Attitude Versus General Habit:

Antecedents of Travel Mode Choice1

BAS VERPLANKEN? HENKAARTS, AD VAN KNIPPENBERG

University of Nijmegen. The Netherlands

CARINA VAN KNIPPENBERG

Bureau Goudappel Coffeng, Deventer, The Netherlands

A model of travel mode choice is tested by means of a survey among 199

inhabitants of a village. Car choice behavior for a particular journey is predicted
from the attitude toward choosing the car and the attitude toward choosing an
alternative mode (i.e., train), on the one hand, and from general car habit, on the
other hand. Unlike traditional measures of habit, a script-based measure was used.
General habit was measured by travel mode choices in response to very global
descriptions of imaginary journeys. In the model, habit is predicted from the degree
of involvement with the decision-making about travel mode choice for the particular
journey (decisional involvement) and from the degree of competition in a household
with respect to car use. The model proves satisfactory. Moreover, as suggested by
Triandis (1977), there is a tradeoff between attitude and habit in the prediction of
behavior: When habit is strong the attitude-behavior relation is weak, whereas when
habit is weak, the attitude-behavior link is strong.

Introduction

One of the characteristics of modem industrialized society is the high

degree of physical mobility. Transport is, therefore, an essential element of
daily life in our Western culture. We enjoy the luxury of being able to choose
from a large set of travel modes. In several Western countries, however, there
is a growing concern that the rapid growth of car traffic may pose serious
problems such as congestion and pollution. Therefore, attempts are made by
local and national authorities to regulate individuals’ choices of travel mode,
that is, to get people to use public transport more frequently or to use their cars
more efficiently, for example, by promoting car pools.

‘This study was supported by a grant from the Ministry of Transport and Public Works,
The Netherlands. The authors wish to thank Birgit van Delft, Benno Labree, Hanneke Karthaus,
Pieter Meima, Dries Vermeulen, and Miranda Wildenbeest for their contribution in collecting
the data, and Ola Svenson for his comments on the manuscript.
2Correspondence concerning this article should be addressed to Bas Verplanken, University
of Nijmegen, Department of Social Psychology, PO Box 9104, 6500 HE Nijmegen, The
Netherlands; e-mail: verplanken@psych.kun.nl.

Journal ofApplied Social Psychology, 1994. 24, 4, pp. 285-300.

Copyright 0 1994 by V. H. Winston & Son, Inc. All rights reserved.
286 VERPLANKEN ET AL.

Several regulatory strategies may be considered to influence individuals’

travel behavior (Geller, Winett, & Everett, 1982). Some measures contain
reinforcers to accomplish this goal, for example, providing free passage by
designating priority lanes for car pools (Geller, Kalsher, Rudd, & Lehman,
1989). Other measures rely on economic principles and imply punishment
for unwanted behavior such as increasing parking prices, or raising taxes on
car use. An alternative approach is to influence travel mode behavior by means
of educational programs. For example, information campaigns are used to
induce positive attitudes toward the use of public transport, to emphasize
unfavorable environmental effects of private car use, or to promote energy
conservation.
Most strategies aimed at behavioral changes are implicitly or explicitly
based on the assumption that behavior is the result of a tradeoff between
perceived costs and benefits. On the cost side of the use of private cars are, for
example, monitary costs, loss of time and annoyance when confronted with
traffic congestion, or the probability of accidents. Benefits may include expe-
rienced comfort, freedom, or cargo capacity. Expectancy-value models of
attitudes describe how perceptions of such attributes combine into a general
evaluative judgment (i.e., an attitude; e.g., Fishbein & Ajzen, 1975). Attitudes,
then, are supposed to guide behavior through the mediating role of behavioral
intentions (Bagozzi, 1981; Fishbein & Ajzen, 1975).

