Insufficiently Complimentary?: Underestimating the Positive Impact of Compliments Creates a Barrier to Expressing Them

Abstract

Compliments increase the well-being of both expressers and recipients, yet people report in a series of surveys giving fewer compliments than they should give, or would like to give. Nine experiments suggest that a reluctance to express genuine compliments partly stems from underestimating the positive impact that compliments will have on recipients. Participants wrote genuine compliments and then predicted how happy and awkward those compliments would make recipients feel. Expressers consistently underestimated how positive the recipients would feel but overestimated how awkward recipients would feel (Experiments 1-3, S4). These miscalibrated expectations are driven partly by perspective gaps in which expressers underestimate how competent—and to a lesser extent how warm—their compliments will be perceived by recipients (Experiments 1-3). Because people’s interest in expressing a compliment is partly driven by their expectations of the recipient’s reaction, undervaluing a compliment creates a barrier to expressing them (Supplemental Experiments S2, S3, S4). As a result, directing people to focus on the warmth conveyed by their compliment (Experiment 4) increased interest in expressing it. We believe these findings may reflect a more general tendency for people to underestimate the positive impact of prosocial actions on others, leading people to be less prosocial than would be optimal for both their own and others’ well-being.

Full text

In press, Journal of Personality and Social Psychology
Insufficiently Complimentary?: Underestimating the Positive Impact
of Compliments Creates a Barrier to Expressing Them
Xuan Zhao
Stanford University, Department of Psychology
Nicholas Epley
University of Chicago, Booth School of Business
Correspondence author: Xuan Zhao (xuanzhao@stanford.edu)
Author note: We thank Bryan Baird, Yvita Bustos, Alice Fried, Amanda Hopcroft, Anita Joshi,
Yena Kim, Jasper Lee, Donald Lyons, Jake Martin, Robyn Myers, John Pezzuto, Ruth Selipsky,
Nancy Shrestha, Michelle Wang, Qing Wang, Xueyan Xie, and Zhi Zhang for assistance
conducting experiments. We are grateful to James Dungan, Michael Kardas, Nadav Klein, and
Amit Kumar for feedback during early stages of this research, and to staff at the Garfield Park
Conservatory for coordinating data collection. This research was funded by the Neubauer Family
Faculty Fellowship and the University of Chicago Booth School of Business.
All data, materials, and preregistration forms can be accessed via a repository at the Open
Science Framework: https://osf.io/ypk5g/
© 2021, American Psychological Association. This paper is not the copy of record and may
not exactly replicate the final, authoritative version of the article. Please do not copy or cite
without authors' permission. The final article will be available, upon publication, via its
DOI: 10.1037/pspa0000277

INSUFFICIENTLY COMPLIMENTARY
!
!
2!
Abstract
Compliments increase the well-being of both expressers and recipients, yet people report in a
series of surveys giving fewer compliments than they should give, or would like to give. Nine
experiments suggest that a reluctance to express genuine compliments partly stems from
underestimating the positive impact that compliments will have on recipients. Participants wrote
genuine compliments and then predicted how happy and awkward those compliments would
make recipients feel. Expressers consistently underestimated how positive the recipients would
feel but overestimated how awkward recipients would feel (Experiments 1-3, S4). These
miscalibrated expectations are driven partly by perspective gaps in which expressers
underestimate how competent—and to a lesser extent how warm—their compliments will be
perceived by recipients (Experiments 1-3). Because people’s interest in expressing a
compliment is partly driven by their expectations of the recipient’s reaction, undervaluing a
compliment creates a barrier to expressing them (Supplemental Experiments S2, S3, S4). As a
result, directing people to focus on the warmth conveyed by their compliment (Experiment 4)
increased interest in expressing it. We believe these findings may reflect a more general
tendency for people to underestimate the positive impact of prosocial actions on others, leading
people to be less prosocial than would be optimal for both their own and others’ well-being.
Keywords: social connection, compliments, happiness, well-being, prosocial behavior

INSUFFICIENTLY COMPLIMENTARY
!
!
3!
Insufficiently Complimentary?: Underestimating the Positive Impact
of Compliments Creates a Barrier to Expressing Them
“—and I will remark here that the happy phrasing of a compliment is one of the
rarest of human gifts and the happy delivery of it another.”
- Mark Twain (1907), p. 130
Thinking and speaking are guided by different psychological processes (Goldin-Meadow
et al., 2008; Pinker, 2007; Slobin, 1996), meaning that Twain was right to distinguish between
generating a compliment and delivering it. Thinking well of oneself is also among the more
satisfying thoughts a person can have (Baumeister & Leary, 1995; Diener & Diener, 1995;
DeNeve & Cooper, 1998; Jones, 1973; Leary, 2007), meaning that Twain was also right to note
the happiness that can quickly follow from receiving a nicely-worded compliment (Fea &
Brannon, 2006; Izuma et al., 2008; Swann et al., 1987). However, here we suggest that happily
phrased compliments are actually quite common, but their delivery is rarer than would be
optimal to maximize both one’s own and others’ well-being. Specifically, we hypothesize that
people systematically underestimate how positively a compliment will be received, which in turn
creates a psychological barrier that keeps people from delivering some of the compliments they
generate about others. We predict that people would be more interested in expressing the
compliments that come to their mind if they accurately understood just how positive their
compliments made others feel.
Insufficiently Complimentary?
Giving compliments is an important part of everyday conversation, drawing both
strangers and friends closer to each other through the warmth conveyed in kind words (Knapp et
al., 1984; Rees-Miller, 2011). Giving compliments is also a readily available way to increase

INSUFFICIENTLY COMPLIMENTARY
!
!
4!
both one’s own and another person’s mood and strengthen a relational bond compared to the
effort and expense of other well-documented mood-brightening behaviors such as spending
money on others (Izuma et al., 2008; Dunn et al., 2014), writing a gratitude letter (Chaudhry &
Loewenstein, 2019; Kumar & Epley, 2018; Lambert et al., 2010), or performing random acts of
kindness (Curry et al., 2018). A compliment can come to mind almost effortlessly, take only a
moment to deliver, come at no financial cost, and be expressed to anyone ranging from a
complete stranger to a spouse. Compliments are likely to be highly valued by recipients because
they affirm a recipient’s positive self-view (Kwang & Swann, 2010; Vonk, 2002) and signal
warmth and acceptance from another person (Leary & Baumeister, 2000; Marigold et al., 2007).
These facts raise an important empirical question: Do people express the compliments
that come to their minds as often as they should to maximize both their own and others’ well-
being? Anecdotal evidence suggests that people sometimes withhold genuine compliments that
they could easily express (e.g., Chapman, 1995). Indeed, a series of four surveys we conducted
suggest that withholding compliments might be somewhat widespread even though people
recognize that giving more compliments could be desirable. In these surveys (Mage = 35.92,
SDage = 10.96, 44% female), we asked online participants in a series of survey to rate how often
they engaged in a variety of prosocial activities, including expressing compliments to a person
they felt close to, or with whom they had a satisfying or dissatisfying relationship (N = 96, 97,
100, 193, See Supplemental Materials, Surveys 1-4). Participants reported how often they
engaged in each prosocial behavior compared to how often they thought they should, or would
like to, on scales ranging from -3 (a lot less often than I think I should [would like to]) to 3 (a lot
more often than I think I should [would like to]). Regardless of how often they reported
communicating with the target, or how satisfied they were with their relationship, participants

INSUFFICIENTLY COMPLIMENTARY
!
!
5!
consistently reported expressing compliments significantly less often than they thought they
should (ps < .001), or would like to (p = .003). This self-assessed deficit was larger for
expressing compliments than for other prosocial behaviors, including expressing gratitude,
providing emotional support, giving advice, and providing helpful criticism (ps < .05; see Survey
1). In a final survey (N = 103, Mage = 36.52, SDage = 10.58, 31% female; U.S. residents), we
asked participants to report how often they expressed versus withheld criticisms to specific close
others, as well as how often they expressed versus withheld compliments. People reported
withholding criticisms 62.2% of the time, on average, but also reported withholding compliments
36.4% of the time. It is perhaps not surprising that people keep their negative thoughts about
others to themselves, but these results suggest that people may be keeping a meaningful number
of their positive thoughts to themselves as well.
Is whatever reluctance people feel about expressing a compliment that comes to their
mind fully warranted?
Miscalibrated Expectations as a Barrier to Prosociality?
Deliberate choices tend to be guided at least in part by the expected value of an action
(Behrens et al., 2009; Bentham, 1825; Decker et al., 2016). As the economist Gary Becker noted
in his Nobel Prize address (1993), “The [rational] analysis assumes that individuals maximize
welfare as they conceive it,” such that, “their behavior is forward-looking.” Conceptions of
reality, however, are not to be confused with reality itself. Psychologists have documented many
ways in which misconceptions of reality can undermine a perfectly rational person’s ability to
maximize his or her own well-being (Kahneman et al., 1997; Read, 2007; Wilson & Gilbert,
2005).

INSUFFICIENTLY COMPLIMENTARY
!
!
6!
We presume that decisions to express compliments can also be guided by an assessment
of expected value. Because a prosocial act is intended to positively impact another person, its
expected value is likely to come from its presumed positive impact on the recipient, which
subsequently guides people’s decisions on interpersonal actions. Underestimating the positive
impact of expressing compliments that come to mind could therefore serve as a barrier to
expressing them more often in everyday life, thereby diminishing the well-being of both
expressers and recipients. Some existing research suggests that people may indeed undervalue
the positive consequences of connecting with others, such as how much they will enjoy talking
with strangers (Epley & Schroeder, 2014), how positive they will feel spending money on others
compared to themselves (Dunn et al., 2008), how much others will like them after a conversation
(Boothby et al., 2018), and how much others will enjoy talking about shared (versus
extraordinary but unshared) experiences (Cooney et al., 2014).
More closely related to our current hypotheses, one series of recent experiments found
that those expressing gratitude tended to underestimate how positive their recipient would feel
and overestimate how awkward the recipient would feel, creating a psychological barrier to
expressing gratitude in close relationships (Kumar & Epley, 2018). Like expressing gratitude,
giving compliments can also be a prosocial action in which people share positive thoughts with
another person, meaning that it may encounter a similar psychological barrier when people
consider giving a compliment. Unlike expressing gratitude, giving compliments is a much
broader prosocial act that can focus on any positive attribute of the recipient, while gratitude
typically involves giving thanks to another person for some meaningful benefit received by the
expresser. Gratitude can therefore communicate evaluations of credit and responsibility
(Chaudhry & Loewenstein, 2019), which simple compliments typically do not. Somewhat

INSUFFICIENTLY COMPLIMENTARY
!
!
7!
surprisingly, while expressing gratitude has been commonly presumed to be a prosocial behavior
in the literature, giving compliments have typically been studied as self-interested actions meant
to enhance one’s own image in the eyes of others through strategic self-presentation and
ingratiation (e.g., Gordon, 1996; Jones, 1964; Seiter, 2007; Stern & Westphal, 2010; Vonk, 2001,
2002). Studying compliments allows us to understand whether there is a consistent set of
psychological barriers that might inhibit prosociality across a wide range of behaviors, or if the
barriers vary across meaningfully different prosocial acts.
One research program developed concurrently with our own suggests that barriers to
prosociality might indeed be more widespread. Specifically, Boothby & Bohns (2020) reported
three experiments in which university students were asked to compliment a stranger on campus
with either a scripted or unscripted compliment, and they found that participants tended to
underestimate how positive and overestimate how negative their recipients would feel. Our
research extends these results in at least five ways. First, our experiments focus on genuine and
meaningful compliments expressed between people in existing relationships, where
miscalibration may be somewhat unexpected, rather than between random strangers where
compliments were mostly focused on visible physical characteristics (e.g., clothes and
accessories). Second, we test the robustness and reliability of our effects using diverse samples
of participants recruited primarily from community populations, rather than relying primarily on
university samples. Third, we test the impact of receiving compliments against a no-compliment
control condition in order to test whether people simply underestimate how positively (and
overestimate how negatively) others feel at baseline, or if they misunderstand the impact of
compliments, in particular. Fourth, we test a comprehensive explanation in multiple experiments
for why people might undervalue compliments, and we compare this account against plausible

INSUFFICIENTLY COMPLIMENTARY
!
!
8!
alternative interpretations. Finally, we test in four experiments how people’s expectations are
related to their interest in actually expressing a compliment, including two experiments testing
how calibrating expectations of a recipient’s reaction influences people’s interest in expressing
compliments.
Specifically, we predict that people undervalue the positive impact of compliments on
recipients because expressers and recipients evaluate distinct aspects of the same compliment
differently in a way that corresponds to their own unique perspective on it. In particular, actors
tend to focus on competence when evaluating their own interpersonal actions while observers
tend to focus relatively more on the actor’s warmth (Abele & Wojciszke, 2007; Bruk et al., 2018;
Fiske et al., 2007; Wojciszke et al., 1998). Those expressing a compliment may therefore be
concerned about the specific words or phrases they are using to express their feelings while the
recipients are attending more to the prosocial meaning and intent of the feelings being expressed.
This could lead expressers to especially underestimate how positively their compliment will be
evaluated in terms of its competence, compared to its warmth, consistent with prior research
studying expressions of gratitude (Kumar & Epley, 2018). Because compliments convey a great
deal of warmth to the recipient, this difference would lead expressers to underestimate the
positive impact (and overestimate the negative impact) of a compliment on their recipients. This
perspective gap also predicts that shifting an expresser’s attention to the warmth conveyed
through their compliment would lead to more calibrated expectations about the recipient’s
reaction.
In addition, the positive impact of compliments may be easy to undervalue because they
are specifically directed to a unique recipient, meaning that the compliments are likely to be
uniquely valued by the recipient in a way that is difficult for the expresser, or third-party

INSUFFICIENTLY COMPLIMENTARY
!
!
9!
observers, to fully appreciate (Van Boven et al., 2013). This predicts that even third-person
observers (who are not receiving an actual compliment directed at them) will also underestimate
how warm and competent the compliments will be perceived by recipients, and will therefore
also underestimate how positive compliment recipients will feel (c.f., Boothby & Bohns, 2020).
We test this hypothesis in two supplemental experiments (Experiments S1a & S1b) that we
discuss following Experiment 1.
Beyond explaining interpersonal misunderstanding, our theory that mistaken expectations
create a barrier to expressing compliments also suggests that the more people expect a
compliment will be positively received, the more interested they will be in expressing it.
Aligning expressers’ expectations with recipients’ actual experiences should therefore increase
their willingness to express compliments to their potential recipients. Testing this hypothesis is
critical for understanding the causal role of miscalibrated expectations in people’s interest in
expressing versus withhold information in social relationships.
Overview of Experiments
In a series of nine experiments, we tested our hypotheses that people tend to
underestimate the positive impact that compliments will have on recipients (Experiments 1–3),
that this miscalibration is guided by a perspective gap in evaluations of competence versus
warmth between expressers and recipients (Experiments 1–3, Supplemental Experiments S1a &
S1b), and that people’s expectations guide decisions to express or withhold compliments
(Experiment 4, Supplemental Experiments S2, S3, S4). We preregistered all experiments except
Supplemental Experiment S1b (due to an oversight). We report all methods and measures
completely in each experiment, including target sample sizes and whether any data were
excluded from analyses. All pre-registered analyses are reported in the Supplemental Materials.

INSUFFICIENTLY COMPLIMENTARY
!
!
10!
We note and explain more significant deviations in the primary analyses in the text where
appropriate. All experimental materials, data, and preregistration forms can be accessed online:
https://osf.io/ypk5g/?view_only=47a263380604485a803db70007930d42.
Experiment 1: Miscalibrated Expectations?
We recruited pairs of people visiting a public park and randomly assigned half to a
compliment condition and the other half to a control condition. In the compliment condition, one
member of each pair (the expresser) wrote three compliments and then predicted how the other
member (the recipient) would feel after reading the compliments, thereby allowing us to test our
main hypothesis that expressers would significantly underestimate the positive impact (and
overestimate the negative impact) of expressing compliments to another person. One member of
each pair in the control condition (the predictor) simply predicted how the other member (the
target) would report feeling on a survey, thereby allowing us to test an alternative interpretation
that people simply underestimate how positive others feel in general (Ong, Goodman, & Zaki,
2018). Expressers and recipients in the compliment condition also evaluated their compliments
in terms of warmth and competence, to test whether differences in perceptions of these attributes
could at least partly explain expressers’ systematic misunderstanding of recipients’ experiences.
Method
Participants. We stationed experimenters behind a table with a poster advertising an
“Interpersonal Relationship Study” in a public park located in an ethnically diverse
neighborhood in a large U.S. city. The experimenters recruited pairs visiting the park together.
Participants received a small novelty gift in exchange for their participation. We targeted a
sample size at 200 participants (50 pairs per condition) and recruited through the end of our last
scheduled shift as we approached that target. A total of 210 participants (Mage = 36.84, SDage =

INSUFFICIENTLY COMPLIMENTARY
!
!
11!
14.38; rangeage = 19–91; 62% female) completed the experiment. One pair did not write three
compliments as instructed, yielding 52 pairs in each condition.
The experimental survey asked one participant in each pair to report their relationship
type and received responses from all but two participants. Our pairs consisted of friends (N =
30), family (N = 10), romantic couples (N = 35), and married/engaged couples (N = 27) who had
had known each other for an average of 10.78 years (ranging from 2 weeks to 76 years; SD =
13.79 years).
Procedure. Experiment 1 employed a 2 (condition: compliment vs. control) × 2
(perspective: expected vs. actual) between-participants design. Pairs of participants were
randomly assigned to either the compliment condition or the control condition. In the
compliment condition, one person was then assigned to be the compliment expresser and the
other to be the compliment recipient. In the control condition, one person was assigned to be the
predictor and the other to be the target. In both conditions, these roles were referred to only as
“Person A” and “Person B” in the materials and instructions.
Compliment condition. The experimenter first gave the expresser a tablet to begin the
study in private while instructing the recipient to wait outside the expresser’s sight until the
expresser completed his/her task. Expressers then read that they were to write down three
compliments they could give to their partner, focusing on “positive things you have noticed but
have not, for whatever reason, had a chance to compliment your partner on yet,” and learned
that their partner would later read their compliments before filling out a short survey. Expressers
then wrote their three compliments. When finished, expressers predicted how their study partner
would report feeling after reading their compliments: how positive/negative their partner would
feel, how pleasant their partner would feel, and how awkward their partner would feel. The next

INSUFFICIENTLY COMPLIMENTARY
!
!
12!
four items measured expressers’ expectations of how their compliments would be perceived by
recipients in terms of warmth and competence (based on items from Fiske et al., 2007). Two
items measured perceived warmth: “How sincere will your study partner perceive your
compliments to be?,” and “How warm will your study partner perceive your compliments to
be?” Two items measured perceived competence (Kumar & Epley, 2018): “How articulate will
your study partner perceive your compliments to be?,” and “To what extent will your partner
consider that you were able to provide compliments using words that were ‘just right’?” Each
item was presented on a separate page with a scale of 0 (not at all) to 10 (extremely), except for
the positive/negative item, which had a scale of -5 (much more negative than normal) to 5 (much
more positive than normal) with 0 (no different than normal) as the midpoint.
Finally, expressers reported their current relationship quality with the recipients on two
bipolar scales, one measuring how close they felt to their partner on a scale ranging from -5
(feels like we’re miles apart) to 5 (feels like we’re really close), and another measuring how
satisfied they were with their relationship on a scale ranging from -5 (extremely dissatisfied) to 5
(extremely satisfied). Expressers then reported how often they gave compliments to, and received
compliments from, the recipients on two 7-point scales ranging from -3 (a lot less often than I
think I should) to 3 (a lot more often than I think I should), with 0 (exactly as often as I think I
should) as the midpoint. Finally, expressers reported how frequently they communicated with
each other (from a few times per day to once or twice per month), their gender, and their age.
Once the expresser finished, the experimenter retrieved the recipient who was then seated
in private and shown their expresser’s three compliments on a tablet. Recipients then provided
responses on all of the same items predicted by the expresser but rephrased for the recipient’s

INSUFFICIENTLY COMPLIMENTARY
!
!
13!
perspective. Finally, they responded to the same additional survey items described above, and
described their relationship type and length with the expresser.
Once the recipient finished, each pair was reunited, debriefed, thanked, and dismissed.
Control condition. The control condition procedure was similar to the compliment
condition except that no compliments were exchanged, and hence one person (the predictor) in
each pair was simply predicting how the other person (the target) would report feeling at that
moment. Because predictors in the control condition did not give any compliments, they also did
not answer the four items evaluating their compliments in terms of perceived warmth and
competence.
Results
Positive Mood. To test whether expressers underestimated the positive impact of their
compliments, we first created a composite score of expected and actual positive mood by adding
5 to the negative/positive mood item and then averaged it together with the pleasantness item
(rexpected = .65; ractual = .62; ps < .001).
1
A 2 (condition: compliment, control) × 2 (perspective:
expected vs. actual) repeated measures ANOVA with perspective as a repeated measure yielded
a significant main effect of condition, F(1,102) = 7.11, p = .009, η2p = .065, and perspective,
F(1,102) = 12.67, p < .001, η2p = .11, qualified by a marginally significant interaction between
condition and perspective, F(1,102) = 3.64, p = .059, η2p = .034.
2
As predicted, compliment
!
1
Before conducting Experiment 1, we preregistered combining pleasantness and awkwardness into a
composite index, but results of this and all subsequent experiments indicated that pleasantness was strongly
correlated with positive mood, but was not meaningfully correlated with awkwardness (e.g., Experiment 1:
rexpresser = -.08, p = .43; rrecipient = -.24, p = .015). We therefore report the positive mood composite including
pleasantness and analyzed awkwardness separately for all experiments. We preregistered this analysis plan in
subsequent experiments once this consistent pattern became clear to us.
2
!Fitting a linear mixed model with condition and perspective as fixed effects and participant pairs as random
effects yielded the same conclusions as the pre-registered repeated measures ANOVA we report here.
Subsequent experiments preregistered linear mixed models for these analyzes, but we report the simpler
ANOVA results throughout the main text for ease of presentation and because they yield the same conclusions.!

