This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Serious Games, is properly cited. The complete bibliographic information, a link to the original publication on http://games.jmir.org, as well as this copyright and license information must be included.
Obesity is a growing global issue that is linked to cognitive and psychological deficits.
This preliminary study investigated the efficacy of training to improve inhibitory control (IC), a process linked to overeating, on consumption and cognitive control factors.
This study utilized a multisession mobile phone–based intervention to train IC in an overweight and obese population using a randomized waitlist-control design. A combination of self-assessment questionnaires and psychophysiological measures was used to assess the efficacy of the intervention in terms of improved general IC and modified food consumption after training. Attitudes toward food were also assessed to determine their mediating role in food choices. A total of 58 participants (47 female) completed 2 assessment sessions 3 weeks apart, with 2 weeks of intervention training for the training group during this time. The groups did not differ in baseline demographics including age, body mass index, and inhibitory control.
Inhibitory control ability improved across the training sessions, with increases in P3 amplitude implying increased cognitive control over responses. Inhibitory control training was associated with increased healthy and reduced unhealthy food consumption in a taste test and in the week following training, as measured by the Healthy Eating Quiz and the food consumption test. Cognitive restraint was enhanced after training for the training but not the waitlist condition in the Three-Factor Eating Questionnaire, implying that attempts to avoid unhealthy foods in the future will be easier for the training group participants.
Inhibitory control training delivered via a purpose-designed mobile phone app is easy to complete, is convenient, and can increase cognitive restraint and reduce unhealthy food consumption.
Australian New Zealand Clinical Trials Registry ACTRN12616000263493; http://www.ANZCTR.org.au/ACTRN12616000263493.aspx (Archived by WebCite at http://www.webcitation.org/6ioHjGING)
The prevalence of obesity has reached extreme proportions on a global scale. More than 2.1 billion people—one-third of the world’s adults, and one-fourth of children and adolescents—are classified as overweight or obese [
Inhibitory control refers to the ability to suppress dominant or automatic responses and allows for self-regulation by constraining thoughts and actions that may interfere with the completion of goal-directed actions [
There is increasing evidence that the neural mechanisms underlying executive functions are responsive to training and that general IC may be conceptualized as a “muscle” that can be strengthened with exercise [
Intervention studies indicate that IC training is an effective tool in reducing food intake and consumption of high-calorie foods in individuals with low trait IC [
Training tasks that have a food-specific, as opposed to a general, focus (using nonfood stimuli), have been found to yield greater improvements in food consumption behaviors [
Event-related potentials (ERPs) are derived from the electroencephalogram (EEG) and provide an objective index of the sensory and cognitive processing stages involved in processing a stimulus event. In the context of this study, ERPs allow consideration of the neural correlates of inhibition elicited during Go/Nogo tasks, with the N2 and P3 ERP components showing sensitivity to inhibitory demands. The N2 component typically occurs between 200 and 300 milliseconds after stimulus onset [
This study was a pilot study focused on researching the efficacy of an app-based intervention and its effect on short-term weight-related goals (ACTRN12616000263493). It utilized a randomized waitlist-control design and a novel IC intervention to train inhibition toward unhealthy foods and approach (or “disinhibition”) toward healthy foods in individuals identified as overweight or obese. Multisession IC training was employed, as more intensive exercise of neural inhibitory “muscles” should produce greater improvements in IC. This study also took a novel approach by (1) evaluating food consumption over a 1-week period, rather than immediately after training, (2) priming dominant automatic responses (or “disinhibition”) toward healthy foods to increase healthy food consumption, and (3) combining behavioral and psychophysiological measures. Attitudes toward dieting and food were also measured. It was hypothesized that participants in the training condition would consume less unhealthy and more healthy foods than the waitlist condition in the laboratory consumption test and over the 1-week period after training.