Attitudes Versus Habits

Travel mode behavior can be considered as being based on attitudes toward

available mode options, in particular on the attitude toward the most promising
alternative in the choice set. This perspective has been widely recognized in the
field of travel mode choice research (e.g., Dobson, Dunbar, Smith, Reibstein,
& Lovelock, 1978; Gilbert & Foerster, 1977; Hartgen, 1974; Recker & Golob,
1976; Reibstein, Lovelock, & Dobson, 1980). Apart from the question of which
strategy is followed in choosing between options, the behavior that is ulti-
mately chosen is, at least to some extent, supposed to be based on a process of
reasoning, in which costs and benefits are considered (cf. Fishbein & Ajzen,
1975). In the domain of travel mode choice behavior, however, this approach
fails to recognize an important aspect of mode choice, that is, its repetitive
character. We rarely consider travel mode for a particular journey for the first
time. First, in a more specific sense, we often make exactly the same journey
under the same circumstances in the same way time and again. In any subse-
quent journey, we then need not deliberate on travel mode choice at all. Second,
in a more general sense, we may have a broad range of experiences with all
sorts of journeys and all kinds of travel modes in a variety of circumstances.
 ATTITUDE VERSUS GENERAL HABIT 287

When confronted with trips we are sufficiently familiar with, such experiences
may enable us to arrive at travel mode choices in a rather mindless, habitual
manner. In other words, on such occasions, travel mode choice is a matter of
habit, rather than reasoned action. Habits are relatively stable behavioral
patterns, which have been reinforced in the past. Habits are executed without
deliberate consideration, and result from automatic processes, as opposed to
controlled processes like consciously made decisions (cf. Shiffrin & Schneider,
1977).
The relationship among an attitudinal component (in the form of behav-
ioral intention), habit, and behavior was formalized by Triandis (1977). In
his model, the probability of an act (Pa)is a weighted function of habit (H)
and behavioral intention (I), multiplied by facilitating conditions Q, that is,
the presence or absence of conditions that facilitate the performance of the
act (e.g., the individual’s ability to perform the act). This relationship is
expressed as:

The weights Wh and wi are supposed to be negatively correlated. New behavior

is assumed to be completely under the control of intentions. As the behavior
occurs more frequently, Wh increases while wi decreases. In other words, the
model implies a tradeoff in importance between habit and behavioral intention
with respect to the probability of an act. Or, as Triandis (1977) expresses this:
“...when a behavior is new, untried, and unlearned, the behavioral-intention
component will be solely responsible for the behavior, while, when the behav-
ior is old, well learned, or overlearned and has occurred many times before in
the organism’s life span, it is very likely to be under control of the habit
component” (p. 205).
Evidence for the predictive power of past behavior and habit with respect
to future behavior, next to attitude and intention, has cumulated the past one
and a half decade (e.g., Bagozzi, 1981; Bentler & Speckart, 1979; Fredricks &
Dossett, 1983; Kahle, 1984; Kahle, Klingel, & Kulka, 1981; Landis, Triandis,
& Adamopoulos, 1978; Mittal, 1988; Triandis, 1977; Wittenbraker, Gibbs, &
Kahle, 1983). This especially concerns daily behaviors of a repetitive nature.
Wittenbraker et al. (1983), for example, investigated seat belt usage among
psychology students. They measured, among other aspects, self-reported be-
havior, habit, and intention in a longitudinal cross-lagged panel design. Their
measure of habit was a stronger predictor of subsequent behavior than behav-
ioral intention. Mittal (1988) broadened the scope with respect to the role of
habit in a study about seat belt usage among adult residents by measuring not
only habit of seat belt use behavior but also habit of not using seat belts. The
288 VERPLANKEN ET AL.

results of that study suggest that both the use and nonuse habit (nonredun-
dantly) predict behavior, next to attitude and intention. Moreover, the nonuse
habit interacted with intention in its prediction of behavior.
Like seat belt use, the repetitive nature of travel mode choice makes a
role of habit in this behavior domain very likely (cf. Banister, 1978). Apart
from attitudes, therefore, habit may be included in models predicting travel
mode choice. It may be predicted that (a) the more favorable the attitude to-
ward a particular travel mode (and the more unfavorable the attitude toward
alternative modes), the more likely is the choice of that mode and (b) the more
strongly the habit of using a particular mode, the more likely is the choice of
that mode. It is worth noting that, in addition to the independent effects of
attitude and habit hypothesized above, the argument of Triandis (1977) sug-
gests that there is also an interaction effect of attitude and habit on choice:
When habit is strong, the attitude-behavior relation will be weak and, con-
versely, when habit is weak, the attitude-behavior relation will be strong (i.e.,
habit serves as a moderator variable; cf. Baron & Kenney, 1986).