INSUFFICIENTLY COMPLIMENTARY
!
!
14!
expressers significantly underestimated how positive their recipients would feel (Ms = 7.79 vs.
8.81, respectively), F(1,51) = 20.19, p < .001, but predictors in the control condition did not
underestimate how positive their target would feel (Ms = 7.50 vs. 7.81, respectively), F(1,51) =
1.08, p = .30 (see Figure 1A). This pattern suggests people do not simply underestimate others’
mood in general, as might be expected if predictions of others’ experiences were simply
regressive (Moore & Healy, 2008), but that people instead underestimate their positive response
to receiving compliments, in particular. Indeed, expressers did not expect their recipients to feel
more positive than predictors expected their targets in the control condition to feel, F(1,102)
= .80, p = .37, even though reading compliments indeed made recipients feel significantly more
positive than targets in the control condition, F(1,102) = 12.18, p = .001.
Figure 1. Expected and actual positive mood (Panel A) and awkwardness (Panel B) in the
compliment and control conditions in Experiment 1. Error bars reflect ± 1 standard errors.
Awkwardness. To test whether expressers also overestimate a potential negative cost of
expressing compliments, we conducted the same analysis on evaluations of awkwardness. This
yielded a significant main effect of perspective, F(1,102) = 8.00, p = .006, η2p = .073, qualified
5
6
7
8
9
10
Compliment Control
Positive Mood
Expected Actual
0
1
2
3
4
5
Compliment Control
Awkwardness
Expected Actual
(A)
(B)

INSUFFICIENTLY COMPLIMENTARY
!
!
15!
by a marginally significant interaction, F(1,102) = 3.48, p = .065, η2p = .033. As predicted,
expressers overestimated how awkward their recipient would feel (Ms = 3.46 vs. 2.06,
respectively), F(1,51) = 8.97, p = .004, but predictors in the control condition did not (Ms = 2.33
vs. 2.04, respectively), F(1,51) = .60, p = .44 (see Figure 1B). Expressers expected that their
compliments would make their recipients feel more awkward than predictors expected their
targets to feel, F(1,102) = 4.71, p = .032. In reality, reading compliments did not make
recipients feel any more awkward than targets in the control condition, F(1,102) = .002, p = .97.
Warmth and competence. Finally, to test whether miscalibrated expectations of
recipients’ reactions could stem from differences in evaluations of competence versus warmth of
the compliments, we averaged the two warmth items (r = .51, p < .001) and two competence
items (r = .69, p < .001) to create composites for both expressers and recipients. A 2 (attribute:
warmth vs. competence) × 2 (perspective: expresser vs. recipient) repeated measures ANOVA
revealed a significant main effect for attribute, F(1,51) = 79.89, p < .001, η2p = .61, and a
significant main effect of perspective, F(1,51) = 44.66, p < .001, η2p = .47, qualified by a
significant interaction, F(1,51) = 11.18, p = .002, η2p = .18. As shown in Figure 2, expressers
underestimated the perceived competence of their compliments (expresser: M = 6.62, SD = 1.55;
recipient: M = 8.59, SD = 1.70; F(1,51) = 42.44, p < .001, η2p = .45) more than their perceived
warmth (expresser: M = 8.12, SD = 1.35; recipient: M = 9.16, SD = 1.17; F(1,51) = 22.36, p
< .001, η2p = .30). These results are consistent with prior research on more lengthy and detailed
expressions of gratitude (Kumar & Epley, 2018), and the competence result is consistent with
concurrently conducted research investigating compliments to strangers (Boothby & Bohns,
2020). This suggests a potentially robust tendency for those who express kind thoughts to be

INSUFFICIENTLY COMPLIMENTARY
!
!
16!
overly self-critical about how well they are expressing their feelings by getting the words “just
right” when anticipating how positive their compliments might make another person feel.
Indeed, expressers’ expectations of how warm and how competent their compliments
would be perceived to be were strongly correlated with their expectations on how positive their
recipient would feel (rs = .62 and .57, ps < .001). Recipients’ evaluations of warmth and
competence were strongly correlated with their positive mood as well (rs = .76 and .62, ps
< .001). Interestingly, awkwardness was not correlated with either the perceived warmth or
competence conveyed in the compliment from either perspective, perhaps suggesting that both
expected and experienced awkwardness come more from the delivery of the compliment rather
than about attributes conveyed by the compliment itself (ps > .18).
Figure 2. Expected (i.e., expresser) and actual (i.e., recipient) ratings of how warm and
competent the compliments were perceived to be in Experiment 1. Error bars reflect ± 1
standard errors.
Mediation analysis. An exploratory mediation analysis suggests that miscalibrated
expectations of competence and warmth could account for a statistically significant proportion of
variance in the discrepancy between expressers’ expected and recipients’ reported positive
5
6
7
8
9
10
Warmth Competence
Ratings
Expresser Recipient

INSUFFICIENTLY COMPLIMENTARY
!
!
17!
mood. Because expressers and recipients were nested in pairs, we employed a within-subjects
mediation analysis using MEMORE macro in SPSS (Montoya & Hayes, 2017) and performed a
5,000 resampling bootstrap with warmth and competence as simultaneous mediators. As shown
in Figure 3, expressers underestimated how warm and how competent their compliments would
be perceived by their own recipient (bs = 1.96 and 1.04, respectively, ps < .001), both of which
were significantly correlated with their underestimation of their recipient’s positive mood
(competence: b = .36, p < .001; warmth: b = .36, p = .004). Although competence yielded a
directly larger mediation effect than warmth (.71 and .38, 95% CIs = [.24, 1.19] and [0, .99],
respectively), these effect sizes did not differ significantly from each other (95% CI = [-1.09,
.62]).
Figure 3. Mediational analysis testing the relationship between miscalibrated expectations of
perceived competence and warmth of the compliments and miscalibrated expectations of
recipients’ positive moods in Experiment 1.
Compliment frequency. Replicating results in the introductory surveys, participants
again reported giving compliments less often than they thought they should (M = -.76, SD =
1.34), t(207) = -8.23, p < .001, d = .57 (see Supplemental Materials).
Perspective
(0 = expresser;
1 = recipient)
underestimating
Warmth
underestimating
Competence
underestimating
Positive Mood
.38 [0, .99]
1.04***
(c’ = -.07, p= .77)
c= 1.02, p< .001
1.96***
.36**
.36***
.71 [.24, 1.19]

INSUFFICIENTLY COMPLIMENTARY
!
!
18!
Third-party observers
Because compliments are kind words specifically directed to a recipient, they are likely to
be perceived as uniquely warm and competent from the recipient’s perspective. If so, then not
only would those expressing compliments underestimate their value to recipients, but so would
third-party observers who simply read the compliments and anticipated how an actual recipient
would feel. We tested this hypothesis in two additional experiments whose details are described
in full in the Supplemental Materials (Experiments S1a and S1b).
Specifically, we recruited new participants from the same location, at the same time of
day and location as in Experiment 1, and then yoked them to pairs from Experiment 1. These
observers read the three compliments written by their yoked expresser and predicted the
recipient’s reactions either from their own perspective (Experiment S1a), or after imagining
themselves as the expresser or the recipient (Experiment S1b). Results indicated that third-party
observers in both experiments, regardless of the perspective they were asked to take,
underestimated how positive the actual recipients would feel to the same degree as the actual
expressers. Third-party observers also underestimated how competent and how warm the
compliments would be perceived. Exploratory mediation analysis further showed that
underestimating the perceived warmth of those compliments—but not the perceived
competence—significantly mediated observers’ tendency to underestimate recipients’ positive
experience.
Interestingly, these results among third-party observers differ from those recently
reported by Boothby & Bohns (2020, Study 4), where participants who imagined someone else
receiving a compliment from a stranger (e.g., “I like your shirt”) anticipated the recipient to
feeling just as positive, on average, as the actual recipients and were thus more calibrated than

INSUFFICIENTLY COMPLIMENTARY
!
!
19!
the compliment expressers. We note, however, that their experiment differed from ours in at
least two important ways: First, while third-party observers can easily appreciate a generic
compliment on a stranger’s appearance, meaningful compliments between close friends or family
members are likely more contextual and are grounded in shared memories, making the
interpersonal warmth less accessible to outsiders. Second, observers in their experiment did not
have access to the exact compliments shared by expressers. Instead, these hypothetical observers
considered the abstract notion of giving a compliment without focusing on the precise
compliment expressed. Our participants, in contrast, evaluated exactly the same compliments
shared by expressers, and hence are able to evaluate the precise words exchanged in the
compliment. The discrepancy between our findings and that of Boothby & Bohns is also
consistent with our proposed mechanism of differing focus of competence and warmth between
compliment expressers and recipients (see Experiments 3 and 4).
Experiment 2: Alternative Interpretations
Companions who have known each other for an average of 10 years surely understand a
great deal about each other, but the results from Experiment 1 indicate that they do not fully
understand the impact that their kind words will have on each other. Those who expressed three
compliments undervalued the positive impact that their words would have on their recipient,
significantly underestimating their positive mood and overestimating their feeling of
awkwardness. This misunderstanding was unique to expectations about compliments as
participants in the control condition—who did not exchange compliments—did not
systematically misunderstand their partners’ positive mood or feelings of awkwardness. As a
result, significant miscalibration emerged in expectations of others’ reactions to a compliment,
rather than from simply making overly regressive predictions of others’ emotional experiences.

INSUFFICIENTLY COMPLIMENTARY
!
!
20!
Nevertheless, there are three alternative interpretations of these results, which we
designed Experiment 2 to test directly. First, our procedure of giving three compliments to
another person is potentially unusual, and hence not representative of what might be a more
typical experience of giving a single compliment to another person. We therefore asked
compliment expressers in Experiment 2 to write either one or three compliments to their
recipient to test if this meaningfully affects expressers’ expectations or recipients’ experiences.
Second, another perspective gap that could create miscalibrated expectations between expressers
and recipients is a “curse of knowledge” (Camerer, Loewenstein, & Weber, 1989). That is,
because expressers know of the praiseworthy qualities in their partner that deserve a compliment,
they may assume their kind words are equally known to their recipient, and hence obvious and
unsurprising. Prior research on gratitude (Kumar & Epley, 2018) found that those expressing
gratitude significantly underestimated how surprised recipients would be in receiving their letter
of gratitude and how surprised they would be by its content, but also found that these judgments
were not related to interest in expressing gratitude. We tested this potential curse of knowledge
mechanism in Experiment 2 by examining whether compliment expressers underestimate how
surprised recipients would be by their compliment. Finally, compliment expressers may worry
that sharing kind words with another person could disrupt the status equilibrium in their
relationship, making the expressers seem relatively weak or submissive. People may expect that
expressing a compliment could lower their perceived status in the eyes of a recipient due to
social comparison (Sezer, Prinsloo, Brooks, & Norton, 2020), or it could appear to raise the
status of the recipient and potentially be perceived as disingenuous or ingratiating (Vonk, 1998,
1999; Westphal & Stern, 2007). We assessed this possibility by measuring how relatively weak

INSUFFICIENTLY COMPLIMENTARY
!
!
21!
and submissive versus strong and dominant expressers expected their compliment-giving
behavior to be perceived, and compared that against recipients’ actual evaluations.
Finally, both compliment expressers and recipients reported their mood before and after
giving or receiving compliments, thereby allowing us to assess the relative causal impact of both
giving and receiving compliments on people’s moods. People tend to feel happier when
performing prosocial actions. One experiment indicated that people who give a stranger a
compliment also report being in a more positive mood after doing so (Boothby & Bohns, 2020),
indicating that any barriers to expressing compliments could also cause people to miss
opportunities to increase one’s own well-being. Experiment 2 allows us to test which experience
creates the most positive mood: giving compliments or receiving compliments.
Method
Participants. We aimed to collect 50 acquainted pairs per condition after exclusions.
Due to social distancing policies from COVID-19, we conducted this experiment online and
advertised via the Internet to potential participants in our research center’s subject pool of
students and community members recruited from the local area as well as from across the United
States. All participants who signed up for this experiment were assigned to the role of expresser,
who then provided the contact information of the recipients who would receive their
compliments. Because this experimental design yields imperfect response rates from recipients
(see also Kumar & Epley, 2018), we continued collecting data from expressers beyond our
targeted number until we had responses from at least 50 recipients in each condition. Expressers
spent approximately ten minutes, on average, on their task and received a $3 Amazon gift card in
exchange for their participation. Recipients spent two to three minutes, on average, to complete
their survey and did not receive financial compensation.

INSUFFICIENTLY COMPLIMENTARY
!
!
22!
A total of 138 people participated as expressers (Mage = 29.69, SDage = 12.58; rangeage =
19–68; 68.8% female), and 112 recipients responded to our survey (an 81% response rate; Mage =
31.92, SDage = 13.71; rangeage = 18–74; 62.0% female)
3
. We excluded one pair from all analyses
because the expresser did not write a compliment. We excluded two more pairs from the
following analyses where recipients reported unusually negative moods both before receiving
their compliments (-2.45 and -2.92 SDs below the mean) and after receiving their compliments
(both -4.37 SDs below the mean), which made them statistical outliers
4
. This yielded 109 pairs
in the final analyses. These pairs consisted of friends (N = 48), family (N = 31),
married/engaged couples (N = 20), romantic couples (N = 9), and colleagues (N = 1). The
expressers reported knowing their recipient for an average of 13.42 years (ranging from 8
months to 45 years; SD = 10.81 years).
Procedure. This experiment employed a 2 (perspective: expresser vs. recipient) × 2
(compliment number: one vs. three) between-participants design. We first asked expressers to
identify someone they could compliment in their life and then briefly described their relationship
type and how long they have known their recipient. Participants then reported their own mood
by responding to the item, “How positive/negative do you feel right now?,” using the same scale
as in Experiment 1.
!
3
We also collected demographic information on participants’ racial/ethnic background and highest level of
education. Participants ranged across racial/ethnic groups (34% Asian, 14% Black, 10% Hispanic, 31% White, 4%
multiracial or other, 7% did not respond) and had various amounts of educational experience (40% bachelor’s
degree, 30% part of college, 19% post-graduate degrees, 7% high school, and 4% less than high school or did not
report).
4
!We did not think to include the standard practice of excluding statistical outliers more than 3SDs from the mean in
our preregistration as we had not observed them in other experiments in this project. In retrospect, recipients’
unusually negative mood in those two instances—even before receiving the compliment—might be related to a
procedural change as we attempted to recruit online during COVID-19. In all other experiments, we recruited
participant simultaneously at time that was convenient for both participants, whereas expressers in this experiment
volunteered their recipient’s email as they completed their portion and we later reached out to the nominated
recipients to recruit them. Importantly, including these two pairs in all analyses does not change any primary
findings (see Footnote 4 as an example). !

INSUFFICIENTLY COMPLIMENTARY
!
!
23!
Expressers then read the same compliment-writing prompt as in Experiment 1, adjusted
to ask participants to write either one compliment or three compliments. Expressers then
reported their mood again on the same scale as above, followed by the same warmth-related and
competence-related items as in Experiment 1. To test the potential role of a “curse-of-
knowledge” in explaining miscalibrated expectations, expressers then reported the extent to
which they expected their recipient “to report already being aware that you thought this way
about him/her?,” and the extent to which their recipient would “report being surprised by the
content of your compliment?,” on scales from 0 (not at all) to 10 (extremely). To test expressers’
beliefs about the impact of giving compliments on their perceived power and status in the
relationship, expressers indicated the extent to which their compliment-giving behavior would
make them appear relatively strong and dominant or weak and submissive to their recipient, on a
scale ranging from -5 (very week and submissive) to 5 (very strong and dominant) with 0 (neither
submissive nor dominant) as the midpoint. Finally, expressers reported gender, age,
race/ethnicity, and the highest education level.
We emailed recipients on the following day and informed them that they were given a
compliment by their expresser (with name specified) in our study. After consenting to
participate, recipients reported their current mood on the same scale used by expressers.
Recipients then read their compliment(s) and reported how positive and how awkward they felt,
how warm and competent they perceived the compliment(s) to be, how surprised they were by
the compliment(s), and how strong or weak they perceived their expresser’s compliment-giving
behavior to be on the same scales that expressers used. Finally, recipients responded to the same
demographic questions as above.
Results

INSUFFICIENTLY COMPLIMENTARY
!
!
24!
Positive mood. Following our pre-registered analysis plan, we created a composite
measure of positive mood after confirming the strong correlation between the negative/positive
mood and the pleasantness item (rexpresser = .67; rrecipient = .73; ps < .001) and then conducted a
linear mixed model regression with perspective (expresser vs. recipient) and compliment number
(one vs. three) as fixed effects and participant pairs as random effects. Results indicated a
significant main effect of perspective, F(1,107) = 10.16, p = .002, η2p = .086, a marginally
significant main effect of compliment number, F(1,107) = 3.44, p = .066, η2p = .031, and a
nonsignificant perspective by compliment number interaction, F(1,107) = 0.17, p = .68, η2p =
.001. Replicating Experiment 1, expressers significantly underestimated how positive their
recipients would feel (expresser: M = 7.49, SD = 1.40; recipient: M = 8.04, SD = 1.50),
regardless of whether they were giving one compliment or three compliments
5
.
Awkwardness. The same linear mixed model regression on awkwardness indicated only
a significant main effect of perspective, F(1,107) = 25.86, p < .006, η2p = .20, with expressers
significantly overestimating how awkward their recipients would feel (expresser: M = 3.04, SD =
2.78; recipient: M = 1.54, SD = 2.14) regardless of whether they were giving one or three
compliments to recipients.
Warmth and competence. As in Experiment 1, we averaged the two warmth items (r
= .78, p < .001) and the two competence items (r = .75, p < .001) to create composite scores for
each attribute and subjected them to a 2 (perspective: expresser vs. recipient) × 2 (compliment
number: one vs. three) × 2 (attribute: warmth vs. competence) linear mixed model regression
with a random intercept for each pair, a random intercept for perspective within each pair, and a
!
5
Including the two pairs where recipients reported unusually negative mood throughout the study yielded a
significant main effect of perspective, F(1,109) = 5.59, p = .020, η2p = .049, but neither a significant main effect of
compliment number, F(1,109) = 1.81, p = .18, η2p = .016, nor any interaction, F(1,109) = 0.84, p = .36, η2p = .008.

