University of Southern California Dissertations and Theses

Illusory motives, elusive habits: lay theories of everyday behavior

(USC Thesis Other)

Illusory motives, elusive habits: lay theories of everyday behavior

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Running head: LAY THEORIES OF BEHAVIOR 1

ILLUSORY MOTIVES, ELUSIVE HABITS: LAY THEORIES OF EVERYDAY
BEHAVIOR

by

Asaf Mazar

A Thesis Presented to the
Faculty of the USC Dornsife College of Letters, Arts and Sciences
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
Master of Arts

(SOCIAL PSYCHOLOGY)

May 2020

LAY THEORIES OF BEHAVIOR 2

Abstract
Causal explanations of habitual, automatic actions are prone to inaccuracy simply because
people have only limited awareness of automatic processes. Thus, they may mistakenly explain
habit performance as a response to salient but incidental inner states such as their current mood.
We tested this misattribution hypothesis in an online experiment, a field study, and an ecological
momentary assessment study. With all three methodologies, people’s attributions for their
actions were poorly calibrated to the actual predictors of these actions. Habit strength
outperformed or matched inner states in predicting behavior, whereas participants’ attributions
for their behavior gave greater or equal weight to inner states over habit. Thus, lay perceptions of
behavior overweight salient inner states and discount automatic processes. We discuss the
implications of these findings for self-regulation and the limitations of self-report measures.
Keywords: habit, misattribution, automaticity, lay theories, self-insight, introspection
illusion

LAY THEORIES OF BEHAVIOR 3

Table of Contents
Abstract……………………………………………………………………………………………2
Introduction………………………………………………………………………………………..4
Attributions are Based on Introspections………………………………………………….5
Experimental Evidence of Misattribution…………………………………………………6
Misattribution in Everyday Behavior……………………………………………………...8
The Present Research…………………………………………………………………….10
Study 1…………………………………………………………………………………………...12
Method…………………………………………………………………………………...12
Results……………………………………………………………………………………15
Discussion………………………………………………………………………………..17
Study 2…………………………………………………………………………………………...18
Method…………………………………………………………………………………...19
Results……………………………………………………………………………………20
Discussion………………………………………………………………………………..22
Study 3…………………………………………………………………………………………...23
Pilot………………………………………………………………………………………25
Method…………………………………………………………………………………...25
Results……………………………………………………………………………………30
Discussion………………………………………………………………………………..34
General discussion……………………………………………………………………………….35
References………………………………………………………………………………………..43
Tables…………………………………………………………………………………………….49
LAY THEORIES OF BEHAVIOR 4

Figures……………………………………………………………………………………………55

LAY THEORIES OF BEHAVIOR 5

Illusory Motives, Elusive Habits: Lay Theories of Everyday Behavior
Why do we act the way we do? Decades of research on attribution and lay theories show
that humans naturally form mental models to make sense of their experience (Beruchashvili &
Moisio, 2013; Kelley, 1973). For example, you might have a sense that you overeat when
stressed or that your mood is better on days when you have slept well. Many of these theories are
accurate enough to be useful; in other cases, however, they may lead us astray.
In the present article, we examine the extent to which attributions to two possible causes
of behavior—habits and inner states—capture the actual effects of these causes. We reasoned
that automatic influences on behavior often function outside of conscious awareness (Moors &
de Houwer, 2006), whereas inner states are more accessible and tend to be overvalued as causes
of our own behavior (see introspection illusion, Pronin, 2009). Consequently, we expected that
people’s explanations for why they perform everyday actions would overweight the influence of
inner states and underweight the influence of habit automaticity. We report three studies that
compared people’s explanations of the causes of repeated behaviors with the actual influence of
their inner states and habits.
In misattribution’s long history, it is hardly novel to argue that people are often mistaken
in their explanations for their own actions. However, we know little about the pervasiveness of
such attributions in daily life, especially given the extent to which habits and other forms of
automaticity guide everyday behaviors (Amodio & Ratner, 2011; Wood, Quinn, & Kashy, 2002).
The everyday behaviors in our research provide an especially strong test of misattribution, given
that people have many opportunities in daily life to observe behaviors like purchasing and
consuming food and drink and to form accurate inferences about the reasons why they do (or do
not) perform them.
LAY THEORIES OF BEHAVIOR 6

Attributions are Based on Introspections
When making attributions, self-predictions, and other judgments that rely on self-
knowledge, people favor introspections about their personal thoughts, feelings, and emotions (the
introspection illusion, Pronin, 2009). For example, people tend to describe themselves in terms
of inner states but describe others in terms of more observable qualities (McGuire & McGuire,
1986). Similarly, people judge their behavior based on their thoughts and intentions but judge
others by observing their actions (Pronin, Berger, & Molouki, 2007).
When explaining their own behavior, people favor inner experiences for informational as
well as motivational reasons. From an informational perspective, people over-attribute to inner
states simply because these are chronically salient and accessible. Given immediate access to a
rich network of cognitive and affective experiences, these inner experiences are readily available
to serve as attributions. Unless a contextual influence is particularly salient, it may escape
detection and therefore leave inner states as the main available explanation (e.g. Schwarz &
Clore, 1983).
Despite the appeal of an informational account, accessibility and salience by themselves
do not wholly explain our tendency to rely on inner states to make sense of our experience. Even
when others’ introspections are made highly accessible, people do not rely on them as much as
when making self-judgments (Pronin & Kugler, 2007). Thus, it seems that over-reliance on
introspection is additionally influenced by several motivational biases.
First, undue attribution to inner states may be motivated by an illusion of control, or the
tendency to overestimate personal control over external events (Langer, 1975). People infer that
they are in control over even completely aleatory events, such as coin tosses and die rolls (see
review by Presson & Benassi, 1996). Because people overestimate their control over external
LAY THEORIES OF BEHAVIOR 7

events, they may favor inner states that are accessible to conscious awareness, such as thoughts,
moods, and intentions.
Overvaluation of inner experience may be motivated further by a need to maintain a
consistent view of the self. As outlined in cognitive dissonance theory, actions that are not
sufficiently justified can create a negative tension state (Aronson, 1969). Given that people so
readily interpret their overt actions as intentional (Wegner & Wheatley, 1999), the simple
thought of acting unintentionally may cause negative affect. In line with this idea, people
experience more guilt with accidental than intentional wrongdoing (McGraw, 1987), suggesting
that unintentional accidents may at times be more aversive than wrongheaded decisions. In order
to reduce the aversive state of not knowing the reasons for one’s actions (a.k.a., explanatory
vacuum), people may confabulate post-hoc reasons to explain their behavior (Oettingen, Grant,
Smith, Skinner, & Gollwitzer, 2006; Parks-Stamm, Oettingen, & Gollwitzer, 2010).
Experimental Evidence of Misattribution
Research on over-estimating internal states comes mostly from highly controlled settings
in contexts that obscure the actual causes of behavior. In Wilson & Nisbett’s (1978) classic
study, for example, shopping mall patrons chose their favorite stocking out of four possibilities.
Unbeknownst to participants, the four pairs were identical. Nevertheless, participants' choices
were not randomly distributed: most of them chose the final pair of stockings that they inspected,
which these researchers attributed to a “habit of ‘shopping around,’ holding off on choice of
early-seen garments…in favor of later seen garments” (Nisbett & Wilson, 1977, p. 244).
Participants’ own explanations of their choices identified a plethora of reasons but failed to
recognize the effect of choice-order habits.
LAY THEORIES OF BEHAVIOR 8

Causes of behavior are especially difficult to identify when automaticity is involved.
Automaticity is a multi-dimensional construct characterized by features such as being
unintentional, efficient, stimulus-driven, and most importantly for the purposes of this paper -
nonconscious (Moors & de Houwer, 2006). Because many everyday habits involve automatic
processes relatively inaccessible to conscious awareness (Wood & Rünger, 2016), habitual
behaviors may be misattributed to salient but irrelevant inner states.
Limited awareness has already been identified as key to a classic form of automaticity—
priming. In a typical priming paradigm, a concept, goal, or behavior is subtly made accessible
and thereby shapes subsequent actions and judgments. For example, participants in one study
who were exposed to names of a family member or friend more strongly endorsed goals
associated with that person (Shah, 2003). Following the situated inference model, these effects
are induced by misattribution (Loersch & Payne, 2011): Priming occurs when active mental
content (e.g., thought of mom) is attributed to one’s own inner thoughts and feelings (e.g.,
achievement goals). Similar ideas can be found in the active-self model of priming, which asserts
that external primes exert influence by activating and interacting with components of the self
(Wheeler, DeMarree, & Petty, 2007).
In a potent example of priming misattribution, Bar-Anan, Wilson, & Hassin (2010)
implicitly primed participants with goals while providing explicit reasons that could explain
participants’ behavior. In one example (Study 4), participants who were primed with a money-
earning goal were more likely to choose a trivia game that featured irrelevant pictures of U.S
dollars. After they had already made their choice, participants were informed randomly that their
game was challenging or easy. Participants then misattributed their choice to this information
LAY THEORIES OF BEHAVIOR 9

after the fact, so that participants who supposedly chose the challenging game reported higher
liking for challenges.
Misattribution in Everyday Behavior
Thus, previous studies have demonstrated misattribution in controlled settings designed
to elicit misattribution. Less is known about whether misattribution occurs naturally in routine
forms of behavior such as habits. Habits are automatic context-response associations formed as
people repeat a behavior in a stable context (Mazar & Wood, 2018; Wood, 2017). For example,
when first moving to a new location, one might deliberately navigate from home to work,
deciding what transit to take and when to leave. With repetition, however, this process can
become habitual, so that each step of the journey is cued by its usual preceding step. Eventually
commutes involve only minimal attention and decision making. Many of our daily actions
become automated in this way: An experience sampling study in which participants reported
what they were thinking, doing, and feeling every hour revealed that around 43% of daily
behaviors were performed habitually, in that they were repeated frequently in the same location,
usually while participants were thinking about something else (Wood et al., 2002).
Even though people have considerable experience with their habits, they may still
misattribute their causes due to being unaware of the cuing mechanism that guides habits. They
may overlook habit cues in favor of more informationally accessible and motivationally
gratifying introspections. In support, more habitual runners believed that their running was
driven by their intentions, even though these intentions were not mentally associated with their
behavior (Neal, Wood, Labrecque, & Lally, 2012). Similarly, people with stronger habits for
watching TV news, exercising, and taking the bus reported greater certainty in their intentions,
LAY THEORIES OF BEHAVIOR 10

