1

Exploring Decision-Making in Dietary Tasks

Aanshika Goswamy 24510001 , Adit Rambhia 21110009, Gautham Pullatt 24510047,

Teesha Saluja 24510140 , Utsav Agarwal 24510144

Indian Institute of Technology, Gandhinagar

CG504 Research Methods in Cognitive Science

Dr. Pratik Mutha

26th November 2024

 2

Introduction

Decision-making is a fundamental cognitive process that involves evaluating available

options and selecting the most advantageous option. A specific subset of decision-making,

self-control, is particularly crucial in scenarios where short-term temptations conflict with

long-term goals. The study of self-control is vital not only for understanding behavioural

economics and psychology but also for addressing practical challenges like obesity, addiction,

and financial planning.

Self-control in decision-making has been studied extensively through behavioural

paradigms, highlighting its susceptibility to situational and individual factors. However,

understanding the neurobiological underpinnings of self-control is still an evolving field. Hare et

al. (2009) made significant contributions by linking self-control to activity in two key brain

regions: the ventromedial prefrontal cortex (vmPFC) and the dorsolateral prefrontal cortex

(DLPFC). They proposed that successful self-control arises from the DLPFC's ability to

modulate the value signals encoded in the vmPFC, allowing individuals to prioritise long-term

goals over immediate gratification.

Hare et al. utilised a dietary decision-making task to examine how individuals weigh

health and taste attributes when choosing food items. Participants rated various food items for

their perceived healthiness and tastiness before choosing between them and a neutral reference

food. This experimental paradigm was particularly effective in isolating the neural mechanisms

of self-control because it introduced situations where participants had to choose between liked

but unhealthy foods and disliked but healthy ones. Based on their decisions, participants were

classified into self-controllers (SCs) and non-self-controllers (NSCs), enabling the researchers to

contrast neural activity between these groups.

 3

The study revealed that activity in the vmPFC consistently correlated with participants'

goal values, regardless of self-control status. However, the vmPFC integrated taste and health

information in self-controllers, whereas it primarily reflected taste in non-self-controllers.

Additionally, the DLPFC showed heightened activity during successful self-control trials and

was functionally connected to the vmPFC, highlighting its role in regulating value-based

decisions.

Our project builds upon this framework by applying the insights from Hare et al. to a

dietary decision-making task involving additional behavioural and statistical analyses.

Specifically, we aim to explore how hunger, time of day, and other contextual factors influence

decision-making and self-control. By collecting data on variables such as the time elapsed since

the last meal and the hunger index, we seek to understand how physiological states modulate

self-control in dietary decisions. Furthermore, we tried to replicate critical findings from Hare et

al. (2009), such as classifying participants into self-controllers and non-self-controllers and using

reaction time analyses to identify challenging decisions.

The research objectives of this project include:

Replication of Findings

Replicating the classification criteria and behavioural patterns observed by Hare et al. to

distinguish between self-controllers and non-self-controllers.

Statistical Insights

Employing t-test, ANOVA and regression analyses to test hypotheses about the

interaction between self-control, physiological states, and decision outcomes.

 4

Visualisation and Reporting

Presenting data through meaningful visualisations and statistical reports to elucidate

trends and differences across groups.

The decision to focus on dietary choices as the experimental paradigm is driven by its

relevance to real-world self-control challenges. Dietary decisions offer a tangible context where

individuals frequently encounter conflicts between short-term and long-term goals, making it an

ideal domain to study self-control mechanisms.

This project aligns with the broader goals of cognitive science by bridging behavioural

and neuroscientific perspectives. It also provides an opportunity to apply theoretical knowledge

to experimental design, data collection, and statistical analysis. By leveraging the methodological

rigour and insights from Hare et al., we aim to contribute to the understanding of self-control and

its practical implications in promoting healthier and more intentional decision-making.

 5

Relevant Literature

This literature review synthesises current research on dietary decision-making and

self-control, focusing on key themes relating to neural mechanisms, behavioural frameworks,

individual differences, paradoxical effects of indulgent cues, cognitive processes in food choice,

and roles of environmental and social factors. Combining findings across these domains thus

gives an integrated understanding of how individuals can navigate the complex interplay

between taste, health, and self-regulation.