INSUFFICIENTLY COMPLIMENTARY
!
!
25!
random intercept for attribute within each pair. Replicating Experiment 1, this analysis yielded a
significant main effect for perspective, F(1,107) = 62.08, p < .001, η2p = .37, and a significant
main effect of attribute, F(1,107) = 48.21, p < .001, η2p = .31, qualified by a significant
perspective × attribute interaction, F(1,107) = 23.34, p < .001, η2p = .18 (all other effects were
nonsignificant, ps > .66). Compliment expressers underestimated the perceived warmth of their
compliments (expresser: M = 7.76, SD = 1.78; recipient: M = 8.85, SD = 1.50; F(1,107) = 33.26,
p < .001, η2p = .30), but underestimated the perceived competence of their compliments to a
significantly larger extent (expresser: M = 6.73, SD = 2.12; recipient: M = 8.58, SD = 1.49;
F(1,107) = 74.42, p < .001, η2p = .45).
Consistent with Experiment 1, expressers’ expectations about the perceived warmth and
competence of their compliments were significantly correlated with their expectations of their
recipients’ positive mood (rs = .53 and .52, ps < .001). Recipients’ evaluations of the warmth
and competence of the expressers’ compliments were correlated with their positive mood as well
(rs = .44 and.43, ps < .001). Unlike in Experiment 1, warmth and competence evaluations were
negatively correlated with awkwardness from both the expressers’ perspective (rs = -.32 and
-.16, ps < .001 and = .089) and the recipient’s perspective (rs = -.37 and -.36, ps < .001).
Mediation analysis. As shown in Figure 4, an exploratory mediation analysis following
the same procedure used in Experiment 1 indicated that underestimating how competent one’s
compliment(s) was perceived accounted for a significant proportion of the variance in
expressers’ underestimation of their recipient’s positive mood. Unlike in Experiment 1, the
indirect effect of underestimating warmth was nonsignificant when entered as a simultaneous
mediator with underestimating competence, likely because it was overshadowed by the larger
indirect effect of the competence path (see more details in Supplemental Materials).

INSUFFICIENTLY COMPLIMENTARY
!
!
26!
Figure 4. Mediational analysis testing the relationship between miscalibrated expectations of
perceived competence and warmth of the compliments and miscalibrated expectations of
recipients’ positive moods in Experiment 2.
Surprise. To assess whether expressers misunderstood the extent to which their
compliments were already known to their recipient, we first calculated composite scores of
surprise from the two items (rexpresser = .44; rrecipient = .41; ps < .001; the “already aware” item was
reverse-coded)
6
and then analyzed them in a linear mixed model regression. We observed
nonsignificant effects of perspective, compliment number, and their interaction, ps > .41,
indicating that expressers did not underestimate how surprised their recipients would be
(expresser: M = 6.12, SD = 2.08; recipient: M = 6.17, SD = 2.29). Additional exploratory
analyses revealed a small and marginally significant correlation between expected surprise and
positive mood for expressers (r = .16, p = .098), and a nonsignificant correlation between
surprise and positive mood for recipients (r = .08, p = .40). Similar to results reported with
!
6
We preregistered our intention to calculate a composite score of surprise after confirming that the two items were
highly correlated (i.e., r >= .50). Given that they were only moderately correlated, we performed the same statistical
analysis on each item separately and found no significant effects for either the first item (ps > .58) or the second
item, ps > .14. Due to the lack of meaningful differences, we reported composite scores here for brevity.
Perspective
(0 = expresser;
1 = recipient)
underestimating
Warmth
underestimating
Competence
underestimating
Positive Mood
.18 [-.10, .45]
1.09***
(c’ = -.06, p= .80)
c= .55, p= .002
1.84***
.17
.23*
.42 [.09, .83]

INSUFFICIENTLY COMPLIMENTARY
!
!
27!
expressions of gratitude (Kumar & Epley, 2018), a “curse of knowledge” does not seem to
explain why expressers undervalue the positive impact of expressing compliments.
Relative status. To assess the potential impact of giving compliments on perceptions of
strength versus weakness, and hence relative status in a relationship, we analyzed participants’
ratings on this item in the same linear mixed model as above. Results indicated only a
significant main effect of perspective, F(1,107) = 10.43, p = .002, η2p = .090, indicating that
expressers underestimated how strong and dominant the recipients would perceive their
compliment-giving to be (expresser: M = 0.67, SD = 1.42; recipient: M = 1.28, SD = 2.12).
However, exploratory analyses comparing expressers’ expectations against the neutral baseline
(i.e., 0: “neither submissive nor dominant”) showed that even expressers expected their
compliment-giving behavior to be perceived as somewhat strong, on average, rather than weak,
t(108) = 4.92, p < .001. Finally, an exploratory mediation analysis testing status as an alternative
mediator found that differences in perceived status did not account for a significant proportion of
variance in the difference between expected versus experienced positive moods (see
Supplemental Materials for more details).
Mood change. To assess the extent to which giving and receiving compliments affected
participants’ mood, we analyzed participants’ self-reported mood (linearly transformed to a 0-10
scale) in a 2 (perspective: expresser vs. recipient) × 2 (compliment number: one vs. three) × 2
(time: before vs. after compliments) linear mixed model. Results revealed a significant main
effect of perspective, F(1,107) = 12.79, p < .001, η2p = .11, a significant main effect of time,
F(1,107) = 221.78, p < .001, η2p = .68, and a significant perspective × time interaction, F(1,107)
= 6.60, p = .012, η2p = .058. Neither the main effect of compliment number nor any other
interactions were statistically significant, ps > .11. As predicted, expressing compliments made

INSUFFICIENTLY COMPLIMENTARY
!
!
28!
the expressers feel significantly more positive (before: M = 6.29, SD = 1.78; after: M = 8.29, SD
= 1.51; F(1,107) = 173.94, p < .001, η2p = .62). Receiving compliments also made the recipients
feel significantly more positive (before: M = 5.75, SD = 2.06; after: M = 7.29, SD = 1.78;
F(1,107) = 109.10, p < .001, η2p = .46). The significant interaction indicates that the increase in
positive mood was significantly larger for recipients than for expressers. This result is
inconsistent with results reported by Klein, Nault, & Sezer (2020), who found that those asked to
compliment a stranger on a character description felt more positive after giving their compliment
than the recipients did. We discuss this result in more detail in the General Discussion.
Discussion
As in Experiment 1, people underestimated how positive their compliments would make
a recipient feel, regardless of whether they were passing along only one or three compliments,
and even though they had known their recipient for an average of more than 13 years.
Mediational analyses indicated that expressers’ tendency to underestimate how competent their
compliment would seem to their recipients, and to a lesser extent how warm they would seem,
was significantly related to their tendency to underestimate the recipients’ positive mood after
reading their compliment. In contrast, alternate mediation models found no statistical evidence
that expressers’ miscalibrated expectations stemmed from either differences in evaluations of
surprise or relative status. This overall pattern of results suggests that focusing on concerns
about how one is conveying a compliment—its “happy phrasing,” to use Twain’s term—instead
of focusing on the meaning behind one’s compliment—the warmth conveyed by one’s
phrasing—may exacerbate expressers’ misunderstandings of their recipient’s reactions. This
then predicts that shifting expressers’ attention to focus more intently on the warmth conveyed
by their compliments should lead to more calibrated expectations of the recipient’s reaction. We

INSUFFICIENTLY COMPLIMENTARY
!
!
29!
designed Experiment 3 to test this hypothesis by manipulating expressers’ attention to either
competence or warmth of their compliments and then measuring how they expected their
recipient to react to their compliments.
Experiment 3: Shifting Attention to Calibrate Expectations
We asked one person in a pair—the expresser—to write three compliments to the other
person—the recipient. Before anticipating how the recipient would feel after reading the
compliments, some expressers were directed to focus on the warmth conveyed by their
compliments (warmth-focused), others on the competence conveyed by their compliments
(competence-focused), and others received no direction (unfocused). We predicted that those
directed to focus on the warmth conveyed by their compliments would expect their recipients to
react more positively than those directed to focus on competence, thereby becoming more
calibrated in their expectations of their recipients’ reactions because they are evaluating their
compliments by focusing on similar attributes as the recipients themselves (Eyal & Epley, 2010).
If expressers are naturally focused first and foremost on the competence-related attributes of
their compliments, then directing their attention to only competence-related aspects should not
significantly alter their expectations compared to a condition where they evaluate both attributes
without any further instruction to focus on one attribute or the other.
Method
Participants. We conducted this experiment in the same public park, using the same
recruiting method, as in Experiment 1. We targeted a sample of 50 pairs in each condition, for a
total of 300 participants. Our final sample included 49 pairs in the warmth-focused condition, 52
pairs in the competence-focused condition, and 49 pairs in the unfocused condition (Mage =
36.54; SDage = 16.34, rangeage = 18–88; 65% female). Six additional pairs started the experiment

INSUFFICIENTLY COMPLIMENTARY
!
!
30!
but were excluded from analyses because they left the park before finishing (3 pairs, 1 in each
condition), lacked English proficiency (1 in the competence-focused condition), accidentally
failed to read the compliments (1 in the warmth-focused condition), or indicated
misunderstanding of the scales upon completion (1 in the competence-focused condition).
Procedure. Experiment 3 employed a 3 (focus: warmth-focused, competence-focused,
unfocused) × 2 (perspective: expresser, recipient) between-participants design. Participant pairs
were randomly assigned to one of the focus conditions, and participants within each pair were
assigned to be the expresser or the recipient. The procedure was similar to the compliment
condition of Experiment 1 except for two changes. First, after writing three compliments and
before predicting how their recipient would feel, the expressers were presented with the
compliments they just wrote on a separate screen and were either provided with an instruction
that manipulated the focus of their evaluation (warmth-focused or competence-focused
conditions) or given no special instruction (unfocused condition). Specifically, we instructed
participants in the warmth-focused condition to, “think about the degree to which your messages
are warm and sincere, paying close attention to the overall impressions you meant to convey in
your compliments.” We instructed participants in the competence-focused condition to, “think
about the degree to which your messages are articulate and well-written, paying close attention
to the exact words you chose to include in each of your sentences.”
Second, we adjusted the order and phrasing of the survey items to strengthen the
evaluative focus manipulation. In the warmth-focused condition, expressers first answered a
one-item question that asked them to evaluate the warmth expressed by their compliments
(“How warm and sincere would you rate the compliments you wrote to be?”). In the
competence-focused condition, expressers first answered a one-item question that asked them to

INSUFFICIENTLY COMPLIMENTARY
!
!
31!
evaluate the competence of their compliments (“How articulate and well-written would you rate
the compliments you wrote to be?”). Participants in both conditions then predicted their
recipient’s positive mood and feeling of awkwardness. In the unfocused condition, expressers
evaluated both the warmth and the competence of their compliments (order counterbalanced
across pairs) before predicting their recipients’ positive mood and awkwardness, but with no
explicit request to consider one attribute or the other. Finally, similar to Experiment 1,
expressers finished their task by reporting their current relationship quality with the recipient,
their compliment frequency, and demographic information (gender and age).
Although our primary interest was the expressers’ expectations, we also adjusted the
recipients’ item order and phrasing. After first reporting their moods, the recipients evaluated
either warmth or competence in the warmth-focused or competence-focused conditions,
respectively, or answered both in the unfocused condition (order counterbalanced).
Results
Positive mood. To test our hypotheses that directing expressers’ focus to the warmth of
their compliments, but not competence, can create more calibrated expectations of recipients’
reactions, we analyzed the data in a 3 (condition: warmth-focused, competence-focused,
unfocused) × 2 (perspective: expresser, recipient) linear mixed model with participant pairs as
random effects. Results revealed a significant effect of perspective, F(1,147) = 48.00, p < .001,
η2p = .25, indicating that expressers tended to underestimate how positive their recipients felt
after receiving their compliments.
7
Most relevant to the current experiment, we also found a
significant main effect of condition, F(2,147) = 4.80, p = .010, η2p = .061, qualified by a
!
7
We preregistered analyses using t-tests and ANOVAs on difference scores but later realized that a linear
mixed model was a better statistical approach and hence report the better analysis. Both approaches yield the
same conclusions.

INSUFFICIENTLY COMPLIMENTARY
!
!
32!
marginally significant interaction between perspective and condition, F(2,147) = 3.00, p = .053,
η2p = .039. Although compliment expressers significantly underestimated how positive their
recipient would feel across all three conditions (warmth-focused: Ms = 8.50 vs. 8.93;
competence-focused: Ms = 7.71 vs. 8.62; unfocused: Ms = 7.82 vs. 8.95), simple effects tests
showed that their expectations were relatively more calibrated in the warmth-focused condition
(F(1,48) = 7.14, p = .01, η2p = .028) than those in the competence-focused condition (F(1,51) =
15.58, p < .001, η2p = .12) or the unfocused condition (F(1,48) = 27.33, p < .001, η2p = .17).
Also as predicted, expressers’ expectations differed significantly across experimental conditions,
F(2,147) = 6.47, p = .002, η2p = .081, with participants in the warmth-focused condition
expecting their recipient to feel more positive than those in the competence-focused (F(1,147) =
11.14, p = .001) or unfocused conditions (F(1,147) = 8.13, p = .005). Recipients’ positive mood,
in contrast, did not vary significantly across conditions, F(2,147) = 1.42, p = .25, η2p = .019.
These results are shown in Figure 5(A).
Figure 5. Participants’ expected and actual positive mood (Panel A) and awkwardness (Panel B)
in the warmth-focused, competence-focused, and unfocused conditions in Experiment 2. Error
bars reflect ± 1 standard errors.
5
6
7
8
9
10
Warmth-focus Competence-focus Unfocused
Positive Mood
Expresser Recipient
0
1
2
3
4
5
Warmth-focus Competence-focus Unfocused
Awkwardness
Expresser Recipient
(A)
(B)

INSUFFICIENTLY COMPLIMENTARY
!
!
33!
Awkwardness. A similar 3 (condition: warmth-focused, competence-focused, or
unfocused) × 2 (perspective: expresser, recipient) linear mixed model on feelings of
awkwardness revealed a main effect of perspective, F(1,147) = 16.79, p < .001, η2p = .10, a
nonsignificant main effect of condition, F(2,147) = .29, p = .75, and a nonsignificant interaction,
F(2,147) = 1.12, p = .33. Expressers in the warmth-focused condition did not significantly
overestimate how awkward their recipient would feel (F(1,48) = 1.59, p = .21), whereas
expressers in both the competence-focused and the unfocused conditions did so (competence-
focused: Ms = 2.79 vs. 1.40; F(1,51) = 8.95, p = .004, η2p = .065; unfocused: Ms = 2.92 vs. 1.67;
F(1,48) = 7.51, p = .009, η2p = .0.50). These results are shown in Figure 5(B).
Warmth and competence. As in Experiments 1 and 2, we examined how expressers
and recipients might have systematically evaluated the same compliments differently. We only
analyzed responses in the unfocused condition given that they evaluated both attributes of the
same compliments. A 2 (attribute: warmth vs. competence) × 2 (perspective: expresser vs.
recipient) linear mixed model on the expressers’ and recipients’ evaluations of the compliments
again showed a significant main effect for attribute, F(1,48) = 71.95, p < .001, η2p = .60, and a
significant main effect of perspective, F(1,48) = 66.98, p < .001, η2p = .58, qualified by a
significant interaction, F(1,48) = 45.71, p < .001, η2p = .49. Expressers underestimated recipients
evaluations of competence (F(1,48) = 77.26, p < .001, η2p = .62) significantly more than their
evaluations of warmth (F(1,48) = 24.00, p < .001, η2p = .33). Replicating previous results,
expressers judged the competence of their compliments significantly less favorably than the
recipients did, suggesting that expressers were overly concerned about not getting the words
“just right” or being articulate enough compared to the recipients’ very positive evaluations.

INSUFFICIENTLY COMPLIMENTARY
!
!
34!
Compliment frequency. Finally, consistent with other surveys, participants again
reported giving compliments significantly less often than they thought they should (M = -.89, SD
= 1.30), t(299) = -11.83, p < .001, d = .68 (see Supplemental Materials).
Discussion
Experiment 3 replicated the primary results from Experiments 1 and 2, and also provides
direct causal evidence that a perspective gap in evaluations of the competence versus warmth
conveyed by compliments could at least partly explain expressers’ miscalibrated expectations of
recipients’ reactions, such that aligning the perspectives could calibrate expressers’ expectations
judgment. Specifically, shifting expressers’ attention to focus on the warmth conveyed by their
compliments made expressers’ expectations more calibrated, and significantly more positive,
compared to expressers who focused on the competence conveyed by their compliments or
evaluated both dimensions. These results, along with other recent research (Boothby & Bohns,
2020; Kumar & Epley, 2018), suggest that an undue focus on competency, relative to warmth,
may contribute to a broader tendency for people to underestimate the positive impact that their
prosocial actions will have on others.
Although shifting expressers’ attention in Experiment 3 to focus on the warmth-related
dimensions of their compliments increased the calibration of their expectations, it did not fully
eliminate their tendency to underestimate how positive their recipient would feel entirely. This
occurs presumably because expressers also underestimate how warm their recipients will
perceive compliments to be.
Experiment 4: Updating Expectations to Change Intentions?
Taken together, five experiments (Experiments 1, 2, 3, S1a, S1b) provide consistent
evidence that people systematically misunderstand the overall positive impact of compliments on

INSUFFICIENTLY COMPLIMENTARY
!
!
35!
recipients. We believe expressers’ miscalibrated expectations are important because they may
create a misplaced psychological barrier to expressing compliments more often in daily life.
Prior research has documented somewhat related interpersonal misunderstandings, speculating
about their importance but rarely testing these implications directly (e.g., Epley & Schroeder,
2014; Flynn & Lake, 2008; Kumar & Epley, 2018; Sandstrom & Dunn, 2014). Experimentally
testing the consequences of mistaken expectations would therefore provide a meaningful
advance on prior research and reveal how such miscalibration creates undue reluctance to
express compliments in daily life.
As a first step, we investigated how people’s expectations of recipients’ reactions were
related to their interest in expressing compliments to others. If expectations about a recipient’s
reaction at least partly guide people’s interest in expressing compliments, then they should be
more interested in giving a compliment when they expect the recipient to respond more
favorably. Identifying the link between expectations and intentions is critical for establishing the
plausibility of our broader hypothesis that miscalibrated expectations about others’ reactions
could lead people to deliver fewer compliments than might be optimal in their daily lives. We
summarize two initial tests of these hypotheses here in the main text, and provide full details in
the Supplemental Materials.
In Experiment S2, we asked online participants to generate compliments for five different
people in their lives, to predict how positive and awkward each recipient would feel and how
warm and competent their compliments would be perceived, and then to indicate how interested
they were in delivering their compliment to each person. As predicted, potential expressers’
expectations of how positive and how awkward their compliments would make their recipient
feel were significantly correlated with their reported interest in expressing that compliment, and

