Driven to snack: simulated driving increases subsequent consumption

When individuals eat while distracted, they may compensate by consuming more a  terwards. Here, we examined the e  ect o  eating while driving, and explored potential underlying mechanisms. Participants ( N = 116, 73.3%  emale) were randomly allocated to complete a driving simulation (distraction condition) or to watch someone else drive (control condition) while consuming 10g (50.8 kcal) o  potato chips. A  terwards, participants rated the taste intensity and hedonic experience, reported stress levels, and were then given the opportunity to eat more chips. As hypothesized, participants consumed more chips a  ter the driving simulation. Stress levels were higher in the driving compared to control condition, but were inversely related to consumption amount, ruling out stress as explanatory mechanism. Saltiness ratings di  ered between the driving and passive viewing condition, only when controlling  or stress. The current  ndings converge with earlier work showing that distracted eating can drive overconsumption, which in turn can lead to long-term health implications. Limitations, implications, and potential directions are discussed

quently exposed to situational stimuli likely to distract them rom the sensory experience o consumption.Pre-packaged oods and drinks are increasingly popular and are oten consumed while people engage in other activities such as listening to music, using their mobile devices, or commuting.Especially when consumption takes place under cognitively taxing conditions, such as while driving, behind a computer, or while looking at one's smartphone, this practice is likely to increase the amount consumed.For example, children and adults that watch television during their meals have been ound to consume more ood during (Blass et al., 2006;Crespo et al., 2001;Dubois et al., 2008), and ollowing (Higgs & Woodward, 2009) the consumption occasion.Likewise, eating while driving, and while listening to loud music, has been ound to be less eective in reducing people's desire to eat (Ogden et al., 2016), and to promote overcon-sumption (Spence & Shankar, 2010;Staord & Dodd, 2013;Stroebele & de Castro, 2006).Finally, a meta-analysis examining the eect o distraction during consumption on the amount o ood consumed revealed a positive association between these actors (however, one study included in this analysis may have biased the overall eect size Robinson et al., 2013).
Several mechanisms have been proposed to explain why distracted eating promotes overconsumption, such as reduced awareness o the amount consumed and reduced memory o ood intake (Robinson et al., 2013;Oldham-Cooper et al., 2011) and compensatory responses to stress (Reichenberger et al., 2018;Torres & Nowson, 2007).There is also growing evidence to suggest that the positive link between distraction and consumption amount may be explained by reduced taste perception.For instance, a number o experiments have demonstrated that distraction reduces the taste or odor intensity o sweet, sour, and

