Received: 10 May 2021 Revised: 31 May 2021 Accepted: 8 June 2021

DOI: 10.1002/pmh.1524

RESEARCH ARTICLE

The Habitual Tendencies Questionnaire: A tool for

psychometric individual differences research

Smriti Ramakrishnan1 | Trevor W. Robbins2,3 | Leor Zmigrod2,3

1
School of Clinical Medicine, University of
Cambridge, Cambridge, UK
Abstract
2
Department of Psychology, University of Habits are automatic responses to learned stimuli or contextual cues that are
Cambridge, Cambridge, UK insensitive to goals. Although habits may allow for automated behaviours that
3
Behavioural and Clinical Neuroscience increase efficiency in our daily lives, an over-reliance on habits has been
Institute, University of Cambridge,
Cambridge, UK
suggested to contribute to disorders such as obsessive–compulsive disorder
(OCD). There are currently few established measures of individual differences
Correspondence in habitual tendencies. To fill this gap, the present study generated and
Leor Zmigrod, Behavioural and Clinical
Neuroscience Institute, University of validated a novel 11-item scale, the Habitual Tendencies Questionnaire (HTQ),
Cambridge, Cambridge, UK. to measure individual differences in habitual tendencies in the general popula-
Email: lz343@cam.ac.uk
tion. In Study 1, factor analysis revealed three underlying subcomponents of
Funding information the HTQ: Compulsivity, Preference for Regularity, and Aversion to Novelty,
Wellcome Trust, Grant/Award Number: with Compulsivity showing the strongest association with subclinical OCD
104631/z/14/z; Gates Cambridge
symptomatology. Study 2 validated the HTQ and replicated the findings of
Scholarship
Study 1 in a larger sample, and explored relationships with other personality
traits. The results emphasise the importance of measuring individual variation
in habitual thinking styles, illustrating that different facets of habitual tenden-
cies may contribute to diverse behavioural and clinical outcomes. The present
investigation provides a new, reliable way of measuring habitual tendencies
and has important implications for future explorations into the nature of indi-
vidual differences from a dimensional perspective to psychiatry.

1 | INTRODUCTION cognitive dispositions and behavioural outcomes, and so

developing an effective questionnaire that taps into
Living creatures are ‘bundles of habits’ as observed by such individual differences is valuable across the
James (1890, p. 3), and indeed, humans quickly learn psychological sciences. Scholars have noted the impor-
to repeat and perpetuate responses when faced with tance of conceptual and methodological assessment
recurring contextual cues. Nonetheless, not all individ- tools of habitual behaviour for the progress of
uals are equally ‘habitual’, with some individuals habit research and application (De Houwer, 2019;
exhibiting strong tendencies towards routine and Gardner, 2015; Luigjes et al., 2019; Mazar &
compulsivity in their daily lives, whereas others Wood, 2018), and so the present investigation seeks to
naturally reject routine and repetition and opt for more provide a measure of domain-general personality
varied change instead. Individual differences in habitual tendencies towards habits that will allow scientists to
tendencies may underpin or reflect a large range of map out the mind prone to habits.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2021 The Authors Personality and Mental Health Published by John Wiley & Sons Ltd.

Personal Ment Health. 2021;1–17. wileyonlinelibrary.com/journal/pmh 1

2 RAMAKRISHNAN ET AL.

Verplanken and Aarts (1999, p. 104) defined habits as create the HTQ, we first conducted a thorough literature
‘learned sequences of acts that have become automatic review of existing measures of habits and sought to evalu-
responses to specific cues, and are functional in obtaining ate their strengths and weaknesses. Two established
certain goals or end-states’. However, this definition does scales measuring habits that reflect two schools of
not take into account other aspects of habits, such as thought on habits are the Self-Report Habit Index (SRHI)
compulsive and addictive behaviours, which are not nec- (Verplanken & Orbell, 2003) and the Creature of Habit
essarily goal-directed. More recent definitions of habits Scale (COHS) (Ersche et al., 2017, 2019). Verplanken and
include ‘automatic behavioural responses to environmen- Orbell (2003, p. 1314) developed the SRHI, a 12-item
tal cues, thought to develop through repetition of behav- self-report index of habit strength, to reflect their
iour in consistent contexts’ (Lally & Gardner, 2013, argument that “habit is a psychological construct, rather
p. 137) and ‘representations of stimulus–response links than simply past behavioral frequency”. Consequently,
that do not refer to goals, and are in a sense directly the SRHI aims to focus on features of habit such as a
elicited by the environmental states or stimuli or con- history of repetition, automaticity and expressing one's
texts’ (Robbins & Costa, 2017, p. 1201). These suggest identity, rather than on past behavioural frequency. In
that habits are not goal-directed, but instead emphasise this scale, a particular behaviour, X, is followed by 12 dif-
their stimulus–response nature. Indeed, it is now widely ferent options from which the participant must choose,
accepted that habitual behaviours are not mediated by such as ‘...I do frequently’ or ‘...I do without thinking’.
goal pursuit (Wood & Neal, 2007). Although a behaviour Nonetheless, the behaviours used for X in two out of four
may originally have been motivated by goal pursuit, once experiments in this study related to modes of transport,
it has been established as a habit, the goal is no longer which may not be representative of an individual's
needed to motivate the behaviour (Ersche et al., 2017). dependence on habits in general. In another experiment,
Despite a long tradition of theorising about the nature participants were asked to list some of their own habits,
of habit, from James to modern neuroscience, some have which they performed either daily or weekly. Although
suggested that habit is an ‘empty construct’. This is using habits unique to each participant ensured that the
because many of these studies used past behavioural fre- behaviours were relatable, this is a very time-consuming
quency as a measure of habit, and statistical relationships method and thus would not be feasible to use in many
between past and future behaviours are ambiguous as research designs. Furthermore, these behaviours were
they may be influenced by confounding variables that are assumed to be habitual based on their frequency and reg-
not measured. However, Verplanken and Aarts (1999, ularity, which seems to contradict the authors' argument
p. 102) argued against the notion of habit as an ‘empty that habit is not exclusively past behavioural frequency.
construct’. Instead, they suggested that different para- Additional self-report scales that build on the SRHI
digms are needed in order to understand habits more from the clinical literature assess domain-specific habits,
fully, and that ‘habits are not only response programs, such as habits with regards to alcohol use (Grodin et al.,
but may have far-reaching consequences for our 2019; Piquet-Pessôa et al., 2019), smoking habits (Ray
cognitive functioning, for instance the way we perceive et al., 2020), hoarding (Maier, 2004) and physical activity
situations and process information’. This emphasises the habits (Hagger, 2019). Nevertheless, an overemphasis on
importance of studying habits, not only in themselves but the domain-specificity of habits, and measuring them
also as indicators of cognitive functions and various exclusively in domains deemed clinically aberrant can
personality traits. Habits can be beneficial by improving lead to a neglect of the mapping of what makes a mind
efficiency in our daily lives and increasing the availability prone to habits, regardless of the domain in which these
of cortical processing capacity for novel, important habits operate (alcohol, hoarding, smoking, etc.). This
situations (Robbins & Costa, 2017). However, excessive opens up key empirical and theoretical questions about
reliance on habits can be detrimental to behavioural the nature of habitual thinking and how these are instan-
plasticity and can contribute to the development of tiated neurally. Consequently, in the present study, we
disorders of compulsivity (Gillan et al., 2016) such as concurred with Verplanken and Orbell's view that habit
obsessive–compulsive disorder (OCD; Gillan et al., 2015) is a psychological construct that includes behaviours with
and substance dependence (Everitt & Robbins, 2016; a history of repetition, automaticity and expressing
Sjoerds et al., 2013). one's identity. Here, we take this one step further and
In order to study habits, we must be able to reliably consider habits as also encompassing attitudes, beliefs
measure them. Therefore, the current investigation and thinking styles. Therefore, we set out to develop an
aimed to develop a validated, representative scale to mea- easy-to-administer tool to measure all of these aspects of
sure individual variation in dependence on habits: the habits, with items that are representative of habits in
Habitual Tendencies Questionnaire (HTQ). In order to general, and as universally relatable as possible.
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 3