INSUFFICIENTLY COMPLIMENTARY
!
!
36!
that expected positive mood was a significantly stronger predictor of expressers interest’ than
was expected awkwardness. In addition, expressers’ expectations of how warm and how
competent the compliments would be perceived were again correlated with expectations of the
recipient’s positive mood and awkwardness, which were in turn positively correlated with their
interest in sharing their compliments with recipients.
To test the extent to which these expectations might guide people’s choices of which
compliments to share with others, we conducted Experiment S3 in which we recruited
acquainted pairs in the same public park described in Experiment 1. We asked those randomly
assigned to be expressers to generate three compliments about the other person, to predict how
this recipient would react to each compliment, and then to choose only one compliment to
actually share. As predicted, participants expected that the compliment they chose to share (M =
7.89; SD = 1.42) would make their recipient feel significantly more positive than the
compliments they did not choose to share (M = 7.08; SD = 1.48), F(1,52) = 23.83, p < .001, η2p =
.31, but they expected similar levels of awkwardness (p = .26).
Taken together, Experiments S2 and S3 suggest that people’s expectations of how
positive their compliments would make the recipients feel at least partly guided their interest in
expressing them. If people systematically underestimate how positive their kind words will
make others feel, then this could create a misplaced barrier to expressing compliments more
often in daily life. This predicts that reducing this psychological barrier by calibrating people’s
expectations should increase their willingness to express compliments to their potential
recipients.
We tested this hypothesis by attempting to manipulate people’s expectations of a
recipient’s reaction to a compliment in two different ways. As a first attempt in Experiment S4

INSUFFICIENTLY COMPLIMENTARY
!
!
37!
(see Supplemental Materials for full details), we informed some compliment expressers that
people were prone to underestimating the positive impact of compliments on receivers, and then
measured the number of compliments they generated and shared with a recipient. We compared
the number of compliments expressers shared in this informed condition against the number of
compliments expressers shared in a separate control condition in which expressers received no
information about people’s tendency to underestimate the positive impact of compliments. This
experiment tested whether learning about our experimental results changed the number of
compliments expressers would share. Consistent with our hypothesis, expressers who were
informed that people tended to underestimate the positive impact of their compliments shared
marginally more compliments than those who were not informed (Ms = 5.64 and 4.77, SDs =
2.82 and 2.51), Mann-Whitney U = 1496, p = .08, d = 0.33. As in Experiment 2, both expressers
and recipients reported feeling more positive after either expressing or receiving their
compliments, although the increase in positive mood was again larger for recipients.
Given that directly informing people about experimental results had only a relatively
small effect on people’s expectations, we utilized a design similar to Experiment 3 in which we
directed participants’ attention to focus on either warmth or competence to test how
manipulating attention would affect their interest in expressing compliments. If underestimating
recipients’ positive reaction to a compliment creates a barrier to expressing them more often,
then participants in Experiment 4 who focus on the warmth conveyed by their compliments
should be more interested in expressing a compliment than those who are focused on its
competence.
Method

INSUFFICIENTLY COMPLIMENTARY
!
!
38!
Participants. We targeted a sample of 100 participants in each of the two experimental
conditions and recruited from the same participant pool as in Experiment 2. A total of 208
participants—102 in the warmth-focused condition and 106 in the competence-focused
condition—completed this experiment (Mage = 24.71, SDage = 6.73, rangeage = 18–54; 72%
female).
Procedure. Due to the COVID-19 pandemic, we conducted this experiment remotely by
sending a study link to participants. Upon opening the study link, participants read the
compliment-writing instruction which prompted them to write down one compliment they could
give to someone in their life who did not currently live in the same household. We encouraged
participants to write a compliment that they could deliver using existing channels of
communication. Participants then identified a potential compliment recipient and described their
relationship type and length. The relationships people identified included friendship (N = 135),
family (N = 41), romantic relationship (N = 15), married/engaged couples (N = 7), professional
relationship (N = 9), and followership (N = 1). Expressers reported knowing their recipient for
an average of 8.87 years (SD = 9.28). Participants then wrote down one compliment they could
give to this recipient, and reported their current mood on the same positive/negative scale used in
the preceding experiments.
Participants randomly assigned to the warmth-focused condition then saw a page titled,
“How warm and sincere is your compliment?”, whereas those randomly assigned to the
competence-focused condition saw a page titled, “How articulate and well-written is your
compliment?”. The first paragraph indicated that participants had an opportunity to actually
deliver the compliment they had just written.
Participants in the warmth-focused condition then read the following paragraph:

INSUFFICIENTLY COMPLIMENTARY
!
!
39!
“When deciding whether or not to express a compliment, it is important to consider the spirit
of compliment-giving: at its core, giving compliments is a matter of sharing warm, sincere,
and friendly thoughts to another person in your life, so that they know how positively you
feel about them. Therefore, when deciding whether or not to express your compliment,
“[display the compliment here]”, please focus on the extent to which your compliment was
intended to be warm, sincere, and friendly. How warm was your compliment intended to
be?”
In contrast, participants in the competence-focused condition read the following:
“When deciding whether or not to express a compliment, it is important to consider the spirit
of compliment-giving: finding just the right words to articulate your thoughts about another
person in your life, so that they know precisely how you feel. Therefore, when deciding
whether or not to express your compliment, “[display the compliment here]”, please focus
on the extent to which you were able to get the words just right, sounding clear, intelligent,
and articulate. How competently were you able to express your thoughts through your
compliment?
As part of this manipulation, participants then rated either the warmth or the competence of their
compliments (depending on their assigned condition) on the same item as in Experiment 3.
On the next page, participants answered, “How interested are you in actually sending this
exact compliment, “[compliment]”, to [recipient] right now?”, with the actual compliment and
recipient name inserted using their own responses from an earlier part of the survey. The
response scale ranged from 0 (not at all interested) to 10 (extremely interested).
On the next page, participants learned that they had a 2-minute window in which they
could deliver the compliment that they wrote, if they would like to, using whatever channels of
communication they normally used for interacting with their recipient. We disabled the advance

INSUFFICIENTLY COMPLIMENTARY
!
!
40!
button on this page for two minutes and provided a text template that included the recipient’s
name, the compliment, and a brief sentence describing the research study that they could directly
copy and send (or modify before sending, if they chose to). On the next page, participants
indicated whether or not they actually sent their compliment just now (“Yes, I did” or “No, I did
not”). To encourage honesty, the instruction assured participants that their response would not
influence their payment or their standing in our participant pool. If participants indicated having
sent their compliment, then we asked them to indicate how they sent it in an open-ended text
box. If participants reported that they did not send it, then we asked them to explain why they
did not send their compliment. Finally, participants recalled what they were asked to focus on
when deciding whether or not to send their compliment in an open-ended question (i.e., the
manipulation check), reported their demographic information, and were debriefed.
Results and Discussion
We first confirmed that compliments in the warmth-focused and the competence-focused
conditions were of comparable length (p = .41) and required a similar amount of time for the
expressers to write (p = .49)
8
.
As predicted, participants in the warmth-focused condition reported being significantly
more interested in expressing their compliments (M = 6.05, SD = 2.93) than participants in the
competence-focused condition (M = 4.71, SD = 3.01), t(206) = 3.26, p = .001, d = .45.
Consistent with our prediction, focusing on how “warm, sincere, and friendly” their compliment
!
8
The p values are calculated with all compliments. However, a few expressers wrote considerably longer
compliments or spent far more time than normal on the compliment-writing page (e.g., one expresser proceeded
after 10 hours with a compliment of 9 words), so in another set of analyses, we excluded outliers that were three
standard deviations above the means in respective conditions and confirmed that compliments in the warmth-
focused and competence-focused conditions were comparable in both word counts (Ms = 25.88 and 23.21 words,
SDs = 28.16 and 17.22, respectively; t(202) = 0.35, p = .72) and writing duration (Ms = 109.90 and 121.14 seconds,
SDs = 64.08 and 71.24, respectively; t(180) = -1.12, p = .27).

INSUFFICIENTLY COMPLIMENTARY
!
!
41!
was led expressers to be more interested in expressing it compared to focusing on how “clear,
intelligent, and articulate” the compliment was.
Because we asked participants to evaluate the warmth or competence of their compliment
to help them focus on the respective constructs, we also conducted linear regression analysis to
examine whether such evaluations were related to their interest in expressing a compliment. As
expected, expectations of the warmth and competence conveyed by their compliments were
related to their interest in expressing their compliment (warmth: B = .53, SE = .18, t = 2.92, p
= .004; competence: B = .86, SE = .10, t = 8.20, p < .001). Consistent with results in Experiment
3, participants in the competence-focused condition evaluated the competence of their
compliments significantly more critically (M = 5.74, SD =2.19) than those evaluating the warmth
of their compliments in the warmth-focused condition (M = 8.34, SD = 1.54), t(206) = 9.88, p
< .001, d = 1.38.
Despite having generated compliments that could be easily shared in the moment,
participants did not seem especially interested in sharing their compliment during the experiment
as only 49.5% of participants reported actually sending their compliment in the 2-minute window
we provided. Specifically, 53.9% of participants in the warmth-focused condition (55 out of
102), compared to 45.3% in the competence-focused condition (48 out of 106), reported sending
their compliments. The direction of this difference was consistent with our prediction, but was
statistically nonsignificant, χ2 = 1.23, p = .27, w = .12.
To better understand why our experimental manipulation created a sizeable effect on
participants’ interest in expressing their compliment but a weaker effect on the percentage of
participants who actually shared their compliment in this procedure, we conducted an
exploratory analysis to investigate the relationship between these two variables. A logistic

INSUFFICIENTLY COMPLIMENTARY
!
!
42!
regression analysis on people’s compliment-giving behavior (0 = did not send; 1 = sent) revealed
a statistically significant relationship between reported interest in expressing a compliment and
actually expressing the compliment, b = 0.44, z = 6.76, p < .001. However, interest accounted
for only 22.4% of the variance in people’s actual behavior (McFadden R2 = .22), indicating that
whether people sent a compliment or not in the window of opportunity provided in our
experiment was largely driven by additional factors.
Analyzing people’s verbal explanations revealed that the most common reported reason
for withholding compliments was timing. Specifically, 58.6% of those who did not send their
compliments (N = 58 out of 99 responses) mentioned that they either knew the other person
would be preoccupied at that moment, or they felt that sending compliments would seem
“random” and “out of the blue” and/or preferred to deliver the compliment face-to-face or when
a more suitable occasion naturally arose. Other participants explained that they felt “weird”,
“embarrassed”, or “uncomfortable” about sending their compliments or were afraid the other
person would react negatively (15.2%); they preferred to rewrite the compliments using more
effective language (8.1%); they found compliment-giving too “emotional” or uncharacteristic of
their normal interactions (5.0%); they felt lazy or tired (5.0%); they lacked a channel to reach the
other person during the study (4.0%); or they thought the compliments were already known to
the intended recipient (4.0%). We offered participants a two-minute window in which to send
their compliment because it enabled us to easily measure behavior, but it may not have been
optimal for measuring actual compliment-giving in daily life.
Finally, a pre-registered analysis on only those participants who passed what turned out
to be a very restrictive manipulation check (99 out of our 208 participants provided warmth-
related or competence-related keywords in the open-ended recall question) yielded larger main

INSUFFICIENTLY COMPLIMENTARY
!
!
43!
effects, suggesting that a stronger experimental manipulation would have yielded larger
behavioral effects. Specifically, the effect of condition on reported interest in expressing a
compliment was considerably larger in this restricted sample than it was in the entire sample,
t(97) = 4.04, p < .001, d = .81 (d = .45 in the full sample); moreover, 56.9% of those in the
warmth-focused condition reported having sent their compliment compared to 41.7% in the
competence-focused condition (compared to 53.9% vs. 45.3% in the full sample), χ2 = 1.72, p =
.19.
These results suggest that underestimating recipients’ positive evaluations of a
compliment, and hence their positive reaction to it, could create a barrier to expressing
compliments in daily life. Reducing this psychological barrier by encouraging people to focus
on the warmth conveyed by their compliment therefore increased their interest in expressing a
compliment that they could express to another person. Future research will have to assess
whether offering a longer time period for expressing a compliment, or measuring actual behavior
using a different study design, would increase the correspondence between reported interest and
actual expressions of compliments. Future research can also assess whether participants’
apparent concern about getting the timing or occasion “just right” for their compliment is
warranted, or if it represents another manifestation of expressers underestimating how competent
their compliment-giving will be perceived by recipients. Note that our experiments measured
expressers’ expectations of how competently their compliment would be evaluated by
recipients—the “happy phrasing of a compliment” to use Twain’s words from our opening
quote—but did not measure how competently their delivery of the compliment would be
evaluated—what Twain referred to as a compliment’s “happy delivery.” Our research suggests

INSUFFICIENTLY COMPLIMENTARY
!
!
44!
that that expressers’ concerns about the competent delivery of a compliment could also be
miscalibrated.
General Discussion
“The deepest principle in human nature is the craving to be appreciated.”
-William James (1896/1920)
People care a great deal about how they are evaluated by others, leaving people feeling
good when they find out that they are appreciated by others (Leary & Baumeister, 2000). Unlike
William James, who may well have recognized the full power of feeling appreciated, our data
suggest that people may instead think more like Mark Twain, believing the happy phrasing and
receipt of a compliment to be somewhat rare. In fact, our experiments suggest a reliable
tendency to underestimate how positive signs of appreciation—in this case expressing
compliments—will make recipients feel. Across multiple experiments (Experiments 1-4), pairs
of friends, family, and romantic couples consistently underestimated how positive, and
overestimated how awkward, their recipients would feel upon receiving their compliments. Our
experiments indicate that these miscalibrated expectations stem from expressers’ overly critical
view of how competent, and to a lesser extent how warm, their compliments would be viewed by
the recipients, creating a perspective gap between those who receive a compliment and those
who are anticipating its consequences. We also found that third-party observers, like the
expressers, also underestimated the positive impact of compliments on a recipient (Supplemental
Experiments S1a and S1b). These miscalibrated expectations matter in daily life because they
are likely to guide people’s interest in expressing versus withholding genuine compliments that
come to their mind (Supplemental Experiments S2 and S3). Undervaluing compliments could

INSUFFICIENTLY COMPLIMENTARY
!
!
45!
therefore create a psychological barrier to expressing them more often in daily life, such that
more accurately recognizing a recipients’ positive response could increase interest in expressing
compliments more often (Experiments S4 and 4).
Prior research has primarily framed the motivation to express compliments through a lens
of self-presentation and ingratiation (e.g., Jones, 1964; Seiter, 2007; Vonk, 2001). Although
people may sometimes attempt to manipulate others through flattery and false praise, people also
care deeply about others’ well-being. We assume that at least some compliments are prosocially
motivated efforts to make another person feel good by expressing authentic positive thoughts.
However, our theorizing does not preclude the possibility that giving compliments can also make
the expressers feel good. In fact, as with other prosocial acts that make the actors feel good
(Dunn et al., 2014; Curry et al., 2018), participants giving compliments in our Experiments 1 and
S4 also felt more positive after doing so than they did moments before (see also Boothby &
Bohns, 2020). Recent research focusing on exchanging compliments between strangers even
suggests that giving compliments might make people feel better than receiving compliments
(Klein et al., 2020). Two of our experiments, in contrast, found that receiving compliments
enhanced positive mood more than giving compliments, possibly because our compliments were
shared within established relationships and hence might have been more meaningful and
powerful to the recipients. Regardless of its exact magnitude, existing research consistently
indicates that giving compliments enhances a person’s own well-being. This effect could come
from expressers’ conscious focus on positive thoughts about a relationship partner (Emmons &
McCullough, 2003; Fredrickson, Cohn, Coffey, Pek, & Finkel, 2008; Lyubomirsky & Layous,
2013), but it could also come from expecting to achieve the goal of making a recipient feel
positive by giving a compliment (Dunn et al., 2014). A misplaced reluctance to share kind

INSUFFICIENTLY COMPLIMENTARY
!
!
46!
thoughts with others could leave people being less prosocial than would be optimal for their own
well-being.
Although our experiments provide strong converging evidence across multiple
experiments, our compliment-giving procedure is constrained in one important way: expressers
were prompted to write compliments in one sitting, whereas everyday life affords repeated
interactions with recipients in which compliments are generated and shared spontaneously. This
feature of our procedure may raise two concerns. First, expressers may have been forced to
generate low-quality or insincere compliments in our experiments. However, expressers in
Experiment 3 who could write as many compliments as they wished wrote a median number of
five to another person, suggesting that the request to write multiple compliments was not likely
forcing expressers to write more compliments than they would have been able to write otherwise.
Second, prompted compliments may lead to different attributional inferences than spontaneously
generated compliments, either among expressers or recipients. A prompted comment might be
expected to be perceived as less sincere by expressers, and/or might actually be perceived as less
sincere by recipients. We note, however, that our pattern of results replicated consistently across
contexts where the number of compliments generated was fixed versus freely chosen by the
expresser, and recipients’ evaluations were also nearly at the ceiling of the positive affect
measure across all experiments, suggesting that they fully appreciated the compliments.
Although we doubt that the prompted nature of our experimental procedures is meaningfully
influencing evaluations of sincerity among either expressers or recipients, future research should
test the robustness of our results in more spontaneous compliment-giving contexts.

INSUFFICIENTLY COMPLIMENTARY
!
!
47!
From miscalibration to misconnection
Connecting with others in a positive way tends to feel good, such that relatively prosocial
actions can sometimes leave people feeling even happier than relatively selfish ones (Dunn et al.,
2008). Performing prosocial actions intended to benefit others feels good at least partly to the
extent that they actually achieve the actor’s intended goal. An effective act of kindness that
leaves another person better off makes the giver feel happier, while an ineffective act of kindness
that leaves another person feeling indifferent also makes the giver feel less positive. Given how
easy a genuine compliment is to generate, and how positive it can make recipients feel when
expressed, it is somewhat surprising that our participants consistently reported giving fewer
compliments either than they would like to give or than they should give. We believe a tendency
to systematically underestimate how positively others will respond to a compliment can help to
explain this surprising sentiment. If a person believes a genuine compliment will leave another
person feeling relatively indifferent, or feeling somewhat awkward, then they might be
somewhat reluctant to give it. That these expectations tend to be systematically miscalibrated
suggests that people’s expectations may create a somewhat mistaken barrier to prosocial actions.
The tendency for people to underestimate the perceived competence of their kind words
to a greater extent than perceived warmth also suggests that underestimating the positive impact
of prosocial actions is likely to be a somewhat widespread phenomenon. Prosocial actions lead
recipients to feel good at least partly because they convey warmth from another person. If
prosocial actors instead tend to be relatively more focused on how competently they are
executing the action, then a very wide range of prosocial acts are likely to leave the average
recipient feeling better than the prosocial actors expect. Indeed, prior research has documented a
similar tendency to underestimate how positively others will respond to relatively elaborate and

INSUFFICIENTLY COMPLIMENTARY
!
!
48!
deliberate expressions of gratitude (Kumar & Epley, 2018), and to honesty (Levine & Cohen,
2018). We would hypothesize similar miscalibrated expectations in reactions to expressions of
social support, random acts of kindness, deep and meaningful conversation, or any other act
where the primary impact on recipients is an expression of positive intentions or liking (i.e.,
warmth). Miscalibrated expectations could create barriers to a wide variety of behaviors that
would otherwise strengthen social bonds and enhance well-being. Importantly, our hypotheses
do not suggest that people simply misunderstand the impact that their own actions have on
others. We would not expect systematically miscalibrated expectations about how others will
respond to actions defined by their competence, such as a boss’s reaction to a report, a diner’s
reaction to a prepared meal, an audience’s reaction to a presentation, or a reader’s reaction to a
journal article.
In addition to underestimating the positive impact of prosocial acts, our mechanism also
predicts the inverse: underestimating the negative impact of antisocial acts on others. Behaviors
defined by their lack of warmth, such as deception, insults, or rejection, could have created a
more negative impact on recipients than antisocial actors expect. Although direct support for
this hypothesis is lacking, recent research has suggested that people underestimate the severity of
other people’ pain from social exclusion when not actively experiencing such pain for
themselves (Nordgren, Banas, & MacDonald, 2011), and they also underestimate how negatively
others judge them for deliberately concealing negative information (John et al., 2016). Other
experiments indicate that those on the receiving end of a social action are sensitive to an actor’s
intention, such that recipients report feeling more pain when they were intentionally harmed than
unintentionally harmed (Gray & Wegner, 2008). Investigating actors’ expectations in these
antisocial contexts promises both important theoretical contributions and practical implications.