Most dietary decision-making is neurobiological, and a key player in integrating taste and

health aspects of a food item is the ventromedial prefrontal cortex, vmPFC. Hare et al. (2009)

have shown that successful control of oneself requires the dorsolateral prefrontal cortex, dlPFC,

to modulate the activity of the vmPFC to prioritise considerations based on health rather than

immediate desires for hedonic gratification. These findings were expanded by Sullivan et al.

(2015), showing that the speed at which health and taste attributes are evaluated as necessary is

crucial; quicker health evaluations correlate with stronger self-control. Rangel (2013) emphasises

the interaction of several brain regions, such as the nucleus accumbens and dlPFC, which

mediate the conflict between reward-seeking and long-term goals cumulatively. These studies

indicate that dietary decisions are not isolated acts but outcomes of dynamic neural processes

that integrate sensory, cognitive, and emotional inputs.

Behavioural studies focus on the cognitive and contextual factors that influence dietary

decisions. Kuijer et al. (2008) conceptualised dieting as a form of behavioural decision-making,

with self-control central to resisting high-calorie temptations. This work emphasises the role of

deliberate strategies and pre-commitment in maintaining dietary goals. Masterson et al. (2019)

also examined the impact of dietary cognitive restraint. Their study suggests that participants
 6

with higher levels of restraint showed stronger preferences for healthier foods and faster reaction

times in the choice tasks. These behavioural tendencies underscore that the absence of dietary

restraint impairs the regulation of food-related impulses. In a relatively novel approach, Sullivan

et al. (2019) demonstrated that exposure to indulgent food items could sometimes enhance

self-control by reminding people about their long-term goals. Such paradoxes illustrate the

complex relationship between temptation and decision-making processes within food contexts.

Dietary decision-making varies based on significant individual variability in self-control

capacities, preferences and priorities. Pearce et al. (2020) explored children's food choices,

noting that the weight assigned to health versus taste attributes varied among individuals.

Individuals with a strong health focus demonstrated more consistent dietary restraint,

emphasising the significance of personal values and priorities in self-regulation. Rosenthal and

Dietl (2022) examined how different forms of control and decentering ability influence dietary

choices. They observed that individuals with higher self-control were better equipped to resolve

such conflicts. Decentering, a mindfulness-based cognitive strategy, was positively related to

dietary discipline, suggesting potential interventions for enhancing control.

Where cues related to indulgent items are typically viewed as temptations, sometimes

these cues will enhance self-control. Sullivan et al. (2019) demonstrated that exposure to such

foods activated cognitive associations with long-term health goals, leading to more disciplined

dietary choices. Similarly, Masterson et al. (2019) found that such cues could speed reaction

times for healthier options, suggesting a facilitative effect under certain conditions. This paradox

challenges more traditional models of temptation, suggesting that indulgent cues motivate

healthier behaviours when people have solid goal-setting frameworks. Such findings open the
 7

possibility of intervention design, where interventions could strategically use indulgent imagery

to support dietary goals.

Attribute weighting and decision-making heuristics are other cognitive mechanisms that

play a significant role in dietary self-control. Sullivan et al. (2015) demonstrated that the relative

processing speed of taste and health attributes affects decision outcomes, with faster health

processing linked to better self-regulation. This again underscores the importance of promoting

health-conscious thinking during food-related decisions. Masterson et al. (2019) proposed that

cognitive restraint, a strategy for pre-commitment, could work to align food choices with

long-term goals. Such strategies could include creating environments where healthy options are

more accessible or mental frameworks based on health ahead of indulgence.

Environmental and social contexts themselves dominate food and drink behaviours.

Pearce et al. (2020) concluded that what children eat is influenced more by healthiness and taste,

the availability of healthy foods, and cues from parents than by personal preferences. Social

modelling also affects preferences through exposure to other people's meal choices. Kuijer et al.

(2008) noted that environmental predictors such as tempting food cues and stressors could hinder

a person's self-control. Interventions targeting these environmental triggers, like restructuring

food environments to favour healthier options, might enhance more consistent dietary restraint.

The reviewed literature provides insight into designing effective interventions for

promoting dietary self-control. Rangel (2013) highlighted the importance of stimulating neural

mechanisms through cognitive training or mindfulness practice in strengthening self-regulation

circuits. Rosenthal and Dietl (2022) proposed that enhancements in mindfulness and decentering

ability may empower subjects to resist temptations and maintain healthier diets. The paradoxical

effects of indulgent cues, as described by Sullivan et al. (2019), suggest that strategic exposure to
 8

such cues could motivate individuals to adhere to dietary goals. In addition, cognitive processes,

such as pre-commitment strategies and habit formation, can enhance dietary decision-making

across diverse populations.