Another recent scale developed to measure habits is to create a multidimensional self-report scale of habitual
the COHS (Ersche et al., 2017). Two main subscales were tendencies in healthy individuals. We consulted the fol-
identified by the authors: routine and automaticity, lowing theoretically-adjacent constructs:
reflecting two different features of habits. A potential
limitation of the COHS is that more than half of the • Intolerance of uncertainty or ambiguity, defined as ‘the
items are food-related, and these items make up the tendency to perceive (i.e. interpret) ambiguous situations
majority of the automaticity subscale. It is possible that as sources of threat’ (Stanley Budner, 1962, pp. 29–30),
an individual's food-related habits are not representative with ambiguous situations being described as those
of their habitual tendencies in general, as many people ‘which cannot be adequately structured or categorised
are habitual in their eating behaviours (van't Riet by the individual because of the lack of sufficient cues’.
et al., 2011), but not necessarily in other aspects of their Frenkel-Brunswik, cited in Bar-Tal (1994), suggested that
daily lives. Therefore, we avoided including food-related intolerance of ambiguity is a preference for familiarity,
or domain-specific habit items in the HTQ. symmetry, definiteness and regularity, all of which seem
To the best of our knowledge, no scale exists that to reflect qualities of habits.
measures behaviours, attitudes, beliefs and thinking • Need for cognitive closure, defined as ‘an answer on a
styles relating to habits. Incorporating these dimensions given topic, any answer … compared to confusion and
was important in order to reflect the psychological litera- ambiguity’ (Kruglanski, 1990, p. 337).
ture indicating phenomenological and conceptual distinc- • Need for cognitive structure, defined as ‘the desire for
tions among attitudes, behaviours, beliefs and cognitive clear and firm knowledge concerning a given
styles (Ajzen, 1989; Armitage & Christian, 2003). Here, in topic, as opposed to ambiguity, doubt, or confusion’
the context of tapping into habitual tendencies, we (Bar-Tal, 1994, p. 46).
considered these four dimensions as follows: behaviour as • Routines, defined as ‘familiar action patterns that
reflecting individual differences in dependence on involve regularity, which are likely to be performed on
routines or habits in daily life, attitude as the desire for a daily basis’ (Ersche et al., 2017, p. 77)
structure or order in life (which might make individuals • Automaticity, with automatic actions being defined as
routine-prone), belief as beliefs about the value of having those that are ‘initiated by environmental cues without
routines or habits (not about the personal self but in gen- a deliberate intention, and they may even continue
eral) and thinking style in terms of a compulsive thinking without the involvement of conscious control’ (Ersche
style that is susceptible to habitual or non-goal-directed et al., 2017, p. 78)
behaviour. All these dimensions may play major roles in • Compulsivity, defined as ‘the tendency to repeat over
the development and maintenance of habits, as well as and over a certain kind of behavior despite its
potentially contributing to associations between habits inappropriateness, and to be unable to inhibit the
and other aspects of cognition such as personality traits, behavior’ (Bari & Robbins, 2013, p. 52). Compulsivity
psychopathology and cognitive functions. Therefore, the has further been described as the ‘manifestation of an
HTQ aims to encompass all of these aspects of habits. imbalance between the brain's goal-directed and habit-
Our criteria for the HTQ were such that it should learning systems’ (Gillan et al., 2016, p. 828), and as ‘a
consist of items that are conceptually representative of maladaptive perseveration of behaviour’ (Robbins
the habitual tendencies construct in general, as per our et al., 2012, p. 83), contributing to the use of habit as a
definitions and descriptions; be relatable to everyday life model of compulsivity.
for individuals across the general population; and be
quick and easy to administer. Study 1 sought to create a All of the above constructs reflect different character-
new scale to measure individual differences in habitual istics of habitual tendencies and thus were included as
tendencies, the HTQ, and Study 2 aimed to validate and keywords in our literature search.
replicate the HTQ in a larger sample, and to explore the
relationships between the HTQ and the COHS (Ersche
et al., 2017). 3 | METHODS

3.1 | Participants
2 | S T UDY 1
For Study 1, we recruited 165 participants, each of whom
In order to construct the HTQ, we conducted an exten- were paid $4.50 for their participation in the study,
sive literature review of existing measures in order to through Amazon Mechanical Turk (MTurk) online
identify potential items that could be used and adapted platform, which is well established for obtaining general
4 RAMAKRISHNAN ET AL.

population samples (Cheung et al., 2017). Of these,

35 (21.2%) were removed prior to data analysis due to
failure of attention checks and repeat participation in the
study identified via duplicated IP addresses. The
130 remaining participants consisted of 49% males, 50%
females and 1% other, between the ages of 22 and
73 (M = 39.527, SD = 12.120). All participants were
based in the United States. The sample identified as 72%
White, 11% Mixed ethnicity, 8% Black or African
American, 6% Asian, 2% American Indian or Alaska
Native, and 1% Hispanic/Latino. The highest levels of
educational attainment of the sample population were as
follows: 1% had achieved less than a high school degree,
13% had graduated high school, 19% had completed some
school but did not have a degree, 15% had completed a
2-year Associate degree in college, 38% had completed
a 4-year Bachelor's degree in college, 12% had a Master's
degree, and 2% had a Doctoral or Professional degree.
Ethical approval for the study was obtained from the
Department of Psychology Ethics Committee of the
University of Cambridge. Electronic informed consent
was obtained from all participants before beginning the
survey, in line with the Declaration of Helsinki (1964),
and participants were informed that they may terminate
their participation in the study at any point.

3.2 | Scale development

The development of the HTQ followed a rigorous process

of item selection (see Figure 1 for flowchart of scale
development). Following a thorough literature review, FIGURE 1 Flowchart of scale development
we selected a series of keywords relating to habits and
used these to search for relevant existing scales in
Google Scholar. These keywords were as follows: uniqueness; relevance to our four proposed aspects of the
‘cognitive closure’, ‘cognition’, ‘uncertainty’, ‘ambiguity’, habitual tendencies construct; conceptual clarity; and
‘cognitive structure’, ‘habit*’, ‘routine*’, ‘automatic*’, applicability to current, everyday life. The HTQ aims to
‘goal-directed’, and ‘compulsiv*’. We also used a citation capture four distinct aspects of habits: behaviour, attitude,
search, in order to maximise the number of scales identi- belief and thinking style. Therefore, as we selected items
fied. We then pooled all the items from each of the scales for our scale, we categorised each item into one of these
found. Twenty-seven scales were identified, resulting in a four subscales, ensuring a minimum of seven items per
total of 618 items. We used a process of elimination to nar- subscale (see supporting information).
row down the number of potential items for the HTQ.
Firstly, the full versions of scales were removed, where
validated shortened or revised versions existed, as were 3.3 | Measures
scales consisting entirely of items irrelevant to the HTQ.
Nineteen scales then remained. Next, where factor load- We administered the 37-item HTQ scale, along with the
ings were available, items with factor loadings below 0.4 additional measures and cognitive tasks, in the form of
were removed, to ensure that the remaining items were an electronic survey. Items from the HTQ were rated on
representative, and following this, any duplicate items 7-point Likert scales ranging from ‘Strongly disagree’ to
were removed in order to achieve nonredundancy. This ‘Strongly agree’ and were randomised across factor
resulted in a pool of 401 items. Finally, we selected 37 of categories. In order to measure subclinical OCD symp-
these items for our scale. For each item, we considered its tomatology, we used the 18-item revised version of the
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 5

Obsessive–Compulsive Inventory (OCI) (Foa et al., 2002), the second of these purposes, in order to validate the
which was rated on 5-point Likert scales ranging from HTQ, and to create a shorter, revised version of the HTQ,
‘Not at all’ to ‘Extremely’, and had a high Cronbach's α consisting of the items most representative of the
value of 0.954. Example items included: ‘I repeatedly habitual tendencies construct.
check doors, windows, drawers, etc.’ and ‘I frequently Exploratory factor analysis (EFA) was carried out
get nasty thoughts and have difficulty in getting rid of using maximum likelihood as the factor extraction
them’. In order to measure intolerance of uncertainty, we method, as recommended by Costello and Osborne
used the 12-item short version of the Intolerance of (2005) (see Table S1 for factor loadings). We expected the
Uncertainty Scale (IUS) (Carleton et al., 2007), which different aspects of habits to be intercorrelated and thus
was rated on 5-point Likert scales ranging from ‘Not at used oblique oblimin rotation with parallel analysis. Four
all characteristic of me’ to ‘Entirely characteristic of me’, factors were obtained, supported by the scree plot and
and had a high Cronbach's α value of 0.912. Example path diagram. We then applied some a priori decision
items included ‘Unforeseen events upset me greatly’ and criteria (in line with past research, e.g., Krumrei-
‘The smallest doubt can stop me from acting’. In order to Mancuso & Rouse, 2016), in order to select which items
measure autism-spectrum traits, we used the 10-item would be included in further analyses. These were as
short version of the Autism Quotient, AQ-10 (Allison follows: items must have a minimum factor loading of
et al., 2012), which was rated on 4-point Likert scales 0.4, which resulted in the removal of nine items (HTQ11,
ranging from ‘Definitely disagree’ to ‘Definitely agree’, HTQ12, HTQ13, HTQ17, HTQ 20, HTQ21, HTQ25,
and had an adequate Cronbach's α value of 0.684. HTQ32 and HTQ34); items must not cross-load onto their
Example items included ‘I often notice small sounds alternative factors above 0.3, which resulted in the
when others do not’ and ‘I find it easy to work out what removal of a further eight items (HTQ 3, HTQ5, HTQ14,
someone is thinking or feeling just by looking at HTQ15, HTQ22, HTQ23, HTQ29 and HTQ31). EFA was
their face’. The survey also included two interspersed then run again in order to avoid skew due to the removed
measures of attention to ensure that participants were items, and three factors were obtained (see Table S2).
concentrating on their responses to the questions (‘I am One further item (HTQ2) was subsequently removed as
paying attention to this survey. I strongly agree’). its factor loading was below 0.4, and thus, it did not fulfil
our inclusion criteria. A third EFA was then carried out,
and the three-factor structure was maintained with
4 | R E SUL T S 19 items.
Examination of the items in each of the three factors
All statistical analyses were conducted using JASP revealed that each factor reflected a distinct aspect of
(Version 0.12.2; JASP Team, 2020), SPSS (Version 27.0; habitual tendencies. Factor 1 encompassed items
IBM Corp, 2020) and R Studio (RStudio Team, 2020). related to compulsivity and very clearly reflected the
The HTQ scores based on the 37-item version thinking style dimension of habitual tendencies. Factor
followed an approximately normal distribution according 2 encompassed items related to a preference for regularity
to the Shapiro–Wilk test (p = 0.955), with minimal and routines, mirroring the attitude dimension of
skewness ( 0.090) and kurtosis (0.136), and Cronbach's α habitual tendencies: desire for structure or order in life.
was calculated to be 0.903, with a 95% confidence interval Factor 3 encompassed items related to an aversion to new
(CI) [0.878, 0.926]. experiences or change, reflecting habitual behaviours.