INSUFFICIENTLY COMPLIMENTARY
!
!
49!
One particularly timely context for such investigation is social media, where antisocial behaviors
including hate speech, Internet trolling, and public shaming are both common and impactful
(Ronson, 2015). One possible explanation for the frequency of antisocial behaviors online is that
the distance between the actor and recipient of an antisocial action makes actors especially likely
to underestimate the harm they inflict on their recipient.
Moderating Miscalibration and Compliment-Giving
We have documented that underestimating how positive a compliment will make a
recipient feel could serve as meaningful psychological barrier for giving compliments more often
and provided evidence for one explanation of miscalibrated expectations based on differences in
perceptions of competence and warmth (Abele & Wojciszke, 2007; Bruk et al., 2018; Fiske et
al., 2007; Wojciszke et al., 1998). Of course, our research does not identify or address all of the
factors that could moderate the magnitude of miscalibration or the frequency of compliment-
giving in everyday life.
One potentially surprising feature of our experimental results is that miscalibrated
expectations emerged among people in long-term relationships, suggesting that interpersonal
misunderstanding is not simply something that occurs between strangers (Boothby & Bohns,
2020). Even in long-term relationships, people do not seem to fully appreciate the positive
impact that their kind words can have on their relationship partners. We believe this
misunderstanding is likely to be maintained over time because miscalibrated expectations about
social interactions can keep people from engaging in the very behaviors that would otherwise
calibrate their expectations (Epley & Schroeder, 2014). If a person believes that compliments
are largely met with indifference by others, then they may be reluctant to give them and hence
never find out that their expectations could be wrong. Simply gaining experience by giving more

INSUFFICIENTLY COMPLIMENTARY
!
!
50!
compliments and observing people’s reactions may create more calibrated expectations and
perhaps encourage more positive expressions as well. Those who give compliments more
frequently, whether due to their personality, occupation, or random assignment to experimental
condition, should also be more calibrated in predicting others’ reactions. Miscalibrated
expectations could be perpetuated by a lack of experience that could calibrate their expectations.
The frequency of compliment-giving also varies across cultures. Americans, for instance,
reportedly give more compliments than both English and non-English speakers in other countries
(Fujimura-Wilson, 2014). Although our participant samples were diverse on some dimensions
(such as age and ethnicity), we did not test our hypotheses around the world, and so we cannot
test how our effects may have been moderated by nationality or geography. We suggest that
cross-cultural differences in the expression of compliments may stem from differences in
expressers’ expectations rather than in recipients’ experiences of compliments. In contrast to
individualistic cultures, the heightened focus on hierarchy and relationship status in collectivistic
cultures may focus expressers’ attention even more intensely on how they are conveying a
compliment rather than on the prosocial intent and warmth behind the compliment itself.
Cultural norms can be guided not only by people’s actual experience, but also by their beliefs
and expectations about others, whether calibrated or miscalibrated (Miller & Prentice, 1994).
Investigating the extent to which variance in prosocial norms across cultures is driven by
differences in expectations or actual experiences is a critical topic for future research.
Although we have suggested that undue attention to competence-related aspects of one’s
compliment can lead expressers to underestimate how positive their compliments will make
recipients feel, compliments do vary meaningful in the warmth they convey. In particular, some
compliments are sincere and genuine while others are insincere and merely flattering. Because

INSUFFICIENTLY COMPLIMENTARY
!
!
51!
we were primarily interested in understanding barriers to expressing genuine compliments, we
did not investigate either expectations about or experiences of disingenuous compliments. Jones
(1964) noted that those trying to enhance their own image in the eyes of others by using self-
presentational tactics, such as flattery, face a dilemma: ingratiators want to use tactics that make
them liked by others, but the most extreme tactics may also be the most likely to be detected as
insincere. Someone aiming to flatter another person therefore needs to pass along their
compliments without being detected as insincere. Existing empirical evidence, however,
suggests that this may create a dilemma primarily in the minds of the ingratiators themselves.
Empirical evidence suggests that recipients may be relatively insensitive to ulterior motives and
may happily accept obvious flattery (Chan & Sengupta, 2010; Fogg & Nass, 1996; Vonk, 2002;
Westphal & Stern, 2007). Research also suggests that people are fairly poor at detecting lies
from speech alone (Bond & Depaulo, 2006), suggesting that people may have difficulty
recognizing the difference between sincere and insincere compliments. And yet, because people
are aware of their own intentions behind a compliment, they may expect those intentions will be
more transparent to observers than they actually are (Gilovich, Savitsky, & Medvec, 1998).
People may be reluctant to flatter others with insincere compliments because they overestimate
the likelihood that their insincerity will be detected. At this point, however, we know of no
research that examines how accurately compliment recipients discount for an expressers’ actual
sincerity, or how accurately expressers can anticipate discounting in their recipients.
Finally, people may be reluctant to express compliments more often in daily life out of a
concern that expressing more compliments to another person may also make them seem less
authentic, or may make their compliments less effective. One compliment may be taken as
authentic and make the recipient feel good, but a second might feel a little less good, a third even

INSUFFICIENTLY COMPLIMENTARY
!
!
52!
less good, and so on. This concern seems plausible and somewhat grounded in psychological
reality—for instance, as additional exposures to certain types of stimuli can lead to adaptation,
thereby diminishing the intensity of emotional reactions (Frederick & Loewenstein, 1999;
Wilson & Gilbert, 2008). However, recent evidence has suggested that people may
underestimate how much they enjoy repeated experiences (O’Brien, 2019), thus overestimating
the speed with which they will adapt to repeated exposures of the same stimulus. Adaptation
also arises in response to repeated exposures of the same stimuli over time, so recipients might
be expected to adapt to the very same compliment over time, but it is not clear that receiving
different compliments would lead to adaptation given that each would be conveying a uniquely
positive message. Finally, adaptation rates vary across stimuli. Concerns about belonging seem
to be a basic need that may need to be repeatedly satisfied, meaning that people may not tire
hearing repeated affirmations about themselves or reassurances that they are valued and liked by
others (Leary & Baumeister, 2000). In a series of experiments following a similar procedure as
Experiment 1 (Zhao & Epley, in press), one person within a pair of close friends, romantic
partners, or family members was randomly assigned to be the expresser and the other person to
be the recipient. Expressers wrote five compliments for their recipient, who was then shown one
compliment each day over the course of the week. People who did not know the details of the
compliments expected recipients to feel progressively less positive over the course of the week.
By contrast, recipients actually felt equally positive over the course of the week, showing no
evidence of adaptation, and generally feeling more positive than both expressers and third-party
observers expected. Considerably more research is needed to understand how expectations of
adaptation may also create a reluctance to be routinely prosocial.

INSUFFICIENTLY COMPLIMENTARY
!
!
53!
Giving compliments in relationships makes both expressers and recipients feel more
positive than they expect to feel, suggesting that people’s relationships might be a little better off
if they were less reluctant to pass along kind thoughts when they had them. Surely there is some
balance to be found between giving compliments too rarely and giving them too frequently. Our
experiments do not identify that critical tipping point. They also do not suggest that people
should spend all of their time passing along every single kind thought they have to others.
Instead, our experiments simply suggest that the optimal frequency of compliment-giving is
likely to be at least a little more often than people may currently be giving.

INSUFFICIENTLY COMPLIMENTARY
!
!
54!
REFERENCES
Abele, A. E., & Wojciszke, B. (2007). Agency and communion from the perspective of self
versus others. Journal of Personality and Social Psychology, 93(5), 751–763.
Bakdash, J. Z., & Marusich, L. R. (2017). Repeated measures correlation. Frontiers in
Psychology, 8(MAR), 1–13.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal
attachments as a fundamental human motivation. Psychological Bulletin, 117(3), 497–529.
Becker, G. S. (1993). The economic way of looking at behavior. Journal of Political Economy,
101(3), 385–409.
Behrens, T. E., Hunt, L. T., & Rushworth, M. F. (2009). The computation of social behavior.
Science, 324(5931), 1160-1164.
Bentham, J. (1825). The Rationale of Reward. London: John and H. L. Hunt.
Bond, C. F., & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and Social
Psychology Review, 10(3), 214–234.
Boothby, E. J., & Bohns, V. K. (2020). Why a Simple Act of Kindness Is Not as Simple as It
Seems: Underestimating the Positive Impact of Our Compliments on Others. Personality
and Social Psychology Bulletin.
Boothby, E., Cooney, G., Sandstrom, G., & Clark, M. (2018). The liking gap in conversation: Do
people like us more than we think? Psychological Science, 29(11), 1742-1756.
Bruk, A., Scholl, S. G., Bless, H., & Gilbert, E. (2018). Beautiful mess effect: Self-other
differences in evaluation of showing vulnerability. Journal of Personality and Social
Psychology, 115(2), 192–205.

INSUFFICIENTLY COMPLIMENTARY
!
!
55!
Chan, E., & Sengupta, J. (2010). Insincere flattery actually works: A dual attitudes perspective.
Journal of Marketing Research, 47(1), 122–133.
Chapman, G. (1995). The five love languages: How to express heartfelt commitment to your
mate. Chicago, IL: Northfield Publishing.
Chaudhry, S. J, & Loewenstein, G. (2019). Thanking, apologizing, bragging, and blaming:
Responsibility exchange theory and the currency of communication. Psychological Review.
Cohn, A., Maréchal, M. A., Tannenbaum, D., & Zünd, C. L. (2019). Civic honesty around the
globe. Science, 365(6448), 70–73.
Cooney, G., Gilbert, D. T., & Wilson, T. D. (2014). The unforeseen costs of extraordinary
experience. Psychological Science, 25(12), 2259–2265.
Crocker, J., Canevello, A., & Brown, A. A. (2017). Social motivation: Costs and benefits of
selfishness and otherishness. Annual Review of Psychology, 68(1), 299–325.
Curry, O. S., Rowland, L. A., Van Lissa, C. J., Zlotowitz, S., McAlaney, J., & Whitehouse, H.
(2018). Happy to help? A systematic review and meta-analysis of the effects of performing
acts of kindness on the well-being of the actor. Journal of Experimental Social Psychology,
76, 320–329.
Decker, J. H., Otto, A. R., Daw, N. D., & Hartley, C. A. (2016). From creatures of habit to goal-
directed learners: Tracking the developmental emergence of model-based reinforcement
learning. Psychological Science, 27(6), 848–858.
DeNeve, K. M., & Cooper, H. (1998). The happy personality: A meta-analysis of 137 personality
traits and subjective well-being. Psychological bulletin, 124(2), 197.
DeSteno, D. (2015). Compassion and altruism: How our minds determine who is worthy of help.
Current Opinion in Behavioral Sciences, 3, 80–83.

INSUFFICIENTLY COMPLIMENTARY
!
!
56!
Diener, E., & Diener, M. (1995). Cross-cultural correlates of life satisfaction and self-esteem.
Journal of Personality and Social Psychology, 68, 653– 663.
Dunn, E. W., Aknin, L. B., & Norton, M. I. (2008). Spending money on others promotes
happiness. Science, 319(5870), 1687–1688.
Dunn, E. W., Aknin, L. B., & Norton, M. I. (2014). Prosocial spending and happiness: Using
money to benefit others pays off. Current Directions in Psychological Science, 23(1), 41–
47.
Emmons, R. A., & McCullough, M. E. (2003). Counting blessings versus burdens: An
experimental investigation of gratitude and subjective well-being in daily life. Journal of
Personality and Social Psychology, 84(2), 377–389.
Epley, N., Keysar, B., Van Boven, L., & Gilovich, T. (2004). Perspective taking as egocentric
anchoring and adjustment. Journal of Personality and Social Psychology, 87(3): 327–39.
Epley, N., & Schroeder, J. (2014). Mistakenly seeking solitude. Journal of Experimental
Psychology: General, 143(5), 1980–1999.
Eyal, T., & Epley, N. (2010). How to seem telepathic: Enabling mind reading by matching
construal. Psychological Science, 21(5), 700–705.
Eyal, T., Steffel, M., & Epley, N. (2018). Perspective mistaking: Accurately understanding the
mind of another requires getting perspective, not taking perspective. Journal of Personality
and Social Psychology, 114(4), 547–571.
Fea, C. J., & Brannon, L. A. (2006). Self-objectification and compliment type: Effects on
negative mood. Body Image, 3, 183–188.
Fehr, E., & Fischbacher, U. (2003). The nature of human altruism. Nature, 425(6960), 785–791.

INSUFFICIENTLY COMPLIMENTARY
!
!
57!
Fiske, S. T., Cuddy, A. J. C., & Glick, P. (2007). Universal dimensions of social cognition:
warmth and competence. Trends in Cognitive Sciences, 11(2), 77–83.
Flynn, F. J., & Lake, V. K. B. (2008). If you need help, just ask: Underestimating compliance
with direct requests for help. Journal of Personality and Social Psychology, 95(1), 128–143.
Fogg, B. J., & Nass, C. (1997). Silicon Sycophants: The effect of computers that flatter.
International Journal of Human-Computer Studies, 46(5), 551–561.
Fredrickson, B. L., Cohn, M. A., Coffey, K. A., Pek, J., & Finkel, S. M. (2008). Open hearts
build lives: Positive emotions, induced through loving-kindness meditation, build
consequential personal resources. Journal of Personality and Social Psychology, 95(5),
1045–1062.
Frederick, S., & Loewenstein, G. (1999). Hedonic adaptation. In D. Kahneman & E. Diener
(Eds.), Well-being: The foundations of hedonic psychology (pp. 302–329). New York, NY:
Russell Sage.
Fujimura-Wilson, K. (2014). A cross-cultural study of compliments and compliment responses in
conversation. English and English and American Literature, 49, 19–36.
Gilovich, T., Savitsky, K., & Medvec, V. H. (1998). The illusion of transparency: Biased
assessments of others’ ability to read one’s emotional states. Journal of Personality and
Social Psychology, 75(2), 332–346.
Goldin-Meadow, S., So, W. C., Özyürek, A., & Mylander, C. (2008). The natural order of
events: How speakers of different languages represent events nonverbally. Proceedings of
the National Academy of Sciences, 105(27), 9163-9168.
Gordon, R. A. (1996). Impact of ingratiation on judgments and evaluations: A meta-analytic
investigation. Journal of Personality and Social Psychology, 71(1), 54–70.

INSUFFICIENTLY COMPLIMENTARY
!
!
58!
Gray, K., & Wegner, D. M. (2008). The sting of intentional pain. Psychological Science, 19(12),
1260–1262.
Izuma, K., Saito, D. N., & Sadato, N. (2008). Processing of social and monetary rewards in the
human striatum. Neuron, 58(2), 284–294.
James, W. (1896/1920). Familiar letters of William James – II. The Atlantic Monthly. Retrieved
from https://www.theatlantic.com/past/docs/issues/96may/nitrous/jamii.htm.
John, L. K., Barasz, K., & Norton, M. I. (2016). Hiding personal information reveals the worst.
Proceedings of the National Academy of Sciences, 113(4), 954–959.
Jones, E. E. (1964). Ingratiation. New York: Appleton-Century-Crofts.
Jones, S. C. (1973). Self- and interpersonal evaluations: Esteem theories versus consistency
theories. Psychological Bulletin, 79(3), 185–199.
Kahneman, D., Wakker, P. P., & Sarin, R. (1997). Back to Bentham? Explorations of
experienced utility. The Quarterly Journal of Economics, 112(2), 375–406.
Knapp, M. L., Hopper, R., & Bell, R. A. (1984). Compliments: A descriptive taxonomy. Journal
of Communication, 34, 12–31.
Korchmaros, J. D., & Kenny, D. A. (2001). Emotional closeness as a mediator of the effect of
genetic relatedness on altruism. Psychological Science, 12(3), 262–265.
Kumar, A., & Epley, N. (2018). Undervaluing gratitude : Expressers misunderstand the
consequences of showing appreciation. Psychological Science, 29(9), 1423–1435.
Kwang, T., & Swann, W. B. (2010). Do people embrace praise even when they feel unworthy? a
review of critical tests of self-enhancement versus self-verification. Personality and Social
Psychology Review, 14(3), 263–280.

INSUFFICIENTLY COMPLIMENTARY
!
!
59!
Lambert, N. M., Clark, M. S., Durtschi, J., Fincham, F. D., & Graham, S. M. (2010). Benefits of
expressing gratitude. Psychological Science, 21(4), 574–580.
Leary, M. R. (2007). Motivational and emotional aspects of the self. Annual Review of
Psychology, 58, 317-344.
Leary, M. R., & Baumeister, R. F. (2000). The nature and function of self-esteem: Sociometer
theory. Advances in Experimental Social Psychology, 32, 1–62.
Levine, E. E., & Cohen, T. R. (2018). You can handle the truth: Mispredicting the consequences
of honest communication. Journal of Experimental Psychology: General, 147(9), 1400–
1429.
Lyubomirsky, S., & Layous, K. (2013). How do simple positive activities increase well-being?
Current Directions in Psychological Science, 22(1), 57–62.
Marigold, D. C., Holmes, J. G., & Ross, M. (2007). More than words: Reframing compliments
from romantic partners fosters security in low self-esteem individuals. Journal of
Personality and Social Psychology, 92(2), 232–248.
Miller, D. T., & Prentice, D. A. (1994). Collective errors and errors about the collective.
Personality and Social Psychology Bulletin, 20(5), 541–550.
Nordgren, L. F., Banas, K., & Macdonald, G. (2011). Empathy gaps for social pain: Why people
underestimate the pain of social suffering. Journal of Personality and Social Psychology,
100(1), 120–128.
Nowak, M. A. (2006). Five rules for the evolution of cooperation. Science, 314(5805), 1560–
1563.
O’ Brien, E. (2019). Enjoy it again: Repeat experiences are less repetitive than people think.
Journal of Personality and Social Psychology, 116(4), 519–540.

INSUFFICIENTLY COMPLIMENTARY
!
!
60!
Ong, D. C., Goodman, N. D., & Zaki, J. (2018). Happier than thou? A self-enhancement bias in
emotion attribution. Emotion, 18(1), 116–126.
Pinker, S. (2007). The stuff of thought: Language as a window into human nature. New York,
NY: Penguin.
Read, D. (2007). Experienced utility: Utility theory from Jeremy Bentham to Daniel Kahneman.
Thinking and Reasoning, 13 (1), 45-61.
Ross, M., & Sicoly, F. (1979). Egocentric biases in availability and attribution. Journal of
Personality and Social Psychology, 37(3), 322–336.
Ronson, J. (2015). So you’ve been publicly shamed. New York, NY: Penguin.
Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic
motivation, social development, and well-being. The American Psychologist, 55(1), 68–78.
Sandstrom, G. M., & Dunn, E. W. (2014). Social interactions and well-being: The surprising
power of weak ties. Personality and Social Psychology Bulletin, 40(7), 910–922.
Seiter, J. S. (2007). Ingratiation and gratuity: The effect of complimenting customers on tipping
behavior in restaurants. Journal of Applied Social Psychology, 37(3), 478–485.
Slobin, D. I. (1996). From “thought and language” to “thinking for speaking”. In J. J. Gumperz
& S. C. Levinson (Eds.), Rethinking linguistic relativity: Studies in the social and cultural
foundations of language, No. 17 (pp. 70–96). New York: Cambridge University Press.
Swann, W. B., Griffin, J. J., Predmore, S. C., & Gaines, B. (1987). The cognitive-affective
crossfire: When self-consistency confronts self-enhancement. Journal of Personality and
Social Psychology, 52(5), 881–889.
Tamir, D. I., & Mitchell, J. P. (2013). Anchoring and adjustment during social inferences.
Journal of Experimental Psychology: General, 142(1), 151–162.