Several interconnected themes, such as those in the reviewed literature, are vital for

understanding dietary self-control. Neural mechanisms are crucial in showing how the

interacting brain regions mediate conflicts between considerations related to health, taste

preferences, and self-regulation. Behavioural frameworks, instead, are framed around the

importance of cognitive restraint, external cues, and pre-commitment strategies in shaping

dietary behaviours. Individual differences further emphasise variability in self-control capacities,

mental strategies, and personal priorities that reflect individuals' varied approaches toward

dietary regulation. Interestingly, the paradoxical effects of indulgent cues challenge traditional

views of temptation, revealing their potential to reinforce self-control under specific conditions.

Cognitive processes, such as attribute weighting and decision-making heuristics, are also crucial,

as they have been found to influence the prioritisation of health versus taste attributes in food

choices. There are also robust external influences shaped by environmental and social factors

that determine dietary behaviours and open avenues for intervention focused on these specific

areas. From these consolidated findings, the review demonstrates the sophistication of

self-control in dietary decision-making and points to areas for future research and interventions

designed to enhance self-control and correct unhealthy eating habits.

 9

Methodology

Experimental Paradigm

The Dietary Decisions Task is a well-established paradigm designed to investigate how

individuals evaluate and make decisions about food based on attributes such as health and taste.

It examines the cognitive processes and trade-offs involved in dietary choices. It operationalises

two fundamental constructs: a) food cue reactivity, reflecting the influence of external food cues

on decision-making, and b) self-control, evaluated through the ability to override immediate

sensory gratification (tastiness) in favour of long-term benefits (health). Our study focused

exclusively on self-control to better understand the cognitive mechanisms underlying

health-based dietary decisions.

The experiment consisted of two rating phases and a decision-making phase. Participants

rated 50 food items for their perceived healthiness and taste in two separate blocks in the first

two phases. The order of these blocks was counterbalanced across participants to eliminate order

effects. In the health rating block, participants rated the healthiness of each food item without

considering its taste, using a five-point scale ranging from Very Unhealthy (coded as -2) to Very

Healthy (coded as 2). In the taste rating block, participants rated the taste of each food item while

disregarding its healthiness on a five-point scale ranging from Very Bad (coded as -2) to Very

Good (coded as 2). Food items within each block were presented randomly, and participants

were given up to four seconds to make their ratings. Each trial was separated by a randomly

assigned inter-trial interval (ITI) lasting between four and fifteen seconds.

Following the two rating phases, a reference item was selected for each participant. This

reference was a food item rated as neutral on both health and taste scales. In cases where no item

met these criteria, an item rated neutral on the taste scale and healthy on the health scale was
 10

selected instead. The reference item was then used in the decision-making phase, where

participants made binary choices between the reference item and each of the 50 previously rated

food items. In each trial, the reference item was presented alongside a comparison item, and

participants indicated their preference using a five-point scale: Strong No (=choose reference),

No (=choose reference), Neutral, Yes (=choose shown item), or Strong Yes(=choose shown

item). Again, participants were given up to four seconds to make their decisions, and trials were

separated by ITIs ranging from four to fifteen seconds. Responses that exceeded the time limit

across all phases were excluded from the analysis.

This experimental paradigm systematically examined how individuals prioritise health

and taste in dietary decisions while controlling for extraneous factors such as presentation order.

The task ensured robust data collection by incorporating neutral reference items, providing

valuable insights into the cognitive mechanisms underlying food-related self-control.

The Protocol and Dataset Details

Our dataset was constructed following the guidelines outlined in the project brief,

ensuring robust and meaningful data collection. The key aspects are described below:

Participants

Data for the study were collected from 15 participants, comprising five females and ten

males, with a mean age of 19.4 years. All participants met the inclusion criteria and were

included in the final analysis. The participants were categorised into two groups based on their

decision-making behaviour (a robust subject classification described under the “Additional

Measures” heading): self-controllers (SC) and non-self-controllers (NSC). The SC group

 11

consisted of 7 participants, including two females and five males, with a mean age of 18.85

years, ranging from 18 to 20 years. These individuals consistently relied on health considerations

when making decisions, often prioritising healthier and less nutritious alternatives. In contrast,

the NSC group comprised 8 participants, including three females and five males, with a mean

age of 19.875 years and an age range of 17 to 23 years. Participants in this group predominantly

based their choices on taste preferences, demonstrating a lesser emphasis on the health attributes

of the options presented.

