschburger, Gmeiner, relate with these measures in Study 1, these associations did Morawietz, & Rinck, 2018). Among other explanations, one not replicate in Study 2 and Study 3. One reason for finding reason may be that these tasks do not involve naturalistic no or inconsistent correlations with approach bias scores approach and avoidance behaviors, which may hinder trans- may be their insufficient reliability. In the current studies, lation of training effects into real-world behavior. There- decision time approach bias scores were unreliable. While fore, future studies may examine whether touchscreen-based 1 3 Psychological Research of approach and avoidance reactions. Journal of Experi- approach–avoidance trainings may be more effective for mental Psychology: General, 137, 262–281. https ://doi. modifying eating behavior. Such trainings would then need org/10.1037/0096-3445.137.2.262. to be rigorously pitted against conventional techniques to Eerland, A., Guadalupe, T. M., Franken, I. H. A., & Zwaan, R. A. reveal the best practices to modify consumption behaviors (2012). Posture as index for approach-avoidance behavior. PLoS One, 7(2), e31291. https://doi.or g/10.1371/journal.pone.00312 91 . with approach–avoidance interventions. Ferentzi, H., Scheibner, H., Wiers, R., Becker, E. S., Lindenmeyer, J., Beisel, S., & Rinck, M. (2018). Retraining of automatic action ten- Acknowledgements Open access funding provided by Paris Lodron dencies in individuals with obesity: A randomized controlled trial. University of Salzburg. This work was supported by the European Appetite, 126, 66–72. https://doi.or g/10.1016/j.appet.2018.03.016 . Research Council (ERC) under the European Union’s Horizon 2020 Kakoschke, N., Kemps, E., & Tiggemann, M. (2015). Combined research and innovation program (ERC-StG-2014 639445 NewEat). effects of cognitive bias for food cues and poor inhibitory con- The authors would like to thank Elena Fischer, Gillian Kleveman, Doris trol on unhealthy food intake. Appetite, 87, 358–364. https ://doi. Olbrich, Julian Ornig, Alexandra Peitl, Karin Scheibenpflug, Judith org/10.1016/j.appet .2015.01.004. Schumilas, Magdalena Spitzbart, and Elisabeth Weingartner for col- Kakoschke, N., Kemps, E., & Tiggemann, M. (2017). Approach bias lecting the data. modic fi ation training and consumption: A review of the literature. Addictive Behaviors, 64, 21–28. https ://doi.org/10.1016/j.addbe Open Access This article is distributed under the terms of the Crea- h.2016.08.007. tive Commons Attribution 4.0 International License (http://creat iveco Krishna, A., & Eder, A. B. (2018). No effects of explicit approach- mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- avoidance training on immediate consumption of soft drinks. tion, and reproduction in any medium, provided you give appropriate Appetite, 130, 209–218. https ://doi.or g/10.1016/j.appe t credit to the original author(s) and the source, provide a link to the .2018.08.023. Creative Commons license, and indicate if changes were made. Laham, S. M., Kashima, Y., Dix, J., & Wheeler, M. (2015). A meta- analysis of the facilitation of arm flexion and extension move- ments as a function of stimulus valence. Cognition and Emotion, 29, 1069–1090. https ://doi.org/10.1080/02699 931.2014.96809 6. Leeman, R. F., Nogueira, C., Wiers, R. W., Cousijn, J., Serafini, K., References DeMartini, K. S., & O’Malley, S. S. (2018). A test of multisession automatic action tendency retraining to reduce alcohol consump- Becker, D., Jostmann, N. B., & Holland, R. W. (2018). Does approach tion among young adults in the context of a human laboratory bias modification really work in the eating domain? A commen- paradigm. Alcoholism: Clinical and Experimental Research, 42, tary on Kakoschke et al. (2017). Addictive Behaviors, 77, 293– 803–814. https ://doi.org/10.1111/acer.13613 . 294. https ://doi.org/10.1016/j.addbe h.2017.02.025. Lender, A., Meule, A., Rinck, M., Brockmeyer, T., & Blechert, J. Becker, D., Jostmann, N. B., Wiers, R. W., & Holland, R. W. (2015). (2018). Measurement of food-related approach–avoidance biases: Approach avoidance training in the eating domain: Testing the Larger biases when food stimuli are task relevant. Appetite, 125, effectiveness across three single session studies. Appetite, 85, 42–47. https ://doi.org/10.1016/j.appet .2018.01.032. 58–65. https ://doi.org/10.1016/j.appet .2014.11.017. Loeber, S., Grosshans, M., Herpertz, S., Kiefer, F., & Herpertz, S. Blechert, J., Meule, A., Busch, N. A., & Ohla, K. (2014). Food-pics: an C. (2013). Hunger modulates behavioral disinhibition and atten- image database for experimental research on eating and appetite. tion allocation to food-associated cues in normal-weight controls. Frontiers in Psychology, 5(617), 1–10. https ://doi.org/10.3389/ Appetite, 71, 32–39. https: //doi.org/10.1016/j.appet. 2013.07.008. fpsyg .2014.00617 . Maas, J., Keijsers, G. P. J., Rinck, M., Sharbanee, J. M., Vroling, Brockmeyer, T., Hahn, C., Reetz, C., Schmidt, U., & Friederich, H.-C. M. S., & Becker, E. S. (2017a). Implicit action tendencies and (2015). Approach bias and cue reactivity towards food in people evaluations in unwanted snacking behavior. International Jour- with high versus low levels of food craving. Appetite, 95, 197– nal of Cognitive Therapy, 10, 79–91. https ://doi.or g/10.1521/ 202. https ://doi.org/10.1016/j.appet .2015.07.013. ijct.2017.10.1.79. Carbine, K. A., Rodeback, R., Modersitzki, E., Miner, M., LeChemi- Maas, J., Keijsers, G. P. J., Rinck, M., Tanis, J., & Becker, E. S. (2015). nant, J. D., & Larson, M. J. (2018). The utility of event-related Does a dieting goal affect automatic cognitive processes and their potentials (ERPs) in understanding food-related cognition: A sys- trainability? Cognitive Therapy and Research, 39, 378–389. https tematic review and recommendations. Appetite, 128, 58–78. https ://doi.org/10.1007/s1060 8-014-9658-0. ://doi.org/10.1016/j.appet .2018.05.135. Maas, J., Woud, M. L., Keijsers, G. P. J., Rinck, M., Becker, E. S., Cousijn, J., Goudriaan, A. E., & Wiers, R. W. (2011). Reaching out & Wiers, R. W. (2017b). The attraction of sugar: an associa- towards cannabis: approach-bias in heavy cannabis users predicts tion between body mass index and impaired avoidance of sweet changes in cannabis use. Addiction, 106, 1667–1674. https ://doi. snacks. Journal of Experimental Psychopathology, 8, 40–54. https org/10.1111/j.1360-0443.2011.03475 .x. ://doi.org/10.5127/jep.05241 5. Cring, C. (2017). Approach and avoidance tendencies: Using smart- Machulska, A., Zlomuzica, A., Adolph, D., Rinck, M., & Margraf, phones to study behavioral responses and food desires. Master’s J. (2015). “A cigarette a day keeps the goodies away”: smokers thesis at Leiden University, Leiden. show automatic approach tendencies for smoking—but not for de Houwer, J., Crombez, G., Baeyens, F., & Hermans, D. (2001). On food-related stimuli. PLoS One, 10(2), e0116464. ht t ps : // d oi . the generality of the affective Simon effect. Cognition and Emo- org/10.1371/journ al.pone.01164 64. tion, 15, 189–206. https ://doi.org/10.1080/02699 93012 5883. Meule, A., & Hormes, J. M. (2015). Chocolate versions of the Food Dickson, H., Kavanagh, D. J., & MacLeod, C. (2016). The pulling Cravings Questionnaires. Associations with chocolate expo- power of chocolate: Effects of approach–avoidance training on sure-induced salivary flow and ad libitum chocolate consump- approach bias and consumption. Appetite, 99, 46–51. https ://doi. tion. Appetite, 91, 256–265. h t t p s : / / d o i . o r g /1 0 . 1 0 1 6 / j . a p p e t org/10.1016/j.appet .2015.12.026. .2015.04.054. Eder, A. B., & Rothermund, K. (2008). When do motor behav- Meule, A., Lender, A., Richard, A., Dinic, R., & Blechert, J. (2019). iors (mis)match affective stimuli? An evaluative coding view Approach–avoidance tendencies towards food: Measurement on 1 3 Psychological Research a touchscreen and the role of attention and food craving. Appetite, Schroeder, P. A., Lohmann, J., Butz, M. V., & Plewnia, C. (2016). 137, 145–151. https ://doi.org/10.1016/j.appet .2019.03.002. Behavioral bias for food reflected in hand movements: a prelimi- Meule, A., Lutz, A., Krawietz, V., Stützer, J., Vögele, C., & Kübler, nary study with healthy subjects. Cyberpsychology, Behavior, A. (2014). Food-cue affected motor response inhibition and self- and Social Networking, 19, 120–126. https ://doi.org/10.1089/ reported dieting success: a pictorial affective shifting task. Fron -cyber .2015.0311. tiers in Psychology, 5(216), 1–17. https ://doi.org/10.3389/fpsyg Seibt, B., Neumann, R., Nussinson, R., & Strack, F. (2008). Movement .2014.00216. direction or change in distance? Self- and object-related approach– Parsons, S. (2018). splithalf: robust estimates of split half reliability. avoidance motions. Journal of Experimental Social Psychology, figshare. https ://doi.org/10.6084/m9.figsh are.55591 75.v3. 