939534

research-article2020
PSSXXX10.1177/0956797620939534Fazio, SherryEffect of Repetition Across Development

ASSOCIATION FOR
Research Article PSYCHOLOGICAL SCIENCE
Psychological Science

The Effect of Repetition on Truth 2020, Vol. 31(9) 1150–1160

© The Author(s) 2020
Article reuse guidelines:
Judgments Across Development sagepub.com/journals-permissions
DOI: 10.1177/0956797620939534
https://doi.org/10.1177/0956797620939534
www.psychologicalscience.org/PS

Lisa K. Fazio and Carrie L. Sherry

Department of Psychology and Human Development, Vanderbilt University

Abstract
According to numerous research studies, when adults hear a statement twice, they are more likely to think it is true
compared with when they have heard it only once. Multiple theoretical explanations exist for this illusory-truth effect.
However, none of the current theories fully explains how or why people begin to use repetition as a cue for truth. In
this preregistered study, we investigated those developmental origins in twenty-four 5-year-olds, twenty-four 10-year-
olds, and 32 adults. If the link between repetition and truth is learned implicitly, then even 5-year-olds should show
the effect. Alternatively, realizing this connection may require metacognition and intentional reflection, skills acquired
later in development. Repetition increased truth judgments for all three age groups, and prior knowledge did not
protect participants from the effects of repetition. These results suggest that the illusory-truth effect is a universal effect
learned at a young age.

Keywords
truth, repetition, illusory truth, children, misinformation, open data, open materials, preregistered

Received 10/23/19; Revision accepted 4/30/20

When deciding whether statements are true or false in statements are more easily processed and understood
laboratory settings, adults who hear a statement twice than novel statements and that this ease of processing,
are more likely to think that it is true compared with or fluency, is used as a signal for truth in a similar way
when they have heard it only once. This illusory-truth that it is used as a signal for memory and other judg-
effect has been found in over 100 studies over the past ments (e.g., Alter & Oppenheimer, 2006; Jacoby, Kelley,
40 years (see Dechêne, Stahl, Hansen, & Wänke, 2010, Brown, & Jasechko, 1989; Jacoby, Kelley, & Dywan,
for a meta-analytic review). The effect occurs with trivia 1989; Tversky & Kahneman, 1973; Unkelbach, 2007;
facts (Hasher, Goldstein, & Toppino, 1977), political Unkelbach & Greifeneder, 2013). Thus, other manipula-
opinions (Arkes, Hackett, & Boehm, 1989), and even tions that increase ease of processing, such as high-
false news headlines (Pennycook, Cannon, & Rand, contrast font colors or predictable rhymes, also increase
2018). It also occurs with statements that contradict truth ratings (McGlone & Tofighbakhsh, 2000; Reber &
people’s existing knowledge—for example, “The skirt Schwarz, 1999; Silva, Garcia-Marques, & Mello, 2016).
that Scottish men wear is called a sari” (Fazio, 2020; Finally, recent research suggests that repetition may
Fazio, Brashier, Payne, & Marsh, 2015). These findings increase the cohesiveness of the concepts represented
have taken on new importance in the modern world, in the statement in memory and that this cohesiveness
where misinformation and false statements are repeat- is used as a signal for truth (Unkelbach & Rom, 2017).
edly presented by politicians and on social media. Manipulations that increase the cohesiveness of the
Multiple theories exist about why repetition increases
perceived truth (see Unkelbach, Koch, Silva, & Garcia-
Marques, 2019, for a review). One explanation is that Corresponding Author:
Lisa K. Fazio, Vanderbilt University, Department of Psychology and
repetition increases the familiarity of a statement, and Human Development, 230 Appleton Place, No. 552, Jesup 105,
people believe that things they have heard before are Nashville, TN 37203
likely true (Arkes et al., 1989). A second is that repeated E-mail: [email protected]
Effect of Repetition Across Development 1151