Participation in the study was entirely voluntary, with informed consent obtained from all

participants before the commencement of the experiment. Before engaging in the experimental

task, a comprehensive pre-experiment screening was conducted to gather additional contextual

data from the participants. This screening required participants to report the time elapsed since

their last meal, providing a measure of their fasting duration, which influences hunger-driven

decision-making. Participants were also asked to rate their current hunger level on a scale of 1 to

10, with 1 indicating minimal hunger and 10 representing extreme hunger. This self-reported

hunger index offered a critical subjective assessment of their physiological state during the

experiment. Additionally, the specific time of day during the experiment was recorded to account

for potential diurnal variations in decision-making behaviour or hunger patterns. This thorough

screening process ensured a robust foundation for interpreting the decision-making behaviours

observed during the study while also enabling an analysis of how individual characteristics and

states might interact with the experimental task.

 12

Stimuli

The dataset for the study comprised ratings and decisions for 50 distinct food items

carefully selected to capture a broad spectrum of tastiness and healthiness. The stimuli, presented

as high-resolution images, were displayed in a randomised order using standardised software to

minimise potential biases arising from order effects. The experimental setup utilised the Dietary

Choice Task, a validated paradigm on the ExpFactory platform. Participants rated food items on

a five-point Likert scale for taste ("Very Bad" to "Very Good") and healthiness ("Very

Unhealthy" to "Very Healthy"). In the decision phase, they chose between food items and a

neutral reference, with selections and reaction times recorded. This approach enabled a detailed

analysis of decision-making influenced by health and taste.

Additional Measures

This study classified participants as self-controllers (SC) or non-self-controllers (NSC)

based on their behaviour during the experimental task rather than their self-reported dietary

habits. This classification adhered to three predefined criteria adapted from prior methodologies.

The first criterion required participants to exhibit self-control in more than 50% of the trials

requiring such behaviour, such as rejecting "Liked-Unhealthy" items or selecting

"Disliked-Healthy" items. For this analysis, responses of "Strong No" and "No" were categorised

as rejections, while "Strong Yes" and "Yes" were classified as acceptances. The second criterion

involved a linear regression analysis of decision strength on health and taste ratings, where

self-controlled participants needed a higher regression coefficient for health than for taste. The

third criterion required that the goodness-of-fit (R²) of the regression for decision strength based
 13

on health ratings be greater than the R² for decision strength based on taste ratings for

participants to be classified as self-controlled.

Unlike the original study, which required participants to meet all three criteria to be

classified as SC, this study adopted a more flexible approach, categorising participants as SC if

they fulfilled at least two criteria. This adaptation was implemented for two primary reasons.

First, none of the participants in the dataset fulfilled all three criteria for SC, making the original

strict classification method unsuitable for this population. Second, this adjustment allowed for a

more balanced distribution of participants between the SC and NSC groups, ensuring that the

dataset was not overly skewed toward one category. Such balance is essential for robust

statistical analysis and enhances the interpretability of the results. Notably, participants who met

all three criteria for NSC were classified as NSC, maintaining the rigour of classification for this

group. This modified criterion facilitated a more comprehensive analysis of self-control

behaviours, accommodating the variability observed within the participant pool while preserving

the study's analytical integrity.

Additionally, in the analysis, trials where participants rated items as "neutral" on either

the health or taste dimension—indicating that they perceived the item as neither particularly

healthy nor unhealthy or neither particularly tasty nor untasty—were excluded from further

categorisation. These neutral responses, coded as 0, reflected a lack of solid preference or

aversion and were not classified within the four primary categories ("Liked-Healthy,"

"Liked-Unhealthy," "Disliked-Healthy," and "Disliked-Unhealthy"). As such, these items were

omitted from the data analysis to maintain clarity in interpreting the participants' explicit

preferences. Similarly, in the decision-making task, trials where participants chose a "Neutral"

response rather than a definitive "Yes" or "No" were excluded from the analysis of decision
 14

behaviour across the four categories. This approach ensured that the study focused solely on

clear and intentional decision patterns, providing a more precise understanding of participants'

preferences and their relationship to health and taste considerations.