Take-home Message
In this study, people consumed more potato chips ater eating chips while completing a driving simulation than in a control condition.We had hypothesized that this was due to lowered perceived taste intensity o the potato chips eaten while distracted, but this was only the case when we controlled or stress.Diferences in perceived stress did not explain the diferences in subsequent consumption amount between the conditions.bitter solutions, and salty snacks (Homann-Hensel et al., 2017;Liang et al., 2018;van der Wal & van Dillen, 2013), and even promotes increased consumption (Morris et al., 2020;van der Wal & van Dillen, 2013).Participants who were distracted by a working memory task while preparing lemonade at their preerred concentration opted or greater amounts o syrup and consumed more salty buttered crackers than participants who experienced minimal distraction (van der Wal & van Dillen, 2013).Additionally, compared to mildly distracted participants, highly distracted participants exhibited reduced neural taste processing during tasting, while they consumed more during a subsequent ad libitum ood test (Dui et al., 2020).More generally, several recent studies have pointed to the importance o sensory perception, in particular taste intensity, or expectations o ullness and later portion selection (as reviewed in Forde, 2018).Furthermore, salt intensity predicted ad libitum intake, even when the oods were equally liked (Bolhuis et al., 2012).However, to our knowledge, no studies have examined both the eect o distraction on perceived taste intensity and palatability o the ood consumed and how this inuences later consumption.Furthermore, previous studies on distracted tasting have used distractions that were either not very ecologically valid (e.g.working memory task van der Wal & van Dillen, 2013; Dui et al., 2020;Liang et al., 2018) or not very cognitively demanding (e.g.listening to music, Stroebele & de Castro, 2006).Accordingly, the aim o the present study is to investigate the proposed eect using a more ecologically valid distractor -to examine whether eating while driving promotes increased consumption, and whether this eect is explained by reduced taste intensity.
Increased stress levels may provide an alternative explanation or the eect o distracted consumption on increased consumption.That is, it is plausible that driving may imbue stress (Antoun et al., 2017).For example, participants completing a driving simulation were more stressed when driving themselves than when the simulation was o a sel-driving car, evidenced by a higher skin potential response and heart rate (Zontone et al., 2020).Elevated stress levels have been linked to both increased and reduced ood intake (Reichenberger et al., 2018;Torres & Nowson, 2007).For instance, ego threat leads to increased snack intake in one study (Wallis & Hetherington, 2004) but lower snack intake in another (Wallis & Hetherington, 2009), depending on the type o snacks oered and restrained and emotional eating style.Another actor that may inuence whether stress has a positive or negative eect on ood intake may be the severity o the stress (Torres & Nowson, 2007).Thus, we additionally examined the potential role o stress in compensatory consumption ollowing distracted eating (snacking while driving).
Societal and technological developments have increased the requency in which oods (particularly high-calorie snacks Hirschberg et al., 2016) are consumed while driving (Food-Shopper Monitor, 2018;Stutts et al., 2005), thus making this an ideal and realistic scenario in which to test this eect.Furthermore, although multiple studies have ound that eating while driving negatively inuences driving perormance (Dingus et al., 2016;Irwin et al., 2014;Young et al., 2008), the reverse question o whether driving inuences eating has so ar not been addressed.
The driving context was chosen to be demanding so as to require attention (rather than just routine), and to be representative o everyday demanding driving contexts (e.g., driving on an unamiliar road, or city trafc during rush hour).We expected that driving would thus induce stress, and mental load.As a result o this higher demand, we hypothesized that driving, relative to control (passive viewing), decreases the perceived taste intensity o salty potato chips.At the same time we expected that it would lead to greater chip consumption aterwards.As noted, we were less certain about the role o stress in this mechanism, as previous research has observed both increased and decreased consumption ollowing stress.Thereore, we examined the possibility o both a positive and a negative relationship between stress and subsequent consumption.Furthermore, we also explored the eects o distraction on the hedonic aspects o taste.We hypothesized that distraction decreases perceived taste intensity but may not aect hedonic ratings, as the hedonic value o consumption varies greatly between individuals but is stable within individuals and as this has not been consistently linked with actual consumption (DiFeliceantonio et al., 2018;McCrickerd & Forde, 2015;Tang et al., 2014).Thereore, we did not think it likely that the hypothesized eect o distraction on consumption could be explained by changes in hedonic ratings.Moreover, we explored whether participants' driving experience was a potential moderator o our proposed eects o distraction on taste perception and consumption since this may a-ect how demanding and stressul the driving manipulation was or each participant.Finally, since some previous studies have ound that the eects o distraction on consumption vary with individual dierences in restrained eating (Boon et al., 2002;Ogden et al., 2016), this was included as a control variable.

Participants and Design
One hundred nineteen English speaking Leiden University students in possession o a driver's license (car) participated in exchange or course credit or money (€3.50) and were randomly assigned to a simulated driving or control condition.Smoking or having allergies were exclusion criteria.Participants were requested not to eat and to only drink water two hours prior to the start o the study.O the sample, three cases were excluded because they ell outside the proposed age range o 18-30 years (ages 45 and 60 years, > 3 SDs rom the mean; or one participant age was not known).An additional three participants initiated but did not complete the study and were thereore also excluded rom urther analyses.Repeating the analyses including these participants did not change the results.The remaining sample or analyses thus consisted o 116 participants (30 men, mean age 22.30 years, SD = 4.98 years) evenly distributed over the two conditions (n = 58 each).The two groups did not dier on the number o men and women, age, nationality, or total Restrained Eating Score (see Supplemental Table 4).
The main dependent variables were taste intensity o the potato chips and the amount o calories consumed.In addition, stress levels and hedonic ratings were considered.Individual dierences in driving experience and restrained eating were examined as potential moderators.The research questions and procedure were approved by the ethical committee o the Leiden University Psychology Institute (CEP19-0301/146).All procedures per-ormed were in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.Inormed consent was obtained rom all individual participants involved in the study.The overall design, research question and hypotheses were specied in the ethics proposal prior to the start o data collection.The ethics proposal, raw data and analysis script can be obtained rom: os.io/twg9r/.