4.1 | Factor analysis 4.2 | Item selection for a shortened scale

Factor analysis is a statistical dimensionality reduction In order to create an easy-to-administer scale, we sought
method for empirically identifying the structure underly- to shorten it. As manifest in Figure 2, we decided to select
ing a variety of measurements (Thompson, 2007). up to four items per factor, choosing the items that
Thompson further states that factor analysis is used for loaded most strongly on those factors in the EFA. This
three main purposes: (1) ‘empirically creating a theory of resulted in the selection of 11 items, which were
structure’, (2) ‘evaluating whether factored entities clus- subjected to another EFA (see Table 1 and Figure S1). As
ter in a theoretically expected way’ and (3) ‘estimating expected, three factors emerged, supported by the scree
latent variables scores (i.e., factor scores) that are then plot and path diagram (see Figure S1), and all items
used in subsequent statistical analyses … in place of the continued to load onto the same factors as they had
measured factored entities’. We used factor analysis for previously done.
6 RAMAKRISHNAN ET AL.

Descriptive statistics and reliability analysis were scale—see Appendix A), as well as for each factor
then carried out on the final HTQ scale (all subsequent individually. The HTQ scores continued to follow a nor-
mentions of the HTQ refer to this final, 11-item mal distribution according to the Shapiro–Wilk test
(p = 0.369), with minimal skewness ( 0.019) and kurtosis
( 0.080), and the mean total score was 34.546 (maximum
possible score = 66, range = 9–55), with standard devia-
tion 9.042 (see Figure 3). For the 11-item scale, Cronbach's
α was 0.764, with 95% CI [0.699, 0.820]. Cronbach's α
values showed good reliability for each factor, or subscale.
These were 0.878 for Compulsivity; 0.770 for Preference
for Regularity; and 0.733 for Aversion to Novelty.

4.3 | Construct validity

As evident in Table 2, all three HTQ subscales showed

significant and strong positive correlations with the
11-item HTQ (with r values above 0.5), but only weak,
mostly non-significant correlations with each other (with
r values less than 0.5). This corroborates the factor
analysis in suggesting that each subscale is representative
of a distinct aspect of the habitual tendencies construct.
In order to evaluate the relationships between the
F I G U R E 2 Flowchart of item selection for final 11-item HTQ and relevant behavioural outcomes, we assessed the
Habitual Tendencies Questionnaire (HTQ) correlations between the HTQ and subclinical OCD

TABLE 1 Exploratory factor analysis of final 11-item Habitual Tendencies Questionnaire

Component loadings

Items Factor 1 Factor 2 Factor 3 Uniqueness

HTQ 37: I tend to dwell on the same issues 0.91 0.05 0.02 0.18
HTQ 36: I mentally fixate on certain issues and cannot 0.88 0.02 0.03 0.22
move on
HTQ 35: The same thoughts often keep going through 0.84 0.00 0.06 0.27
my mind over and over again
HTQ 33: I tend to repeat actions because I keep 0.59 0.10 0.17 0.63
doubting that I have done them properly
HTQ 10: I like to have a regular, unchanging schedule 0.02 0.70 0.10 0.46
HTQ 9: There is comfort in regularity 0.10 0.67 0.06 0.58
HTQ 27: A good job has clear guidelines on what to do 0.01 0.67 0.09 0.59
and how to do it
HTQ 1: I hate it when my routines are disrupted 0.14 0.68 0.10 0.43
HTQ 30: I look forward to new experiences R 0.03 0.02 0.92 0.14
HTQ 26: Life is boring if you never take risks and 0.06 0.00 0.67 0.56
always play it safe R
HTQ 7: When eating at restaurants, I like to try new 0.04 0.02 0.54 0.71
dishes rather than ones I have tried before R
Correlation with factor 1 1.00
Correlation with factor 2 0.17 1.00
Correlation with factor 3 0.13 0.33 1.00

Note: R = reversed item.

THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 7

F I G U R E 3 Descriptive statistics for final,

11-item Habitual Tendencies Questionnaire
(HTQ): distribution plot and boxplot

TABLE 2 Correlation matrix of the Habitual Tendencies Questionnaire and OCD traits, including Pearson's correlations and Bayes
factors

HTQ aversion
HTQ HTQ compulsivity HTQ regularity to novelty
HTQ Pearson's r —
BF10 —
HTQ compulsivity Pearson's r 0.728*** —
BF10 4.897  10 19
—
HTQ regularity Pearson's r 0.672*** 0.161 —
BF10 2.863  10 15 0.573 —
HTQ aversion to novelty Pearson's r 0.577*** 0.065 0.278** —
BF10 1.335  10 10
0.143 17.309 —
OCI Pearson's r 0.484*** 0.598*** 0.146 0.103
BF10 2.293  106 1.474  10 11 0.425 0.215

Note: BF < 3 = Anecdotal evidence; BF < 10 = Moderate evidence; BF < 30 = Strong evidence; BF < 100 = Very strong evidence; BF > 100 = Extremely
strong evidence.
Abbreviations: HTQ, Habitual Tendencies Questionnaire; OCI, Obsessive–Compulsive Inventory.
*p < 0.05. **p < 0.01. ***p < 0.001.

symptomatology. The Pearson's correlations for these correlation). Here, we found that the relationship between
variables were computed (see Table 2). As evident in HTQ Compulsivity and the OCI possesses an extremely
Table 2, there was a significant positive correlation large Bayes Factor of 1.474  1011 (see Table 2), indicating
between the HTQ and OCI scales (r = 0.484, p < 0.001). that the observed data iare 1.474  1011 times more likely
Within the three subscales of the HTQ, the Compulsivity under H1 than H0. See supporting information for analysis
subscale contributed the most to this association of the associations of the HTQ with intolerance of
(see Table 2 and Figure 4), as it showed the strongest uncertainty and autism spectrum traits.
correlation with the OCI (r = 0.598, p < 0.001), whereas We then conducted a two-step hierarchical linear
the Preference for Regularity and Aversion to Novelty regression with the three subscales of the HTQ as predic-
subscales were not significantly correlated with the OCI. tors of subclinical OCD symptomatology, and age, gender
The Pearson's r effect sizes of 0.484 and 0.598 are and educational attainment as covariates. Of the demo-
relatively large, as per the individual differences research graphic variables, only age was a significant predictor of
guidelines set out by Gignac and Szodorai (2016). subclinical OCD symptomatology (β = 0.233, t(127)
To complement the Pearson's correlations, we also = 2.616, p = 0.010), and of the three HTQ subscales,
examined the Bayes Factors (see Table 2), which quantify only HTQ Compulsivity emerged as a significant
the evidential strength in favour of a significant correlation predictor of subclinical OCD symptomatology (β = 0.546,
given the present data (H1, the alternative hypothesis), or t(127) = 7.385, p < 0.001). The demographic variables
in favour of no significant correlation given the present explained 6.5% of the variance in subclinical OCD symp-
data (H0, the null hypothesis). In line with the guidelines tomatology (R2 = 0.065, F(3, 124) = 2.853, p = 0.040),
by Wagenmakers et al. (2018), a Bayes Factor (BF10) above but addition of the three subscales of the HTQ in step
100 indicates ‘extreme evidence’ for H1 (significant 2 increased the R2 term to 0.386, accounting for a further
8 RAMAKRISHNAN ET AL.

relationships between the HTQ, its three subscales, and

existing measures of subclinical traits of clinical disorders
related to maladaptive habits. Most notably, significant,
strong positive correlations were found between HTQ
Compulsivity and subclinical OCD symptomatology
(as measured by the OCI). These findings suggest that the
HTQ as a whole is representative of a range of different
habitual tendencies, and its individual subscales may be
used to explore the variable ways in which habits are
distributed across different subclinical traits of clinical
disorders. For example, the relationship between HTQ
Compulsivity and the OCI implies that thinking style,
rather than other features such as behaviour or attitudes
towards the value of habits, is specifically related to sub-
F I G U R E 4 Scatter plot showing correlations between the
clinical OCD symptomatology. Hierarchical regression
Obsessive–Compulsive Inventory (OCI) and the compulsivity
subscale of the Habitual Tendencies Questionnaire (HTQ) demonstrated that the three subscales of the HTQ explain
a significant proportion of the variance in subclinical
OCD symptomatology, and furthermore, revealed HTQ
32.1% of the variance in subclinical OCD symptomatol- Compulsivity to be a significant predictor of subclinical
ogy (R2 = 0.386, F(3, 121) = 12.704, p < 0.001). OCD symptomatology. As such, the HTQ may be used as
a validated measure of individual differences in habitual
tendencies, and its subscales may have an important role
4.4 | Interim discussion in predicting subclinical traits in the general population.