Antecedents of Habit

In the tradition of behaviorism, habit formation is considered a learning

process (e.g., Hull, 1943). Increments of habit strength are accomplished by
repetitions of reinforcements. For example, every time a particular journey is
made by car, the satisfaction of using the car reinforces car choice behavior
and, thus, contributes to the strength of a car choice habit. Next to rein-
forcement of the particular behavior, however, repetitively and satisfactorily
making the same decision over and again may have another consequence
namely, that there is gradually less need to consider alternative choices. In
other words, the decision makers’ involvement with that particular decision
decreases as they make the same decision repeatedly. When this motivation to
evaluate alternatives diminishes, there is less need to deliberate actively about
the pros and cons of options, search for information (internally or externally),
or otherwise expend mental effort in preparation for the particular trip. In the
present study, we attempted to investigate such motivation, which will be
referred to as decisional involvement (cf. Houston & Rothschild, 1978; Mittal,
1989; Mittal & Lee, 1989). Decisional involvement is expected to be nega-
tively related to habit.
Obviously, in order for habit to develop, physical circumstances must be
favorable. In the present context of travel mode choice, development of a car
habit, for example, is directly dependent upon the availability of a car, having
a drivers license, and the number of competitors in a household who make use
of the same car (e.g., Banister, 1978).
 ATTITUDE VERSUS GENERAL HABIT 289

Operationalizations of Habit

The measurement of habit is difficult (cf. Mittal, 1988). Whereas behavior-

ists measure habit by the number of times the behavior has been reinforced
(e.g., Hull, 1943), social psychologists usually measure habit by having respon-
dents report on the frequency with which they have performed the behavior in
the past (e.g., Bagozzi, 1981; Landis et al., 1978; Triandis, 1977). Wittenbraker
et al. (1983), measuring seat belt use habit, asked respondents to indicate,
“How many times in the last two weeks when driving a car have you put on a
seat belt by force of habit” (p. 41 1). Mittal (1988) makes an important point
concerning such operationalizations when he states that “Repeated occurrence
is necessary for the formation of habit, but it is not habit itself’ (p. 997). Prior
and later behavior may correlate because other determinants than habit (e.g.,
external conditions) may operate on both occasions (Beck & Ajzen, 1991;
Ronis, Yates, & Kirscht, 1989). Mittal(l988) stresses the fundamental feature
of habit-that is, its occurrence without awareness, and measures habit of
use and nonuse of seat belts by asking respondents to provide self-reports of
their awareness of using and not using seat belts (e.g., “During the past few
weeks, when I got into my car, I was not even aware and I put on my seatbelt”
Mittal, 1988, p. 1001). We agree with Mittal’s (1988) objection to the use of
(self-reported) frequency of previous behavior as a measure of habit. On the
other hand, retrospectively reporting on internal processes (or the lack of
internal processes) is a very difficult task (cf. Nisbett & Wilson, 1977), which
makes such a measure less attractive.
Another aspect concerning the measurement of habit that has not been
addressed so far is that habit may not be restricted to one particular behavior in
a specific context (cf. the concept of stimulus generalization in learning theory;
e.g., Hull, 1943). For example, in explaining car choice, the choice may be
determined by general habit of car use (i.e., no matter which journey, one
chooses the car), rather than by (or above) habit with respect to one particular
journey. In that case, predicting behavior by general habit is more interesting
than focusing on journey-specific habit.
In this study, we have taken a different approach to the measurement of
habit than has been done by other researchers. We have attempted to construct
a measure of car choice habit that: (a) does not involve frequency of past
behavior, (b) is more general than with respect to one particular journey,
(c) does not rely on retrospective introspection, and (d) is easy to administer,
at least in a structured interview setting. We have presented our respondents
with ten short statements that very globally indicate imaginary journeys (e.g.,
“Suppose you go to the beach with some friends”). As a measure of car choice
habit, we then counted how many times the car was chosen as travel mode from
290 VERPLANKEN ET AL.

a set of six options (i.e., bus, bicycle, cab, car, train, or walking). The assump-
tion behind this operationalization is that, when respondents are asked to
respond to global stimuli, which are presented with a minimum of detail, there
is little possibility engaging in making elaborate tradeoffs between the pros and
cons of various travel modes. Rather, we expect that respondents’ responses are
guided by preexisting schemas or scripts about mode choice in general (cf.
Kahle, 1984; Kahle et al., 1981). When habit with respect to car choice is
strong, the choice of the car as the travel mode will dominate such schemas,
which is supposed to be reflected in a relatively large number of car choices
across the ten presented journeys.