This dataset design adheres to the standards and requirements specified in the project

brief while facilitating meaningful statistical analysis and comparison with prior research ( Hare

et al., 2009).

Statistical Analysis and Justification

Exploratory Tasks

In the exploratory analysis, we aimed to identify the tastiest, untastiest, healthiest, and

unhealthiest food items rated by participants. The statistical approach involved calculating the

average item-wise rating across all participants for both the taste and health dimensions. This

method provided an aggregate measure reflecting the collective perception of each food item.

For the tastiest and untastiest items, the taste ratings of each food item were averaged

across participants. The food item with the highest average rating was identified as the tastiest,

while the one with the lowest average rating was identified as the untastiest. Similarly, for the

healthiest and unhealthiest items, the health ratings of each item were averaged, with the highest

and lowest averages corresponding to the healthiest and unhealthiest items, respectively.

Averages were used for the statistical analysis as it provides an objective measure of each

item's perceived healthiness and tastiness based on the collective input of all participants. Using

average ratings, the analysis minimised the influence of individual rating biases and allowed us

to identify the most extreme cases regarding taste and health perceptions. This approach also
 15

facilitated further comparisons in subsequent phases of the experiment, ensuring consistency in

interpreting participants' choices and preferences.

Easy analysis

To explore whether there was a difference in the time taken for decision-making between

healthy and unhealthy food items, we conducted a dependent samples t-test. This test compared

the mean decision times for trials involving healthy items to those involving unhealthy items

among the same participants. The dependent samples t-test was chosen as it accounts for the

paired nature of the data, ensuring that differences are analysed within individuals, which

improves statistical power and controls for variability between participants.

Mandatory analysis

To analyse differences in response time for decision-making between self-controlled and

non-self-controlled individuals across the categories of disliked unhealthy, disliked healthy, liked

unhealthy, and liked healthy food items, we performed a mixed-method ANOVA. This approach

allowed us to examine both within-subject factors (the four food categories) and the

between-subject factor (self-control status: self-controlled vs. non-self-controlled).

The mixed-method ANOVA was appropriate for this analysis as it accounted for the

repeated measures nature of the within-subject factor while simultaneously testing for

group-level differences. Interaction effects between self-control status and food category were

also evaluated to determine whether response patterns differed based on self-control. This

method comprehensively understood how self-control and food category influenced

decision-making response times.

 16

We used regression analysis to examine the relationship between self-control and hunger

to determine if hunger (measured by the hunger index and hours since the last meal) negatively

affects self-control. In this model, self-control was treated as the dependent variable. At the same

time, hunger was measured through two independent variables: the hunger index and the number

of hours since the participant's last meal. The regression aimed to assess whether these hunger

measures significantly predicted self-control during the dietary decision-making task.

By including the hunger index and hours since the last meal as predictors, the analysis

provided insights into how varying levels of hunger might influence participants' ability to resist

immediate gratification (tastiness) in favour of healthier choices. The regression helped

determine if hunger, as operationalised in this way, significantly affected self-control during the

task.
 17

Results

Exploratory Tasks

As part of the exploratory analysis, participants rated the food items on scales of taste and

healthiness, providing insights into their subjective perceptions. Based on average ratings, the

tastiest items were strawberries and white grapes, achieving a high mean rating of 1.4, reflecting

their strong appeal across participants. Conversely, the least tasty item was orange jello, which

received a notably low average rating of -0.0667, indicating general disfavour.

Regarding healthiness, white grapes emerged as the highest-rated item, with a mean

healthiness score of 1.8, aligning with their status as a natural and nutritious choice. In contrast,

Twix, Reese’s, and Doritos Ranch were the least healthy items, each receiving an average health

rating of -1.2, highlighting their perception as unhealthy, indulgent snack options. These results

delineate between the items perceived as most and least desirable based on taste and health

attributes, offering valuable context for understanding participant preferences.

Easy analysis

A paired samples t-test was conducted to determine whether there was a significant

difference in reaction times for decision-making on healthy versus unhealthy food items.
 18

Figure 1. Comparison of reaction times for decision-making on healthy and unhealthy food
items, showing no significant difference

Results showed no significant difference between reaction times on healthy food items

(M = 1887.53, SD = 257.39) and unhealthy food items (M = 1921.06, SD = 273.45), t(16) =

-0.54, p = 0.596, Cohen’s d = -0.1314044.

