Article Critique

Research Article

When Questions Change Behavior The Role of Ease of Representation Jonathan Levav1 and Gavan J. Fitzsimons2

1 Columbia University and

2 Duke University

ABSTRACT—In three experiments, we examined the mere-

measurement effect, wherein simply asking people about

their intent to engage in a certain behavior increases the

probability of their subsequently engaging in that behav-

ior. The experiments demonstrate that manipulations that

should affect the ease of mentally representing or simu-

lating the behavior in question influence the extent of the

mere-measurement phenomenon. Participants who were

asked about their intention to engage in various behaviors

were more likely to engage in those behaviors than par-

ticipants not asked about their intentions in situations in

which mentally simulating the behavior in the intention

question was relatively easy. We tested this ease-of-repre-

sentation hypothesis using both socially desirable and so-

cially undesirable behaviors, and our dependent variables

comprised both self-reports and actual behaviors. Our

findings have implications for survey research in various

social contexts, including assessments of risky behaviors by

public health organizations.

People are often asked to predict their likelihood of engaging in

a behavior in the near or distant future. For instance, political

pollsters survey potential voters about their likelihood of voting

during election years; market researchers survey customers

about their likelihood of purchasing a product; public-health

officials survey people about their likelihood of engaging in safe

sex. The implicit assumption in virtually all survey research is

that the act of responding to the question does not affect the

respondent’s probability of subsequently engaging in the be-

havior. As Fishbein and Ajzen (1975) commented, ‘‘If one wants

to know whether or not an individual will perform a given be-

havior, the simplest and probably most efficient thing that one

can do is to ask the individual whether he intends to perform that

behavior’’ (p. 369).

Although making such predictions might be ‘‘simple,’’ it is not

benign. Sherman (1980) showed that errors in predictions of

future behavior can be ‘‘self-erasing’’: People who had predicted

compliance with socially desirable behaviors were more likely

to subsequently engage in those behaviors than were people in a

control group, who had made no predictions about the behaviors.

Similarly, Greenwald, Carnot, Beach, and Young (1987) re-

ported a 25% increase in voting probability for people who had

been asked whether they intended to vote in the following day’s

election. Interestingly, questions about behaviors for which

people possess negative attitudes (e.g., socially undesirable

behaviors) lead to a decrease in the propensity to engage in those

behaviors (Sherman, 1980).

The self-erasing nature of errors in predictions even extends

to predicted purchases of very large items, such as automobiles

(Morwitz, Johnson, & Schmittlein, 1993). In a study conducted

on a nationally representative sample of more than 40,000

participants, asking a simple question about purchase intent

increased actual rates of automobile purchase in the following

6 months more than 35%. Morwitz et al. labeled this phenom-

enon the mere-measurement effect, as merely measuring inten-

tions changed respondents’ behavior.

The mere-measurement effect has been attributed to in-

creased accessibility of an attitude toward the target behavior in

the intention question (Feldman & Lynch, 1988; Morwitz &

Fitzsimons, 2004). For instance, Fitzsimons and Morwitz (1996)

found that asking a category-level intent question about the

likelihood of buying an automobile in the next 6 months led to

a systematic pattern of behavior at the subcategory level. Re-

spondents who had experience in the category (i.e., who were

automobile owners) were substantially more likely to purchase a

new automobile of the brand that they currently owned than

were respondents who were not asked the intent question. For

Address correspondence to Jonathan Levav, Columbia University, Graduate School of Business, Uris Hall, Room 509, 3022 Broadway, New York, NY 10027, e-mail: [email protected].

P S Y C H O L O G I C A L S C I E N C E

Volume 17—Number 3 207Copyright r 2006 Association for Psychological Science

nonowners, the purchase increase associated with answering the

intent question was observed for brands with a large share of the

market. The authors’ explanation was that although current car

owners’ attitude toward their automobile brand is most likely to

be positive and accessible, nonowners have positive and ac-

cessible attitudes toward frequently advertised brands. Morwitz

and Fitzsimons (2004) obtained similar results in a laboratory

setting when they manipulated attitude accessibility using un-

familiar brands of Canadian candy bars for which participants

did not have preexisting attitudes.