44, 713–720. https ://doi.org/10.1016/j.jesp.2007.04.013. Parsons, S., Kruijt, A.-W., & Fox, E. (2018). Psychological Science Shen, H., Zhang, M., & Krishna, A. (2016). Computer interfaces and needs a standard practice of reporting the reliability of cognitive the “direct-touch” effect: Can iPads increase the choice of hedonic behavioural measurements. PsyArXiv. https ://doi.org/10.31234 / food? Journal of Marketing Research, 53, 745–758. https ://doi. osf.io/6ka9z . org/10.1509/jmr.14.0563. Paslakis, G., Kühn, S., Grunert, S., & Erim, Y. (2017). Explicit and Warschburger, P., Gmeiner, M., Morawietz, M., & Rinck, M. (2018). implicit approach vs avoidance tendencies towards high vs low Evaluation of an approach–avoidance training intervention for calorie food cues in patients with obesity and active binge eating children and adolescents with obesity: A randomized placebo- disorder. Nutrients, 9(10), 1–16. https ://doi.org/10.3390/nu910 controlled prospective trial. European Eating Disorders Review, 1068. 26, 472–482. https ://doi.org/10.1002/erv.2607. Paslakis, G., Kühn, S., Schaubschläger, A., Schieber, K., Röder, K., Wiers, C. E., Kühn, S., Javadi, A. H., Korucuoglu, O., Wiers, R. W., Rauh, E., & Erim, Y. (2016). Explicit and implicit approach vs. Walter, H., & Bermpohl, F. (2013). Automatic approach bias avoidance tendencies towards high vs. low calorie food cues in towards smoking cues is present in smokers but not in ex-smok- patients with anorexia nervosa and healthy controls. Appetite, 107, ers. Psychopharmacology, 229, 187–197. https://doi.or g/10.1007/ 171–179. https ://doi.org/10.1016/j.appet .2016.08.001.s0021 3-013-3098-5. Phaf, R. H., Mohr, S. E., Rotteveel, M., & Wicherts, J. M. (2014). Wiers, C. E., Stelzel, C., Park, S. Q., Gawron, C. K., Ludwig, V. U., Approach, avoidance, and affect: a meta-analysis of approach- Gutwinski, S., & Bermpohl, F. (2014). Neural correlates of alco- avoidance tendencies in manual reaction time tasks. Frontiers hol-approach bias in alcohol addiction: the spirit is willing but the in Psychology, 5(378), 1–16. h ttp s : //d oi. or g/ 10. 3389 /fp sy g flesh is weak for spirits. Neuropsychopharmacology, 39, 688–697. .2014.00378. https ://doi.org/10.1038/npp.2013.252. Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear Zech, H. G. (2015). The Mobile Approach–Avoidance Task. Master’s of spiders. Journal of Behavior Therapy and Experimental Psy- thesis at Leiden University, Leiden. chiatry, 38, 105–120. https://doi.or g/10.1016/j.jbtep.2006.10.001 . Rotteveel, M., & Phaf, R. H. (2004). Automatic affective evaluation Publisher’s Note Springer Nature remains neutral with regard to does not automatically predispose for arm flexion and extension. jurisdictional claims in published maps and institutional affiliations. Emotion, 4, 156–172. https://doi.or g/10.1037/1528-3542.4.2.156. Schoenmakers, T., Wiers, R. W., & Field, M. (2008). Effects of a low dose of alcohol on cognitive biases and craving in heavy drink- ers. Psychopharmacology, 197, 169–178. https://doi.or g/10.1007/ s0021 3-007-1023-5. 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Psychological Research Springer Journals

Measuring approach–avoidance tendencies towards food with touchscreen-based arm movements

Loading next page...
 
/lp/springer-journals/measuring-approach-avoidance-tendencies-towards-food-with-touchscreen-tWGwu0PLLi

References (41)

Publisher
Springer Journals
Copyright
Copyright © 2019 by The Author(s)
Subject
Psychology; Psychology Research
ISSN
0340-0727
eISSN
1430-2772
DOI
10.1007/s00426-019-01195-1
Publisher site
See Article on Publisher Site

Abstract

Most tasks for measuring automatic approach–avoidance tendencies do not resemble naturalistic approach–avoidance behav- iors. Therefore, we developed a paradigm for the assessment of approach–avoidance tendencies towards palatable food, which is based on arm and hand movements on a touchscreen, thereby mimicking real-life grasping or warding movements. In Study 1 (n = 85), an approach bias towards chocolate-containing foods was found when participants reached towards the stimuli, but not when these stimuli had to be moved on the touchscreen. This approach bias towards food observed in grab movements was replicated in Study 2 (n = 60) and Study 3 (n = 94). Adding task features to disambiguate distance change through either corresponding image zooming (Study 2) or emphasized self-reference (Study 3) did not moderate this effect. Associations between approach bias scores and trait and state chocolate craving were inconsistent across studies. Future studies need to examine whether touchscreen-based approach–avoidance tasks reveal biases towards other stimuli in the appetitive or aversive valence domain and relate to relevant interindividual difference variables. Introduction examples include manikin tasks, in which a symbol of a person needs to be moved towards or away from the tar- Humans typically approach appetitive stimuli and try to get stimulus (de Houwer, Crombez, Baeyens, & Hermans, avoid aversive ones. To bypass limitations of self-report 2001), and joystick-based tasks, in which the target stimulus measures of such partially automatic approach–avoidance needs to be moved towards or away from oneself by pulling tendencies towards environmental stimuli, several behavioral or pushing a joystick (Rinck & Becker, 2007). reaction time tasks have been developed. Two prominent While such tasks represent established and relatively well-validated tools for measuring approach–avoidance ten- dencies, studies that aimed at demonstrating an approach Electronic supplementary material The online version of this bias towards (high-calorie) food produced mixed findings. article (https ://doi.org/10.1007/s0042 6-019-01195 -1) contains Specifically, the majority of studies that used joystick- supplementary material, which is available to authorized users. based tasks did not demonstrate an approach bias towards (high-calorie) food relative to reactions to control stimuli or * Adrian Meule adrian.meule@sbg.ac.at found such a bias only in certain subgroups of participants (Brockmeyer, Hahn, Reetz, Schmidt, & Friederich, 2015; Department of Psychology, University of Salzburg, Kakoschke, Kemps, & Tiggemann, 2015; Maas, Keijsers, Hellbrunner Straße 34, 5020 Salzburg, Austria et al., 2017; Maas, Keijsers, Rinck, Tanis, & Becker, 2015; Center for Cognitive Neuroscience, University of Salzburg, Maas, Woud, et al., 2017; Machulska, Zlomuzica, Adolph, Salzburg, Austria Rinck, & Margraf, 2015; Paslakis, Kühn, Grunert, & Erim, Schoen Clinic Roseneck, Prien am Chiemsee, Germany 2017; Paslakis et al., 2016). We previously demonstrated an Department of MultiMedia Technology, Salzburg University approach bias towards chocolate-containing food in a pre- of Applied Sciences, Puch, Austria dominantly young, female sample, but this bias was only Institute of Psychology, University of Goettingen, Göttingen, found when stimulus categories (food vs. objects) were Germany explicitly associated with approach–avoidance instructions Behavioural Science Institute, Radboud University Nijmegen, (Lender, Meule, Rinck, Brockmeyer, & Blechert, 2018). Nijmegen, The Netherlands Vol.:(0123456789) 1 3 Psychological Research One reason for these inconsistencies may be that (dragging time). Across three studies, we used pictures approaching or avoiding real items such as food involves of chocolate-containing foods and pictures of non-edible reaching, grasping, and moving these objects, which are objects, which we also employed in our previous studies movements that are not executed during the above-men- (Lender et al., 2018; Meule et al., 2019). We expected that tioned computerized tasks. Recent research has started to participants in the present studies would show an approach examine how to measure approach–avoidance tendencies bias towards food (e.g., shorter reaction times in pull food based on more naturalistic movements. Eerland, Guadalupe, than pull objects trials) in our touchscreen-adapted task. This Franken, and Zwaan (2012), for example, used posture on a bias may be reflected in grab movements (which would be balance board as an index for approach–avoidance behaviors. in line with the findings by Schroeder et al., 2016), in drag Recently, Schroeder, Lohmann, Butz, and Plewnia (2016) movements (which would correspond to manikin or joystick demonstrated an approach bias towards food based on grasp- tasks, where no grab movements are required), or both. ing hand movements in a virtual reality setting. Shen, Zhang, In Study 1, participants were instructed to respond to and Krishna (2016) have also demonstrated the relevance either pictures of food or objects by reaching towards them of directly, naturalistically interacting with food stimuli in and moving them to the opposite side of the