Procedure
Beore engaging in the experiment, participants were seated behind a desk with a laptop on which a short introductory text was displayed that inormed them that the study was about multitasking while driving.Ater providing inormed consent, participants next reported their driving experience.Following this, they were randomly assigned to the driving or control condition (see Driving simulation or details), and asked to sit in the driver's seat o the simulator where they were provided instructions or the driving simulation.Participants were then provided with a bowl o potato chips (10 grams, 50.8 kcal) and instructed to consume them all during the driving simulation.The chips used were Lays Classic salted potato chips.All participants consumed the entire 10 grams/50.8kcal.Participants then completed the driving simulation.Following  the driving simulation, participants returned to the desk to report their ratings, stress levels, age, sex, and ethnicity on the laptop.Participants were then instructed to wait in the room while the experimenter collected debrieng orms rom the adjacent room (wait time held constant at three minutes), and that i they wanted, they could eat the rest o the potato chips (the remaining 15 g (76.2 kcal) rom the 25 g party bag, in a bowl on the same desk).The Netherlands Nutrition Centre states 30 g as the average portion size o chips in the Netherlands (Voedingscentrum, n.d.).Participants were told that these potato chips were let over rom the party bag and that they were ree to consume them all.Finally, all participants were debrieed, thanked, and compensated or their participation.

Driving Simulation
To create a realistic and demanding driving context, a set-up was built that consisted o a chair, steering wheel, pedals, and a 23-inch at screen (see Fig. 1).A PlayStation 3 and the game Gran Turismo (Yamauchi, 2013) were used to simulate a realistic driving experience.Participants were seated in the driving chair and it was explained how they could speed up, break and steer.Participants were asked to drive three laps on the Twin Ring Motegi course that consisted o two straight sections, a large bend and 2 sharp bends.Participants were told they should drive as well as they could.Driving the three laps took three minutes on average.I a participant took longer than 10 minutes to complete the laps the simulation was stopped, however, none o the participants took longer than 10 minutes to drive the three laps.To create a comparable situation in the control condition, the same driving simulator was used.The participants in the control group acted as co-driver/passenger and did not actually drive themselves.Instead, a three-minute recorded video was played, showing the exact same three laps o the Twin Ring Motegi course that the participants drove in the experimental driving condition.

Driving Experience
Three questions addressed participants' driving experience: 'How many years do you have your driver's license?', 'How oten do you drive on average per week?' and 'How many kilometers did you cover on average in the last year?'.The three items were answered on vepoint Likert scales.These included respectively, driving years ranging rom 1 (up to 1 year), increasing with each scale point with 1 year to a maximum o 5 (over 7 years); driving requency ranging rom 1 (once a week) increasing with each scale point with one time per week to a maximum o 5 (7 times per week); and driving distance ranging rom 1 (1,000km per year) increasing with each scale point with 1,000km per year to a maximum o 5 (7,000 km per year).

Taste Intensity
Participants rated the potato chips on three items relating to taste intensity, namely 'saltiness', 'sourness', and 'sweetness', on sevenpoint Likert scales ranging rom 1 (not at all) to 7 (very).Sweetness and sourness ratings were included as catch trials, to establish that participants were not merely guessing when assessing the potato chips' avor.Furthermore, the sweetness and sourness ratings serve as a baseline measure since we do not expect them to dier between conditions.

Hedonic Rating
Participants next rated the potato chips on three more items relating to hedonic experience, namely 'quality', 'tastiness', and 'crunchiness', on the same seven-point Likert scales ranging rom 1 (not at all) to 7 (very).

Stress levels
Participants were asked ve questions pertaining to their experiences o stress during the simulation: 'How relaxed were you during the driving simulation?' (reversed), 'How much did you have the eeling that you were in control during the driving simulation?' (reversed), 'How rushed did you eel during the driving simulation?', 'How nervous were you during the driving simulation?', and 'How well did you perorm during the driving simulation?' (reversed).All questions were answered on a six-point Likert scale ranging rom 1 (not at all) to 6 (very).

Calories Consumed
The number o calories consumed was determined by weighing the bowl with the remaining chips once the participant had let and subtracting this rom its initial weight.The weight in gram was then multiplied by the amount o kcal/g (5.08).