A Conceptual Model of Car Choice Behavior

In the present study, a model of car choice behavior is tested. Self-reported

car choice is predicted by the attitude toward choosing a car and the attitude
toward choosing a train as a travel mode for a particular journey, on the one
hand, and the measure of general car choice habit, on the other hand. Although
attitudes and habit are supposed to be correlated, it is hypothesized that both
attitude components and habit have unique predictive value with respect to car
choice behavior. Furthermore, it is expected that there is a tradeoff between the
weights of car attitude and car habit in the prediction of car choice (i.e., a
unique contribution of their interaction). A relatively high correlation between
attitude and behavior is expected for individuals who have a weak car-choice
habit, whereas a low correlation is expected for those with a strong habit.
Finally, car habit is predicted from decisional involvement concerning mode
choice, on the one hand, and car availability, on the other hand.

Method

Respondents

A structured interview was conducted by trained interviewers among 258

adult residents of a village. This village is located exactly in between two cities
(further referred to as City A and City B; the distance to either city is approxi-
mately 5 miles), and it was expected that the residents of the village frequently
visit one or both towns. A highway, which is easily accessible from the village,
runs between the two cities. A railway line also serves the village frequently.
The target behavior is choosing the car as the travel method for a shopping trip
to either of the two cities.
The questionnaire contained items measuring car choice behavior, attitudes
toward choosing the car and train for the particular trip, general car-choice
 ATTITUDE VERSUS GENERAL HABIT 291

habit, and a scale for decisional involvement. The subjects also responded to
several demographic items such as age, gender, number of cars available in the
household, number of household members having a drivers license, and an
estimate of the annual mileage. The questionnaire contained a number of other
additional items, which are not relevant here.
For the analyses to be reported here only those respondents were selected
who had a drivers license, and had a car at their disposal ( N = 199). The ages
ranged from 19 to 65 ( M = 39.9). There were 92 males and 107 females. The
self-reported mean annual mileage is 9,956.

Measures

Car choice behavior was measured by asking the respondents how often
during the past 2 months they had traveled by car in order to go shopping in
either one of the two cities (cf. Mittal, 1988; Wittenbraker et al., 1983).
Responses were given on a 4-point scale ranging from never (1) to very
often (4).
Attitude toward traveling by car for shopping in either one of the two cities
was measured by averaging two bipolar 7-point items ranging from very bad to
very good and very unfavorable to very favorable respectively ( r = 0.58, p <
.OOl). The same items were used to measure attitude toward traveling by train
for shopping in either one of the two cities ( r = 0.64, p < .OOl).
General habit of choosing the car as the travel mode was measured by
presenting respondents with ten imaginary situations calling for a choice of
travel mode (e.g., “Going to the beach with some friends, Going for sporting
as leisure activity, Going to visit friends in the village, Going for shopping after
work”). The respondents were required to choose between six possible travel
modes for each particular journey. The six modes presented were bicycle, bus,
cab, car, train, and walking. The car choice habit measure was calculated by
counting the number of times the car was mentioned as the travel mode. Thus,
the measure could have values ranging from 0 to 10. The measure correlated
moderately, however significantly, with the respondents’ annual mileage ( r =
.37, p < .OOl). There is also a sizable correlation ( r = .55, p < .001) with one
additional item, which reflected some form of habit. It was worded “Whether I
go to work or go shopping, I almost always travel by car,” and it was accom-
panied by a 7-point disagree-agree scale.
Decisional involvement was measured with eight 7-point Likert-type agree-
disagree items. Because decisional involvement is supposed to be situation-
specific, the subjects were presented with the stem, “When I go shopping in
City A or City B, .” The items were, respectively: “I want to expend
effort to find out which travel method is the best, I think about which travel
292 VERPLANKEN ET AL.

method I want to use, there is no doubt in my mind about how I will travel, I
want to know in detail which pros and cons various travel methods have, I think
it is useless to expend time and energy to find out which travel method is most
suitable, I want to prepare the journey well before, I don’t need to deliberate
about how I want to travel because I already know, and I use the first method
of transportation that comes to mind.” The eight items form a reliable scale
(coefficient alpha = 0.83). After reversing the third, fifth, seventh, and eighth
item, the items were averaged into a score representing decisional involvement.
Car availability was measured by the ratio of the number of cars available
in the household (usually one) and the number of household members having
a drivers license. This measure is widely used in travel behavior research to
reflect the amount of competition in a household with respect to car use.