INSUFFICIENTLY COMPLIMENTARY
!
!
61!
Twain, M. (1907/2010). Mark Twain’s own autobiography: The chapters from the north
american review. Madison, WI: University of Wisconsin Press.
Twain, M. (1966). Mark Twain’s satires and burlesques. Berkeley, CA: University of California
Press.
Rees-Miller, J. (2011). Compliments revisited: Contemporary compliments and gender. Journal
of Pragmatics, 43, 2673–2688.
Van Boven, L., Loewenstein, G., Dunning, D., & Nordgren, L. F. (2013). Changing places: A
dual judgment model of empathy gaps in emotional perspective taking. Advances in
Experimental Social Psychology, 48, 117–171.
Van Lange, P. A. M., Otten, W., De Bruin, E. M. N., & Joireman, J. A. (1997). Development of
prosocial, individualistic, and competitive orientations: Theory and preliminary evidence.
Journal of Personality and Social Psychology, 73(4), 733–746.
Vonk, R. (2001). Aversive self-presentations. In R. M. Kowalski (Ed.), Behaving badly: Aversive
interpersonal behaviors (pp. 79–155). Washington, DC: American Psychological
Association.
Vonk, R. (2002). Self-serving interpretations of flattery: Why ingratiation works. Journal of
Personality and Social Psychology, 82(4), 515–526.
Westphal, J. D., & Stern, I. (2007). Flattery will get you everywhere (Especially if you are a
male caucasian): How ingratiation, boardroom behavior, and demographic minority status
affect additional board appointments at U.S. companies. Academy of Management Journal,
50(2), 267–288.
Wilson, T. D., & Gilbert, D. T. (2005). Affective forecasting: Knowing what to want.
Psychological Science, 14(3), 131–134.

INSUFFICIENTLY COMPLIMENTARY
!
!
62!
Wilson, T. D., & Gilbert, D. T. (2008). Explaining away: A model of affective adaptation.
Perspectives on Psychological Science, 3, 370– 386.
Wojciszke, B., Bazinska, R., & Jaworski, M. (1998). On the dominance of moral categories in
impression formation. Personality and Social Psychology Bulletin, 24(12): 1251–1263.
Zhao, X., & Epley, N. (in press). Kind words do not become tired words: Undervaluing the
positive impact of frequent compliments. Self and Identity.

Supplemental Materials for
Insufficiently Complimentary?: Underestimating the Positive Impact
of Compliments Creates a Barrier to Expressing Them
This PDF file includes:
Introductory Surveys: Compliments in Everyday Life ..................................................................... 2
Surveys 1-4: Comparing to Ideal Standards ................................................................................ 2
Survey 5: Expressing vs. Withholding Compliments .................................................................. 6
Supplemental Experiments .............................................................................................................. 9
Supplemental Experiments S1a and S1b: Third-Person Observers ............................................ 9
Supplemental Experiment S2: Correlating Expectations with Intentions ................................. 15
Supplemental Experiment S3: To Share, or Not to Share? ....................................................... 19
Supplemental Experiment S4: Updating Expectations to Change Intentions? ......................... 24
Additional Analysis ........................................................................................................................ 30
Experiment 1 ............................................................................................................................. 30
Experiment 2 ............................................................................................................................. 32
Experiment 3 ............................................................................................................................. 34
Experiment 4 ............................................................................................................................. 35
References ...................................................................................................................................... 37
All data, materials, and preregistration forms can be accessed via a repository at the Open
Science Framework: https://osf.io/ypk5g/?view_only=47a263380604485a803db70007930d42

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
2
Introductory Surveys: Compliments in Everyday Life
Surveys 1-4: Comparing to Ideal Standards
Do people express the compliments that come to their minds as often as they should to
maximize both their own and others’ wellbeing? The results of four online surveys suggested
that withholding compliments might be somewhat widespread even while most people report
thinking that they should, or would like to, express compliments more often than they do
currently.
Methods
In four surveys, participants were recruited on Amazon Mechanical Turk (Ns = 96, 97,
100, & 193, Mage = 35.92, SDage = 10.96, 61% female, U.S. residents) and rated how often they
engaged in a variety of prosocial activities with another person in their personal life to whom
they felt close (Surveys 1, 2, and 4) or with whom they had a satisfying or dissatisfying
relationship (Survey 3; relationship satisfaction presented as a within-participants factor with
order counterbalanced). All surveys included seven prosocial behaviors: giving compliments
(“praise or validate their qualities, behaviors, or merits”), expressing gratitude (“show your
appreciation for what they have done for you”), expressing support (“listen to them, validate
their feelings, or simply be there to show your care”), giving advice (“suggest ideas, solutions, or
resources to help them deal with challenging situations”), taking perspective (“consider a
specific situation or the world more general from their unique point of view”), being completely
honest (“state your true opinions, thoughts, or feelings to them in complete honesty”), and
providing helpful criticism (“offer observations and solutions to help them make improvements
in particular areas”). In addition, Survey 1 also included three potentially negative behaviors that
were absent from other surveys—i.e., providing unhelpful criticism (“express negative emotions

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
3
and disapproval of their behaviors, thoughts, or characters”), telling prosocial lies (“make false
statements with the intention of misleading and benefitting them, such as making them feel
better”), and telling selfish lies (“make false statements with the intention to advance your self-
interest, such as protecting your image or avoiding making them angry at you”).
For each of these behaviors, participants reported how often they engaged in each
behavior compared to how often they thought they should (Surveys 1–4), or would like to
(Survey 4 only), on scales ranging from -3 (a lot less often than I think I should [would like to])
to 3 (a lot more often than I think I should [would like to]; In Survey 1, participants also had the
option “I have never done this to them,” which was frequently selected in response to the three
potentially undesirable behaviors and was eliminated in later surveys). Next, participants
reported how often the other person engaged in each behavior compared to how often they
thought the other person should (Surveys 1–4), or how often they would like the other person to
(Survey 4 only). In Surveys 1–3, participants also reported how they expected their relationship
quality with the other person to change if they or the other person engaged in each of the
aforementioned activities more often. Finally, recipients reported their gender, age,
race/ethnicity, and highest education level.
Results and Discussion
As shown in Figures S1, S2, and S3, participants across surveys consistently reported
giving compliments significantly less often than they should (p’s < .001) or would like to (p
= .003) compared to their ideal level. Comparing people’s self-assessment on compliment-
giving against other prosocial behaviors also revealed that people believed that their compliment-
giving behavior suffered from one of the biggest gaps between ideals and reality among all the
prosocial behaviors. For instance, people believed that they should express gratitude and

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
4
emotional support more often as well, but the deficit appeared to be somewhat less severe than
giving compliments. And people often indicated that they should not give advice or express
criticisms—even helpful ones—more often than they already did. Finally, comparing people’s
assessments of their own behaviors (dark bars in Figures S1–S3) against their assessments of
other people’s behaviors (light bars in Figures S1–S3) shows that people thought they had a
greater deficit than other people in compliment-giving, yet such a self-critical view disappeared
when they considered relationships that were dissatisfying in nature.
Figure S1. Participants’ assessments of their own (dark bars) and other people’s (light bars)
behavior frequency in Surveys 1 and 2 (data combined). Means are displayed in the graph and in
the table below. Scales range from -3 (a lot less often than I think I/the other person should) to 3
(a lot more often than I think I/the other person should) with 0 (exactly as often as I think I/the
other person should) as the midpoint. Error bars reflect ± 1 standard errors. Asterisks in the
table indicate significance tests against the midpoint of 0 (the ideal standard), and asterisks above
the bars indicate comparisons between one’s own and other people’s behaviors. * p < .05. ** p
< .01. *** p < .001. Same as below.
Give
compliments
Express
gratitude
Express
support Give advice Take
perspective
Give helpful
critcism
Be
completely
honest
Self behavior -0.508 -0.365 -0.23 0.016 -0.161 -0.121 0.041
Other behavior -0.109 -0.169 -0.135 0.021 -0.414 -0.212 -0.164
-1.5
-1
-0.5
0
0.5
***
***
**
*
*
***
*
*
***
**

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
5
Figure S2. Participants’ assessments of their own and other people’s behavior frequency in
satisfying (Panel A) and unsatisfying relationships (Panel B) in Survey 3.
Figure S3. Participants’ assessments of their own and other people’s behavior frequency in the
“should” condition (Panel A) and the “would like to” condition (Panel B) in Survey 4.
Give
compliments
Express
gratitude
Express
support Give advice Take
perspective
Give helpful
critcism
Complete
honesty
Self behavior -0.43 -0.37 -0.28 0.06 -0.01 -0.12 -0.11
Other behavior -0.14 -0.18 -0.16 0 -0.23 -0.11 -0.04
-1
-0.5
0
0.5
Give
compliments
Express
gratitude
Express
support Give advice Take
perspective
Give helpful
critcism
Complete
honesty
Self behavior -0.63 -0.52 -0.46 -0.42 -0.56 -0.48 -0.67
Other behavior -0.95 -1.23 -1.1 -0.28 -1.53 -0.75 -0.67
-2
-1.5
-1
-0.5
0
Give
compliments
Express
gratitude
Express
support Give advice Take
perspective
Give helpful
critcism
Complete
honesty
Self behavior -0.474 -0.242 -0.105 0.200 0.032 -0.084 -0.042
Other behavior -0.137 -0.105 -0.011 0.232 -0.326 0.042 0.095
-1
-0.5
0
0.5
Give
compliments
Express
gratitude
Express
support Give advice Take
perspective
Give helpful
critcism
Complete
honesty
Self behavior -0.367 -0.449 -0.184 -0.153 -0.214 -0.286 -0.225
Other behavior -0.122 -0.082 -0.276 0.071 -0.388 0.000 -0.061
-1
-0.5
0
0.5
***
** **
(A)
(B)
(A)
(B)
***
**
*
***
***
***
***
**
**
***
***
***
***
***
***
***
*
***
***
***
***
**
*
*
*
**
***
**
*

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
6
Survey 5: Expressing vs. Withholding Compliments
People may report thinking they should, or would like to, express compliments more
often, but it is unclear whether this is because people cannot generate positive thoughts about
another person in the first place or because they sometimes refrain from expressing a compliment
that they already have in mind. To understand whether people withhold expressing compliments
even when they already have them in mind, we conducted another online survey on to measure to
what extent people refrained from communicating positive and negative social evaluations to the
recipients in the form of expressing compliments or criticisms.
Methods
We recruited participants on MTurk and targeted a sample size of 100 after exclusion,
and a total of 118 participants completed this study (Mage = 36.52, SDage = 10.58, 31% female).
Among them, 15 were excluded because they failed to pass our preregistered attention check (we
requested that percentage estimates on expressing vs. withholding a thought should add up to
100%, yet these participants’ estimates considerably deviated from this requirement in both
positive or negative thoughts, thus revealing a lack of attention to our study instruction.)
In this survey, we first asked participants to identify one person to whom they were close
and then report their opinions on how often they expressed or withheld their positive or negative
thoughts about that person. Specifically, participants were prompted to consider that, “Every
time when you see [name] or think of [name] and a positive characteristic or action comes to
your mind, you can either choose to share that positive thought by giving a compliment to
[name], or you might not share it with [name] for a variety of different reasons.” Then, they
responded to two separate items about sharing versus withholding (i.e., “When a positive

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
7
characteristic or action comes to mind about [name], what percentage of the time do you [share it
by giving a compliment to/keep it to yourself and not share it with] [name]?” ) on scales from
0% to 100% with a 10% increment. They were also prompted to confirm that their percentages
of expressing versus withholding a thought added up to 100%, although this rule was not forced
due to technical limitations. Next, similar to other surveys, participants reported a) how often
they complimented the other person compared to how often they thought they should, and b)
how they expected engaging in this behavior more often would change their relationship.
Questions related to criticism were phrased using identical formats and were grouped into a
separate block, and the presentation order of the compliment block and the criticism block was
counterbalanced across participants.
Results and Discussion
One of the primary research questions we aimed to answer is whether people ever refrain
from expressing their positive and negative evaluations about other people when such thoughts
occurred to them. To this end, we compared people’s estimated percentages of verbal expression
against the null hypothesis of people expressing thoughts 100% of the time. Results confirmed
that participants indeed withhold both compliments and criticisms from close others, reporting
that they expressed their compliments 63.6% of the time on average (SD = 25.9%, t(102) = -
14.27, p < .001) and their criticism 37.8% of the time (SD = 26.4%, t(102) = -23.94, p < .001).
Comparing between compliments and criticisms further suggested that people expressed their
compliments more often than their criticisms, t(102) = 8.62, p < .001, d = 0.85.
Similar to Surveys 1–4, we tested the extent to which people believed that they expressed
compliments and criticisms as often as they should. However, unlike the results in any other
pretest surveys (Surveys 1–4) or experiments (Experiments 1a and 2, Supplemental Experiment

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
8
S3), participants reported giving compliments to the other person approximately as often as they
thought they should and did not perceive a deficit of compliment-giving (M = -0.18, SD = 1.38,
t(102) = -1.35, p = .18). They also reported giving criticisms as often as they thought they
should (M = -0.22, SD = 1.41, t(102) = -1.60, p = .11), and we found no significant difference
between expressing compliments and expressing criticisms, t(102) = 0.19, p = .85. In light of the
consistent evidence in other surveys and experiments that people believed they did not give
enough compliments in their everyday lives, the lack of a similar result in this survey is
surprising. We can only speculate at this point about the factors responsible for this discrepancy.
One caveat about the percentage estimates in this survey is that we only asked people to
consider one person they felt close to and estimate the extent to which they would share their
thoughts with that specific person. However, people are more likely to share their thoughts
freely with close others than with the majority of people in their everyday lives. Therefore, the
percentages of thoughts people would withhold from most other people in their social lives are
likely to be considerably higher than the numbers estimated in this survey.
Taken together, our results showed that even within the most intimate relationships and
during the most intimate times, people still forgo some opportunities to express compliments to
others they are close to.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
9
Supplemental Experiments
Supplemental Experiments S1a and S1b: Third-Person Observers
Participants in Supplemental Experiment S1a read compliments written by expressers in
Experiment 1nd predicted how their recipients would feel after reading them. Participants in
Experiment S1b did the same after being asked to explicitly adopt either the expresser’s or the
recipient’s perspective. Although it would have been optimal to randomly assign these
participants to the role of observer as part of Experiment 1, we think comparisons with the
expressers and recipients from Experiment 1 are reasonable given that we recruited from exactly
the same location, at the same time of day, and during the same days of the week. Nevertheless,
this methodological imperfection is worth acknowledging upfront.
We predicted that observers in Experiment S1a would underestimate how positive the
actual recipients from Experiment 1would feel, just as the expressers did. Experiment S1b tested
whether this miscalibration could be reduced by explicitly trying to put oneself in the recipient’s
shoes through perspective taking (Eyal, Steffel, & Epley, 2018), or whether the positive
experience of receiving a compliment was uniquely felt by the actual recipient.
Method
Participants. We conducted Supplemental Experiments S1a and S1b in the same public
park as Experiment 1nd employed a similar recruitment setup, except that we recruited
individuals rather than pairs. We targeted a sample size of 104 participants for each
experiment—two yoked to each of the 52 pairs in the compliment condition in Experiment 1.
Four participants in both experiments were later excluded from data analyses because they were
either yoked to a pair that already had two observers or ended up being the only observer for the
yoked pair (due to a coordination error among multiple experimenters), yielding 100 participants

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
10
in the final sample for both Experiment S1a (Mage = 37.82, SDage = 16.97, rangeage = 18–78; 58%
female) and Experiment S1b (Mage = 36.17, SDage = 12.67, rangeage = 19–86; 64% female).
Experiment S1a Procedure. Participants each received a tablet to begin the study in
private. Their survey first summarized the compliment condition from Experiment 1(including
the exact compliment-writing prompt) and then presented the three compliments written by the
expresser from their yoked pair. Next, participants predicted how the recipient of those
compliments would report feeling using the same items from Experiment 1. Because we thought
some participants might have difficulty recalling the compliments after only seeing them once,
but we also wanted to replicate the procedure from Experiment 1s closely as possible, we
experimentally manipulated whether participants saw the compliments only once or saw them
next to each prediction item in the survey. This manipulation had no meaningful influence on
participants’ responses to any of the items (paired ts < 1.25, ps > .22), so we collapsed responses
across these conditions in the following analyses.
Experiment S1b Procedure. The procedure was identical to that in Experiment S1a
except that all compliments were presented only once, and that we experimentally manipulated
the perspective participants were encouraged to adopt. Specifically, we asked participants to
vividly imagine that they were participating in the described study as either the expresser or the
recipient. Accordingly, participants who played the role of an imagined expresser predicted the
yoked recipient’s responses while participants who played the role of an imagined recipient
reported how they thought they would have responded if they had been the recipient.
Participants also reported how difficult it was to imagine being the expresser or recipient on a
scale from 0 (not at all difficult) to 10 (extremely difficult), and the extent to which they felt able
to imagine being the expresser or recipient on a scale from 0 (not at all able) to 10 (extremely

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
11
able). Participants reported have relatively little difficulty imagining their roles and did not
differ significantly between expresser and recipient conditions (Ms = 2.26 & 2.24, SDs = 2.81 &
2.57, respectively), t = .04, p = .97, and they also felt able to imagine their assigned perspective
across conditions (Ms = 7.62 & 7.88, SDs = 2.76 & 2.09, respectively), t = -.53, p = .60.
Results
Because participants in both Experiments S1a and S1b were yoked to pairs of
compliment expressers and recipients in Experiment 1, we conducted paired t-tests to compare
their responses to those of the actual expressers and recipients.
Positive mood. As shown in Figure S4(A), neutral observers in Experiment S1a
significantly underestimated how positive the recipients would feel, t(49) = -3.35, p = .002, d = -
0.47, just as the expressers in Experiment 1 did.
In Experiment S1b, participants also significantly underestimated how positive the
recipients would feel both when imagining themselves as expressers, t(49) = -2.27, p = .027, d =
-.32, and when imagining themselves as recipients, t(49) = -3.46, p = .001, d = -.49. In fact,
their expectations did not differ significantly from those of the actual expressers, ts < 1, and
manipulating participants’ perspective as the expresser or the recipient did not significantly
influence their expectations, either, t(49) = 1.01, p =.32, d = .14. These results suggest that the
strong positive mood that results from receiving a compliment could not be recognized by simply
trying to imagine oneself in the recipient’s perspective.
Awkwardness. As shown in Figure S4(B), neutral observers in Experiment S1a
significantly overestimated how awkward the recipients would feel, t(49) = 4.74, p < .001, d =
0.67, just as the expressers in Experiment 1 did. In fact, observers’ expectations did not differ
significantly from the actual expressers’ expectations from Experiment 1, ts < 1.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
12
However, results were somewhat different for expected awkwardness in Experiment S1b.
Imagined expressers overestimated how awkward the actual recipients from Experiment 1would
feel, t(49) = 2.92, p = .005, d = .41, which did not differ significantly from the actual expressers
from Experiment 1, t(49) = 0.33, p = .74, d = .05. Imagined recipients, in contrast, were
calibrated in their expectations of how awkward the actual recipients would actually feel in
Experiment 1, t(49) = -0.10, p = .92, d = -.02. As a result, those who imagined being expressers
anticipated that their recipient would feel more awkward than those who imagined being
recipients, t(49) = 3.28, p = .002, d = .46. This unexpected divergence in the effect of
perspective on expectations of positive mood versus awkwardness again suggests—as we
indicated in Experiment 1—that expectations of a recipient’s positive mood may be influenced
more by the precise content of the compliments, while expectations of awkwardness may be
influenced more by context in which the compliments are delivered.
Figure S4. Positive mood (Panel A) and awkwardness (Panel B) predicted by neutral observers
in Experiments S1a and perspective-taking observers in Experiment S1b compared to those by
the actual expressers and recipients in the compliment condition in Experiment 1. Error bars
reflect ± 1 standard errors.
5
6
7
8
9
10
Observer (Exp.3a) Imagined-E Imagined-R Expected Actual
Positive Mood
5
6
7
8
9
10
Imagined (Exp.3b) Actual (Exp.1)
Positive Feeling
Expresser
Recipient
10
9
8
7
6
5
Positive Mood
Expresser
Recipient
Neutral
(Exp. S1a)
Imagined
(Exp. S1b)
Actual
(Exp. 1)
10
9
8
7
6
5
(A)
0
1
2
3
4
5
Observer (Exp.3a) Imagined-E Imagined-R Expected Actual
Awkward Feeling
0
1
2
3
4
5
Imagined (Exp.3b) Actual (Exp.1)
Expresser
Recipient
Awkwardness
Neutral
(Exp. S1a)
Imagined
(Exp. S1b)
Actual
(Exp. 1)
10
9
8
7
6
5
Expresser
Recipient
(B)