Furthermore, an exploratory hypothesis suggested that participants in the training compared with the waitlist condition would show (1) improved performance (reflected by less inhibition errors and faster reaction times) and (2) enhanced N2 and P3 amplitudes, after training in an untrained, non–food-specific Go/Nogo task. Previous research has focused on assessing the generalization of IC abilities from general to specific, or from general to general stimuli; this research is seeking to assess how this process may occur in the opposite direction. Additionally, in accordance with previous research showing changes in attitudes toward food and dieting after weight-based interventions [
Participants (n=58; females=47, 81%) were recruited from the general population in the Wollongong (Australia) area through flyers distributed in the community. Flyers called for individuals who wanted to improve their eating habits, and the intervention was described broadly as “brain training,” with no specific mention of IC. Participants were required to be older than 13 years (range 19-61, mean 36.48, SD 14.22), have a BMI higher than 25 (mean 29.54, SD 4.05), and possess an iOS device in order to access the training app. Participants were screened for neurological disorders and normal hearing and vision. Participants were excluded if they did not complete both experimental sessions (n=6), leaving a final sample of 52 participants (females=41).
CONSORT (Consolidated Standards of Reporting Trials) Flow diagram. Representation of the progress of participants through the trial. T1: time 1; T2: time 2.
The Barratt Impulsiveness Scale (BIS) is a 30-item questionnaire rated on a 4-point Likert scale and includes 3 subscales: attentional, motor, and nonplanning [
The training was delivered via the NoGo iOS app, developed by Neurocog Solutions Pty Ltd, Australia. The NoGo app contains IC training in 3 domains (unhealthy eating, smoking, alcohol consumption)—the training mechanism and parameters were designed by SJ informed by IC training literature. The training task was based on modified Go/Nogo and stop-signal tasks, with each “game” containing (1) Go and Nogo trials in which the image remained the same when the reaction time deadline (RTD) timer appeared next to the image and started to count down and (2) stop trials in which the image could change from healthy to unhealthy after the timer appeared. Participants were instructed to tap the images of healthy food as quickly and accurately as possible before the RTD expired and to refrain from tapping images of unhealthy food. Each game consisted of 30 trials, using stimuli drawn from a pool of food-specific images. For the purpose of creating a clear divide between different food types, the presented stimuli were in 2 categories: healthy (eg, fruits, vegetables) and unhealthy (eg, chips, doughnuts) foods (
Participants were required to play 10 games per day for 14 consecutive days with each game taking approximately 1 minute to complete. The difficulty level of the game increased according to past performance by reducing the RTD and increasing the number of images presented at one time (maximum 12 images). Go stimuli were presented on 70%-90% of occasions, varying randomly each game to ensure individuals were responding genuinely to presented stimuli and were not prepreparing responses based on previous patterns of stimulus presentation.
A log of performance data including reaction time, game level, correct responses, and errors was stored locally and accessed by researchers at the end of the training period. For analysis purposes, and in order to assess potential improvements in NoGo app performance as game play progressed, the training data of each individual participant were split into 3 sessions. Sessions 1, 2, and 3 comprised an equal number of games for each participant (eg, if a participant completed 120 games, session 1 would consist of data collected from games 1 to 40, session 2 would consist of games 41 to 80, and session 3 would consist of games 81 to 120). Reaction time, game level, and correct/incorrect response scores were averaged for each participant across each session to create 3 data points, which were used to assess improvements in game-play performance across all participants. Although participants in the training condition were expected to play a total of 140 games, an arbitrary cutoff of 90 games was selected, reflecting a 65% adherence rate. Participants who failed to play 90 games over the 14-day period were excluded from further analysis.
Example of the NoGo app environment with “unhealthy” (doughnut, orange soft drink) and “healthy” (radish, capsicum, water, orange) image categories. This example shows level 6—there are 6 images shown simultaneously, with the active image (requiring a response or not) indicated by the reaction time deadline timer (below doughnut).