network, such as deeper processing, increase the illu-

sory-truth effect (Unkelbach & Rom, 2017). All three Statement of Relevance
theories are related and somewhat overlapping; increas-
ing the cohesiveness of the representation through In a world where misinformation is prevalent and
deeper processing also likely increases ease of process- spreads rapidly throughout society, it is essential
ing, and repetition increases familiarity, fluency, and to be able to distinguish truth from falsehood. Yet
cohesiveness. It is likely that all of these processes play prior research demonstrates that people often use
a role in the illusory-truth effect. unreliable cues, such as repetition, to judge truth
For all three theories, however, a key unexplained rather than more accurate cues, such as their prior
step is how people first begin to associate familiarity, knowledge or the source of the information. In
fluency, or a coherent referential network with truth. this study, we examined the developmental origins
Examining the development of the repetition-induced of this illusory-truth effect. Among 5-year-olds,
truth effect can help answer this question. One possibil- 10-year-olds, and adults, we found that repeated
ity is that our brains naturally and automatically track statements were more likely to be judged as true
the relevant statistics and quickly realize the correlation compared with statements presented only one
between these proximal cues and truth. That is, through time. In addition, prior knowledge did not protect
implicit or statistical learning, people realize that state- against this effect. Even adults were more likely to
ments that are familiar, cohesive, and fluent are also say that the false statement “a calf is a baby horse”
likely to be true (e.g., Aslin, 2017; Frost, Armstrong, was true when it was repeated. These findings
Siegelman, & Christiansen, 2015; Thiessen, Kronstein, suggest that the illusory-truth effect occurs across
& Hufnagle, 2013). Note that this is similar to the learn- development and for a variety of statements. We
ing-from-feedback account provided by Unkelbach and should all be wary of the dangers of repeating false
Greifeneder (2013). If this is the case, children should information.
begin to show the effect as soon as they understand
the concept of truth. Alternatively, the effect may occur
only for older children who have more experience different internal cues and later memory or belief. For
intentionally reflecting on their experiences and what example, 7- and 8-year-olds do not yet realize that
is true or false. That is, through metacognition, people there is an inverse relation between study time and
may develop a naive theory that these proximal cues later recall (items that require more effort to encode
are a signal for truth (e.g., Schwarz, 2010). are less likely to be recalled on a later test; Koriat,
Ackerman, Lockl, & Schneider, 2009a, 2009b). These
Children’s Metacognitive Abilities and metacognitive abilities continue to improve during ado-
lescence; older teens are better able to monitor the
Their Understanding of Truth
accuracy of their task performance than younger teens
Children begin to notice inaccurate statements at a (Weil et al., 2013).
young age. Sixteen-month-olds look longer at speakers There is some preliminary evidence that young chil-
who mislabel familiar objects (Koenig & Echols, 2003), dren may use processing fluency as a signal for truth.
and preschoolers can accurately indicate whether a When guessing where a ball was hidden, 4- and 5-year-
speaker is labeling objects correctly or incorrectly (Koenig, olds were more likely to rely on the testimony of a
Clément, & Harris, 2004). By the age of 5, children use character who spoke fluently compared with a disfluent
their knowledge of the world to judge the reality status character whose speech was overlaid with digital noise
of novel creatures. Novel animals with typical features (Bernard, Proust, & Clément, 2014). However, there
(e.g., “A Jarpat is kind of like a lion. It roars.”) are judged were a number of differences between this paradigm
as real, and novel animals with atypical or impossible and the typical illusory-truth effect. The most important
features (e.g., “A Calak is kind of like a horse. It climbs is that the spoken messages were in direct conflict, so
trees.”) are judged as not real (Lopez-Mobilia & Woolley, the child was forced to side with only one of the char-
2016). Similarly, 5- and 6-year-old children reject the acters. When forced to choose between a fluent and
idea that novel animals can accomplish counterintuitive disfluent speaker, young children relied on the fluent
actions (e.g., “Spalts can jump very high. They can jump testimony, but if the speakers were presented sequen-
so high that they can jump over hills.”; Lane & Harris, tially, the children may have believed both speakers.
2015). Thus, it is unclear how young children will react in the
However, 5-year-olds are still developing their meta- illusory-truth paradigm, in which each statement is pre-
cognitive skills (see Schneider, 2008, for a review). In sented individually and children are asked whether it
particular, they are still learning the relation between is true or false.
1152 Fazio, Sherry

Knowledge-Conditional Model
Knowledge Respond
K Retrieved Accurately

Rely on Respond
F Fluency “True”
1−K No Knowledge
Retrieved G Respond
Guess True
“True”
Do Not Rely
1−F on Fluency
Respond
Guess False
1−G “Not True”

Fluency-Conditional Model
Rely on Respond
F Fluency “True”

Knowledge Respond
K Retrieved Accurately
1−F Do Not Rely
on Fluency G Respond
Guess True
“True”
No Knowledge
1−K Retrieved
Respond
Guess False
1−G “Not True”

Fig. 1. Knowledge-conditional and fluency-conditional models of illusory truth. Parameter

values represent the probability that the cognitive process contributes to the behavior. K
knowledge; F fluency; G guess “true.”

Effects of Prior Knowledge model (Fig. 1, top), people primarily use their prior
knowledge to make truth judgments; only when that
A secondary goal of this research was to replicate previ- knowledge search fails do people rely on fluency. In
ous research on the effects of prior knowledge on the contrast, in the fluency-conditional model (Fig. 1, bot-
illusory-truth effect. Studies have shown that repetition tom), people are able to rely on fluency alone to make
increases perceived truth, even when people have prior their judgments (without consulting prior knowledge).
knowledge that contradicts the repeated falsehood Note that although we use the term fluency in the mod-
(Fazio, 2020; Fazio et  al., 2015), and that repetition els, participants may be relying on any cue that increases
affects belief similarly for both plausible and implausible with repetition (fluency, familiarity, or cohesion). With
statements (Fazio, Rand, & Pennycook, 2019). To con- adults, the fluency-conditional model provided a good fit
firm and extend these findings, we designed our stimuli for participants’ responses, but the knowledge-conditional
so that each age group would have relevant prior knowl- model did not (Fazio et al., 2015). In the current study,
edge for some statements and limited knowledge about we used the same two models to examine the response
others. patterns of both children and adults.
This manipulation also allowed us to examine how
knowledge and repetition interact to produce truth
judgments. Our previous research suggests that people
do not always consult their prior knowledge when mak-
The Current Study
ing truth judgments and instead rely only on fluency The current study examined the effect of repetition on
or other proximal signals (Fazio et al., 2015). That work the truth ratings of 5-year-olds, 10-year-olds, and col-
contrasted two different processing-tree models. In lege students. Our main question was whether 5- and
each model, the parameters represent the probability 10-year-old children use repetition as a cue for truth.
that a specific unobserved cognitive process contributes As a secondary interest, we also examined whether
to the observed behavior. In the knowledge-conditional prior knowledge protects against the illusory-truth
Effect of Repetition Across Development 1153