even though those intentions did not predict their future behavior (Ji & Wood, 2007). Therefore,
people with strong habits for a behavior are likely to misattribute it to inner states.
What about people with weaker habits? Non-habitual behavior may be more amenable to
the influence of inner states, as people often act on their desires and intentions (Gardner, 2015;
Ouellette & Wood, 1998). However, even attributions for such acts may be inaccurate. Although
people have access to a rich source of personally introspected experiences, they also are
influenced by a-priori, culturally shared causal theories (Nisbett & Bellows, 1977). For example,
a person may attribute overeating to bad moods based on a common theory that connects the
two. In one study, college students tracked their daily mood and sleep patterns for five weeks and
estimated that their mood was better on days when they slept well, even though actual within-
person associations between sleep and mood were close to zero (Wilson, Laser, & Stone, 1982).
Moreover, participants’ judgments were no more accurate than judgments made by unrelated
students, even though participants had closely monitored their own daily experiences for over a
month. Therefore, despite cultural valuing of introspective experiences, even attributions of
nonhabitual behavior to inner states may be only weakly related to the actual effects of those
states.
Among the many internal experiences plausibly explaining actions, mood states have
emerged in numerous lines of research on attribution and misattribution. For example, smokers
in one study reported that their smoking was triggered by negative affect, even though in-the-
moment assessments of affect revealed little association between emotions and subsequent
smoking (Shiffman et al., 1997; see also Shiffman et al., 2002). Self-reported emotional eaters
show a similar gap between judged and actual effects of mood (Adriaanse, Prinsen, de Witt
Huberts, de Ridder, & Evers, 2016). After an episode of overeating, such eaters retrospectively
LAY THEORIES OF BEHAVIOR 11

exaggerated the amount of negative affect they felt before overeating—a judgment that favored a
coherent self-view over accurate introspection. Together, these lines of research suggest that lay
explanations of behavior often implicate mood, and that mood may be given more credit than
merited by its actual impact on behavior.
The Present Research
The present research compares the actual effects of habit strength and inner states on
behavior with the effects people perceive them to have. Given the informational and motivational
biases detailed above, we expected lay theories to overestimate the effect of inner states on
behavior and underestimate the effect of habit. Our analysis thus has novel, broad-reaching
implications, suggesting that misattribution occurs not only as a result of experimentally
constructed paradigms like priming but instead is inherent in our experience of everyday life.
We first tested this prediction in a hypothetical scenario in which participants imagined a
positive, negative, or neutral life event and then reported their food purchase intentions. This
design provided a highly controlled test of our hypothesis bolstered by random assignment to
mood-inducing conditions. We then examined the ecological validity of this finding in a field
study that tracked actual food or drink purchases following a mood-inducing event (a difficult
exam). Our third study extended beyond this emotionally-evocative event paradigm and used
ecological momentary assessment (Shiffman, Stone, & Hufford, 2008) to track the actual and
perceived effects of habit and fatigue on coffee drinking in everyday contexts.
We have two main predictions. First, we hypothesized that habit strength will predict
food and drink purchase and consumption better than inner states. These behaviors are likely to
be performed frequently in similar contexts and therefore become habitual (Wood & Rünger,
2016). Previous research has supported the notion that eating is mostly automatic and habitual
LAY THEORIES OF BEHAVIOR 12

(Cohen & Farley, 2007; Moldovan & David, 2012; Riet, Sijtsema, Dagevos, & De Bruijn, 2011).
To state this hypothesis formally:
H1: Habit strength will predict drinking and eating better than inner states such as mood
and fatigue.
Second, although people repeatedly experience food and drink choices and their
antecedents, and presumably have ample opportunity to understand what drives them, we
hypothesized that accessible and motivationally congenial inner states would overshadow habit
as potential causes of behavior. Therefore, we predicted that participants, regardless of whether
they have strong or weak habits, will attribute their behavior to inner states more than habit. In
essence, we asked participants to duplicate with their attributions the predictive models for
behavior we computed in hypothesis one. As a formal statement:
H2: Participants’ attributions for their own behavior will assign more importance to inner
states (e.g. mood, fatigue) than habit.
Taken together, these two hypotheses suggest a mismatch between the actual and perceived
effects of habit and inner states. Such a mismatch could take two forms. In the strong version of
these hypotheses, habit would outperform inner states in predicting behavior, but attributions
would implicate inner states more than habit. Such a pattern would suggest a stark discrepancy
between actual and perceived causes of behavior. In the weak version of these hypotheses, either:
(a) habit and inner states predict behavior equally well but attributions implicate inner states
more than habit, or (b) habit predicts behavior better than inner states but attributions implicate
both factors equally. Such patterns would both capture the basic prediction that attributions
overweight the importance of inner states and underweight the importance of habit.

LAY THEORIES OF BEHAVIOR 13

Study 1
This online experiment formed an initial test of our hypotheses using reported intentions
instead of actual behavior as our key outcome. We experimentally manipulated participants’
mood by randomly assigning them to read a job application scenario with either a successful,
failed, or neutral outcome. Participants then reported intentions to buy coffee in that scenario,
their attributions for these intentions, and their habit strength for buying coffee. Following the
first two hypotheses, we anticipated divergent patterns for actual and attributed causes of buying
intentions: Actual buying intentions should be predicted more strongly by habit compared with
mood, whereas participants’ attributions should weight mood more than habit.
Intentions are, of course, a proxy for actual behavior, and they may not be influenced by
mood and habitual associations in the same way as actual performance. Nonetheless, we used
this initial study as a preliminary assessment of the viability of comparing participants’
attributions for their behavioral inclinations with predictive models of these tendencies.
Method
Power Analyses. R code used for the power analyses for all studies can be found in the
following link: https://osf.io/5xfsm/?view_only=ba7cacb9c9e047c1ad60fdd137f6d84a. In
studies 1 and 2, we conducted analyses for regression analyses (for the behavior analyses) as
well as t-tests (for the attribution analyses). Power analysis for a regression slope with a small-to-
medium effect size, f
2
= .10 revealed that 109 participants were needed to achieve 80% power.
Power analyses for a t-test with a difference in means of 5 and a standard deviation of 15
revealed that 72 participants were needed to achieve 80% power.

Participants. Amazon’s Mechanical Turk was used to recruit 150 U.S. residents (55
female, 95 male, mean age = 32.9 [SD = 8.97], range 19-68).
LAY THEORIES OF BEHAVIOR 14

Procedure. After giving informed consent, participants read and imagined themselves in
a scenario that began as follows:
You recently interviewed for a job you've been hoping to get for a long time. The
job is highly engaging and offers great salary and benefits. You feel that the
interview went fairly well, but the position is highly competitive as many other
applicants applied for it as well. While you're out of the house the next morning,
you receive a phone call....
In the successful scenario condition, the phone call informed participants that they were offered
the position. In the failed scenario condition, the call informed participants that they were not
offered it. In the control condition, the call informed participants that they would be notified of
the outcome in two weeks. All participants then rated their intentions to buy coffee after the
phone call. They then completed the attribution measure, manipulation check (PANAS), habit
strength measures (BFCS and SRBAI), and demographic information.
Measures.
Purchase intentions. Participants were asked: “After hanging up the phone, you pass by
your regular coffee shop - the one you frequent the most. How likely are you to go inside and get
a cup of coffee?” Answers ranged from 0% (certainly would not purchase coffee in that
situation) to 100% (would certainly purchase coffee in that situation).
Attribution. Participants responded to the following prompt: “Please rate how important
[habit] and [mood] would be in driving your decision to get or not get coffee in this situation.
Rate the percent to which your decision was driven by each factor.” They then rated mood (“my
feelings and emotions at the time”) and habit (“my habits and patterns of behavior in the past”)
on a scale from 0% (behavior was not influenced by this factor) to 100% (behavior would be
LAY THEORIES OF BEHAVIOR 15

completely determined by this factor). The sum was constrained to equal 100%. Thus, a score of
50 represented equal attribution to mood and habit, scores between 51-100 represented greater
attribution to mood than habit, and scores between 0-49 represented greater attribution to habit
than mood.
Habit strength. Participants completed two measures of habit strength. The Behavioral
Frequency and Context Stability measure (BFCS; Ji & Wood, 2007) measures habit strength as
the product of repetition and context stability. Participants were asked “how often do you get
coffee at [your favorite] coffeeshop?” (once a month or less/ several times a month/ several times
a week/ every day or multiple times a day). The Self-Report Behavioral Automaticity Index
includes a subset of four items that measure habit strength in terms of perceived automaticity
from the Self-Report Habit Index (SRBAI; Verplanken & Orbell, 2003). On 5-point scales
ranging from “strongly disagree” to “strongly agree,” participants answered items such as:
“getting coffee at [my favorite] coffee shop is something that I do without thinking.” The four
items were averaged to create a single composite SRBAI score for each participant (α = .96).
Manipulation check. To confirm that the manipulation was effective in eliciting positive
and negative affect, participants completed an abbreviated set of items from the Positive And
Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988). The scales measured
four positive emotions (“happy,” “excited,” “proud,” and “surprised”) and four negative
emotions (“sad,” “disappointed,” “distressed,” and “ashamed”). Because one item, “surprised,”
had a low item-whole correlation, r = .44 compared with rs = ~.9 for the other items in the scale,
this item was dropped from subsequent analyses. Both scales showed high reliability (αs = .93
and .91 for positive and negative emotion, respectively).

LAY THEORIES OF BEHAVIOR 16

Results
Descriptive statistics and correlations between key variables appear in Tables 1 and 2.
Suggesting that random assignment was successful, habit strength, age, sex ratio, and ethnicity
did not differ by condition, all ps > .1. Furthermore, participants across conditions did not differ
in the strength of their coffee drinking habits, with scores on the BFCS and SRBAI revealing
moderate automaticity of coffee drinking. Suggesting that the mood manipulation was
successful, participants in the job success scenario reported more positive affect and less
negative affect than the other 2 conditions, whereas participants in the job failure scenario
reported more negative affect and less positive affect than the other 2 conditions (all p’s < .03).
Intentions to drink coffee in the scenario were favorable on average (M = 68.16, SD = 28.34).
Attributions were close to the midpoint of the scale (M = 53.07, SD = 27.48). Our two measures
of habit – the BFCS and SRBAI – substantially correlated with each other (r = .59, p < .001).
Because the hypothesis testing results below were unchanged regardless of the habit strength
measure used, subsequent results are reported using BFCS due to its higher predictive validity
(Labrecque & Wood, 2015).
To test the first hypothesis, we fitted a linear regression model predicting purchase
intentions from habit strength, interview outcome (dummy coded as comparisons with the
neutral condition), and interactions between interview outcome and habit strength. As predicted,
habit strength strongly predicted purchase intentions, b = 2.41, SE = 0.95, t(144) = 2.52, p =
.0125 (see Table 3). Experimental condition, however, had little to no effect on intentions.
Participants in the success condition did not differ in their intentions from those in the control
condition, b = 4.60, SE = 5.24, t(146) = 0.87, p = .381. An unexpected marginal effect indicated
that participants in the failure condition reported slightly lower coffee purchase intentions
LAY THEORIES OF BEHAVIOR 17

compared with participants in the control condition, b = -9.40, SE = 5.26, t(146) = -1.78, p =
.076. Habit strength did not significantly interact with interview outcomes, both ps > .1.
To assess the second hypothesis that participants will attribute to mood more than habit,
we used a one-sample t-test to evaluate whether the mean of attribution scores was different from
50, representing equal attribution to mood and habit. Although the overall sample mean did not
differ from the scale midpoint, t(149) = 1.36, p = .172, we conducted an additional analysis with
only the success and failure outcome conditions, which were exposed to emotionally evocative
content. As predicted, these participants assigned greater weight to affect compared with habit in
influencing their intentions, t(100) = 2.42, p = .017.
Finally, we conducted several exploratory analyses to evaluate additional patterns in our
results. First, we replicated our main regression analysis by substituting PANAS ratings for
interview outcome condition with PANAS ratings, based on the assumption that PANAS scores
might measure mood more reliably than experimental condition. Thus, the resulting regression
model predicted purchase intentions from habit strength and each participant’s positive and
negative PANAS ratings. As in the original analysis, habit strength significantly predicted
purchase intentions, b = 2.22, SE = 0.56, t(144) = 3.91, p < .001. Also as hypothesized, negative
affect had little effect, b = -1.37, SE = -0.05, t(144) = -0.59, p = .551, and did not interact with
habit strength, b = -0.86, SE = 0.64, t(144) = -1.34, p = .181. Unexpectedly, positive affect did
predict purchase intentions, b = 4.90, SE = 1.84, t(144) = 2.65, p = .008. Furthermore, positive
affect interacted with habit strength, such that participants with weaker habits were more likely
to intend to purchase when feeling good, b = -1.11, SE = .51, t(144) = -2.16, p = .031. This
pattern broadly replicates earlier research in which inner states predicted behavior mostly for
people who did not have strong habits (Gardner, 2015; Ouellette & Wood, 1998).
LAY THEORIES OF BEHAVIOR 18