Study 1 has developed and validated a reliable, represen-

tative 11-item scale to measure individual differences in 5 | STUDY 2
habitual tendencies in the general population, the HTQ.
A total of 618 items from 27 existing scales were pooled Study 1 developed the HTQ and examined its relation-
through keyword and citation searches. Item selection ships with subclinical OCD symptomatology in an
took place through a process of elimination using a priori exploratory way, demonstrating that its Compulsivity
decision criteria, by considering factor loadings, non- subscale acts as a predictor of OCD traits in a sample of
redundancy, relevance to our four proposed aspects of the general population. In order to replicate and extend
the habitual tendencies construct, conceptual clarity and the findings of Study 1, we conducted a second study.
applicability to current everyday life (see Figure 1). EFA Study 2 aimed to reproduce the positive association
resulted in the subdivision of the HTQ into 3 subscales, between HTQ Compulsivity and OCD traits found in
namely, Compulsivity, Preference for Regularity and Study 1, and furthermore, to examine how the HTQ
Aversion to Novelty (see Tables 1,S1 and S2). The HTQ relates to a recent measure of habitual tendencies, the
as a whole encompassed the four aspects of habits we COHS (Ersche et al., 2017), in order to determine
originally proposed: behaviour, attitude, belief and think- whether the HTQ is representative of the habitual ten-
ing style, highlighting that habits are more than merely dencies construct. The aims of Study 2 were as follows:
past behavioural frequency (Verplanken & Orbell, 2003). (1) to replicate the three-factor structure of the HTQ
However, although the three subscales obtained after fac- obtained in study 1 in a larger, independent sample;
tor analysis represented three distinct aspects of the (2) to replicate the relationship between the HTQ and
habitual tendencies construct, these subscales differed subclinical OCD symptomatology (as measured by the
somewhat from our four originally-proposed theoretical OCI); and (3) to explore associations between the HTQ
aspects (see supporting information, Extended 37-Item and a recent measure of habitual tendencies, the COHS
Habitual Tendencies Questionnaire). The clustering of (Ersche et al., 2017).
items into the three distinct HTQ subscales provides a Furthermore, Study 2 was preregistered on the Open
new way of dividing the habitual tendencies construct Science Framework at the following link: https://osf.io/
into its component parts and allows the different aspects 3ag79/?view_only=0e0478eb848b477180d25e8b175edad9.
of habits to be studied separately. The construct validity Some changes were made to Study 1 after the preregistra-
of the HTQ was then demonstrated by exploring tion, namely, that the number of items in the HTQ was
the Pearson's correlations and Bayes factors of the reduced from 20 to 11, in order to create an even shorter
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 9

scale, resulting in a three-factor structure rather than a Department of Psychology Ethics Committee of the
four-factor structure. In Study 2, we did not analyse the University of Cambridge. Electronic informed consent was
data for cognitive flexibility, binge eating, alcohol obtained from all participants before beginning the survey,
addiction, smoking habits or apathy in relation to the in line with the Declaration of Helsinki (1964), and
HTQ, as it was beyond the scope of the present paper. participants were informed that they may terminate their
participation in the study at any point.

6 | METHODS
6.2 | Measures
6.1 | Participants
We administered the 11-item HTQ rated on 7-point
In order to collect a well-powered participant pool, we Likert scales ranging from ‘Strongly disagree’ to
preregistered a power analysis that would allow us to ‘Strongly agree’, along with the additional measures and
evaluate the relationship between habitual tendencies and cognitive tasks, in the form of an electronic survey hosted
relevant behavioural outcomes. In order to estimate the by Qualtrics Survey Software. As in Study 1, these
expected effect sizes, we relied on previous work conducted consisted of the revised OCI (Foa et al., 2002), which had
by Ersche et al. (2017) on associations between habitual a high Cronbach's α value of 0.944 as well as the COHS
tendencies and OCD symptomatology. Specifically, Ersche (Ersche et al., 2017), which had a high Cronbach's α
et al. (2017) found a correlation of r = 0.265 between the value of 0.902. The survey also included two interspersed
Routine subscale of the COHS and scores on the OCI. The attention checks, as in Study 1.
power analysis indicated that a sample of 287 would be
needed to detect an equivalent effect size (α = 0.05,
power = 0.90, r = 0.265). We oversampled by 34.8% 7 | RESULTS
(98 participants) to have a total sample size of 385 due to
the high prevalence of repeated IP addresses and bot 7.1 | Replicating the scale structure
responses in our sample. Each participant was paid $4.50
for their participation in the study, through MTurk online As shown in Figure 5, the HTQ scores followed an
platform. Of these, 126 (32.7%) were removed prior to data approximately normal distribution according to the
analysis in line with guidance from Meade and Craig (2012) Shapiro–Wilk test (p = 0.323), with minimal skewness
due to: failure of one or both attention checks (n = 28), ( 0.124) and kurtosis (0.015). The mean total score on
being identified as a bot via repeated answers in an the HTQ was found to be 35.927 (maximum possible
open-answer feature of the survey (n = 62), poor English score = 66, range = 6–63), with standard deviation
proficiency identified by lack of understanding through 10.082, and a good Cronbach's α of 0.810, with 95% CI
irrelevant or incoherent answers to other features of the [0.774, 0.843]. Cronbach's α values for each of the sub-
survey (n = 25), repeat participation in the study identified scales were 0.822 for Compulsivity, 0.777 for Preference
via duplicated IP addresses (n = 8), and finally one or more for Regularity, and 0.694 for Aversion to Novelty. Next,
missing answers on the HTQ (n = 3). The 259 remaining we conducted a Confirmatory Factor Analysis (CFA),
participants consisted of 56% males, 43% females and which provides indicators of model fit to help researchers
1% other/unspecified, between the ages of 19 and decide whether a model should be rejected or revised in
73 (M = 37.372, SD = 11.280). All participants were based light of new data (Brown, 2015). The CFA was then car-
in the United States. The sample identified as 68.3% White, ried out on the 11-item HTQ (see Figure S2 and
13.5% Black or African American, 5.8% Mixed ethnicity, Table S3), which indicated that the three-factor structure
4.6% Asian, 4.6% Hispanic/Latino, 1.2% American Indian was adequate [χ 2(41, 259) = 104.901, p < 0.001, root
or Alaska Native, 0.4% Native American/Pacific Islander, mean square error of approximation (RMSEA) = 0.078
1.2% other, 0.4% unspecified. The highest levels of educa- [0.059, 0.096], standardized root mean square residual
tional attainment of the sample population were as follows: (SRMR) = 0.055, comparative fit index (CFI) = 0.934;
0.4% had achieved less than a high school degree, 13.1% Tucker–Lewis Index (TLI) = 0.912].
had graduated high school, 22.0% had completed some
school but did not have a degree, 13.9% had completed a
2-year Associate degree in college, 43.2% had completed 7.2 | Construct validity
a 4-year Bachelor's degree in college, 6.6% had a Master's
degree, and 0.8% had a Doctoral or Professional degree. As evident in Table 3, all three HTQ subscales showed
Ethical approval for the study was obtained from the significant and strong positive correlations with the
10 RAMAKRISHNAN ET AL.