effect. The experiment consisted of two phases. During (repetition: new, repeated) 2 (statement truth: truth,
an exposure phase, participants rated true and false falsehood) 3 (knowledge level: preschool, elementary
statements as “interesting” or “not interesting.” Then, school, middle school) 3 (age: 5-year-olds, 10-year-
during the truth phase, participants were asked to judge olds, adults) mixed design. Repetition, statement truth,
whether statements were “true” or “not true.” Some of and knowledge level were all manipulated within sub-
these statements were new, and some were repeated jects, and age was a between-subjects factor.
from the exposure phase. To examine the interaction
between the effects of prior knowledge and repetition,
we varied the difficulty of the statements. For some,
Materials
even the 5-year-olds should have prior knowledge to We selected 48 nature facts as stimuli. To ensure that
confirm or disprove the statements. For others, the truth some facts would be known by all of the participants
status was likely unknown even to the adults. If the and others would be unknown, we selected items from
connection between repetition and truth is learned rela- three versions of the game Brain Quest. One third of
tively quickly and implicitly, then all age groups should the statements came from the preschool and kindergar-
be more likely to rate the repeated statements as true ten games (Feder & Bishay, 2016b, 2016c), one third
than the new statements. If, however, people slowly came from the game designed for third graders (Feder
develop a naive theory of truth through reflection and & Bishay, 2012), and the rest came from the game
metacognition, then repetition may not have an effect designed for seventh graders (Feder & Bishay, 2016a).
until later in development. Because of the limited number of nature-related facts
in the seventh-grade game, we added eight nature facts
from an online article (Stryker, 2014) and two from a
Method set of general-knowledge norms (Tauber, Dunlosky,
Rawson, Rhodes, & Sitzman, 2013). For each correct
Participants
fact, we created a matching falsehood that referred to
The participants were twenty-four 5-year-old children a plausible but incorrect alternative (e.g., “A spider has
(age: M 5.39, SD 0.27), twenty-four 10-year-old six legs”). Sample statements can be seen in Table 1,
children (age: M 10.57, SD 0.29), and 32 adults. We and the full set of materials is available online (https://
preregistered our intention to recruit 25 participants for osf.io/hw3gy).
each age group, but we decided to stop at 24 in order To counterbalance statement truth and repetition
to have even numbers in each condition. Our past across participants, we divided the 16 statements in
research with a binary true/not-true decision yielded each knowledge level into four sets of four statements.
an illusory-truth effect size (d) of 0.58 (Fazio et  al., For each participant, we presented one set as new
2015). A power analysis using G*Power suggested that truths, one set as new falsehoods, one set as repeated
for a one-sided t test (predicting more “true” judgments truths, and one set as repeated falsehoods. Which set
for repeated statements), 20 participants per group appeared in each format was counterbalanced across
would be required to detect an effect of that size with participants.
80% power and an alpha of .05 (Faul, Erdfelder, Lang,
& Buchner, 2007). We recruited a larger-than-expected
number of adult participants because an overzealous Procedure
research assistant posted extra time slots. The children For the child participants, informed consent was
were recruited using state birth records from Nashville, obtained from the parent or guardian, and verbal or
Tennessee, and from community events. Each child written assent was obtained from each participant. The
received a small toy worth less than $5 as compensation. entire experiment was presented on a touch-screen
The adult participants were recruited from Vanderbilt computer, and both the instructions and the statements
University’s participant pool and received course credit were presented verbally. (The instructions were pro-
for their participation. An additional four 5-year-olds vided by the experimenter, and the statements were
were excluded for missing at least two of the three prac- presented by the computer.)
tice questions (a preregistered exclusion criterion). The experiment began with the exposure phase. The
children were introduced to a digital cartoon character
named Ruby the Robot, who was going to tell them
Design about animals and nature. They were told that “Ruby
The experiment was conducted in the laboratory during knows a lot about some things, and not a lot about
a single session and consisted of two phases: an expo- other things. So, some of the things Ruby tells you will
sure phase and a truth phase. The experiment had a 2 be true, and some will be not true.” The child’s job was
1154 Fazio, Sherry

Table 1. Sample True and False Statements From Each Knowledge Level

True False
Knowledge level: preschool
A calf is a baby cow. A calf is a baby horse.
Tomatoes grow above ground. Potatoes grow above ground.
A sheep is the animal that wool comes from. A goat is the animal that wool comes from.
Knowledge level: elementary school
The dodo is an animal that is extinct. The emu is an animal that is extinct.
Reptiles are cold-blooded animals. Birds are cold-blooded animals.
Porcupines raise their quills to defend themselves. Porcupines shoot their quills to defend themselves.
Knowledge level: middle school
Cumulonimbus clouds often bring thunderstorms. Cirrus clouds often bring thunderstorms.
All clownfish are born male. All clownfish are born female.
Antarctica is the largest desert on Earth. The Sahara is the largest desert on Earth.