screen. That a food choice task. In their studies, participants performed is, when target stimuli were displayed near the top (distal) the same task either on a touchscreen device or on a desktop edge of a horizontally positioned touchscreen, they had to be computer. The authors found what they termed a “direct- pulled (approached) and when target stimuli were displayed touch effect” such that participants who performed the task at the bottom (proximal) edge of the screen, they had to be on a touchscreen made more hedonic food choices than those pushed (avoided). In Study 2, we examined whether intro- who performed the task with a non-touch interface. Finally, ducing a zooming effect would facilitate approach–avoidance Zech and colleagues implemented an approach–avoidance inclinations. As in joystick tasks (Rinck & Becker, 2007), task (AAT) on a smartphone that could be moved closer or the picture would enlarge in pull trials and would shrink in further away to simulate naturalistic grasping or rejection push trials. In Study 3, we examined whether manipulating movements (Cring, 2017; Zech, 2015). self-reference would modulate approach–avoidance move- Due to the popularity and wide availability of touch- ments. Specifically, we tested whether presenting a mani- screen-based devices such as smartphones and tablet com- kin (representing the participant) at the bottom half of the puters, we have recently examined an implementation of an screen would facilitate approach–avoidance inclinations and AAT on a touchscreen monitor (Meule, Lender, Richard, whether presenting a manikin at the top half of the screen Dinic, & Blechert, 2019). Here, participants had to pull would reverse response patterns. As approach bias towards or push pictures of chocolate-containing foods or objects food was related to higher trait or state food craving in pre- towards or away from themselves by dragging the pictures vious studies (Brockmeyer et al., 2015; Lender et al., 2018; to the top or bottom of a screen, which was horizontally Meule et al., 2019), we explored whether trait chocolate positioned in front of them. An approach bias towards these craving as well as state chocolate craving before and after foods, however, was only found in individuals who reported the task were associated with approach–avoidance tenden- that they frequently crave chocolate. Moreover, while this cies across all three studies. Although relationships between paradigm represents an AAT with more naturalistic arm craving and approach biases have not been consistently movements in terms of pushing and pulling the pictures, it found in the literature, positive relationships may provide still deviates from real-life approach and avoidance behav- an indication of convergent validity of our new paradigm. iors. For example, approaching food may involve reaching towards that food first and then pulling it closer and avoid- ing food may involve first grabbing it and then moving it Study 1 away from oneself. Therefore, we modified our previous touchscreen-based paradigm such that both grabbing and Methods dragging movements would be required. In the current studies, we thus tested a paradigm in which Participants Eighty-five individuals (82.4% female, participants had to lean forward and reach out to distal n = 70) participated in this study. Mean age was 22.1 years stimuli and then drag these stimuli towards themselves (i.e., (SD = 2.63) and mean body mass index was 22.5  kg/m approach) and to grasp proximal stimuli and then drag these (SD = 3.23). Mean hunger ratings on a scale from 0 = not stimuli away from themselves (i.e., avoidance) on a touch- hungry at all to 10 = very hungry were 4.66 (SD = 2.69). screen monitor. Through this setup, we were able to die ff ren- Thirty participants (56.7% German, n = 17; 40.0% Austrian, tiate between the time participants needed to reach the target n = 12; 3.30% other citizenship, n = 1) were tested at the Uni- stimulus (grabbing time) and the time participants needed versity of Salzburg, Austria. Fifty-five participants (96.4% to move the stimulus towards or away from themselves German, n = 53; 3.60% other citizenship, n = 2) were tested 1 3 Psychological Research at the Radboud University Nijmegen, The Netherlands. Par- When participants reached to the wrong picture, this picture ticipants’ sex (χ = 0.03, p = 0.861), hunger ratings, and could not be moved (i.e., a trial could only be completed by (1) age (both ts < 1.26, ps > 0.211) did not differ between the grabbing the correct picture and moving it to the other side two study centers. of the screen). The task consisted of two blocks and partici- AAT The AAT included 16 pictures displaying choco- pants were instructed to respond to the food pictures in one late-containing foods and 16 pictures displaying non-edible block and to the objects pictures in the other block. Block objects, which were obtained from the food-pics database order was counterbalanced across participants. Within each (Blechert, Meule, Busch, & Ohla, 2014). Picture numbers block, each picture was presented four times at the top and in the food-pics database are 004, 079, 107, 111, 137, 140, four times at the bottom (in randomized order), totaling 128 162, 163, 165, 168, 189, 286, 289, 465, 500, 510 (chocolate trials in one block. Thus, the task consisted of 256 trials pictures), and 1004, 1015, 1045, 1056, 1059, 1095, 1146, in total, and participants had to pull food, push food, pull 1188, 1212, 1226, 1227, 1260, 1265, 1279, 1283, 1293 (neu- objects, and push objects in 64 trials each. tral pictures). Food and objects pictures did not differ in Questionnaires The German, chocolate-adapted ver- color, size, brightness, contrast, complexity, recognizability, sion of the Food Cravings Questionnaire–Trait–reduced or familiarity (all ts < 1.27, ps > 0.214). Each picture had a (FCQ–T–r; Meule & Hormes, 2015) was used to measure resolution of 96 dpi (600 × 450 pixels). The pictures have the frequency and intensity of chocolate cravings in general. been previously used in a joystick-based AAT (Lender et al., The scale has 15 items which are scored from 1 = never to 2018). 6 = always. Internal reliability was α = 0.961 in the current The task was programmed in unity (https://unity 3d.com ) study. The German, chocolate-adapted version of the Food and displayed on a 23-inch iiyama ProLite T2336MSC-B2 Cravings Questionnaire–State (FCQ–S; Meule & Hormes, touchscreen monitor with a resolution of 1920 × 1080 pix- 2015) was used to measure the intensity of current choco- els. Participants were seated in front of a table on which the late craving and hunger before and after the AAT. The scale touchscreen monitor was positioned in portrait orientation has 15 items (12 items for the chocolate craving subscale with an angle of approximately 15° relative to the horizontal and 3 items for the hunger subscale) which are scored from table top (Fig. 1). Each trial started with presentation of a 1 = strongly disagree to 5 = strongly agree. Internal reli- hand symbol in the center of the screen. When participants abilities of the craving subscale were α = 0.939 before and placed five fingers on the hand symbol, two pictures simul- α = 0.950 after the task. Internal reliabilities of the hunger taneously appeared on the screen, which were either a food subscale were α = 0.902 before and α = 0.929 after the task. picture on the top (distal part of the screen) and an object Procedure The study was approved by the institutional picture on the bottom (proximal part of the screen) or vice review board of the University of Salzburg. Participants versa (Fig. 2a). Participants were instructed to reach for the were recruited and tested at the University of Salzburg and target picture. Instructions then read “When the [target] pic- at Radboud University. A few days prior to the laboratory ture is in the lower half [of the screen], push it away from testing session, participants completed an online survey, yourself. When the [target] picture is in the upper half [of which included the FCQ–T–r. In the laboratory testing ses- the screen], pull it towards you.”. When participants reached sion, participants signed informed consent, provided the the target picture, the other picture disappeared so that the sociodemographic information, and completed the FCQ–S. target picture could be moved to the other side of the screen. They then performed the AAT and, subsequently, completed the FCQ–S again. Participation was reimbursed with course credits. Data analyses Trials in which participants lifted their hand too early or reached to the wrong picture were excluded from analyses (4.25% of all trials). We differ- entiated two types of reaction times: the time between picture onset until participants reached the target stim- ulus (grabbing time) and the time participants needed to move the target stimulus to the border of the screen (dragging time). Bootstrapped split-half reliability esti- mates for each condition (pull food, push food, pull objects, push objects) were obtained using the average function of the R package splithalf version 0.5.2 (Par- sons, 2018) performing 5000 random splits. Reliability Fig. 1 Experimental setup in all three studies. Participants sat in front estimates were r = 0 . 