The Healthy Eating Quiz (HEQ) is a food frequency questionnaire containing 70 items relating to the frequency of healthy food consumption over the past 7 days, rated on a Likert scale from “never” to “5 or more times a week.” It includes 8 subgroups: fruits, vegetables, meat proteins and vegetarian proteins, grains, water, dairy, and extras (ie, sauces and spreads). Participants were asked to rate approximately how many servings of items within these food groups they had consumed in the past week. Food frequency questionnaires are widely used to assess eating habits [
The food consumption test (FCT) was used to examine immediate changes in food consumption after the IC training. It was based on the bogus taste tests used in previous research but without the associated element of deception [
Participants completed the 51-item TFEQ [
Two auditory Go/Nogo tasks that varied in RTD were used to assess general IC ability. Tones were presented at 1100 Hz and 2000 Hz for 200 milliseconds, with an interstimulus interval (ISI) of either 2500 milliseconds (longer RTD) or 1250 milliseconds (shorter RTD). These specific time intervals were pilot-tested for this study in an adult population. Shorter RTDs increase task demands by inducing rapid responding to Go stimuli making the inhibition of motor responses on infrequent “Nogo” trials more difficult and increase the likelihood of errors [
In the first experimental session (around 90 minutes) participants gave informed consent and then completed the BIS, Go/Nogo tasks, FCT, HEQ, and TFEQ. Participants also completed a passive image-viewing task, not reported here. After completing this session, participants in the training condition were given access to the training app and instructed to play 10 games a day for 14 consecutive days. Participants performed a trial game with the researchers to provide them with the opportunity to ask questions and understand the game format. Participants were then sent reminders to play the game via email on days 1, 7, and 14 to ensure compliance. During the second session, approximately 3 weeks since session 1 and 1 week since the end of the training (average time between sessions was 4.5 weeks but did not differ between conditions), individuals repeated the same procedure, game data were obtained from the training participants, and the waitlist-control condition received access to the training app. Participants were entered into a prize draw to win 1 of 4 A$100 gift vouchers after completing the study, as reimbursement for their time.
Electroencephalographic data were recorded at 9 sites (F3, Fz, F4, C3, Cz, C4, P3, Pz, and P4) using a 19-site cap and a Nuamps amplifier (Compumedics, Melbourne Australia). Vertical and horizontal electro-oculogram was recorded using electrodes placed above and below the left eye and near the outer canthus of each eye. Linked ears were used as a reference, and an electrode located between Fz, Fp1, and Fp2 acted as a ground electrode, with impedances kept below 5 kΩ. Signals were amplified 19 times with a band-pass down 3 dB at 0.01 and 100 Hz. Raw EEG data were visually inspected and sections of muscle artifact were removed. Data were low-pass filtered down 3 dB at 24 Hz and divided into epochs from −100 milliseconds prestimulus to 900 milliseconds poststimulus. Epochs were excluded if they contained amplitudes outside ±100 µV at any EEG site. The number of accepted epochs for Go and Nogo stimuli were compared between conditions, with no differences evident. ERPs were calculated separately for correct Go and Nogo trials. Grand mean ERPs were visually inspected to identify major peaks. Peak quantification was completed via a computer algorithm, which allows for the automatic identification of the maximum or minimum voltage occurring within a specified latency window. For N2, peak identification at all other sites was locked to the largest negativity at Fz in the 190- to 300-millisecond latency window, whereas the P3 was locked to the largest positivity at Pz in the 300- to 580-millisecond window as per Johnstone et al [
One-way analysis of variance (ANOVA) was used to assess baseline differences between training and waitlist conditions and to assess NoGo training app performance data. Additionally, time (1, 2) × condition (waitlist, training) mixed design ANOVA was used to assess the FCT, HEQ, and TFEQ, with planned follow-up analysis splitting data by condition and assessing the effect of time. ERP component latency analysis was restricted to the frontal midline location (Fz), as the N2 and P3 components showed a frontal maximum. Go and Nogo ERP data were subject to a condition × time × stimulus (Go, Nogo) × RTD (longer, shorter) mixed factorial ANOVA. ERP amplitude analyses included an additional lateral (left, midline, right) factor. Planned contrasts within the lateral factor compared the right and left hemispheres. Here we focus on interactions between time and condition, and time × condition × laterality for ERP amplitude analysis. Alpha level was set to .05 for all analyses.
Baseline analysis results for training and waitlist conditions.