to listen to each statement and decide whether it was that you have already heard today and some new
“interesting” or “not interesting.” The children indicated things.” For each statement, the children first decided
their choice by pressing the relevant button on the whether it was “true” or “not true” and then whether
touch screen (see Fig. 2). For the first few trials, the they were “very sure” or “not so sure.” To ensure that
“interesting” and “not interesting” buttons were labeled the children understood the task and the rating scale,
by the experimenter to ensure that the younger children we first presented three very simple non-nature-related
knew which was which. Twenty-four of the 48 state- practice statements from the preschool and kindergar-
ments were presented during the exposure phase. After ten Brain Quest games (e.g., “A hat goes on your feet”).
the exposure phase, participants had a quick break, Following our preregistration, we excluded any partici-
during which they worked on mazes or connect-the- pant who incorrectly identified the truth of two or more
dots activities for approximately 90 s. practice statements. After the practice items, partici-
After the filler task, participants were asked to listen pants were presented with all 48 statements.
to more statements from the same cartoon character. The adult study was similar, but the instructions were
They were reminded that “Ruby knows a lot about some presented visually on the screen rather than verbally
things and not a lot about other things, so some of what by the experimenter. At the start of the experiment,
you hear will be true and some will be not true.” The adults were told that “This adult study was created from
children were also told, “Ruby will tell you some things a child study which is why Ruby, our talking robot, is
used.” They then were given the exact same instructions
and completed the exact same procedures as the child
participants, although they responded using a mouse
instead of the touch screen. In addition, instead of solv-
ing mazes, the adults worked on visuospatial puzzles
between the exposure and truth phases.

Results
All data and supplemental analyses are available online
at OSF, along with our preregistration of the primary
analyses and sample size (https://osf.io/hw3gy).

Do 5- and 10-year-old children use

repetition as a cue for truth?
Fig. 2. Screenshot from the exposure phase. Participants listened As shown in Figure 3, all three age groups were more
to a statement that was spoken aloud by Ruby the Robot and then
indicated whether the statement was “interesting” or “not interesting”
likely to rate repeated statements as true than new
on the touch screen. Screenshots from the truth phase are available statements. As detailed in our preregistration, the pri-
online at https://osf.io/hw3gy. mary analysis was a 2 (repetition: new, repeated) 2
Effect of Repetition Across Development 1155

5-Year-Olds 10-Year-Olds Adults

1.0

0.8
Proportion Rated True

0.6

0.4

0.2

0.0

New Repeated New Repeated New Repeated

Fig. 3. Proportion of new and repeated statements rated as true by each age group. Each
dot represents one participant. The filled diamond represents the group mean, and error
bars represent standard errors.

(statement truth: truth, falsehood) 2 (age: 5-year-olds, slightly larger for false statements than true statements.
10-year-olds, adults) analysis of variance (ANOVA) on No other main effects or interactions were significant,
the proportion of statements that were rated as true. As largest F 0.90, p .412.
would be expected, true statements (M .78, 95% con- Similar results were found when we recoded partici-
fidence interval, or CI [.75, .81]) were more often pants’ responses into a 4-point scale (1= very-sure false,
rated as true than were false statements (M .52, 95% 2 not-so-sure false, 3 not-so-sure true, 4 very-sure
CI [.49, .55]), F(1, 77) 255.19, p .001, p2 .77. We true). The full ANOVA results are presented in the Sup-
also replicated the typical illusory-truth effect; repeated plemental Analyses file at https://osf.io/hw3gy, but our
statements (M .69, 95% CI [.66, .73]) were more likely key findings included an overall illusory-truth effect,
to be rated as true than new statements (M .61, 95% F(1, 77) 36.06, p .001, p2 .32: Repeated statements
CI [.58, .64]), F(1, 77) 32.37, p .001, p2 .30. In (M 3.06, 95% CI [2.99, 3.14]) were given higher truth
addition, there was no interaction between repetition ratings than new statements (M 2.85, 95% CI [2.76,
and age, F(2, 77) 0.28, p .754, p2 .01; the size of 2.92]), and there was no interaction between age and
the illusory-truth effect was similar across the three age repetition, F(2, 77) 0.89, p .413, p2 .02.
groups. (Note, however, that the study was underpow- Finally, we examined whether repetition significantly
ered to detect differences in the size of the effect increased the proportion of statements rated as true for
between age groups.) The age groups did differ in their each of the three age groups. Following the analyses
ability to distinguish between the true and false facts, outlined in our preregistration, we found a significant
F(2, 77) 30.96, p .001, p2 .45. As shown in Table illusory-truth effect for the 5-year-olds (new: M .61,
2, 10-year-olds and adults were better able to distin- repeated: M .68), t(23) 2.76, p .011, 95% CI for
guish between the true and false statements than the mean difference [.02, .14], d 0.56; 10-year-olds
5-year-olds. In addition, there was a marginal interac- (new: M .60, repeated: M .70), t(23) 3.20, p .004,
tion between repetition and statement truth, F(1, 77) 95% CI for the mean difference [.04, .16], d 0.65;
2.98, p .088, p2 .04. Probably because of ceiling and adults (new: M .62, repeated: M .70), t(31)
effects for the easier items, the illusory-truth effect was 3.96, p .001, 95% CI for the mean difference [.04,
1156 Fazio, Sherry