9 0 – 0 . 9 1 ( S p e a r m a n – B row n - c o r- of a table on which a touchscreen monitor was positioned in portrait rected r = 0.95–0.96) for grabbing time and r = 0.98 orientation with an angle of approximately 15° sb 1 3 Psychological Research Fig. 2 Representative pull trials in a food block in Study 1 (a), Study 2 (b), and Study 3 (c). Each trial began with the display of a hand symbol in the center of the screen. When participants touched this symbol with five fingers, two pictures appeared at the top and bottom of the screen. Participants were instructed to either respond to pictures with food or to pictures with non-edible objects and to move pictures at the top towards themselves (to the bottom of the screen) and to move pictures at the bottom away from them- selves (to the top of the screen). The picture disappeared and the next trial started when the picture reached the opposite border of the screen. In Study 1, all participants performed the same task (a). In Study 2, one group of participants performed the task with a zoom feature and one group of participants performed the task as in Study 1 (b). In Study 3, one group of participants performed the task with a manikin displayed at the bottom, one group of partici- pants performed the task with a manikin displayed at the top, and one group of participants performed the task as in Study 1 (c). Note that the arrows were not used in the task but are pre- sented here for illustration 1 3 Psychological Research (Spear man–Brown-cor rected r = 0.99) for dr agging Results sb time. In line with joystick-based AAT studies (Rinck & Becker, 2007), median reaction times of all trials as a Grabbing time A significant main effect of stimulus function of condition were calculated for each partici- [F = 56.7, p < 0.001, η = 0.403] indicated that partici- (1,84) p pant. Analyses of variance for repeated measures with pants reacted faster to food (M = 779  ms, SD = 96.8) than trial type (pull vs. push) and stimulus (food vs. objects) as to objects (M = 827  ms, SD = 95.9). This effect, however, within-subjects factors were run separately for grabbing was qualified by a significant interaction trial type × stimu- times and dragging times. To examine correlates of AAT lus [F = 5.45, p = 0.022, η = 0.061]. Following up this (1,84) p performance, approach bias scores were calculated sepa- interaction effect with paired t tests was inconclusive, as rately for grabbing time and dragging time (approach bias grabbing times in pull versus push trials were not signifi- score = [reaction time for pushing food − reaction time for cantly different for either stimulus category (both t s < 1.87, pulling food] − [reaction time for pushing objects − reac- ps > 0.065) and were faster for food versus objects in both tion time for pulling objects]). Thus, positive values indi- pull and push trials (both ts > 5.26, ps < 0.001). The pattern cate an approach bias towards chocolate-containing food of the means, however, suggests that grabbing objects in pull and negative values indicate an avoidance bias from choc- trials was slightly slower than in push trials, potentially due olate-containing foods, relative to non-edible objects. For to basic motor movement characteristics (i.e., more shoulder this approach bias score, reliability estimates using the muscle activity necessary when reaching towards the distal difference-of-difference function of splithalf were r = 0.50 side of the touchscreen). Taking this object-related move- (Spearman–Brown-corrected r = 0.66) for grabbing time ment pattern as a reference, this slowing was not observed sb and r = 0.43 (Spear man–Brown-cor rected r = 0.59) for for grabbing food in pull versus push trials, which might sb dragging time. hint at a facilitation of the grab movement in pull trials due to food approach (Fig. 3a). The main effect of trial type was not significant [F = 1.26, p = 0.265, η = 0.015]. (1,84) p Fig. 3 Grabbing times as a function of trial type (push vs. pull) and time was not included in grabbing times in Study 2 and Study 3. stimulus (food vs. objects) in Study 1 (a), Study 2 (b), and Study 3 Therefore, grabbing times are longer in Study 1 than in Study 2 and (c). Note that grabbing times in Study 1 include the time participants Study 3 and include a main effect of stimulus (i.e., that participants needed to recognize the pictures and decide to which picture they had were faster for food than objects across trial types), which was simi- to reach (i.e., the time between picture onset and the moment when larly found in Study 2 and Study 3 when decision time was analyzed participants lifted their hand off the starting position). This decision separately. Error bars represent standard errors of the mean 1 3 Psychological Research Dragging time A significant main effect of trial type For this, we used a between-subjects design where one group [F = 19.4, p < 0.001, η = 0.187] indicated that partici- of participants performed the same task as in Study 1 and (1,84) p pants were faster in push (M = 408  ms, SD = 108) than in another group of participants performed the task with a pull trials (M = 423 ms, SD = 108). The main effect of stimu- zoom feature. lus [F = 1.57, p = 0.214, η = 0.018] and the interaction An additional change compared to Study 1 concerns the (1,84) p trial type × stimulus [F = 0.80, p = 0.373, η = 0.009] calculation of reaction times. In Study 1, we calculated grab- (1,84) p were not significant. bing time as the time between picture onset and reaching the Correlates of approach bias scores Grabbing time target stimulus. However, this conflated the time participants approach bias scores did not correlate with trait choco- needed to recognize the pictures and decide whether they late craving, current chocolate craving or hunger before or have to reach to the picture at the top or bottom and the after the task (all rs < − 0.125, ps > 0.257). Dragging time time participants needed to reach to the target stimulus. To approach bias scores did not correlate with current hunger remedy this, we differentiated between three reaction times before or after the task (both rs < −  0.106, ps > 0.337) but in Study 2: the time between picture onset and the moment correlated positively with trait chocolate craving (r = 0.250, when participants lifted their hand off the screen (decision p = 0.021) and current chocolate craving before (r = 0.239, time), the time between the moment when participants lifted p = 0.028) and after the task (r = 0.217, p = 0.046). their hand off the screen and when they reached the target stimulus (grabbing time), and—as in Study 1—the time par- Conclusion ticipants needed to move the target stimulus to the border of the screen (dragging time). This allowed us to conduct a Study 1 revealed an approach bias towards food as indicated more fine-grained analysis of action preparation and motor by the trial type × stimulus interaction. This effect was not movement execution effects, in line with previous research found for drag movements and is, therefore, in line with that indicated that approach–avoidance biases might emerge the findings by Schroeder et al. (2016) who examined grasp prior to the execution of the actual motor movement (Rot- movements towards food in a virtual reality setting. When teveel & Phaf, 2004). examining correlates of approach bias scores, however, higher approach bias scores for drag but not grab movements Methods were related to higher trait and state chocolate craving, in line with findings previous studies (Brockmeyer et al., 2015; Participants Sixty women participated in this study at Lender et al., 2018; Meule et al., 2019). In sum, although the the University of Goettingen, Germany. Mean age was data seemed promising for a first touchscreen-based imple- 23.5  years (SD = 2.88) and mean body mass index was mentation of an AAT, there was a need for replication and 21.3 kg/m (SD = 2.44). Most participants had German citi- clarification, which motivated Study 2. zenship (93.3% German, n = 56; 6.70% other citizenship, n = 4). Mean food deprivation (i.e., time since participants’ last meal) was 3.08 h (SD = 3.26). Study 2 AAT The t ask and apparatus were equal to Study 1, except that half of participants performed the task that included Study 1 revealed an approach bias towards food in grab a zooming effect when dragging the stimuli on the screen movements, but post hoc tests when comparing the sin- (Fig. 2b). In pull trials, picture size increased