ArticlePDF Available

Abstract

Holding a strong goal intention ("I intend to reach Z!") does not guarantee goal achievement, because people may fail to deal effectively with selfregulatory problems during goal striving. This review analyzes wether realization of goal intentions is facilitated by forming an implementation intention that spells out the when, where, and how of goal striving in advance ("If situation Y is encountered, then I will initiate goal-directed behavior X!"). Findings from 94 independent tests showed that implementation intentions had a positive effect of medium-to-large magnitude (d= .65) on goal attainment. Implementation intentions were effective in promoting the initiation of goal striving, the shielding of ongoing goal pursuit from unwanted influences, disengagement from failing courses of action, and conservation of capability for future goal striving. There was also strong support for postulatad component processes: Implementation intention formation both enhanced the accessibility of specified opportunities and automated respective goal-directed responses. Several directions for future research are outlined.
IMPLEMENTATION INTENTIONS
AND GOAL ACHIEVEMENT: A
METAANALYSIS OF EFFECTS
AND PROCESSES
Peter M. Gollwitzer
Paschal Sheeran
Holding a strong goal intention (‘‘I intend to reach Z!’’) does not guaran-
tee goal achievement, because people may fail to deal eVectively with self
regulatory problems during goal striving. This review analyzes whether
realization of goal intentions is facilitated by forming an implementation
intention that spells out the when, where, and how of goal striving in
advance (‘‘If situation Yis encountered, then I will initiate goaldirected
behavior X!’’). Findings from 94 independent tests showed that imple-
mentation intentions had a positive eVect of mediumtolarge magnitude
(d¼.65) on goal attainment. Implementation intentions were eVective in
promoting the initiation of goal striving, the shielding of ongoing goal
pursuit from unwanted influences, disengagement from failing courses of
action, and conservation of capability for future goal striving. There was
also strong support for postulated component processes: Implementation
intention formation both enhanced the accessibility of specified opportu-
nities and automated respective goaldirected responses. Several directions
for future research are outlined.
I. Introduction
Understanding what factors determine whether people succeed or fail in
achieving desired outcomes is a fundamental concern in both basic and
applied psychology. Most theories of motivation and selfregulation con-
verge on the idea that setting a behavioral or outcome goal is the key act of
willing that promotes goal attainment (e.g., Ajzen, 1991; Atkinson, 1957;
69
ADVANCES IN EXPERIMENTAL Copyright 2006, Elsevier Inc.
SOCIAL PSYCHOLOGY, VOL. 38 All rights reserved.
DOI: 10.1016/S0065-2601(06)38002-1 0065-2601/06 $35.00
Bandura, 1991; Carver & Scheier, 1998; Gollwitzer, 1990; Locke & Latham,
1990). The basic assumption is that the strength of a person’s intention deter-
mines respective accomplishments (Austin & Vancouver, 1996; Gollwitzer &
Moskowitz, 1996; Oettingen & Gollwitzer, 2001; Sheeran, 2002). Although
accumulated research supports this idea (e.g., Armitage & Conner, 2001;
Sheeran, 2002; Sutton, 1998),there is also contrary evidence that gives credence
to the proverb that ‘‘the road to hell is paved with good intentions’’ (Orbell &
Sheeran, 1998; Sheeran, 2002). To address this issue, Gollwitzer (1993, 1996,
1999) proposed that successful goal achievement is facilitated by a second act
of willing that furnishes the goal intention with an if–then plan specifying
when, where, and how the person will instigate responses that promote goal
realization. These plans are termed implementation intentions.
Implementation intentions appear to be eVective at enhancing the likeli-
hood of goal achievement. However, the eVectiveness of if–then planning has
been reviewed only in narrative (e.g., Gollwitzer, 1999; Gollwitzer, Bayer, &
McCulloch, 2005) and smallscale quantitative (e.g., Koestner, Lekes,
Powers, & Chicoine, 2002b; Sheeran, 2002) reports to date, and a comprehen-
sive evaluation of implementation intention eVects and processes is overdue.
The aim of this review is to quantify the overall impact of implementation
intention formation on goal achievement using metaanalytic techniques. In
addition, this chapter tests the eVectiveness of implementation intentions in
relation to diVerent selfregulatory problems and goal domains and assesses
potential moderators of implementation intention eVects. Finally, the impact
of implementation intentions on theoretically specified component processes is
examined to understand why implementation intentions may help people
obtain outcomes that they desire.
II. Goal Intention Strength and Goal Achievement
Goal intentions are selfinstructions to attain certain outcomes or perform
particular behaviors and typically take the format of ‘‘I intend to reach Z!’’
They are derived from beliefs about the feasibility and desirability of actions
and end states (e.g., Ajzen, 1991; Atkinson, 1957; Bandura, 1991, 1997;
Brehm & Self, 1989; Carver & Scheier, 1998; Heckhausen, 1991; Locke &
Latham, 1990; Vroom, 1964) and represent the culmination of the decision
making process (Gollwitzer, 1990). Goal intentions signal the end of delib-
eration about what actions to perform or outcomes to reach; they imply a
commitment to act that may vary in strength (Ajzen, 1991; Gollwitzer, 1990;
Sheeran, 2002; Webb & Sheeran, 2005a).
70 PETER M. GOLLWITZER AND PASCHAL SHEERAN
In traditi onal theories of goal pursuit, goal intent ions are co nstrued as the
most imm ediat e and impor tant predicto r of atta inment. For inst ance, pre-
emine nt accoun ts of goal directed beh avior, such as control theory ( Car ver &
Sc heier, 1982, 1998 ), social cogn itive theory (Bandu ra, 1991, 1997 ), and goal
sett ing theo ry ( Locke & Lat ham, 1990 ), models of attitude– behavior rela-
tions , such as the theori es of reasoned acti on ( Fishbei n, 1980; Fishbei n &
Ajz en, 1975 ) an d planned beh avior ( Ajz en, 199 1), and the model of interpe r-
son al be havior (Triand is, 1980 ), as wel l as theori es of health relat ed behav-
ior, such as protection motivation theory (Rogers, 1983) and the prototype/
willingness model (Gibbons, Gerrard, Blanton, & Russell, 1998), each accord
goal intentions a central role in their theorizing about action. Accordingly,
research has been concerned for several decades with the factors that deter-
mine strong intentions—the assumption being that intention strength is a
good predictor of intention realization.
This assumption seems to be supported by metaanalyses of correlational
studies in which participants’ goal intentions (e.g., ‘‘I intend to perform
behavior W!’’ or ‘‘I intend to achieve outcome Z!’’) are measured at one
timepoint and behavior is measured at a later timepoint. For example, re-
views of the theory of reasoned action (Kim & Hunter, 1993; Sheppard,
Hardwick, & Warshaw, 1988; van den Putte, 1993), the theory of planned
behavior (Armitage & Conner, 2001; Godin & Kok, 1996; Hausenblas, Carron,
& Mack, 1997), and protection motivation theory (Floyd, PrenticeDunn, &
Rogers, 2000; Milne, Sheeran, & Orbell, 2000), as well as metaanalyses
of particular behaviors (e.g., condom use, Sheeran & Orbell, 1998; physical
activity, Hagger, Chatzisarantis, & Biddle, 2002), indicate that strength of
intention typically explains 20–35% of the variance in goal achievement.
To gain insight into the overall strength of intention–behavior consistency in
this type of research, Sheeran (2002) conducted a metaanalysis of 10 meta
analyses of the intention–behavior relation. Findings showed that intentions
accounted for 28% of the variance in behavior, on average, across 422
studies involving 82,107 participants. According to Cohen’s (1992) power
primer, R
2
¼.28, constitutes a ‘‘large’’ eVect size, which suggests that intentions
are ‘‘good’’ predictors of behavior—as traditional theories of goal pursuit have
supposed.
However, bivariate correlations between goal intentions and future be-
havior may overestimate the strength of intention–behavior relations
because it is possible that future behavior and goal intentions are both
determined by selfperceptions of past behavior (Bem, 1972). The implica-
tion is that analyses should control for previous performance in order to
determine to what extent goal intentions are associated with behavior
change.Sutton and Sheeran (2003) conducted a metaanalysis along these
lines. Sampledweighted average correlations between past behavior, goal
METAANALYSIS OF IMPLEMENTATION INTENTIONS 71
intentions, and future behavior were computed from 51 studies involving
8166 participants and then used as inputs for a hierarchical regression
analysis. Findings indicated that, not surprisingly, past behavior was a good
predictor of future behavior on the first step of the equation and accounted
for 26% of the variance. Entering goal intentions on the second step was
associated with a significant increment in the variance explained in future
behavior (R2
change ¼:07). These findings suggest that goal intentions have
significant associations with future behavior even when previous perfor-
mance is taken into account. However, the eVect size for goal intentions is
smalltomedium rather than large.
Even correlational analyses that statistically control for past behavior in
estimating the goal intention–goal achievement relation are problematic,
however, because it is always possible that a third variable is responsible
for the observed associations. To eliminate this alternative explanation of
intention–behavior consistency, it is necessary to experimentally manipulate
goal intentions and then determine whether this manipulation produces a
significant diVerence in subsequent goal attainment. Webb and Sheeran
(in press a) tested this idea in a recent metaanalysis. They identified 47
studies (N¼8802) that (1) were successful at inducing statistically significant
diVerences in goal intentions between experimental versus control partici-
pants and (2) followed up participants in order to measure diVerences in
subsequent goal attainment. Findings showed that the mean diVerence
in goal intention strength produced by the experimental manipulations had
an eVect size of mediumtolarge magnitude (d¼.66). Findings also indi-
cated that manipulating goal intention strength engendered a significant
diVerence in goal achievement. However, the eVect size was smallto
medium only; dwas .36 that equates to R
2
¼.03. Thus, producing significant
changes in goal intention strength only generates a modest change in goal
achievement. This finding indicates that there is a substantial ‘‘gap’’ between
people’s goal intentions and their subsequent attainment.
A converging line of research has decomposed the intention–behavior
relation in terms of a 2 (goal intention: to act vs. not to act) 2 (goal
achievement: acted vs. did not act) matrix (McBroom & Reid, 1992; Orbell
& Sheeran, 1998; Sheeran, 2002). This decomposition provides insight into
the sources of consistency and discrepancy between intentions and action.
Consistency is attributable to participants who intend to act and subsequent-
ly act (termed ‘‘inclined actors’’) and to participants who do not intend to act
and do not act (‘‘disinclined abstainers’’). Discrepancies between intentions
and action, on the other hand, can be attributed to participants who intend
to act but do not act (‘‘inclined abstainers’’) and to participants who do not
intend to act but end up acting (‘‘disinclined actors’’). A review by Sheeran
(2002) found that inclined abstainers, rather than disinclined actors, were
72 PETER M. GOLLWITZER AND PASCHAL SHEERAN
princi pally responsi ble for the intention–beh avior ‘‘ga p.’’ The media n pro-
por tion of parti cipants who intende d to but did not act was 47%, wher eas
the media n proporti on of pa rticipant s who did not intend to act but subse-
que ntly acted was only 7%. Thes e findings would seem to confirm that the
pro verbia l ro ad to hell is paved with goo d intention s—bare ly more than
one half of pe ople who intende d to act were successful at trans lating those
intent ions into actio n.
In sum, it app ears that the single act of willing involv ed in form ing a goal
intent ion is not su Y cient to en sure go al achieve ment. The implici t assum p-
tion in tradi tional models of goal pur suit—that go al intentions fashio ned
from appropri ate evaluat ion of feasib ility and desir ability con siderati ons
sati sfactori ly acc ount for the intens ity of go al striv ing—is not strong ly
sup ported by the evidence. Cl early, so me add itional psych ologica l concepts
are ne eded (1) to underst and why peop le often be come inclined abstainers
rather than inclin ed actors and (2) to develop self regula tory strategi es to
he lp peo ple ‘‘br idge’’ the gap be tween their intent ions and their behavior .
III. Self Regulat ion of Goal Striv ing
Recent resear ch on goals has de monstrated that varia bles other than stre ngth
of goal intent ion a Vect the intensit y of goal striving and rate of g oal attain-
men t ( Gollwi tzer & Mosko witz, 1996; Oet tingen & Goll witzer, 2001 ). Some
goal theories focus on the implications of particular goal contents and struc-
tural features. For instance, people who set themselves learning goals rather
than performance goals are better at dealing with failure experiences and,
co nsequentl y, sh ow more persi stent an d success ful goa l pursu it (Dweck,
2000 ). Higgins (2000) demo nstrates that peop le who pursue their go als using
means that have a natural fit to the content of the goal have a better chance of
goal attainment. For example, people with promotion goals (that focus on
gain and achievement) are more likely to realize those goals using eagerness
means whereas prevention goals (that focus on safety and security) are more
likely to be realized by vigilance means. Other important distinctions between
types of goals have been drawn by Locke and Latham (1990) (e.g., specific vs.
‘‘do your best’’ goals), Bandura (1991) (e.g., proximal vs. distal goals), and
Deci and Ryan (1991) (e.g., goals based on needs for autonomy, competence,
and social integration vs. goals based on other needs). All of these theories
construe features related to the content and structure of set goals as critical in
determining the likelihood of goal achievement.
Other goal theories assume that setting a goal (of whatever kind) is only a
first step en route to goal realization. A key impetus for selfregulation
METAANALYSIS OF IMPLEMENTATION INTENTIONS 73
research on goals is the model of action phases (Gollwitzer, 1990; Heckhausen,
1991; Heckhausen & Gollwitzer, 1987) that construes goal attainment in terms
of solving a number of consecutive tasks. Goal setting is viewed as merely the
first of these tasks—with planning how to achieve the goal, getting started, and
successfully completing goal striving as equally important subsequent tasks.
The model of action phases seeks to provide a comprehensive temporal
account of goal pursuit. Four diVerent consecutive action phases are postu-
lated by the model. The first, predecisional, phase starts from the assumption
that people have many more wishes and desires than they can possibly
realize. Here people’s task is to deliberate about the desirability and feasibil-
ity of their various wishes in order to choose which ones will be turned
into binding goals. The model agrees with classic motivational notions
(e.g., Atkinson, 1957; Fishbein & Ajzen, 1975; Lewin, 1926) that people
commit to those goals in which attainment is perceived as both highly
desirable and feasible. However, the model of action phases also states that
goal attainment is not yet secured by the act of goal setting (i.e., by having
formed strong goal intentions). Rather, goal accomplishment requires in
addition that the individual eVectively regulates the actual striving for the
goal (i.e., engages in eVective goal implementation).