The normality assumption was tested using the Shapiro-Wilk test and was met, W = 0.91,

p = 0.083. A descriptive plot indicated slight differences in mean reaction times between

conditions, but these were not statistically significant. Hence, this sample's reaction times for

healthy and unhealthy decisions appear comparable.

 19

Mandatory analysis

The subjects were classified into self-controlled (SC) and non-self-controlled (NSC)

based on the abovementioned criterion.

Figure 2. The percentage of time participants chose specific foods over a reference item,
comparing SC (blue) and NSC (red) groups.

The graph illustrates the percentage of time participants chose specific foods over a

reference item, comparing two groups: SC (blue) and NSC (red). The four categories of food

preferences analysed are Disliked-Healthy, Liked-Healthy, Liked-Unhealthy, and

Disliked-Unhealthy. Each bar on the graph represents the mean percentage of choices, with error

bars indicating the standard error of the mean.

In the Disliked-Healthy category, contrary to expectations, participants from the SC

groups chose disliked but healthy foods less frequently than those from the NSC groups.

Although both groups demonstrated a high preference for Liked-Healthy foods, the SC group

had a slightly higher selection rate. In the Liked-Unhealthy category, participants from both

groups chose the liked but unhealthy foods almost 75% of the time, indicating a high and nearly
 20

identical preference for these foods across both groups. Finally, in the Disliked-Unhealthy

category, participants from the NSC groups chose disliked and unhealthy foods approximately

25% of the time, which is higher than the SC group. The standard error bars on each bar

represent the variability or precision of the sample mean estimate.

The results of the Repeated Measures ANOVA are divided into two main sections:

Within Subjects Effects and Between Subjects Effects, each reporting on critical metrics such as

Sum of Squares, degrees of freedom (df), Mean Square, F-value, p-value, partial eta squared (η²),

and omega squared (ω²).

Before performing the Repeated Measures ANOVA on the dataset, the required

assumption checks were conducted to ensure the validity of the analysis.

Figure 3. Levene's test results and Q-Q plot of standardised residuals for ANOVA assumption
checks, assessing homogeneity of variances and normality of residuals, respectively
 21

Figure 4. ANOVA table obtained from JASP by running Repeated Measures ANOVA.

Within Subjects Effects revealed that the factor Taste showed a near-significant effect,

with F(1,6)=5.882963 and p=0.051478, indicating a marginal influence of taste on participants'

choices. The partial eta squared (η²) of 0.027494 and omega squared (ω²) of 0.032491 suggest a

small effect size. However, the interaction between Taste and Group was not significant

(F(1,6)=0.779705, p=0.411210), indicating no significant interaction between taste and group

preferences.

Similarly, the factor Health (F(1,6)=0.405308, p=0.547864) and its interaction with

Group (F(1,6)=1.395856, p=0.215355) were not significant, suggesting that health did not

significantly influence the participants' choices, nor did it interact with group preferences.

Additionally, the three-way interaction of Taste, Health, and Group was also not significant
 22

(F(1,6)=0.118770, p=0.742131), indicating no significant combined effect of these factors on

participants' choices.

In the Between Subjects Effects, the factor Group was not significant (F(1,6)=0.344981,

p=0.573835), suggesting no overall significant difference between the groups. The partial eta

squared (η²) of 0.013878 indicates a minimal effect size. Overall, these results suggest that while

there are some minor effects of taste on participants' choices, the factors of health and group do

not have significant effects, nor do their interactions.

We utilised logistic regression to examine the relationship between self-control and

hunger, explicitly investigating whether fasting before the experiment had a negative impact on

self-control. Two separate tests were run.

Figure 5. Logistic regression results, including model summaries and coefficients for the
intercept of Hours before meal and Hunger index.
 23

Figure 5 represents the logistic regression analysis results divided into two sections:

Model Summary - Group and Coefficients. In the Model Summary - Group section, three models

(M₀, M₁, and M₂) are compared using metrics such as deviance, AIC, BIC, degrees of freedom

(df), likelihood ratio tests (Δχ² and p-value), and various pseudo-R-squared measures (McFadden

R², Nagelkerke R², Tjur R², and Cox & Snell R²). Model M₀ serves as the baseline, while Model

M₁ includes predictors for hours before meal and hunger index, and Model M₂ possibly includes

additional predictors or interactions. None of the models significantly improve fit compared to

the baseline, as indicated by the non-significant Δχ² values (e.g., M₁: Δχ² = 0.010774, p =

0.917329; M₂: Δχ² = 0.112635, p = 0.737164). Furthermore, the pseudo-R-squared values are

very low, suggesting minimal explanatory power.