by 20% dur- gle conditions were not clear. No approach bias was found ing the drag movement and—when the picture reached for drag movements. Thus, Study 2 examined whether the border of the screen—picture size increased threefold strengthening approach–avoidance associations with the within 500 ms and disappeared. In push trials, picture size executed arm movements would provide a more clear-cut decreased by 20% during the drag movement and—when pattern of results. Most joystick-based AAT implementa- the picture reached the border of the screen—picture size tions use a zooming feedback to facilitate perceiving pull decreased to zero within 500 ms. and push movements as approach and avoidance behavior Questionnaires As in Study 1, the chocolate version (Laham, Kashima, Dix, & Wheeler, 2015). That is, picture of FCQ–T–r was used to measure trait chocolate craving size increases in pull trials and decreases in push trials. (α = 0.953), and the chocolate version of the FCQ–S was This zooming feature might be crucial for the emergence used to measure current chocolate craving (α = 0.903 before of approach–avoidance biases (Phaf, Mohr, Rotteveel, & and α = 0.942 after the task) and hunger (α = 0.850 before Wicherts, 2014). Thus, Study 2 examined whether intro- and α = 0.899 after the task). ducing a zooming effect would produce an approach bias Procedure The study was approved by the institutional towards food when dragging the pictures on the screen and review board of the Institute of Psychology at the Univer- whether the grabbing bias of Study 1 could be strengthened. sity of Goettingen. Participants were recruited and tested at 1 3 Psychological Research the University of Goettingen. In the laboratory testing ses- A significant main effect of group [ F = 5.88, p = 0.018, (1,58) sion, participants signed informed consent and completed η = 0.092] indicated that participants in the group with the the FCQ–S. They were then randomly assigned to the AAT zoom feature (M = 439 ms, SD = 174) were faster than par- either with or without the zooming feature. After the AAT, ticipants in the group without the zoom feature (M = 550 ms, they completed the FCQ–S again as well as the FCQ–T–r SD = 179). No other effects were significant (all F s < 3.29, and other questionnaires that are not reported here. Partici- ps > 0.074). pation was reimbursed with course credits or € 8. Dragging time There were no significant effects (all Data analyses Participants in the group with (n = 30) and Fs < 0.96, ps > 0.331). without (n = 30) the zooming feature did not differ in food Correlates of approach bias scores Decision time and deprivation, age, trait chocolate craving, current chocolate grabbing time approach bias scores did not correlate with craving, or hunger before the task (all ts < 0.81, ps > 0.423). trait chocolate craving, current chocolate craving or hun- Trials in which participants lifted their hand too early or ger before or after the task (all rs between −  0.169 and reached to the wrong picture were excluded from analy- 0.138, ps > 0.195). Dragging time approach bias scores ses (4.11% of all trials). Bootstrapped split-half reliability did not correlate with current chocolate craving or hunger estimates in the four conditions (pull food, push food, pull before or after the task (all rs between − 0.143 and 0.023, objects, push objects) were r = 0.94–0.96 (r = 0.97–0.98) ps > 0.276), but correlated negatively with trait chocolate sb for decision time, r = 0.95–0.96 (r = 0.97–0.98) for grab- craving (r = − 0.361, p = 0.005). sb bing time, and r = 0.98–0.99 (r = 0.99) for dragging time. sb Median reaction times were submitted to analyses of vari- Conclusion ance for repeated measures with group (zoom vs. no zoom) as between-subjects factor, and trial type (pull vs. push) Study 2 replicated the trial type × stimulus interaction and stimulus (food vs. objects) as within-subjects factors. and, thus, the approach bias towards food found in Study Approach bias scores were calculated for each reaction time 1. Importantly, Study 2 provided additional mechanistic as in Study 1. Reliability estimates using the difference- of- insights: the differentiation between decision time and grab- difference function of splithalf were r = − 0.26 (r = − 0.39) bing time indicated that the main effect of stimulus for grab- sb for the decision time approach bias score, r = 0.41 (r = 0.58) bing time found in Study 1 might be attributed to the fact sb for the grabbing time approach bias score, and r = 0.65 that participants were faster to recognize or categorize the (r = 0.78) for the dragging time approach bias score. food pictures than the objects pictures. Therefore, they start sb their motor movements in response to food earlier, regard- Results less of where the stimulus is located. Thus, the approach bias found in Study 1 and Study 2 is restricted to the actual grab Decision time A significant main effect of stimulus movement and—in contrast to the finding by Rotteveel and [F = 11.6, p = 0.001, η = 0.167] indicated that partici- Phaf (2004)—is not reflected in the action preparation stage. (1,58) p pants reacted faster to food (M = 337 ms, SD = 153) than to Adding the zooming feature did not affect reaction times objects (M = 382 ms, SD = 161). A significant main effect as a function of trial type and/or stimulus category. Instead, of group [F = 6.11, p = 0.016, η = 0.095] indicated main effects of the zooming feature emerged for decision (1,58) p that participants in the group without the zoom feature and grabbing times that are hard to interpret (as these trial (M = 314 ms, SD = 153) reacted faster than participants in phases should not be affected by zooming). It may be spec- the group with the zoom feature (M = 405  ms, SD = 132). ulated that these effects could be due to a slightly longer No other effects were significant (all Fs < 2.22, ps > 0.142). inter-trial interval in the zoom group (because of the picture Grabbing time As in Study 1, the interaction trial fade-out after reaching the border of the screen). Finally, type × stimulus was significant [F = 6.16, p = 0.016, in contrast to Study 1, dragging time approach bias scores (1,58) η = 0.096]. Yet again, following up this interaction effect were not correlated with state chocolate craving, and even with paired t tests was inconclusive as grabbing times in pull negatively related to trait chocolate craving. versus push trials were not significantly different for either stimulus category (both ts < 1.85, ps > 0.069) and did not dif- fer for food versus objects in either trial type (both ts < 1.52, Study 3 ps > 0.133). Similar to Study 1, however, grabbing objects in pull trials was slightly slower than grabbing objects in Study 2 replicated the approach bias towards food as push trials and this direction reversed for foods: they were reflected in grab movements and no approach bias as grabbed faster in pull than in push trials at a descriptive level reflected in drag movements. Yet, it might still be that the (Fig. 3b). Thus, this crossed interaction again points to an lack of finding an approach bias towards food when dragging approach bias towards food as reflected in grab movements. the pictures on the screen may be because participants did 1 3 Psychological Research not associate the executed arm movements with approach- as in Study 1 and as the no zoom group in Study 2; Fig. 2c). ing and avoiding the stimuli. Therefore, we aimed to ensure Participants in the two manikin groups were instructed that that participants actually perceive the required movements the person symbol represented themselves (“A person sym- as approach and avoidance behavior in Study 3. In the litera- bol will be displayed at the top/bottom of the screen. This is ture, this has been achieved by explicitly labeling responses you!”). That is, in the group with the manikin at the bottom, as towards and away from oneself (Eder & Rothermund, the instruction to move the target stimulus towards or away 2008). In fact, it has been found that response patterns can from themselves corresponded to the actual position of the even be reversed by changing response labels. For example, participant in front of the touchscreen monitor. In the group Seibt, Neumann, Nussinson, and Strack (2008) found oppo- with the manikin at the top, however, this was reversed: the site compatibility effects when using inverse instructions instruction to move the target stimulus towards themselves regarding self- versus object-related reference points. There- now corresponded to dragging the stimuli to the top of the fore, to exclude the possibility that the lack of an approach screen and the instruction to move the stimulus away from bias in dragging time might be because participants did themselves now corresponded to dragging the stimuli to the not perceive the downward dragging as moving the stimuli