Once a person has committed to a goal, she makes the transition to the
second action phase, preactional. Here the goalrelevant task is to initiate
goaldirected behaviors successfully. This may be straightforward when
the respective actions have become routinized through frequent and consis-
tent performance in stable situational contexts. However, matters are likely
to be more complex when people are unfamiliar with, or imprecise about, the
respective goaldirected actions and contexts of performance. In these cir-
cumstances, people are likely to benefit from fashioning plans that spell out
when, where, and how to implement goaldirected behaviors.
The initiation of actual performance of the respective goaldirected beha-
viors marks the transition to the third, actional, phase. The task to be
accomplished during this phase pertains to responding flexibly and adap-
tively to contextual threats to goal progress so that goal striving is not
derailed prematurely. In other words, the key actional task is to bring
the respective goaldirected activity to a successful conclusion by shielding
it from distractions and temptations that could potentially disrupt goal
striving.
In the final action phase, postactional, the task is to evaluate goal achieve-
ment both in terms of degree of attainment (‘‘Did I do as well as I had
hoped?’’) and quality of attainment outcomes (‘‘Was it worth doing?’’). This
process involves comparing what has been achieved with one’s original
wishes and desires, and it may at times imply eVortful disengagement from
the goal (if further striving is inappropriate). Thus, goal completion is likely
74 PETER M. GOLLWITZER AND PASCHAL SHEERAN
to provide valuable information that can feed back into evaluations of the
feasibility and desirability of future courses of action; people return to the
position of deliberating about their various wishes and desires from which
they started.
A. PROBLEMS EN ROUTE TO GOAL COMPLETION
The foregoing discussion suggests that merely forming a goal intention does
not guarantee goal achievement as people often face problems en route to
goal completion. So what are these challenges, and how can people tackle
them successfully by using selfregulatory strategies? We propose that the
following four problems may prevent people from reaching their goals.
1. Failing to Get Started
The first problem that can undermine goal attainment is failing to get started
with goal striving. A number of factors militate against getting started on
one’s goals. The first has to do with remembering to act. When a behavior is
not part of one’s routine, or when one has to postpone acting until a suitable
opportunity presents itself, one can easily forget to perform the intended
behavior. This is because situational demands on attention and memorial
resources may serve to reduce the activation level of a focal goal intention
compared to other intentions (Einstein & McDaniel, 1996). Dealing with
many things at once or becoming preoccupied by a particular task can make
it diYcult to remember to act on one’s goals, especially when the intended
behavior is new or unfamiliar. Empirical support for this explanation of
intention–behavior discrepancies comes from retrospective reports by in-
clined abstainers. For example, 70% of participants who had intended to
perform a breast selfexamination but failed to do so, oVered ‘‘forgetting’’ as
their reason for nonperformance (Milne, Orbell, & Sheeran, 2002; Orbell,
Hodgkins, & Sheeran, 1997). Similarly, metaanalysis has shown that the
longer the time interval between measures of goal intentions and goal
achievement, the less likely it is that intentions are realized (Sheeran &
Orbell, 1998). These findings speak to the idea that remembering to act
can be a vital but diYcult task.
Even if one remembers what one intends to achieve, there is a second
problem that may need to be resolved, namely, seizing the opportunity to
act. This problem is especially acute when there is a deadline for performing
the behavior or when the opportunity to act is presented only briefly. In
these circumstances, people may fail to initiate goaldirected responses either
because they fail to notice that a good time to get started has arrived or they
METAANALYSIS OF IMPLEMENTATION INTENTIONS 75
are unsure how they should act when the moment presents itself. For
instance, Oettingen, Ho
¨nig, and Gollwitzer (2000, Study 3) showed that
considerable slippage can occur even when people have formed strong goal
intentions to perform a certain behavior at a particular time. In one of their
experimental conditions, participants were provided with diskettes contain-
ing four arithmetic tasks and formed goal intentions to perform these tasks
on their computers at a particular time each Wednesday morning for the
next 4 weeks. The program on the diskette recorded the time that partici-
pants started to work on the task from the clock on participants’ computers.
Findings indicated that the mean deviation from the intended start time was
8 hours, that is, a discrepancy of 2 hours on average for each specified
opportunity. Similar findings were obtained by Dholakia and Bagozzi
(2003, Study 2) using a ‘‘short fuse behavior’’ paradigm in which partici-
pants’ task was to evaluate a website that could be accessed only during a
short time window. Here, only 37% of participants who formed goal inten-
tions were successful at accomplishing the task (see also Gollwitzer &
Brandsta
¨tter, 1997). In sum, people may not get started with goal striving
because they fail to seize suitable opportunities to act.
Third, there are also many instances in which people remember their
‘‘good’’ intention (e.g., to order a low fat meal) and recognize that an
opportune moment is upon them (e.g., it is lunchtime at one’s usual restau-
rant) but, nonetheless, they fail to initiate action (e.g., because ‘‘I just didn’t
fancy the low fat meal!’’). This problem has to do with overcoming an initial
reluctance to act. Initial reluctance is likely to arise when people have
decided to initiate a behavior that involves a tradeoVbetween attractive
longterm consequences versus less attractive shortterm consequences. For
example, a strong goal intention to order the low fat meal might have been
formed on the basis of longerterm cognitive considerations (e.g., the low fat
meal is perceived as ‘‘healthy’’ or ‘‘beneficial’’); however, one might not
have anticipated how the shortterm aVective considerations would occupy
attention at the moment of action (e.g., the low fat meal is perceived as
‘‘unsatisfying’’ or ‘‘tasteless’’ at the critical juncture). Such dilemmas be-
tween the head and the heart are commonplace (e.g., Loewenstein, Weber,
Hsee, & Welch, 2001; Metcalfe & Mischel, 1999; Trafimow & Sheeran,
2004). Evidence for this explanation of intention–behavior discrepancies
comes, for instance, from the field of sexual health in which findings show
that young people may, in ‘‘the heat of the moment’’ of a sexual encounter,
have problems overcoming reluctance to practice safer sex (e.g., Abraham
et al., 1999; Sheeran, White, & Phillips, 1991; Wight, 1992). Overcoming
initial reluctance is also a significant problem in several other domains (e.g.,
environmental, consumer, and academic goals).
76 PETER M. GOLLWITZER AND PASCHAL SHEERAN
2. Getting Dera iled
The goals of inter est to social and healt h psychologi sts are not usuall y
discr ete one shot actio ns but sequences of actio n that requ ire continuous
striv ing a nd repeat ed beha vioral perfor mance to be acco mplished. The
pro blem with such striving is that many situati onal contexts or self states
are not conducive to intent ion realizat ion but inst ead hold the potential to
de rail an ongoing go al pursuit. Thus , the person’ s self regu latory task is to
shiel d goal stri ving from unwanted infl uences.
Shielding one’s goal striving is necessary under the following circumstances.
First, shielding is called for when conicting attention and behavioral responses
could make people stray oV course. For example, despite making good initial
progress with one’s goal intention to nish a report, one maynd one’s
attention wandering and feel compelled to join colleagues whom one hears
gathering around the water cooler. In these instances, spontaneous attention to
distracting stimuli may have to be suppressed in order to complete the goal.
Such suppression may not be easy when the distractions are vivid, arousing, or
highly valenced because, as the literature on cravings has shown (Kavanagh,
May, & Andrade, 2005), people are liable to elaborate desirable stimuli through
mental imagery. However, failing to control the attention paid to enticing
stimuli (opportunities related to competing goal pursuits) can greatly under-
mine achievement of the focal (task) goal—as was demonstrated by Mischel
and Pattersons (1978; Patterson & Mischel, 1976) c la ss ic s tu d ie s o n r es is ta nc e
to temptation (see also Gollwitzer & Schaal, 1998).
Of course, it may not be enough to suppress unwanted attention responses
to appealing distractions in order to reach one’s goal. Often, it will be
necessary to suppress behavioral responses. For example, the person who
succeeded in enacting her goal intention to order the low fat meal at lunchtime
still has to forego the chocolate dessert after dinner if the superordinate goal
intention is to lose weight. If an unwanted behavior possesses features of
automaticity, it should be especially diYcult to control (Aarts & Dijskterhuis,
2000a,b; Sheeran et al., 2005a; Verplanken & Aarts, 1999; Wood, Quinn, &
Kashy, 2002). Keeping such behavioral responses in check merely by forming
the respective goal intention may not be suYcient, as research on weight loss
and smoking cessation has shown (e.g., COMMIT Research Group, 1995;
Garner & Wooley, 1991). Findings from a recent metaanalysis (Ouellette &
Wood, 1998) are also consistent with this idea. Goal intentions emerged as
much poorer predictors of future action when antagonistic behaviors had
been performed frequently and consistently in relevant contexts (see also
Verplanken, Aarts, van Knippenberg, & Moonen, 1998). In sum, controlling
interfering unwanted attention and behavioral responses makes an important
METAANALYSIS OF IMPLEMENTATION INTENTIONS 77
diVerence to whether one’s goaldirected eVorts warrant the designation
‘‘inclined actor’’ versus ‘‘inclined abstainer.’’
There is a second circumstance in which shielding an ongoing goal pursuit
becomes crucial (Gollwitzer et al., 2005). So far, our discussion of
controlling unwanted attention and behavioral responses has assumed that
people have some knowledge and awareness of what sorts of obstacles
(distractions, temptations, barriers) the environmental context is likely to
present, when those obstacles are likely to arise, and what kind of unwanted
responses those obstacles typically generate. Knowing what might happen,
when it might happen, and how it might aVect us thus appears to be a
prerequisite for the successful use of any strategy of suppression. However,
there is a route to eVectively shielding an ongoing goal pursuit that does not
require the anticipation of a situational threat or its impact on goal striving.
The existence of such an alternative strategy is crucial as, more often than
not, we are not in a position to consciously anticipate the occurrence of
obstacles and the working of habits, or in what form and intensity these will
threaten ongoing goal pursuits. The following three social psychological
phenomena exemplify what we have in mind when we speak of obstacles to
an ongoing goal pursuit that are not anticipated by the individual: deindivi-
duation eVects on social loafing, the impact of loss frames on negotiation
outcomes, and nonconscious priming of antagonistic goals.
Social loafing eVects occur when people are asked to work in groups in
which performance outcomes cannot be checked at an individual level
(Karau & Williams, 1993; Latane
´, Williams, & Harkin, 1979). Under these
circumstances, people show reduced eVort and performance compared to
situations in which individual outcomes can be identified. The problem is
that people are unlikely to have insight into the negative impact of the group
setting and task instructions on their performance. Not surprisingly, there-
fore, having the goal intention to perform well is not suYcient to overcome
social loafing (Gollwitzer & Bayer, 2000).
A similar issue arises in relation to the impact of loss versus gain frames on
negotiation outcomes (De Dreu, Carnevale, Emans, & van de Vliert, 1995;
Neale & Bazerman, 1985). In a typical experiment, pairs of participants are
asked to negotiate the distribution of some finite resource (e.g., land on an
island). The framing of the negotiation is manipulated by either providing
participants with information about how many points they lose by giving
up elements of the resource (loss frame) or by telling participants how
many points they gain by receiving these elements (gain frame). Findings
indicate that cognitive loss frames lead to comparatively unfair agreements
about resources and also hinder integrative solutions. Participants are not
aware of the negative impact of loss frames on their negotiation behavior.
78 PETER M. GOLLWITZER AND PASCHAL SHEERAN
Again, merely forming a goal intention to engage in fair and cooperative
negotiation with one’s partner fails to be suYcient to overcome the negative
impact of loss framing (Tro
¨tschel & Gollwitzer, 2004).
Finally, people are unaware of the fact that their behavior is often guided
by goals that have become activated directly by the situational context at
hand. Automotive theory (Bargh, 1990; Bargh & Gollwitzer, 1994) pro-
poses that goals that have a history of being acted upon in a parti-
cular situation have the potential to become directly activated by this
critical situation without the need for conscious intent. Studies have used
priming techniques to show that activated chronic goals have predictable
eVects on the intensity of goal striving. For example, participants who
completed scrambled sentences designed to prime achievement goals per-
formed better on a word puzzle task as compared to controls (Bargh,
Gollwitzer, LeeChai, Barndollar, & Tro
¨tschel, 2001). Moreover, extensive
debriefing indicated that participants had no awareness of either the activa-
tion of the goal or its impact on respective performance. Direct goal activation
has serious implications for realizing one’s goal intentions when the situation-
al context activates a goal that is antagonistic to the focal goal. Consistent
with this idea, Gollwitzer, Sheeran, Tro
¨tschel, and Webb (2004d) found that
participants who had formed a goal intention to drive carefully in a driving
simulator exhibited greater speed and more errors when they had been primed
with the automotive of ‘‘moving fast’’ compared to when the primed auto
motive was to ‘‘move slow.’’ Clearly, therefore, blocking the adverse contex-
tual threat posed by situationally activated antagonistic goals constitutes an
important challenge in shielding an ongoing goal pursuit.
The discussion so far only refers to derailments of goal striving by unan-
ticipated unwanted influences that originate in the environment. But such
unanticipated unwanted influences can also originate within the person
(Gollwitzer et al., 2005). This is the third circumstance in which the shielding
of an ongoing goal pursuit is needed—when detrimental selfstates threaten
goal attainment. The negative consequences of the following three selfstates
on goal striving may serve as examples: the eVects of mood on stereotyping,
the influence of selfdefinitional incompleteness on social sensitivity, and the
impact of egodepletion on subsequent task performance.
Being in a good mood signals to the self that one’s current situation
is unproblematic, and thus information processing is less elaborate or
systematic as compared to being in a bad mood (Schwarz, 1990; Schwarz,
Bless, & Bohner, 1991). Consequently, people are more liable to stereo-
typing when they are in a good mood than when they are in a bad mood
(Bless, 1997; Bless & Fiedler, 1995). The impact of positive mood on stereo-
typing target persons is diYcult to anticipate by the layperson and, thus,
METAANALYSIS OF IMPLEMENTATION INTENTIONS 79
should be diYcult to control. Indeed, Gollwitzer and Bayer (2000) found
that merely having the goal intention to form nonstereotypical impressions
did not attenuate the goodmoodeVect on increased stereotyping.
Symbolic selfcompletion theory (Wicklund & Gollwitzer, 1982) proposes
that when people who are highly committed to an identity goal (e.g., becom-
ing a lawyer) obtain negative feedback about their accomplishments in the
respective domain, they experience a sense of selfdefinitional incomplete-
ness. This is a highly aversive selfevaluative state that is associated with
compensatory eVorts to show oValternative symbols or indicators of the
aspired to identity in front of other people (e.g., by wanting to talk about
one’s achievements). Consequently, incomplete individuals tend to become
absorbed in selfsymbolizing activities and thus neglect the thoughts and
feelings of an audience; their interactions with others exhibit social insensi-
tivity (Gollwitzer & Wicklund, 1985). Gollwitzer and Bayer (2000) observed
that this eVect cannot be ameliorated by explicitly assigning participants the
goal of taking the perspective of their interaction partners.