Data Preparation
All data preparation steps and statistical analyses were perormed in R (R Core Team, 2019) and can be retrieved rom the OSF page: os.io/twg9r/.Distribution o the variables was examined by visual inspection, Shapiro-Wilks test and Levene's test.Since some o the variables were skewed, robust regression using the rlm unction o the R package MASS was used throughout or consistency.
Robust regression was used to examine the dierences between conditions unless otherwise specied (see Results).For each dependent variable (taste intensity, hedonic rating and number o calories consumed) we rst estimated a ull actorial model that included main eects and interaction o the experimental actor (Driving, Control) and Driving Experience.I the interaction term was not statistically signiicant, subsequently models with only the main eects o the experimental actor and Driving Experience were estimated.Aterwards, we calculated the Bayes Inormation Criterion (BIC) in order to see which model perormed best.
Reliability analysis revealed saltiness was poorly associated with sourness and sweetness (Cronbach's α = 0.31), as expected, and so these ratings were thereore examined separately.The items assessing hedonic rating and stress showed adequate reliability (Cronbach's alphas o respectively .69 and .79)and were thereore averaged into two overall scores.The three items that assessed driving experience were only moderately associated (Cronbach's α = 0.54), but driving distance correlated signicantly with both requency (r = 0.48) and years o license (r = 0.34), with the latter two being uncorrelated (r = 0.06).Even though each item thus seemed to tap into a somewhat di-erent aspect o driving experience, they were nevertheless averaged to orm a broad index o driving experience.
Subsequent t-test analyses conrmed that driving experience in years, requency and distance did not vary across conditions (ts < 1.42, ps > 0.153, so that these could be incorporated as moderator variables into the regression models or taste ratings and consumption.Table 1 depicts the raw means and standard deviations o the three Driving Experience items (years, requency and distance) as a unction o condition.Since driving experience was highly skewed towards the lower end (see Figure 3a in the Supplement section), quartile scores were used in the analyses.

Efects o Driving on Calories consumed
Table 2 and Figure 2 depict the mean and standard deviation/error or potato chips consumed in kcal during the driving manipulation and the ollow-up ree consumption test as a unction o condition (Driving; Control).Inspection o the histograms revealed that the number o calories consumed was not normally distributed, but had a bimodal distribution with many observations at the scale extremes (50.8, 127.0; see supplemental Fig. S6.b or histograms per condition).More specically, during the ree consumption period 52% o participants consumed no chips and 20% o participants consumed all o the chips.Given how many participants consumed the maximum amount o chips available, it is likely that the mean consumption amount would have been higher i it had not been restricted (i.e., censoring eect is likely).There-ore, we applied Tobit regression analyses1 (i.e., censored regression models Tobin, 1958), using the R package censReg (Henningsen, 2010).
This Tobit regression analysis with calories consumed as dependent variable and main eects and interaction o the experimental actor (Driving, Control) and Driving Experience showed no signicant interaction term, so a model with only main eects was estimated.There was a signicant main eect o the driving manipulation, b = -19.43,p = 0.026.As hypothesized, participants consumed more potato chips when driving (M = 84.3kcal, SE = 4.11), compared to passively watching the same route (M = 72.9kcal, SE = 3.24).Driving Experience did not have a main eect.The BIC or the model with only main eects was lower than the model with the interaction term (BIC 3.15).

Efects o Driving on Taste Intensity
Table 2 and Figure 2 depict the means and standard deviations or all taste intensity ratings as a unction o condition (Driving; Control).
Robust regression analyses incorporating main eects and interaction o the experimental actor (Driving, Control) and Driving Experience were conducted to examine the eects on saltiness ratings.The BIC or the model with only main eects was lower than the model with the interaction term (BIC 2.67).Contrary to our rst hypothesis, we did not observe a signicant main eect o condition, b = 0.42, SE = .23,t(111) = 1.84, p = 0.067.As predicted, participants rated the potato chips as less salty when they were driving themselves (M = 4.43, SE = 0.16), than when they were attending a recording o the same route being driven by someone else (M = 4.74, SE = 0.15), but this dierence did not reach the threshold or sig-nicance.Control analyses conrmed that the driving manipulation likewise did not signicantly impact participants' sourness and sweetness ratings (ts < 0.3, ps > .54)with very similar intensity ratings across conditions (see Fig. 2).The potato chips were generally perceived to be minimally sweet (M = 2.01, SD = 1.15) and sour (M = 1.72,SD = 0.96).Taken together, even though the intensity ratings showed the expected pattern, we ound no robust proo that driving interered with participants' processing o the saltiness o the chips.
There was no main eect or interaction e-ect o Driving Experience. Figure 2 Mean o taste intensity ratings, hedonic ratings and total calories o chips consumed per condition."Hedonic rating" here reects the mean o the "quality", "crunchy" and "tasty" ratings.Total amount o calories includes the standard amount o 50.8 kcal o chips eaten during the manipulation, as indicated by the horizontal black line.Error bars reect standard error.