bottom of the screen. towards them and the upward dragging as moving the stimuli Questionnaires As in Study 1 and Study 2, the chocolate away from them, we manipulated self-reference in Study 3. version of FCQ–T–r was used to measure trait chocolate For this, we used a between-subjects design where one craving (α = 0.929), and the chocolate version of the FCQ–S group of participants performed the task as in Study 1 but, was used to measure current chocolate craving (α = 0.897 additionally, a manikin representing the participant and the before and α = 0.923 after the task) and hunger (α = 0.883 participant’s first name were displayed at the bottom of the before and α = 0.910 after the task). screen. As this clearly labeled downward dragging as mov- Procedure The study was approved by the institutional ing the stimuli towards oneself and upward dragging as mov- review board of the University of Salzburg. Participants ing the stimuli away from oneself, this manipulation was were recruited and tested at the University of Salzburg. In expected to facilitate an approach bias towards food ree fl cted the laboratory testing session, participants signed informed in dragging time. Another group of participants performed consent and completed the FCQ–S. They were then ran- the task with the manikin and name displayed at the top of domly assigned to the AAT either with the manikin at the the screen, which was expected to reverse response patterns bottom, with the manikin at the top, or without the manikin. (Seibt et al., 2008). This group allowed investigating which After the AAT, they completed the FCQ–S again as well as type of distance cue would dominate approach bias: if the the FCQ–T–r and other questionnaires that are not reported physical location of a target stimulus at the distal side of here. Participation was reimbursed with course credits. the screen speeded responses to food in pull trials despite Data analyses Participants in the group with the manikin moving the food away from the symbolic self, a dominance at the bottom (n = 31), the group with the manikin at the of physical over symbolic cues can be inferred. Finally, a top (n = 32), and the group without the manikin (n = 31) did control group of participants performed the task as in Study not differ in sex (χ = 1.17, p = 0.585), food deprivation, (2) 1, that is, without the manikin and name displayed on the age, trait chocolate craving, or current chocolate craving screen. or hunger before the task (all Fs < 1.73, ps > 0.182). Trials in which participants lifted their hand too early or reached Methods to the wrong picture were excluded from analyses (4.37% of all trials). Bootstrapped split-half reliability estimates Participants Ninety-four individuals (74.5% female, n = 70) in the four conditions (pull food, push food, pull objects, participated in this study at the University of Salzburg, Aus- push objects) were r = 0.98–0.99 (r = 0.99) for decision sb tria. Mean age was 23.4 years (SD = 4.74) and mean body time, r = 0.97–0.98 (r = 0.98–0.99) for grabbing time, sb mass index was 22.8 kg/m (SD = 4.30). Most participants and r = 0.98–0.99 (r = 0.99) for dragging time. Median sb had German (55.3%, n = 52) or Austrian (36.2%, n = 34) citi- reaction times were submitted to analyses of variance for zenship (8.51% other citizenship, n = 8). Mean food depriva- repeated measures with group (manikin at the bottom vs. tion was 3.11 h (SD = 2.94). manikin at the top vs. no manikin) as between-subjects AAT The t ask and apparatus were equal to Study 1, except factor, and trial type (pull vs. push) and stimulus (food vs. that one group of participants performed the AAT with a objects) as within-subjects factors. Approach bias scores manikin and the participant’s first name displayed at the bot- were calculated as in Study 1 and Study 2. Reliability esti- tom of the screen. Another group of participants performed mates using the difference- of-difference function of splithalf the AAT with the manikin and name displayed at the top of were r = 0.03 (r = 0.04) for the decision time approach bias sb the screen. A third group of participants performed the AAT score, r = 0.56 (r = 0.72) for the grabbing time approach sb without the manikin and name (i.e., the same task version 1 3 Psychological Research bias score, and r = 0.59 (r = 0.72) for the dragging time drag movements. Approach bias scores were not related to sb approach bias score. trait or state chocolate craving. Results Additional analyses Decision time A significant main effect of stimulus As the post hoc tests for following up the interaction effects [F = 10.9, p = 0.001, η = 0.107] indicated that partici- (1,91) p for grabbing time in Studies 1–3 were inconclusive, we pants reacted faster to food (M = 312  ms, SD = 168) than explored whether merging data across studies would pro- to objects (M = 347  ms, SD = 186). The three-way inter- vide a more clear-cut picture. For this, we merged grabbing action group × trial type × stimulus was also significant times of Study 2 and Study 3 (as grabbing time in Study 1 [F = 3.91, p = 0.023, η = 0.079]. Following up this inter- (2,91) p included decision time), leading to a combined sample size action by comparing approach bias scores between groups of n = 154. An analysis of variance for repeated measures revealed that approach bias scores were higher in the group again yielded a significant interaction trial type × stimulus with the manikin at the bottom (M = 6.06  ms, SD = 30.8) [F = 11.5, p = 0.001, η = 0.070]. Paired t tests indi- (1,153) p than in the group with the manikin at the top (M = − 14.2 ms, cated that grabbing food (M = 478 ms, SD = 199) was faster SD = 26.3; t = 2.81, p = 0.007). Approach bias scores in (61) than grabbing objects (M = 498 ms, SD = 219) in pull trials the group without a manikin did not significantly differ from [t = 2.09, p = 0.038]. Grabbing times for food and objects (153) the other two groups (both ts < 1.80, ps > 0.076). No other did not differ in push trials [t = 0.67, p = 0.504]. (153) effects were significant (all Fs < 3.69, ps > 0.057). Grabbing time As in Study 1 and Study 2, the interaction trial type × stimulus was significant [F = 5.38, p = 0.023, (1,91) Discussion η = 0.056] and, again, following up this interaction effect with paired t tests was inconclusive as grabbing times in pull The aim of the current studies was to develop a paradigm versus push trials were not significantly different for either for measuring approach–avoidance tendencies towards food stimulus category (both ts < 1.30, ps > 0.199) and did not dif- with arm movements on a touchscreen. Across all three stud- fer for food versus objects in either trial type (both ts < 1.46, ies, an approach tendency towards food (relative to non- ps > 0.149). Similar to Study 2, however, grabbing objects edible objects) was found when participants had to reach in pull trials was slightly slower than grabbing objects in towards the stimuli. Specifically, when stimuli were located push trials and this direction reversed for foods: they were distally—that is, when participants had to reach out to them grabbed faster in pull than in push trials at a descriptive level in preparation to move the stimuli towards them—there was (Fig. 3c). Thus, this crossed interaction again points to an a speeding of grabbing food compared to non-edible objects. approach bias towards food as reflected in grab movements. No such approach bias was found for the speed of dragging No other effects were significant (all Fs < 1.89, ps > 0.157). the stimuli towards or away from oneself. Thus, results differ Dragging time A significant main effect of stimulus from conventional tasks that measure approach or avoid- [F = 4.42, p = 0.038, η = 0.046] indicated that partici- (1,91) p ance biases by requiring participants to move a manikin or pants moved objects (M = 434 ms, SD = 121) faster than food the stimuli on a computer screen (de Houwer et al., 2001; (M = 453 ms, SD = 164). No other effects were significant Rinck & Becker, 2007). However, they are in line with find- (all Fs < 2.04, ps > 0.156). ings from a virtual reality study in which an approach bias Correlates of approach bias scores Decision time, grab- towards food was reflected in grasping movements towards bing time, and dragging time approach bias scores did not stimuli (Schroeder et al., 2016). correlate with trait chocolate craving, current chocolate crav- ing or hunger before or after the task (all rs between − 0.185 Decision time and 0.074, ps > 0.073). Due to the fine-grained measurement of a composite, multi- Conclusion stage, approach–avoidance behavior, our study series gives insights beyond demonstrating an approach bias towards Study 3 again replicated the trial type × stimulus interac- food. First, the present setup allowed for differentiating deci- tion for grabbing time and, thus, the