11-item HTQ (with r values above 0.5), but only significant positive correlation between the HTQ and
moderate correlations with each other (with r values less OCI scales (r = 0.258, p < 0.001). Within the three sub-
than 0.5). This corroborates the factor analysis in scales of the HTQ, the Compulsivity subscale contrib-
suggesting that each subscale is representative of a dis- uted the most to this association (see Table 3 and
tinct aspect of the habitual tendencies construct. Figure 6), as it showed the strongest correlation with
In order to evaluate the relationships between the the OCI (r = 0.461, p < 0.001), whereas the Preference
HTQ and subclinical OCD symptomatology, we for Regularity and Aversion to Novelty subscales were
computed the Pearson's correlations for these variables not significantly correlated with the OCI. The Pearson's
(see Table 3). As evident in Table 3, there was a r effect sizes of 0.258 and 0.461 are typical and
relatively large, respectively, as per the individual differ-
ences research guidelines set out by Gignac and
Szodorai (2016).
To complement the Pearson's correlations, we also
examined the Bayes Factors (see Table 3), which demon-
strated that that the relationship between HTQ Compul-
sivity and the OCI possesses an extremely large Bayes
Factor of 5.094  1011 (see Table 3), indicating that the
observed data is 5.094  1011 times more likely under H1
(significant correlation) than H0 (no correlation). As this
Bayes Factor value is above 100, it indicates ‘extreme
evidence’ for H1, in line with the guidelines from
Wagenmakers et al. (2018).
We then explored the relationships between the HTQ
F I G U R E 5 Distribution plot for 11-item Habitual Tendencies and the COHS, a recently-developed self-report measure
Questionnaire (HTQ) of habits. Descriptive statistics revealed that the

T A B L E 3 Correlation matrix of the Habitual Tendencies Questionnaire, Creature of Habit Scale and OCD traits, including Pearson's
correlations and Bayes factors

HTQ HTQ HTQ aversion COHS

HTQ compulsivity regularity to novelty OCI COHS total routine
HTQ r —
BF10 —
HTQ compulsivity r 0.747*** —
BF10 1.188  10 44
—
HTQ regularity r 0.782*** 0.314*** —
BF10 2.284  10 51
44,224.531 —
HTQ aversion to r 0.685*** 0.198** 0.465*** —
novelty BF10 8.676  10 33
12.917 2.355  10 12
—
OCI r 0.258*** 0.461*** 0.095 0.080 —
BF10 404.171 5.094  10 11
0.240 0.173 —
COHS total r 0.347*** 0.276*** 0.302*** 0.175** 0.220*** —
BF10 348,786.925 987.477 7300.442 3.245 24.957 —
COHS routine r 0.312*** 0.218*** 0.309*** 0.154* 0.190** 0.881*** —
BF10 15,295.251 27.060 12,651.433 1.391 5.728 1.882  10 76
—
COHS r 0.273*** 0.253*** 0.190** 0.141* 0.181** 0.806*** 0.430***
automaticity BF10 788.434 211.415 6.454 0.887 3.746 1.326  10 53
3.020  109

Note: BF < 3 = Anecdotal evidence; BF < 10 = Moderate evidence; BF < 30 = Strong evidence; BF < 100 = Very strong evidence; BF > 100 = Extremely
strong evidence.
Abbreviations: COHS, Creature of Habit Scale; HTQ, Habitual Tendencies Questionnaire; OCI, Obsessive–Compulsive Inventory.
*p < 0.05. **p < 0.01. ***p < 0.001.
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 11

7.3 | Interim discussion

Study 2 has replicated the findings of Study 1 and

validated the 11-item HTQ in a larger sample. CFA
demonstrated that the three-factor structure of the HTQ
was adequate, thus validating the 11 items subdivided
into the three subscales for the final version of the HTQ.
The Bayes factors and Pearson's correlations for the
relationships between the HTQ, its subscales, subclinical
OCD symptomatology, and the COHS, an existing
measure of habitual tendencies, were then examined in
order to determine the construct validity of the HTQ.
There was a significant positive correlation between HTQ
Compulsivity and the OCI, as in Study 1, suggesting that
F I G U R E 6 Scatter plot showing correlations between the
individuals prone to compulsivity show increased sub-
Obsessive–Compulsive Inventory (OCI) and the Compulsivity
clinical OCD symptomatology. Similarly, a significant
subscale of the Habitual Tendencies Questionnaire (HTQ)
positive correlation was found between HTQ Regularity
and the Routine subscale of the COHS, suggesting that
they tap into similar constructs. Hierarchical regression
COHS followed an approximately normal distribution, demonstrated that the three subscales of the HTQ explain
according to the Shapiro–Wilk test (p = 0.146), with min- a significant proportion of the variance in subclinical
imal skewness ( 0.266) and kurtosis ( 0.243). The mean OCD symptomatology, and furthermore, revealed HTQ
total score on the COHS was 87.599 (SD = 18.049), and it Compulsivity to be a significant predictor of subclinical
had a good Cronbach's α of 0.902, 95% CI = [0.884, OCD symptomatology, replicating the findings of Study
0.918]. Cronbach's α values for each of the subscales were 1. Therefore, the HTQ may be used as a validated tool to
0.887 for Routine, and 0.863 for Automaticity. CFA was measure individual variation in habitual tendencies, and
carried out on the COHS, which indicated that the two- its subscales may be valuable in predicting subclinical
factor structure was borderline acceptable (χ 2(323) traits in the general population.
= 743.635, p < 0.001, RMSEA = 0.073 [0.066, 0.080],
SRMR = 0.069, CFI = 0.824; TLI = 0.809). As evident in
Table 3, there were significant positive correlations 8 | DISCUSSION
between the HTQ and the COHS (r = 0.347, p < 0.001),
as well as their individual subscales. The largest of these The present study has developed and validated a key
correlations was between HTQ Regularity and COHS research tool for measuring individual differences in
Routine (r = 0.309, p < 0.001), suggesting that they may habitual tendencies, the HTQ. Through a rigorous pro-
represent similar constructs. cess of selection, 11 items were chosen for the final ver-
We then conducted a two-step hierarchical linear sion of the HTQ, and factor analysis revealed that these
regression with the three subscales of the HTQ as items clustered into three factors, representing three dis-
predictors of subclinical OCD symptomatology, and tinct aspects of the habitual tendencies construct: Com-
age and gender as covariates. Of the demographic vari- pulsivity, Preference for Regularity, and Aversion to
ables, only age was a significant predictor of subclini- Novelty. The three-factor structure of the HTQ was
cal OCD symptomatology (β = 0.283, t(249) = 4.602, reliably maintained across two independent samples,
p < 0.001), and of the three HTQ subscales, HTQ including a preregistered replication, and a combined
Compulsivity emerged as the most significant predictor summative analysis (supporting information), and was
of subclinical OCD symptomatology (β = 0.443, t(249) shown to be able to discriminate between various fea-
= 7.262, p < 0.001). The demographic variables alone tures of habitual tendencies (encompassing behaviours,
explained 7.9% of the variance in subclinical OCD attitudes, beliefs and thinking styles) in healthy
symptomatology (R2 = 0.079, F(2, 247) = 10.601, populations. Participants' scores on the Compulsivity
p < 0.001), but addition of the three subscales of the subscale of the HTQ consistently showed a significant
HTQ in Step 2 increased the R2 term to 0.259, strong positive correlation with their subclinical OCD
accounting for a further 18% of the variance in sub- symptoms, suggesting that individuals prone to compul-
clinical OCD symptomatology (R2 = 0.259, F(3, 244) sive thoughts and actions in their normative daily lives
= 17.029, p < 0.001). show increased subclinical OCD symptomatology.
12 RAMAKRISHNAN ET AL.