Finally, egodepletion refers to the phenomenon that exerting selfcontrol
on an initial task produces a temporary reduction in people’s capacity for
selfcontrol that is reflected in poor performance on a subsequent task
(Baumeister, Bratlavsky, & Muraven, 1998; Muraven & Baumeister, 2000).
For example, Baumeister et al. (1998, Experiment 1) showed that partici-
pants who had to eat radishes instead of tempting chocolate during an initial
task, persisted for less time on a subsequent unsolvable puzzles task than did
participants who were allowed to eat the chocolate during the initial task
(these participants did not have to exert selfcontrol). Apparently, ego
depletion can undermine task performance even when people have strong
goal intentions to perform well. Consistent with this idea, Webb and Sheeran
(2003) found that egodepleted participants and nondepleted controls exhib-
ited substantive diVerences in puzzle task performance. However, both
groups reported devoting equivalent eVort to the puzzle task and had
equivalent desire to quit. Thus, egodepletion would seem to be an important
factor in reducing the intensity of goal striving and one in which appropriate
goal intentions are not necessarily an eVective defense.
3. Not Calling a Halt
Initiating and shielding goal striving from unwanted influences are crucial
for successfully reaching goal completion. However, there is a third problem
that needs to be resolved, namely, disengaging from goal striving that
has become unproductive (Wrosch, Scheier, Carver, & Schulz, 2003). Dis-
engagement may be straightforward when goal monitoring indicates satis-
factory progress, or attainment of desired outcomes. However, a good deal
80 PETER M. GOLLWITZER AND PASCHAL SHEERAN
of research indicates that it is very diYcult to disengage from an ongoing
goal pursuit when selfdefensive concerns are activated. Researchers have
studied such failure to disengage under varying labels, such as sunk costs
(e.g., Arkes & Blumer, 1985), entrapment (e.g., Brockner, Rubin, & Lang,
1981), and escalation of commitment (e.g., Tan & Yates, 2002). However,
the basic conceptualization of the phenomenon is similar (Bragger, Hantula,
Bragger, Kirnan, & Kutcher, 2003). Again, mere goal intentions to halt a
failing course of action are often insuYcient as has been shown by work
using standard escalation paradigms (Henderson, Gollwitzer, & Oettingen,
2004).
4. Overextending Oneself
There is a fourth problem in goal striving that has to do with the fact that
people have to pursue multiple goals (e.g., Austin & Vancouver, 1996;
Carver & Scheier, 1998; Gollwitzer & Moskowitz, 1996). Thus, over-
extending oneself in an ongoing goal pursuit is likely to jeopardize the
achievement of subsequent important goals. Accordingly, eVective self
regulation of goal striving needs to conserve the person’s capability to
successfully engage in subsequent goal pursuit once striving for the initial
goal has ended. However, action control by goal intentions makes people
vulnerable to overextension.
A good example is the phenomenon of egodepletion in which assigning
participants goal intentions to perform well on an initial task that requires
selfcontrol is associated with reduced selfregulatory capability (and dimin-
ished performance) on a subsequent task (Baumeister et al., 1998). The
wellknown ironic eVects of mental control (Wegner, 1994) constitute anoth-
er instance in which goal intentions can produce overextension on an
initial task and thereby diminish future capability. For example, Macrae,
Bodenhausen, Milne, and Jetten (1994) assigned participants the goal inten-
tion of forming a nonstereotypical impression of a homeless person (or not)
and asked them to provide a written statement of their impression. After a
5minute filler task, participants were asked to evaluate homeless people in
general on semantic diVerential scales that included five stereotypical adjec-
tive pairs (e.g., drunk–sober, busy–lazy). Findings indicated that goal inten-
tions were successful in producing less stereotypical impressions of the
person on the initial task compared to controls. However, on the subsequent
rating task, goal intention participants gave more stereotypical evaluations
of homeless people in general. That is, goal intentions to suppress stereo-
types produced a rebound eVect. In sum, achieving desired outcomes on the
basis of mere goal intentions has costs in terms of undermining the success of
subsequent goal pursuits.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 81
B. FORM ING IMPLEMEN TAT ION INTENT IONS: A STR ATEGY
FOR EFFEC TIVE SELF REGU LATIO N OF GOA L STRIVING
The idea tested in the present meta an alysis is that impl ement ation inten-
tions (i.e., if–then plans) facilitate e Vective self regula tion of goal striving .
Impl ement ation intentions shou ld enhance pe ople’s abili ty to initiat e, main-
tain, disenga g e from , an d undertake further goal striving a nd thereby
increa se the likelihood that strong goal intentions are realized success fully.
In other words, this form of planning is expecte d to bridge the intent ion–
be havior gap .
Impl ement ation intent ions are if–then plans that connect goo d opportu-
niti es to act with cognitive or be haviora l responses that are e V ective in
acco mplishing one’s goals. W hereas go al intent ions specify what one wants
to ach ieve (i. e., ‘‘I intend to reach Z!’’), implement ation intention s sp ecify
both the behavior that one will perfor m in the service of goal achieve ment
an d the situatio nal context in whi ch one will enact it (i.e., ‘‘I f situ ation Y
oc curs, then I will initiat e goal directed behavior X !’’). Thus, go al intentions
an d implement ation intent ions can easily be dist inguis hed on the ba sis of
their content an d struc ture; a goal intention refers to what one intends to
ach ieve, wher eas an impl ementation intent ion specifies when , where , and how
one intends to achieve it.
To form an implement ation intention, the person mu st (1) identify a
respo nse that wi ll promot e goal attainme nt and (2) antic ipate a su itable
oc casion to initiat e that respon se. For inst ance, a pos sible implement ation
intent ion in the servi ce of the goal intention to do more exerci se would link
an ap propria te behavior (e.g., take the stairway instea d of the elevator) to a
suit able situatio nal co ntext (e.g., standi ng in front of the entranc e to the
elevat or a t work). As a con sequence, a strong men tal link is creat ed be tween
the critical sit uation of waiti ng for the elevato r and the goal directed
respo nse of walking upstai rs.
Select ing suita ble opportun ities to enact g oal direct ed respo nses entails
that people antic ipate situ ations in which it would be fitti ng to exec ute go al
direct ed responses . The critical sit uation specified in one’s plan can involv e
an intern al cue (e.g ., a strong feeling) or an exter nal cue (e.g ., a particular
place, object , person, or point in time). The cues can either be related to
good opportunities to act (i.e., it is easy to perform actions that are in-
strumental for reaching the goal) or to anticipated obstacles to goal striving.
Thus, cue selection can focus on initiating and stabilizing the goal striving at
hand or on shielding it from particular anticipated obstacles.
Forming an implementation intention also involves the selection of an
eVective goaldirected behavior. In line with the theory of goal systems
(Kr uglanksi et al., 2002; Sha h, Krugl anksi, & Friedm an, 2003 ), it is assum ed
82 PETER M. GOLLWITZER AND PASCHAL SHEERAN
that for any given goal, various routes to goal attainment are available.
Accordingly, the specification of the thencomponent of an implementation
intention can take many diVerent forms. For instance, not only can an
implementation intention specify one of the many behaviors that lead to
goal attainment, it can also specify the suppression of one of the many
responses that prevent goal attainment. In addition, the specification of the
goaldirected responses can either focus on the initiation or the maintenance
of goal striving. Finally, the thencomponent of an implementation intention
may specify ignoring those stimuli that have the potential to instigate un-
wanted attention or behavior responses that could derail an ongoing goal
pursuit (see Appendix 1 for sample implementation intentions).
C. COMPONENT PROCESSES OF
IMPLEMENTATION INTENTIONS
The mental links created by implementation intentions are expected to
facilitate goal attainment on the basis of psychological processes that relate
both to the anticipated situation (specified in the ifcomponent of the plan)
and the specified response (the plan’s thencomponent). As forming imple-
mentation intentions implies the selection of a critical future situation, it is
assumed that the mental representation of this situation becomes highly
activated (Gollwitzer, 1999). The person who forms an implementation
intention selects a situation that is ripe for action to achieve the goal; the
person is therefore perceptually ready to encounter this critical situation.
This idea implies that processing information about the critical situation is
highly proficient (Gollwitzer, 1993; Gollwitzer, Bayer, Steller, & Bargh,
2004a; Webb & Sheeran, 2004). That is, compared to those who merely form
a respective goal intention, people who form implementation intentions
should exhibit increased accessibility of the critical cue, and thus should be
better able to detect the cue and discriminate the cue from other similar
stimuli. For instance, Webb and Sheeran (2004) used a classic illusion
paradigm from the psychology of language to investigate cue detection by
goal intentions as compared to implementation intentions. Participants were
asked to form the goal intention to count the instances of the letter fin the
following piece of text: ‘‘Finished files are the/result of years of scientific/
study combined with the/experience of years.’’ (Line breaks are marked by
back slashes.) The illusion resides in the fact that most people count only
three fs because they miss the fin the three instances of the word ‘‘of.’’
However, when participants furnished the goal intention with a respective
implementation intention (i.e., ‘‘And as soon as I see the letter f, then I’ll
add one more to my count!’’) their detection of the diYculttoidentity fs
METAANALYSIS OF IMPLEMENTATION INTENTIONS 83
impr oved. Additional experiments showed that this impr oved detection of
crit ical cues did not have co sts in term s of false alarm s or reduced pe rfor-
man ce on identifyi ng nonc ritical stimuli , even when these stimu li wer e quite
sim ilar to the critical cu e (i. e., ambigu ous cues).
Increas ed acce ssibility of the specified situatio n should a lso faci litate
spo ntaneous atten tion to the cu e an d engender bette r recal l of the cue
(Gol lwitze r et al., 2004a) . Spont aneous attention was demo nstrated in a
dichotic listening experiment. When the critical cues specified in implemen-
tation intentions were presented in the nonattended channel, participants’
shadowing performance (i.e., repeating the words presented in the attended
ch annel) de clined. Goll witzer et al. (2004a ) also showed in a cu ed recall
experiment that the situations specified in if–then plans are better remem-
bered compared to alternative good opportunities to act. In sum, forming an
implementation intention should induce heightened sensitivity to the critical
situation at each stage of information processing such that people are better
able to detect, attend to, and remember specified cues when these cues are
encountered later.
Specifying that one will perform a particular goaldirected response in
the thencomponent of a plan, at the critical moment stipulated in the if
component of the plan, involves a strategic abdication of eVortful action
control. This is because forming an implementation intention delegates con-
trol of behavior from the self to specified situational cues that directly elicit
action (i.e., implementation intentions create ‘‘instant habits’’) (Gollwitzer,
1999). Forming an if–then plan means that the person commits herself in
advance to acting as soon as certain contextual constraints are satisfied. Once
that situation is encountered, action initiation should proceed swiftly and
eVortlessly and without requiring the person’s conscious intent. Accordingly,
the execution of a behavior specified in an implementation intention should
exhibit features of automaticity as identified by Bargh (1992, 1994).
Automaticity commonly characterizes highly overlearned activities
(e.g., driving a car, typing) including the operation of habits (Aarts &
Dijksterhuis, 2000a,b; Sheeran et al., 2005a; Wood et al., 2002). Action
control by implementation intentions seems to exhibit three features of
automatic processes: immediacy, eYciency, and lack of conscious intent.
Immediacy has been tested by means of response latencies (e.g., Webb &
Sheeran, 2004) and the temporal proximity of actual performance to
the time of performance specified in the implementation intention (e.g.,
Gollwitzer & Brandsta
¨tter, 1997; Oettingen et al., 2000, Experiment 3).
For instance, Gollwitzer and Brandsta
¨tter had research participants watch
a video presentation of a presumed Nazi who expressed racial slurs and
mark good opportunities to speak up. Participants all formed the goal
intention to counterargue at opportune moments when watching the video
84 PETER M. GOLLWITZER AND PASCHAL SHEERAN
a second time. A subset of participants also formed implementation inten-
tions by mentally linking these critical situations with respective counter-
arguments. Only having marked critical opportunities (mere goal intention
condition) was less eVective in promoting the immediate initiation of counter-
arguments compared to having also formed implementation intentions (i.e.,
if–then plan participants were much faster in using the marked opportunities).
The eYciency of implementation intention eVects is supported by studies
that varied cognitive load either through selection of the sample (e.g.,
schizophrenic patients, heroin addicts under withdrawal) or by experimental
manipulations using dual task paradigms (e.g., Brandsta
¨tter, Lengfelder, &
Gollwitzer, 2001; Lengfelder & Gollwitzer, 2000). For instance, Brandsta
¨tter
et al. (2001) assigned heroin addiction patients the task of writing a curricu-
lum vitae within a set time period. Forming implementation intentions that
specified exactly when and where to get started with this task helped not only
control participants (i.e., heroin users who were no longer experiencing
withdrawal symptoms) to meet this task but also those participants who still
showed withdrawal symptoms. Apparently, the eVect of implementation
intentions on task achievement did not interact with the cognitive load (drug
urge) experienced by the participants. Evidence for eYciency of action
control by implementation intentions was also observed in experiments in
which participants had to perform two tasks at the same time. The second-
ary task in these studies was always a Go/No Go task, whereas the primary
task was either a memorization task or a tracking task (Brandsta
¨tter et al.,
2001, Studies 3 and 4; Lengfelder & Gollwitzer, 2001, Study 2). The imple-
mentation intention was linked to performing the Go/No Go task and the
diYculty of the primary task was varied (easy vs. diYcult). The beneficial
eVects of implementation intentions on performance in the secondary
task were not qualified by an interaction with the primary task indicating
that the operation of implementation intentions is eYcient (i.e., independent
of cognitive load). Moreover, better performance was observed in the pri-
mary task during those phases of the secondary task that were guided by
implementation intentions (i.e., a transfer of freed resources).
Finally, there is evidence that the eVective operation of implementation
intentions does not require that people be consciously aware of either the
anticipated critical situation or the respective goal intention (e.g., Bayer,
Moskowitz, & Gollwitzer, 2004; Sheeran, Webb, & Gollwitzer, 2005c).