approach bias towards sion time—that is, the time between stimulus onset and start food found in Study 1 and Study 2. Although manipulating of the hand movement (release of the start button)—from the self-reference did affect decision times, it did not change subsequent two movement stages grabbing and dragging. approach bias towards food as reflected in grab movements In contrast to the findings by Rotteveel and Phaf ( 2004), or the absence of approach bias towards food as reflected in those ‘planning times’ (implicated in our decision time) did 1 3 Psychological Research not carry a bias that would point to a facilitated prepara- internal reliability estimates were generally low for grab- tion of affectively compatible movements. As indicated by bing time and dragging time approach bias scores as well, stimulus type main effects, however, decision times were some were in the acceptable range (Parsons, Kruijt, & Fox, faster for foods compared to objects in Study 2 and Study 2018). Similarly, it appears that findings on the relationship 3 (irrespective of trial type). One reason for this might be a between approach biases and craving are rather inconclu- higher degree of attentional capture of appetitive food rela- sive. For example, either trait food craving (Brockmeyer tive to other stimuli (Carbine et al., 2018). Yet, as we did et  al., 2015), state food craving (Lender et al., 2018), or not measure attentional processes (e.g., eye movements) increases in craving during performing an AAT (Dickson, directly, differences in the physical characteristics of food Kavanagh, & MacLeod, 2016) have been reported to cor- and objects pictures—although those were well-matched— relate with an approach bias towards food. When looking cannot be fully excluded. Another reason may be ‘classifi- at the wider literature that include studies using alcohol-, cation speed’: because participants were instructed to react tobacco-, and cannabis-related AATs, relationships between to either food or non-edible objects, they had to categorize approach biases and craving have also been found incon- the pictures to identify the target stimuli and execute the sistently (Cousijn, Goudriaan, & Wiers, 2011; Schoenmak- required movement. Using such instructions typically leads ers, Wiers, & Field, 2008; Wiers et al., 2013, 2014). Thus, to faster response latencies to food versus neutral stimuli in future studies are needed that clarify whether approach bias simple reaction time tasks because the food category is more towards food can be found independent of food craving or specific than the more diverse category of neutral objects whether it rather relates to trait, state, or changes in (i.e., (Loeber, Grosshans, Herpertz, Kiefer, & Herpertz, 2013; cue-induced) food craving. Meule et al., 2014). Future directions Grabbing and dragging time Interpretation of results is limited to the stimuli and partici- When differentiating between decision and grabbing time, pant characteristics in the current study. That is, results may Study 2 and Study 3 converged in showing that approach be different for other stimulus categories (e.g., other appeti- bias towards food only emerged during grabbing motor tive stimuli such as savory foods or alcoholic beverages) and movements and not during dragging. Facilitated grabbing in other samples (e.g., clinical samples such as individuals on the background of comparable decision and dragging with eating disorders or obesity). Moreover, future stud- times suggest that stimulus–response compatibility effects ies need to examine whether approach bias towards food potentially driven by approach biases emerge during motoric as reflected in grab movements in our paradigm relates to control of grab movements. Future research might follow- actual consumption of these foods. up on this in other AAT implementations, for example, in An important next step would be to evaluate whether balance board or virtual reality studies (Eerland et al., 2012; modifying the current paradigm to a training (e.g., by con- Schroeder et al., 2016) or other setups (e.g., computer gam- sistently presenting food stimuli at the bottom and control ing, 3D navigation). Dragging was not modulated by stimu- stimuli at the top) results in decreased approach bias and lus or trial type, suggesting that ‘securing’ food from a distal intake of the avoided foods. Previous findings on modify - to a proximal position is not particularly biased, at least in ing behavior using approach–avoidance trainings have been our setup. Yet, the very high internal reliabilities suggest that ambiguous. For example, while joystick-based trainings there was not much variation in dragging times across trials. (e.g., repeatedly avoiding appetitive stimuli in terms of Likely, participants operated at maximal speed in all trials, push movements) have been found to reduce approach ten- which may have concealed stimulus and trial type effects. dencies towards appetitive stimuli, effects on actual intake are less consistent (Becker, Jostmann, & Holland, 2018; Correlates of approach bias scores Kakoschke, Kemps, & Tiggemann, 2017). That is, sev- eral studies did not find that an avoidance training reduced When examining correlates of approach biases, however, actual consumption of alcoholic beverages (Leeman et al., no consistent associations were obtained. Grabbing time 2018), soft drinks (Krishna & Eder, 2018), or high-calorie approach bias did not correlate with trait and state chocolate foods (Becker, Jostmann, Wiers, & Holland, 2015; Dickson craving and although dragging time approach bias did cor- et al., 2016; Ferentzi et al., 2018; Warschburger, Gmeiner, relate with these measures in Study 1, these associations did Morawietz, & Rinck, 2018). Among other explanations, one not replicate in Study 2 and Study 3. One reason for finding reason may be that these tasks do not involve naturalistic no or inconsistent correlations with approach bias scores approach and avoidance behaviors, which may hinder trans- may be their insufficient reliability. In the current studies, lation of training effects into real-world behavior. There- decision time approach bias scores were unreliable. While fore, future studies may examine whether touchscreen-based 1 3 Psychological Research of approach and avoidance reactions. Journal of Experi- approach–avoidance trainings may be more effective for mental Psychology: General, 137, 262–281. https ://doi. modifying eating behavior. Such trainings would then need org/10.1037/0096-3445.137.2.262. to be rigorously pitted against conventional techniques to Eerland, A., Guadalupe, T. M., Franken, I. H. A., & Zwaan, R. A. reveal the best practices to modify consumption behaviors (2012). Posture as index for approach-avoidance behavior. PLoS One, 7(2), e31291. https://doi.or g/10.1371/journal.pone.00312 91 . with approach–avoidance interventions. Ferentzi, H., Scheibner, H., Wiers, R., Becker, E. S., Lindenmeyer, J., Beisel, S., & Rinck, M. (2018). Retraining of automatic action ten- Acknowledgements Open access funding provided by Paris Lodron dencies in individuals with obesity: A randomized controlled trial. University of Salzburg. This work was supported by the European Appetite, 126, 66–72. https://doi.or g/10.1016/j.appet.2018.03.016 . Research Council (ERC) under the European Union’s Horizon 2020 Kakoschke, N., Kemps, E., & Tiggemann, M. (2015). Combined research and innovation program (ERC-StG-2014 639445 NewEat). effects of cognitive bias for food cues and poor inhibitory con- The authors would like to thank Elena Fischer, Gillian Kleveman, Doris trol on unhealthy food intake. Appetite, 87, 358–364. https ://doi. Olbrich, Julian Ornig, Alexandra Peitl, Karin Scheibenpflug, Judith org/10.1016/j.appet .2015.01.004. Schumilas, Magdalena Spitzbart, and Elisabeth Weingartner for col- Kakoschke, N., Kemps, E., & Tiggemann, M. (2017). Approach bias lecting the data. modic fi ation training and consumption: A review of the literature. Addictive Behaviors, 64, 21–28. https ://doi.org/10.1016/j.addbe Open Access This article is distributed under the terms of the Crea- h.2016.08.007. tive Commons Attribution 4.0 International License (http://creat iveco Krishna, A., & Eder, A. B. (2018). No effects of explicit approach- mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- avoidance training on immediate consumption of soft drinks. tion, and reproduction in any medium, provided you give appropriate Appetite, 130, 209–218. https ://doi.or g/10.1016/j.appe t credit to the original author(s) and the source, provide a link to the .2018.08.023. Creative Commons license, and indicate if changes were made. Laham, S. M., Kashima, Y., Dix, J., & Wheeler, M. (2015). A meta- analysis of the facilitation of arm flexion and extension move- ments as a function of stimulus valence. Cognition and Emotion, 29, 1069–1090. https ://doi.org/10.1080/02699 931.2014.96809 6. Leeman, R. F., Nogueira, C., Wiers, R. W., Cousijn, J., Serafini, K., References DeMartini, K. S., & O’Malley, S. S. (2018). A test of multisession automatic action tendency retraining to reduce alcohol consump- Becker, D., Jostmann, N. B., & Holland, R. W. (2018). Does approach tion among young adults in the context of a human laboratory bias modification really work in the eating domain? A commen- paradigm. Alcoholism: Clinical and Experimental Research, 42, tary on Kakoschke et al. (2017). Addictive Behaviors, 77, 293– 803–814. https ://doi.org/10.1111/acer.13613 . 