The present study highlights the importance of grouping of OCD symptoms into the different factors
recognising that habits are composed of different facets reflects the dissociation between obsessions (as Factor
that manifest in the daily lives of individuals to varying 1 consisted of symptoms relating to thinking style), and
degrees, and of making distinctions between these facets. compulsions (as Factors 2 and 3 consisted of symptoms
This is consistent with recent research such as that of relating to behaviours), and the distinct neural correlates
Hardwick et al. (2019), who found a difference between of these different dimensions reinforce the notion that
the formation of habits and their expression. They pro- OCD, and the habitual tendencies underlying it, are com-
pose that a stimulus triggers the preparation of a posed of various dimensions.
response, but that this response is not enacted immedi- An important future direction may be to explore
ately. Therefore, a more appropriate, goal-directed action whether the different components of the HTQ map onto
may replace the prepared response before it can be distinct neural circuitries in a similar way, as subclinical
initiated. In Hardwick et al.' (2019) study, participants OCD symptomatology is associated with HTQ Compul-
practised a visuomotor association task for 4 days. They sivity, but not with the other HTQ subscales.
then learned a new association, but when forced to Extensive evidence from neuroscientific studies of
respond rapidly, habitually expressed the old association. experimental animals and neuroimaging studies in
This demonstrates a dissociation between habit humans has supported the concept of dual systems of
formation and expression, which may be reflective of the behavioural control: a goal-directed system, implicating
different aspects of habitual tendencies encompassed by the ventromedial prefrontal cortex (vmPFC) and
the HTQ, such as thinking style (HTQ Compulsivity) or caudate nucleus, and a habit system that recruits the
attitude (HTQ Preference for Regularity), and behaviour putamen and premotor regions of cortex (Balleine &
(HTQ Aversion to Novelty). O'Doherty, 2010). Compulsivity as measured by question-
Elucidating the underlying components of habits has naire scales in a large sample of adolescents has been
important implications for our understanding of the linked to reduced white matter in dorsomedial and
antecedents of clinical disorders involving excessive dorsolateral PFC regions, especially including the
habits, such as OCD. Dissociations similar to those of anterior cingulate cortex and the ventral striatum.
Hardwick et al. (2019) have been made in relation to Moreover, compulsive behaviour in addiction and OCD,
different aspects of OCD, which may arise, in part, as a as measured respectively by the Obsessive–Compulsive
result of an over-reliance on habits, as well as deficits in Drug Use Scale (OCDUS) and the Yale–Brown
goal-directed control (e.g., Gillan et al., 2016). The Obsessive–Compulsive Scale (YBOCS) has been linked to
ego-dystonic nature of OCD means that patients possess structural changes or dysconnectivity of the ventromedial
the knowledge that their behaviour is irrational, and this and orbitofrontal PFC (Ersche et al., 2011; Meunier
has been demonstrated experimentally using a contin- et al., 2012). One interpretation of these findings is that
gency degradation task (Vaghi et al., 2019). Although underactivity in these PFC circuits leads either to an
patients with OCD showed exaggerated responding imbalance in the goal-directed versus habit systems, or a
compared with healthy controls, their action-outcome dysregulated control over the striatal habit system
contingency knowledge was intact, implying a divergence (e.g., Hardwick et al., 2019), thus linking compulsivity to
between their actions or behaviours; and their enhanced habits. Future studies should aim to link these
knowledge, or thinking style. Furthermore, it has been neural studies with experimental measures of habit
suggested that the obsessions (represented by habit learning and compulsivity scales, such as OCD with habit
learning) and compulsions (represented by habit perse- scales such as HTQ or COHS, to validate the laboratory
verance) underlying OCD may themselves be attributable test paradigms against habitual behaviour in the real
to distinct systems and even neural circuitries (Robbins world. For example, Ersche et al. (2021) showed that a
et al., 2019). Revisiting a PET study conducted by Rauch shift to habitual control, as assessed with contingency
et al. (1998), which explored the neural correlates of degradation procedure, was impaired in chronic cocaine
factor-analysed OCD symptoms, provides support for abusers, and additionally that contingency degradation
this. Religious, aggressive and sexual obsessions, and performance was positively related to the automaticity
checking compulsions (Factor 1 in Rauch et al., 1998) score on the COHS, which in turn in this group, was
were positively associated with bilateral striatal activity, significantly related to reductions in glutamate turnover
whereas symmetry and ordering symptoms (Factor 2 in in the putamen.
Rauch et al., 1998) were negatively associated with Additionally, the present study demonstrated robust
right caudate nucleus activity. Washing and cleaning individual differences in habitual tendencies, which may
symptoms (Factor 3 in Rauch et al., 1998) were positively help to explain past inconsistencies between animal and
associated with activity in several prefrontal areas. The human research, and between empirical and theoretical
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 13

work. De Wit et al. (2018) attempted to induce habits in be a role for the administration of other cognitive tasks,
human participants using five outcome devaluation tasks, such as the Alternative Uses Task (Guilford, 1967;
but this was unsuccessful, leading them to conclude that Ionescu, 2012; Zmigrod et al., 2019), along with the HTQ,
these tasks are mainly a measure of goal-directed control, to improve our understanding of how individual
and thus compulsive individuals perform less well in variation in cognitive inflexibility may moderate habitual
these tasks due to impaired goal-directed control rather tendencies, and contribute to disorders involving these
than overactive habit learning. However, individual dif- traits (Ramakrishnan et al., in prep). Indeed, this is
ferences in habitual tendencies were unaccounted for in particularly important given the diversity of definitions
most of these experiments or deliberately cancelled out. offered for compulsivity and the endeavour to create a
We propose that individual differences in susceptibilities dimension-based psychiatric approach to compulsive
to habitual tendencies may have a significant impact on disorders and behaviour (Albertella et al., 2019; Dajani &
the findings, or lack thereof, in studies such as this one. Uddin, 2015; Luigjes et al., 2019).
A study conducted by Luijten et al. (2020) provides sup- Another important future direction would be to explore
port for this view. This study found that although there relationships between the HTQ and various clinical disor-
was no difference in habitual versus goal-directed control ders associated with habits (Gillan et al., 2014; Gillan &
between smokers and non-smoking controls in outcome Robbins, 2014; Gillan & Sahakian, 2015). This could be
devaluation tasks, individual differences in nicotine achieved by administering the HTQ to those with clinically
dependence within the smoking group were positively diagnosed OCD, as well as to alternative populations of
correlated with habitual responding after appetitive individuals with disorders involving compulsivity, such as
instrumental learning, modelling positive reinforcement. addictions and binge eating disorders, which have been
This emphasises the importance of individual differences suggested to involve ‘deficits in goal-directed control and
in this field of research and suggests that individual associated over-reliance on habits’ (Gillan et al., 2016,
variation in susceptibilities to habits must be taken into p. 836). Another disorder that may be of interest to study
account in order to effectively manipulate habitual in relation to habitual tendencies is autism and autism
tendencies. spectrum disorders, as the present study found a significant
The results obtained in the present study possess positive correlation between the HTQ and the Autism
important implications for future research and interven- Quotient (Allison et al., 2012), (see supporting information,
tion. As the present study used an online convenience Additional Analyses and Table S5) and there is often
sample, replication of these findings in countries other comorbidity between OCD and autism, as well as overlap
than the United States would be useful in order to in their symptomatology (Leyfer et al., 2006). Thus, future
explore whether the present findings are consistent investigations may identify convergences and divergences
across cultural contexts. In addition, the 11-item HTQ in patterns of habitual tendencies across different
may be used in individual difference research on clinical disorders.
habitual tendencies and their associations with other To conclude, the present study developed and
constructs, such as personality traits and political views validated a novel, representative measure of habitual
(e.g., Zmigrod, 2020; Zmigrod et al., 2015, 2018, 2020). It tendencies, the HTQ, which has good reliability and
may also be fruitful to extend the present findings by validity. The HTQ may prove useful in future research
using existing behavioural measures of habits, such as into habitual tendencies, including in relation to
the Fabulous Fruit Game (de Wit et al., 2007) and compulsivity disorders such as OCD, potentially contrib-
outcome devaluation paradigms (e.g., Gillan et al., 2011), uting to the development of interventions targeting the
as well as measures of goal-directed control, such as con- maladaptive habits proposed to underlie OCD. To return
tingency degradation paradigms (e.g., Vaghi et al., 2019), to William James's Habit (1890, pp. 3–4), it ‘thus appears
in conjunction with the HTQ in order to strengthen the that habit covers a very large part of life, and that one
reliability and validity of the present findings and to engaged in studying the objective manifestations of mind
contribute to theories that try to understand the causal is bound at the very outset to define clearly just what its
mechanisms that make some individuals more suscepti- limits are’.
ble to habitual tendencies, and more specifically, to
compulsive thinking. It has been suggested that many of A C KN O WL ED G EME N T S
the supposed behavioural measures of habits in fact We would like to acknowledge Bushra Zafar, Paul
measure impaired goal-directed control (De Wit Matthews and Chessie Broadhurst for research assistance
et al., 2018), and as such, there may be a need for the and help in developing the Habitual Tendencies
development of novel behavioural paradigms that Questionnaire. This research was made possible through
measure habits more specifically. In addition, there may a Gates Cambridge Scholarship and a Junior Research
14 RAMAKRISHNAN ET AL.