Bayer et al. (2004) demonstrated that conscious awareness of the specified
situation was redundant in two experiments that used subliminal priming of
respective cues. In one study, participants were asked to classify a series of
geometric figures (e.g., circles, ellipses, squares) as rounded or angular
objects by leftor rightbutton–press responses. All participants formed
the goal intention to classify the objects as fast and accurately as possible.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 85
Impl ement ation intent ion participant s wer e in add ition asked to make the
foll owing plan: ‘‘And if I see a triang le, then I will press the respective button
imm ediate ly!’’ Thi s implement ation intention led to fast er classificat ion
respo nses for trian gles. Impo rtantly, classificat ion performan ce on all angu-
lar figures was faci litated when these fig ures wer e preced ed by a subli minal
tri angle prime compared to a co ntrol prim e (the percen tage symbo l, %). No
such eVects were observed for goal intention participants.
Mo reover, She eran et al. (2005c ) sho wed that peop le need not be con-
sciously aware of the underlying goal intention for implementation intention
eVects to occur. All participants formed the conscious goal intention to solve
puzzles from the Wechsler Adult Intelligence Scale (WAIS III) as accurately
as possible. Half of the participants also formed an implementation inten-
tion in relation to another dimension of performance, namely, to solve the
puzzles as quickly as possible. This implementation intention manipulation
was then crossed with a priming procedure that activated the goal of
responding quickly outside of awareness. Speed and accuracy of responses
to the puzzles was then measured. Even though participants reported no
awareness of the primed goal during debriefing, findings indicated that
responses were fastest when participants were primed to respond quickly
and had formed respective implementation intentions. This study shows that
conscious intent is not required to observe implementation intention eVects
on performance.
In sum, the evidence on component processes suggests that people
can enhance rates of goal completion obtained by conscious and eVortful
guidance of behavior (action control by goal intentions) by strategically
switching to automated selfregulation of goal striving (action control by
implementation intentions).
IV. Present Review
Accumulated research indicates that there is a substantial gap between
people’s goal intentions and their goal achievement. This is because forming
a goal intention does not prepare people suYciently for dealing with self
regulatory problems in initiating, maintaining, disengaging from, or over-
extending oneself in goal striving. Forming an implementation intention,
on the other hand, spells out the when, where, and how of goal striving
in advance. If–then plans are therefore thought to enhance the accessibility
of the specified critical situation and induce automatic execution of the
specified response. The consequence is that people should remember to act,
seize good opportunities, overcome initial reluctance, suppress unwanted
86 PETER M. GOLLWITZER AND PASCHAL SHEERAN
responses, block detrimental selfstates and adverse contextual influences,
and successfully disengage from goals without costs to selfregulatory capa-
bility. Goal striving should be regulated eVectively, and goal achievement
should thereby be facilitated.
The present review tests these ideas using metaanalysis. First, we assess
the overall impact of implementation intention formation on goal achieve-
ment. We evaluate potential moderators of implementation intention eVects
and test whether implementation intentions are eVective in promoting per-
formance in diVerent domains of attainment. Second, we test the eVective-
ness of implementation intentions in overcoming selfregulatory problems
that have to do with initiating goal striving, shielding goals from unwanted
influences, disengaging from failing goals, and conserving selfregulatory
capability. Finally, we calibrate the eVect sizes for the component (if–then)
processes of implementation intentions.
A. METHOD
1. Sample of Studies
Several methods were used to generate the sample of studies: (1) computer-
ized searches were conducted on social scientific and medical databases
(PsychINFO, Social Science Citation Index and Conference Papers Index
[Web of Knowledge], Medline, Index Medicus, and Dissertation Abstracts
International Online) from January 1990 to December 2003 using the key-
words implementation intention(s) and plan(s), (2) references in each article
identified above were evaluated for inclusion, and (3) authors were contacted
and requests were made for unpublished studies and studies in press.
Studies were included in the review if (a) the implementation intention
formed by participants specified the performance of a goaldirected response
upon encountering an internal or external critical cue and (b) a statistical
association between the formation of an implementation intention and an
outcome variable could be retrieved (or obtained). Using these criteria, 94
tests of the relationship between implementation intentions and goal achieve-
ment could be included in the metaanalysis. The focal goal and eVect size for
each test are presented in Table I. The 94 independent tests come from 63
reports (these reports are preceded by an asterisk in the reference list).
2. MetaAnalytic Strategy
The eVect size estimate used here was d, which is the diVerence between the
means for two groups divided by a pooled standard deviation and corrected
for small sample bias (Hedges & Olkin, 1985). We subtracted the control
METAANALYSIS OF IMPLEMENTATION INTENTIONS 87
TABLE I
S TUDIES OF THE IMPACT OF IMPLEMENTATION INTENTION F ORMATION ON
GOAL ACHIEVEMENT
Author(s) Goal Nd
Aarts, Dijksterhuis, and Midden (1999) Collect a coupon/latencies to cues 40 .80
Ajzen, Czasch, and Flood (2002) Rate TV newscasts 102 .47
Armitage (2004) Eat a lowfat diet 126 .30
Bagozzi, Dholakia, and Basuroy (2003) Personal goals 153 .68
Bamberg (2000) Public transport use 90 .45
Bamberg (2002) Organic food purchase 160 .32
Bayer, Jaudas, and Gollwitzer (2002) Task switch 40 .85
Bayer, Moskowitz, and Gollwitzer (2004) Study 1 Retaliation behavior 61 .98
Bayer et al. (2004) Study 2 Classification of geometric shapes 61 .49
Brandsta¨ tter et al. (2003) Initiation of vocational retraining 126 .72
Brandsta¨ tter, Lengfelder, and Gollwitzer (2001) Study 1 Write a curriculum vitae 41 1.32
Brandsta¨ tter et al. (2001) Study 2 Go/No Go task 61 .93
Brandsta¨ tter et al. (2001) Study 3 Go/No Go task 68 .63
Brandsta¨ tter et al. (2001) Study 4 Go/No Go task 33 1.10
Bru¨ wer, Bayer, and Gollwitzer (2002) Simon eVect task 34 2.20
Bulgarella, Oettingen, and Gollwitzer (2003) Numerical judgment dilemma 34 .80
Chasteen, Park, and Schwarz (2001) Study 1 Prospective memory task 68 .82
Dewitte, Verguts, and Lens (2003) Study 1 Ten personal goals 15 .10
Dewitte et al. (2003) Study 2 Ten personal goals 14 .10
Dewitte et al. (2003) Study 3 Ten personal goals 16 .18
Dholakia and Bagozzi (2003) Study 1 Reading assignment 102 .56
Dholakia and Bagozzi (2003) Study 2 Visit website 138 .43
Dholakia and Bagozzi (2003) Study 3 Visit website 179 .82
Dholakia and Bagozzi (2002) Study 1 Product/service purchase 131 .49
Dholakia and Bagozzi (2002) Study 2 Personal goals 169 .41
DieVendorf and Lord (2003) Human resource task 170 .41
Einstein, McDaniel, Williford, Pagan, and Dismukes (2003) Prospective memory task 48 .24
Gillholm, Ettema, Sellart, and Garling (2004a) Study 1 Mundane activities 28 1.01
Gillholm et al. (1999) Study 2 Mail response forms 48 .02
Gollwitzer and Bayer (2004a) Study 1 Degree of perspective taking 34 1.16
Gollwitzer and Bayer (2004a) Study 2 Social loafing task 42 .75
Gollwitzer and Bayer (2004a) Study 3 Moodinduced gender
stereotyping
54 .80
Gollwitzer, Bayer, Steller, and Bargh (2002) Study 1 Cue detection 46 .82
Gollwitzer et al. (2002) Study 2 Dichotic listening task 55 .72
Gollwitzer et al. (2002) Study 3 Recall of specified cues 79 .87
Gollwitzer and Brandsta
¨tter (1997) Study 1 Personal goals 70 .43
Gollwitzer and Brandsta
¨tter (1997) Study 2 Complete a written report 36 .90
Gollwitzer and Brandsta
¨tter (1997) Study 3 Counter arguments to racist remarks 60 .52
Gollwitzer, Sheeran, and Seifert (2004c) Study 3 Solving law cases 55 1.10
Gollwitzer, Sheeran, Tro
¨tschel, and Webb (2004d) Study 1 Driving simulation dilemma 69 .98
Gollwitzer et al. (2004d) Study 2 Accessibility of drinking behavior 72 .49
Gollwitzer et al. (2004d) Study 3 Helping behavior dilemma 60 1.25
Gollwitzer et al. (2004d) Study 4 Cooperation in resource dilemma 60 .90
Gollwitzer, Tro
¨tschel, Bayer, and Sumner (2004e) Study 1 Stereotype rebound task 30 1.81
Gollwitzer et al. (2004e) Study 2 Anagram performance 31 1.12
Henderson, Gollwitzer, and Oettingen (2004) Study 1 Escalation of commitment task 87 .52
Henderson et al. (2004) Study 2 Escalation of commitment task 96 .54
Henderson et al. (2004) Study 3 Escalation of commitment task 187 .43
Holland, Aarts, and Langendam (in press) Recycling behaviors 54 1.42
Koestner, Downie, Horberg, and Hata (2002a) Personal goals 106 .39
Koestner, Lekes, Powers, and Chicoine (2002b) Study 1 Personal goals 106 .39
Koestner et al. (2002b) Study 2 New Year resolutions 38 .32
Koole and Van’t Spijjker (2000) Write a report 80 .75
Lengfelder and Gollwitzer (2001) Study 2 Go/No Go task 67 .80
Lippke and Ziegelmann (2002) Exercise 88 .18
Milne, Orbell, and Sheeran (2002) Exercise 248 1.25
Milne and Sheeran (2002a) Testicular selfexamination 432 .43
continues
Milne and Sheeran (2002b) Persistence with boring task 66 1.35
Milne and Sheeran (2002c) Visit study skills website 183 .87
MurgraV, White, and Phillips (1996) Binge drinking 102 .68
Oettingen, Ho¨ nig, and Gollwitzer (2000) Study 2 Compose a CV 20 1.25
Oettingen et al. (2000) Study 3 Complete a project 25 .80
Orbell, Hodgkins, and Sheeran (1997) Breast self examination 155 1.22
Orbell and Sheeran (1999) Study 3 Reduce snack food consumption 111 .61
Orbell and Sheeran (1999) Study 4 Reduce snack food consumption 93 .45
Orbell and Sheeran (2000) Recovery of functional activities 64 .70
Prestwich, Lawton, and Conner (2003a) Personal goals 79 .54
Prestwich, Lawton, and Conner (2003b) Exercise 58 .68
Rise, Thompson, and Verplanken (2003) Exercise and recycling 112 1.58
Schaal (1993) Arithmetic task 40 .93
Sheeran and Milne (2003) Reduce snack good consumption 129 .49
Sheeran and Orbell (1999) Study 1 Vitamin supplement use 78 .35
Sheeran and Orbell (1999) Study 2 Vitamin supplement use 37 .59
Sheeran and Orbell (2000) Cervical cancer screening 104 .58
Sheeran and Silverman (2003) Attendance at workplace safety training 271 .52
Sheeran and Webb (2003) Stroop performance 48 1.62
Sheeran, Webb, and Gollwitzer (2005c) Study 1 Independent study 85 .35
TABLE I Continued
Author(s) Goal Nd
Sheeran et al. (2005c) Study 2 Performance on puzzle task 40 .70
Sniehotta, Scholtz, and Schwarzer (2002a) Cardiac rehabilitation exercise/training 74 .61
Sniehotta, Scholtz, and Schwarzer (2002b) Physical activity 65 .70
Steadman and Quine (2000) Vitamin supplement use 174 .32
Steadman and Quine (in press) Testicular self examination 75 .54
Stephens and Conner (1999) Resist taking up smoking 124 .37
Tro¨ tschel and Gollwitzer (2004) Study 1 Framing e Vects in negotiation 43 .82
Tro¨ tschel and Gollwitzer (2004) Study 2 Framing e Vects in negotiation 57 .87
Verplanken and Faes (1999) Healthy eating 102 .47
Webb and Sheeran (2003) Study 1 Persistence with unsolvable puzzles 32 .87
Webb and Sheeran (2003) Study 2 Stroop performance 57 .87
Webb and Sheeran (2004) Study 1 Letter identification 54 .63
Webb and Sheeran (2004) Study 2 Number identication 42 .75
Webb and Sheeran (2004) Study 3 Number identication 53 .82
Webb and Sheeran (in press, b) Study 1 Personal goals 646 .75
Webb and Sheeran (in press, b) Study 2 Academic performance 129 .56
Williams (2003) Return postcard 60 .56
group mean from the mean for the implementation intention group so that a
positive dvalue indicates the benefit in performance conferred by forming an
implementation intention. The average eVect size was computed by averag-
ing the dvalues with each dweighted by the reciprocal of its variance. As a
test of significance, 95% confidence intervals were computed around each
mean. Where studies reported r,t,F,
2
, or contingency tables, we trans-
formed values into ds using the formulas supplied by Hedges and Olkin
(1985) and Hunter (1990). Homogeneity of eVect sizes was tested by means
of the Qstatistic that has an approximate chisquare distribution with k1
degrees of freedom, kbeing the number of eVect sizes. When Qis significant
(p<.05), eVect sizes are heterogeneous.
3. Multiple Measures and Multiple Tests
Several papers contained data from more than one sample or reported eVect
sizes for multiple measures of an independent variable or multiple measures of
the dependent variable. We tried to take advantage of the richness of these
data without violating the assumption of independence that underlies the
validity of metaanalysis. Data from independent samples were, therefore,
treated as separate units. In the case of multiple measures of independent or
dependent variables, the average dwithin each study was the unit of analysis.
Where studies contained multiple nonindependent samples, we used the
conservative strategy of computing the weighted average eVect size and using
the smallest Nin the analysis in order to determine the overall eVect size
for that study (Sheeran, Abraham, & Orbell, 1999). For example, Holland,
Aarts, and Langendam (in press) examined the impact of forming implemen-
tation intentions on objective measures of recycling old paper (N¼54, d¼
1.32) and recycling plastic cups (N¼109, d¼1.50). The eVect size used to
represent this study is the weighted average of the two eVects (d¼1.42), and
the sample size is 54.
B. RESULTS
1. Overall EVect Size
The overall impact of forming implementation intentions on goal achievement
was d¼.65 based on k¼94 tests that involved 8461 participants. This eVect had
a 95% confidence interval from .60 to .70. According to Cohen’s (1992) power
primer, d¼.20 is a ‘‘small’’ eVect, d¼.50 is a ‘‘medium’’sized eVect, whereas
d¼.80 is a ‘‘large’’ eVect. Thus, the eVect size that characterizes the impact of
if–then planning on goal achievement is of mediumtolarge magnitude.