294. https ://doi.org/10.1016/j.addbe h.2017.02.025. Lender, A., Meule, A., Rinck, M., Brockmeyer, T., & Blechert, J. Becker, D., Jostmann, N. B., Wiers, R. W., & Holland, R. W. (2015). (2018). Measurement of food-related approach–avoidance biases: Approach avoidance training in the eating domain: Testing the Larger biases when food stimuli are task relevant. Appetite, 125, effectiveness across three single session studies. Appetite, 85, 42–47. https ://doi.org/10.1016/j.appet .2018.01.032. 58–65. https ://doi.org/10.1016/j.appet .2014.11.017. Loeber, S., Grosshans, M., Herpertz, S., Kiefer, F., & Herpertz, S. Blechert, J., Meule, A., Busch, N. A., & Ohla, K. (2014). Food-pics: an C. (2013). Hunger modulates behavioral disinhibition and atten- image database for experimental research on eating and appetite. tion allocation to food-associated cues in normal-weight controls. Frontiers in Psychology, 5(617), 1–10. https ://doi.org/10.3389/ Appetite, 71, 32–39. https: //doi.org/10.1016/j.appet. 2013.07.008. fpsyg .2014.00617 . Maas, J., Keijsers, G. P. J., Rinck, M., Sharbanee, J. M., Vroling, Brockmeyer, T., Hahn, C., Reetz, C., Schmidt, U., & Friederich, H.-C. M. S., & Becker, E. S. (2017a). Implicit action tendencies and (2015). Approach bias and cue reactivity towards food in people evaluations in unwanted snacking behavior. International Jour- with high versus low levels of food craving. Appetite, 95, 197– nal of Cognitive Therapy, 10, 79–91. https ://doi.or g/10.1521/ 202. https ://doi.org/10.1016/j.appet .2015.07.013. ijct.2017.10.1.79. Carbine, K. A., Rodeback, R., Modersitzki, E., Miner, M., LeChemi- Maas, J., Keijsers, G. P. J., Rinck, M., Tanis, J., & Becker, E. S. (2015). nant, J. D., & Larson, M. J. (2018). The utility of event-related Does a dieting goal affect automatic cognitive processes and their potentials (ERPs) in understanding food-related cognition: A sys- trainability? Cognitive Therapy and Research, 39, 378–389. https tematic review and recommendations. Appetite, 128, 58–78. https ://doi.org/10.1007/s1060 8-014-9658-0. ://doi.org/10.1016/j.appet .2018.05.135. Maas, J., Woud, M. L., Keijsers, G. P. J., Rinck, M., Becker, E. S., Cousijn, J., Goudriaan, A. E., & Wiers, R. W. (2011). Reaching out & Wiers, R. W. (2017b). The attraction of sugar: an associa- towards cannabis: approach-bias in heavy cannabis users predicts tion between body mass index and impaired avoidance of sweet changes in cannabis use. Addiction, 106, 1667–1674. https ://doi. snacks. Journal of Experimental Psychopathology, 8, 40–54. https org/10.1111/j.1360-0443.2011.03475 .x. ://doi.org/10.5127/jep.05241 5. Cring, C. (2017). Approach and avoidance tendencies: Using smart- Machulska, A., Zlomuzica, A., Adolph, D., Rinck, M., & Margraf, phones to study behavioral responses and food desires. Master’s J. (2015). “A cigarette a day keeps the goodies away”: smokers thesis at Leiden University, Leiden. show automatic approach tendencies for smoking—but not for de Houwer, J., Crombez, G., Baeyens, F., & Hermans, D. (2001). On food-related stimuli. PLoS One, 10(2), e0116464. ht t ps : // d oi . the generality of the affective Simon effect. Cognition and Emo- org/10.1371/journ al.pone.01164 64. tion, 15, 189–206. https ://doi.org/10.1080/02699 93012 5883. Meule, A., & Hormes, J. M. (2015). Chocolate versions of the Food Dickson, H., Kavanagh, D. J., & MacLeod, C. (2016). The pulling Cravings Questionnaires. Associations with chocolate expo- power of chocolate: Effects of approach–avoidance training on sure-induced salivary flow and ad libitum chocolate consump- approach bias and consumption. Appetite, 99, 46–51. https ://doi. tion. Appetite, 91, 256–265. h t t p s : / / d o i . o r g /1 0 . 1 0 1 6 / j . a p p e t org/10.1016/j.appet .2015.12.026. .2015.04.054. Eder, A. B., & Rothermund, K. (2008). When do motor behav- Meule, A., Lender, A., Richard, A., Dinic, R., & Blechert, J. (2019). iors (mis)match affective stimuli? An evaluative coding view Approach–avoidance tendencies towards food: Measurement on 1 3 Psychological Research a touchscreen and the role of attention and food craving. Appetite, Schroeder, P. A., Lohmann, J., Butz, M. V., & Plewnia, C. (2016). 137, 145–151. https ://doi.org/10.1016/j.appet .2019.03.002. Behavioral bias for food reflected in hand movements: a prelimi- Meule, A., Lutz, A., Krawietz, V., Stützer, J., Vögele, C., & Kübler, nary study with healthy subjects. Cyberpsychology, Behavior, A. (2014). Food-cue affected motor response inhibition and self- and Social Networking, 19, 120–126. https ://doi.org/10.1089/ reported dieting success: a pictorial affective shifting task. Fron -cyber .2015.0311. tiers in Psychology, 5(216), 1–17. https ://doi.org/10.3389/fpsyg Seibt, B., Neumann, R., Nussinson, R., & Strack, F. (2008). Movement .2014.00216. direction or change in distance? Self- and object-related approach– Parsons, S. (2018). splithalf: robust estimates of split half reliability. avoidance motions. Journal of Experimental Social Psychology, figshare. https ://doi.org/10.6084/m9.figsh are.55591 75.v3. 44, 713–720. https ://doi.org/10.1016/j.jesp.2007.04.013. Parsons, S., Kruijt, A.-W., & Fox, E. (2018). Psychological Science Shen, H., Zhang, M., & Krishna, A. (2016). Computer interfaces and needs a standard practice of reporting the reliability of cognitive the “direct-touch” effect: Can iPads increase the choice of hedonic behavioural measurements. PsyArXiv. https ://doi.org/10.31234 / food? Journal of Marketing Research, 53, 745–758. https ://doi. osf.io/6ka9z . org/10.1509/jmr.14.0563. Paslakis, G., Kühn, S., Grunert, S., & Erim, Y. (2017). Explicit and Warschburger, P., Gmeiner, M., Morawietz, M., & Rinck, M. (2018). implicit approach vs avoidance tendencies towards high vs low Evaluation of an approach–avoidance training intervention for calorie food cues in patients with obesity and active binge eating children and adolescents with obesity: A randomized placebo- disorder. Nutrients, 9(10), 1–16. https ://doi.org/10.3390/nu910 controlled prospective trial. European Eating Disorders Review, 1068. 26, 472–482. https ://doi.org/10.1002/erv.2607. Paslakis, G., Kühn, S., Schaubschläger, A., Schieber, K., Röder, K., Wiers, C. E., Kühn, S., Javadi, A. H., Korucuoglu, O., Wiers, R. W., Rauh, E., & Erim, Y. (2016). Explicit and implicit approach vs. Walter, H., & Bermpohl, F. (2013). Automatic approach bias avoidance tendencies towards high vs. low calorie food cues in towards smoking cues is present in smokers but not in ex-smok- patients with anorexia nervosa and healthy controls. Appetite, 107, ers. Psychopharmacology, 229, 187–197. https://doi.or g/10.1007/ 171–179. https ://doi.org/10.1016/j.appet .2016.08.001.s0021 3-013-3098-5. Phaf, R. H., Mohr, S. E., Rotteveel, M., & Wicherts, J. M. (2014). Wiers, C. E., Stelzel, C., Park, S. Q., Gawron, C. K., Ludwig, V. U., Approach, avoidance, and affect: a meta-analysis of approach- Gutwinski, S., & Bermpohl, F. (2014). Neural correlates of alco- avoidance tendencies in manual reaction time tasks. Frontiers hol-approach bias in alcohol addiction: the spirit is willing but the in Psychology, 5(378), 1–16. h ttp s : //d oi. or g/ 10. 3389 /fp sy g flesh is weak for spirits. Neuropsychopharmacology, 39, 688–697. .2014.00378. https ://doi.org/10.1038/npp.2013.252. Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear Zech, H. G. (2015). The Mobile Approach–Avoidance Task. Master’s of spiders. Journal of Behavior Therapy and Experimental Psy- thesis at Leiden University, Leiden. chiatry, 38, 105–120. https://doi.or g/10.1016/j.jbtep.2006.10.001 . Rotteveel, M., & Phaf, R. H. (2004). Automatic affective evaluation Publisher’s Note Springer Nature remains neutral with regard to does not automatically predispose for arm flexion and extension. jurisdictional claims in published maps and institutional affiliations. Emotion, 4, 156–172. https://doi.or g/10.1037/1528-3542.4.2.156. Schoenmakers, T., Wiers, R. W., & Field, M. (2008). Effects of a low dose of alcohol on cognitive biases and craving in heavy drink- ers. Psychopharmacology, 197, 169–178. https://doi.or g/10.1007/ s0021 3-007-1023-5. 1 3

Journal

Psychological ResearchSpringer Journals

Published: May 4, 2019

There are no references for this article.