In the Coefficients section, the intercepts for all models are non-significant, indicating no

substantial baseline effect. For Model M₁, the predictors - hours before meal and hunger index

are insignificant, with p-values of 0.895796 and 0.732454, respectively. Similarly, in Model M₂,

these predictors remain non-significant, with p-values of 0.780645 for hours before meal and

0.796401 for hunger index. These findings suggest that the included predictors do not

significantly contribute to explaining the outcome variable. The analysis indicates that the

models do not significantly improve upon the baseline, and the predictors do not directly impact

the outcome.
 24

Figure 6. Logistic regression results, including model summaries and coefficients for the
intercept of Hours before meal.

Figure 6 represents the logistic regression analysis results, divided into Model Summary -

Group and Coefficients sections, indicating that including the predictor variable Hours before

meal does not significantly improve the model's explanatory power. In the Model Summary -

Group section, comparing the null model (M0​) with the model including Hours before meal (M1​),

the likelihood ratio test (Δχ2=0.005801Δ, p=0.939291) shows no significant improvement in fit.

The pseudo-R-squared values (e.g., McFadden R² = 0.000280) remain very low, suggesting

minimal explanatory power. The Coefficients section reveals that the intercepts in both models

(M​0 and M1​) are not statistically significant, with p-values of 0.796401 and 0.832203,

respectively. Moreover, the predictor Hours before a meal in M1​also fails to show significance (p

= 0.939286), with a negligible effect size. These results collectively suggest that the predictor

Hours before meal does not significantly contribute to explaining the outcome variable, and the

models have very limited explanatory power.

 25

Discussion

The present study extended the foundational work by Hare et al. (2009) by exploring the

role of self-control in dietary decision-making and incorporating additional contextual variables

like hunger index, age, time of day, and hours from the last meal. While previous research

primarily focused on the interplay between neural mechanisms and behavioural self-control, our

study highlighted how physiological states and external contexts shape self-control in everyday

decisions.

A critical goal of this research was to replicate the criteria for distinguishing between

self-controllers (SCs) and non-self-controllers (NSCs) established by Hare et al. (2009). We

successfully classified participants using behavioural patterns such as preference consistency and

response time, corroborating prior findings that SCs prioritise health attributes over taste during

decision-making. The decision-making paradigm further validated the link between

health-oriented considerations and self-control by replicating similar behavioural patterns.

However, a notable divergence from Hare et al. was the adaptation of the classification criteria

due to the dataset's characteristics. This adjustment underscores the importance of contextual

flexibility in experimental design, ensuring relevance and rigour in diverse participant

populations.

The mixed-method ANOVA revealed that taste, health and their interaction with

self-control status were not statistically significant. Interestingly, SCs showed a slightly higher

preference for liked-healthy items than NSCs, highlighting a nuanced interplay between

self-control and intrinsic motivation toward healthier options. In our analysis, hunger did not

correlate significantly negatively with self-control. Neither the hours elapsed since the last meal

nor the self-reported hunger index significantly predicted self-control behaviours, as indicated by
 26

logistic regression analyses. These findings challenge the widely held assumption that hunger

universally compromises self-control, suggesting that its impact may vary based on individual

differences or the specific decision-making context.

The study faced several methodological challenges, including limited sample size and

variability in participant behavioural patterns. While necessary for balanced analysis, the

flexibility in classification criteria for SCs and NSCs introduces potential variability that could

influence interpretations. Furthermore, although methodologically justified, excluding neutral

responses from analyses may limit the comprehensive understanding of ambiguous or uncertain

decision-making scenarios. These limitations emphasise the need for larger, more diverse

participant samples in future research to enhance generalizability.

The study's implications extend beyond theoretical contributions to practical domains like

public health and behavioural interventions. Understanding the factors influencing dietary

decisions can inform strategies for promoting healthier eating habits. For instance, interventions

could focus on enhancing the salience of health attributes or mitigating the impact of hunger on

impulsive choices. Additionally, tailoring strategies based on individual differences in

self-control may improve the efficacy of interventions aimed at addressing obesity, addiction,

and other self-regulatory challenges.