Explorations o Hedonic Rating and Drivinginduced Stress as Alternative Explanation
We also explored whether driving altered hedonic aspects o the consumption experience.
A robust regression model with hedonic rating as dependent variable and main eects and interaction o the experimental actor (Driving, Control) and Driving Experience was estimated.Table 2 depicts the means and standard deviations or the three hedonic ratings as a unction o condition (Driving; Control).These showed that participants rated hedonic aspects no di-erent in the driving condition (M = 5.08, SE = 0.41) than the control condition (M = 4.95, SE = 0.42, b = -0.18,p = 0.782).When the items were analyzed separately, this did not yield any sig-nicant dierences either (ps>0.356).Finally, driving experience did not signicantly impact hedonic rating (p=0.344)nor was there a signiicant interaction between Driving Experience and condition on hedonic rating (p = 0.125).
We next examined whether the eects o driving on perception and consumption resulted rom driving-induced stress as opposed to distraction.Table 3 depicts the means and standard deviations or the ve stress ratings as a unction o condition (Driving; Control).These revealed that all ve items were aected by the driving manipulation; participants were signicantly less relaxed, and elt signicantly more in control, rushed, nervous, and per-orming well while driving than while in the passive viewing condition, ts>2.93,ps<0.020.This conrms that driving compared to passive viewing heightened participants' stress levels.There was no interaction between Driving Experience and the driving manipulation on perceived stress levels (p = 0.17).
To test whether the eect o condition on the amount o ood consumed could be explained by the dierence in experienced stress, a Tobit regression analysis was perormed with calories consumed as the dependent variable and the main eects and interactions o the experimental actor (Driving, Control), stress, and Driving Experience.Since the ull actorial did not show a signicant eect o the interaction term with Driving Experience, subsequently a model was estimated with calories consumed as the dependent variable and  the main eects and interactions o the experimental actor (Driving, Control) and stress and only a main eect or Driving Experience (BIC 9.43).The analysis showed a main eect or both conditions, b = -25.82,SE = 32.14, t =-2.60, p = 0.009, and stress, b = -15.83,SE = 7.60, t = -2.08,p = 0.037.Interestingly, the e-ect o stress on consumption was negative, which means that participants who elt more stressed ate less.This suggests that increases in stress did not explain increased consumption ollowing driving.Additionally, there was a signicant interaction eect o driving manipulation and stress on calories consumed: b = 18.22,SE = 9.29, t = 1.96, p = 0.038.Although there was no signicant main eect o stress on consumption when the analyses were done in the respective conditions, a Fischer r to z comparison conrmed that the slopes o the eect o stress on calories consumed in the driving and control condition were dierent, Z = 2.00, p = 0.05.In the driving condition stress had a stronger negative eect on calories consumed (r = -0.37)compared to the control condition (r = -0.12).There was no signicant main eect o stress on consumption when the analyses were done in the respective conditions.When added as covariate to the overall regression model, stress did not explain the main eect o driving on calories consumed.
In conclusion, stress had a negative eect on consumption.So, even though participants elt more stressed in the driving condition, stress did not account or the dierence in calories consumed between the driving and control condition.
There was no eect o stress on saltiness ratings or any interactions between stress and condition or driving experience on saltiness ratings (all ps >0.35).However, when stress was entered into the model, the eect o con-dition on saltiness ratings became signicant (b = 0.54, SE = 0.26, t(111) = 2.035, p = 0.0454). 2 In order to examine the relationships between all actors o interest and to take into account that driving experience and stress are assessed by multiple items, we used structural equation modeling (SEM) to estimate path models using the lavaan package in R (Rosseel, 2012).Figure 6 in the supplementary section depicts the model that was tested.Calories consumed was the dependent variable, driving condition was included as a predictor and taste perception (saltiness ratings) and stress were assessed as possible mediators between driving condition and calories consumed.Stress and driving experience were modeled as latent variables.Furthermore, gender was added to the model as a control variable.Since lavaan does not support interactions with latent variables, no interactions were modeled.Model t indices show a poor t or the model (CFI = 0.69; SRMR = 0.12; RMSEA = 0.14 (90% CI: 0.12 to 0.17).The model showed a signicant eect o driving condition on calories consumed (b = 13.62,SE = 6.14, p = 0.027) but did not indicate stress or saltiness ratings as a mediator o this eect via a direct or indirect path (Figure 6; analysis script on OSF).