92 PETER M. GOLLWITZER AND PASCHAL SHEERAN
a. Moderators of Implementation Intentions EVects. We examined
methodological moderators of the relationship between implementation
intentions and goal achievement in order to ensure that the eVect sizes for
if–then plans were not exaggerated by weaker methods (e.g., correlational
rather than experimental designs). The homogeneity test encouraged a
search for moderators as there was significant variability in the eVect
sizes obtained in individual studies, Q(93) ¼173.46, p<.001. Moderator
analyses were conducted for three methodological factors: type of sample,
study design (correlational vs. experimental), and measurement of goal
attainment (selfreport vs. objective).
Table II shows that most tests of implementation intention eVects were
conducted among university students (k¼79) though eight tests sampled
members of the public and there were two tests of children/young people.
Four tests were conducted with physically ill people, and there were three
tests among people with psychological problems (schizophrenic patients,
frontal lobe patients, and heroin addicts). Findings showed that, excluding
TABLE II
IMPACT OF METHODOLOGICAL FACTORS ON EFFECT SIZES FOR IMPLEMENTATION INTENTIONS
Factor Nk d 95% CI Q
Sample
General public 1076 8 .58 [.45, .70] 14.09*
Children/young adults 144 2 .47 [.14, .85] 2.38
People with physical illness 291 4 .52 [.28, .77] 3.66
People with psychological
problems
Schizophrenic patients 20 1 1.01
Braininjured patients 34 1 .87
Heroin addicts 41 1 1.32
University students 6855 79 .65 [.61, .70] 147.93***
Design
Correlational 1688 11 .70 [.61, .82] 20.23*
Experimental 6773 83 .65 [.61, .70] 151.59***
Measurement
Selfreport 4488 36 .63 [.58, .70] 80.96***
Objective 3973 58 .67 [.61, .74] 92.32**
Publication status
Unpublished 3759 46 .67 [.61, .72] 75.13***
Published 4702 48 .65 [.59, 70] 98.23***
*p<.05, **p<.01, ***p<.001.
Note: The sum of kequals 96 because data from two diVerent samples were disaggregated in
two studies (Lengfelder & Gollwitzer, 2001, Study 2; Brandsta
¨tter et al., 2001, Study 2).
N¼sample size; k¼number of independent eVects; d¼eVect size; CI ¼confidence interval;
Q¼homogeneity statistic.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 93
people with psychological problems, eVect sizes were of medium size and
equivalent magnitude across samples, Q(3) ¼4.01, p>.25. Implementation
intentions appeared to have stronger eVects for people with psychological
problems compared to the other groups; this diVerence proved significant
when subgroup comparisons were conducted (ds¼1.10 and .66, respective-
ly), Q(1) ¼4.54, p<.04. This finding suggests that forming implementation
intentions is especially beneficial to goal attainment among people who have
diYculties with regulating their behavior.
Findings indicated that implementation intentions were similarly eVective
whether the study design was correlational or experimental (ds¼.70 and
.65, respectively), Q(1) ¼1.93, p> .16. Moreover, the impact of implemen-
tation intentions on goal achievement was not exaggerated by overreliance
on selfreport measures of behavior. Implementation intentions had similar
eVects whether or not the outcome was measured objectively (d¼.67) or by
selfreport (d¼.63), Q(1) ¼2.18, p¼.14.
We also examined whether publication status was associated with the
strength of observed implementation intention eVects. Fortynine percent
of the eVects that could be included in the review were unpublished (k¼46),
and it is possible that unpublished tests may be of poorer methodological
quality than are published tests (Rosenthal, 1984). This could mean that the
overall estimate of eVect size is inflated. However, there was no diVerence in
eVect sizes from published versus unpublished tests (ds¼.65 and .67,
respectively), Q(1) ¼1.53, p¼.22.
b. EVect Sizes for DiVerent Goal Domains. To test the generality of
implementation intention eVects, values of dwere computed for diVerent
goal domains. We drew on the classification of domains used in Kim and
Hunter’s (1993) comprehensive metaanalysis of the impact of topic on
attitude–behavior relations in order to categorize the goals (Canary &
Seibold, 1984). EVects were available for seven of the domains identified
by Kim and Hunter (health, academic, consumer, environmental, prosocial,
antiracist, and laboratory tasks). An eighth category was personal goals
because several studies asked participants to nominate their own desired
outcomes that were then furnished with implementation intentions (or not).
Findings indicated that implementation intentions had medium or large
eVects for all domains (Table III). The goals examined most frequently related
to laboratory tasks (k¼38) and health (k¼23). There were large eVects for
antiracist, prosocial, and environmental behaviors (ds¼.87, 1.01, and 1.12,
respectively). There were mediumtolarge eVects for laboratory tasks and
academic achievement (ds¼.70 and .72, respectively) and mediumsized
eVects for consumer behaviors, health behaviors, and personal goals. Overall,
Table III indicates that implementation intentions have reliable eVects for a
wide range of goal domains.
94 PETER M. GOLLWITZER AND PASCHAL SHEERAN
2. EVect Sizes for DiVerent SelfRegulatory Problems
Table IV presents the eVect sizes for implementation intentions for self
regulatory problems associated with initiating goal striving, shielding goals
from unwanted influences, disengaging from failing goals, and conserving
selfregulatory capability. Three problems that militate against action initia-
tion are remembering to act, seizing opportunities, and overcoming initial
reluctance. There were k¼11, 20, and 21 tests of these problems, respective-
ly. For all three problems, eVect sizes for implementation intentions were
of mediumtolarge magnitude (ds¼.54, .61, and .65, respectively). These
findings indicate that implementation intentions help to ensure that people
(1) do not forget to perform intended actions, (2) do not miss good oppor-
tunities to initiate action, and (3) do not fail to act because they are swayed
by shortterm considerations. The overall eVect size was d¼.61 indicating
that implementation intention formation makes an important diVerence to
whether or not people initiate goal striving successfully.
We identified three selfregulatory tasks in relation to shielding goals from
unwanted influences, namely, suppressing unwanted responses, blocking
detrimental selfstates, and blocking adverse contextual influences. Imple-
mentation intentions proved beneficial for all three tasks. First, implemen-
tation intentions had a large eVect on suppressing unwanted attention
responses (d¼.90) and had a medium eVect on suppressing unwanted
behavioral responses (d¼.54). Second, implementation intentions had a
large eVect on goal achievement even when participants were in a detrimen-
tal selfstate (d¼1.10). Implementation intentions promoted performance
TABLE III
METAANALYSIS OF IMPLEMENTATION INTENTION EFFECTS FOR DIFFERENT GOAL DOMAINS
Goal domain Nkd 95% CI
2
Consumer 291 2 .41 [.16, .65] 0.50
Environmental 256 3 1.12 [.85, 1.42] 17.07***
Antiracist 144 3 .87 [.52, 1.25] 5.30
Prosocial 254 5 1.01 [.72, 1.28] 1.91
Academic 836 9 .72 [.56, .87] 7.42
Personal 1391 11 .58 [.47, .70] 14.11
Health 2861 23 .59 [.52, .67] 47.92***
Laboratory 2428 38 .70 [.61, .79] 59.90*
*p<.05, **p<.01, ***p<.001.
Note: N¼sample size; k¼number of independent eVects; d¼eVect size; CI ¼confidence
interval; Q¼homogeneity statistic.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 95
TABLE IV
METAANALYSIS OF IMPLEMENTATION INTENTION EFFECTS FOR DIFFERENT SELFREGULATORY
TASKS IN GOAL STRIVING
Selfregulatory tasks Nkd 95% CI Q
Initiating Goal Striving
Remembering to act 983 11 .54 [.45, .72] 16.79
Seizing opportunities 2270 20 .61 [.52, .70] 23.75
Overcoming initial reluctance 2588 21 .65 [.56, .72] 72.82***
Overall 5841 52 .61 [.56, .67] 114.21***
Shielding Goal Striving from Unwanted Influences
Suppressing unwanted responses
Attention responses 184 3 .90 [.61, 1.25] 5.75
Behavioral responses 559 5 .54 [.35, .70] 1.59
Blocking detrimental selfstates 248 5 1.10 [.80, 1.39] 4.40
Blocking adverse contextual influences 405 8 .93 [.70, 1.16] 2.22
Overall 1396 21 .77 [.67, .87] 27.60
Disengaging from Futile Goal Striving 370 3 .47 [.26, .70] 0.22
Conserving SelfRegulatory Capability 93 3 1.28 [.77, 1.76] 3.08
*p<.05, **p<.01, ***p<.001.
Note: N¼sample size; k¼number of independent eVects; d¼eVect size; CI ¼confidence
interval; Q¼homogeneity statistic.
when participants had incomplete selfdefinitions (d¼1.12, k¼2), were
egodepleted (d¼1.22, k¼2), or were in a good mood and therefore liable
to stereotyping (d¼.80, k¼1). Third, a large eVect was obtained for
implementation intentions when goal achievement was blocked by adverse
contextual influences (d¼.93). Forming an implementation intention
meant that participants were able to overcome the characteristic impacts
of deindividuation on social loafing (d¼.80, k¼1), loss frames on negotia-
tion (d¼.85, k¼2), and situational activation of goals that were antago-
nistic to the focal goal striving (d¼0.98, k¼5). In sum, implementation
intentions are eVective in blocking adverse contextual influences.
As well as initiating goal striving and shielding ongoing goal pursuits from
unwanted influences, people must also disengage from goal striving when such
striving is no longer productive. Three studies tested the eYcacy of imple-
mentation intentions in helping people disengage from failing goals. Findings
indicated that implementation intention eVects were of approximately medium
size (d¼.47).
The final selfregulatory problem is whether implementation intention
formation conserves people’s capability for future goal striving. Findings
showed that, even when the experimental settings involved two phases and
the initial task was known to engender performance deficits on the subse-
quent task, implementation intentions still had a large eVect on performance
96 PETER M. GOLLWITZER AND PASCHAL SHEERAN
(d¼1.28). Participants who formed implementation intentions to control
initial performance did not exhibit egodepletion or stereotype rebound. In
both cases, eVect sizes for implementation intentions were positive and large
(ds¼.87, and 1.81, respectively).
3. Component Processes of Implementation Intentions
Forming implementation intentions should activate the mental representa-
tion of the specified cues (ifcomponent) and automate responding to these
cues (as specified in the thencomponent). Table V shows that implemen-
tation intentions had large eVects on the detection, discrimination, and
accessibility of critical cues (ds¼.72, .82, and .95, respectively) and on the
attention paid to, and memory for, those cues (ds¼.72 and .87, respective-
ly). The overall eVect size for processes related to the ifcomponent of the
plan was large (d¼.80) indicating that implementation intentions are
associated with highly proficient processing of critical cues.
There were 7, 7, and 3 tests, respectively, of the immediacy and eYciency
of, and redundancy of conscious intent for action control by implementation
intentions. Implementation intentions showed large eVects for each of these
three key features of automaticity. If–then plans produced more immediate
responding (d¼.77), were eYcient with respect to cognitive resources (d¼
.85), and proceeded without the need for conscious intent (d¼.72). These
findings provide strong support for the postulated automaticity of action
control induced by implementation intentions.
TABLE V
METAANALYSIS OF COMPONENT PROCESSES OF IMPLEMENTATION INTENTIONS
Component process Nkd 95% CI Q
Ifcomponent
Cue detection 100 2 .72 [.30, 1.16] 0.23
Cue discrimination 53 1 .82
Cue accessibility 40 1 .95
Attention to cue 55 1 .72
Memory for cue 79 1 .87
Overall 327 6 .80 [.56, 1.04] 0.89
Thencomponent
Immediacy 363 7 .77 [.49, .98] 2.09
EYciency 278 7 .85 [.58, 1.12] 26.45
Lack of intent 122 3 .72 [.39, 1.07] 1.62
*p<.05, **p<.01, ***p<.001.
Note: N¼sample size; k¼number of independent eVects; d¼eVect size; CI ¼confidence
interval; Q¼homogeneity statistic.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 97
C. DISC USSIO N
Fin dings from 94 studi es involv ing mo re than 8000 participant s indica ted
that the e Vect size associ ated with the impac t of implemen tation intent ion
form ation on goal attainme nt is d ¼ .65, an e V ect of medium to large
magn itude ( Cohen, 1992 ). This e Vect size is impr essi ve because d ¼ .65
repres ents the di Verence in goal achieve ment engendered by furni shing a
go al intention with a respective implement ation intent ion compared to the
form ation of a go al intention on its own. The implicat ion is that if–t hen
plan ning sub stantial ly increa ses the likelihood of attaining one’s goals.
Several features of the meta analys is serve to underli ne the e V ectiven ess of
impl ement ation intent ions in promot ing go al achieve ment. First , it is unlike-
ly that the revie w su V ers from the ‘‘file drawer pro blem’’ (e.g., Rosenth al,
1984 ) as 49% of the included tests were unpublishe d. Moreover, publ ication
stat us had no impac t on the e Vect size obtaine d for impl ement ation inten-
tions . Second , 88% of tests involv ed experi menta l designs (i.e., rand om
assi gnment of participant s to implemen tation intention formati on) that
increa se confiden ce in the findings. It was also the case that the e Vect sizes
obta ined for experi menta l versus correl ational studi es wer e equ ivalent (un-
like meta analyses of the impac t of goal intent ions on goal achieve ment in
whi ch exp erimental tests sho w much weake r e Vects compared to correl ation-
al tests; Webb & She eran, in press a). Third, the compo sition of the sampl e
gen erally did not moderat e implemen tation intent ion e V ects. If–th en plans
wer e simila rly e V ective in promot ing goal achieve ment among studen ts,
member s of the general public, and people wi th physica l illness. Fourth,
when we used Kim an d Hun ter’s (1993) syste m to class ify dom ains of
atta inment , implement ation intent ions wer e sh own to have medium or large
e V ects for a wide varie ty of go als. Finally, the e Vec tiveness of implemen ta-
tion intentions was not exagger ated by ov erreliance on self report measur es
of behavior . The e Vect size for implement ation intent ions was of equ ivalent
magnitude in studies in which objective measures of performance were used.
In sum, implementation intentions seem to have benefited goal achievement
no matter how one looks at the data.