Results

Descriptive Statistics and Correlations

In Table 1, the means, standard deviations, and intercorrelations of the

measured variables are shown. It appears that respondents relatively often
choose the car as a travel mode for the particular shopping trip. Attitudes
toward car use and train use with respect to the shopping trip are generally
favorable, whereas decisional involvement is relatively low. Most households
have one car available, which is used by either one or two drivers license
holders. The habit measure indicated choice of the car on about half of the
hypothetical trips, indicating a moderate habit on average.

Path Analysis

The operationalizations of car choice behavior, attitudes toward car and

train, car-choice habit, decisional involvement, and car availability were sub-
jected to path analysis in order to test the model outlined in the introduction
section. In Figure 1, the model is presented, and the standardized path coeffi-
cients are indicated.
The antecedents of car choice behavior accounted for 47.8% of its variance.
Furthermore, all paths in the model had significant coefficients. Both the two
attitudes and habit had unique contributions in the prediction of behavior. As
was expected, attitude was not independent of habit, as was indicated by
weights around .30. Habit was satisfactorily predicted by decisional involve-
ment, on the one hand, and car availability, on the other hand. An overall test
of fit between the data and the model was obtained by using LISREL VII, in
which the analysis was based on the covariance matrix. The indicators of fit
Table 1

Mean, Standard Deviation, and Intercorrelations of Car Choice Behavior, Car Attitude, Train
Attitude, Car Habit, Decisional Involvement, and Car Availability
~

Measure and range A4 SD 2 3 4 5 6

-
1. Car choice behavior (1 -4) 3.37 0.99 0.57' -0.29' 0.54' -0.32' 0.13 2
L
2. Car attitude (1-7) 4.98 - -0.13 0.32' -0.18a 0.14a U
1.47 rn
3. Train attitude (1 -7) 4.56 1.52 - -0.29' 0.17a -0.09
4. Car habit (0-10) 5.58 2.18 - -0.36' 0.22'
5. Decisional involvement (1-7) 2.92 1.23 - 0.00
6. Car availability (>O) 0.69 0.28

ap < .05. bp < .01. 'p .001.

294 VERPLANKEN ET AL

attitude toward
train-choice

-29 attitude toward .44 car-choice

car-choice behavior

decisional general
involvement ' -.36 car-choice habit

y p availability

Figure I . Path analysis for antecedents of car choice behavior (the numbers are
standardized path coefficients; all coefficients are statistically significant at
.001-level, except the path between train attitude and car choice behavior,
which is significant at .05-level)

suggested that the model fits the data well, ~ ~ ( =87.89,

) p = .444; Adjusted
Goodness of Fit Index = .965.3

Investigating the Tradeoff Between Attitude and Habit

In order to test the hypothesized tradeoff between car attitude and car
choice habit in their prediction of car choice behavior, a stepwise hierarchical
multiple regression analysis was performed, in which behavior is predicted by
attitude, habit, and the Attitude x Habit interaction term. All variables were
first standardized. The prediction of car choice behavior by car attitude (R2 =
.33) was significantly improved by including car choice habit (R2 = .47;
F-change = 47.16, p < .001). Inclusion of the interaction term after attitude and
habit yielded a modest, however significant, improvement of the prediction (R2
= .49; F-change = 8.32, p < .005). In Table 2, the nature of the interaction is
shown. The sample was split as close as possible at the 33rd and 67th percen-
tiles of the distribution of the habit measure. Then, the correlation between car

3The directions of the paths are theory driven.