Demographics | Training | Waitlist | |||||
Mean | SD | Mean | SD | ||||
Age, years | 35.2 | 14.1 | 38.3 | 14.8 | 0.56 | .46 | |
Body mass index | 29.7 | 4.4 | 29.2 | 3.5 | 0.19 | .67 | |
Return time, days | 33.5 | 15.5 | 31.2 | 12.5 | 0.34 (1,46) | .56 | |
Attentional | 17.9 | 3.4 | 17.0 | 2.8 | 1.03 | .32 | |
Motor | 21.6 | 3.8 | 22.3 | 3.7 | 0.40 | .53 | |
Nonplanning | 23.5 | 3.9 | 22.6 | 5.2 | 0.54 | .46 |
a BIS: Barratt Impulsiveness Scale.
For participants completing the NoGo training, 1-way repeated measures ANOVA was used to assess changes in reaction time across the 3 training session blocks. The main effect of time was significant (
The main effect of time was also significant for correct Go responses (
The total HEQ score showed a time x condition interaction (
Calorie consumption was calculated for grapes, mixed unsalted nuts, M&M's, and plain potato chips separately. Then, the “healthier” foods (nuts and grapes) and “less healthy” foods (chips and M&M's) were added together to create 2 food groups for analysis. The condition × food × time ANOVA was significant (
The disinhibition subscale did not show a time x condition interaction. A significant interaction was found for the hunger subscale (
For the shorter RTD Go/Nogo task, the time x condition interaction was not significant for correct Go responses or reaction time, or Nogo errors. No significant interactions were found for the same measures on the longer RTD Go/Nogo task.
Grand mean ERPs to Go and Nogo stimuli for each condition and time are presented separately for the longer RTD (
Grand mean event-related potentials at Fz for Go and Nogo stimuli in the longer reaction time deadline task. W1: waitlist time 1; W2: waitlist time 2; T1: = training time 1; T2: training time 2.
Grand mean event-related potentials at Fz for Go and Nogo stimuli in the shorter reaction time deadline task. W1: waitlist time 1; W2: waitlist time 2; T1: training time 1; T2: training time 2.
P3 amplitude changes in the Go/Nogo tasks. P3 amplitude at sites F3 and F4 for the Go/Nogo tasks, with each reaction time deadline and time shown separately.
The key significant effects for both groups are presented in
Mean scores for each outcome; only statistically significant results are shown.
Variable | Participant group | |||
Training | Waitlist | |||
Baseline | Time 2 | Baseline | Time 2 | |
Healthy Eating Quiz | 36.96 | 42.42 | 39.58 | 35.88 |
FCTa: healthy food | 63.17 | 31.59 | - | - |
FCT: unhealthy food | 234.68 | 76.50 | - | - |
TFEQb: hunger | - | - | 5.58 | 6.88 |
TFEQ: cognitive restraint | 9.27 | 11.50 | - | - |
a FCT: food consumption test.
b TFEQ: Three-Factor Eating Questionnaire.
This pilot study used a randomized waitlist-control design to examine the efficacy of food-specific IC training in overweight and obese individuals. Traditional measures were used to assess food consumption (HEQ, FCT) and attitudes toward food and dieting (TFEQ). As a measure of general IC ability, Go/Nogo performance data and ERPs were also assessed. Key findings included increased healthy food consumption outside laboratory environments, decreased unhealthy food consumption during testing, and an increase in scores on the cognitive restraint scale for the training group. Condition-specific changes in the P3 ERP component indicated that the long-term, food-specific, IC training had a generalizable effect on IC-related processing, at least in the context of this preliminary research.