Table 2. Average Proportion of Statements Rated as True, Split by Age Group, Knowledge
Level, Repetition, and Statement Truth

Knowledge level 5-year-olds 10-year-olds Adults

and statement
truth New Repeated New Repeated New Repeated
Preschool
True .75 (.27) .82 (.19) .93 (.12) .98 (.07) .91 (.15) .95 (.12)
False .53 (.29) .59 (.31) .22 (.20) .35 (.26) .23 (.22) .38 (.33)
Elementary school
True .65 (.23) .71 (.28) .83 (.19) .89 (.16) .85 (.18) .95 (.14)
False .57 (.21) .68 (.23) .43 (.21) .55 (.31) .46 (.21) .46 (.25)
Middle school
True .55 (.29) .59 (.33) .65 (.23) .70 (.27) .64 (.30) .73 (.24)
False .58 (.27) .67 (.24) .54 (.26) .72 (.27) .65 (.28) .77 (.21)
M
True .65 (.19) .71 (.22) .80 (.10) .85 (.10) .80 (.14) .87 (.11)
False .56 (.19) .65 (.18) .40 (.15) .54 (.19) .45 (.15) .54 (.20)

Note: Standard deviations are given in parentheses.

.12], d 0.70. Similar results with the full scale are The three age groups did differ in their ability to
reported in the Supplemental Analyses file at https:// distinguish the true and false statements at each knowl-
osf.io/hw3gy. edge level. As reflected by interactions between age
and knowledge level, F(4, 154) 3.72, p .006, p2
Does prior knowledge protect against .09, and statement truth by age by knowledge level,
F(4, 154) 7.54, p .001, p2 .16, the 5-year-olds were
the illusory-truth effect? less able to distinguish between true and false state-
Next, we examined whether the effect of repetition ments than the 10-year-olds and adults. This was par-
depended on participants’ prior knowledge. As shown ticularly true for the preschool statements. As shown
in Table 2, repetition increased perceived truth for state- in Table 2, adults’ and 10-year-olds’ truth ratings were
ments across all three knowledge levels: preschool, ele- very similar for statements at each knowledge level.
mentary school, and middle school. This was confirmed Most importantly, repetition did not interact with
by a preregistered 2 (repetition: new, repeated) 2 knowledge level, either on its own, F(2, 154) 0.20,
(statement truth: truth, falsehood) 3 (knowledge level: p .818, p2 .00, or in conjunction with age, F(4, 154)
preschool, elementary school, middle school) 2 (age: 0.27, p .894, p2 .01. As in prior research (Fazio,
5-year-olds, 10-year-olds, adults) ANOVA on the pro- 2020; Fazio et al., 2015), prior knowledge did not pro-
portion of statements that were rated as true. Similar tect against the illusory-truth effect. Repetition affected
analyses using the full rating scale are presented in the all statements equally.
Supplemental Analyses file at https://osf.io/hw3gy. To further confirm that the effect of repetition did
First, our selection of stimuli on the basis of knowl- not vary with knowledge, we conducted a preregistered
edge level was successful. Participants had the most 2 (repetition: new, repeated) 2 (statement truth: truth,
knowledge about the preschool statements, followed falsehood) 3 (knowledge level: preschool, elementary
by the elementary school statements, and then the school, middle school) ANOVA on the proportion of
middle school statements. This pattern was reflected in statements that were rated as true within each age
an interaction between statement truth and knowledge group. The interaction between repetition and knowledge
level, F(2, 154) 112.72, p .001, p2 .59. Participants level was not significant for the 5-year-olds, F(2, 46)
were able to easily distinguish between true and false 0.09, p .915, p2 .00; the 10-year-olds, F(2, 46) 0.09,
preschool statements (true: M .89, 95% CI [.85, .93]; p .911, p2 .00; or the adults, F(2, 62) 0.59, p
false: M .39, 95% CI [.34, .43]) and could distinguish .559, p2 .02. The full results of the ANOVAs are pre-
between true and false elementary school statements sented in the Supplemental Analyses file at https://osf
(true: M .81, 95% CI [.77, .85]; false: M .53, 95% .io/hw3gy. As shown in Table 2, repetition increased
CI [.49, .57]), but true and false middle school adults’ truth judgments even for falsehoods that con-
statements were rated similarly (true: M .64, 95% tradicted preschool-level knowledge (e.g., “A wasp is
CI [.60, .69]; false: M .66, 95% CI [.62, .70]). an insect that makes honey”; new: M .23, repeated:
Effect of Repetition Across Development 1157

Table 3. Parameter Estimates for the Fluency-Conditional Model With

the Fluency Parameters Equivalent Across the Age Groups

Parameter 5-year-olds 10-year-olds Adults

F: new .17 [.10, .24] .17 [.10, .24] .17 [.10, .24]
F: repeated .33 [.27, .39] .33 [.27, .39] .33 [.27, .39]
K: preschool .31 [.17, .44] .88 [.80, .95] .80 [.72, .88]
K: elementary school .07 [ .06, .20] .50 [.38, .63] .58 [.47, .69]
K: middle school .00 [ .12, .12] .06 [ .06, .19] .00 [ .10, .10]
G .52 [.46, .58] .56 [.48, .64] .60 [.54, .67]

Note: Values shown in brackets are 95% confidence intervals. “F: new” refers to
reliance on fluency for new statements; “F: repeated” refers to reliance on fluency
for repeated statements; “K: preschool” refers to reliance on knowledge for
preschool statements; “K: elementary school” refers to reliance on knowledge for
elementary school statements; “K: middle school” refers to reliance on knowledge
for middle school statements; and “G” refers to “guess ‘true.’”