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
13
Warmth and competence. Observers in Experiment S1a significantly underestimated
how warm the actual recipients would perceive the compliments to be (M = 7.13, SD = 1.44),
t(49) = -8.64, p < .001, d = -1.22, and also expected them to be perceived as less warm than the
actual expressers expected, t(49) = -3.49, p = .001, d = -.49. Observers also significantly
underestimated how competent the actual recipients would perceive the compliments to be (M =
6.82, SD = 1.40), t(49) = -6.00, p < .001, d = -.85, but did not differ from the actual expressers’
expectations of perceived competence, t(49) = 0.73, p = .001.
Participants’ expected evaluations of warmth and competence in Experiment S1b did not
differ significantly based on the perspective they adopted, ts (49) = 1.28 and .48, respectively,
ps > .2, ds = .18 and -.07. Imagined expressers (M = 7.73, SD = 1.83) and imagined recipients
(M = 7.25, SD = 2.03) both significantly underestimated how warm the actual recipients would
perceive the compliments to be, ts(49) = -4.55 and -6.30, ps < .001, d = -.64, but imagined
expressers did not differ significantly from the actual expressers in Experiment 1, t(49) = -1.20,
p = .24, d = -.17. Imagined expressers (M = 7.01, SD = 2.33) and imagined recipients (M = 7.21,
SD = 2.00) also underestimated how competent the actual recipients would perceive the
compliments to be, ts(49) = -3.68 and -4.60, ps < .001, d = -.64, but imagined expressers again
did not differ from the actual expressers in Experiment 1, t(49) = 0.85, p = .40, d = .12.
Mediation analyses. Exploratory mediation analyses further showed that in both third-
person observer studies, underestimating the perceived warmth of those compliments—but not
the perceived competence—could account for a statistically significant proportion of variance in
the discrepancy between observers’ expected and recipients’ reported positive mood. See Figure
S5(A) for mediation results in Experiment S1a and Figure S5(B) for mediation results in
Experiment S1b.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
14
Figure S5. Results from the exploratory mediation analysis in Experiment S1a (neutral observer;
Panel A) and Experiment S1b (perspective-taking observer; Panel B).
Discussion
When third-person observers predicted how the compliments would make the recipients
feel by either taking a neutral perspective (Experiment S1a) or imagining themselves as the
expressers (Experiment S1b), they underestimated the positive impact of receiving compliments
just like the actual expressers did. Even those who explicitly imagined being the recipients of
the compliments (Experiment S1b) still failed to fully appreciate how positive the actual
recipients felt. Such results supported our hypothesis that compliments are kind words
specifically targeted at recipients and are therefore difficult to anticipate and make recipients feel
uniquely positive.
In addition, expressers in both Experiments S1a and S1b overestimated the potentially
negative outcomes of giving compliments (i.e., feeling awkward), while those who imagined
receiving the compliments reported feeling a similar level of awkwardness as the actual
recipients. We did not anticipate this result, but it again suggests that evaluations of a recipient’s
positive mood and awkwardness might be guided by somewhat different processes.
Perspective
(0 = observer;
1 = recipient)
underestimating
War m t h
underestimating
Competence
underestimating
Positive Mood
1.15 [.43, 2.36]
2.02***
(c’ = -.73, p= .018)
c= .86, p= .002
1.73***
.57***
.26*
.44 [-.14, 1.17]
(A)
Perspective
(0 = observer;
1 = recipient)
underestimating
War m t h
underestimating
Competence
underestimating
Positive Mood
1.32 [.62, 2.81]
1.94***
(c’ = -.47, p= .11)
c= .94, p= .001
1.37***
.68***
.06
.09 [-.29, .68]
(B)

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
15
Supplemental Experiment S2: Correlating Expectations with Intentions
If expectations about a recipient’s reaction at least partly guide people’s interest in
expressing compliments, then they should be more interested in sharing a compliment with
someone they expect will respond favorably. We tested this hypothesis in Experiment S2 by
asking participants to generate compliments for several different people in their lives, to predict
how positive and awkward each recipient would feel as well as how warm and competent their
compliments would be perceived, and then to state how interested they were in delivering their
compliment to each person. We predicted that participants’ anticipated reactions would be
related to their interest in expressing the compliment. Specifically, people would be more
interested in expressing a compliment if they expected the compliment to make the recipient feel
more positive and less awkward. Testing this hypothesis is critical for establishing the
plausibility of our broader hypothesis that miscalibrated expectations about others’ reactions
could lead people to deliver fewer compliments than might be optimal in their daily lives.
Method
Participants. Online participants were recruited for a study on interpersonal
communication from the same population as in Experiment 2. Participants completed this
experiment in exchange for $3. We aimed to recruit 100 participants and received completed
data from 97 (Mage = 31.61, SDage = 12.85, rangeage = 19–68; 75% female); three other
participants signed up but did not start the study.
Procedure. Participants were first instructed to think of people in their daily lives that
they thought highly of in each of five different categories (i.e., friends, family,
colleagues/classmates, mentors/advisors, and acquaintances). Participants were prompted to
identify one target person in each category by writing down that person’s first name or initials.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
16
After identifying all five targets (order randomized), participants read the same compliment-
writing prompt as in Experiment 1 and typed one compliment for each recipient in a separate
textbox. On the next pages, participants read each compliment they just wrote, imagined
expressing that compliment to the target, indicated “how warm and sincere” and “how articulate
and well-written” they expected the target to perceive the compliment to be. Next, they reported
how they expected the recipient to feel on the same items in Experiment 1 (i.e.,
positive/negative, pleasant, awkward), and how interested they would be in expressing the
compliment on a scale ranging from 0 (not at all interested) to 10 (extremely interested). After
completing this same procedure for each of the targets, participants reported their demographic
information and were debriefed.
Results
Positive mood and awkwardness. To examine how participants’ expectations on the
target’s positive mood and awkwardness were related to their compliment-giving intentions, we
constructed two linear mixed models with interest in expressing compliments as the outcome
variable, participant-specific intercepts as random effects, and either expected positive mood or
expected awkwardness as the predictor. This analysis allows us to identify the common within-
participant association (regression slope) across five relationship categories among participants.
We found that expressers’ expectations of the recipients’ positive mood significantly
predicted how interested they were in expressing their compliments, β = .48, t = 12.88, p < .001.
Expectations of how awkward the recipient would feel, in contrast, negatively predicted their
behavioral intention, β = .27, t = -6.21, p < .001. These relationships supported our hypotheses
that the more positive and less awkward people expected their compliments to make their
recipients feel, the more interested they were in expressing their compliment.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
17
Because expressers’ expectations on how positive and how awkward their recipients
would feel were also correlated (r = -.25, p < .001), we next constructed a linear mixed model
with both expected positive mood and expected awkwardness as simultaneous predictors to
estimate their unique contribution in expressers’ compliment-giving intentions. We found a
significant effect of both expected positive mood, β = .44, t = 11.83, p < .001, and expected
awkwardness, β = -.17, t = -4.27, p < .001, suggesting that both expectations uniquely predict
participants’ interest in expressing a potential compliment. More importantly, positive mood
showed a considerably larger impact than awkwardness, as revealed by its larger standardized
regression coefficient, difference score = .28, 95% CI = [.16, .39], t = 4.65, p < .001.
Warmth and competence. Next, we examined how expressers’ expectations of the
perceived warmth and competence of a compliment were related to expectations of how positive
and how awkward the recipient would feel upon receiving it. Entering warmth or competence as
the predictor in a linear mixed model showed that both significantly predicted how positive
expressers expected their recipients to feel (warmth: β = .69, t = 21.41, p < .001; competence: β
= .62, t = 12.48, p < .001); in addition, both warmth or competence significantly predicted how
awkward expressers expected their recipients to feel (warmth: β = -.20, t = -4.66, p < .001;
competence: β = -.20, t = -4.16, p < .001). Furthermore, consistent with results in Experiments 1
and 2, both warmth and competence evaluations showed stronger impacts on expected positive
mood than on expected awkwardness, ps < .001.
Finally, to assess how expected warmth and competence evaluations might ultimately
influence how interested people would be in expressing a compliment, we calculated the
relationship between expected warmth and competence evaluations and compliment-giving
intention in separate linear mixed models. We found that expressers’ expectations on both how

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
18
warm and how competent their compliment would be seen was both strongly correlated with
their intention in expressing it (warmth: β = .51, t = 13.81, p < .001; competence: β = .51, t =
12.48, p < .001). Further, to examine potential relationships among expected warmth and
competence evaluations of a compliment, expected positive mood and awkwardness of the
recipient, and people’s behavioral intention in expressing the compliment, we preregistered a set
of analysis that employed a mediation-style model with positive mood and awkwardness as
simultaneous “mediators” and warmth and competence as the “causal variables.” This set of
analysis conceptually followed the four steps of a mediation analysis (Kenny, 2018) but
employed linear mixed models to calculate standardized regression coefficients, accounting for
the repeated measures. Our results showed that how warm people expected their compliment to
be seen had a significant “direct effect” on their interest in expressing it (β = .31, t = 6.26, p
< .001). Similarly, expected competence also had a significant “direct effect” on behavioral
intention (β = .29, t = 6.19, p < .001) (see Figures S6(A) and S6(B)). It is important to
recognize, however, that our research method and statistical model cannot establish a causal
model (Fiedler et al., 2011; 2018); the results only showed that the data is consistent with a
causal model based on our theorization.
Warmth
Positive Mood
Awkwardness
Behavioral
Intention
c= .51, t = 13.81, p< .001
c' = .31, t = 6.26, p< .001
a= .69, t = 21.41,
p< .001
a= -.20, t = -4.66,
p< .001
b= .24, t= 78,
p< .001
b= -.14, t= -3.85,
p< .001
(A)

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
19
Figure S6. Relationship among expected warmth and competence evaluation, expected positive
mood and awkwardness, and compliment-giving intention in Supplemental Experiment S2.
Standardized regression coefficients were calculated from linear mixed models.
Discussion
Consistent with our hypothesis, people’s expectations of how positive and how awkward
their compliments would make the recipient feel were correlated with their intention to express
the compliment. Furthermore, expected positive mood showed a much stronger relationship than
expected awkwardness. We made no predictions about which of these effects would be larger,
and we do not know from a single experiment whether this result is reliable or idiosyncratic to
the hypothetical context and procedure. Supplemental Experiment S3 thus provides another
test—with actual compliment-giving behavior—to compare the extent to which expected
positive mood and expected awkwardness guide potential expressers’ decisions to express (or
withhold) compliments.
Supplemental Experiment S3: To Share, or Not to Share?
In this experiment, we asked one person in an acquainted pair to generate three
compliments about the other person, predict how the recipient would react to each compliment,
and then choose only one compliment to share. We compared each expresser’s expected positive
Competence
Positive Mood
Awkwardness
Behavioral
Intention
c= .51, t = 12.48, p< .001
c' = .29, t = 6.19, p< .001
a= .62, t = 15.04,
p< .001
a= -.20, t = -4.16,
p< .001
b= .31, t= 7.12,
p< .001
b= -.14, t= -3.73,
p< .001
(B)

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
20
mood and awkwardness of the one compliment they decided to share against the other two they
withheld to assess how expectations might be related to decisions about sharing versus
withholding compliments.
Method
Participants. We conducted this experiment in the same public park, and using the same
procedure, as Experiment 1. We targeted a sample of 50 pairs and recruited through the end of
our last scheduled shift as we approached that target, yielding a total of 106 participants (Mage =
38.86, SDage = 15.00, rangeage = 19–76; 61% female). We excluded three additional pairs due to
their having misunderstood the experimental procedure (n = 2) or left the park before completing
the experiment (n = 1).
Procedure. Participants in each pair were first randomly assigned to be either the
compliment expresser or the recipient. The experimenter then gave a tablet to the expresser to
begin the study while instructing the recipient to wait out of sight until the expresser finished.
The expresser then started reading the experimental materials on the tablet, which began with an
introduction page explaining that their task was to write down three compliments they could give
to the recipient and then choose one to share. Expressers received the same compliment-writing
prompt as in Experiment 1 and typed three separate compliments into three textboxes. They
were then presented with each compliment on a separate page and reported how they expected
their recipient would feel after reading each compliment on the same items as expressers in
Experiment 1(i.e., positive/negative, pleasant, awkward). After responding to all three
compliments, expressers were prompted to choose one compliment to share with their recipient.
Finally, expressers completed a short survey measuring their relationship satisfaction,
compliment frequency, and demographic information, as in Experiment 1.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
21
The experimenter then brought the recipient back while the expresser moved out of sight,
seated the recipient in private, and handed them the tablet. After reading the expresser’s
compliment-writing prompt, the recipient was presented with the compliment that their expresser
decided to share, reported how they felt on the same items predicted by the expressers (rephrased
when necessary), and finished with a short survey measuring their relationship satisfaction,
compliment frequency, and demographic information.
Results
Expressed vs. withheld compliments. We compared expressers’ expectations of their
recipients’ reactions between the shared and the withheld compliments. To account for the
correlation among expectations reported by the same expresser, we conducted a mixed-model
ANOVA with one a priori Helmert contrast that compared the mood expectations for the one
expressed compliment against the two withheld compliments. Consistent with our hypothesis,
expressers expected that their shared compliment would make the recipients feel more positive
(M = 7.89; SD = 1.42) than the compliments they did not share (M = 7.08; SD = 1.48), F(1,52) =
23.83, p < .001, η2p = .31 (see Figure S7(A)). In contrast, expressers did not expect their shared
compliments to make the recipient feel less awkward than the unshared compliments (Ms = 2.72
and 2.45; SDs = 2.69 and 2.63, respectively), F(1,52) = 1.29, p = .26 (see Figure S7(B)). The
absence of an effect on awkwardness again suggests that expectations about awkwardness are
not guided by the specific content of the compliment being expressed, but more likely about the
context in which the compliments are shared, in which case each of the compliments would be
expected to feel similarly awkward.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
22
Figure S7. Expressers’ expectations on how positive (Panel A) and awkward (Panel B) their
shared and withheld compliments would make their recipient feel in Supplemental Experiment
S3. Error bars reflect ± 1 standard errors.
Because participants wrote each compliment on a separate page, we also examined
whether expressers invested different amounts of time writing the compliments they eventually
shared versus eventually withheld. Differential investment in the compliment could serve as an
alternative interpretation of expressers’ decisions to share compliments. However, results
showed that expressers actually spent a similar amount of time writing compliments they decided
to express (M = 56.28 seconds, SD = 38.76) and withhold (M = 53.83 seconds, SD = 34.88),
F(1,52) = .16, p = .69, η2p = .003.
Calibration. Finally, we analyzed whether expressers were miscalibrated in their
expectations of recipients’ emotional reactions. Results showed that expressers’ did not
significantly underestimate how positive their recipients would feel (expressers: M = 7.89, SD =
1.42; recipients: M = 8.18, SD = 1.46), t(52) = 1.25, p = .22, but they indeed overestimated how
awkward the recipients would feel (expressers: M = 2.72, SD = 2.69; recipients: M = 1.58, SD =
2.32), t(52) = -2.68, p = .010, d = -.37. Based on comparisons with some of the preceding
experiments, we believe this weaker miscalibration pattern on positive mood may stem from
5
6
7
8
9
10
Shared Unshared
Positive Mood
0
1
2
3
4
5
Shared Unshared
Awkwardness
(A) (B)
Figure S4

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
23
differences in the recipients’ evaluations compared to the prior experiments due to an unintended
side effect of our experimental design (see the Discussion section of this experiment).
Compliment frequency. Participants again reported giving compliments less often than
they thought they should (M = -1.04, SD = 1.33), one-sample t(105) = -8.03, p < .001, d = -.78.
Responses did not differ between expressers and recipients, paired t(52) = 1.15, p = .26.
Discussion
Expressers who chose which of three compliments to share with a recipient tried, using
Mark Twain’s words, to achieve a “happy delivery” by sharing the compliment they believed
would make the recipients feel most positive. Taken together, Supplemental Experiments S2 and
S3 provided converging evidence that people’s expectations of how positive their compliments
would make the recipients feel guided their decisions about expressing them. These experiments
also suggested that expected awkwardness might be a weaker guide in these decisions, showing
only weak predictive power in Experiment S2 and not differing between shared and withheld
compliments in Experiment S3.
When expressers wrote three compliments but only shared one, they did not significantly
underestimate their recipient’s positive mood. Intriguingly, comparing expressers’ expectations
and recipients’ experiences in this experiment to the same roles in Experiments 1a and 2, which
were conducted in the same park with identical recruitment methods among the same visiting
population, reveals that expressers’ expectations were very similar in these prior experiments.
By contrast, recipients in Experiment S3 reported feeling less positive than their counterparts in
Experiments 1–4. Although these exploratory analyses have to be interpreted with caution
because participants were not randomly assigned at the level of the experiment, they do suggest
that the recipient’s experience may be what differs in this procedure compared to preceding

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
24
experiments. This could have come from an unintended design feature that affected recipients’
perception—they were informed that expressers chose to share one compliment and withhold
two prior to reporting their reactions. Because people judge others harshly for concealing
information (John, Barasz, & Norton, 2016), recipients may have felt a little less positive
because they knew that two compliments had been intentionally withheld.
Supplemental Experiment S4: Updating Expectations to Change Intentions?
If underestimating recipients’ positive reaction to a compliment creates a barrier to
expressing them, then those who are informed that their expectations may be mistaken should be
more interested in expressing compliments to a recipient. We tested this hypothesis in
Experiment S4 by informing some participants of our experiments results demonstrating that
people tend to underestimate how positively others react to a compliment, while others received
no information. We then measured the number of compliments participant wrote to a recipient,
measured how these expressers expect their participants to feel, and measured expressers’ own
mood. We then presented these compliments to their intended recipient, and measured the
recipients’ actual reactions. In addition to testing how altering expressers’ expectations might
influence their willingness to express compliments, it also measures whether expressing as well
as receiving compliments enhances positive mood.
Method
Participants. We conducted this experiment in the same public park as in Experiment 1,
using the same recruiting method for acquainted pairs. We targeted a sample of 50 pairs in each
condition, for a total of 200 participants. Our final sample included 47 in the informed condition
and 53 in the uninformed condition (Mage = 35.31, SDage = 14.33, rangeage = 18–83; 60% female;

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
25
uneven distribution due to coordination issue among multiple experimenters in the field). We
excluded four additional pairs for not completing the experiment (2 in the informed condition
and 1 in the uninformed condition) or for losing the Internet connection during the procedure (1
in the uninformed condition).
Procedure. We randomly assigned pairs of participants to either the informed or the
uninformed condition, with one participant in each pair assigned to be the compliment expresser
and the other the compliment recipient.
We seated the expressers in a private location and provided them a table to start the
study. Expressers first reported their mood on a scale ranging from -5 (much more negative than
normal) to 5 (much more positive than normal), with 0 (no different than normal) as the
midpoint. Participants were then shown the compliment-writing prompt adapted from
Experiment 1 with the additional emphasis that they were “free to write as many or as few
compliments as [they] would like,” and could “write as many as [they wanted], including none at
all.” The critical manipulation was then delivered on the next page titled, “What is this study
about?” In the uninformed condition, expressers simply learned that we had already conducted a
series of experiments on compliment-giving in the park and would like to invite them to take part
in this experiment. In the informed condition, we added another paragraph describing results
from our previous experiments, which read:
“After writing their compliments, we asked the compliment-writers to predict how their
recipients would feel. Although the compliment-writers expected their recipients to feel
happy upon receiving compliments, the recipients actually felt EVEN HAPPIER than the
writers expected. This means that those who gave compliments underestimated the
POSITIVE IMPACT their compliments could have on their recipients.”