Discussion
Study 1 examined the effects of mood and habit strength on coffee purchase intentions in
a hypothetical job interview scenario with either a success, failure, or neutral outcome.
Regardless of the interview’s outcome, participants with stronger habits had stronger intentions
to purchase. Nevertheless, when explaining the reasons for their intended behavior, participants
assigned equal or greater weight to their mood than to habit. These findings support the
hypothesis that people underestimate the role of automatic influences such as habit in favor of
explanations involving internal states. By randomly assigning participants to the different mood
conditions, this study provided a highly controlled test of our hypothesis.
Despite this support for our hypotheses using manipulated job outcome as an indicator of
mood, the results were not as clear in analyses with experienced emotion. Because participants’
purchase intentions were predicted in our models by good feelings (i.e., PANAS scores), both
intentions and attributions were influenced similarly by mood states. However, it is worth
keeping in mind that this first study design elicited simulated coffee purchase decisions in a
hypothetical choice scenario. Because hypothetical paradigms omit or minimize the context cues
that trigger habit performance (e.g., the sight of a familiar storefront), they may underestimate
the effects of habit compared with assessments of actual behavior (Herziger & Hoelzl, 2016).
Moreover, intentions are often only weakly predictive of habitual behaviors such as the coffee
purchases examined in the present study (Webb & Sheeran, 2006; Wood, 2017). Thus, in our
second study, we tested our ideas in a more ecologically valid setting using actual, naturally-
occurring behavior as the primary outcome.

LAY THEORIES OF BEHAVIOR 19

Study 2
In Study 2, we measured the effects of habit and mood on whether students got food or
drink after exams. We evaluated purchases in this context because students were likely to
experience a salient inner state (positive or negative mood, presumably based on their exam
performance) that could explain their purchase. In addition, they should have had sufficient
opportunity to repeat this behavior in the past, thereby potentially forming a food- or drink-
purchase habit after exams. Thus, this was an ideal context in which to evaluate the actual mood
or habit predictors of behavior and compare these with participants’ beliefs about these
determinants of their behavior.
Study participants provided data at three time points: Several days before the exam (Time
1), they completed a survey that included a measure of habit strength for purchasing food or
drink after exams. Immediately after completing the exam (Time 2), they reported their current
mood. Finally, between 6 and 48 hours after the exam (Time 3), they reported whether they had
bought something to consume immediately after the exam as well as how important were habit
and mood in driving that purchase. To test our first hypothesis about the actual causes of food or
drink purchase, we used logistic regression to predict likelihood of purchase from habit strength
and post-exam mood. We expected habit strength to predict purchases after the exam better than
mood. To test the second hypothesis concerning participants’ attributions, we compared within-
person differences between mood and habit attributions for food and drink purchases (as in Study
1). We anticipated that participants would judge their mood to be more important than habit in
driving their purchase decisions.

LAY THEORIES OF BEHAVIOR 20

Method
Participants. We recruited 258 undergraduate psychology students from the University
of Southern California for a study on “students’ experience before and after exams.” Of these, 62
participants failed to complete surveys at one or more of the three time-points included in the
study, and thus were not included in the analyses (76% retention rate). To ensure that
participants’ mood ratings reflected concurrent post-exam mood and not retrospective
recollections, we excluded 23 additional participants who reported their mood more than four
hours after the start of the exam. Nevertheless, post-hoc sensitivity analyses revealed that
including these participants did not notably impact the results. Thus, the final sample size for
analyses consisted of 173 participants (120 female, 51 male, 2 genderqueer/other, mean age =
20.83 [SD = 4.13], range 18-58).
Procedure. Participants completed surveys at three timepoints. The first survey (Time 1),
completed between two and four days before the exam, included a variety of irrelevant measures
(to disguise its purpose) along with measures of habit strength for buying food or drink after
exams. The second survey (Time 2), which included our mood measure as well as an exploratory
measure of perceived test performance, was emailed to participants when the exam began.
Participants were instructed to complete this survey immediately after finishing the exam and
exiting the classroom. The final survey (Time 3), completed between 6-48 hours after the exam,
assessed whether participants purchased food or drink right after the exam as well as their
attributions for this purchase.
Measures.
Habit strength (Time 1). Participants rated how often they performed a number of
behaviors right after taking an exam, defined as “up to two hours after completing the exam,”
LAY THEORIES OF BEHAVIOR 21

including “getting something to eat or drink outside of home.” Ratings were made on 5-point
scales ranging from “never” to “always. As in Study 1, the SRBAI provided an additional
measure of habit strength in which participants rated on 7-point scales (strongly disagree -
strongly agree) whether buying something to eat or drink immediately after an exam was
something that they do: “automatically,” “without thinking,” “without having to consciously
remember,” and “before [they] realize that [they’re] doing it.”
Mood (Time 2). Mood was rated in a visual analog scale with 100 points in which scale
anchors were a sad face (1) on the left end of the scale, a neutral face at the midpoint (50), and a
happy face on the right end (100). Participants moved the slider from the scale’s midpoint to
whichever point represented their current mood.
Food or drink purchase (Time 3). After the exam, participants answered yes/no to the
question “did you get something to eat or drink outside of home after the recent midterm exam
(up to two hours after completing the exam)?”
Attribution measure (Time 3). Participants rated how important were habit and mood in
driving them to get something to eat or drink after the exam. Each factor was rated on a 0-100%
scale on which 0% indicated that the factor had no effect on their behavior and 100% indicated
that their behavior was solely determined by that factor. Participants were instructed to not allow
the sum of both ratings (habit and mood) to total more than 100%.
Test performance (Time 2). In this exploratory item, participants indicated the grade that
they expected to receive on the exam from 0-100.
Results
Descriptive statistics and correlations between key variables appear in Table 4.
Approximately half (51%, or 89 out of 173) of participants reported purchasing something to eat
LAY THEORIES OF BEHAVIOR 22

or drink after their exam. Scores for both habit strength measures (BFCS and SRBAI) suggested
moderate to strong purchasing habits in our sample (MBFCS = 3.73, SD = 0.76; MSRBAI = 3.44, SD
= 1.59). These two habit measures were only moderately correlated, r = .42, p < .001.
Nevertheless, choice of habit measure did not have a noticeable impact on the results, and so all
analyses below use the BFCS. Mean post-exam mood was close to the midpoint of the scale,
although individual scores ranged from 0 to 100 (M = 57.22, SD = 25.42), reflecting the
expected variation in student feelings after a difficult exam. Finally, attribution ratings were both
around the scale’s midpoint, indicating that both habit and mood were considered plausible
drivers of food and drink purchases in this context.
To test the first hypothesis, a logistic regression predicted food or drink purchase (yes/no)
from habit strength (using the BFCS), post-exam mood, and an interaction between habit
strength and mood. Habit strength significantly predicted purchases, OR = 1.81, 95% CI [1.09,
3.01], p = .022, but mood did not, OR = 1.00, 95% CI [0.98, 1.01], p = .546. In addition, no
interaction was evident between mood and habit strength, OR = 1.00, 95% CI [0.99, 1.02], p =
.842.
The second hypothesis concerns participants’ explanations for their food purchases rather
than lack of food purchases, and therefore the primary analysis on attributions was restricted to
the 89 participants who did purchase something to eat or drink after the exam. To test whether
these participants attributed their behavior to mood more than habit, a paired-samples t-test
compared each participant’s mood attribution rating with their habit attribution rating. A positive
difference score represents attribution to habit over mood, and a negative score represents
attribution to mood over habit. No significant difference emerged between the two, Mdifference =
0.05, t(88) = 0.01, p = .989. Sensitivity analyses revealed a similar pattern when all participants
LAY THEORIES OF BEHAVIOR 23

were included, Mdifference = -3.67, t(172) = -1.14, p = .255.
Exploratory Analyses.
In addition, we conducted an exploratory logistic regression analysis predicting food or
drink purchase (yes/no) from habit strength and test performance. Although we assumed that
perceived test performance would be associated with mood, this variable captures a different
aspect of students’ momentary experience in that it is an external outcome, analogous to the job
interview outcome in Study 1. As in our main model, habit strength strongly predicted food or
drink purchase after the exam, OR = 1.75, 95% CI [1.23, 2.47], p = .002. Test performance
marginally predicted purchases, so that participants with worse perceived test performance were
slightly more likely to purchase food or drink after the exam, OR = 0.73, 95% CI [0.52, 1.02], p
= .063. Thus, although we found no evidence for an effect of mood on behavior in the main
model, perceived test performance had a marginal effect. This effect, however, was substantially
smaller than the effect of habit strength, which is evident in the much-larger odds-ratio for habit,
as well as the non-overlapping confidence intervals for each effect. In sum, these analyses
replicate the primary findings of Study 1, in that test performance effects on behavior were
dwarfed in magnitude by habit strength.
Discussion
In Study 2, we compared how well habit strength and in-the-moment mood predicted
students’ food or drink purchases after a challenging exam. Supporting our first hypothesis,
participants with stronger purchase habits were more likely to purchase food or drink after the
exam, whereas participants’ mood had little impact. Thus, behavior in this study was largely a
function of habit strength.
LAY THEORIES OF BEHAVIOR 24