General Discussion
In this study we aimed to build on previous studies that ound that distraction increased 2 Restrained Eating as measured by the Restrained Eating Scale (Polivy et al., 1978) was examined as a potential moderator as well.There was no dierence in Restrained Eating between the driving (M=13.0;SD=5.82) and control condition (M = 11.9;SD = 4.76 consumption by examining possible explanations o the eect in a practically relevant setting.To do so, in a simulated driving experiment, we examined whether snacking while driving would result in greater consumption aterwards.We urthermore investigated whether this eect could be explained by reduced taste intensity while driving or by drivinginduced stress.
In support o our predictions, participants who engaged in the driving simulation while consuming potato chips, consumed more potato chips during a ollow-up ree consumption test than participants who merely watched a recording.There were some indications that the driving simulation lowered the saltiness ratings o the chips.Other sensory ratings and hedonic ratings were unaected by driving.
Many dierent (complementary) explanations have been proposed or the mechanism which makes people consume more ater or during distracted consumption, including reduced memory or ood intake or health goals, disrupted inuence o satiation, and dishabituation (Forde, 2018;Robinson et al., 2013).In the current study, we ound some indications that lowered taste perception may be an interesting component to consider when studying the mechanism behind overconsumption ater distracted eating.
Our nding that distraction may reduce perceptions o saltiness supports previous literature demonstrating this eect (van der Wal & van Dillen, 2013; Liang et al., 2018;Dui et al., 2020).As taste intensity has been ound to correlate negatively with ood intake (Forde

Original Purpose
In this study, we aimed to examine the efect o eating while driving, and potential underlying mechanisms.We hypothesized that eating while driving would reduce taste perception, which would in turn cause participants to overconsume aterwards to compensate.Based on previous research, we expected that taste perception, but not hedonic preerence, would be diminished during distracted consumption.
We urthermore wanted to examine the efect o stress experienced during the driving simulation as an alternative explanation.
et al., 2013), lowered experienced taste intensity during distracted eating may lead to increased ood consumption.Furthermore, when distracted, taste inormation may not be processed in a way that leads to satisaction or satiation.For example, consuming a high calorie drink under high perceptual load led to lower satiety than when the same drink was consumed under low perceptual load (Morris et al., 2020).Future studies could examine the eect o distracted consumption on satiation/satiety and how this relates to taste perception and other outcomes.
Perceived stress was examined as an alternative explanation o the eect o distracted consumption on subsequent consumption.Whereas participants reported more stress ater driving than ater watching someone else drive, sel-reported stress yielded an opposite eect on consumption, with participants consuming ewer rather than more potato chips.The phenomenon that acute stress can reduce ood intake has been attributed to physiologic changes that occur ater acute stress and that might be expected to temporarily reduce ood intake, e.g., slowed gastric emptying and shiting o blood rom the gastrointestinal tract to muscles (Torres & Nowson, 2007).
Several previous studies have ound an e-ect o restrained eating on the relationship between stress and consumption (Wallis & Hetherington, 2004;Wallis & Hetherington, 2009).However, we did not nd an eect o restrained eating on consumption or any interaction between restrained eating, stress or driving manipulation.This could possibly be explained by the act that restrained eating scores in our sample were low.
In conclusion, higher perceived stress was associated with lower consumption o potato chips.Thereore, the nding that the driving manipulation increased intake could be not accounted or by the driving induced stress.
A strength o the current study is the use o a realistic and practically relevant distractor and consumption situation.This study aimed or a control condition that matched the sensory input during driving and thus only diered rom the experimental condition in the mental load and stress induced.As a result, the participants in the control condition were probably still somewhat distracted and this might have created a conservative test o our hypotheses.
However, this way, any dierences in consumption and perceived taste intensity between the conditions could be attributed to dierences in the availability o mental capacity.It is possible that the smaller eect sizes have caused our study to be underpowered to detect the eect o driving condition on saltiness ratings.Future research could examine variations in mental load and stress urther by comparing dierent levels o distraction during consumption, e.g., high distraction, low distraction, no distraction and targeted attention through mindul eating instructions in a larger sample.
The current study also has its limitations.Whereas standardization o the consumption amount during the driving manipulation allowed us to examine dierences in compensatory consumption, one limitation o the study is the limited amount that could be consumed later.The mean dierence in the amount o potato chips consumed ater driving or passively watching was only 11 kcal.However, these 11 kcal were consumed in addition to the 50.8 kcal that participants already ate during the driving distraction or passively viewing.In addition, a substantial proportion o the sample consumed the entire additional 15 grams or 76 kcal, which might indicate that they would have consumed more had they had the opportunity to do so.To urther examine the magnitude and practical relevance o the compensatory consumption eect, uture studies could examine ad libitum intake ollowing distracted eating.
Although participants were requested not to eat in the two hours prior to the start o the study, subjective hunger was not assessed.However, since participants were randomly assigned to the driving or control condition, possible variability in hunger status is unlikely to have caused the dierence in subsequent consumption between conditions.
We did not assess how much experience with playing video games participants had.In addition to driving experience, this may have aected how challenging the driving simulation was or participants.
Lastly, the relatively young age, low driving experience, high education level and unbalanced gender ratio o our sample limits the generalizability o our results.Furthermore, ethnicity was not assessed.Future studies could extend our ndings in broader samples that are more representative o the general population.