1. Implementation Intentions and SelfRegulatory Tasks in Goal Striving
Whereas traditional theories of goal pursuit assumed that strong goal inten-
tions are a suYcient determinant of goal achievement (e.g., Ajzen, 1991;
Atki nson, 1957 ; Fishbei n, 1980; Locke & Latham, 1990 ; Roger s, 1983), the
present research started from the position that there is a substantial gap
between intentions and action as there are numerous problems of goal striving
that need to be solved even if people hold strong binding goals (Gollwitzer,
98 PETER M. GOLLWITZER AND PASCHAL SHEERAN
1990 , 1993, 1996, 1999 ; Heckh ausen & Go llwitzer, 1987; Goll witzer et al.,
2005 ; Orbell & She eran, 1998; Sheeran, 2002 ). We analyze d the intent ion–
behavior gap in terms of four key selfregulatory tasks: initiating goal striving,
shielding ongoing goal pursuit from unwanted influences, disengaging from
unproductive goal striving, and conserving selfregulatory capability. Find-
ings indicated that if–then planning facilitated initiation of goal striving no
matter whether getting started was an issue of remembering to act, seizing
good opportunities, or overcoming initial reluctance.
Although fewer studies were conducted on the issue of shielding goal striving,
the beneficial eVects of implementation intentions were also strong. Forming
if–then plans helped with diVerent problems of maintaining an ongoing wanted
(focal) goal pursuit. The implementation intentions used were geared either at
suppressing unwanted attention and behavioral responses, or toward spelling
out the focal goal striving and thereby blocking detrimental selfstates and
adverse contextual influences. It is worth noting that various detrimental self
states and adverse contextual influences have been scrutinized and the aware-
ness of their presence varied between studies, as did the awareness of their
potential negative impact on the person’s goal striving.
Three studies investigated disengagement from failing courses of action
(Henderson et al., 2004) using standard escalation of commitment para-
digms (i.e., escalation of commitment was induced by instigating the jus-
tification motive; e.g., Bobocel & Meyer, 1994). Even though it is well
established that it is very diYcult to overcome strong selfjustification con-
cerns and thus halt escalation of commitment, implementation intention
formation was eVective in doing so and produced an eVect of approximately
medium size. Thus, implementation intentions provide a useful means for
successfully bringing futile goal striving to a close.
The final selfregulatory task in goal striving is conserving capability for
pursuing subsequent goals once an initial goal has been completed. Whereas
action control by goal intentions has been shown to generate ironic rebound
and egodepletion eVects for subsequent task performance, action control by
implementation intentions did not produce such costs. In other words,
controlling goal striving with implementation intentions allows people to move
on to subsequent goal striving without these selfregulatory handicaps; com-
pared to selfregulation by goal intentions, selfregulation by implementation
intentions conserves rather than diminishes capability for further goal striving.
2. Psychological Processes Underlying Implementation Intention EVects
Several studies explored the ifand thencomponent processes of imple-
mentation intentions. Findings strongly support the postulated mechanisms
(Gollwitzer, 1999). Apparently, specifying a situational cue in the ifcomponent
METAANALYSIS OF IMPLEMENTATION INTENTIONS 99
of an implemen tation intention creat es a height ened acti vation of the respec-
tive menta l repres entat ion of the situ ation. The implied ease of accessibil ity
co uld be observed in respect ive lexical de cision, pe rceptua l detect ion, cue
discr imina tion, an d memor y performanc es. M oreover, imple mentation inten-
tion pa rticipant s atten ded to specified critical cues even when the cues were
presen ted on the nona ttende d c hannel in a dicho tic listening task. Altho ugh
studi es used di Verent paradig ms to assess height ened acti vation, the e Vect
sizes obtaine d were uni formly large . These findings strong ly suggest that
ha ving selected a situati onal cue for acting tow ard one ’s go al, it is hard for
the person to overlook this opportunit y. This co ntrasts wi th the predica ment
of the person who has only formed a goal intention and thus needs to actively
search for and iden tify good oppor tunities to act ( Sheeran , Milne , Webb,
& Gollwi tzer, 2005b ).
At the same time, specif ying an e Vective go al direct ed response in the then
co mponent of the plan en dows the control of this response wi th features of
au tomatici ty. The three features of automa ticity that have been analyze d in
various di Veren t studi es are imm ediacy, e Y ciency, and lack of awareness
(i.e., co nscious intent is not requir ed). For all three feat ures, e Vect sizes were
large . Appare ntly, furni shing goal intentions wi th imple mentation intentions
switche s the mode of goal directed be havior from hesit ant to imm ediate,
from e Vortf ul to e Y cient, and from a conscious intent to act to direct
respo nse elicita tion by the situatio n. Whereas the person who ha s only
form ed a goal intent ion still has to deliberate in situ abo ut what g oal direct ed
respo nse to unde rtake an d/or energi ze the self to perfor m it, form ing an
impl ement ation intent ion means de ciding these issue s in advance, thereby
de legating the control of goal directed behavior to specified situati onal cue s.
Once these cues are encou ntered, acti on initiation is tri ggered au tomatica lly.
One might wonder whet her a chan ge in motivati onal fact ors (stren gth of
respect ive goal intentions an d/or self e Y cacy) due to if–then plan formati on
may ex plain impl ement ation intention e V ects on goal attainme nt—i n add i-
tion to, or even instead of, the pos tulated c omponent process es. However, at
least tw o lines of resear ch co ntradict this idea. First, studies that measur ed
stre ngth of go al intentions (comm itment) or self e Ycacy both be fore and
after respective implementation intention inductions found no evidence that
if–then plan formation increased scores on these variables in either within
participants analyses (diVerences within the if–then plan group over time) or
betweenparticipants analyses (diVerences between the if–then plan and
control group at either timepoint) (e.g., Brandsta
¨tter et al., 2001, Study 1;
M ilne e t a l., 2002 ; Oetting en et al., 2000, Study 2; Orbell et al., 199 7; Sheeran
& Orbell, 1 999 ; She eran et al., 2005c, Stud y 1). Second, implemen tation
intention formation enhanced rates of goal attainment even when partici-
pants had extremely high scores on goal intention and selfeYcacy prior to
100 PETER M. GOLLWITZER AND PASCHAL SHEERAN
plan form ation. For instan ce, Sheeran and Orbel l (2000) found that if–then
plan ning increa sed attend ance for cervic al cancer screening even though the
preint ervent ion means for the if–then plan group wer e 4.60 and 4.63, respec-
tive ly, on 1–5 scales. It is impla usible to attribut e the obs erved 33% impr ove-
men t in atten dance behavior among impl ement ation intent ion participant s
to postm anipula tion increa ses in goal intentions or self e Y cacy (for equ iva-
lent findi ngs see also Sheeran & Orbell, 1999; Ver planken & Fae s, 1 999).
Thes e results, toget her wi th findin gs showi ng that implement ation intent ion
e V ects do not ex hibit the tempor al decli ne of moti vation al inter venti ons
(e.g ., Sheeran & Silverman, 2003; She eran et al., 2005b ) and actual ly show
strong er e Vects for di Y cult to impl ement as compared to easy to implement
go als (e.g., Gollwitzer & Br andsta¨ tter, 1 997, Study 1), all indica te that
increases in goal intention strength and selfeYcacy as a consequence of
if–then plan formation cannot explain implementation intention eVects on
goal achievement.
3. Implementation Intentions in Everyday Life
Two findings help to clarify when implementation intention formation is
likely to especially benefit goal attainment. First, if–then planning has a
significantly larger eVect size among people who are known to have pro-
blems with action control (e.g., frontal lobe patients, schizophrenics). Sec-
ond, implementation intentions exhibit a noticeably large eVect size in tasks
that are known to overextend people’s capability to regulate their behavior
(i.e., in egodepletion and ironic rebound paradigms; d¼1.28). These
findings speak to the idea that the presence of problems in goal striving is
an important determinant of the strength of implementation intention
eVects. If the set goal is extremely easy to initiate and pursue, then simply
forming the respective goal intention could satisfactorily facilitate goal
achievement; in such instances, it is possible that forming an implementation
intention may confer little additional benefit. If, on the other hand, person
characteristics or task features make it diYcult to execute goaldirected
behaviors, then it is especially advantageous to engage in if–then planning.
That is, forming implementation intentions is most likely to benefit goal
achievement when regulating the behavior is diYcult or people have chronic
diYculties in regulating their behavior.
In the light of this analysis, and the overall support obtained in this review
for the beneficial impact of implementation intentions on goal achievement,
can we conclude that if–then planning will facilitate such attainment under
any circumstances? In other words, is forming implementation intentions a
foolproof selfregulatory strategy of goal striving? In everyday life, people
may fail to form eVective implementation intentions due to unfortunate
METAANALYSIS OF IMPLEMENTATION INTENTIONS 101
specifications of opportunities and goaldirected responses. For instance, a
person may identify an opportunity that hardly ever arises (e.g., when one
rarely has the choice between walking vs. taking the elevator), or an oppor-
tunity in which it turns out to be impossible to act toward one’s goal (e.g.,
one’s boss insists that you ride the elevator together to discuss work).
Similarly, a person may specify a behavior that has limited instrumentality
with respect to reaching the goal (e.g., taking the stairs instead of the
elevator is unlikely, on its own, to achieve the superordinate goal of reducing
weight) or a behavior that, in reality, proves impossible for the person to
perform (e.g., walk up 60 flights of stairs to one’s oYce).
In addition, if–then plans may not be very eVective because opportunities
and responses are not specified precisely. For example, a plan that specifies
‘‘eat healthily’’ in the thencomponent and ‘‘tomorrow’’ in the ifcomponent
has hardly spelled out an unambiguous opportunity to act or a specific goal
directed response to initiate—the person still has to identify a particular
behavior to perform in a particular situation to facilitate goal achievement
(e.g., order a salad at lunch time tomorrow in my usual restaurant). Having
to thus deliberate about when, where, and what to do in situ means that the
person is unlikely to garner much benefit from the enhanced activation of
critical cues or automation of responding conferred by forming precise
if–then plans; the person seems no better oVthan having merely formed
the goal intention to ‘‘eat healthily tomorrow.’’ In sum, implementation
intention formation should prove useful in promoting goal achievement
provided components of the plan are precise (i.e., deliberation about appro-
priate opportunities and responses is not required in situ), viable (i.e., the
specified situation will be encountered, the specified response can be per-
formed), and instrumental (i.e., the specified situation permits action, and
the specified response facilitates goal achievement). How often the if–then
plans fashioned in people’s everyday lives satisfy these conditions is an
empirical issue.
Finally, it is important that people who specify obstacles in the ifcompo-
nent of their implementation intentions select those barriers and distractions
that most hinder goal completion. In other words, it matters that people
specify those obstacles that do indeed undermine goal striving. Research has
demonstrated that the mental exercise of juxtaposing the desired future with
the present negative reality (i.e., mental contrasting) is a particularly eVective
strategy for discovering powerful barriers and hindrances that stand in the
way of realizing desired outcomes (Oettingen, 2000; Oettingen et al., 2001).
Accordingly, inviting people to engage in mental constrasting prior to
if–then planning should ensure that people gear their implementation in-
tentions to precisely those obstacles that present the greatest obstruction to
goal attainment.
102 PETER M. GOLLWITZER AND PASCHAL SHEERAN
4. Do Implem entat ion Inten tions Engender Rigid Goal Striving?
Ass uming that impl ement ation intent ions create strong links be tween antic i-
pa ted situati ons an d goal directed behavior s, does this mean that implemen-
tatio n intent ion form ation unde rmin es perfor mance when flexibl e goal
striv ing is call ed for? The idea that impl ement ation intention s could engen-
de r costs in terms of rigidit y has at least three aspect s. First , action con trol
by implemen tation intent ions cou ld be rigi d in the sense that goal striving no
long er takes into accoun t the state (act ivation, stre ngth) of parti cipant s’ goal
intent ions. Research does not support this concern, however. Several studies
have shown that goal intentions moderate the impact of implementation inten-
tions on goal attainment such that strong eVects of ifthen plans only emerge
when participants hold strong respective goal intentions (e.g., Koestner et al.,
2002b; Orbell et al., 1997; Sheeran et al., 2005c, Study 1). Similarly, studies that
either activated (Bayer et al., 2004; Sheeran et al., 2005c, Study 2) or deactivated
(Seehausen, Bayer, & Gollwitzer, 1994, cited in Gollwitzer, 1996)relevantgoal
intentions indicate that implementation intentions only aVected performance
when the respective goal intention was activated. For example, Sheeran et al.
(2005c) showed that an ifthen plan to enhance speed of responding on a puzzle
task only aVected response times when the goal intention to respond quickly
had been primed in the situation. Thus, action control by implementation
intentions does not involve a mechanistic elicitation of action in the presence
of environmental cues but rather respects the presence versus absence of
activated strong goal intentions. Apparently, the automaticity instigated by
ifthen plans is goald epe nd ent ( Bargh, 1992, 1994)concerns that ifthen
plans could engender rigid adherence to a course of action that does not serve
a person’s goals seem unfounded.
The second aspect of rigi dity concerns the possibili ty that implemen tation
intent ions could facilitate one aspect of goal striv ing but do so at the expense
of other aspect s of goal striving . That is, forming an if–then plan to promot e
one dimension of perfor mance could consu me self regu latory resourc es
an d thereby eng ender inflexib le perfor mance on other dimens ions; as a con-
sequ ence, ov erall goal attainme nt might be comprom ised. Again, evidence
seems to co ntradict this idea. For exampl e, a lthough Sheeran et al. (2005c )
found that implementation intentions enhanced response times on a puzzle
task, accuracy of responses was not compromised. Similarly, Gollwitzer
and Bayer (2000) showed that implementation intentions not only increased
the number of solutions generated in a creativity task but also enhanced
the conceptual variety of those solutions. These findings indicate that action
control by implementation intentions does not induce rigidity in terms of
inevitable tradeoVs between dimensions of performance (speed vs. accuracy,
quantity vs. quality). Rather, the automation of one aspect of goal striving
METAANALYSIS OF IMPLEMENTATION INTENTIONS 103
seems to free up cognitive capacity such that other aspects of the focal
striving are not compromised, and even can be enhanced (Brandsta
¨tter
et al., 2001).
The third aspect of potential rigidity concerns whether implementation
intention participants refrain from using alternative good opportunities to
act toward the goal by insisting on acting only when the critical situation
specified in the ifpart of the implementation intention is encountered. Several
features of if–then plans suggest that such rigid adherence to specified oppor-
tunities is unlikely. Because implementation intentions respect the activation
and strength of participants’ superordinate goal intentions, participants who
have formed if–then plans should still be sensitive to the issue of identifying
good opportunities to act. Moreover, because action control by implementa-
tion intentions is eYcient and conserves selfregulatory capability, if–then
planners should be in a good position to eVectively process information about
alternative opportunities, and to seize those opportunities judged suitable for
execution of behavior. In sum, implementation intentions do not seem to
engender rigid selfregulation in terms of mechanistic situational control,
performance tradeoVs, or neglecting suitable alternative opportunities to
move toward the goal.