 ATTITUDE VERSUS GENERAL HABIT 295

Table 2

Correlation Between Car Choice Attitude

and Car Choice Behavior at Three Levels
of Car Habit

Weak habit ( n = 57) 0.65b

Moderate habit ( n = 76) 0.41b
Strong habit ( n = 66) 0.2P

"p < .05. bp < .001.

attitude and behavior was calculated for the three subgroups. This correlation
was stronger among the subgroup with relatively weak habit compared with
those who indicated a relatively strong habit. The difference between attitude-
behavior correlations of the strong versus weak habit subgroups was signifi-
cant, z = 2.67, p < .01. In other words, when habit was weak the attitude and
behavior measures shared 43% of their variance, whereas when habit was
strong, the attitude and behavior measures only shared 8% of their ~ a r i a n c e . ~

Discussion

The results of the present study suggest that habit is an important determi-
nant of travel mode choice. Car choice behavior was satisfactorily predicted
from attitudes toward choosing available options, on the one hand, and car
choice habit, on the other hand. Evidently, predicting mode choice behavior
from attitudes alone is not optimal because car choice habit also explains 14%
of variance. In this respect, mode choice resembles other highly repetitive
behavior, for which the predictive role of habit has been demonstrated (e.g.,
Kahle et al., 1981; Landis et al., 1978; Mittal, 1988; Wittenbraker et al., 1983).
It is worth emphasizing that the habit measure reported here is of a general

4 0 n e reviewer suggested an interesting analysis concerning the role of decisional

involvement in the tradeoff between the weights of attitude and habit in the prediction of
behavior. Theoretically, one would predict a relatively high weight given to attitude compared
with habit at a high level of decisional involvement, whereas a relatively high weight of habit
compared with attitude would be expected at a low level of decisional involvement. This
interaction was examined by inspecting the respective weights of attitude and habit within the
high versus low tertiles of the involvement distribution. Although the pattern of results are in
the expected direction (high involvement: beta-attitude = .45, beta-habit = .39; low involvement:
beta-attitude = .32, beta-habit = .52), the differences between groups do not reach statistical
significance.
296 VERPLANKEN ET AL.

nature. That is, rather than measuring a situation-specific habit (i.e., the habit
of choosing the car for this particular shopping trip), our measure attempted to
tap a generalized habit of choosing the car as the travel mode. Consequently,
the habit component is more powerful because it applies to a large variety of
mode choice contexts. Together with Mittal’s (1988) identification of use
versus nonuse habit, the present distinction between specific and general habit
may constitute a useful qualification of the construct of habit for future re-
search .
The regression results with respect to the interaction between attitude and
habit in the prediction of behavior suggest that a tradeoff between attitude and
habit exists (cf. Bagozzi, 1981; Triandis, 1977). When habit is weak, the
attitude-behavior link is stronger than when habit is strong. On the other hand,
among the strong habit subgroup, a small but significant correlation of .28
between attitude and behavior indicated that attitudes still guide these indi-
viduals’ behavior to some extent, even if the behavior was repetitively ex-
hibited. When the behavior is frequently repeated, it is, of course, quite
conceivable that the reasons that originally resulted into the onset of the
behavior still retain much of their validity. When respondents are asked to
report their attitudes, these reasons become salient once more. Ronis et al.
(1989) discussed other mechanisms that may contribute to an attitude-behavior
correlation when behavior is repetitive. For example, when behavior is repeat-
edly exhibited, attitudes may be inferred from self-perceptions of that behavior
(Bem, 1967). However, the present data suggest that, although some correla-
tion remains when habit is strong, attitudes reflect behavior to a much greater
extent when habit is weak, which makes a self-perception explanation less
likely.
As was discussed in the introduction section, various operationalizations of
habit are reported in the literature. The instrument we use to measure habit does
not rely on respondents’ abilities to remember past behavior or to remember
psychological processes that are supposed to reflect habit. Rather, the measure
is based on current schematic thinking, which is supposed to be guided by the
strength of the habit component (cf. Kahle, 1984). In this respect, our measure
has an advantage compared to previously used measures. A point of criticism
might be raised concerning whether the measure reflects an attitude, rather than
a habit. Although we do not have external validation data concerning our
measure, we think the measure reflects habit, and not, or only partially, atti-
tude, for a number of reasons. First, the measure component behaves in the
model as was hypothesized on theoretical grounds. Second, its correlation with
attitude is moderate and can be expected to be higher if the measure reflects car
choice attitude. Moreover, the occurrence of the (predicted) interaction of habit
and attitude in their prediction of behavior would be unlikely if the habit
 ATTITUDE VERSUS GENERAL HABIT 297