Our central question, from the second hypothesis, is whether participants recognized
these patterns. And the findings are clear—they did not. Participants’ beliefs about the causes of
their actions were in stark contrast to the actual determinants of behavior. Participants judged
that food or drink purchases were due to habit and mood relatively equally. Thus, they may have
been correct about habits driving purchase, but they also believed that they purchased food and
drink after an exam out of mood—presumably to celebrate when in a good mood or to
compensate for being in a bad mood. Although this second finding did not confirm the strong
version of Hypothesis 2, in which participants’ attributions favor mood and other salient inner
states over habit, the overall pattern revealed poorly calibrated attributions in the anticipated
direction. That is, participants’ attributions overly emphasized their happy or sad feelings after
the exam while discounting habit.
Study 3
The first two studies established that lay perceptions overestimate the influence of mood
on behavior while underestimating the influence of habit. However, these studies tested
explanations for a single behavior immediately following an emotionally evocative experience. It
is perhaps understandable that, when focused on a recent event like learning about a potential job
or taking a difficult test, people overly rely on their feelings. Being preoccupied with the
implications of the event, they overestimate the extent to which their powerful feelings are
driving their actions.
Study 3 provided a stronger test of the idea that people overlook the power of habit when
explaining everyday actions. It assessed a mundane behavior, coffee drinking, that is not usually
associated with strong emotions. Our participants were occasional or habitual coffee drinkers
who should have considerable experience with the conditions under which they drink coffee. We
LAY THEORIES OF BEHAVIOR 25

assessed their coffee drinking in response to fatigue and low energy or to habit, which our pilot
data identified as two of the most common factors believed to drive coffee drinking (see below).
To capture experiences and explanations as they naturally unfold in daily life, we used a
combination of survey measures and EMA. Participants first completed intake surveys, including
measures of habit strength and attributions for their own coffee drinking. Then, over the course
of a week, they reported every two hours on their fatigue and coffee drinking.
As in previous studies, we tested the actual associations between causes (i.e., habit,
fatigue) and behavior (i.e., coffee drinking), and then compared these with the perceived
associations reflected in participants’ causal attributions. Following our first two hypotheses and
prior study results, we anticipated that habit would be the primary predictor of coffee drinking,
but that explanations would focus more on fatigue.
This design improved on our first two studies in several ways. First, the influence of
fatigue on coffee drinking was estimated from the lagged associations between fatigue at a given
prompt and coffee drinking by the next prompt (i.e., in the following two hours). This lagged
design tests our hypotheses without the response biases associated with concurrently reporting a
predictor and outcome, such as fatigue judgments being shaped by lay theories (e.g., “I drank
coffee, therefore I must have been tired earlier;” see Mitchell, Thompson, Peterson, & Cronk,
1997). In addition, we elicited attributions twice: once at the beginning of the study (as in the
first two studies), and a second time at the end, when participants were financially incentivized
to provide accurate attributions of their coffee drinking during the study week. Pre-registered
hypotheses and analysis plans for this study can be found in the following link:
http://aspredicted.org/blind.php?x=xn3cg8.

LAY THEORIES OF BEHAVIOR 26

Pilot
To assess common lay beliefs about the causes of coffee drinking, 40 University of
Southern California undergraduates (22 male, 16 female, 2 genderqueer or other) rated the
importance of six potential causes of coffee drinking. On a 5-point scale ranging from “not at all
important” to “extremely important,” participants indicated “how important is [X] in driving
people to drink coffee?” The drivers assessed were fatigue (“tiredness or low energy”), habit
(“habit or behavior routines”), thirst, taste, social motives (“spending time with friends”) and
having coffee after a meal. Fatigue was the most important driver of coffee drinking (M = 4.05,
SD = 0.95), followed by taste (M = 3.57, SD = 1.03), habit (M = 3.50, SD = 1.21), social motives
(M = 3.17, SD = 1.08), having coffee after a meal (M = 2.12, SD = 1.22), and thirst (M = 1.70,
SD = 0.91). A paired-samples t-test comparing fatigue and habit attributions (within
participants) revealed that participants attributed coffee drinking to fatigue significantly more
than habit, t(39) = 2.71, p = .009.
Method
Power Analyses. In the current study, we maximized power by using an intensive
longitudinal design that measured each participant multiple times. We calculated power for our
multilevel logistic regression by simulating a dataset with standardized regression coefficients of
0.3 and 0.2 (corresponding to odds-ratios of 1.35 and 1.22) for a level 2 and level 1 variable,
respectively. We then simulated sampling from this dataset 1,000 times. This demonstrated that
50 participants were sufficient to achieve 90% power for our between-subject variable (habit
strength) and 99.5% power for our within-person variable (fatigue).
For participating in a study on students’ daily experiences, 151 University of Southern
California students received course credit or monetary compensation. The full selection criteria
LAY THEORIES OF BEHAVIOR 27

were: (a) speaking English fluently, (b) owning a smartphone with internet access, (c) being 18
or older, and (d) drinking coffee once a week or more often. Out of our initial sample, 35
participants reported drinking coffee only once or less during the study period and were therefore
excluded from subsequent analyses. Two additional participants answered less than 50% of
EMA prompts and were excluded as well. Thus, the final sample size for analyses consisted of
114 students (87 female, 27 male, mean age = 20.8 [SD = 2.8], range = 18-33), which is slightly
smaller than our preregistered target of 120.
To minimize attrition, amount of compensation was linked to compliance. Paid
participants received $20 if they completed 80%-100% of EMA prompts, $15 if they completed
50%-80%, and $5 if they completed less than 50%. Participants who received course credit had a
similar 3-tier compensation system.
Procedure
The study involved three parts: An in-person intake session, an EMA tracking period
starting a day after the intake session, and a final survey completed a day after the end of the
EMA tracking period.
Intake session. After informed consent, participants completed measures of coffee
drinking habit strength (BFCS and SRBAI), coffee drinking attributions, and demographics. In
addition, to obscure the purpose of the study and limit reactivity, participants also answered an
identical set of measures about soft drinks.
Participants then learned about the EMA procedure and final survey. They formed
implementation intentions to overcome obstacles for completing EMA prompts (e.g., “if my
phone beeps when I am with people, then I will excuse myself and answer the prompt;”
Adriaanse, Vinkers, De Ridder, Hox, & de Wit, 2011).
LAY THEORIES OF BEHAVIOR 28

EMA period. EMA data were collected for five weekdays (i.e., Monday-Friday), with 8
prompts per day at regular two-hour intervals from 8am to 10pm. Each prompt included decoy
items to obscure the purpose of the study, including location (e.g., home, campus) and
temperature (hot, cold, or comfortable). Participants then reported how tired they were, whether
they drank coffee in the past two hours, and whether they drank soft drinks in the past two hours.
Participants also completed an exploratory mood item and an open-response item in which they
briefly described their current situation.
In addition, because fatigue immediately after waking up may be particularly important
for coffee drinking, we asked participants to complete an additional EMA prompt immediately
after they got out of bed in the morning. These morning prompts included the same items as the
regular EMA prompts as well as an item asking whether they had already drunk coffee since
waking up. This last item ensured that we measured the prospective effect of waking fatigue on
coffee drinking, rather than the concurrent association between the two (which, as explained
above, could potentially bias lay theories).
At the end of the first day of EMA prompts, participants with response rates of 50% or
above were informed of their rough level of compliance via email (50-75% or 75-100%). Those
with less than 50% compliance were contacted by phone and/or text message to address potential
technical difficulties that might have led to low compliance.
Final survey. The final survey tested an exploratory measure of habit strength based on
the specific coffee drinking and non-drinking incidents students reported (see context-specific
habit measure below).
Follow-up survey. Shortly after the study concluded, participants again completed the
attribution measure (from the intake session). To provide an especially strong test of our second
LAY THEORIES OF BEHAVIOR 29

hypothesis that people will misattribute to fatigue, they were offered a monetary incentive of $3
if they accurately estimated the effects of fatigue and habit strength on their own coffee drinking
during the study week. Moreover, to more closely match the perceived and actual effects of habit
and fatigue on coffee drinking, we asked participants to report how frequently they drank coffee
in any 2-hour period during the study, conditional on the level of fatigue that they reported
immediately before this time period. This measure and related analyses are explained in detail
below, under ancillary analyses.
Measures
Additional measures for this study, as well as missing data analyses, are included in the
supplemental materials.
Intake measures
Habit strength. As in Study 2, habit strength was measured using the BFCS and SRBAI,
with getting a coffee as the target behavior.
Attribution. On scales ranging from 0 – 100%, participants rated the extent to which their
coffee drinking was driven by habit (“my past behavior and habits”) and then by fatigue (“my
energy levels and tiredness”). Anchors of 0% indicated that coffee drinking was unaffected by a
factor and 100% that coffee drinking was completely determined by a factor. Participants were
instructed that the sum of both ratings should not exceed 100%.
EMA measures
Fatigue. Participants rated how tired they were on a 6-point scale ranging from “not at
all” to “extremely” (Hooff, Geurts, Kompier, & Taris, 2007).
Coffee drinking. Participants indicated whether they started drinking coffee in the past
two hours. This ensured that a single coffee consumed over a long period of time was
LAY THEORIES OF BEHAVIOR 30

categorized as one episode. Answer choices were: “No,” “Yes – 1 drink,” “Yes – 2 drinks,” and
“Yes – 3 drinks.” The latter three categories were collapsed to create a binary (drink/did not
drink) indicator of coffee drinking.
Final survey
Context-specific habit measure. To test a novel measure of habit, we used an automated
script to randomly select seven EMA reports that did not directly precede coffee drinking and up
to seven that directly preceded coffee drinking. For each situation description, participants rated
how often they drank coffee in that situation as well as the strength of their intentions to drink
coffee in that situation. A full description of the measure can be found in the supplemental
materials.
Single-event attribution measure. This measure was included to account for possible
differences in attribution for a specific instance of a behavior as opposed to behavior in general.
Participants were presented with a context description of their most recent coffee drinking event
(see supplemental materials for details) and rated to what extent habit and fatigue contributed to
their coffee drinking in that situation. Item wording and answer choices were the same as the
main, intake attribution measure. To ensure that participants could recall the coffee drinking
event in question, they reported whether they remembered the event, with possible responses
being “yes,” “no,” and “not sure.”
Follow-up
Participants completed the same attribution measure used in the intake survey. In
addition, participants read the following prompt: “Using the items below, estimate how often you
drank coffee following each tiredness score. For each tiredness score, estimate what was the
likelihood of you drinking coffee in the 2 hours after mini-surveys in which you reported the
LAY THEORIES OF BEHAVIOR 31

corresponding tiredness score.” Participants then rated, for each fatigue score from 1-6, their
conditional coffee drinking likelihood from 0%-100%. Participants then estimated how many
times they reported each tiredness score during the study, from 0 to 40 (the maximum number of
prompts that could be answered during the study).
Results
Means, standard deviations, and between-person correlations for key variables appear in
Table 5. The 114 participants (level-2 sample size) produced 4323 individual observations
(level-1 sample size), corresponding to an average response rate of 33.5 out of 40 EMA prompts
(84%). On average, participants drank coffee a little over five times during the five-day EMA
period (M = 5.34, SD = 3.26), or approximately once a day. Scores on both the BFCS and
SRBAI suggested moderate coffee drinking habits. The BFCS and SRBAI strongly correlated
with each other (r = .74). As in Studies 1 and 2, choice of habit measure did not have a
noticeable impact on the results, and thus analyses are reported using the BFCS.
Main analyses. Results were analyzed using the following multilevel model:

Level 1:
𝑐𝑜𝑓𝑓𝑒𝑒 𝑖𝑗
= 𝛽 0𝑗 + 𝛽 1𝑗 𝑓𝑎𝑡𝑖𝑔𝑢𝑒 _𝑐𝑚𝑐 𝑖𝑗
+ 𝑒 𝑖𝑗

Level 2:

𝛽 0𝑗 = 𝛾 00
+ 𝛾 01
ℎ𝑎𝑏𝑖𝑡 𝑗 + 𝛾 02
𝑚𝑒𝑎𝑛 _𝑓𝑎𝑡𝑖𝑔𝑢𝑒 𝑗 + 𝛾 03
𝑙𝑎𝑦𝑏𝑒𝑙𝑖𝑒𝑓 𝑗 + 𝑢 0𝑗
𝛽 1𝑗 = 𝛾 10
+ 𝛾 11
ℎ𝑎𝑏𝑖𝑡 𝑗 + 𝛾 12
𝑙𝑎𝑦𝑏𝑒𝑙𝑖𝑒𝑓 𝑗 + 𝑢 1𝑗

Where i and j represent observations (i) nested within persons (j); coffee refers to whether the
participant did or did not report drinking coffee in the following prompt (i.e. a lead indicator of
coffee drinking meant to capture the prospective association between fatigue and coffee drinking
LAY THEORIES OF BEHAVIOR 32

in the following two hours); fatigue_cmc is a person mean-centered fatigue rating at each EMA
prompt, computed by subtracting each participant’s mean level of fatigue from each fatigue
ratings, so that positive values reflect higher-than-average fatigue for that person whereas
negative values reflect lower-than-average fatigue for that person; mean_fatigue is each person’s
mean level of fatigue, habit is each participant’s habit strength score, and laybelief is each
person’s attribution of coffee drinking to mood. To allow the within-person association between
fatigue and coffee drinking to vary between participants, the model included random slopes for
fatigue (𝛽 1𝑗 in the model equation).
Model estimates for the primary multilevel model are shown in Table 6. Due to
convergence issues with the original frequentist model, we re-specified the main model as a
Bayesian model. Because we did not want to impose restrictive priors on the results, we
specified uninformative priors for all slopes (a prior mean of 0 with a standard deviation of
1,000; Klauenberg, Wübbeler, Mickan, Harris, & Elster, 2015).
A plot of predicted coffee drinking likelihood based on model estimates appears in Figure
1. The inter-class correlation was .10, meaning that person-level variance explained 10% of the
overall variance in coffee drinking. To test whether habit predicted coffee drinking better than
fatigue (H1), we compared the standardized coefficients for habit strength and person-mean
centered fatigue (𝛽 1𝑗 and 𝛾 01
in the model). Habit strength predicted coffee drinking, OR = 1.40,
95% CI [1.18, 1.61]. Fatigue (within-person) predicted coffee drinking as well, OR = 1.31, 95%
CI [1.11, 1.56]. None of the other effects in the model (between-person fatigue, fatigue
attribution, and interactions between within-person fatigue, habit strength, and fatigue
attribution) predicted coffee drinking.
LAY THEORIES OF BEHAVIOR 33

To test an alternative explanation by which the fatigue-coffee drinking association only
holds early in the day, we computed a separate multilevel analysis in which waking fatigue (as
measured in morning prompts) predicted coffee drinking at the first prompt of each day. Our
final sample size for this analysis consisted of 102 participants (level-2 sample size) and 313
responses (level-1 sample size, corresponding to participant-days). Out of the original 354
responses, 41 were excluded because participants reported that they already drank coffee by the
time that they completed the prompt. Early morning fatigue did not predict coffee drinking at the
following prompt, OR = 1.08, 95% CI [0.78, 1.49], p = .650. Thus, when focusing on waking
fatigue and the first coffee-drinking report of the day, we did not find an association between
fatigue and subsequent coffee drinking.
To test whether participants attributed coffee drinking to fatigue more than habit (H2), we
computed a paired-samples t-test comparing the within-person difference between each
participant’s fatigue and habit attributions. Providing support for the strong version of this
hypothesis, fatigue attributions were significantly higher than habit attributions M difference = 32.73,
t(113) = 8.23, p < .001. Additional analyses for Hypothesis 2 can be found below.
Exploratory analyses.
In the final survey, 79 out of 114 eligible participants (69%) reported recalling their most
recent coffee drinking event, as displayed in the single-event attribution measure. Echoing results
from our main attribution measure, a paired-samples t-test revealed that participants attributed
their most recent coffee drinking event to fatigue more than habit, Mdifference = 24.67, t(78) = 4.77,
p < .001. Our follow-up survey, which was completed by 81 participants (71%), financially
incentivized them to report attributions that accurately described their coffee drinking patterns
during the study week. Again, a paired-samples t-test compared each participant’s fatigue
LAY THEORIES OF BEHAVIOR 34

attribution rating with their habit attribution rating. Once again, fatigue attributions were
significantly higher than habit attributions, M difference = 15.91, t(80) = 3.13, p = .002. Thus,
although incentives reduced the difference between habit and fatigue in attributions of coffee
drinking (Mdifference ≈ 16 when incentivized compared with Mdifference ≈ 33 in the non-incentivized
measure), incentivized participants still overwhelmingly rated fatigue as more important than
habit.
Additionally, our analyses yielded several findings of interest for which we did not have
a-priori analysis plans. As mentioned above, our findings revealed no evidence for a prospective
association between fatigue and coffee drinking. One may wonder, however, whether the
opposite prospective association is true – that is, whether drinking coffee made our participants
less tired. To answer this question, we constructed a Bayesian multilevel model predicting
fatigue at a given timepoint from coffee drinking reported at the previous timepoint. Analyses
revealed a modest effect of coffee drinking on fatigue, b = -0.11, 95% CI [-0.01, -0.22], so that
participants reported slightly less fatigue if they drank coffee in the previous 2-4 hours. Of
course, it may be that 2-4 hours is too long of a time lag in order to detect a tonic effect of
caffeine. However, this relation can also be tested with concurrent measures by predicting
fatigue from reports of coffee drinking in the previous two hours. Even in these concurrent
reports of fatigue and coffee consumption, the association increased only slightly, b = -0.22, 95%
CI [-0.08, -0.36]. Thus, coffee drinking only helped to alleviate fatigue to a modest degree, even
in concurrently assessed measures for which one could expect an association based on lay
theories alone (i.e. “I drank coffee so I must be less tired”).
In addition, we replicated the main MLM model using the exploratory context-specific
measure of habit that participants completed during the final survey. As noted above, this
LAY THEORIES OF BEHAVIOR 35

measure randomly sampled participants’ situation descriptions for coffee drinking events and
non-events and obtained ratings for each situation of their coffee drinking frequency and
automaticity. The MLM model predicted coffee drinking at each prompt (yes/no) from fatigue
(within-person) and habit (situation-specific coffee drinking frequency). This analysis has the
advantage of operationalizing both habit and fatigue as within-person variables, making them
more comparable. Providing even stronger evidence for our first hypothesis than the more
standard measures of habit strength, habit operationalized as context-specific frequency
predicted coffee drinking, OR = 1.68, 95% CI [1.19, 2.27], but fatigue did not, OR = 1.20, 95%
CI [0.94, 1.47]. It should be noted, however, that this context-specific measure has not yet been
validated in other research, and so these conclusions are necessarily preliminary.
Discussion
In this third study, participants explained the causes of their coffee drinking and then
tracked their momentary fatigue and coffee drinking over the course of a week. This allowed us
to compare participants’ explanations for a mundane everyday action against the actual
predictors of that action. Consistent with the first two studies and with Hypothesis 1, the
association between fatigue and coffee drinking over the course of a typical week was equal to
the association between habit strength and coffee drinking. Supporting Hypothesis 2, participants
mainly attributed their coffee drinking to fatigue rather than habit, and this inaccuracy in
attribution persisted even when participants were incentivized for accuracy or asked about a
specific coffee drinking event rather than coffee drinking in general.
Notably, we found evidence for misattribution in Study 3 despite several design features
that should have reduced misattribution. First, this study’s sample consisted of occasional to
frequent coffee drinkers. Thus, they should have had frequent opportunities to observe their
LAY THEORIES OF BEHAVIOR 36

coffee drinking and understand its antecedents. Indeed, the frequency of their consumption (once
a day on average) should in itself serve as a plausible cue that their behavior is habitual. Second,
unlike our first two studies that assessed attributions after a salient life challenge (job interview,
difficult test), this third study revealed that people underestimate habit and overestimate inner
states even when behavior is not preceded by a salient, emotionally evocative event. Finally, this
study included attribution measures that incentivized participants to report the antecedents of
their own coffee drinking accurately, with accuracy measured against their actual self-reported
data from the EMA period. Even with a financial incentive and an objective criterion, however,
participants’ attributions were biased in favor of inner states.
General Discussion
The present paper compared the actual and perceived effects of habit and mood on eating
and drinking behavior in everyday life. In Study 1, a hypothetical scenario design tested the
effects of mood and habit on food purchase intentions in a controlled setting. In Study 2, we
extended this finding to a field study using naturally occurring food and drink purchases and a
momentary assessment of mood. Finally, in Study 3, we used EMA to track momentary fatigue
and coffee drinking and juxtaposed the actual and perceived effects of fatigue and habit on
coffee drinking.
In all studies, habit equaled or surpassed inner states in predicting behavior. This finding
aligns with past literature suggesting that eating and drinking tend to be performed automatically
and habitually (Cohen & Farley, 2007; Moldovan & David, 2012). Our participants’ attributions,
however, tended to emphasize inner states more than habit. Thus, participants’ lay theories
overemphasized inner states and underemphasized habit. This pattern is congruent with the
notion that habitual behavior, lacking an immediate valid explanation, prompts misattribution to
LAY THEORIES OF BEHAVIOR 37

more salient inner states (Neal et al., 2012). Participants may have favored mood due to cultural
beliefs about relevant introspected states (Pronin, 2009) as well as general information- and
motivation-based tendencies to interpret actions as goal-directed (Pronin, Wegner, McCarthy, &
Rodriguez, 2006).
The present paper tested these ideas in several ways. First, we used concurrent
assessment to obtain accurate estimates of behavior in Studies 2 and 3. Because recall for recent
events is less likely to be biased and subject to forgetting (Shiffman et al., 2008), and because the
behaviors that we examined (food and drink purchases, coffee drinking) were presumably neutral
in terms of social desirability, we can assume that our behavioral outcome measures were
reasonably accurate. This, in turn, allowed us to assess the accuracy of participants’ attributions
by comparing them to a precise behavioral benchmark. Second, the results in all studies
replicated using three distinct measures of habit: a self-report measure of behavioral repetition
and context stability (BFCS); a self-report measure of experienced automaticity (SRBAI); and in
Study 3, an exploratory context-specific habit measure tapping participants’ history of repetition
in a specific situation. The convergent findings across these different measures of habit strength
lends additional support for our claims.
Our findings extend understanding of the explanatory vacuum by suggesting that
misattribution is not limited to contexts in which automatic behavior elicits negative affect, such
as violating a personal health goal (e.g. Adriaanse, Weijers, Ridder, Huberts, & Evers, 2014).
Because the behaviors tested in the present research are presumably neutral in terms of social
desirability, it is unlikely that they violated a personal standard. It may be that the state of being
unable to explain a behavior is aversive in and of itself, so that simply asking participants to
LAY THEORIES OF BEHAVIOR 38