Finally, there may be a further fourth issue related to rigidity, this one
having to do with how people deal with having acted on a faulty if–then
plan. We do not know yet what happens when people recognize that they
have formed an if–then plan that failed to lead to goal attainment (or even
produced negative outcomes). Do people stubbornly adhere to the faulty
if–then plan, or readily modify the ifand thencomponents of that plan, or
do they even completely refrain from forming if–then plans? Also one
wonders how the explicitness of the failure feedback and the strength of
the respective goal intention aVect whether people will adhere to or modify
the plan, or stay away from planning altogether.
5. Future Research on If–Then Plans
Although 94 independent tests of implementation intention eVects on goal
achievement were examined in this chapter, further research is warranted to
exploit the benefits of implementation intentions in facilitating goal attain-
ment and to enhance understanding of this mode of action control. Findings
from 52 and 21 studies, respectively, showed that implementation inten-
tions facilitated initiation of goal striving and eVectively shielded ongoing
goal pursuits from unwanted influences. However, there were fewer studies
that addressed selfregulatory problems in disengaging from futile goal
striving and conserving capability for future goal striving. Even considering
the 21 tests to do with the problem of getting derailed, additional studies
104 PETER M. GOLLWITZER AND PASCHAL SHEERAN
would help to corroborate the eYcacy of if–then plans in dealing with the
various aspects of the respective selfregulatory tasks (i.e., suppressing un-
wanted responses, blocking detrimental selfstates, and blocking adverse
contextual influences).
The same reasoning applies to research in diVerent goal domains and
using diVerent samples. Most studies to date used laboratory tasks, and
there have been relatively few applications to consumer, environmental,
antiracist, and prosocial behaviors. Similarly, the 23 tests in relation to
health goals predominantly concerned the initiation of healthprotective
behaviors (e.g., exercise, cancer screening). However, healthrisk behaviors,
such as smoking, excess alcohol consumption, and poor diet, are major
contributors to mortality and morbidity in Western societies (Belloc, 1973;
Breslow & Enstrom, 1980). How well implementation intentions can help
people to assiduously avoid these actions constitutes an important avenue
for future investigation. Previous studies also mainly used undergraduate
samples, and although sample type did not generally moderate implementa-
tion intention eVects, further tests among more representative groups would
enhance the generality of the present analysis. The finding that people with
chronic problems in action control (e.g., schizophrenics) were especially
likely to benefit from implementation intention formation is encouraging
and provides grounds for further rigorous tests of if–then planning inter-
ventions among other clinical samples (e.g., ADHD children, depressed
individuals). More generally, although the present metaanalysis shows that
implementation intentions are eVective in enabling people to translate their
‘‘good’’ intentions into action, the review also reveals considerable scope
for further tests in relation to longstanding selfregulatory problems (e.g.,
control of pain or stress), underresearched samples (e.g., people with physical
illness), and new domains of application (e.g., educational, organizational,
and clinical settings). In whatever context people’s goal intentions are found
to fall short of their goal achievement, applied psychologists might do well
to consider deploying if–then plans to promote eVective selfregulation of
goal striving.
There is also room for further theoretical integration of the concept of
implementation intentions with theories of motivation (e.g., Bandura, 1997)
and willpower (e.g., Metcalfe & Mischel, 1999). For instance, with respect to
motivation, future studies may want to explore whether implementation
intentions can be used to elevate selfeYcacy beliefs (e.g., ‘‘And if I run into
problems with any of my homework, then I will tell myself ‘I can do it!’ ’’).
With respect to willpower, implementation intentions can be used to turn oV
the hot system and activate the cool system when selfcontrol is needed. For
example, a person who wants to cope better with unpleasant social encoun-
ters could use implementation intentions to reduce feelings of frustration
METAANALYSIS OF IMPLEMENTATION INTENTIONS 105
and anger (e.g., ‘‘And if I run into an obnoxious person, then I will try to
understand this person as if I was a therapist!’’). What distinguishes this
approach from past research on implementation intentions is the fact that
the thencomponent of the if–then plan does not specify one particular goal
directed response, but rather focuses on changing motivationrelevant beliefs
and/or selfregulatory systems that can ultimately facilitate the performance
of multiple and various goaldirected responses.
The present review obtained strong support for the component processes
postulated to underlie implementation intention eVects. Implementa-
tion intentions showed large eVects on processes to do with heightened
activation of the critical situation (accessibility, detection, discrimination,
attention, memory) and automation of the goaldirected response (immedia-
cy, eYciency, redundancy of intent). However, it would be valuable to con-
duct mediation analyses to explore whether these processes are indeed
responsible for the positive eVects of implementation intention formation
on rates of goal achievement. One study that conducted this type of analysis
measured the accessibility of situational cues specified in participants’ if–then
plans in a lexical decision task and subsequently measured rates of goal
attainment (Aarts et al., 1999). Findings indicated that cue accessibility
mediated the impact of implementation intention formation on goal comp-
letion. Recently, Webb and Sheeran (2005c) extended this paradigm to in-
vestigate the mediational role of both cue accessibility and the strength of
cue–response links forged by if–then planning. In one experiment, partici-
pants had the goal intention to collect a coupon from a specified location as
part of a series of laboratory tasks. A subset of participants also formed
an implementation intention that specified the location for collecting the
coupon in the ifcomponent and the action of coupon collection in the then
component. Subsequently, an ostensibly unrelated lexical decision task had to
be performed that assessed the accessibility of the critical cues (location
words) and the accessibility of the target behavior when subliminally primed
by the critical cues (i.e., the word ‘‘collect’’ preceded by location words).
Findings indicated that implementation intention formation increased the rate
of coupon collection (goal achievement) as well as the accessibility of both
location cues and locationprimed target behavior (i.e., the strength of the link
between the ifand thencomponents of the plan). Most important, implemen-
tation intention eVects on goal attainmentwere mediated by cue accessibility as
well as the strength of respective cue–response links. These findings are
consistent with the postulated theoretical mechanisms. Further tests are needed
to explore the mediational role of the other hypothesized processes (e.g.,
immediacy, eYciency, and redundancy of conscious intent), however.
Moderators of implementation intention eVects also warrant investigation.
There are two aspects to moderation here. First, individual diVerences could
106 PETER M. GOLLWITZER AND PASCHAL SHEERAN
either enhance or reduce the impact of implementation intentions on goal
achievement. Individuals with personal attributes that make regulating their
behavior more diYcult, for instance, might especially benefit from implemen-
tation intention formation. Thus, people who score highly on measures of
procrastination, distractability, or selfdefensiveness may show higher rates
of goal attainment when they form if–then plans compared to people who
obtain low scores on these measures. On the other hand, individuals who
spontaneously form implementation intentions may garner less advantage
from inductions designed to prompt plan formation. Individual diVerences in
conscientiousness, planfulness, or need for cognition could predict spontane-
ous if–then planning. It is also possible that individual diVerence variables
could be identified that render if–then planning counterproductive. For
instance, people who are poor at reality monitoring could form plans that
are antithetical to eVective goal striving. Similarly, people who set too much
store by adherence to plans (e.g., perfectionist individuals) may be prone to
selfevaluative ruminations that undermine the eVective operation of their
plans. Thus, standard individual diVerence variables could have an impor-
tant influence on whether and how well implementation intentions are
formed and how much of an eVect they have on goal achievement.
If one conceives of personality in terms of ‘‘intraindividually stable,
if...then..., situation–behavior relations’’ (Mischel & Shoda, 1995, p. 248),
the question of how personality and if–then planning work together in the
selfregulation of goal striving may get even more interesting. Let us assume
that a person has the goal to reduce aggression in relating to others, and
he also knows about his respective situation–behavior profile (i.e., he knows
what kind of social situations elicit aggressive responses in him and which
social situations allow him to stay calm and collected). Given this goal and
knowledge, the person can now tailor his implementation intentions to
those critical situations specifying any of the following goaldirected
responses: ‘‘...then I will not get aggressive!’’ or ‘‘...then I will stay calm
and collected!’’ or ‘‘...then I will ignore this situation!’’ Thus, it seems
possible that people could maximize the selfregulatory benefits of forming
implementation intentions by taking into account their unique chronic
if–then (situation–behavior) profiles and specify implementation intentions
exactly where they are needed. Exploring interactions between chronic
and strategic situation–behavior links constitutes a promising direction for
future studies.
The second aspect of moderation concerns degree of plan formation and
refers both to the activation level of the ifand thencomponents of the plan
and to the strength of the mental link between the ifcomponent and the
thencomponent of the plan. These features of implementation intentions are
responsible for the enhanced identification of specified contextual cues and
METAANALYSIS OF IMPLEMENTATION INTENTIONS 107
au tomated acti on con trol in the presence of these cues, thus de termining
how wel l if–th en plans facilitate goa l atta inment. The implicat ion of varia-
tions in de gree of plan form ation is that procedu res that en hance the
acti vation level of critical cue s or the strength of cue–res ponse associati ons
sho uld thereby increa se the impac t of implemen tation intentions on goal
striv ing and go al completion. To date, only a small number of studies have
tested this aspect of moderation. For instance, Gollwitzer et al. (2004a)
manipulated the strength of participants commitment to their implementation
intentions presuming to thereby strengthen cueresponse links. Findings from
a cued recall paradigm showed that the high commitment group had superior
memory for selected opportunities compared to the low commitment group.
Similarly, Milne and Sheeran (2002c) manipulated cognitive rehearsal by hav-
ing some participants concentrate on the cue–response link during plan forma-
tion; participants wrote down their plan to visit a particular website twice, and
were instructed to concentrate on the link between the situation and action
when they were writing the plan the second time. Another implementation
intention group also wrote down their plan twice, but were instructed to take
the second writeup of the plan with them and put it in a prominent place at
home as a reminder. Findings indicated that participants who rehearsed the
cue–response link were more likely to act on their plans compared to both
participants who wrote their implementation intention on a reminder note and
a control group who did not form implementation intentions (rates of visiting
the Web site were 87%, 40%, and 20%, respectively).
Future studies should examine the eVectiveness of strategies to aid encod-
ing of if–then plans (e.g., diVerent types of cognitive rehearsal, surprise recall
tasks or plan reminders) and strategies to increase commitment to these
plans (e.g., inducing anticipated regret about not following one’s plan or
making one’s commitment public) in order to ensure that opportunities are
highly accessible and opportunity–action links are strong. Some individuals
are likely to be more in need of such strategies than others because people
diVer in their ability to generate strong if–then links when asked to form
implementation intentions (Gollwitzer, Grant, & Oettingen, 2004b). But
even if people’s original if–then links are weak, it seems possible to strength-
en these links by having people act repeatedly on their implementation
intentions. Research by Orbell and Verplanken (2005) observed that when-
ever participants performed repeated actions (e.g., flossing one’s teeth) on
the basis of an implementation intention, they reported experiencing features
of habitual action control (e.g., I do it without thinking, I start doing it
before I realize I’m doing it, I do it automatically, I do it without having to
think consciously, It would require eVort not to do it, ...) more so than
participants who performed the repeated action on the basis of a mere goal
intention. Further research along these lines would be valuable in order to
108 PETER M. GOLLWITZER AND PASCHAL SHEERAN
ensure that implementation intentions are as eVective as possible in facilitat-
ing the realization of goal intentions for particular people.
V. Conclusions
Goal intentions are not always successfully translated into behavior because
merely making a commitment to attain a goal does not necessarily prepare
people for dealing eVectively with selfregulatory problems in goal striving.
This chapter tested the idea that goal striving could benefit from a second act
of willing—the formation of if–then plans—that focuses on the enactment of
goal intentions. A metaanalysis of 94 studies showed that forming an
implementation intention makes an important diVerence to whether or not
people achieve their goals. This finding was robust across variations in study
design, outcome measurement, and domains of goal attainment. Moreover,
if–then planning facilitated goal striving no matter what selfregulatory
problem was at hand. Mediumtolarge eVects were obtained in relation to
initiating goal striving, shielding goals from unwanted influences, disen-
gaging from failing goals, and preserving selfregulatory capability for future
goal striving. There was also strong support for the if–then component
processes. People who form implementation intentions are in a good posi-
tion to recognize opportunities to act and respond to these opportunities
swiftly and eVortlessly. Thus, this chapter shows that the concept of imple-
mentation intentions is valuable both in understanding the processes of goal
attainment and in providing a selfregulatory strategy to help people reach
their goals. Notwithstanding the selfregulatory benefits of implementation
intentions demonstrated here, there is considerable scope for further re-
search to exploit the potential of if–then planning and to understand how
implementation intentions can best be deployed to facilitate intention reali-
zation. Such research would seem to be a worthwhile goal pursuit for both
basic and applied psychologists.
Acknowledgments
We thank Icek Ajzen, Christopher J. Armitage, Paul Dholakia, Rob Holland, Richard
Koestner, Sonia Lippke, Andrew Prestwich, Falko Sniehotta, Liz Steadman, and Jochen Philipp
Ziegelman for access to unpublished findings, and Tom Webb, Ute Bayer, Anja Achtziger, Inge
Schweiger Gallo, Alex Jaudas, Georg Odenthal, Caterina Gawrilow, and Tanya Faude for
comments on an earlier draft of the manuscript.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 109
Preparation of this chapter was supported in part by National Institute of Health grant
(R01–67100) and the Center for Research on Intentions and Intentionality at the University of
Konstanz, Germany.
Appendix I
EXAMPLES OF POSSIBLE IMPLEMENTATION INTENTIONS
GEARED AT RESOLVING THE FOUR PROBLEMS OF
GOAL STRIVING
1. Failing to get started
a. Remembering to act
To achieve the goal intention of sending a birthday card on time:
And if I walk by the institute’s mail box, then I will drop in my card!
b. Seizing opportunities
To achieve the goal intention of complaining about poor service:
And if I see the manager walk into the restaurant, then I will go over
to him and complain about the poor service!
c. Overcoming initial reluctance
To achieve the goal intention of completing course work on time:
And if it is Saturday morning at 10 a.m., then I will sit down at my
computer and make an outline for my essay!
2. Getting derailed
a. Suppressing unwanted attention responses
To achieve the goal intention of behaving calmly in the face of scary
spider pictures: And if I see a spider, then I will ignore it!
b. Suppressing unwanted behavioral responses
To achieve the goal intention of behaving calmly in the wake of
being insulted: And if I feel my anger rise, then I will tell myself to
stay calm and not aggress back!
c. Blocking detrimental selfstates
To block the negative influence of egodepletion on solving diYcult
anagrams: And if I have solved one anagram, then I will immediately
move onto the next one!
d. Blocking adverse contextual influences
To block the negative influence of loss framing on negotiation out-
comes when having to share an attractive commodity (e.g., a fictitious
island in the Lake of Constance): And if I receive a proposal on how to
share the island, then I will oVer a cooperative counterproposal!