measure were another measure of attitude. Furthermore, some validating items

were present in the questionnaire-for example, self-reported amount of an-
nual mileage and the item that reflects self-reported habit. The habit measure
correlates substantially with both items ( r = .37,p < .001, and r = .55,p< .001,
respectively). Even more important, correlations of car choice attitude with
these two items are much lower (Y = .11, ns, and r = .22, p < .O 1, respectively).
In summary, we believe that our measure has the potential to serve as an
indication of general habit. Additional studies are necessary to validate this
measure. The measurement of habit will always be difficult, however, or as
Mittal (1988) expresses: “It is in the nature of the concept that an entirely
satisfactory measure of it may remain unavailable” (p. 1001). This is especially
true as long as verbal measures are used, calling for respondents to think about
the issue, which is in fact incompatible with the habit construct.
An important condition for habit to develop is that individuals are able to
repeat their behavior. To acquire the habit of choosing the car, a car should at
least be available and one should have a license (cf. Banister, 1978). Car
availability (as measured by the number of cars divided by the number of
driving licenses) appears to be modestly, but significantly, related to the
habit measure. In this study, a second antecedent of habit was investigated-
that is, decisional involvement. Decisional involvement refers to the salience
or relevance of the decision (cf. Zaichkowsky, 1985). In the context of con-
sumer purchase behavior Mittal and Lee (1989) distinguish between product
involvement and purchase involvement or brand-decision involvement.
Whereas the former type of involvement refers to a consumer’s interest in
a product class, the latter form concerns an individual’s interest in focusing
on making a decision. Our scale of decisional involvement parallels Mittal and
Lee’s (1989) concept of brand-decision involvement in the domain of travel
mode choice. The present data suggest that when individuals are motivated to
engage in the decision-making process, lower levels of habit are obtained,
whereas a low motivational level increases habit (cf. Ronis et al., 1989).
The present study certainly has its limitations. The use of questionnaires,
for example, limits the types of measures that can be used. By using trained
interviewers to collect the data, however, we attempted to obtain the highest
possible quality of data. A serious limitation is that the study comprises a cross
section at one point in time. The correlational data do not allow us, therefore,
to draw conclusions on the causality of the constructs measured. Other designs
(e.g., longitudinal and/or experimental) are necessary in order to investigate the
dynamic processes underlying the relationships among behavior, attitude,
habit, and decisional involvement.
The results may have implications for strategies aimed at changing travel
mode behavior. First, it seems useful to define target groups in terms of their
298 VERPLANKEN ET AL.

level of habit. The behavior of those with strong habit may be especially
difficult to change with traditional persuasive communications aimed at chang-
ing attitudes. As the present results suggest, in that case, attitudes are only
weakly related to behavior and, consequently, such attempts cannot be ex-
pected to succeed. The model tested in this study may lead to the consideration
of approaches that are primarily aimed at changing the habit component. This
might be attempted directly, for example, by employing technical measures
that reduce the attractiveness of car travel or by providing incentives for the
target behavior. Alternatively, increasing the level of decisional involvement
(cf. Geller et al., 1989) may constitute a promising strategy. Higher levels of
decisional involvement may result in consideration of alternative travel modes
and may thus decrease the strength of habit. Once habit strength is reduced,
persuasive communications aimed at changing attitudes may be more success-
ful in terms of behavioral changes. Especially if infrastructure has been im-
proved (e.g., new bus lines) but old habits have remained, strategies focusing
on increasing decisional involvement and decreasing habit may be promising
perspectives. The goal, then, is not to change notorious car drivers into rigid
users of public transport, but to increase the awareness that travel mode choice
may often involve some level of deliberation and decision-making. Such a
strategy may thus promote a more flexible travel choice behavior.

References

Bagozzi, R. P. (1981). Attitudes, intentions and behavior: A test of some

key hypotheses. Journal of Personality and Social Psychology, 41,
607-627.
Banister, D. (1978). The influence of habit formation on modal choice: A
heuristic model. Transportation, 7, 5-18.
Baron, R. M., & Kenney, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistic considerations. Journal of Personality and Social Psychology, 51,
1173-1182.
Beck, L., & Ajzen, I. (1991). Predicting dishonest actions using the theory of
planned behavior. Journal of Research in Personality, 25,285-301.
Bem, D. J. (1 967). Self-perception: An alternative interpretation of cognitive
dissonance phenomena. Psychological Review, 74, 183-200.
Bentler, P. M., & Speckart, G. (1979). Models of attitude-behavior relations.
Psychological Review, 86,452-464.
Dobson, R. de P., Dunbar, F., Smith, C . J., Reibstein, D. J., & Lovelock, C. H.
(1978). Structural models for analysis of travel attitude-behavior relation-
ship. Transportation, 7, 35 1-363,
 ATTITUDE VERSUS GENERAL HABIT 299