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
26
Expressers were reminded on the next page to write separate compliments into separate
textboxes and were given the option to add a new textbox on the same page if they wanted to
write one more compliment—a procedure they repeated until they were finished writing as many
compliments as they wanted, and then clicked a button labeled “I’m done with writing
compliments” to submit all of the compliments at once. Expressers then predicted their
recipients’ reactions on the same mood items as in Experiment 1 (i.e., positive/negative, pleasant,
awkward), and for those who wrote at least one compliment, they predicted how “warm and
sincere” and “articulate and well-written” their recipients would perceive their compliment(s) to
be. Finally, they finished their survey by again reporting how positive/negative they felt on the
same 11-point bipolar scale that they used at the beginning of the experiment.
Once the expresser finished, the experimenter sent the expresser out of sight and retrieved
the recipient, who was then seated in private and given a tablet to start the experimental survey.
Recipients first reported how positive/negative they felt on the same 11-point scale used by
expressers. They were then presented with the expresser’s compliment-writing prompt and all
compliments that were written for them. Recipients then reported their actual mood, followed by
the warmth and competence evaluations (if they had received at least one compliment). Once
finished, each pair was reunited, debriefed, thanked, and dismissed.
Results
Number of compliments. Expressers wrote an average of 5.64 compliments in the
informed condition (SD = 2.82) and 4.77 compliments in the uninformed condition (SD = 2.51).
Shapiro-Wilk tests indicated that the distribution of compliment numbers violated the normality
assumption (informed: W = .96, p = .08; uninformed: W = .90, p < .001), so we deviated from
our preregistered plan to conduct a t-test and performed a one-tailed non-parametric Mann-

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
27
Whitney U test to examine whether expressers in the informed condition wrote more
compliments than those in the uninformed condition. Consistent with our hypotheses, expressers
in the informed condition wrote significantly more compliments than those in the uninformed
condition, U = 1496, p = .040, d = 0.33.
Calibration. As in the preceding experiments, we created a composite measure of
positive mood (rexpresser = .67, rrecipient = .55, ps < .001) and subjected it to a 2 (perspective:
expresser vs. recipient) × 2 (condition: informed vs. uninformed) repeated measures ANOVA.
Results showed a significant main effect of perspective, F(1, 96) = 17.50, p < .001, η2p = .15,
indicating that expressers again underestimated how positive their recipients would feel. We
observed neither a significant main effect of condition, F(1, 96) = 2.18, p = .14, nor a significant
interaction between perspective and condition, F(1, 96) = .67, p = .41, indicating that
information we provided to expressers did not significantly influence expressers’ expectations of
their recipients’ mood, nor did it affect how recipients actually reacted. Expressers
systematically underestimated how positive the recipients would feel upon receiving the
compliments regardless of conditions (expressers: M = 7.86, SD = 1.34; recipients: M = 8.57, SD
= 1.48). A similar 2 × 2 ANOVA on awkwardness likewise yielded a significant main effect of
perspective, F(1, 96) = 7.78, p = .006, η2p = .08, but neither a main effect of condition, F(1, 96)
= .08, p = .77, nor an interaction between perspective and condition, F(1, 96) = .15, p = .70. As
in the prior experiments, expressers in both conditions systematically overestimated how
awkward their recipients would feel (expressers: M = 3.16, SD = 3.01; recipients: M = 2.12, SD =
2.60).
Next, we conducted a 2 (perspective) × 2 (condition) × 2 (attribute) repeated measures
ANOVA on expected and actual warmth and competence evaluations. This revealed a main

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
28
effect of perspective, F(1, 96) = 68.22, p < .001, η2p = .42, a main effect of attribute, F(1, 96) =
102.95, p < .001, η2p = .52, yet no main effect of condition, F(1, 96) = 1.87, p = .18, η2p = .02.
No interaction terms involving condition reached statistical significance, ps > .16. Replicating
results from the previous studies, we found a significant interaction between perspective and
attribute, F(1, 96) = 21.37, p < .001, η2p = .18, indicating that expressers underestimated how
favorably their compliments would be rated in terms of competence (expressers: M = 6.19, SD =
2.56; recipients: M = 8.58, SD = 1.58) even more than in terms of warmth (expressers: M = 8.08,
SD = 1.73; recipients: M = 9.38, SD = 1.05).
Mood change. Similar to Experiment 2, we conducted a 2 (time: before vs. after) × 2
(condition: informed vs. uninformed) × 2 (role: expresser vs. recipient) repeated measures
ANOVA on participants’ reported mood. This revealed a main effect of time, F(1, 96) = 125.63,
p < .001, η2p = .57, a main effect of condition, F(1, 96) = 4.62, p = .034, η2p = .05, and a
nonsignificant effect of perspective, F(1, 96) = 2.11, p = .15, η2p = .02. Among all interaction
terms, only the interaction between time and perspective was statistically significant, F(1, 96) =
16.38, p < .001, η2p = .15. Simple effects tests indicated that expressers reported feeling more
positive after writing compliments (M = 3.10, SD = 1.59) than before (M = 2.38, SD = 1.83), F(1,
96) = 20.72, p < .001. Recipients also reported feeling more positive after receiving
compliments (M = 3.85, SD = 1.71) than before (M = 2.20, SD = 2.01), F(1, 96) = 146.11, p
< .001. Even though both roles felt more positive at the end of the experiment than they did at
the beginning, the significant interaction between time and perspective again indicates that
receiving compliments created a larger increase in positive mood than expressing compliments.
Discriminative accuracy. Expressers again had only moderate insight in how their own
recipient would feel compared to other recipients. Correlation analyses between expressers’

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
29
expectation and recipients’ experiences revealed a significant correlation for positive mood in the
informed condition (rinformed = .31, p = .026) and a marginally significant correlation in the
uninformed condition (runinformed = .25, p = .091), but nonsignficant correlations for awkwardness
in both conditions (rinformed = .17, p = .23; runinformed = . 09, p = .56). We likewise observed no
meaningful discrimination accuracy either in terms of warmth (rinformed = .02, p = .91; runinformed =
-.04, p = .80) or competence (rinformed = .14, p = .33; runinformed = .08, p = .60).
Discussion
After being informed about people’s general tendency to underestimate the positive
impact of their compliments on recipients, expressers wrote slightly more compliments than
those who received no additional information. This suggests that expecting a more positive
impact of one’s compliments on others could encourage people to express more compliments in
daily life. This information did not, however, lead our expressers to become more calibrated in
predicting how positive (or how awkward) their own recipient would feel upon receiving their
compliments. Additional exploratory analyses showed that although expressers in the informed
condition did expect their recipient to feel marginally more positive than expressers in the
uninformed condition (Ms = 8.09 and 7.61, SDs = 1.35 and 1.30), t(96) = 1.78, p = .079, d = .36,
recipients in the informed condition also ended up feeling slightly—although nonsignificantly—
more positive than those in the uninformed condition (Ms = 8.66 and 8.47, SDs = 1.41 and 1.57),
t(96) = .65, p = .52, d = .13. These results suggest that simply informing people of existing
experimental results may be insufficient to notably change people’s expectations, and hence does
not provide a strong test of how manipulating experiments could affect interest in expressing
compliments. We therefore conducted Experiment 4 (reported in the main text) to provide what
we expected would be a stronger test of our hypotheses. Experiment S4 also replicated the

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
30
finding from Experiment 2 that both expressing and receiving compliments increases positive
mood, but that the increase is larger when receiving compliments than when expressing them.
Additional Analysis
Experiment 1
Warmth and competence. We examined how expressers and recipients weighted
evaluations of warmth and competence in their expectations of the recipient’s mood or the
recipient’s actual experiences. Multiple linear regression analysis showed that expressers’
expectations of the recipient’s positive mood were significantly predicted by their expectations
on both the perceived warmth (β = .45, t = 3.43, p = .001) and perceived competence (β = .29, t
= 2.26, p = .028) of the compliments. The two regression coefficients did not differ significantly
from each, F(1,49) = .43, p = .51. This suggests that expressers could have relied somewhat
similarly on warmth and competence evaluations when thinking about the recipients’ positive
mood. In contrast, recipients’ reports of their own positive mood weighed somewhat more
heavily on the perceived warmth of the compliments (β = 0.62, t = 5.28, p < .001) than the
perceived competence (β = 0.22, t = 1.86, p = .069), F(1,49) = 3.53, p = .066, suggesting that
recipients’ positive mood may have been driven more by the compliments’ perceived warmth
than perceived competence. For awkwardness, neither competence nor warmth was correlated
with expressers’ expectations (ps > .19) or recipients’ actual experiences (ps > .53), suggesting
that the expected and experienced awkwardness may not come from the specific words used in
the compliments themselves. These results suggest, at least for positive mood, that expressers
and recipients weight evaluations of competence and warmth somewhat differently.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
31
Relationship quality. A 2 (condition: compliment vs. control) × 2 (perspective:
expected vs. actual) mixed-model ANOVA on participants’ self-reported relationship quality
yielded no significant effects, indicating that participants did not differ across conditions or
perspectives in their evaluation of their relationships (all Fs < 1). Giving a series of compliments
made the recipients feel positive in the moment, but it did not cause them to reassess the overall
quality of their long-term relationships.
Relationship length. We preregistered a secondary analysis to explore the relationship
between how long expressers knew their recipient and their degree of miscalibration on the
recipient’s positive mood. This analysis showed only a weak, and statistically nonsignificant
negative relationship between relationship length and miscalibration both when using the raw
data, r = -.22, p = .12, and a logarithmic transformation of relationship length to normalize the
distribution, r = -.20, p = .16. Being in a relatively longer relationship did not meaningfully
increase expressers’ ability to predict recipients’ reactions to their compliments.
Compliment frequency. Although we had no particular hypothesis regarding how the
experimental conditions would influence people’s self-assessment on compliment-giving, we
conducted an exploratory 2 (condition: compliment vs. control) × 2 (perspective: expected vs.
actual) mixed-model ANOVA that yielded a significant main effect of condition, F(1,102) =
4.66, p = .033, η2p = .044, no main effect of perspective, F(1,102) = .47, p = .50, and a significant
interaction, F(1,102) = 5.13, p = .026, η2p = .048. Simple effects tests indicated that recipients in
the compliment condition reported feeling more deficient in their compliment-giving than
expressers did (Ms = -1.23 vs. -.69, respectively), F(1,51) = 4.45, p = .040, η2p = .041, whereas
predictors and targets in the control condition reported similar ratings (Ms = -.71 vs. -.42,
respectively), F(1,51) = 1.22, p = .27. Furthermore, expressers in the compliment condition and

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
32
predictors in the control condition reported similar levels of compliment-giving, F(1,102) = .01,
p = .94, while compliment recipients, who had just experienced the positive impact of
compliments, reported feeling more deficient in their compliment-giving than targets in the
control condition, F(1,102) = 9.01, p = .003, η2p = .081.
Discrimination accuracy. Systematic miscalibration between the expressers’
expectations and the recipients’ actual experiences does not necessarily mean that the expressers
have no insight in their recipients’ experience. They could still have some above-chance
understanding of how their own recipient will respond compared to other recipients, resulting in
a strong correlation between expressers’ and recipients’ ratings even if they are systematically
underestimating the overall positivity of their recipient’s experience (Epley & Dunning, 2006;
Gagne & Lydon, 2004). However, the correlations between expressers’ and recipients’ ratings
provided only modest support for such discrimination accuracy among expressers. We observed
a correlation significantly larger than zero only when expressers predicted the recipients positive
mood (r = .37, p = .007), and observed nonsignificant correlations when expressers predicted the
recipients’ feelings of awkwardness (r = .22, p = .11), evaluations of warmth (r = .22, p = .12), or
evaluations or competence (r = .11, p = .45). Predictors in the control condition showed modest
positive correlation in predicting their recipient’s feelings of awkwardness (r = .30, p = .033), yet
not positive moods (r = .18, p = .20). Therefore, expressers might have some limited insight into
their recipient’s unique reactions, but it is far from perfect.
Experiment 2
Warmth and competence: mediation analysis. According to our exploratory mediation
analysis, when both underestimating warmth and underestimating competence were entered as
simultaneous mediators, the competence path showed a significant indirect effect (indirect effect

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
33
= .42; bootstrapped 95% CI = [.09, .83]), yet the warmth path was nonsignificant (indirect effect
= .18; bootstrapped 95% CI = [-.10, .45]). However, this does not necessarily suggest that
underestimating warmth could not have an indirect effect on underestimating positive mood—in
fact, when entered as the only mediator, underestimating warmth still accounted for a statistically
significant proportion of variance in the discrepancy between expressers’ expected and
recipients’ reported positive mood (c = .55, p = .002; c’ = .20, p = .30; indirect effect = .35;
bootstrapped 95% CI = [.14, .58]), suggesting that the mediation effect of underestimating
warmth could have been overshadowed by the competence path when both were simultaneously
included in the same model.
Relative status: mediation analysis. An exploratory mediation analysis found that
differences in perceived status did not mediate the difference between expressers’ expected and
recipients’ reported positive mood (see Figure S8).
Figure S8. Mediational analysis in Experiment 2 with relative status as the mediator.
Discriminative accuracy. We calculated to what extent expressers’ expectations were
correlated with their recipients’ actual responses. We observed a small to medium correlation in
positive mood (r = .22, p = .019), awkwardness (r = .24, p = .012), perceived warmth (r = .29, p
= .002), and perceived competence (r = .28, p = .003), and a large correlation when expressers
Perspective
(0 = expresser;
1 = recipient)
underestimating
Status
underestimating
Positive Mood
.03 [-.09, 16]
.61**
(c’ = .52, p= .006)
c= .55, p= .002
.05

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
34
predicted how surprised recipients would be (r = .55, p < .001) and how submissive/dominant
their compliment-giving behavior would be seen as (r = .46, p < .001). Overall, expressers
showed some insights into their recipients’ unique reactions but were far from perfect.
Correlation between expressers’ expectation and their own mood. Expressers’
expectations of how positive their recipient would feel after reading the compliment(s) were
positively correlated with their own mood after writing, r = .39, p < .001. In comparison, their
expectations were only marginally correlated with recipients’ actual responses, r = .17, p = .084.
The former correlation coefficient was shown to be larger than the latter (Pearson and Filon’s z =
1.99, p = .047; Diedenhofen & Musch, 2015), suggesting that expressers’ own mood was
strongly correlated with their expectations of how positive their recipients would feel.
Experiment 3
Compliment frequency. Consistent with results in Surveys 1-4, participants on average
reported giving compliments to their study partner significantly less often than they thought they
should on a scale of -3 (a lot less often than I think I should) to 3 (a lot more often than I think I
should) (M = -.89, SD = 1.30), t(299) = -11.83, p < .001, d = .68, and they felt that their study
partner gave them compliments about as often as they should (M = -.06, SD = .06, p = .29).
Hence, participants felt that how often they gave compliments to their study partner was less
ideal than how often they received compliments from them, t(299) = -8.84, p < .001, d = -.51.
On people’s self-assessed deficit in expressing compliments, an exploratory 3 (condition)
× 2 (perspective) mixed-model ANOVA further showed a marginally significant main effect of
perspective, F(1,147) = 3.05, p = .083, η2p = .020, a marginally significant main effect of
condition, F(2,147) = 2.79, p = .065, η2p = .037, and no interaction between perspective and
condition, F(2,147) = 1.55, p = .22. Overall, expressers considered themselves to be more

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
35
deficient in giving compliments than recipients did (Ms = -1.01 vs. -.76, SDs = 1.28 vs. 1.30,
respectively). Across conditions, participants in the unfocused condition reported the highest
level of self-perceived deficiency in giving compliment (M = -1.08, SD = 1.22), followed by
participants in the competence-focused condition (M = -.93, SD = 1.15) and then the warmth-
focused condition (M = -.64, SD = 1.48).
Discrimination accuracy. As in Experiment 1, we tested the extent to which expressers
had some unique insight into their own recipient’s experience compared to that of other
recipients by assessing the correlation between expressers’ expectations and recipients’ actual
evaluations. Mirroring the calibration findings discussed above, expressers’ expectations in the
warmth-focused condition were significantly correlated with their recipient’s reported positive
mood (r = .36, p = .01) and experienced awkwardness (r = .36, p = .01), indicating some above-
chance accuracy in understanding their recipient’s unique perspective on their compliments.
Expressers’ expectations in the competence-focused and unfocused conditions, in contrast, were
nonsignificantly correlated with the recipient’s actual experience of positive mood (rs = .13
& .19, respectively, ps = .35 & .18) and awkwardness (rs = .07 & .19, ps = .64 & .19).
Experiment 4
As described in the method section of the main text, we asked participants to rate how
“warm and sincere” or how “articulate and well-written” their compliments were in their
respective conditions in order to focus them on the warmth or competence of their compliments.
We predicted that expressers’ interest in expressing a compliment and their actual behavior
would be positively correlated with such warmth and competence evaluations. We preregistered
a set of secondary analysis to test the above relationships by constructing (logistic) linear

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
36
regression models with either warmth or competence rating as the predictor, and either interest or
actual behavior as the outcome variable.
For self-reported interest in expressing a compliment, we found that how competent
expressers rated their compliment to be predicted their interest in expressing it, B = .86, SE = .10,
t = 8.20, p < .001. Likewise, how warm they rated their compliment to be also predicted their
interest in expressing it, B = .53, SE = .18, t = 2.92, p = .004.
For their actual behaviors, we found that how competent expressers rated their
compliment to be also significantly predicted their actual compliment-giving behavior, B = .061,
SE = .022, t = 2.82, p = .006. By contrast, how warm expressers rated their compliment to be did
not predict their actual behavior, B = -.012, SE = .032, t = -0.37, p = .72.
Given that participants were instructed to focus on either warmth or competence in this
experiment, and that we did not have a third condition where no prompt was provided, we could
not directly test if focusing on the competence of a compliment reduced people’s interest in
expressing their compliment compared to an unprompted condition, or if focusing on warmth
increased their interest in expressing it. However, Experiment 2 indeed found expressers in the
unfocused condition responded in a way that was similar to participants in the competence-
focused condition, suggesting that people naturally focus on the competence of their
compliments when unprompted. Combined with consistent evidence on how expectations guide
people’s behaviors across Supplemental Experiments S2 to S4, we believe that an undue focus
on competency, relative to warmth, can therefore create a barrier for people to expressing kind
words to other people.

INSUFFICIENTLY COMPLIMENTARY (SUPPLEMENTAL MATERIALS)
37
References
Diedenhofen, B., & Musch, J. (2015). Cocor: A comprehensive solution for the statistical
comparison of correlations. PloS one, 10(4), e0121945.
Epley, N. & Dunning, D. (2006). The mixed blessings of self-knowledge in behavioral
prediction: Enhanced discrimination but exacerbated bias. Personality and Social
Psychology Bulletin, 32(5), 641–655.
Eyal, T., Steffel, M., & Epley, N. (2018). Perspective mistaking: Accurately understanding the
mind of another requires getting perspective, not taking perspective. Journal of Personality
and Social Psychology, 114(4), 547–571.
Fiedler, K., Harris, C., & Schott, M. (2018). Unwarranted inferences from statistical mediation
tests – An analysis of articles published in 2015. Journal of Experimental Social
Psychology, 75, 95–102.
Fiedler, K., Schott, M., & Meiser, T. (2011). What mediation analysis can (not) do. Journal of
Experimental Social Psychology, 47(6), 1231–1236.
Gagne, F. M., & Lydon, J. (2004). Bias and accuracy in close relationships: An integrative
review. Personality and Social Psychology Review, 8(4), 322–338.
John, L. K., Barasz, K., & Norton, M. I. (2016). Hiding personal information reveals the worst.
Proceedings of the National Academy of Sciences, 113(4), 954–959.