Although this empirical evidence is consistent with increased

accessibility as an explanation for the mere-measurement effect,

accessibility alone seems an incomplete explanation in light of

the relatively ephemeral nature of semantic primes (Bargh,

Gollwitzer, Lee-Chai, Barndollar, & Troetschel, 2001), as well as

evidence that mere-measurement effects in the financial-service

industry peak approximately 6 months following the intent

survey (Dholakia & Morwitz, 2002). This increase cannot be

explained by attitude accessibility alone. Indeed, Sherman’s

(1980) original explanation for his finding was that participants

had engaged in unspecified ‘‘pre-behavioral cognitive work’’ (p.

219).

In the experiments we report here, we investigated the nature

of the cognitive work that people engage in when responding to

intent questionnaires. We conjecture that intention questions

trigger the use of a simulation heuristic (Kahneman & Tversky,

1982), such that respondents mentally represent the target be-

havior and the instances in which they might engage in that

behavior. Our ease-of-representation hypothesis posits that the

effect of measuring intentions to engage in a behavior on sub-

sequent behavior is an increasing function of the ease with

which the behavior is mentally represented by the respondent.

Respondents may interpret ease of representation as reflecting

likelihood of the behavior, as suggested by research that links

ease of retrieval with perceived likelihood (Anderson & Godfrey,

1987; Schwarz & Vaughn, 2002; Tversky & Kahneman, 1973);

this ease might, in turn, spur an implementation intention

(Gollwitzer, 1999). Thus, intention questions lead to two related

mental operations: representation of the target behavior and

assessment of how easily the representation came about.

Questions about easy-to-represent behaviors should lead to

more pronounced mere-measurement effects relative to ques-

tions about harder-to-represent behaviors.

We tested our hypothesis in three experiments. When possi-

ble, we held constant the accessibility of the attitude toward the

target behavior, while manipulating ease of representation. Our

dependent variables included participants’ actual choices, as

well as self-reported behaviors.

EXPERIMENT 1

In our first test of the ease-of-representation hypothesis, we

manipulated the self-relevance of the intention question. We

predicted that respondents would find it easier to imagine

themselves engaging in a behavior than to imagine an average

classmate engaging in the behavior (at least for behaviors that

respondents were likely to have experienced previously). There-

fore, we expected that respondents asked about their own be-

havior would show a pronounced mere-measurement effect

relative to control participants who were not asked the intention

question, but that participants asked about an average class-

mate’s behavior would show either a smaller or no mere-mea-

surement effect.

Method

One hundred forty-five executive M.B.A. students were ran-

domly assigned to one of three conditions. In the control con-

dition (n 5 46), participants were asked to indicate their

likelihood of reading for pleasure in the next 2 weeks. In the self-

intent condition (n 5 51), participants were asked about their

likelihood of flossing their teeth in the next 2 weeks. Finally, in

the other-intent condition (n 5 48), participants were asked to

indicate the likelihood that one of their classmates would floss

his or her teeth in the next 2 weeks.

Two weeks following the initial questionnaire, the same par-

ticipants were asked to report how many times they had flossed

and how many times they had read for pleasure in the preceding

2 weeks.

Results

The data conformed to our predictions. Participants in the self-

intent condition reported flossing on a significantly greater

number of occasions than did control participants (6.25 vs.

4.11), t(96) 5 2.06, prep 5 .89, d 5 0.42. Thus, a mere-mea-

surement effect was obtained when the respondent was the actor

in question. In contrast, this pattern did not emerge in the other-

intent condition (4.23 vs. 4.11), t(93) 5 0.13. Furthermore, the

difference between the self- and other-intent conditions was

significant, t(98) 5 2.02, prep 5 .88, d 5 0.41. There were no

significant differences in the number of times participants in the

three conditions reported reading for pleasure. The pattern of

data supports our ease-of-representation hypothesis because

respondents who were expected to experience ease in imagining

a behavior showed a pronounced mere-measurement effect

relative to control participants, but participants who were ex-

pected to have more difficulty imagining the behavior did not.

To bolster our ease-of-representation account, we adminis-

tered a follow-up questionnaire to a sample from a similar

population. Participants were asked to indicate on a scale from 1

(extremely difficult) to 10 (extremely easy) either how easy it was

to imagine themselves flossing (n 5 37) or how easy it was to

imagine one of their classmates flossing (n 5 36). As expected,

the results indicated that it was easier for participants to imagine

themselves flossing (M 5 8.76) than a classmate flossing (M 5

5.58), t(71) 5 6.04, prep 5 .99, d 5 1.43. Nevertheless, it is

208 Volume 17—Number 3

When Questions Change Behavior

possible that in addition to manipulating ease, our self-rele-

vance manipulation manipulated self-investment. The subse-

quent experiments overcame this limitation.

EXPERIMENT 2

In this experiment, we manipulated ease of representation by

varying the question frame. Participants in the treatment con-

ditions were asked either a straightforward, positively framed

question about their intent to engage in a behavior (intent con-

dition) or one of two questions about the opposite intent: like-

lihood of not engaging in the behavior (negation condition) or

likelihood of avoiding it (avoidance condition).

On the basis of previous research, we expected that an intent

question about a target behavior for which people possess a

negative attitude would lead to a decrease in the propensity to

engage in that behavior (Sherman, 1980). Therefore, we ex-

pected that participants in the intent condition would be less

likely to engage in a negative behavior than would participants

in the control group. We expected this effect to be magnified in

the avoidance condition because the congruence between peo-

ple’s negative attitude and the avoidant behavior would make

the avoidant behavior easy to represent.

In contrast, despite the fact that both the avoidance and the

negation frames asked participants about their likelihood of

engaging in the opposite of the target behavior, we expected that

negation-frame participants would exhibit the same likelihood

of engaging in the negative behavior as their intent-frame

counterparts. This prediction was derived from Johnson-Laird’s

research on comprehension and reasoning (Johnson-Laird,

1983; Johnson-Laird, Legrenzi, Girotto, Legrenzi, & Caverni,

1999), which suggests that when individuals interpret dis-

course, they construct mental representations of what is true in a

proposition. Negations are not mentally construed because they

increase the load on working memory, which renders their rep-

resentation difficult. Instead, information about falsity is typically

treated as a ‘‘mental footnote’’ that is soon ‘‘forgotten’’ (Johnson-

Laird et al., 1999, p. 66). Consequently, we expected that the

negation frame would not facilitate a representation of avoidant

behavior, but that instead the negation information would be

forgotten and the question would be spontaneously recoded into a

positively framed (intent) statement. This recoding would then

give rise to the same behavior as in the intent condition.

Method

Ninety-nine undergraduates participated in this experiment.

Upon arrival in the lab, they completed a 10-question ‘‘market

research survey’’ about various consumption habits. The target

intent question, which appeared last, concerned consumption of

fatty foods in the following 1-week period. Participants were

randomly assigned to experimental conditions in which they

were asked to indicate their likelihood (on a 7-point scale) of (a)

consuming fatty foods in the coming week (intent condition; n 5

23), (b) not consuming fatty foods (negation condition; n 5 25),

or (c) avoiding consumption of fatty foods (avoidance condition;

n 5 26). In a fourth, control condition (n 5 25), participants

were asked about their likelihood of consuming orange drinks in

the coming week. All participants then proceeded with an hour-

long set of unrelated experiments.

As the session came to a close, respondents were informed

that their last task would be a taste test. They entered a separate

room where they were offered two snacks: mini rice cakes (low-

fat snack) and mini chocolate-chip cookies (high-fat snack).

They received a form and were instructed to consume either

snack in order to evaluate its taste. Participants’ choices were

recorded surreptitiously.

Results

The results conformed to our predictions. Whereas nearly all

(92%) participants in the control condition chose to eat the

cookies over the rice cakes, this propensity dropped equally in

the intent (65%) and the negation (68%) conditions, w2(1, N 5 48) 5 5.21, prep 5 .92, w 5 �.70, and w2(1, N 5 50) 5 4.5, prep 5 .90, w 5 �.75, respectively. In the avoidance condition, in which a representation of avoidant behavior was facilitated

by the wording of the problem, the propensity to eat cookies fell

much more dramatically (38%). The drop was significant rela-

tive to the control condition, w2(1, N 5 51) 5 15.99, prep 5 .99, w 5 �.32; the intent condition, w2(1, N 5 49) 5 3.50, prep 5 .86, w 5 �.04; and the negation condition, w2(1, N 5 51) 5 4.46, prep 5 .90, w 5 �.07.

To bolster our assertion that negation is more difficult to

represent than avoidance, we conducted a follow-up manipu-

lation-check study. A separate sample of the same participant

population was randomly assigned to the intent (n 5 29), ne-

gation (n 5 26), and avoidance (n 5 26) conditions. Participants

again answered a series of questions that were ostensibly part of

a market-research questionnaire. The target question again

appeared last, but this time was preceded by an unrelated ne-

gation question to be used as a covariate (‘‘How likely are you to

not purchase sneakers in the next three months?’’). Participants’

response time was measured for both the covariate and the target

questions. We included the covariate question to partial out the

additional time that it might take respondents to read the target

negation question; any remaining variance in response time

could therefore be attributed to difficulty of representation. We

assumed that easy-to-represent behaviors would be associated

with faster reaction times than harder-to-represent behaviors

(i.e., reaction times were an indirect measure of ease). Following

the covariate and target questions, participants were asked to

rate explicitly how easy it was to imagine the target behavior,

using a scale from 1 (extremely difficult) to 7 (extremely easy).

The results of the follow-up study support our interpreta-

tion of the results of the main experiment. Participants’ log-

Volume 17—Number 3 209

Jonathan Levav and Gavan J. Fitzsimons

transformed response times were significantly greater in the ne-

gation condition (M 5 7.00 s) than in either the intent condition

(M 5 5.13 s), F(1, 77) 5 10.20, prep 5 .98, d 5 0.73, or the

avoidance condition (M 5 5.32 s), F(1, 77) 5 5.29, prep 5 .92, d

5 0.52, but did not differ between the intent and avoidance

conditions (F < 1). (Note that this analysis included the covariate;

the results were also significant without the covariate.) We in-

terpret these results to mean that the behavior was easier to im-

agine—and therefore elicited quicker responses—in the intent

and avoidance questions than in the negation question. This in-

terpretation is supported by the direct measurements of ease.

Participants reported greater difficulty imagining the target be-

havior in the negation condition (M 5 4.08) than in either the

intent condition (M 5 5.38), F(1, 78) 5 9.89, prep 5 .98, d 5

0.71, or the avoidance condition (M 5 5.12), F(1, 78) 5 5.96, prep 5 .93, d 5 0.55, which in turn did not differ from each other, F(1,

78) 5 0.41.

Note that the manipulation used in this experiment varied

content in addition to ease of representation. Not only did ne-

gation prove more difficult to represent than avoidance, but

because the negation question was mentally transformed, the

mental representation itself was also different in the two con-

ditions. The work of Johnson-Laird et al. (1999) hints at a causal

link between ease and content: The mental transformation may

occur because representing negations taxes working memory and

is therefore difficult. Consequently, the pattern of ease observed

does not map directly onto the pattern of behavior. Although

participants in the negation and intent conditions behaved

equivalently, it was the participants in the avoidance and intent

conditions who indicated equivalent levels of ease. Thus, the

pattern of behavior observed in the main study reflects differ-

ences both in ease of representation and in content. In Experi-

ment 3, ease of representation was varied, while content

remained constant across conditions.

EXPERIMENT 3

In this experiment, we tested the ease-of-representation hy-

pothesis by manipulating both the regularity of the target be-

havior and the frequency referenced in the question. By

definition, regular behaviors occur at regular frequencies (e.g.,

daily), so we expected that assessing the likelihood that regular

behaviors will occur at regular frequencies should be relatively

easy. In contrast, thinking about performing a regular behavior

at an irregular frequency should be more difficult (Menon,

1993). For instance, a respondent asked to predict the likelihood

of engaging in a once-a-day activity eight times in the coming

week would have to speculate whether there might be a day when

the behavior would occur more than usual. No such uncertain-

ties arise for irregular behaviors because their frequency is not

tethered to regular intervals—one might be just as likely to

perform an irregular activity eight times in 1 week as two times

or seven times. Hence, frequency regularity should affect the

ease of representation for regularly occurring target behaviors,

but not for irregularly occurring behaviors.

We therefore expected a pronounced mere-measurement ef-

fect for a regular target behavior when the question frame ref-

erenced a regular frequency, but not when the question frame

referenced an irregular frequency. By contrast, we expected the

extent of the mere-measurement effect to be independent of

frequency regularity for items concerning irregular behaviors. It

is noteworthy that for regular behaviors, questions referencing

an irregular frequency, rather than a regular frequency, might

actually cause respondents to think more or ‘‘harder’’ because

they would need to surmise when the unusual occurrence might

take place. If so, a simple attitude-accessibility explanation

would predict a pronounced mere-measurement effect in this

condition because the elaboration required to answer the

question should increase accessibility.

Method

Sixty-three undergraduates were randomly assigned to one of

four experimental conditions in a two-by-two factorial design. In

the manipulation of target behavior, participants were asked to

indicate their likelihood (on a 7-point scale) of either (a) flossing

in the coming week or (b) reading for pleasure in the coming

week. The frequency-frame manipulation consisted of two lev-

els: regular and irregular. In the regular-frequency conditions,

participants were asked about the target behavior occurring

either 7 or 21 times in the coming week (i.e., in a regular fre-

quency). In the irregular-frequency conditions, participants

were asked about the target behavior occurring either 2 or 8

times in the coming week (i.e., in an irregular frequency). 1

Note

that flossing is typically considered a regularly occurring be-

havior, but reading for pleasure—especially for undergraduate

students—is typically irregular.

One week later, participants were given a follow-up ques-

tionnaire in which they were asked to report how many times

they had read for pleasure and how many times they had flossed

in the past week (the order of these questions was counterbal-

anced; there were no order effects). Hence, participants whose

target behavior in the initial survey had been reading served as

controls for participants who had been asked initially about

flossing, and vice versa.

Results

We tested the significance of the interaction of target behavior

and frequency frame on reported flossing and reported reading,

separately (see Fig. 1). We expected a significant interaction for

reported flossing, but not for reported reading. As expected, a

significant interaction was observed for reported flossing, F(1,

1 The different instantiations of regular and irregular frequencies were used to

test the robustness of our theory to various frequencies. Responses did not differ across instantiations, so the data were collapsed by regularity.

210 Volume 17—Number 3

When Questions Change Behavior

59) 5 7.92, prep 5 .96, d 5 0.73. Participants asked about their

intent to floss at a regular frequency showed a pronounced mere-

measurement effect, and participants asked about their intent

to floss at an irregular frequency showed a depressed mere-

measurement effect compared with control participants. In

contrast, the identical interaction test for reported reading was

not significant.

A planned contrast comparing reported flossing by partici-

pants whose target behavior was flossing was significant as ex-

pected (Ms 5 5.41 and 1.86 for the regular and irregular

frequency frames, respectively), F(1, 59) 5 11.22, prep 5 .98,

d 5 0.90. Also as expected, in the regular-frequency condition,

participants asked about their intent to floss reported signifi-

cantly greater flossing rates than their counterparts who had not

been asked about their intent to floss (i.e., who had been asked

about their intention to read; M 5 2.86), F(1, 59) 5 5.80, prep 5

.93, d 5 0.63. For the target behavior reading, we simply rep-

licated the mere-measurement effect: Irrespective of frequency

frame, participants asked about reading (M 5 5.66) reported

reading more than control participants (M 5 3.41), although not

significantly so, F(1, 58) 5 1.61, prep 5 .72, d 5 0.33. The

results support the ease-of-representation hypothesis: Regu-

larity of the frequency referenced in the question affected

reported behavior only for regularly occurring behaviors.

GENERAL DISCUSSION

We have presented evidence that the simple act of stating one’s

intent to engage in a behavior is associated with an increased

likelihood of subsequently engaging in the behavior when it is

easy to mentally represent or imagine. Participants asked their

intention to engage in a behavior were more likely to enact the

behavior when mentally simulating it was an easier task. When

possible, attitude accessibility was held constant across con-

ditions, and arguably in one case our predictions and those of an

attitude-accessibility account were in opposition.

Our data offer empirical evidence supporting Sherman’s

(1980) supposition that intention questions prompt respondents

to engage in ‘‘pre-behavioral cognitive work.’’ Using various

manipulations, our experiments shed light on the nature of this

work, and suggest that participants simulate the behavior in the

intent question. Note that this simulation is not necessarily

elaborative. It may instead be the case that the mental repre-

sentation and simulation occur virtually automatically. Indeed,

Fitzsimons and Williams (2000) demonstrated that the mere-

measurement effect is due largely to nonconscious factors. Ease

of representation may be viewed as one of these factors; re-

spondents may use ease as a fluency cue (Bornstein, 1989;

Reber, Winkielman, & Schwarz, 1998).

Even though we focused on ease of representation, our ma-

nipulations may have affected the content of mental represen-

tations, as well as their ease. To some degree this is inevitable—

the content of one’s representation of a behavior that is easy to

represent will necessarily differ from the content of one’s rep-

resentation of a behavior that is difficult to represent. In Ex-

periment 1, not only was a mental representation of oneself

flossing easier to conjure than a mental representation of an

average classmate flossing, but the content of these represen-

tations also differed because the actor differed. In Experiment 2,

as we acknowledged earlier, a negation not only was more dif-

Fig. 1. Number of occasions participants reported flossing (top panel) and reading for pleasure (bottom panel) in the week following the initial survey in Experiment 3. In the initial survey, each participant was asked about his or her intention of engaging in one of the target behaviors a specific number of times (which represented either a regular or an irreg- ular frequency of the behavior). Error bars indicate �1 standard error of the mean.

Volume 17—Number 3 211

Jonathan Levav and Gavan J. Fitzsimons

ficult to represent than avoidance, but also was represented

differently. In Experiment 3, the representation of a regular

behavior at an irregular frequency not only was more difficult

than the representation of the same behavior at a regular fre-

quency, but also was likely to have included imagination of

unusual events in the coming week. Consequently, ease and

content of representation were inextricably linked. We have

focused on ease because it is the most parsimonious explanation

for our results. Furthermore, the manipulation-check data in-

dicate that our experimental treatments exerted a significant

effect on ease.

Although our evidence suggests that ease of representation

influences the question-behavior link, it is unclear whether the

effect of ease is direct or is mediated by additional variables. For

instance, does ease facilitate the formation of implementation

intentions (Gollwitzer, 1999), which in turn lead to the mere-

measurement effect? Support for this conjecture comes from

Anderson and Godfrey (1987), who showed that imagining

oneself—but not other people—enacting a behavioral script

increases one’s expectations of engaging in the behavior. The

increased expectations may spur the formation of implementa-

tion intentions; ease may trigger a stronger intention than dif-

ficulty, which may trigger no intention at all. This hypothesized

relation between ease and implementation intentions may ex-

plain why mere-measurement effects endure beyond the typical

duration of priming effects—it is the implementation intentions

that endure, not the behavioral primes.

Alternatively, does ease affect accessibility? In particular,

one interpretation of the results of Experiment 1 is that re-

spondents in the other-intent condition could conjure fewer

instances of the behavior than could respondents in the self-

intent condition, and as a result of these fewer instances their

attitudes were less accessible. Answers to these questions will

enhance psychologists’ understanding of the question-behavior

link in particular and the attitude-behavior link in general.

The unintended impact of measuring intentions is wide-

ranging, and the unintended change in behavior may be harmful

to the respondent. For example, researchers often query at-risk

populations about their likelihood of engaging in risky or un-

healthy behaviors (e.g., drug use) as a way to assess the need for

prevention programs. Regrettably, respondents’ history of en-

gaging in unhealthy behaviors may facilitate their imagining

repeating such behaviors. In recent research, simply responding

to a question about their likelihood of recreational drug use in

the upcoming 2 months led to increased use among drug users,

but not among nonusers (Williams, Block, & Fitzsimons, in

press). Note that the source of ease in that study was somewhat

different from the source of ease in our Experiments 2 and 3.

Non-drug users in the study by Williams et al. may have found

the intention question difficult to answer because they had no

representation of themselves using drugs whatsoever (just as

other-intent participants in Experiment 1 were unlikely to have

had an accessible representation of their classmate flossing). In

our experiments, however, participants did have accessible

representations of engaging in the target behavior itself (e.g.,

flossing), but they may not have had an easily accessible rep-

resentation of engaging in the behavior in the circumstance

cited in the question (e.g., flossing eight times). The general

robustness of the mere-measurement effect, as well as its sig-

nificantly increased magnitude for behaviors that are easy to

represent and imagine, suggests the need to focus research on

assessment tools that prevent an increase in the probability of

unwanted behaviors.

REFERENCES

Anderson, C.A., & Godfrey, S.S. (1987). Thoughts about actions: The

effects of specificity and availability of imagined behavioral

scripts on expectations about oneself and others. Social Cognition, 5, 238–258.

Bargh, J.A., Gollwitzer, P.M., Lee-Chai, A., Barndollar, K., & Troet-

schel, R. (2001). The automated will: Nonconscious activation and

pursuit of behavioral goals. Journal of Personality and Social Psychology, 81, 1014–1027.

Bornstein, R.F. (1989). Exposure and affect: Overview and meta-

analysis of research 1968–1987. Psychological Bulletin, 106, 265–289.

Dholakia, U.M., & Morwitz, V.G. (2002). The scope and persistence of

mere-measurement effects: Evidence from a field study of cus-

tomer satisfaction measurement. Journal of Consumer Research, 29, 159–167.

Feldman, J.M., & Lynch, J.G., Jr. (1988). Self-generated validity and

other effects of measurement on belief, attitude, intention and

behavior. Journal of Applied Psychology, 73, 421–435. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and be-

havior: An introduction to theory and research. Reading, MA: Addison-Wesley.

Fitzsimons, G.J., & Morwitz, V.M. (1996). The effect of measuring intent

on brand level purchase behavior. Journal of Consumer Research, 23, 1–11.

Fitzsimons, G.J., & Williams, P. (2000). Asking questions can change

choice behavior: Does it do so automatically or effortfully? Journal of Experimental Psychology: Applied, 6, 195–206.

Gollwitzer, P.M. (1999). Implementation intentions: Strong effects of

simple plans. American Psychologist, 54, 493–503. Greenwald, A.G., Carnot, C.G., Beach, R., & Young, B. (1987). In-

creasing voting behavior by asking people if they expect to vote.

Journal of Applied Psychology, 72, 315–318. Johnson-Laird, P.N. (1983). Mental models. Cambridge, England:

Cambridge University Press.

Johnson-Laird, P.N., Legrenzi, P., Girotto, V., Legrenzi, M.S., & Cav-

erni, J.-P. (1999). Naı̈ve probability: A mental model theory of

extensional reasoning. Psychological Review, 106, 62–88. Kahneman, D., & Tversky, A. (1982). The simulation heuristic. In D.

Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under un- certainty: Heuristics and biases (pp. 201–208). Cambridge, En- gland: Cambridge University Press.

Menon, G. (1993). The effects of accessibility of information in memory

on judgments of behavioral frequencies. Journal of Consumer Research, 20, 431–440.

Morwitz, V.G., & Fitzsimons, G.J. (2004). The mere-measurement ef-

fect: Why does measuring intentions change actual behavior?

Journal of Consumer Psychology, 14, 64–73.

212 Volume 17—Number 3

When Questions Change Behavior

Morwitz, V.G., Johnson, E.J., & Schmittlein, D. (1993). Does measuring

intent change behavior? Journal of Consumer Research, 20, 46– 61.

Reber, R., Winkielman, P., & Schwarz, N. (1998). Effects of perceptual

fluency on affective judgments. Psychological Science, 9, 45–48. Schwarz, N., & Vaughn, L.A. (2002). The availability heuristic revis-

ited: Ease of recall and content of recall as distinct sources of

information. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.),

Heuristics and biases: The psychology of intuitive judgment (pp. 103–119). Cambridge, England: Cambridge University Press.

Sherman, S.J. (1980). On the self-erasing nature of errors of prediction.

Journal of Personality and Social Psychology, 39, 211–221.

Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for

judging frequency and probability. Cognitive Psychology, 5, 207– 232.

Williams, P., Block, L.G., & Fitzsimons, G.J. (in press). When asking

questions about health behaviors help versus hurt. Social Influence.

(RECEIVED 8/26/04; REVISION ACCEPTED 9/30/05; FINAL MATERIALS RECEIVED 10/10/05)

Volume 17—Number 3 213

Jonathan Levav and Gavan J. Fitzsimons