M .38), t(31) 2.51, p .018, 95% CI for the mean elementary school items, knowledge for middle school
difference [.03, .26], d 0.44. items, and a guessing parameter. In the baseline mod-
els below, we allowed all six parameters to vary across
the three age groups.
Model testing As expected, given previous research, the knowl-
Finally, we examined how well the knowledge-condi- edge-conditional model provided a poor fit for the data,
tional and fluency-conditional multinomial models fit- G2(18) 54.43, p .001. In contrast, the fluency-con-
ted the participants’ binary true/not true responses. As ditional model fit the data well, G2(18) 22.41, p .214.
a reminder, the knowledge-conditional model assumes Because the analyses above indicated that fluency may
that participants rely on fluency only when they do not affect each of the age groups similarly, we examined
have relevant prior knowledge, whereas the fluency- whether there was a decrease in model fit when we
conditional model assumes that participants sometimes constrained the fluency parameters to be equivalent
rely on fluency, even when they have relevant prior across the three age groups. The resulting model still
knowledge (Fig. 1). Previous studies have found that provided a good fit to the data, G2(22) 28.71, p .153,
adults’ responses are best fitted with a fluency-conditional and this constraint did not significantly reduce the
model (Fazio et al., 2015). model fit, G2(4) 6.30, p .178. Thus, all three age
Using multiTree software (Version 0.46; Moshagen, groups were equally reliant on fluency to make their
2010), we estimated the model parameters by minimiz- decisions. In contrast, constraining knowledge to be
ing the distance between the observed and estimated equivalent across the three age groups while allowing
response frequencies. That difference was measured the fluency and guessing parameters to vary produced
by G 2, and thus smaller G 2 values indicate a better a model that significantly decreased the model fit,
model fit. The null hypothesis is that the model fits the G2(6) 65.70, p .001, and in fact no longer fit the
data, so p values less than .05 indicate a poor fit. As data, G2(24) 88.10, p .001.
in prior work (Fazio et al., 2015), we placed theoreti- Table 3 shows the parameter estimates from the flu-
cally informed constraints on the parameters to con- ency-conditional model with the fluency parameters
serve degrees of freedom. The fluency parameter (F) equivalent across the three age groups. Each parameter
varied across new and repeated items but was con- can be interpreted as the probability that the cognitive
strained to be the same for preschool, elementary process contributes to the observed behavior. Impor-
school, and middle school facts and for truths and tantly, the parameters varied as one would expect, pro-
falsehoods. The knowledge parameter (K) varied viding further evidence for the model. First, participants
across the different knowledge levels but was con- were less likely to rely on their prior knowledge for
strained to be the same for new and repeated items middle school facts than for preschool facts. In addition,
and for truths and falsehoods. Finally, the guessing participants were more likely to rely on fluency for
parameter (G) was constrained to be equivalent for all repeated statements than for novel statements. Finally,
statements. Thus, there were six free parameters— the younger children were less likely to rely on their
fluency for new items, fluency for repeated items, prior knowledge than the older children and adults
knowledge for preschool items, knowledge for (likely because they had less prior knowledge).
1158 Fazio, Sherry