attribute their behavior motivated them to confabulate an explanation. Thus, an explanatory
vacuum may occur under a broader array of conditions than previously demonstrated.
Ancillary Analyses: Correlates of Inner State Attribution
We conducted several additional analyses to test possible boundary conditions or
moderators of our effects. In essence, we tested whether the findings hold for most of our
participants or reflect just the experience of particular segments or subgroups.
Attribution strength. We first considered the extent to which participants attributed their
behavior to inner states. It is possible that participants who made especially strong attributions to
mood or fatigue were more accurate in their assessments. That is, attributions might mirror actual
determinants of behavior for participants who strongly attributed to inner states, but not for those
who made weaker inner state attributions. We were able to test this possibility in Studies 2 and
3—the two studies for which we assessed experienced inner states (mood in Study 2, and fatigue
in Study 3) as well as actual eating and drinking behavior.
In Study 2, the logistic regression model predicting behavior was recomputed with two
added predictors: mood attribution and the interaction between mood attribution and post-exam
mood. If attributions are more accurate for those with stronger attributions, then we should
observe a significant interaction between mood and mood attributions, such that mood would
predict actual behavior better for people who made stronger mood attributions. However, no
such interaction emerged, OR = 1.00, 95% CI [1.00, 1.00], p = .798, and the main effect of mood
attributions was not significant, OR = 1.01, 95% CI [0.99, 1.02], p = .377.
In Study 3, we capitalized on a unique design feature: the intensive tracking of fatigue-
coffee drinking covariation. This allowed us to test whether participants with stronger within-
person associations between fatigue and actual coffee drinking attributed their coffee drinking to
LAY THEORIES OF BEHAVIOR 39

fatigue more than participants with weaker fatigue-coffee drinking associations. Specifically, we
examined the interaction between fatigue attribution scores and the within-person coefficient for
fatigue (𝛾 12
in the model). An odds-ratio larger than one for this term would indicate that people
who attribute their coffee drinking to fatigue exhibit stronger associations between fatigue and
subsequent coffee-drinking. In other words, an OR > 1 would indicate that attributions are
sensitive to the actual within-person lagged fatigue-coffee drinking associations. However, no
significant effects emerged for the interaction between fatigue attributions and within-person
fatigue-coffee drinking associations, OR = 1.00, 95% CI = [0.99, 1.00]. That fatigue attributions
were not associated with actual experience even using this individually tailored measure lends
further credence to the notion of a disconnect between attribution to inner states and experienced
inner states.
Finally, we analyzed Study 3 participants’ estimates of their individual fatigue – coffee
drinking association. We only included estimates for fatigue levels that participants: (a) reported
more than 2 times during the study, and 2) estimated (in the follow-up survey) reporting more
than 2 times during the study. Thus, if a participant reported being “extremely tired” only once
during the study, or estimated (in the follow-up survey) that they reported being “extremely
tired” only once (regardless of how frequently they did so during the study), their estimate was
excluded from analyses. This was done to reduce the impact of extreme, outlying values on the
association estimates. Out of the 81 participants who completed the follow-up survey, 67 (83%)
provided usable data. We then compared these participants’ estimates to their individual within-
person fatigue-coffee drinking association, represented by the random slopes in our main model.
As seen in Figure 2, we did not find any evidence for an association between perceived and
actual fatigue-coffee drinking associations, r(65) = -.12, p = .328.
LAY THEORIES OF BEHAVIOR 40

Strength of internal states. Our research also provides some insight into the
determinants of participants’ attributions for their behavior. Participants might have confabulated
reasons for their behavior primarily when unable to identify the inner states that actually drove
their responses. Such an effect is anticipated by self-perception theory’s claim that we are forced
to infer the reasons for our behavior “to the extent that internal cues are weak, ambiguous, or
uninterpretable” (Bem, 1972, p. 5). In contrast, participants with stronger, more salient inner
states may have based their causal inferences on these states, following temporal precedence
(i.e., if bad mood preceded eating, it must have driven that eating; Wegner & Wheatley, 1999).
To examine whether inner state intensity was correlated with attribution to inner states in
Study 2, we evaluated the relation between post-exam mood and attributions to mood. A linear
regression predicted attribution to mood from both the linear effect of post-exam mood (testing
whether mood attributions are affected by greater or lesser mood positivity), and the quadratic
effect of mood (testing whether mood attributions are dictated by mood intensity, regardless of
valence). Suggesting little influence of temporally salient experiences, no effect for mood
emerged: linear, b = -0.33, 95% CI [-1.08, 0.42], t(86) = -0.87, p = .386; quadratic, b = 0.00,
95% CI [0,0], t(86) = 0.29, p = .769. That is, participants with more extremely favorable moods
or more intense ones did not explain their behavior as due to mood any more than people with
weaker mood states.
In Study 3, we evaluated whether participants who experienced higher fatigue or larger
within-person shifts in fatigue were more likely to attribute their coffee drinking to this inner
state. To assess fatigue variability, we computed each participant’s within-person fatigue
standard deviation. We then examined whether mean fatigue and fatigue variability correlated
with attribution to fatigue. Mean fatigue was not associated with fatigue attributions, r(112) = -
LAY THEORIES OF BEHAVIOR 41

.03, p = .691. Fatigue variability was also not associated with fatigue attributions, r(112) = .01, p
= .870. Thus, attributions were not associated with how tired a participant was on average or
whether a participant experienced stronger fluctuations in fatigue levels. Therefore, in Studies 2
and 3, the salience of inner states was not associated with attribution to inner states. This
suggests that participants’ attributions were not strongly informed by experienced inner states.
In general, these ancillary analyses demonstrate that misprediction of behavior was not
limited to participants with uncertain or unclear understanding of the causes for their behavior.
Furthermore, participants’ attributions were not driven by a lack of clarity about their inner
states. Instead, the miscalibration of reasons for action seems to be a general phenomenon shared
by most of our participants.
Implications
Our present findings have several notable implications. First, the bias we observed in
behavioral attributions may have downstream effects on self-regulation. If people misattribute
the sources of their behavior, then they may select ineffective strategies when regulating that
behavior. For example, they might reduce food purchases by using mood-repair strategies instead
of situational self-control strategies that might be more successful in modifying habits. Indeed, a
similar bias may underlie the tendency to eschew situational self-control strategies despite their
efficacy: the automaticity of situational strategies makes them highly effective but
simultaneously reduces their salience, leading people to overlook these interventions’ potential
(Duckworth, Gendler, & Gross, 2016). Although inaccurate attributions may be harmful to self-
regulation, it is also possible that biased attributions serve to enhance self-regulation. For
example, underestimating external control of behavior may motivate action despite past failures
(Taylor & Brown, 1994).
LAY THEORIES OF BEHAVIOR 42

Our findings also carry notable implications for methodology in behavioral science.
People’s limited insight into their own behavioral antecedents underscores the limitations of self-
reports of behavioral causes. Although critique of introspection is hardly a new issue in
psychology (e.g. Watson, 1913), many published studies continue to rely solely on self-reports of
behavior without comparing them to an independent benchmark (Baumeister, Vohs, & Funder,
2007). We hope that these findings will reinforce a distinction between self-reports as reflections
of reality and self-reports as reflections of lay perceptions.

Sample and Generalizability Considerations
Our sample consisted of Amazon mTurk workers in Study 1, and Southern California
undergraduate students in Studies 2 and 3. Given that all samples were drawn from U.S.
populations, we cannot draw conclusions about generalizability to non-U.S. populations. For
example, people from East Asian cultures tend to attribute more to situational factors than do
Americans (Choi & Nisbett, 1998; Masuda & Kitayama, 2004). This may suggest that east
Asians will attribute to habits (which are situationally cued) more than Americans. However, we
did not have an a-priori hypothesis regarding cultural differences in attribution. We tested this
question in Study 3, in which a plurality of participants identified as Asian, allowing us to test
whether Asian participants differed from white participants in their behavior attributions. We
used independent-samples t-tests to compare participants who identified as Asian (55
participants, 48% of sample) and white (25 participants, 21% of sample) on attribution to fatigue
and attribution to habit. Neither difference was significant, both ps > .1. Thus, we did not
observe an effect of culture on attribution. This non-finding is qualified by the fact that in this
dataset, we could not distinguish between participants who were raised in an East-Asian culture
and participants who were Asian-American.
LAY THEORIES OF BEHAVIOR 43

In terms of gender, most of our Study 1 participants identified as male, whereas most
participants in Studies 2 and 3 identified as female. To examine gender effects, we added a main
effect for gender as well as interactions between gender and inner states into our main regression
models. We did not find any significant main effects of gender or significant interactions
between gender and other predictors, all ps > .1 We then used independent-samples t-tests to
examine gender effects in attribution in all 3 studies. No gender differences emerged, all ps > .1.
In terms of age, only the mTurk sample in Study 1 had enough variability to afford testing for
age effects. However, age in this study was not correlated with attributions, r(148) = -.08, p =
.314.
Conclusion
Although habits pervade much of our daily behavior, they exert their influence
surreptitiously, outside of conscious awareness. Inner states such as mood and fatigue
overshadow habit in lay attributions, even when their actual effect on behavior is minimal. The
present findings shed light on the ways in which people make sense of the automatic and
deliberate processes in their everyday experience. Furthermore, our findings suggest that
researchers interested in the determinants of such behavior pay greater attention to the ways that
these processes influence self-reported attributions. Although self-reported attributions may
certainly be useful in understanding human psychology, a productive science of attribution must
recognize that they can reflect lay theories as much as (or even more than) objective covariation
patterns in daily life.

LAY THEORIES OF BEHAVIOR 44

References
Adriaanse, M. A., Prinsen, S., de Witt Huberts, J. C., de Ridder, D. T. D., & Evers, C. (2016). ‘I
ate too much so I must have been sad’: Emotions as a confabulated reason for overeating.
Appetite, 103, 318–323. https://doi.org/10.1016/j.appet.2016.04.028

Adriaanse, M. A., Vinkers, C. D. W., De Ridder, D. T. D., Hox, J. J., & De Wit, J. B. F. (2011).
Do implementation intentions help to eat a healthy diet? A systematic review and meta-
analysis of the empirical evidence. Appetite, 56, 183–193.
https://doi.org/10.1016/j.appet.2010.10.012

Adriaanse, M. A., Weijers, J., Ridder, D. T. D. D., Huberts, J. D. W., & Evers, C. (2014).
Confabulating reasons for behaving bad: The psychological consequences of
unconsciously activated behaviour that violates one’s standards. European Journal of
Social Psychology, 44(3), 255–266. https://doi.org/10.1002/ejsp.2005

Amodio, D. M., & Ratner, K. G. (2011). A Memory Systems Model of Implicit Social
Cognition. Current Directions in Psychological Science, 20(3), 143–148.
https://doi.org/10.1177/0963721411408562

Aronson, E. (1969). The Theory of Cognitive Dissonance: A Current Perspective. In L.
Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 4, pp. 1–34).
https://doi.org/10.1016/S0065-2601(08)60075-1

Bar-Anan, Y., Wilson, T. D., & Hassin, R. R. (2010). Inaccurate self-knowledge formation as a
result of automatic behavior. Journal of Experimental Social Psychology, 46, 884–894.
https://doi.org/10.1016/j.jesp.2010.07.007

Baumeister, R. F., Vohs, K. D., & Funder, D. C. (2007). Psychology as the Science of Self-
Reports and Finger Movements: Whatever Happened to Actual Behavior? Perspectives
on Psychological Science, 2(4), 396–403. https://doi.org/10.1111/j.1745-
6916.2007.00051.x

Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in experimental
social psychology (Vol. 6, pp. 1-62). New York: Academic Press.
https://doi.org/10.1016/S0065-2601(08)60024-6

Beruchashvili, M., & Moisio, R. (2013). Is Planning an aid or an obstacle? Examining the role of
consumers’ lay theories in weight loss. Journal of Consumer Affairs, 47, 404–431.
https://doi.org/10.1111/joca.12014

Choi, I., & Nisbett, R. E. (1998). Situational Salience and Cultural Differences in the
Correspondence Bias and Actor-Observer Bias. Personality and Social Psychology
Bulletin, 24(9), 949–960. https://doi.org/10.1177/0146167298249003

LAY THEORIES OF BEHAVIOR 45

Cohen, D., & Farley, T. A. (2007). Eating as an automatic behavior. Preventing Chronic
Disease, 5(1). Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2248777/

Duckworth, A. L., Gendler, T. S., & Gross, J. J. (2016). Situational strategies for self-control.
Perspectives on Psychological Science, 11, 35–55.
https://doi.org/10.1177/1745691615623247

Gardner, B. (2015). Analysis of the use of ‘habit’ in understanding, predicting and influencing
health-related behaviour. Health Psychology Review, 9, 277-295.
https://doi.org/10.1080/17437199.2013.876238

Gardner, B., Abraham, C., Lally, P., & Bruijn, G.-J. de. (2012). Towards parsimony in habit
measurement: Testing the convergent and predictive validity of an automaticity subscale
of the Self-Report Habit Index. International Journal of Behavioral Nutrition and
Physical Activity, 9, 102. https://doi.org/10.1186/1479-5868-9-102

Herziger, A., & Hoelzl, E. (2017). Underestimated habits: Hypothetical choice design in
consumer research. Journal of the Association for Consumer Research, 2, 359-370.
https://doi.org/10.1086/694585

Hooff, M. L. M. van, Geurts, S. A. E., Kompier, M. A. J., & Taris, T. W. (2007). “How fatigued
do you currently feel?” convergent and discriminant validity of a single-item fatigue
measure. Journal of Occupational Health, 49, 224–234.
https://doi.org/10.1539/joh.49.224

Ji, M. F., & Wood, W. (2007). Purchase and consumption habits: Not necessarily what you
intend. Journal of Consumer Psychology, 17, 261–276.
https://doi.org/10.1080/10577400701542403

Kelley, H. H. (1973). The processes of causal attribution. American Psychologist, 28, 107–128.
http://dx.doi.org.libproxy2.usc.edu/10.1037/h0034225

Klauenberg, K., Wübbeler, G., Mickan, B., Harris, P., & Elster, C. (2015). A tutorial on
Bayesian Normal linear regression. Metrologia, 52, 878–892.
https://doi.org/10.1088/0026-1394/52/6/878

Labrecque, J. S., & Wood, W. (2015). What measures of habit strength to use? Comment on
Gardner (2015). Health Psychology Review, 9(3), 303–310.
https://doi.org/10.1080/17437199.2014.992030

Langer, E. J. (1975). The illusion of control. Journal of Personality and Social Psychology, 32,
311–328. http://dx.doi.org.libproxy1.usc.edu/10.1037/0022-3514.32.2.311

Loersch, C., & Payne, B. K. (2011). The situated inference model: An integrative account of the
effects of primes on perception, behavior, and motivation. Perspectives on Psychological
Science, 6, 234–252. https://doi.org/10.1177/1745691611406921
LAY THEORIES OF BEHAVIOR 46

Masuda, T., & Kitayama, S. (2004). Perceiver-induced constraint and attitude attribution in
Japan and the US: A case for the cultural dependence of the correspondence bias. Journal
of Experimental Social Psychology, 40(3), 409–416.
https://doi.org/10.1016/j.jesp.2003.08.004

Mazar, A., & Wood, W. (2018). Defining habit in psychology. In B. Verplanken (Ed.), The
Psychology of Habit: Theory, Mechanisms, Change, and Contexts (pp. 13–29).
https://doi.org/10.1007/978-3-319-97529-0_2

McGraw, K. M. (1987). Guilt following transgression: An attribution of responsibility approach.
Journal of Personality and Social Psychology, 53, 247–256.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-3514.53.2.247

McGuire, W. J., & McGuire, C. V. (1986). Differences in conceptualizing self versus
conceptualizing other people as manifested in contrasting verb types used in natural
speech. Journal of Personality and Social Psychology, 51, 1135–1143.
http://dx.doi.org.libproxy1.usc.edu/10.1037/0022-3514.51.6.1135

Mitchell, T. R., Thompson, L., Peterson, E., & Cronk, R. (1997). Temporal Adjustments in the
Evaluation of Events: The “Rosy View.” Journal of Experimental Social Psychology,
33(4), 421–448. https://doi.org/10.1006/jesp.1997.1333

Moldovan, A. R., & David, D. (2012). Features of automaticity in eating behavior. Eating
Behaviors, 13, 46–48. https://doi.org/10.1016/j.eatbeh.2011.09.006

Moors, A., & De Houwer, J. (2006). Automaticity: A theoretical and conceptual analysis.
Psychological Bulletin, 132, 297–326. http://dx.doi.org.libproxy1.usc.edu/10.1037/0033-
2909.132.2.297

Neal, D. T., Wood, W., Labrecque, J. S., & Lally, P. (2012). How do habits guide behavior?
Perceived and actual triggers of habits in daily life. Journal of Experimental Social
Psychology, 48, 492–498. https://doi.org/10.1016/j.jesp.2011.10.011

Nisbett, R. E., & Bellows, N. (1977). Verbal reports about causal influences on social judgments:
Private access versus public theories. Journal of Personality and Social Psychology,
35(9), 613–624. http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-3514.35.9.613

Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on
mental processes. Psychological Review, 84, 231–259.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0033-295X.84.3.231

Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple
processes by which past behavior predicts future behavior. Psychological Bulletin, 124,
54–74. http://dx.doi.org.libproxy1.usc.edu/10.1037/0033-2909.124.1.54

LAY THEORIES OF BEHAVIOR 47

Oettingen, G., Grant, H., Smith, P. K., Skinner, M., & Gollwitzer, P. M. (2006). Nonconscious
goal pursuit: Acting in an explanatory vacuum. Journal of Experimental Social
Psychology, 42, 668–675. https://doi.org/10.1016/j.jesp.2005.10.003

Parks-Stamm, E. J., Oettingen, G., & Gollwitzer, P. M. (2010). Making sense of one’s actions in
an explanatory vacuum: The interpretation of nonconscious goal striving. Journal of
Experimental Social Psychology, 46, 531-542. https://doi.org/10.1016/j.jesp.2010.02.004

Presson, P. K., & Benassi, V. A. (1996). Illusion of control: A meta-analytic review. Journal of
Social Behavior and Personality, 11, 493–510.

Pronin, E. (2009). The introspection illusion. In Advances in Experimental Social Psychology
(Vol. 41, pp. 1–67). https://doi.org/10.1016/S0065-2601(08)00401-2

Pronin, E., Berger, J., & Molouki, S. (2007). Alone in a crowd of sheep: Asymmetric perceptions
of conformity and their roots in an introspection illusion. Journal of Personality and
Social Psychology, 92, 585–595. http://dx.doi.org.libproxy1.usc.edu/10.1037/0022-
3514.92.4.585

Pronin, E., & Kugler, M. B. (2007). Valuing thoughts, ignoring behavior: The introspection
illusion as a source of the bias blind spot. Journal of Experimental Social Psychology, 43,
565–578. https://doi.org/10.1016/j.jesp.2006.05.011

Pronin, E., Wegner, D. M., McCarthy, K., & Rodriguez, S. (2006). Everyday magical powers:
The role of apparent mental causation in the overestimation of personal influence.
Journal of Personality and Social Psychology, 91(2), 218–231.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-3514.91.2.218

Riet, J. van’t, Sijtsema, S. J., Dagevos, H., & De Bruijn, G.J. (2011). The importance of habits in
eating behaviour. An overview and recommendations for future research. Appetite, 57,
585–596. https://doi.org/10.1016/j.appet.2011.07.010

Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being:
Informative and directive functions of affective states. Journal of Personality and Social
Psychology, 45, 513–523. http://dx.doi.org.libproxy1.usc.edu/10.1037/0022-
3514.45.3.513

Shah, J. (2003). Automatic for the people: How representations of significant others implicitly
affect goal pursuit. Journal of Personality and Social Psychology, 84, 661–681.
http://dx.doi.org.libproxy1.usc.edu/10.1037/0022-3514.84.4.661

Shiffman, S., Gwaltney, C. J., Balabanis, M. H., Liu, K. S., Paty, J. A., Kassel, J. D., … Gnys,
M. (2002). Immediate antecedents of cigarette smoking: An analysis from ecological
momentary assessment. Journal of Abnormal Psychology, 111, 531–545.
http://dx.doi.org.libproxy1.usc.edu/10.1037/0021-843X.111.4.531

LAY THEORIES OF BEHAVIOR 48

Shiffman, S., Hufford, M., Hickcox, M., Paty, J. A., Gnys, M., & Kassel, J. D. (1997).
Remember that? A comparison of real-time versus retrospective recall of smoking lapses.
Journal of Consulting and Clinical Psychology, 65, 292–300.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-006X.65.2.292.a

Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological Momentary Assessment.
Annual Review of Clinical Psychology, 4(1), 1–32.
https://doi.org/10.1146/annurev.clinpsy.3.022806.091415

Taylor, S. E., & Brown, J. D. (1994). Positive illusions and well-being revisited: Separating fact
from fiction. Psychological Bulletin, 116, 21–27.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0033-2909.116.1.21

Watson, J. B. (1913). Psychology as the behaviorist views it. Psychological Review, 20, 158–
177. http://dx.doi.org.libproxy2.usc.edu/10.1037/h0074428

Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures
of positive and negative affect: The PANAS scales. Journal of Personality and Social
Psychology, 54, 1063–1070. http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-
3514.54.6.1063

Webb, T. L., & Sheeran, P. (2006). Does changing behavioral intentions engender behavior
change? A meta-analysis of the experimental evidence. Psychological Bulletin, 132, 249–
268. http://dx.doi.org.libproxy1.usc.edu/10.1037/0033-2909.132.2.249

Wegner, D. M., & Wheatley, T. (1999). Apparent mental causation: Sources of the experience of
will. American Psychologist, 54, 480–492.
http://dx.doi.org.libproxy1.usc.edu/10.1037/0003-066X.54.7.480

Wheeler, S. C., DeMarree, K. G., & Petty, R. E. (2007). Understanding the role of the self in
prime-to-behavior Effects: The active-self account. Personality and Social Psychology
Review, 11, 234–261. https://doi.org/10.1177/1088868307302223

Wilson, T. D., Laser, P. S., & Stone, J. I. (1982). Judging the predictors of one's own mood:
Accuracy and the use of shared theories. Journal of Experimental Social Psychology, 18,
537-556. https://doi.org/10.1016/0022-1031(82)90072-5