This study was novel and exploratory in its linking of healthy foods with approach responses in the context of an intensive multisession training regimen. As hypothesized, a reduction in immediate unhealthy food consumption and an increase in healthy food consumption (both in the laboratory and 1 week after training) was found for participants who completed the training task, supporting previous research [
To complement the food consumption data, attitudes toward food were also measured. Participants in the training condition showed increased cognitive restraint, a construct related to self-regulation of food intake on the TFEQ scale. Increased cognitive restraint has been linked to weight loss in dieters and improvements in avoiding unwanted food consumption [
Our study incorporated psychophysiological methods to measure the neural correlates of IC training. Contrary to prediction, N2 amplitude did not increase in the training condition as a result of IC training. This indicates that the IC training did not have a broad effect on the process reflected by the N2 component at either difficulty level. This may be due to the N2 being more closely related to conflict monitoring (a process not targeted by the IC training), as opposed to IC [
Training effects were present for the P3 ERP component, with amplitude increases after training seen in the training but not the waitlist conditions. This effect could be described as “global” as it did not differ based on the stimulus type (ie, Go, Nogo) and was consistent across the longer and shorter RTD Go/Nogo tasks. However, the nature of the condition effect depended on task difficulty; the shorter ISI task places higher demands on IC as a more speeded response is required to Go stimuli and could be considered more difficult. In the easier Go/Nogo task, the training condition showed a substantial increase (of similar magnitude in left and right frontal regions) in P3 amplitude at time 2, whereas the waitlist condition showed a substantial reduction (largest in the left frontal region). In the more difficult Go/Nogo task, the training condition showed a minor increase in P3 amplitude (largest in the right frontal region), whereas the waitlist condition showed a minor reduction (largest in the right frontal region). In both tasks, rapid responding was induced and thus inhibitory “activation” was required to inhibit automatic motor responses [
The results reported here should be considered in light of several limitations, primarily because of the limited nature of this preliminary study. Although all attempts were made to recruit adolescent participants, the final sample only included participants older than 18 years, preventing analysis across different age groups. Further research investigating the ideal age to administer IC training is warranted. Adolescents may experience greater benefits from the training, as IC abilities are known to continue developing into adolescence [
Additionally, the study duration was relatively short and did not use weight loss as an outcome measure. A longer-term study that assesses weight loss and maintenance would be beneficial, as previous research has found mixed results for changes in BMI at follow-up [
Most importantly, this study has focused on self-report measures, which are inherently biased and vulnerable to demand effects. Although the waitlist-control design helps to reduce the effects of random error, the nonblind randomization of participants means that individuals who know they are in a waitlist versus training group may not respond as they normally would. This may explain why the participants in the waitlist condition ate significantly less healthy food in the HEQ at time 2 and may explain the increase in the disinhibition subscale of the TFEQ. This subscale is linked to making poorer health food choices [
Future research must be especially careful in controlling for the expectations and beliefs of both the training and waitlist groups or measure these expectations to examine potential differences that may confound results. Further studies therefore need to consider a way of incorporating a more active control group, which will help make the purpose of the intervention less apparent to participants, reducing bias effects. Alternative food consumption measures, such as food diaries or dietary interviews, would also improve the accuracy of these data, although this was beyond the scope of this study. A larger number of participants and an equal number of male and female participants would improve statistical power.
This study, while preliminary, replicated and extended previous food-related IC training research, by priming both inhibitory responses to unhealthy food and approach responses toward healthy foods. Inhibitory control training was delivered through a purpose-designed, mobile phone app that allowed participants to complete the training at a time and place convenient to them. Further research is required to assess if the observed changes transfer to longer-term real-life contexts, such as weight loss. As it is easy to complete and cheap to obtain, IC training is a promising alternative or addition to existing obesity interventions.
analysis of variance
Barratt Impulsiveness Scale
body mass index
electroencephalogram
event-related potential
food consumption test
Healthy Eating Quiz
inhibitory control
interstimulus interval
reaction time deadline
Three-Factor Eating Questionnaire
We gratefully acknowledge the individuals who participated in this study. Neurocog Solutions Pty Ltd provided the NoGo app to research participants at no cost.
SJ provided the training mechanism and parameters for the NoGo app to Neurocog Solutions Pty Ltd (Australia) at no cost. SJ has no financial interest in the NoGo app. SJ is a co-inventor of intellectual property licensed by the University of Wollongong (UOW) to Neurocog Solutions Pty Ltd and is entitled to a small portion of royalties received by UOW in relation to the sale of any product that uses the UOW intellectual property. This intellectual property makes up a proportion of the intellectual property used in another Neurocog Solutions product (Focus Pocus) but not NoGo. TB and AR have no financial association with Neurocog Solutions Pty Ltd.