Discussion variation in the size of the illusory-truth effect across

participants (De keersmaecker et al., 2020). In addition,
Repetition increased perceived truth across develop- repetition affects belief in political false-news headlines
ment. Five-year-olds, 10-year-olds, and college students equally regardless of whether the headline supports or
were all more likely to judge repeated statements as opposes the reader’s preferred candidate (Pennycook
true than novel statements. These results suggest that et al., 2018). Similarly, in this study, we found that rep-
the connection between repetition and truth is learned etition affects belief across development and prior
at a young age. Instead of requiring a naive theory of knowledge. We conclude that the illusory-truth effect
fluency that is developed through reflection and meta- is a universal effect learned at a young age.
cognition, children learn the relation between repetition
and truth implicitly through their interactions with the Transparency
world around them. Before children are able to reflect
on their internal feeling of familiarity or fluency and Action Editor: D. Stephen Lindsay
Editor: D. Stephen Lindsay
how it relates to the truth of a statement, children use
Author Contributions
repetition as a signal for truth. L. K. Fazio developed the study concept. Data collection
These results fit with prior research suggesting that was supervised by C. L. Sherry. Analyses were conducted
people connect repetition with truth because, in gen- by L. K. Fazio. Both authors contributed to writing the
eral, statements that we hear multiple times are likely manuscript, with L. K. Fazio taking the lead role. Both
to be true. Thus, when we are presented with a situa- authors approved the final manuscript for submission.
tion in which this correlation does not hold (e.g., Declaration of Conflicting Interests
repeated items are more likely to be false), our judg- The author(s) declared that there were no conflicts of
ments change accordingly (Unkelbach, 2007). Our interest with respect to the authorship or the publication
results demonstrate that, just like adults, young children of this article.
are attentive to this connection between repetition and Open Practices
All data and materials have been made publicly available
truth.
via OSF and can be accessed at https://osf.io/hw3gy. The
We also replicated previous findings suggesting that design and analysis plans for the experiment were prereg-
prior knowledge does not protect against the illusory- istered on AsPredicted; a copy of the preregistration is
truth effect (Fazio et al., 2015; Fazio et al., 2019). Across available at https://osf.io/hw3gy. The complete Open
all three age groups, repetition increased truth judg- Practices Disclosure for this article can be found at http://
ments equally for known statements and unknown journals.sagepub.com/doi/suppl/10.1177/0956797
statements. In fact, even college students who heard 620939534. This article has received the badges for Open
false preschool-level statements (e.g., “A wasp is an Data, Open Materials, and Preregistration. More informa-
insect that makes honey”) twice were more likely to tion about the Open Practices badges can be found at
judge them as true compared with when they heard http://www.psychologicalscience.org/publications/
these statements only once. badges.
Multinomial processing-tree models supported these
two conclusions. For all age groups, the best-fitting
model was one in which participants sometimes relied
on fluency alone without consulting their prior knowl- ORCID iD
edge. Thus, repetition can affect truth judgments even Lisa K. Fazio https://orcid.org/0000-0002-0415-4862
when participants have prior knowledge. In addition,
the fluency-first model provided a good fit to the data Acknowledgments
even when the fluency values were restricted to be We thank Taylor Boothby for her help in designing the stimuli
equivalent across the three age groups. Despite large for the experiment.
developmental differences among 5-year-olds, 10-year-
olds, and adults, all three age groups relied on fluency, References
familiarity, or cohesion similarly to make their truth
judgments. Alter, A. L., & Oppenheimer, D. M. (2006). Predicting short-
term stock fluctuations by using processing fluency.
These findings fit with prior evidence that the illu-
Proceedings of the National Academy of Sciences, USA,
sory-truth effect is universal—occurring for all people 103, 9369–9372.
and all stimuli. For example, a number of individual- Arkes, H. R., Hackett, C., & Boehm, L. (1989). The generality
difference measures—including fluid and verbal intel- of the relation between familiarity and judged validity.
ligence, need for closure, experiential and rational Journal of Behavioral Decision Making, 2, 81–94.
processing styles, preference for intuition and delibera- Aslin, R. N. (2017). Statistical learning: A powerful mechanism
tion, and cognitive reflection—all fail to explain that operates by mere exposure. Wiley Interdisciplinary
Effect of Repetition Across Development 1159

Reviews: Cognitive Science, 8(1–2), Article e1373. doi:10.1002/ Koenig, M. A., & Echols, C. H. (2003). Infants’ understanding
wcs.1373 of false labeling events: The referential roles of words
Bernard, S., Proust, J., & Clément, F. (2014). The medium and the speakers who use them. Cognition, 87, 179–208.
helps the message: Early sensitivity to auditory fluency Koriat, A., Ackerman, R., Lockl, K., & Schneider, W. (2009a).
in children’s endorsement of statements. Frontiers in The easily learned, easily remembered heuristic in chil-
Psychology, 5, Article 1412. doi:10.3389/fpsyg.2014.01412 dren. Cognitive Development, 24, 169–182.
Dechêne, A., Stahl, C., Hansen, J., & Wänke, M. (2010). The Koriat, A., Ackerman, R., Lockl, K., & Schneider, W. (2009b).
truth about the truth: A meta-analytic review of the truth The memorizing effort heuristic in judgments of learning:
effect. Personality and Social Psychology Review, 14, 238– A developmental perspective. Journal of Experimental
257. doi:10.1177/1088868309352251 Child Psychology, 102, 265–279.
De keersmaecker, J., Dunning, D., Pennycook, G., Rand, Lane, J. D., & Harris, P. L. (2015). The roles of intuition and
D. G., Sanchez, C., Unkelbach, C., & Roets, A. (2020). informants’ expertise in children’s epistemic trust. Child
Investigating the robustness of the illusory truth effect Development, 86, 919–926.
across individual differences in cognitive ability, need Lopez-Mobilia, G., & Woolley, J. D. (2016). Interactions
for cognitive closure, and cognitive style. Personality and between knowledge and testimony in children’s reality-
Social Psychology Bulletin, 46, 204–215. doi:10.1177/ status judgments. Journal of Cognition and Development,
0146167219853844 17, 486–504.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). McGlone, M. S., & Tofighbakhsh, J. (2000). Birds of a feather
G*Power 3: A flexible statistical power analysis program flock conjointly (?): Rhyme as reason in aphorisms.
for the social, behavioral, and biomedical sciences. Psychological Science, 11, 424–428.
Behavior Research Methods, 39, 175–191. Moshagen, M. (2010). multiTree: A computer program for the
Fazio, L. K. (2020). Repetition increases perceived truth even analysis of multinomial processing tree models. Behavior
for known falsehoods. Collabra: Psychology, 6(1), Article Research Methods, 42, 42–54.
38. doi:10.1525/collabra.347 Pennycook, G., Cannon, T. D., & Rand, D. G. (2018). Prior
Fazio, L. K., Brashier, N. M., Payne, B. K., & Marsh, E. J. exposure increases perceived accuracy of fake news.
(2015). Knowledge does not protect against illusory truth. Journal of Experimental Psychology: General, 147, 1865–
Journal of Experimental Psychology: General, 144, 993– 1880. doi:10.1037/xge0000465
1002. doi:10.1037/xge0000098 Reber, R., & Schwarz, N. (1999). Effects of perceptual fluency
Fazio, L. K., Rand, D. G., & Pennycook, G. (2019). Repetition on judgments of truth. Consciousness and Cognition, 8,
increases perceived truth equally for plausible and 338–342.
implausible statements. Psychonomic Bulletin & Review, Schneider, W. (2008). The development of metacognitive
26, 1705–1710. doi:10.3758/s13423-019-01651-4 knowledge in children and adolescents: Major trends and
Feder, C. W., & Bishay, S. (2012). Brain quest grade 3. New implications for education. Mind, Brain, and Education,
York, NY: Workman Publishing. 2, 114–121. doi:10.1111/j.1751-228X.2008.00041.x
Feder, C. W., & Bishay, S. (2016a). Brain quest grade 7. New Schwarz, N. (2010). Meaning in context: Metacognitive experi-
York, NY: Workman Publishing. ences. In B. Mesquita, L. F. Barrett, & E. R. Smith (Eds.), The
Feder, C. W., & Bishay, S. (2016b). Brain quest kindergarten. mind in context (pp. 105–125). New York, NY: Guilford Press.
New York, NY: Workman Publishing. Silva, R. R., Garcia-Marques, T., & Mello, J. (2016). The dif-
Feder, C. W., & Bishay, S. (2016c). Brain quest preschool. New ferential effects of fluency due to repetition and fluency
York, NY: Workman Publishing. due to color contrast on judgments of truth. Psychological
Frost, R., Armstrong, B. C., Siegelman, N., & Christiansen, M. H. Research, 80, 821–837.
(2015). Domain generality versus modality specificity: Stryker, S. (2014). 79 insane facts that will change the way you
The paradox of statistical learning. Trends in Cognitive look at the animal kingdom. BuzzFeed. Retrieved from
Sciences, 19, 117–125. https://www.buzzfeed.com/samstryker/insane-facts-that-
Hasher, L., Goldstein, D., & Toppino, T. (1977). Frequency will-change-the-way-you-look-at-the-ani
and the conference of referential validity. Journal of Tauber, S. K., Dunlosky, J., Rawson, K. A., Rhodes, M. G.,
Verbal Learning and Verbal Behavior, 16, 107–112. & Sitzman, D. M. (2013). General knowledge norms:
Jacoby, L. L., Kelley, C. M., Brown, J., & Jasechko, J. (1989). Updated and expanded from the Nelson and Narens
Becoming famous overnight: Limits on the ability to avoid (1980) norms. Behavior Research Methods, 45, 1115–1143.
unconscious influences of the past. Journal of Personality Thiessen, E. D., Kronstein, A. T., & Hufnagle, D. G. (2013).
and Social Psychology, 56, 326–338. The extraction and integration framework: A two-process
Jacoby, L. L., Kelley, C. M., & Dywan, J. (1989). Memory account of statistical learning. Psychological Bulletin, 139,
attributions. In H. L. Roediger, III, & F. I. M. Craik (Eds.), 792–814.
Varieties of memory and consciousness: Essays in honour Tversky, A., & Kahneman, D. (1973). Availability: A heu-
of Endel Tulving (pp. 391–422). Mahwah, NJ: Erlbaum. ristic for judging frequency and probability. Cognitive
Koenig, M. A., Clément, F., & Harris, P. L. (2004). Trust in Psychology, 5, 207–232.
testimony: Children’s use of true and false statements. Unkelbach, C. (2007). Reversing the truth effect: Learning the
Psychological Science, 15, 694–698. interpretation of processing fluency in judgments of truth.
1160 Fazio, Sherry

Journal of Experimental Psychology: Learning, Memory, Current Directions in Psychological Science, 28, 247–253.
and Cognition, 33, 219–230. doi:10.1177/0963721419827854
Unkelbach, C., & Greifeneder, R. (2013). A general model Unkelbach, C., & Rom, S. C. (2017). A referential theory
of fluency effects in judgment and decision making. In of the repetition-induced truth effect. Cognition, 160,
C. Unkelbach & R. Greifender (Eds.), The experience 110–126.
of thinking: How the fluency of mental processes influ- Weil, L. G., Fleming, S. M., Dumontheil, I., Kilford, E. J.,
ences cognition and behavior (p. 11–32). New York, NY: Weil, R. S., Rees, G., . . . Blakemore, S.-J. (2013). The
Psychology Press. development of metacognitive ability in adolescence.
Unkelbach, C., Koch, A., Silva, R. R., & Garcia-Marques, T. Consciousness and Cognition, 22, 264–271. doi:10.1016/j
(2019). Truth by repetition: Explanations and implications. .concog.2013.01.004