Conclusion
Using a realistic but lab-controlled driving simulation, the ndings reported here provide additional support or the notion that distracting consumption settings may have long-term health implications, through their contribution to overconsumption o unhealthy products.This pushes the need or a better understanding o what these settings look like in people's daily lives and how consumption settings can be changed.The current research provides some preliminary evidence that taste perception, and especially perceived taste intensity, may be a relevant aspect to consider when examining the mechanism through which distracted eating leads to overconsumption.

Figure 1
Figure1The set-up o the driving simulator used in both the experimental driving and passive viewing control conditions.It consisted o a chair, steering wheel, pedals and a 23-inch at screen.A PlayStation 3 and the game Gran Turismo(Yamauchi, 2013) were used to simulate the actual driving experience.Participants drove (or viewed a recorded video o) three laps on the Twin Motegi Course.

Figure 4
Figure 4 Histograms showing the distributions per condition (Driving and Control) or a. mean Driving Experience, b. mean stress score.

Figure 5
Figure 5 Histograms showing the distributions per condition (Driving and Control) or a. salt intensity rating, b. total amount o calories consumed.

Figure 6
Figure 6Structural Equation Modeling path model.Drive_dist = average kms driven in last year; Drive_req = weekly driving requency; Drive_years = years o having drivers' license; Drive_exp_l = latent variable o driving experience; Gendern = gender, male (0) or emale (1); Salty = how salty the chips were perceived during the experiment; Drive_Condition = experimental condition, either completing a driving simulation (1) or control condition (0); Calories_consumed = the amount o calories consumed ater the driving manipulation; stress_l = latent variable or stress experienced during the driving manipulation; Relaxed = how relaxed participants elt during the driving manipulation; Control = how in control participants elt during the driving manipulation; Rushed = how rushed participants elt during the driving manipulation; Nervous = how nervous participants elt during the driving manipulation; Perormance = how well participants elt they perormed during the driving manipulation.

Table 1
Means and standard deviations o driving experience (in years, distance and requency) as a unction o condition (driving; control).

Table 2
Means and standard deviations o the various taste ratings (1 -not at all to 7 -very) and amount consumed in kcal as a unction o condition (driving; control).

Table 3
Means and standard deviations (between brackets) o the various stress ratings (1-not at all to 7 -very) as a unction o condition (driving; control).
). Restrained Eating did not interact with any o the variables o interest.Adding total Restrained Eating score as a covariate did not change the overall pattern o results, see: os.io/twg9r/.