Wilson, T.D, & Nisbett, R. E. (1978). The accuracy of verbal reports about the effects of stimuli
on evaluations and behavior. Social Psychology, 41, 118–131.
https://doi.org/10.2307/3033572

Wood, W. (2017). Habit in Personality and social psychology. Personality and Social
Psychology Review, 21, 389–403. https://doi.org/10.1177/1088868317720362

LAY THEORIES OF BEHAVIOR 49

Wood, W., Quinn, J. M., & Kashy, D. A. (2002). Habits in everyday life: Thought, emotion, and
action. Journal of Personality and Social Psychology, 83, 1281–1297.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-3514.83.6.1281

Wood, W., & Rünger, D. (2016). Psychology of habit. Annual Review of Psychology, 67, 289–
314. https://doi.org/10.1146/annurev-psych-122414-033417

Wood, W., Tam, L., & Witt, M. G. (2005). Changing circumstances, disrupting habits. Journal
of Personality and Social Psychology, 88, 918–933.
http://dx.doi.org.libproxy2.usc.edu/10.1037/0022-3514.88.6.918

LAY THEORIES OF BEHAVIOR 50

Tables
Table 1
Study 1: Means (standard deviations) For Key Variables by Condition.
Condition Intentions

BFCS

SRBAI

Pos.
affect
Neg.
affect
Habit
attrib.
Mood
attrib.
Attrib
diff

Negative 59.88
(32.63)
6.57
(4.28)
2.66
(1.45)
1.61
(1.09)
2.79
(1.23)
44.30
(29.78)
55.70
(29.78)
11.40
(59.55)
Control 70.61
(22.01)
7.07
(3.95)
2.92
(1.15)
2.66
(1.09)
1.64
(0.84)
54.33
(25.82)
45.67
(25.82)
-8.65
(51.64)
Positive 73.92
(27.87)
6.43
(2.87)
2.88
(1.18)
3.88
(1.16)
1.23
(0.54)
42.39
(25.69)
57.61
(25.69)
15.21
(51.38)
Total 68.16
(28.34)
6.69
(3.73)
2.82
(1.27)
2.73
(1.45)
1.89
(1.13)
46.93
(27.48)
53.07
(27.48)
6.15
(54.96)

Note. Intentions ranged from 0-100, with higher scores reflecting stronger purchase intentions. BFCS
ranged from 1-16 and SRBAI ranged from 1-5, with higher scores on both scales reflecting stronger
purchase habits Pos. affect = positive affect (1-5), with higher numbers reflecting greater positivity. Neg.
affect = negative affect (1-5), with higher numbers reflecting greater negativity. Habit attrib. = habit
attribution (0 – 100), with higher numbers representing stronger attribution to habit, Mood attrib. = mood
attribution (0-100), with higher numbers representing stronger attribution to mood. Attrib diff =
attribution difference score (-100 to +100), with positive scores signifying attribution to affect more than
habit, and negative scores signifying attribution to habit more than affect.

LAY THEORIES OF BEHAVIOR 51

Table 2

Study 1: Correlations Between Key Variables

Variable 1 2 3 4 5

1. Purchase
intentions

2. Habit strength
(BFCS)
.35**
3. Habit strength
(SRBAI)
.32** .59**
4. PANAS –
positive
.37** .21** .25**
5. PANAS –
negative
-.19* -.05 -.02 -.57**
6. Attribution
difference score
a
-.10 -.12 -.15 .10 .02

Note.
a
the difference between each participants’ affect attribution score and habit attribution score.
* p < .05.
** p < .01

LAY THEORIES OF BEHAVIOR 52

Table 3
Study 1: Results of Linear Regression Predicting Behavioral Intentions from Habit Strength
(BFCS) and Experimental Condition.
Predictor b b
95% CI
[LL, UL]
β p-value Fit
Habit
strength
2.41 [0.52, 4.29] .31 .012*
Negative
a
-9.40 [-19.80,
1.00]
-.15 .076
Positive
b
4.60 [-5.76,
14.07]
.07 .381
Habit
strength X
Negative
1.04 [-1.50,
3.59]
.09 .418
Habit
strength X
Positive
-0.96 [-4.12,
2.20]
-.05 .548
R
2
=
.179**
Note. b represents unstandardized regression weights. β indicates the standardized regression
weights. LL and UL indicate the lower and upper limits of a confidence interval, respectively.
a

dummy-coded variable comparing the negative and control conditions.
b
dummy-coded variable
comparing the positive and control conditions.
*p < .05
**p < .01

LAY THEORIES OF BEHAVIOR 53

Table 4

Study 2: Means, Standard Deviations, and Correlations with Confidence Intervals

Variable M SD 1 2 3 4 5

1. Food
purchase
.51
a

2. Habit strength
(BFCS)
3.73 0.76 .23**

3. Habit strength
(SRBAI)
3.44 1.59 .19** .42**

4. Post-exam
mood
57.22 25.42 .00 .07 .07

5. Habit
attribution
48.29 30.45 .10 .06 .10 -.01

6. Mood
attribution
44.62 31.82 .22** .25** .20* -.17* .07

Note. M and SD are used to represent mean and standard deviation, respectively. Score ranges for
the BFCS ranged from 1-5 and SRBAI ranged from 1-7, with higher scores on both scales
reflecting stronger purchase habits. Post-exam mood ranged from 1-100, with higher scores
reflecting more positive affect. Habit attribution and mood attribution ranged from 1-100, with
higher scores reflecting stronger attribution to each factor.
a
Proportion of participants who purchased food after the exam.
*p < .05
**p < .01

LAY THEORIES OF BEHAVIOR 54

Table 5

Study 3: Means, Standard Deviations, and Between-Person Correlations for Key Variables

Variable M SD 1 2 3 4 5 6

1. Habit
strength
(BFCS)
11.48 5.20

2. Habit
strength
(SRHI)
4.05 1.34 .74**

3. Mean
fatigue
3.36 0.80 .02 .15

4. Habit
attribution
34.57 26.68 .45** .53** -.04

5. Fatigue
attribution
68.66 23.06 -.35** -.32** -.15 -.44**

6. Coffee
count
a

5.34 3.26 .50** .49** -.07 .31** -.09

7. Response
rate
32.98 6.00 -.11 -.14 .27* -.13 -.10 .07

Note. M and SD are used to represent mean and standard deviation, respectively. Score ranges for
the BFCS ranged from 1-5 and SRBAI ranged from 1-7, with higher scores on both scales
reflecting stronger purchase habits. Mean fatigue ranged from 1-6, with higher scores reflecting
higher fatigue. Habit attribution and mood attribution ranged from 1-100, with higher scores
reflecting stronger attribution to each factor. Response rate ranged from 0-40 and reflects the
number of prompts answered (out of 40 possible prompts), so that higher scores indicate a higher
response rate.
a
Total number of coffee drinking events reported during the study period.
* indicates p < .05. ** indicates p < .01.

LAY THEORIES OF BEHAVIOR 55

Table 6

Study 3: Coefficient Estimates for Fixed Effects in the Main Multilevel Model

Variable Odds Ratio 95% CI

1. Habit strength 𝛾 01 1.40 1.18-1.61

2. Fatigue (person-mean
centered) 𝛽 1𝑗
1.31 1.11-1.56

3. Fatigue (average person-
level) 𝛾 02
0.85 0.63-1.08

4. Fatigue attribution 𝛾 03 1.00 0.86-1.20

5. Habit strength X fatigue
(person-mean centered)
interaction 𝛾 11𝛽 1𝑗
0.95 0.79-1.10

6. Fatigue (person-mean
centered) X Fatigue
attribution interaction
𝛾 03𝛽 1𝑗
1.00 0.78-1.10

Note. The 95% CI (Credible Interval) represents the range of values that has a 95% chance of
including the population odds ratio.

LAY THEORIES OF BEHAVIOR 56

Figures

Figure 1. Study 3: Likelihood of coffee drinking by the following EMA prompt as a function of
fatigue in the prior prompt and habit strength.

LAY THEORIES OF BEHAVIOR 57

Figure 2. Association between self-reported and actual associations between fatigue and coffee
drinking.

Abstract (if available)

Abstract Causal explanations of habitual, automatic actions are prone to inaccuracy simply because people have only limited awareness of automatic processes. Thus, they may mistakenly explain habit performance as a response to salient but incidental inner states such as their current mood. We tested this misattribution hypothesis in an online experiment, a field study, and an ecological momentary assessment study. With all three methodologies, people’s attributions for their actions were poorly calibrated to the actual predictors of these actions. Habit strength outperformed or matched inner states in predicting behavior, whereas participants’ attributions for their behavior gave greater or equal weight to inner states over habit. Thus, lay perceptions of behavior overweight salient inner states and discount automatic processes. We discuss the implications of these findings for self-regulation and the limitations of self-report measures.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Habits and Friction: Actual and Perceived Effects on Behavior

PDF

The role of good habits in facilitating long-term goals

PDF

Performance and attention novelty slows hedonic adaptation during habit formation

PDF

How perceived moral congruence shapes propensities to engage in pro-group behaviors

PDF

Habits as friend and foe to self-control: the dual roles of routinized behavior

PDF

Fathers’ social-cognitive network connectivity: associations with prenatal father-infant attachment and postpartum parenting behavior

PDF

Selflessness takes time: altruistic (but not cooperative) prosocial behavior increases with decision time

PDF

Powerful guts: how power limits the role of disgust in moral judgment

PDF

Difficulty-as-importance and difficulty-as-impossibility: unpacking the context-sensitivity and consequences of identity-based inferences from difficulty

PDF

Theory of mind processing in expectant fathers: associations with prenatal oxytocin

PDF

Habits and policy: the socio-cognitive foundations of foreign policy stability

PDF

Dynamics of victimization, aggression, and popularity in adolescence

PDF

Evaluating social-cognitive measures of motivation in a longitudinal study of people completing New Year's resolutions to exercise

PDF

Interpersonal coping responses during adolescence: implications for adjustment

PDF

Measuring truth detection ability in social media following extreme events

PDF

A longitudinal study of sexual minority stress and behavioral health in sexual minority adolescents

PDF

Beyond active and passive social media use: habit mechanisms are behind frequent posting and scrolling on Twitter/X

PDF

Difficulty-as-sanctifying: when difficulties build character, purify the self, and elevate the soul

PDF

Bound in hatred: a multi-methodological investigation of morally motivated acts of hate

PDF

Behavioral and neural correlates of core-self and autobiographical-self processes

Asset Metadata

Creator Mazar, Asaf (author)

Core Title Illusory motives, elusive habits: lay theories of everyday behavior

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Psychology

Publication Date 02/04/2020

Defense Date 11/16/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag automaticity,habit,introspection illusion,lay theories,misattribution,OAI-PMH Harvest,self-insight

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wood, Wendy (committee chair), Monterosso, John (committee member), Stone, Arthur (committee member)

Creator Email amazar@usc.edu,mazaras17@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-264811

Unique identifier UC11674925

Identifier etd-MazarAsaf-8144.pdf (filename),usctheses-c89-264811 (legacy record id)

Legacy Identifier etd-MazarAsaf-8144.pdf

Dmrecord 264811

Document Type Thesis

Rights Mazar, Asaf

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA