Microsoft Word - Artful Dodger.docx

The Artful Dodger: Answering the Wrong Question the Right Way

Todd Rogers Michael I. Norton

ideas42, Cambridge, MA, and Analyst Institute, Harvard Business School

Washington, DC

What happens when speakers try to “dodge” a question they would rather not answer by answering a

different question? In 4 studies, we show that listeners can fail to detect dodges when speakers answer

similar—but objectively incorrect—questions (the “artful dodge”), a detection failure that goes hand- in-

hand with a failure to rate dodgers more negatively. We propose that dodges go undetected because

listeners’ attention is not usually directed toward a goal of dodge detection (i.e., Is this person answering

the question?) but rather toward a goal of social evaluation (i.e., Do I like this person?). Listeners were

not blind to all dodge attempts, however. Dodge detection increased when listeners’ attention was diverted

from social goals toward determining the relevance of the speaker’s answers (Study 1), when speakers

answered a question egregiously dissimilar to the one asked (Study 2), and when listeners’ attention was

directed to the question asked by keeping it visible during speakers’ answers (Study 4). We also examined

the interpersonal consequences of dodge attempts: When listeners were guided to detect dodges, they

rated speakers more negatively (Study 2), and listeners rated speakers who answered a similar question

in a fluent manner more positively than speakers who answered the actual question but disfluently (Study

3). These results add to the literatures on both Gricean conversational norms and goal-directed attention.

We discuss the practical implications of our findings in the contexts of interpersonal communication and

public debates.

Keywords: Gricean norms, conversational implicature, goal-directed attention, inattentional blindness

Cite as: Rogers, T., & Norton, M. I. (2011). The Artful Dodger: Answering the Wrong Question the

Right Way.

Journal of Experimental Psychology: Applied, 17

(2), 139-147.

Don’t answer the question you were asked. Answer the question you

wish you were asked.

- Robert McNamara, describing the lessons he learned during his time

as Secretary of Defense during the Vietnam War (Morris, Williams, &

Ahlberg, 2004)

As this opening quote demonstrates, many in public life seek to

master the artful dodge, frequently attempting to wriggle out from

answering questions they would rather avoid. Although perhaps

most grating when performed by politicians, question dodging

occurs in a wide array of other contexts: corporate executives

avoiding reporters’ requests for their expectations for the next fiscal

quarter, employees sidestepping their bosses’ questions as to why

they are late for the third straight day, or spouses evading their

partners’ inquiries as to their whereabouts the previous evening.

Under what conditions does a dodge go undetected, allowing

speakers to escape unscathed? In the studies that follow, we show

that dodges can go undetected when a speaker responds to a question

by offering an answer to a similar question rather than the actual

question asked—provided that the listener’s attention is not

This article was published Online First April 25, 2011.

Todd Rogers, ideas42, Cambridge, MA, and Analyst Institute,

Washington, DC; Michael I. Norton, Department of Marketing, Harvard

Business School.

We thank Max Bazerman, Bill Simpson, Carey Morewedge, and Debby

Kermer for helpful feedback on this article. We also thank Greg Barron,

Natasia de Silva, and Adi Sundaram for help with materials development.

Correspondence concerning this article should be addressed to Todd

Rogers, 1870 North Scott St #649, Arlington, VA 22209. E-mail:

todd.rogers.mail@gmail.com

directed to explicitly assess whether the speaker answered the

question asked. As an illustration, consider a politician who is

asked about the illegal drug problem in America and instead

provides an answer about the need for universal health care: We

suggest that he has engaged in a successful dodge if listeners have

forgotten the question he was asked at the end of his answer.

Successful (and unsuccessful) question dodging can also have

interpersonal consequences; we also explore whether listeners rate

the speaker as positively after a successful dodge as they would

have had he/she been asked the question answered. Indeed, we

propose that in some cases, speakers are evaluated more highly

when they answer a similar question fluently than when they

answer the correct question disfluently.

How is it possible that listeners could fail to notice such

question dodging? We suggest that dodgers mask their deception

by exploiting implicit norms that direct listeners’ attention away

from detecting whether a particular answer truly addresses the

specific question asked. The title of our article is taken from the

Dickens character, the Artful Dodger, who was skillful at

distracting the attention of his victims wit

h conversation as he

picked their pockets (Dickens, 1838/1994); by assuming that

friendly conversation implied a lack of guile, his victims made

themselves vulnerable to his thievery. Indeed, Grice’s (1989)

theory of conversational implicature posits that listeners make

assumptions about the good faith cooperation of speakers. His

“Cooperative Principle” has four constituent maxims:

Communication will (a) contain the appropriate quantity of

information; (b) be of truthful quality; (c) be delivered in an

appropriate manner; and, most crucial to the present investigation,

(d) will be relevant to the topic at hand. Deceptive communication,

in this view, is communication that violates any of these maxims

(see also Burgoon, Buller, Guerrero,

139

Jou

rnal of Experimental Psychology: Applied

2011, Vol. 12, No. 2, 139-147 Public copy created by the Student Social Support R&D Lab

ROGERS AND NORTON

140

Afifi, & Feldman, 1996; McCornack, 1992). In support of this

theory, previous research has demonstrated that speakers prompted

to generate deceptive communication do in fact construct messages

that violate these maxims, listeners rate deceptive communication

as violating these maxims to a greater degree than truthful

communication, and listeners rate communication that violates these

maxims as more deceptive and less honest (Buller & Bur- goon,

1996; McCornack, Levine, Torres, Solowczuk, & Campbell, 1992;

Yeung, Levine, & Nishiyama, 1999).

Whereas previous research has focused on how speakers violate

Gricean norms when producing deceptive communications, we

focus on a different implication of those norms. Because listeners

assume that speakers generally abide by such norms, speakers may

be given the benefit of the doubt when they adhere closely enough

to conversational norms (Burgoon, Blair, & Strom, 2008; Clark &

Clark, 1977; O’Sullivan, 2003). Furthermore, whereas previous

research has demonstrated that egregious violations of

conversational norms are likely to be detected, we suggest that the

same norms allow speakers to push the boundaries of

communication: When speakers answer a question that is similar

enough to the question asked, listeners may fail to notice that the

answer offered is, in fact, irrelevant. In line with previous research,

however, we do suggest that when speakers stray too far—

answering questions that are egregiously dissimilar—listeners are

likely to notice.

Why might listeners be susceptible to dodging? We suggest that

the competing goals of listeners coupled with the limited nature of

their attentional capacity contribute to the failure to detect dodges.

First, a large body of research has documented the limited capacity

of attention, such that orienting toward one stimulus degrades

attention toward others (Cherry, 1953; Miller, 1956; Posner, 1980);

for example, attending to the goal of counting basketball passes

made by a group of people can cause viewers to fail to notice a man

in a gorilla suit walking through the game (Simons & Chabris,

1999). Although attending to whether a speaker has actually

answered a question is clearly an important goal for listeners, it

competes with other, often more automatic, goals. In particular,

research has suggested that when people encounter someone new,

they automatically pursue the social goal of evaluating that person:

Do I like this person? Do I trust this person? (Cuddy, Fiske, & Glick,

2008). Indeed, social interactions can cause people to fail to notice

even jarringly incongruous statements, as when people acquiesce to

requests (“May I cut in line to make a copy . . .”) accompanied by

substantively irrelevant “reasons” (“. . . because I need to make a

copy?”; Langer, Blank, & Chanowitz, 1978).

Given attentional limitations, we suggest that the automaticity of

listeners’ social goal of forming an impression of speakers makes

them vulnerable to failing to notice when a speaker dodges a

question asked by answering a similar question instead. In short, we

suggest that when listeners hear a speaker answer a question, their

assumption that the speaker will follow norms of cooperative

communication makes them vulnerable to failing to notice that the

speaker is not answering the question asked, and is instead pro-

viding an answer to an altogether different question.

Recall the previously described example: A politician’s

messaging consultants have advised him to talk about health care in

an upcoming debate, regardless of the actual questions asked.

Should the politician be asked about health care, his prepared

answer about health care will, of course, be a relevant and correct

response. However, should the question be about the illegal drug

problem, we suggest that the answer about health care is similar

enough that listeners would fail to detect the dodge. Should the

question be about the War on Terror, however, that answer would

be too dissimilar from the question asked, prompting listeners to

detect that the speaker attempted to dodge the question asked. Thus,

the similarity between the question asked and the question answered

is one crucial determinant of dodge detection. But also, as outlined

above, answering a similar question should only go unnoticed when

listeners’ attention is not directed toward dodge detection. We

predicted that focusing attention on whether the politician answers

the question asked should increase detection of dodge attempts.

Thus, the attentional goal of the listener was our second proposed

key factor in determining dodge detection.

Overview of Experiments

We tested the role of these two factors in four experiments,

investigating when dodge attempts will be noticed by listeners,

while also exploring the consequences of dodging on listeners’

interpersonal perceptions of speakers. In Study 1, participants

watched a debate in which a speaker answered the correct question

or answered a similar question. We also varied the listeners’ goals

(social evaluation vs. dodge detection) to examine the impact of

goal

s on dodge detection. In Study 2, participants listened to a

debate in which a speaker answered the correct question, answered

a similar question, or answered a dissimilar question. We assessed

the impact of similarity on dodge detection, as well as the impact of

successful and unsuccessful question dodges on listeners’ social

evaluation of the speaker. In Study 3, we included a condition in

which the speaker answered the correct question but in a disfluent

manner, exploring whether answering a similar question well might

be perceived more positively than answering the correct question

poorly. Finally, in Study 4, we explored a practical intervention to

increase the likelihood of dodge detection—making the question

asked more salient—that shifts listeners’ attention toward dodge

detection goals.

Study 1: Goals, Attention, and Successful Dodges

In Study 1, we explored how social goals contribute to listeners’

failure to detect question dodging. Participants watched an excerpt of

a debate in which a speaker answered the correct question (the

question he was asked) or a similar question (a topically related

question), and they were later asked to identify what question the

speaker had been asked. We expected that participants who viewed a

speaker answering a similar question would frequently fail to realize

that the speaker did not answer the question he had been asked. Most

important for our account, we predicted that the goal that participants

brought to viewing the debate would impact their likelihood of dodge

detection. Some participants were directed to attend to how they

socially evaluated the speaker (social goal), some were directed to

attend to whether the speaker answered the question he was asked

(detection goal), and others were given no specific guidance (no

goal). We expected that the social-goal and no-goal conditions would

result in the same low rates of dodge detection, suggesting that

participants’ default goal under the no- goal condition was, in fact,

social evaluation.

We also expected, however, that diverting attention from the default

social goal toward whether the speaker answered the question asked

would increase dodge detection, suggesting that limited attention is

one of the causes of listeners’ failure to detect dodges. We had two

specific hypotheses. First,

Hypothesis 1: Overall, recall of the question asked will be

lower when speakers answer a similar question relative to

when speakers answer the correct question.

Reflecting the role of attention, our second hypothesis was that

the following:

Hypothesis 2: Listeners directed to attend to whether the

speaker answered the question asked will exhibit increased

ARTFUL DODGER

141

dodge detection compared with listeners directed to attend to

either a social goal or no specific goal.

Method

Participants. Listeners (N = 333, 70% women, M

age

= 47.1

years, SD = 23.2) completed the study online. They were recruited

using an online survey company that compensated participants with

an online currency worth less than $8.

Procedure. In this study, each participant watched a 4-min clip

of a mock political debate designed to simulate an actual televised

debate (Norton & Goethals, 2004). Participants were randomly

assigned to one condition of a 3 (goal: none, social, dodge detection)

× 2 (dodge: correct, similar) between-subjects design.

Participants given no goal were told to pay attention because they

would be asked questions about the clips afterward. Participants

given a social goal were instructed to attend closely to how they felt

about the speakers and whether or not they liked the speakers; they

were told that they would later be asked about their feelings.

Participants given a detection goal were instructed to attend to

whether or not the speakers were answering the questions they were

asked; they were told they would later be asked about the questions

the speakers were asked.

All clips began with a question about education for the first

speaker, to which the speaker provided an answer about education.

The second speaker was then asked to give his opinion about

universal health care or asked to give his opinion about illegal drug

use in America. The second speaker’s response was always the

same, an answer about universal health care (see the Appendix).

Thus, participants either heard the second speaker answer the

correct question or a similar (but incorrect) question.

After listening to the entire clip, listeners completed two

multiple-choice questions recalling the question asked of each

speaker. The answer choices were education, health care, the drug

problem, or the War on Terror.

Results

Pretest. To establish that the answer to the similar question was

indeed recognized as being incorrect, we asked a different group of

participants (n = 48) to read the second speaker’s answer and select

which question best fit that answer, given the same four-option

multiple-choice question. Fully 94% of listeners identified health

care as the question that the second speaker answered, χ

(2) = 78.88,

p < .001. Removed from the difficulty of following a live exchange,

these participants experienced little uncertainty as to which question

fit the answer they had just read, suggesting that listeners who

recalled incorrectly in the results below from the live exchanges

were truly forgetting the question the speaker was asked.

Dodge detection. Before addressing the first two hypotheses, we

first define what constitutes a successful dodge. Success in question

dodging entails listeners being less likely to recall the question

asked of a speaker when the speaker answers a different question

than when the speaker answers the actual question. In the example

discussed above, a successful dodge occurs when the speaker

responds to a question about illegal drug use with an answer about

health care, and listeners fail to identify that the actual question was

about illegal drug use. A dodge is detected, on the other hand, when

listeners recall the actual question asked despite the speaker’s

efforts to dodge it by answering a similar question. We measured

dodge detection by assessing whether listeners selected the correct

multiple-choice response regarding which question the second

speaker was asked. Given that there were only four options, this left

listeners with a 25% chance of randomly selecting the correct

response. Consistent with recent research (Levine, 2001), we

conceptualized dodge detection as occurring on a binary scale

(accurately detected or not) as opposed to on a continuous scale.

First, Hypothesis 1 predicted that, overall, recall of the question

asked would be lower when speakers answered similar questions

compared with when speakers answered the correct question. As

expected, listeners who heard the speaker answer a correct question

were more accurate at recalling the question asked of the second

speaker (85%, n = 177) than were listeners who heard the speaker

answer a similar question (45%, n = 156), χ

(1) = 60.7, p < .001, ϕ

= .43, suggesting that answering a similar question impaired

listeners’ ability to remember the actual question the speaker was

asked. Indeed, although 55% of all listeners offered an incorrect

response when the speaker had been asked about illegal drug use,

40% of all listeners offered the specific incorrect response that the

speaker had been asked about health care— reflecting that the

speaker’s strategy of answering a question about illegal drugs with

an answer about health care had convinced many listeners that he

had actually been asked the question he chose to answer.

Whereas Hypothesis 1 predicted an overall effect of question

dodging, Hypothesis 2 made a prediction about the impact of

specific goals on dodge detection. Hypothesis 2 predicted that

dodge detection would be greater among participants who were

asked whether the speaker answered the question asked compared

with those who were given the social goal or no goal.

Listeners in the no-goal–correct condition were more accurate at

recalling the question asked of the second speaker (88%, n = 51)

than those in the no-goal–similar condition (39%, n = 52), χ

(1) =

27.4, p < .001, ϕ = .52. As expected, results for participants given a

social goal were strikingly similar: Listeners in the social-goal–

correct condition were more accurate (78%, n = 59) than those in

the social-goal–similar condition (25%, n 49), χ

(1) = 30.8, p <

.001, ϕ = .53. Confirming Hypothesis 2, there was no significant

difference in dodge detection among participants in the no-goal and

social-goal conditions when the second speaker answered a similar

question, χ

(1) = 2.3, p = .10.

Hypothesis 2 also predicted that when compared with participants

with no goal or a social goal, participants with a detection goal would

exhibit increased dodge detection when the speaker answered a

similar question. As can be seen clearly in Figure 1, the incidence of

dodge detection in the similar condition when participants were given

a detection goal was significantly higher than detection in both the

no-goal–similar condition (69% vs. 39%), χ

(1) = 10.1, p = .001, ϕ =

.31, and the social-goal–similar condition (69% vs. 25%), χ

(1) =

20.7, p < .001, ϕ = .45.

Study 2: (Dis)similarity and Dodge Detection

Study 1 explored our first key factor contributing to dodge

detection—the goal of the listener—demonstrating that under de-

fault social goal conditions, answering a similar question can result

in successful question dodging. Although Study 1 began to examine

the role of our second key contributor—similarity—in Study 2, we

extended the range of similarity, examining how answering a

dissimilar question can increase dodge detection. Of the three

conditions in Study 2, two were the same as in Study 1: a condition

in which the speaker answers the correct question and a condition

in which the speaker answers a similar question. We also included

a new condition in which the speaker answers an egregiously

dissimilar question: In this new condition, the speaker was asked

about the War on Terror, but he offered the same answer about

universal health care. Thus, we varied whether the speaker

answered the correct question (about health care), a similar question

ROGERS AND NORTON

142

(about the illegal drug use problem), or a dissimilar question (about

the War on Terror). In Study 2, we used audio excerpts from a mock

debate between two speakers, using the same text as in Study 1. We

hypothesized the following:

Hypothesis 3: Dodge detection will be greater when speakers

answer questions that are highly dissimilar to the actual

question asked than when speakers answer questions that are

similar to the actual question asked.

In addition, Study 2 explored not just the extent to which listeners

detected dodge attempts but the interpersonal ramifications of

detecting such dodges. The objective of artful dodgers, of course, is

to avoid answering the actual question and to ensure that listeners

fail to notice. We predicted that when listeners do notice

that speakers have violated the Gricean norm of relevance, the

speaker would suffer interpersonal costs. Hypothesis 3 predicted

that listeners would be more likely to detect dodges when the

speaker answered a dissimilar question than when the speaker

answered a similar question, we predicted the following:

Hypothesis 4: Listeners who hear a speaker dodge a question

by answering a dissimilar question will evaluate the speaker

more negatively than listeners who hear a speaker dodge a

question by answering a similar question.

Method

Participants. Listeners (N = 243, 57% women, M

age

= 43.0

years, SD = 12.7) completed the study in a computer lab as part of

a larger set of studies in exchange for $20.

Procedure. Listeners heard an audio excerpt from a mock debate

that began with a question about education for the first speaker, to

which the speaker provided an answer about education. Listeners

were then randomly assigned to one of three conditions in which the

question asked of the second speaker was varied; as in Study 1, the

second speaker always responded by answering a question about

universal health care. Some listeners heard the speaker asked a

question about health care (correct condition), some heard the

speaker asked about the illegal drug problem (similar condition),

and others heard him asked a question about the War on Terror

(dissimilar condition), before hearing his answer about health care.

After listening to the entire excerpt, listeners evaluated both

speakers on four interpersonal dimensions: how much they trusted

him, how much they liked him, how honest he was, and how capable

Figure 1. The impact of similarity and goals on recall of correct question

(Study 1). Error bars signify one standard error of the mean.

he was, all on 6-point scales (1 = not at all to 6 = very much). We

created a composite index of these items (Cronbach’s α = .95).

Finally, respondents answered the four-option multiple-choice

questions about which question each speaker had been asked, as

in Study 1.

Results

Pretest. To confirm that the question about the illegal drug

problem was more similar to the health care question than the

question about the War on Terror, a separate sample of

participants (n = 48) rated how similar both the illegal drug

problem and the War on Terror were to health care on 7-point

scales. As expected, the illegal drug problem was rated as more

similar to health care (M = 4.90, SD = 1.56) than the War on

Terror (M = 2.98, SD = 1.62), t(47) = 6.14, p < .01, d = 1.20.

Dodge detection. Accuracy in recalling the question asked of

the second speaker varied across conditions, χ

(2) = 15.13, p <

.01, ϕ = .25. However, listeners who heard the speaker answer the

correct question (82%, n = 82) and those who heard him answer a

dissimilar question about the War on Terror (70%, n = 74) were

not significantly different in their ability to recall the actual

question, χ

(1) = 2.81, p = .09. Thus, speakers’ attempts to answer

a dissimilar question prompted listeners to notice that dodge;

rather than believing that the speaker had actually been asked a

question related to the answer he offered (about health care),

participants remembered that he had been asked about the War on

Terror.

In contrast, as in Study 1, listeners who heard the speaker answer

a similar question were significantly worse at recalling the actual

question asked: Just 54% did so, lower than both other conditions,

s(1) > 4.45, ps < .04, ϕs > .16 (see Table 1). Whereas 46% of all

listeners offered an incorrect response, 26% of all listeners offered

the specific incorrect response that the speaker had been asked

about health care, demonstrating that his answer about health care

had led them to believe that he had been asked about health care.

Replicating Hypothesis 1, answering a similar question resulted in

a successful dodge; consistent with Hypothesis 3, dodging the

question by answering a dissimilar question resulted in recall on par

with when the speaker answered the correct question.

Interpersonal evaluations. Ratings of the second speaker on our

composite measure of interpersonal evaluation were impacted by

our manipulation, F(2, 240) = 7.05, p < .005. Supporting

Hypothesis 4, listeners who heard the speaker answer a dissimilar

question rated him more negatively (M = 2.75, SD = 1.39) than

those who heard the speaker answer the correct question (M = 3.46,

SD = 1.22) or a similar question (M = 3.28, SD = 1.06), ts > 2.75,

ps < .01, ds > 1.0 (see Table 1). It is important to note that

evaluations of the speaker who answered the correct question and

those of the speaker who answered a similar question did not differ,

t(167) = 1.02, p > .31, suggesting that dodging by answering a

similar question can be as effective, in interpersonal terms, as

actually answering the correct question. Ratings of the first speaker

were not impacted by our manipulations, F(2, 242) = 0.96, p = .38.

Was the success of the second speaker’s dodge related to

listeners’ interpersonal ratings of him? When the speaker

answered a dissimilar question, listeners’ ratings were highly and

negatively correlated with whether they recalled the actual

question (and thus noticed the speaker’s unsuccessful dodge

attempt), r(74) = -.49, p < .001; a less-than-artful dodge thus had

serious consequences for the would-be dodger. In contrast, no

such relationships emerged when the speaker answered the

correct question, r(82) = .11, p .> 31, or answered a similar

question r(87) = .08, p > .46. These findings suggest that there are

interpersonal costs of unsuccessful dodge attempts, whereas

ARTFUL DODGER

143

artful dodging can effectively sidestep those costs.

Study 3: Answer the Correct Question Poorly,

or Dodge?

The previous studies demonstrated that speakers can get away

with dodging a question without being detected and without social

cost by answering a similar question. In everyday life, people often

attempt to dodge questions when they are not prepared with a good

answer to the question asked—as in the example of politicians

stammering through poorly phrased answers to questions for which

they are not prepared. Indeed, another of Grice’s (1989)

conversational norms suggests that listeners expect speakers to offer

answers in an appropriate manner—one aspect of which is that it

not be delivered in an inappropriate style. In Study 2, we compared

the efficacy of dodging a question by answering a similar question

to bumbling through an answer to the correct question. We expected

that providing a well-delivered answer to a similar question would

result in lower recall of the question asked than answering the

correct question, replicating Hypothesis 1. Moreover, we expected

that interpersonal evaluations of speakers who successfully dodged

questions by answering a similar question would not be

significantly different from evaluations of speakers who answered

the correct question, replicating Hypothesis 4. Study 3 tested an

additional hypothesis:

Hypothesis 5: Listeners who hear a speaker answer the correct

question in a disfluent manner will evaluate the speaker more

negatively than listeners who hear a speaker offer a fluent

answer to a similar question.

Whereas 32% of all listeners offered an incorrect response in the

similar condition, 24% of all listeners offered the specific incorrect

response that the speaker had been asked about health care.

Interpersonal evaluations. Evaluations of the second speaker

were impacted by our manipulations, F(2, 272) = 5.26, p < .01.

Replicating Hypothesis 4, evaluations of the speaker did not differ

whether the speaker provided a good answer to the correct question

(M = 3.31, SD = 1.13) or to a similar question (M = 3.24, SD =

1.28), t(188) = 0.45, p > .65. Evaluations of the speaker who

answered the correct question but did so poorly, however, were

significantly lower than both other conditions (M = 2.78, SD =

1.12), ts > 2.51, ps < .02, ds > 0.38. Thus, supporting Hypothesis 5,

speakers who made an effort to answer the correct question—but

who did so poorly—were rated less positively than those who made

no effort to answer the correct question, instead answering a similar

question well (see Table 2). Ratings of the first speaker did not vary

by condition, F(2, 272) = 1.81, p > .16.

Method

Participants. Listeners (N = 275, 56% women, M

age

= 39.2

years, SD = 15.5) completed the study online. They were recruited

using an online survey company that compensated participants with

an online currency worth less than $8.

Procedure. In this study, listeners were assigned to watch one of

three video clips of a mock political debate. The first two conditions

used same videos as in Study 1: one in which the second speaker

offered an answer to the correct question (which was about health

care) and another in which the second speaker offere

d an answer to

a similar question (which was about the illegal drug problem). We

created a third condition in which the second speaker offered an

answer to the correct question (which was about health care) using

the identical text as the other conditions but while fumbling through

his answer, adding pauses, “um”s, and “uh”s throughout.

Listeners evaluated both speakers on the same four interpersonal

dimensions as in Study 2; we again created a composite index

(Cronbach’s α = .93).

Finally, listeners answered the four-option multiple-choice

question about what question both speakers had been asked, as in

the previous studies.

Results

Dodge detection. Listeners’ ability to identify the question asked

of the second speaker was again impacted by our manipulation, χ

(2)

= 8.50, p< .02, ϕ

= .19. As in study 1, listeners were generally able

to identify that the question asked of the second speaker was about

health care when he answered the correct question, whether he

answered it well (84%, n = 87) or poorly (84%, n = 73; see Table 2).

Recall in both conditions was significantly higher than when he

answered a similar question, where just 67% (n = 83) recalled the

actual question asked, χ

s(1) > 5.36, ps < .03, ϕs > .18, again

replicating Hypothesis 1.

Study 4: A Practical Intervention for Increasing

Dodge Detection

Studies 1 through 3 documented some of the underlying causes

and social consequences of dodge detection. In Study 4, we tested

an intervention that could be used in situations where dodges occur

to increase dodge detection: posting the text of the question asked

of the speakers on the screen during their answers. Indeed, this

strategy has recently been employed by the TV media during some

political debates. This intervention also provides a naturalistic

example of one of our earlier manipulations; in Study 1, participants

provided with a detection goal—those whose attention had been

diverted from a social evaluation goal toward evaluating the content

of the speaker’s answer—showed improved dodge detection. We

expected that posting the text on the screen would serve a similar

attentional function, shifting their focus from merely evaluating the

speaker on social dimensions to reminding listeners to evaluate the

relevance of the answer offered, thereby increasing dodge detection.

To the extent that participants watching a speaker dodge a question

are uncertain as to whether the speaker is in fact dodging, posting

the question on the screen—drawing attention to the dodge

attempt—should reduce that uncertainty, leading to accurate

detection of dodging.

In conditions in which the text of the question did not remain on

the screen, we expected to again replicate Hypothesis 1: Listeners

who heard the speaker answer a correct question would be more

accurate at recalling the actual question asked than listeners who

heard the speaker answer a similar question. We also tested one final

hypothesis in Study 4

ROGERS AND NORTON

144

Hypothesis 6: Listeners who see the text of the question posted

on the screen during the speakers’ answers will exhibit greater

dodge detection than listeners who do not see the text of the

question posted on the screen.

Method

Participants. Listeners (N = 288, 68% women, M

age

= 47.0

years, SD = 11.3) completed the study online. They were recruited

using an online survey company that compensated participants with

an online currency worth $8.

Procedure. In this study, listeners watched the same video clips used

in Study 1. As in Study 1, all clips began with a question about

education for the first speaker, which this speaker answered.

Listeners were then randomly assigned to one of four conditions in a

2 (dodge: correct, similar) × 2 (text on screen: absent, present)

between-subjects design. As before, listeners either watched the

second speaker answer the correct question (about health care) or a

similar question (about the illegal drug problem). As in the other

studies, the second speaker’s response was identical across all

condition, always about health care. In addition, some listeners

watched the videos with the text of the questions present on the

screen during the responses, and others watched with the text absent.

Following the video, participants were redirected to a new screen

that did not have either images from the debate or text; as in Study

1, listeners were asked to recall the questions the speakers had been

asked using the same four-option multiple-choice questions.

Results

Hypothesis 6 predicted that posting the questions asked of the

speakers on the screen while they delivered their answers would

increase dodge detection. This was confirmed: Listeners who heard

the second speaker answer a different question than the one he was

asked were much more likely to recall the actual question asked when

the text was posted on the screen (88%, n = 72) as opposed to when

it was not (39%, n = 71), χ

(1) = 35.7, p < .001, ϕ = 50 (see Figure

2). When the questions were absent from the screen, listeners were

again more likely to identify the second speaker’s actual question

when he answered the correct question (85%, n = 68) than when he

answered a similar question (39%), χ

(1) = 31.0, p < .001, ϕ = .47,

replicating Hypothesis 1. Consistent with the above studies, whereas

61% of all listeners who heard the speaker answer the similar question

failed to correctly recall the question asked of the second speaker,

44% of all listeners in those conditions believed he was asked about

health care. When the text was posted on the screen, however,

respondents who watched the second speaker answer a similar

question were able to recall the actual question (88%) as well as those

who watched him answer the correct question (95%, n = 77), χ

(1)

= 2.4, p = .10.

As predicted, posting the text of the question on the screen—and

therefore diverting attention toward dodge detection—significantly

decreased the second speaker’s ability to successfully dodge

questions.

General Discussion

We might expect that when people dodge a question by answering

a different question, listeners would both notice the dodge and

rate the dodger negatively. To the contrary, we found that

listeners engaged in a default goal of socially evaluating speakers

did not rate speakers poorly when they dodged a question by

answering a similar question—a lack of disapproval that went

hand-in-hand with their failure to detect that the speaker had

dodged. We proposed, and offered evidence in support of, two

key factors in dodge detection: the attentional goal of the listener

and the similarity of the answer offered to the actual question

asked. We demonstrated the role of goals by showing that when

listeners were given a direct goal to attend to the relevance of

speakers’ answers to the questions they were asked (Study 1), or

their attention was drawn to the speaker’s answers by posting the

text of the question on the screen (Study 4), dodge detection

increased. We documented the role of similarity by showing that,

although answering a similar question often went undetected

(Studies 1, 2, 3, and 4), answering a question that was egregiously

dissimilar increased dodge detection (Study 2). Finally, we

explored the interpersonal consequences of successful and

Figure 2. The impact of posting question text on recall of correct question

(Study 4). Error bars signify one standard error of the mean.

unsuccessful dodges: Listeners rated speakers who answered a

similar question as positively as those who answered the correct

question, but they rated those who answered a dissimilar question

and those who answered the correct question in a disfluent manner

more negatively (Studies 2 and 3).

We focused on two factors that

contribute to the failure to detect dodges, but future research should

explore other factors that con- tribute to and moderate the

phenomenon.

First, the current studies examined question dodging from the

perspective of outside observers —viewers of a televised political

debate. In the examples with which we opened the article, however,

we noted that question dodging often occurs in more directly

interpersonal contexts: between romantic partners or between

employers and employees. Future research should explore the

extent to which direct interpersonal contact moderates the

frequency of dodge detection. One view would suggest that the

more interpersonal, the more likely dodges would be detected: An

ARTFUL DODGER

145

employer is likely more motivated and engaged in obtaining the

truth from an employee than a TV viewer causally watching a

debate. At the same time, as reviewed in the introduction,

interpersonal interactions require a great deal of attention and can

lead people to fail to notice even glaringly obvious information

(Langer et al., 1978; Simons & Chabris, 1999). As a result, further

research is needed to explore the impact of variation in the

interpersonal nature of interactions on dodge detection.

Second, individual differences—in both attention and

motivation—are likely to influence dodge detection. Given the

important role of attention in the failure to detect dodges, it is

possible that individual differences in working memory may

moderate listeners’ susceptibility to dodging (Kane & Engle, 2003).

In addition, the inherent motivation of listeners to detect dodges is

likely to play a role. For example, we might expect individuals who

are more politically engaged to be more likely to detect politicians’

efforts to dodge. Although our studies were not designed to test this

question, in Study 1, we did assess whether participants had voted

in the 2008 presidential election, a proxy for political engagement.

It is interesting that a logistic regression revealed no evidence of an

interaction between voting and dodge detection, Wald(1) = 0.20, p

> .65; voters and nonvoters were equally likely to fail to detect

dodges. Still, we expect that a finer grained approach to political

engagement—for example, measuring people’s partisan affiliation

and commitment—would impact detection.

Third, we have proposed and offered evidence for a goal-directed

attentional mechanism underlying dodge detection, but there are

likely other processes that contribute to the failure to notice dodges.

In particular, the role of reconstructive memory processes warrants

further attention. A large body of research has demonstrated that

memories for past events can be strongly influenced by suggestions

that such events occurred; people have been shown to misremember

events ranging from meeting Bugs Bunny at Disney World to

becoming sick after eating particular foods as a child (Bernstein &

Loftus, 2009; Loftus, 1997). Our results suggest that people are

unable to remember an initial question when a speaker answers a

similar question; future research should explore the extent to which

speakers’ efforts to dodge actually change listeners’ memory of the

initial question, such that they might actually create a vivid memory

of the questioner asking the wrong question.

Our findings offer new insight into the impact of Gricean norms in

communication. Whereas much of the previous research in this area

has explored whether and how deceptive communication violates

these norms (e.g., Burgoon et al., 1996; McCornack, 1992), we

explored how listeners’ assumptions about the relevance of

speakers’ answers set the stage for speakers to choose not to answer

the actual question they were asked without negative consequences.

Our results are consistent with an interpretation that listeners adopt

a default goal of evaluating speakers on social dimensions, and that

this default goal leaves listeners susceptible to dodges – although

further research is needed to confirm this causal pathway. At the

same time, however, we showed that when dodges are detected—

as when speakers answer questions that are too dissimilar—listeners

rate them more negatively for their attempt to dodge.

From a practical standpoint, we identified several strategies for

counteracting these limitations. The results of Study 4, for example,

suggest that a practice occasionally used by TV networks during

political debates may increase dodge detection among listeners.

Indeed, recall of the question when the speaker attempted to dodge

by answering a similar question more than doubled, from 39%

without the text to 88% with the text. This is not only statistically

significant but practically significant. Given concerns that voters

are uninformed or misinformed and the many calls for increased

education of voters—from politicians and pundits alike—these

results suggest that very simple interventions can dramatically

increase the extent to which voters are focused on the substance of

politicians’ answers rather than their personal style (Shenkman,

2007; for discussion of the consequences for public policy of

uninformed and misinformed voters, see Caplan, 2007). At the same

time, however, we expect that not all question summaries are

created equally. The question summaries posted during political

debates are often overly vague (i.e., “the economy”), a vagueness

that ironically might facilitate question dodging when the question

answered is similar to that vague summary but different from the

question actually asked.

In many situations, such interventions are unlikely to be feasible.

It would undoubtedly be awkward to hold up a sign indicating the

specific question you expected an acquaintance to be answering, for

example. This is especially unfortunate because the low rates of

recall in our experiments are, if anything, overestimates of people’s

ability to detect dodges. After all, we presented listeners with a

forced-choice four-option question—giving them a 25% chance at

simply guessing the right answer. Obviously, this forced-choice

prompting does not arise naturally in the world. Accordingly,

increasing dodge detection in everyday interactions may be no easy

task. At the same time, however, we note that although we have

focused on the negative aspects of failing to detect dodges, constant

monitoring of potential dodges may be undesirable during many

interactions. For example, successful dodging may prevent needless

social friction in low-consequence interactions, as when someone

asks coworkers for their opinion on a new outfit. In addition, dodge

detection goals may be particularly harmful when people engage in

creative, wide-ranging conversations. Such exchanges are typified

by people making connections that are objectively irrelevant to the

immediate question at hand (De Bono, 1968). Still, our results

suggest that in many cases, dodges cause sought-after and relevant

information to go unspoken, with little awareness and few

consequences.

References

Bernstein, D.m., & Loftus, E.F. (2009). The consequences of false memories

for food preferences and choices. Perspectives on Psychological Science,

4, 135-139.

Buller, D.B., & Burgoon, J. K. (1996). Interpersonal deception theory.

Communication Theory, 6, 203–242.

Burgoon, J. K., Blair, J. P., & Strom, R. E. (2008). Cognitive biases and

nonverbal cue availability in detecting deception. Human

Communication Research, 34, 572–599.

Burgoon, J. K., Buller, D. B., Guerrero, L. K., Afifi, W. A., & Feldman, C.

M. (1996). Interpersonal deception: XII. Information management

dimensions underlying types of deceptive messages. Communication

Monographs, 63, 50–69.

Caplan, B. (2007). The myth of the rational voter: Why democracies choose

bad policies. Princeton, NJ: Princeton University Press.

Cherry, E.C. (1953). Some experiments on the recognition of speech, with

one and with two ears. Journal of the Acoustical Society of America, 25,

975–979.

Clark, H., & Clark, E. (1977). Psychology and language. New York:

Harcourt Brace Jovanovich.

Cuddy, A.J.C., Fiske, S.T., & Glick, P. (2008). Warmth and competence as

universal dimensions of social perception: The Stereotype Content Model

and the BIAS Map. In M. P. Zanna (Ed.), Advances in experimental social

psychology (Vol. 40, pp. 61–149). New York: Academic Press.

De Bono, E. (1968). New think; the use of lateral thinking in the generation

of new ideas. New York: Basic Books.

Dickens, C. (1994). Oliver Twist. New York: Penguin. (Original work

published 1838)

Grice, H. P. (1989). Studies in the way of words. Cambridge, MA: Harvard

University Press.

Kane, M. J., & Engle, R. W. (2003). Working memory capacity and the

control of attention: The contributions of goal neglect, response

competition, and task set to Stroop interference. Journal of Experimental

Psychology: General, 132, 47–70.

Langer, E. J., Blank, A., & Chanowitz, B. (1978). The mindlessness of

ostensibly thoughtful action. Journal of Personality and Social

Psychology, 36, 635–642.

ROGERS AND NORTON

146

Levine, T. R. (2001). Dichotomous and continuous views of deception:

A reexamination of deception ratings in information manipulation theory.

Communication Research Reports, 18, 230–240.

Loftus, E. F. (1997). Creating false memories. Scientific American, 277,

70–75.

McCornack, S.A. (1992). Information manipulation theory, Communication

Monographs, 59, 1-16.

McCornack, S. A., Levine, T. R., Torres, H. I., Solowczuk, K. A., &

Campbell, D. M. (1992). When the alteration of information is viewed as

deception: An empirical test of information manipulation theory.

Communication Monographs, 59, 17–29.

Miller, G.A. (1956). The magical number seven plus or minus two: Some

limits on our capacity for processing information. Psychological Review,

63, 81-97.

Morris, E. (Producer and Director), Williams, M. (Producer), & Ahlberg, J.

(Producer). (2004). The fog of war: Eleven lessons from the life of Robert

S. McNamara [Motion picture]. United States: Sony Pictures Classics.

Norton, M. I., & Goethals, G. R. (2004). Spin (and pitch) doctors: Campaign

strategies in televised political debates. Political Behavior, 26, 227–248.

O’Sullivan, M. (2003). The fundamental attribution error in detecting

deception: The body-who-cried-wolf effect. Personality and Social

Psychology Bulletin, 29, 1316–1327.

Posner, M. I. (1980). Orienting of attention. Quarterly Journal of

Experimental Psychology, 32, 3–25.

Shenkman, R. (2007). Just how stupid are we? Facing the truth about the

American voter. New York: Basic Books.

Simons, D. J., & Chabris, C. F. (1999). Gorillas in our midst: Sustained

inattentional blindness for dynamic events. Perception, 28, 1059–1074.

Yeung, L. N. T., Levine, T. R., & Nishiyama, K. (1999). Information

manipulation theory and perceptions of deception in Hong Kong.

Communication Reports, 12(1), 1–11

Appendix

Q1: What do you see as the major factors that need to be

addressed regarding the American education system?

A1: I’m glad you asked me about this. When I think about

America’s future there are so many important pressing needs, and

education is the root of America’s future. When every child in

America receives a good education, our country will advance even

further to lead the world technologically and productively. The first

two factors that need to be addressed are getting the kids to stay in

school through motivation and positive role models and ensuring that

the time spent in school is not wasted and the curricula in schools are

advanced. Too many children are not being motivated to achieve.

Instead, they are motivated only to do the minimum. We need to

motivate our kids. Teachers should act as mentors and friends, giving

students a sense of self-worth and accomplishment. Parents need to

encourage their kids to do the best they can and instill values and hard

work and achievement in them. Parents and teachers can act as

positive role models, but so can other adults that children have access

to, people like actors and even politicians. If we demonstrate that

working hard pays off, children will be more motivated to stay in

school. In terms of the curricula, schools across the country are

teaching entirely different things. A child should be able to get the

same education across the country. Furthermore, an education should

be practical. Students should learn the skills necessary to go on to

college or to get a job. We need to make sure that classrooms are

productive places and teachers cover material so that our students

have the most knowledge possible when they graduate from high

school and enter the world.

Q2: What are your main concerns about the way in which

Americans receive health care?

Q3: What are your main concerns about the rising drug problem in

America?

Q4: (Study 2 only) What are your main concerns regarding the

War on Terror?

A2: I am glad you asked me about this. There are so many

challenges facing America today. Many of our problems have

arisen because too many Americans cannot afford the care that they

need. Costs are always increasing and if you do not have insurance,

certain necessary procedures can be out of the question. Even if the

government were to distribute aid, it would be difficult to determine

to whom and how much. As the costs of care rise, it becomes more

and more difficult for the government to pay for this care for so

many people. Individual costs go up and insurance costs go up.

While most employers provide insurance, increased costs will have

negative effects on this provision as well, which will undoubtedly

lower the number of insured Americans. It seems to me that the

primary problem is the cost of health care and we need to address

the problem by developing new technologies and processes to lower

the costs. Once we have lowered the costs, we can begin to address

the larger issues of distribution of aid and revamping the systems

that are already in place. However, before we lower the costs, my

main concern is that Americans are not receiving the health care

that they all need because they are not financially able