 27

Conclusion

The nuanced interplay between taste, health, and contextual factors underscores the

complexity of decision-making processes, highlighting the need for multifaceted approaches in

research and practical applications. By bridging behavioural and physiological perspectives, this

work lays the groundwork for more comprehensive models of self-control and its role in

promoting intentional and healthier choices.

The practical implications of these findings are substantial. Insights into dietary

decision-making processes can inform interventions to promote healthier eating behaviours, such

as increasing the salience of health attributes or mitigating hunger's impulsive effects. These

interventions could leverage individual differences in self-control to create tailored strategies for

addressing broader challenges.

Future research should expand the contextual scope by incorporating variables such as

stress, social influences, and longitudinal data to observe behavioural trends over time, enabling

a more comprehensive understanding of decision-making processes. This study lays a foundation

for advancing theoretical and applied domains in promoting healthier and more intentional

dietary choices by bridging behavioural insights with actionable strategies.

 28

References

Hare, T. A., Camerer, C. F., & Rangel, A. (2009). Self-control in decision-making involves the
modulation of the vmPFC valuation system. Science (New York, N.Y.), 324(5927),
646–648. https://doi.org/10.1126/science.1168450

Kuijer, R., de Ridder, D., Ouwehand, C., Houx, B., & van den Bos, R. (2008). Dieting as a case
of behavioural decision making. Does self-control matter? Appetite, 51(3), 506–511.
https://doi.org/10.1016/j.appet.r2008.03.014

Masterson, T. D., Brand, J., Lowe, M. R., Metcalf, S. A., Eisenberg, I. W., Emond, J. A.,
Gilbert-Diamond, D., & Marsch, L. A. (2019). Relationships Among Dietary Cognitive
Restraint, Food Preferences, and Reaction Times. Frontiers in Psychology, 10, 2256.
https://doi.org/10.3389/fpsyg.2019.02256

Pearce, A. L., Adise, S., Roberts, N. J., White, C., Geier, C. F., & Keller, K. L. (2020). Individual
differences in the influence of taste and health impact successful dietary self-control: A
mouse tracking food choice study in children. Physiology & behavior, 223, 112990.
https://doi.org/10.1016/j.physbeh.2020.112990

Rangel A. (2013). Regulation of dietary choice by the decision-making circuitry. Nature

Neuroscience, 16(12), 1717–1724. https://doi.org/10.1038/nn.3561

Rosenthal, J., & Dietl, E. (2022). The role of trait self-control, healthy eating habits and
decentering ability in response to conflict (Version 1). Loughborough University.
https://hdl.handle.net/2134/17711228.v1

Sullivan, N., Hutcherson, C., Harris, A., & Rangel, A. (2015). Dietary Self-Control Is Related to
the Speed With Which Attributes of Healthfulness and Tastiness Are Processed.
Psychological Science, 26(2), 122-134. https://doi.org/10.1177/0956797614559543

Sullivan, N. J., Fitzsimons, G. J., Platt, M. L., & Huettel, S. A. (2019). Indulgent Foods Can
Paradoxically Promote Disciplined Dietary Choices. Psychological science, 30(2),
273–287. https://doi.org/10.1177/0956797618817509
 29

APPENDIX

Appendix A - Data

Master Datasheet:

CG 504: Dietary Decisions Survey (Responses)

Raw data sheet including averages:

RM_ProjectRaw

Data sheet used for Logistic Regression:

Regression data

Data sheet used for ANOVA:

ANOVA DATA

Appendix B - Contribution Report

Aanshika Goswamy (24510001) - Data Cleaning & Preparation, Exploratory & Mandatory
Analysis, Methodology and Results

Adit Rambhia (21110009) - Data Collection, Easy & Exploratory Analysis, and Review of
Literature

Gautham Pullatt (24510047) - Easy & Exploratory Analysis, Introduction and Conclusion.

Teesha Saluja (24510140) - Data Cleaning & Preparation, Exploratory & Mandatory Analysis,
Methodology and Results

Utsav Agarwal (24510144) - Data Cleaning & Preparation, Exploratory & Mandatory Analysis,
Discussion, and Formatting.