110 PETER M. GOLLWITZER AND PASCHAL SHEERAN
3. Not calling a halt
To prevent escalation of commitment to a certain strategy of
performing a general knowledge test: And if I receive disappointing
feedback, then I will switch to a diVerent strategy!
4. Overextending oneself
To prevent the emergence of egodepletion in the wake of controlling one’s
emotions, such as not laughing at amusing cartoons: Andifanamusing
scene is presented, then I will tell myself ‘‘these are just stupid, silly jokes!’’
References
Papers included in the metaanalysis are preceded by an asterisk.
Aarts, H., & Dijksterhuis, A. (2000a). Habits as knowledge structures: Automaticity in goal
directed behavior. Journal of Personality and Social Psychology, 78, 53–63.
Aarts,, H, & Dijksterhuis, A. (2000b). On the automatic activation of goaldirected behavior:
The case of travel habit. Journal of Environmental Psychology, 20, 75–82.
*Aarts, H., Dijksterhuis, A., & Midden, C. (1999). To plan or not to plan? Goal achievement or
interrupting the performance of mundane behaviors. European Journal of Social Psycholo-
gy, 29, 971–979.
Abraham, C., Sheeran, P., Norman, P., Conner, M., Otten, W., & de Vries, N. (1999). When
good intentions are not enough: Modelling postintention cognitive correlates of condom
use. Journal of Applied Social Psychology, 29, 2591–2612.
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision
Processes, 50, 179–211.
*Ajzen, I., Czasch, C., & Flood, M. G. (2002). From intentions to behavior: Implementation
intention, commitment, and conscientiousness. Manuscript under review.
Arkes, H. R., & Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and
Human Performance, 35, 129–140.
*Armitage, C. J. (2004). Implementation intentions and eating a lowfat diet: A randomized
controlled trial. Health Psychology, 23, 319–323.
Armitage, C. J., & Conner, M. (2001). EYcacy of the theory of planned behaviour: A meta
analytic review. British Journal of Social Psychology, 40, 471–499.
Atkinson, J. W. (1957). Motivational determinants of risk taking behavior. Psychological
Review, 64, 359–372.
Austin, J. T., & Vancouver, J. B. (1996). Goal constructs in psychology: Structure, process, and
content. Psycological Bulletin, 120, 338–375.
*Bagozzi, R. P., Dholakia, U. M., & Basuroy, S. (2003). How eVortful decisions get enacted:
The motivating role of decision processes, desires, and anticipated emotions. Journal of
Behavioral Decision Making, 16, 273–295.
*Bamberg, S. (2000). The promotion of a new behavior by forming an implementation inten-
tion: Results of a field experiment in the domain of travel mode choice. Journal of Applied
Social Psychology, 23, 573–587.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 111
*Bamberg, S. (2002). Implementation intention versus monetary incentive comparing the eVects
of interventions to promote the purchase of organically produced food. Journal of Eco-
nomic Psychology, 23, 573–587.
Bandura, A. (1991). Selfregulation of motivation through anticipatory and selfreactive
mechanisms. In R. A. Dienstbier (Ed.), Nebraska symposium on motivation: Perspectives
on motivation (Vol. 38, pp. 69–164). Lincoln, NE: University of Nebraska Press.
Bandura, A. (1997). SelfeYcacy: The exercise of control. New York: W. H. Freeman.
Bargh, J. A. (1990). Automotives: Preconscious determinants of social interaction. In E. T.
Higgins and R. M. Sorrentino (Eds.), Handbook of motivation and cognition: Foundations of
social behavior (Vol. 2, pp. 93–130). New York: Guilford Press.
Bargh, J. A. (1992). The ecology of automaticity: Towards establishing the conditions needed to
produce automatic processing eVects. American Journal of Psychology, 105, 181–199.
Bargh, J. A. (1994). The four horsemen of automaticity: Awareness, eYciency, intention, and
control in social interaction. In R. S. Wyer, Jr. and T. K. Srull (Eds.), Handbook of social
cognition (2nd ed., pp. 1–40). Hillsdale, NJ: Erlbaum.
Bargh, J. A., & Gollwitzer, P. M. (1994). Environmental control over goaldirected action.
Nebraska Symposium on Motivation, 41, 71–124.
Bargh, J. A., Gollwitzer, P. M., LeeChai, A., Barndollar, K., & Tro
¨tschel, R. (2001). The
automated will: Nonconscious activation and pursuit of behavioral goals. Journal of
Personality and Social Psychology, 81, 1014–1027.
Baumeister, R. F., Bratlavsky, E., Muraven, M., & Tice, D. M. (1998). Egodepletion: Is the
active self a limited resource? Journal of Personality and Social Psychology, 74, 1252–1265.
*Bayer, U. C., Jaudas, A., & Gollwitzer, P. M. (2002, July). Do implementation intentions
facilitate switching between tasks?. Poster presented at the International Symposium on
Executive Functions. Konstanz, Germany.
*Bayer, U. C., Moskowitz, G. B., & Gollwitzer, P. M. (2004). Implementation intentions and
action initiation without conscious intent. Manuscript under review.
Belloc, N. B. (1973). Relationship of physical health status and health status. Preventive
Medicine, 2, 67–81.
Bem, D. J. (1972). Selfperception theory. Advances in Experimental Social Psychology, 6, 1–62.
Bless, H. (1997). Stimmung und Denken: Ein Modell zum Einfluvon Stimmung auf Denkprozesse
(Mood and reasoning: A model on the impact of mood on cognitive processes). Bern: Hans
Huber.
Bless, H., & Fiedler, K. (1995). AVective states and the influence of activated general knowledge.
Personality and Social Psychology Bulletin, 21, 766–778.
Bobocel, D. R., & Meyer, J. P. (1994). Escalating commitment to a failing course of action:
Separating the roles of choice and justification. Journal of Applied Psychology, 79, 360–363.
Bragger, J. D., Hantula, D. A., Bragger, D., Kirnan, J., & Kutcher, E. (2003). When success
breeds failure: History, hysteresis, and delayed exit decisions. Journal of Applied Psychology,
88, 6–14.
*Brandsta
¨tter, V., Heimbeck, D., Malzacher, J. T., & Frese, M. (2003). Goals need implemen-
tation intentions: The model of action phases tested in the applied setting of continuing
education. European Journal of Work and Organizational Psychology, 12, 37–59.
*Brandsta
¨tter, V., Lengfelder, A., & Gollwitzer, P. M. (2001). Implementation intentions and
eYcient action initiation. Journal of Personality and Social Psychology, 81, 946–960.
Brehm, J. W., & Self, E. (1989). The intensity of motivation. Annual Review of Psychology, 40,
109–131.
Breslow, L., & Enstrom, J. E. (1980). Persistence of health habits and their relationship to
mortality. Preventive Medicine, 9, 469–483.
112 PETER M. GOLLWITZER AND PASCHAL SHEERAN
Brockner, J., Rubin, J. Z., & Lang, E. (1981). Facesaving and entrapment. Journal of Experi-
mental Social Psychology, 17, 68–79.
*Bru
¨wer, J., Bayer, U. C., & Gollwitzer, P. M. (2002). Intentional control of the Simon eVect.
Poster presented at the International Symposium on Executive Functions. Konstanz,
Germany.
*Bulgarella, C., Oettingen, G., & Gollwitzer, P. M. (2003, February). Reflexive competition and
its selfregulation. Poster presented at the meeting of the Society for Personality and Social
Psychology. Los Angeles, CA.
Canary, D. J., & Seibold, D. R. (1984). Attitudes and behavior: A comprehensive bibliography.
New York: Praeger.
Carver, C. S., & Scheier, M. F. (1982). Control theory: A useful conceptual framework for
personalitysocial, clinical, and health psychology. Psychological Bulletin, 92, 111–135.
Carver, C. S., & Scheier, M. F. (1998). On the selfregulation of behavior. Cambridge, UK:
University Press.
*Chasteen, A. L., Park, D. C., & Schwarz, N. (2001). Implementation intentions and facilitation
of prospective memory. Psychological Science, 12, 457–461.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.
COMMIT Research Group (1995). Community intervention trial for smoking cessation (COM-
MIT): I. Cohort results from a fouryear community intervention. American Journal of
Public Health, 85, 183–192.
De Dreu, C. K. W., Carnevale, P. J. D., Emans, B. J. M., & van de Vliert, E. (1995). Outcome
frames in bilateral negotiation: Resistance to concession making and frame adoption. In W.
Stroebe and M. Hewstone (Eds.), European review of social psychology (Vol. 6, pp. 97–125).
Chichester, UK: Wiley.
Deci, E. L., & Ryan, R. M. (1991). A motivational approach to self: Integration in personality.
Nebraska Symposium on Motivation, 38, 237–288.
*Dewitte, S., Verguts, T., & Lens, W. (2003). Implementation intentions do not enhance all
types of goals: The moderating role of goal diYculty. Current Psychology, 22, 73–89.
*Dholakia, U. M., & Bagozzi, R. P. (2002). Mustering motivation to enact decisions: How
decision process characteristics influence goal realization. Journal of Behavioral Decision
Making, 15, 167–188.
*Dholakia, U. M., & Bagozzi, R. P. (2003). As time goes by: How goal and implementation
intentions influence enactment of shortfuse behaviors. Journal of Applied Social Psycholo-
gy, 33, 889–922.
*DieVendorV, J. M., & Lord, R. G. (2003). The volitional and strategic eVects of planning on
task performance and goal commitment. Human Performance, 16, 365–387.
Dweck, C. S. (2000). Selftheories: Their role in motivation, personality, and development.
Philadelphia, PA: Psychology Press.
Einstein, G. O., & McDaniel, M. A. (1996). Retrieval processes in prospective memory:
Theoretical approaches and some new empirical findings. In M. Bradimonte, G. O.
Einstein, and M. A. McDaniel (Eds.), Prospective memory: Theory and applications
(pp. 115–141). Mahwah, NJ: Erlbaum.
*Einstein, G. O., McDaniel, M. A., Williford, C. L., Pagan, J. L., & Dismukes, R. K. (2003).
Forgetting of intentions in demanding situations is rapid. Journal of Experimental Psychol-
ogy: Applied, 9, 147–162.
Fishbein, M. (1980). Theory of reasoned action: Some applications and implications. In H.
Howe and M. Page (Eds.), Nebraska symposium on motivation, 1979 (pp. 65–116). Lincoln,
NB: University of Nebraska Press.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to
theory and research. Reading, MA: AddisonWesley.
METAANALYSIS OF IMPLEMENTATION INTENTIONS 113
Floyd, D. L., PrenticeDunn, S., & Rogers, R. W. (2000). A metaanalysis of research on
protection motivation theory. Journal of Applied Social Psychology, 30, 407–429.
Garner, D. M., & Wooley, S. C. (1991). Confronting the failure of behavioral and dietary
treatments of obesity. Clinical Psychology Review, 6, 58–137.
Gibbons, F. X., Gerrard, M., Blanton, H., & Russell, D. W. (1998). Reasoned action and social
reaction: Willingness and intention as independent predictors of health risk. Journal of
Personality and Social Psychology, 74, 1164–1180.
*Gillholm, R., Ettema, D., Selart, M., & Garling, T. (1999). The role of planning for intention
behavior consistency. Scandinavian Journal of Psychology, 40, 241–250.
Godin, G., & Kok, G. (1996). The theory of planned behavior: A review of its applications in
health–related behaviors. American Journal of Health Promotion, 11, 87–98.
Gollwitzer, P. M. (1990). Action phases and mindsets. In E. T. Higgins and J. R. M. Sorrentino
(Eds.), The handbook of motivation and cognition (Vol. 2, pp. 53–92). New York: Guilford.
Gollwitzer, P. M. (1993). Goal achievement: The role of intentions. In W. Stroebe and M.
Hewstone (Eds.), European Review of Social Psychology (Vol. 4, pp. 141–185). New York:
Wiley.
Gollwitzer, P. M. (1996). The volitional benefits of planning. In P. M. Gollwitzer and J. A.
Bargh (Eds.), The psychology of action: Linking cognition and motivation to behavior
(pp. 287–312). New York: Guilford.
Gollwitzer, P. M. (1999). Implementation intentions: Strong eVects of simple plans. American
Psychologist, 54, 493–503.
*Gollwitzer, P. M., & Bayer, U. C. (2000, October). Becoming a better person without changing
the self. Paper presented at the Self and Identity Preconference of the Annual Meeting of the
Society of Experimental Social Psychology. Atlanta, Georgia.
Gollwitzer, P. M., Bayer, U. C., & McCulloch, K. C. (2005). The control of the unwanted. In
R. R. Hassin, J. S. Uleman, and J. A. Bargh (Eds.), The new unconscious (pp. 485–515).
New York: Oxford University Press.
*Gollwitzer, P. M., Bayer, U. C., Steller, B., & Bargh, J. A. (2004a). Delegating control to the
environment: Perception, attention, and memory for preselected behavioral cues. Manuscript
under review.
*Gollwitzer, P. M., & Brandsta
¨tter, V. (1997). Implementation intentions and eVective goal
pursuit. Journal of Personality and Social Psychology, 73, 186–199.
Gollwitzer, P. M., Grant, H., & Oettingen, G. (2004b, February). An individual diVerence
measure of the tendency to form implementation intentions. Poster presented at the 5th
Annual Meeting of SPSP. Austin, Texas.
Gollwitzer, P. M., & Moskowitz, G. B. (1996). Goal eVects on action and cognition. In E. T.
Higgins and A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles
(pp. 361–399). New York: Guilford.
Gollwitzer, P. M., & Schaal, B. (1998). Metacognition in action: The importance of implemen-
tation intentions. Personality and Social Psychology Review, 2, 124–136.
*Gollwitzer, P. M., Sheeran, P., & Seifert, E. (2004c). Intentions as symbols: How selfcompletion
aVects the translation of intentions into action. Manuscript under review.
*Gollwitzer, P. M., Sheeran, P., Tro
¨tschel, R., & Webb, T. L. (2004d). The control of behavior
priming eVects by implementation intentions. Manuscript under review.
*Gollwitzer, P. M., Tro
¨tschel, R., Bayer, U. C., & Sumner, M. (2004e). Potential costs of self
regulation by implementation intentions: Rebound and egodepletion eVects? Manuscript
under review.
Gollwitzer, P. M., & Wicklund, R. A. (1985). Selfsymbolizing and the neglect of others’
perspectives. Journal of Personality and Social Psychology, 56, 531–715.
114 PETER M. GOLLWITZER AND PASCHAL SHEERAN