Fellowship by Churchill College, Cambridge, to Leor Cheung, J. H., Burns, D. K., Sinclair, R. R., & Sliter, M. (2017).
Zmigrod, and a Wellcome Trust Senior Investigator Amazon Mechanical Turk in organizational psychology: An
Grant 104631/z/14/z to Trevor Robbins. evaluation and practical recommendations. Journal of Business
and Psychology, 32(4), 347–361. https://doi.org/10.1007/s10869-
016-9458-5
CONFLICT OF INTEREST Costello, A. B., & Osborne, J. (2005). Best practices in exploratory
There are no conflicts of interest. factor analysis: Four recommendations for getting the most
from your analysis. Practical Assessment, Research, and Evalua-
E TH IC S ST A T EME N T tion, 10(1), 7.
Ethical approval for the study was acquired from the Dajani, D. R., & Uddin, L. Q. (2015). Demystifying cognitive flexi-
Department of Psychology Ethics Committee of the bility: Implications for clinical and developmental neurosci-
ence. Trends in Neurosciences, 38(9), 571–578. https://doi.org/
University of Cambridge. In line with the Declaration
10.1016/j.tins.2015.07.003
of Helsinki (1964), electronic informed consent was
De Houwer, J. (2019). On how definitions of habits can complicate
obtained from all participants before beginning the habit research. Frontiers in Psychology, 10, 2642.
survey, and participants were notified that they may De Wit, S., Kindt, M., Knot, S. L., Verhoeven, A. A., Robbins, T. W.,
terminate their participation in the study at any point. Gasull-Camos, J., Evans, M., Mirza, H., & Gillan, C. M. (2018).
Shifting the balance between goals and habits: Five failures in
DATA AVAILABILITY STATEMENT experimental habit induction. Journal of Experimental Psychol-
Data will be provided upon request. ogy: General, 147(7), 1043.
de Wit, S., Niry, D., Wariyar, R., Aitken, M. R. F., & Dickinson, A.
(2007). Stimulus-outcome interactions during instrumental dis-
ORCID
crimination learning by rats and humans. Journal of Experi-
Trevor W. Robbins https://orcid.org/0000-0003-0642- mental Psychology: Animal Behavior Processes, 33(1), 1–11.
5977 https://doi.org/10.1037/0097-7403.33.1.1
Leor Zmigrod https://orcid.org/0000-0001-8270-7955 Ersche, K. D., Barnes, A., Jones, P. S., Morein-Zamir, S.,
Robbins, T. W., & Bullmore, E. T. (2011). Abnormal structure
R EF E RE N C E S of frontostriatal brain systems is associated with aspects of
Ajzen, I. (1989). Attitude structure and behavior. Attitude Structure impulsivity and compulsivity in cocaine dependence. Brain,
and Function, 241, 274. 134, 2013–2024. https://doi.org/10.1093/brain/awr138 PMID:
Albertella, L., Le Pelley, M. E., Chamberlain, S. R., Westbrook, F., 21690575
Fontenelle, L. F., Segrave, R., Lee, R., Pearson, D., & Yücel, M. Ersche, K. D., Lim, T. V., Murley, A. G., Rua, C., Vaghi, M. M.,
(2019). Reward-related attentional capture is associated with White, T. L., Williams, G. B., & Robbins, T. W. (2021).
severity of addictive and obsessive–compulsive behaviors. Reduced glutamate turnover in the putamen is linked with
Psychology of Addictive Behaviors, 33(5), 495. automatic habits in human cocaine addiction. Biological
Allison, C., Auyeung, B., & Baron-Cohen, S. (2012). Toward brief Psychiatry, 89(10), 970–979. https://doi.org/10.1016/j.biopsych.
“red flags” for autism screening: The short autism spectrum 2020.12.009
quotient and the short quantitative checklist in 1,000 cases and Ersche, K. D., Lim, T. V., Ward, L. H., Robbins, T. W., & Stochl, J.
3,000 controls. Journal of the American Academy of Child & (2017). Creature of habit: A self-report measure of habitual rou-
Adolescent Psychiatry, 51(2), 202–212. https://doi.org/10.1016/j. tines and automatic tendencies in everyday life. Personality and
jaac.2011.11.003 Individual Differences, 116, 73–85. https://doi.org/10.1016/j.
Armitage, C. J., & Christian, J. (2003). From attitudes to behaviour: paid.2017.04.024
Basic and applied research on the theory of planned behaviour. Ersche, K. D., Ward, L. H., Lim, T. V., Lumsden, R. J., Sawiak, S. J.,
Current Psychology, 22(3), 187–195. Robbins, T. W., & Stochl, J. (2019). Impulsivity and compulsiv-
Balleine, B. W., & O'Doherty, J. P. (2010). Human and rodent ity are differentially associated with automaticity and routine
homologies in action control: Corticostriatal determinants of on the Creature of Habit Scale. Personality and Individual Dif-
goal-directed and habitual action. Neuropsychopharmacology, ferences, 150, 109493.
35, 48–69. https://doi.org/10.1038/npp.2009.131 Everitt, B. J., & Robbins, T. W. (2016). Drug addiction: Updating
Bari, A., & Robbins, T. W. (2013). Inhibition and impulsivity: Behav- actions to habits to compulsions ten years on. Annual Review of
ioral and neural basis of response control. Progress in Neurobiol- Psychology, 67, 23–50. https://doi.org/10.1146/annurev-psych-
ogy, 108, 44–79. https://doi.org/10.1016/j.pneurobio.2013.06.005 122414-033457
Bar-Tal, Y. (1994). The effect on mundane decision-making of the Foa, E. B., Huppert, J. D., Leiberg, S., Langner, R., Kichic, R.,
need and ability to achieve cognitive structure. European Jour- Hajcak, G., & Salkovskis, P. M. (2002). The obsessive-
nal of Personality, 8(1), 45–58. compulsive inventory: Development and validation of a short
Brown, T. A. (2015). Confirmatory factor analysis for applied version. Psychological Assessment, 14(4), 485–496.
research. New York, NY, USA: Guilford publications. Gardner, B. (2015). A review and analysis of the use of ‘habit'in
Carleton, R. N., Norton, M. P. J., & Asmundson, G. J. (2007). understanding, predicting and influencing health-related
Fearing the unknown: A short version of the Intolerance of behaviour. Health Psychology Review, 9(3), 277–295. https://doi.
Uncertainty Scale. Journal of Anxiety Disorders, 21(1), 105–117. org/10.1080/17437199.2013.876238
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 15