Fishbein, M., & Ajzen, I. (1975). Belief; attitude, intention, and behavior: An
introduction to theory and research. Reading: Addison-Wesley.
Fredricks, A. J., & Dossett, D. L. (1983). Attitude-behavior relations: A com-
parison of the Fishbein-Ajzen and the Bentler-Speckart models. Journal of
Personality and Social Psychology, 4 5 5 0 1-5 12.
Geller, E. S., Kalsher, M. J., Rudd, J. R., & Lehman, G . R. (1989). Promoting
safety belt use on a campus: An integration of commitment and incentive
strategies. Journal of Applied Social Psychology, 19,3-19.
Geller, E. S., Winett, R. A., & Everett, P. B. (1982). Preserving the environ-
ment: New strategies for behavior change. New York: Pergamon Press.
Gilbert, G., & Foerster, J. (1977). The importance of attitudes in the decision
to use mass transit. Transportation, 6, 321-332.
Hartgen, D. T. (1974). Attitudinal and situational variables influencing urban
mode choice: Some empirical findings. Transportation, 3, 377-392.
Houston, M. J., & Rothschild, M. L. (1978). Conceptual and methodological
perspectives on involvement. In S . Jain (Ed.), Research frontiers in mar-
keting: Dialogues and directions (pp. 184- 187). Chicago: American Mar-
keting Association.
Hull, C. L. (1 943). Principles of behavior: An introduction to behavior theory.
New York: Appleton-Century Crofts.
Kahle, L. R. (1984). Attitudes and social adaptation: A person-situation inter-
action approach. Oxford: Pergamon.
Kahle, L. R., Klingel, D. M., & Kulka, R. A. (1981). A longitudinal study of
adolescent attitude-behavior consistency. Public Opinion Quarterly, 45,
402-414.
Landis, D., Triandis, H. C., & Adamopoulos, J. (1 978). Habit and behavioral
intentions as predictors of social behavior. The Journal of Social Psychol-
ogy, 106,227-237.
Mittal, B. (1988). Achieving higher seat belt usage: The role of habit in
bridging the attitude-bahavior gap. Journal of Applied Social Psychology,
18, 993-1016.
Mittal, B. (1 989). Measuring purchase-decision involvement. Psychology and
Marketing, 6 , 147-162.
Mittal, B., & Lee, M. S. (1989). A causal model of consumer involvement.
Journal of Economic Psychology, 10,363-389.
Nisbett, R. E., & Wilson, T. (1977). Telling more than we can know: Verbal
reports on mental processes. Psychological Review, 84,23 1-259.
Recker, W. W., & Golob, T. F. (1976). An attitudinal modal choice model.
Transportation Research, 10,299-3 10.
Reibstein, D. J., Lovelock, C. H., & Dobson, R. de P. (1980). The direction of
causality between perceptions, affect, and behavior: An application to
300 VERPLANKEN ET AL.

travel behavior. Journal of Consumer Research, 6 , 370-376.

Ronis, D. L., Yates, J. F., & Kirscht, J. P. (1989). Attitudes, decisions, and
habits as determinants of repeated behavior. In A. R. Pratkanis, S. J.
Breckler, & A. G. Greenwald (Eds.), Attitude structure and function
(pp. 213-239). Hillsdale, NJ: Erlbaum.
Shifhn, R. M., & Schneider, W. (1977). Controlled and automatic human
information processing: 11. Perceptual learning, automatic attending, and
a general theory. Psychological Review, 84, 127-190.
Triandis, H. C. (1977). Interpersonal behavior. Monterey: Brooks/Cole Pub-
lishing Company.
Wittenbraker, J., Gibbs, B.L., & Kahle, L.R. (1983). Seat belt attitudes, habits,
and behaviors: An adaptive amendment to the Fishbein model. Journal of
Applied Social Psychology, 13,406-42 1.
Zaichkowsky, J.L. (1985). Measuring the involvement construct. Journal of
Consumer Research, 12,341-352.