Gignac, G. E., & Szodorai, E. T. (2016). Effect size guidelines for development and rates of disorders. Journal of Autism and
individual differences researchers. Personality and Individual Developmental Disorders, 36(7), 849–861. https://doi.org/10.
Differences, 102, 74–78. 1007/s10803-006-0123-0
Gillan, C. M., Apergis-Schoute, A. M., Morein-Zamir, S., Luigjes, J., Lorenzetti, V., de Haan, S., Youssef, G. J., Murawski, C.,
Urcelay, G. P., Sule, A., Fineberg, N. A., Sahakian, B. J. & Sjoerds, Z., van den Brink, W., Denys, D., Fontenelle, L. F., &
Robbins, T. W. (2015). Functional neuroimaging of avoidance Yücel, M. (2019). Defining compulsive behavior. Neuropsychol-
habits in obsessive-compulsive disorder. American Journal of ogy Review, 29(1), 4–13.
Psychiatry, 172(3), 284–293. Luijten, M., Gillan, C. M., De Wit, S., Franken, I. H.,
Gillan, C. M., Morein-Zamir, S., Urcelay, G. P., Sule, A., Voon, V., Robbins, T. W., & Ersche, K. D. (2020). Goal-directed and
Apergis-Schoute, A. M., Fineberg, N. A., Sahakian, B. J., & habitual control in smokers. Nicotine & Tobacco Research, 22
Robbins, T. W. (2014). Enhanced avoidance habits in obsessive- (2), 188–195. https://doi.org/10.1093/ntr/ntz001
compulsive disorder. Biological Psychiatry, 75(8), 631–638. Maier, T. (2004). On phenomenology and classification of hoarding:
https://doi.org/10.1016/j.biopsych.2013.02.002 A review. Acta Psychiatrica Scandinavica, 110(5), 323–337.
Gillan, C. M., Papmeyer, M., Morein-Zamir, S., Sahakian, B. J., https://doi.org/10.1111/j.1600-0447.2004.00402.x
Fineberg, N. A., Robbins, T. W., & de Wit, S. (2011). Disruption Mazar, A., & Wood, W. (2018). Defining habit in psychology. In B.
in the balance between goal-directed behavior and habit learn- Verplanken (Ed.), The psychology of habit. Cham: Springer.
ing in obsessive-compulsive disorder. American Journal of Psy- https://doi.org/10.1007/978-3-319-97529-0_2
chiatry, 168(7), 718–726. Meade, A. W., & Craig, S. B. (2012). Identifying careless responses
Gillan, C. M., & Robbins, T. W. (2014). Goal-directed learning and in survey data. Psychological Methods, 17(3), 437–455. https://
obsessive–compulsive disorder. Philosophical Transactions of doi.org/10.1037/a0028085
the Royal Society, B: Biological Sciences, 369(1655), 20130475. Meunier, D., Ersche, K. D., Craig, K. J., Fornito, A., Merlo-
Gillan, C. M., Robbins, T. W., Sahakian, B. J., van den Pich, E. M., Fineberg, N. A., Shabbir, S. S., Robbins, T. W., &
Heuvel, O. A., & van Wingen, G. (2016). The role of habit in Bullmore, E. T. (2012). Brain functional connectivity in stimu-
compulsivity. European Neuropsychopharmacology, 26(5), lant drug dependence and obsessive-compulsive disorder.
828–840. https://doi.org/10.1016/j.euroneuro.2015.12.033 NeuroImage, 59, 1461–1468. https://doi.org/10.1016/j.
Gillan, C. M., & Sahakian, B. J. (2015). Which is the driver, the obses- neuroimage.2011.08.003 PMID: 21871569
sions or the compulsions, in OCD? Neuropsychopharmacology, Piquet-Pessôa, M., Chamberlain, S. R., Lee, R. S., Ferreira, G. M.,
40(1), 247–248. https://doi.org/10.1038/npp.2014.201 Cruz, M. S., Ribeiro, A. P., de Menezes, G., Albertella, L.,
Grodin, E. N., Bujarski, S., Venegas, A., Baskerville, W. A., Yücel, M., & Fontenelle, L. F. (2019). A study on the correlates
Nieto, S. J., Jentsch, J. D., & Ray, L. A. (2019). Reward, relief of habit-, reward-, and fear-related motivations to use alcohol
and habit drinking: Initial validation of a brief assessment tool. in alcohol use disorder. CNS Spectrums, 24(6), 597–604. https://
Alcohol and Alcoholism, 54(6), 574–583. doi.org/10.1017/S1092852918001554
Guilford, J. P. (1967). The nature of human intelligence. New York: Ramakrishnan, S., Robbins, T., & Zmigrod, L. (in prep). Cognitive
McGraw-Hill. Rigidity, Habitual Tendencies, and Obsessive-Compulsive
Hagger, M. S. (2019). Habit and physical activity: Theoretical Symptoms: Individual Differences and Interactions in a
advances, practical implications, and agenda for future Healthy Population.
research. Psychology of Sport and Exercise, 42, 118–129. Rauch, S. L., Dougherty, D. D., Shin, L. M., Alpert, N. M.,
Hardwick, R. M., Forrence, A. D., Krakauer, J. W., & Haith, A. M. Manzo, P., Leahy, L., Fischman, A. J., Jenike, M. A., & Baer, L.
(2019). Time-dependent competition between goal-directed and (1998). Neural correlates of factor-analyzed OCD symptom
habitual response preparation. Nature Human Behaviour, 3(12), dimensions: A PET study. CNS Spectrums, 3(7), 37–43.
1252–1262. https://doi.org/10.1038/s41562-019-0725-0 Ray, L. A., Du, H., Grodin, E., Bujarski, S., Meredith, L., Ho, D.,
IBM Corp. (2020). IBM SPSS statistics for windows, version 27.0. Nieto, S., & Wassum, K. (2020). Capturing habitualness of
Armonk, NY: IBM Corp. drinking and smoking behavior in humans. Drug and Alcohol
Ionescu, T. (2012). Exploring the nature of cognitive flexibility. New Dependence, 207, 107738.
Ideas in Psychology, 30(2), 190–200. Robbins, T. W., & Costa, R. M. (2017). Habits. Current Biology,
James, W. (1890). The principles of psychology, in two volumes. 27(22), R1200–R1206. https://doi.org/10.1016/j.cub.2017.09.060
New York: Henry Holt and Company. Robbins, T. W., Gillan, C. M., Smith, D. G., de Wit, S., &
JASP Team. (2020). JASP (version 0.12) [Computer software]. Ersche, K. D. (2012). Neurocognitive endophenotypes of impul-
Kruglanski, A. W. (1990). Motivations for judging and knowing: sivity and compulsivity: Towards dimensional psychiatry.
Implications for causal attribution. Trends in Cognitive Sciences, 16(1), 81–91. https://doi.org/10.
Krumrei-Mancuso, E. J., & Rouse, S. V. (2016). The development 1016/j.tics.2011.11.009
and validation of the comprehensive intellectual humility scale. Robbins, T. W., Vaghi, M. M., & Banca, P. (2019). Obsessive-
Journal of Personality Assessment, 98(2), 209–221. https://doi. compulsive disorder: Puzzles and prospects. Neuron, 102(1),
org/10.1080/00223891.2015.1068174 27–47. https://doi.org/10.1016/j.neuron.2019.01.046
Lally, P., & Gardner, B. (2013). Promoting habit formation. Health RStudio Team. (2020). RStudio: Integrated development for R.
Psychology Review, 7(sup1), S137–S158. Boston, MA: RStudio, Inc.. http://www.rstudio.com/
Leyfer, O. T., Folstein, S. E., Bacalman, S., Davis, N. O., Dinh, E., Sjoerds, Z., de Wit, S., van den Brink, W., Robbins, T. W.,
Morgan, J., Tager-Flusberg, H., & Lainhart, J. E. (2006). Comor- Beekman, A. T., Penninx, B. W., & Veltman, D. J. (2013).
bid psychiatric disorders in children with autism: Interview Behavioral and neuroimaging evidence for overreliance on
16 RAMAKRISHNAN ET AL.

habit learning in alcohol-dependent patients. Translational Zmigrod, L. (2020). The role of cognitive rigidity in political ideolo-
Psychiatry, 3(12), e337–e337. gies: Theory, evidence, and future directions. Current Opinion
Stanley Budner, N. Y. (1962). Intolerance of ambiguity as a person- in Behavioral Sciences, 34, 34–39.
ality variable 1. Journal of Personality, 30(1), 29–50. Zmigrod, L., Rentfrow, P. J., & Robbins, T. W. (2018). Cognitive
Thompson, B. (2007). Factor analysis. The Blackwell Encyclopedia of underpinnings of nationalistic ideology in the context of Brexit.
Sociology. Proceedings of the National Academy of Sciences, 115(19),
Vaghi, M. M., Cardinal, R. N., Apergis-Schoute, A. M., E4532–E4540.
Fineberg, N. A., Sule, A., & Robbins, T. W. (2019). Action- Zmigrod, L., Rentfrow, P. J., & Robbins, T. W. (2020). The partisan
outcome knowledge dissociates from behavior in obsessive- mind: Is extreme political partisanship related to cognitive
compulsive disorder following contingency degradation. inflexibility? Journal of Experimental Psychology: General,
Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 149(3), 407.
4(2), 200–209. Zmigrod, S., Zmigrod, L., & Hommel, B. (2015). Zooming into crea-
van't Riet, J., Sijtsema, S. J., Dagevos, H., & De Bruijn, G. J. (2011). tivity: Individual differences in attentional global-local biases
The importance of habits in eating behaviour. An overview are linked to creative thinking. Frontiers in Psychology, 6, 1647.
and recommendations for future research. Appetite, 57(3), Zmigrod, L., Zmigrod, S., Rentfrow, P. J., & Robbins, T. W. (2019).
585–596. The psychological roots of intellectual humility: The role of
Verplanken, B., & Aarts, H. (1999). Habit, attitude, and planned intelligence and cognitive flexibility. Personality and Individual
behaviour: Is habit an empty construct or an interesting case of Differences, 141, 200–208.
automaticity? European Review of Social Psychology, 10,
101–134.
Verplanken, B., & Orbell, S. (2003). Reflections on past behavior: A SU PP O R TI N G I N F O RMA TI O N
self-report index of habit strength 1. Journal of Applied Social Additional supporting information may be found online
Psychology, 33(6), 1313–1330. in the Supporting Information section at the end of this
Wagenmakers, E. J., Love, J., Marsman, M., Jamil, T., Ly, A., article.
Verhagen, J., Selker, R., Gronau, Q. F., Dropmann, D.,
Boutin, B., Meerhoff, F., Knight, P., Raj, A., van Kesteren, E.-J.,
van Doorn, J., Šmíra, M., Epskamp, S., Etz, A., Matzke, D., …
Morey, R. D. (2018). Bayesian inference for psychology.
How to cite this article: Ramakrishnan, S.,
Part II: Example applications with JASP. Psychonomic Robbins, T. W., & Zmigrod, L. (2021). The Habitual
Bulletin & Review, 25(1), 58–76. https://doi.org/10.3758/s13423- Tendencies Questionnaire: A tool for psychometric
017-1323-7 individual differences research. Personality and
Wood, W., & Neal, D. T. (2007). A new look at habits and the habit- Mental Health, 1–17. https://doi.org/10.1002/
goal interface. Psychological Review, 114(4), 843–863. https:// pmh.1524
doi.org/10.1037/0033-295X.114.4.843
THE HABITUAL TENDENCIES QUESTIONNAIRE: A TOOL FOR PSYCHOMETRIC INDIVIDUAL DIFFERENCES RESEARCH 17

A P P EN D I X

ELEVEN-ITEM HABITUAL TENDENCIES QUESTIONNAIRE

HTQ compulsivity

HTQ 37: I tend to dwell on the same issues

HTQ 36: I mentally fixate on certain issues and cannot move on
HTQ 35: The same thoughts often keep going through my mind over and over again
HTQ 33: I tend to repeat actions because I keep doubting that I have done them properly

HTQ preference for regularity

HTQ 10: I like to have a regular, unchanging schedule

HTQ 9: There is comfort in regularity.
HTQ 27: A good job has clear guidelines on what to do and how to do it
HTQ 1: I hate it when my routines are disrupted

HTQ aversion to novelty

HTQ 30: I look forward to new experiences R

HTQ 26: Life is boring if you never take risks and always play it safe R
HTQ 7: When eating at restaurants, I like to try new dishes rather than ones I have tried before R