# Overly Shallow?: Miscalibrated Expectations

# Create a Barrier to Deeper Conversation

## Michael Kardas^1 , Amit Kumar^2 , and Nicholas Epley^3

(^1) Kellogg School of Management, Northwestern University
(^2) McCombs School of Business, University of Texas at Austin
(^3) Booth School of Business, University of Chicago
People may want deep and meaningful relationships with others, but may also be reluctant to engage in
the deep and meaningful conversations with strangers that could create those relationships. We hypothe-
sized that people systematically underestimate how caring and interested distant strangers are in one’s
own intimate revelations and that these miscalibrated expectations create a psychological barrier to
deeper conversations. As predicted, conversations between strangers felt less awkward, and created
more connectedness and happiness, than the participants themselves expected (Experiments 1a–5).
Participants were especially prone to overestimate how awkward deep conversations would be com-
pared with shallow conversations (Experiments 2–5). Notably, they also felt more connected to deep
conversation partners than shallow conversation partners after having both types of conversations
(Experiments 6a–b). Systematic differences between expectations and experiences arose because partici-
pants expected others to care less about their disclosures in conversation than others actually did
(Experiments 1a, 1b, 4a, 4b, 5, and 6a). As a result, participants more accurately predicted the outcomes
of their conversations when speaking with close friends, family, or partners whose care and interest is
more clearly known (Experiment 5). Miscalibrated expectations about others matter because they guide
decisions about which topics to discuss in conversation, such that more calibrated expectations encour-
age deeper conversation (Experiments 7a–7b). Misunderstanding others can encourage overly shallow
interactions.
Keywords:self-disclosure, intimacy, social connection, accuracy, social cognition
Supplemental materials:https://doi.org/10.1037/pspa0000281.supp
A number of porcupines huddled together for warmth on a cold day in
winter; but, as they began to prick one another with their quills, they
were obliged to disperse. However the cold drove them together again,
when just the same thing happened...In the same way the need of so-
ciety drives the human porcupines together, only to be mutually
repelled by the many prickly and disagreeable qualities of their nature.
The moderate distance which they at last discover to be the only toler-
able condition of intercourse, is the code of politeness andfine man-
ners.—Schopenhauer, 1851/1964, p. 226
Schopenhauer argued that people want deeper and more inti-
mate relationships but are reluctant to pursue them because they
expect that greater intimacy will be unpleasant. These expecta-
tions, in turn, discourage intimacy, entrenching the“politeness and
fine manners”of small talk as the norm in everyday discourse.
Indeed, everyday conversation seems to be anything but a stream
of deep and meaningful exchanges: Fewer than half of people’s
conversations are substantive and meaningful (Mehl et al., 2010),
creating an entire genre of self-help books promising to help you
make better“small talk”and“chit chat”with those around you.
Thefine manners of small talk aside, empirical research sug-
gests notable benefits of having more“deep talk”in everyday life.
Deep and meaningful conversations strengthen social ties (Aron et
al., 1997; Collins & Miller, 1994), relieve the psychological bur-
dens of secrecy or negative emotional experiences (Pennebaker,
1997 ; Slepian & Moulton-Tetlock, 2019), and speed the develop-
ment of close relationships (Altman & Taylor, 1973; Derlega et
al., 2001). Presumably because positive social relationships bring
happiness and wellbeing (e.g.,Diener & Seligman, 2002), those
who spend more time engaging in deep talk tend to be happier
than those who spend relatively more time in small talk (Mehl et
al., 2010; Milek et al., 2018). With all of the positive outcomes
that spring from deep talk, what keeps people from having more
of it in their everyday lives?
Here we examine the psychological processes guiding conversa-
tions to understand why people choose to discuss relatively deep
versus shallow topics, and whether those choices accurately
Michael Kardas https://orcid.org/0000-0003-3486-
This research was supported by the Neubauer Family Faculty Fellowship
and the Booth School of Business.
Correspondence concerning this article should be addressed to Nicholas
Epley, Booth School of Business, University of Chicago, 5807 South
Woodlawn Avenue, Chicago, IL 60637, United States. Email:epley@
chicagobooth.edu
1

## Journal of Personality and Social Psychology:

## Attitudes and Social Cognition

```
©2021 American Psychological Association
ISSN: 0022-3514 https://doi.org/10.1037/pspa
```
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
anticipate the outcomes of relatively deep and shallow conversa-
tions. We define deep conversations as those in which two people
engage in self-disclosure by revealing personally intimate infor-
mation about their thoughts, feelings, or experiences. We predicted
that people’s decisions to engage in deep conversation are guided
by their expectations about how much their interaction partner
cares about the intimate details of one’s life, and that people are
reluctant to engage in deep talk to the extent they believe their
conversation partner will be indifferent to the conversation’s
content. Perhaps more important, we also predict that people’s
expectations are systematically miscalibrated, such that they
underestimate how much others care about one’s own disclosures
in conversation, and thereby overestimate how awkward and
uncomfortable deep conversations will be. We suggest that these
miscalibrated expectations about others can create a psychological
barrier to engaging in deep conversations more often in everyday
life.
```
## Psychological Barriers to Meaningful Conversation

```
Belonging is a fundamental human need virtually on par with
eating and sleeping (Baumeister & Leary, 1995; Ryan & Deci,
2000 ), and yet previous research suggests that people often forego
opportunities to connect with others that would satisfy this need
and thereby enhance both their own and others’wellbeing. People
can be somewhat reluctant to talk with strangers (Dunn et al.,
2007 ; Epley & Schroeder, 2014; Mallett et al., 2008), give compli-
ments (Boothby & Bohns, 2021; Zhao & Epley, in press), express
gratitude (Kumar & Epley, 2018), perform random acts of kind-
ness (Dunn et al., 2008; Kumar & Epley, 2021), or disclose perso-
nal information (John et al., 2016; Kardas et al., 2021), at least
partly because people in these cases underestimate the positive
impact their social behavior will have on others. Even the briefest
engagements with others, from saying“hello”to a barista (Sand-
strom & Dunn, 2014a,2014b) to making eye contact with a pass-
erby (Wesselmann et al., 2012), can increase the sense of
connection to others, and yet people sometimes forego even these
easy opportunities to engage with others. Misunderstanding how
positively others value social interaction can create a barrier to
engaging with others, thereby creating a systematic barrier to satis-
fying the basic human need for belonging.
Whereas existing research has primarily examined how wisely
people choosewhetherto engage with others, we expand this body
of work by examining how wisely people choosehowto engage
with others in the midst of conversation. Any conversation can
range from relatively shallow and superficial to relatively deep
and intimate. Like many decisions that are guided by some assess-
ment of expected costs and benefits ( Becker, 1993), we predict
that people choose the depth of conversation based partly on how
they expect that an intimate versus shallow conversation is likely
to unfold.
Although there are likely to be multiple mechanisms that influ-
ence people’s expectations about a conversation, we theorize that
people’s expectations about shallow and deep conversation are
guided by at least two inferences. First, expectations should be
guided by the information people expect to share during the con-
versation. Shallow conversations tend to be impersonal and require
relatively little self-disclosure. People are unlikely to feel vulnera-
ble to others’evaluations while discussing impersonal topics like
```
```
the weather. Deep conversations, in contrast, require sharing per-
sonal information about one’s past experiences, preferences, or
beliefs, which could leave people feeling more vulnerable to
others’evaluations (Berenson et al., 2009; Leary, 1983). Existing
research suggests that these fears of vulnerability may be mis-
placed. In one series of experiments, people overestimated how
harshly they would be judged by others when they were in the
midst of some embarrassing blunder or mishap (Savitsky et al.,
2001 ). In another series of experiments, people overestimated how
negatively they would be judged when they revealed personal
imperfections or weaknesses to another person (Bruk et al., 2018).
Finally,strangers talking for thefirst time reported feeling more
positive, less awkward, and more connected to each other after a
relatively deep conversation than they expected, especially when
they were communicating over a relatively more intimate (voice-
based) communication media compared with less intimate (text-
based) media (Kumar & Epley, 2021, Experiment 2). This
research did not include a shallow conversation condition for com-
parison nor did it examine underlying mechanisms that might
explain why people would undervalue deep conversation. Collec-
tively, this existing research suggests that deeper conversations
may not leave us as vulnerable as we might expect.
Second, expectations should also be guided at least in part by
people’s inferences about how much one’s conversation partner
will care about the intimate and meaningful topics that make up
deeper conversation. Whereas a shallow conversation may be
unlikely to feel awkward or uncomfortable regardless of how car-
ing and interested the recipient is, a deep conversation may feel
quite awkward with an uncaring or indifferent partner who cares
little for learning more about us. This predicts that people should
have more positive expectations about deep conversation when
they expect their conversation partner to be caring and interested
than when they expect another person to be uncaring or indiffer-
ent. People should also be more willing to discuss deeper topics
when they expect their partner to be relatively caring and inter-
ested in what one has to say.
Identifying how wisely people choose the depth of their conver-
sation topics requires understanding how accurately people infer
both their own and others’care and interest in the content of deep
conversation. People can directly assess, via introspection, how
much they personally care about what others have to say regarding
deep and intimate topics, but have to infer how much others care
(Epley & Waytz, 2010; Gilbert & Malone, 1995; Jones, 1979; Nis-
bett & Wilson, 1977). People are also deeply motivated to connect
with others (Baumeister & Leary, 1995) and deeply value signs of
trustworthiness and honesty from others, including others’willing-
ness to open up in deeper conversation and their willingness to lis-
ten attentively to one’s own statements (Ames et al., 2012; Cuddy
et al., 2007; Kluger et al., 2021; Kluger & Zaidel, 2013; Levine &
Cohen, 2018). One series of experiments even indicated that peo-
ple underestimate how much attention others pay to them in every-
day life (Boothby et al., 2017), suggesting that people may fail to
fully appreciate how much interest others are likely to take in
them.
This prior research led us to make several predictions about the
accuracy of people’s expectations about care and interest. First,
we predicted that people would assume that they would care more
about the intimate details of what they share in the midst of a deep
conversation than a typical stranger would care about those same
```
## 2 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
intimate revelations. Second, we predicted that people’s beliefs
about their own interest and care would be more calibrated than
their inferences about a stranger’s interest, thereby leading people
to underestimate others’interest and care (but not their own).
Third, we predicted that underestimating the extent to which
others are interested in, and care about, one’s statements in conver-
sation would also lead people to underestimate how connected
they would feel to their conversation partner and how happy they
would feel about the conversation, and to overestimate how nega-
tive the conversations would be. Fourth, because conversations
with seemingly indifferent others could feel awkward when they
require more intimate self-disclosure, we predicted that partici-
pants would overestimate the awkwardness of deep conversation
more than the awkwardness of shallow conversation. Finally,
because one’s friends are known to care more about oneself than
strangers, we expected that people would be more calibrated pre-
dicting the outcomes of their conversations with friends than with
strangers because people would anticipate more positive outcomes
with friends, even when discussing relatively intimate topics.
We believe that misunderstanding others’sociality—the degree
to which others are interested in, and care about, connecting
through conversation—matters because it creates a psychological
barrier to engaging in deeper and more meaningful conversation.
This predicts that people are overly reluctant to engage in deep
conversation because they have miscalibrated expectations about
the consequences of sharing meaningful information with others
whose interest and care is unknown. A simple alternative interpre-
tation is that people do not engage in deeper conversations with
strangers because they do not care about what theotherperson has
to say. If people are indifferent to strangers’responses in deep
conversations, and hence are uninterested in having deeper conver-
sations with them, then none of the hypotheses we articulated
above would be worth testing because beliefs about others’care
would not guide people’s decisions to engage in relatively shallow
versus deep conversation.
We conducted an initial pilot experiment using Amazon’s Me-
chanical Turk (MTurk;N= 188) in which participants imagined
talking with someone they had never met before and were asked
how interested they were in sharing intimate details with a stranger
or learning those same details about a stranger. Specifically, we
randomly assigned participants to one of four roles:Listeners
imagined that they would listen to the other person’s responses to
a series of questions,Askersimagined that they would ask the
other person a series of questions and then listen to the person’s
responses,Discussantsimagined that they would both ask and an-
swer a series of questions with the other person, andAnswerers
imagined that they would answer a series of questions while the
other person listened. Participants then viewed a list of 20 pre-
tested discussion questions that varied in intimacy from very shal-
low and superficial to very deep and intimate (seeAppendix), and
selected thefive questions they preferred for the exchange. We
hypothesized that Discussants and Answerers, who were required
to answer the questions and reveal information about themselves,
would choose less intimate questions on average than Listeners
and Askers, who were simply able to learn about others.
Results supported these hypotheses. Listeners (Mintimacy= 5.06)
and Askers (Mintimacy= 4.71) chose questions higher in average in-
timacy than either Discussants (Mintimacy= 4.04) or Answerers
(Mintimacy= 3.58, ps# .008). Whereas the majority of the
```
```
questions selected by Listeners (M= 3.06 out of 5,p= .017) were
“deep”questions that were more intimate than the median ques-
tion, and Askers (M= 2.74,p= .320) selected equal numbers of
deep and“shallow”questions, Discussants (M= 1.89,p= .003)
and Answerers (M= 1.36,p,.001) chosefewerdeep questions
than shallow questions (seeonline supplemental materialsfor the
full method and results). This suggests that people are indeed
interested in knowing intimate information about others (see also
Hart et al., 2021), but are reluctant to reveal intimate information
about themselves. A follow-up experiment (N= 144) suggested
that this difference did not stem from differences in the perceived
difficulty of generating answers to shallow versus deep questions,
as participants’choices did not differ between a condition in
which they imagined revealing prewritten responses (Mintimacy=
3.79) and one in which they imagined generating answers to the
questions in real time (Mintimacy= 3.83,p= .885). Again, partici-
pants in the conditions that required revealing personal informa-
tion in conversation preferred to discuss less intimate questions
than participants who imagined only listening to the other person’s
responses (Mintimacy= 4.80,ps,.001; seeonline supplemental
materials). These results indicate that these participants were more
interested in getting to know meaningful information about others
than they were in revealing meaningful information to others, sug-
gesting a barrier to engaging in deep conversation that may come
from inferences about how others will respond to these self-disclo-
sures. This article reports a series of experiments testing whether a
reluctance to engage in deep conversation arises partly from mis-
calibrated concerns that strangers will be indifferent toward one’s
self-disclosures, creating more superficial conversations than
might be ideal for one’s own and others’wellbeing.
Of course, conversations can range from very shallow and su-
perficial to very deep and intimate. Our hypotheses do not suggest
that underestimating others’care creates a barrier to having the
deepestconversations. Some topics may be sufficiently deep that
people will prefer to avoid discussing them even with close others
who they expect to be highly caring and considerate. Rather, our
hypotheses suggest that underestimating others’care creates a bar-
rier to havingdeeperconversations. By deeper conversations we
mean those that are more intimate than people’s typical conversa-
tions and are instead closer to the depth of conversation that peo-
ple report wishing they would experience more often. We will test
each component of our reasoning empirically: whether people’s
typical conversations are less intimate than they wish they would
be, whether people underestimate others’care during conversa-
tion, and whether experimentally manipulating people’s beliefs
about others’care influences the depth of conversation that they
prefer.
```
## Overview of Experiments

```
We conducted a series of experiments to test our hypotheses,
and in doing so we recruited diverse samples of participants
including primarily American undergraduate and master’s degree
students (Experiments 2, 3, 4a, 4b, 6a, and 6b), community mem-
bers (Experiments 3 and 5),financial services employees and
executives (Experiments 1a and 1b), international MBA students
from around the world (Experiment 1c), and online participants
(Experiments 7a and 7b). Wefirst tested whether people underesti-
mate the positivity of deep and intimate conversations with
```
## DEEPER CONVERSATION 3

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
strangers by asking participants in Experiments 1a–1c to discuss a
series of deep and intimate questions with a stranger. Participants
reported how they expected to feel after these conversations, and
we then compared participants’expectations against their actual
experiences. Experiments 2–6b compared shallow conversations
against deeper conversations by manipulating the intimacy of par-
ticipants’conversations. Experiment 5 also compared conversa-
tions between strangers versus known family or friends to
examine whether participants are especially likely to misunder-
stand the outcomes of deep conversations with strangers. Experi-
ments 6a and 6b assigned participants to engage inbothshallow
and deep conversations as a more direct test of whether relatively
deeper conversations build stronger connections between strangers
than shallower ones. Finally, Experiments 7a and 7b examined the
extent to which expectations of others’interest and care create a
psychological barrier to deeper conversations.
Our experiments make several novel contributions to our under-
standing of social cognition and interpersonal relationships. First,
previous experiments test whether people appreciate the benefits
of distant social connections with strangers (Dunn et al., 2007;
Epley & Schroeder, 2014; Mallett et al., 2008) but do not measure
or manipulate the intimacy of these conversations. Strangers tend
not to engage in particularly intimate conversations with each
other (Taylor, 1968), meaning that our experiments are an even
stronger test of the potential benefits of distant social connections
(Sandstrom & Dunn, 2014b). Second, a substantial body of litera-
ture examines how people make inferences about others’warmth
(e.g.,Cislak & Wojciszke, 2008; Fiske et al., 2007, 2002; Russell
& Fiske, 2008; Todorov et al., 2009) but very little work examines
the accuracy of these inferences. Our experiments test whether
people systematically underestimate others’care and concern in
the context of deep and meaningful conversations. Third, whereas
ample research indicates that inferences about others’warmth
impactwhetherpeople choose to engage with one another (e.g.,
Adolphs et al., 1998; Cuddy et al., 2007; Todorov, 2008; Woj-
ciszke et al., 1998; see alsoEpley & Schroeder, 2014), our experi-
ments enrich this literature by testing whether a related set of
inferences influenceshow deeplypeople choose to engage. Our
findings open new avenues for research about contexts that may
remove psychological barriers to engaging deeply. Finally,
whereas some experiments suggest that deep and intimate conver-
sations can quickly foster experiences of connectedness between
strangers (e.g.,Aron et al., 1997; Fishman & Gardner, 2017), our
experiments move substantially beyond prior research by trying to
understand why people might be reluctant to engage in these con-
versations in thefirst place. Practically speaking, our research may
offer useful guidance for people trying to behave more wisely to
enhance their own wellbeing and strengthen their social relation-
ships. Everyday conversations, we suggest, could be more superfi-
cial than would be optimal for one’s own wellbeing.
To maximize statistical power, we recruited at least 50 partici-
pants or pairs per condition (Experiments 3, 5, 7a, and 7b), or at
least 30 groups of four participants per condition (Experiments 6a
and 6b), for experiments conducted in laboratory,field, and online
settings, and recruited as many participants as we could in experi-
mentsconducted during presentations (Experiments 1a, 1b, 1c, 2,
4a, and 4b). To test the robustness of our effects, we also con-
ducted multiple replication experiments (Experiments 1a–1c, 4a,
and 4b). We report all measures and manipulations throughout the
```
```
article, and we report all analyses without exclusions in theonline
supplemental materials. All experiments were approved by the
university’s Institutional Review Board and we obtained informed
consent from all participants. Theonline supplemental materials,
surveys, data, and analysis script, as well as preregistrations for
Experiments 1b, 2, 3, 4a, 4b, 5, 6a, 6b, 7a, 7b, and several supple-
mental experiments, can be accessed at tinyurl.com/overly
-shallow-osf.
```
## Experiments 1a–1c: Deep Conversations

```
Wefirst tested whether people consistently underestimate how
positively their deep conversations with strangers would unfold. In
Experiments 1a, 1b, and 1c, we controlled the content of the con-
versations by providing deep and intimate discussion questions.
We hypothesized that deep conversations between strangers would
feel less awkward, and would lead to stronger bonds, greater lik-
ing, and greater happiness, than people anticipated. Although these
experiments do not enable direct comparisons between shallow
and deep conversations, they serve as initial tests of whether
highly intimate conversations are more positive than strangers
expect them to be. We test both shallow and deep conversations in
the experiments that follow.
Furthermore, we began testing why people may misunderstand
the outcomes of deep conversations. We predicted that strangers
would underestimate how interested the other person would be
during the conversation, and that this misunderstanding would
help to explain why deep and intimate conversations unfold more
positively than people expect.
We conducted these three experiments on meaningfully differ-
ent samples of varying sizes to test the robustness and reliability
of any experimental results. Experiment 1a includedfinancial
executives at a conference, 1b included managers and employees
at a largefinancial servicesfirm, and 1c included international
MBA students participating from around the world in an online
session. In all cases, the sample size was determined by the num-
ber of people who attended the session that comprised the experi-
ment. Because of the similarity in both the procedures and results,
we present the methods and results in detail only for Experiment
1a and report summaries of Experiments 1b and 1c in the main
text. Full details of Experiments 1b and 1c are reported in the
online supplemental materials.
```
## Method (Experiment 1a)

## Participants

```
Twenty-five pairs offinancial services executives were recruited
during a session at a management conference (N= 50 individuals;
Mage= 48.92;SDage= 7.55; 18.00% female; 86.00% White). The
executives entered the session unaware that they would be partici-
pating in any experiment.
```
## Procedure

```
Participants opened a survey on their computers or mobile devi-
ces. They read that they would be randomly paired with another
person attending the session who they had not met before and
would answer and discuss four questions with the other person.
```
## 4 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
Participants then read four deep questions, adapted from the“fast
friends”paradigm (Aron et al., 1997):
```
1. For what in your life do you feel most grateful? Tell the
    other participant about it.
2. If a crystal ball could tell you the truth about yourself,
    your life, your future, or anything else, what would you
    want to know?
3. If you were going to become a close friend with the other
    participant, please share what would be important for him
    or her to know.
4. Can you describe a time you cried in front of another
    person?

```
After reading the questions, participants reported their expecta-
tions of the upcoming interaction. First, participants reported how
interested they would be in hearing the other person’s answers
(0 =not at all interested;10=very interested) and also how inter-
ested they expected the other person would be in hearing the par-
ticipant’s own answers (0 =not at all interested;10=very
interested). Participants then predicted how awkward they would
feel during the conversation (0 =not at all awkward;10=very
awkward), how strong a bond they would feel with the other per-
son (0 =weak, like a stranger;10=strong, like a new friend),
how much they would like the other person (0 =not at all;10=
very much), and how happy they would feel about the conversation
with the other person (0 =not at all happy;10=extremely happy).
Research assistants then randomly paired participants together
with someone they did not know in the session (most were unfami-
liar with each other), and gave each participant a card containing
their pair’s ID number and the four discussion questions. Partici-
pants then had a conversation for approximately 10 min and
returned to their original seats tofill out another survey asking
them to report their actual experiences in the conversation. First,
participants reported their own experiences using the same meas-
ures on which they reported their expectations before the conver-
sation. They then reported their perceptions of their partner’s
experiences on these measures. Participants then thought about
times when they had spoken with strangers and reported how often
they wished they engaged in small talk with strangers, and how of-
ten they wished they engaged in deep conversations with strang-
ers, on separate scales ranging from
5(much less often than I do
now)to0(neither less nor more often than I do now)to5(much
more often than I do now). Finally, participants reported demo-
graphic information and were debriefed.
```
## Results (Experiment 1a)

```
The strength-of-bond and liking items were highly correlated in
both expectations (a= .82) and experiences (a= .85) and so we
combined these items to form a connectedness scale. In this and
the following experiments, we computed mean responses across
paired participants for each of the primary measures and then per-
formed analyses at the level of the dyad.
Participants underestimated both their own interest in hearing
from the other person,paired t(24) =
5.92,p,.001, 95%
CIdifference[
2.29,
1.11],d=
1.36, and how interested they
```
```
would perceive their partner to be in hearing from them,paired
t(24) =
9.63,p,.001, 95% CIdifference[
3.38,
2.18],d=

2.14 (CI = confidence interval). Moreover, and consistent
with our hypotheses, participants underestimated the extent to
which their partners would be interested in the content of the
conversation significantly more than they underestimated their
own interest,F(1, 24) = 14.63,p,.001,g^2 p=.38.^1
Our data suggest that these differences between participants’
expectations and their own experiences reflect miscalibrated
beliefs about a deep conversation rather than response biases that
might arise in ratings of their partner’s interest after the conversa-
tion. Specifically, participants’expectations of their partner’s in-
terest significantly underestimated both their partner’s expected
interest before the conversation,paired t(24) =
8.68,p,.001,
95% CIdifference[
2.33,
1.43],d=
1.36, as well as their part-
ner’s self-reported interest after the conversation,paired t(24) =

14.07,p,.001, 95% CIdifference[
4.11,
3.05],d=
3.04 (see
online supplemental materialsfor further analyses).
Participants’conversations also unfolded significantly better than
they anticipated: Participants overestimated how awkward they
would feel,paired t(24) = 8.71,p,.001, 95% CIdifference[2.44,
3.96],d= 1.74, underestimated how connected they would feel to
their partner,paired t(24) =
11.32,p,.001, 95% CIdifference
[
2.34,
1.62],d=
2.26, and underestimated how happy they
would feel about the conversation,paired t(24) =
9.59,p,.001,
95% CIdifference [
2.99, 
1.93], d = 
1.92 (see Figure 1,
top panel).2,
We hypothesized that underestimating the other person’s inter-
est would explain miscalibration on the primary measures. To test
this, we performed within-pairs mediational analyses using Mea-
surement Phase (Expectation, Experience) as the independent vari-
able, partner interest as the mediating variable, and each of the
primary measures as dependent variables in separate analyses
using the MEMORE macro (Montoya & Hayes, 2017). The indi-
rect effects were significant for awkwardness (b=
1.61, 95% CI
[
3.08,
.04]) and connectedness (b=
.79, 95% CI [
1.58,

.13]) but nonsignificant for happiness (b=
.80, 95% CI
[
1.97, .33]).
We also hypothesized that people want to engage in deeper con-
versations than they typically do. Consistent with this hypothesis,
participants wished they engaged in marginally more small
talk with strangers than they typically do (M= .68,SD= 2.72),
```
(^1) For all experiments, we present critical hypothesis tests in the main
text and report all other main effects and interaction effects from ANOVAs
in theonline supplemental materials.
(^2) We conducted exploratory analyses using the actor-partner
interdependence model (Kenny et al., 2006) to test whether miscalibration
between participants’expectations and experiences was more strongly
associated with the participants’own expectations (as our theory would
suggest) or their partners’expectations. Throughout our experiments, we
found consistent evidence that miscalibration was associated with one’sown
expectations, but little evidence that miscalibration was associated with one’s
partner’s expectations (seeonline supplemental materials Table S2).
(^3) We also performed exploratory analyses of gender. Throughout our
experiments, neither miscalibration on the primary measures nor
differences in miscalibration between shallower and deeper conversations
varied reliably between same-gender and mixed-gender pairs (seeonline
supplemental materialsfor details).

## DEEPER CONVERSATION 5

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
one-sample t(49) = 1.77,p= .083, 95% CI [
.09, 1.45],d= .25,
but wished they engaged in significantly more deep conversation
with strangers than they typically do (M= 2.60,SD= 1.93),one-
sample t(49) = 9.54,p,.001, 95% CI [2.05, 3.15],d= 1.35.
Moreover, participants’desire for more deep talk was significantly
stronger than their desire for more small talk,paired t(49) = 3.88,
p,.001, 95% CIdifference[.93, 2.91],d= .55. People may refrain
from having more deep conversations in part because they under-
estimate how well these conversations will unfold.
```
## Method and Results (Experiments 1b and 1c)

```
We also replicated thesefindings in Experiments 1b (N=
pairs) and 1c (N= 56 groups of 2–3 individuals; seeonline
supplemental materialsfor the full method and results): Partici-
pants overestimated how awkward they would feel discussing
deep topics with a stranger yet underestimated how connected
they would feel to the other person and how happy they would
feel about the conversation (seeFigure 1). Experiment 1b also
```
```
Figure 1
Mean Awkwardness, Connectedness, and Happiness Across Measurement Phase
(Expectations Versus Experiences) in Experiments 1a–1c
```
```
Note. Error bars 61 SE.
```
## 6 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
included measures of the other person’s care and interest. In this
experiment, participants significantly underestimated the degree to
which their partner would care about, and be interested in, the con-
tent of the conversation. The extent to which they did so statisti-
cally mediated the extent to which they underestimated their own
feelings of connectedness and happiness in a deep conversation.
```
## Experiment 2: Shallow Versus Deep Conversations

```
Experiments 1a–c suggest that relatively deep conversations
between strangers tend to go better than expected. Specifically,
participants felt less awkward, and felt happier and more con-
nected to their partner, than they anticipated. These results were at
least partly explained by participants’tendency to underestimate
how caring and interested their partner would be in hearing what
they had to say.
We suggest that the tendency to overestimate the conversa-
tion’s awkwardness should be especially strong in deep conver-
sations in which intimate self-disclosure is required, compared
with relatively shallow conversations in which less intimate
self-disclosure is required. We tested this hypothesis in Experi-
ments 2–6b by manipulating the intimacy of conversation. In
Experiment 2, we provided participants with a series of shallow
or deep discussion questions, thereby manipulating the content
of the conversation between pairs. We hypothesized that partic-
ipants in the deep talk condition would overestimate how awk-
ward their conversations would be more than participants in the
shallow talk condition, and that participants in both conditions
would underestimate how connected they would feel to their
conversation partner.
```
## Method

## Participants

```
Eighty-nine pairs of master’s degree students participated dur-
ing a faculty-member presentation as part of a wellness event
on campus (N= 178 individuals after exclusions;Mage= 28.23;
SDage= 1.99; 54.49% female; 53.37% White). Participants were
unaware that they would be participating in an experiment until it
was announced during the presentation. An additional seven par-
ticipants could not be matched with their partner after the session
because either one or three participants reported the same pair ID,
and so were excluded from analyses. Because we analyzed the
data at the level of the dyad rather than at the level of the individ-
ual, we did not reanalyze data with these participants included.
```
## Procedure

```
Participants opened a survey on their computers or mobile devi-
ces and were randomly assigned to condition (signified by A or B)
based on where they were seated in the room. Participants in group
A read the followingfive shallow questions adapted fromAron et
al. (1997):
```
1. When was the last time you walked for more than an
    hour? Describe where you went and what you saw.
2. How did you celebrate last Halloween?
    3. Do you like to get up early or stay up late? Why?
    4. What is the best TV show you’ve seen in the last month?
       Tell your partner about it.
    5. How often do you get your hair cut? Where do you go?
       Have you ever had a really bad haircut experience?

```
Participants in Group B read the followingfive deep questions:
```
1. For what in your life do you feel most grateful? Tell your
    partner about it.
2. If a crystal ball could tell you the truth about yourself,
    your life, your future, or anything else, what would you
    want to know?
3. If you were going to become a close friend with your
    partner, please share what would be important for him or
    her to know.
4. What is one of the more embarrassing moments in your
    life?
5. Can you describe a time you cried in front of another
    person?

```
After reading the questions, participants predicted how awk-
ward they would feel during the conversation, how uncomfortable,
how much they would enjoy the conversation, how strong a bond
they would feel with their conversation partner, how much they
would like their conversation partner, and how well they would
feel they got to know their conversation partner’s“true beliefs,
attitudes, preferences, and character,”on similar 11-point scales as
in Experiments 1a–1c. Participants were then randomly paired
with someone from the same condition (shallow or deep) who
they had not met before. Participants received question cards con-
taining thefive questions and were left to discuss the questions in
pairs for about 10 min.
Participants then returned to their original seats and reported
how they felt during their conversations on the same measures. To
test whether deep conversation reduces momentary feelings of
loneliness, participants then completed the three-item perceived
loneliness scale (Hughes et al., 2004), reporting (a) how often they
feel they lack companionship, (b) how often they feel left out, and
(c) how often they feel isolated from others (hardly evervs.some
of the timevs.often). Because research has indicated that extra-
verts have better-calibrated beliefs about the benefits of extra-
verted social behavior than introverts (Zelenski et al., 2013), we
then asked participants to complete the 10-item personality inven-
tory (Gosling et al., 2003) to test whether calibration varies by per-
sonality. Finally, participants reported demographic information
and were thanked and debriefed.
```
## Results

```
The awkwardness and discomfort items were highly correlated
(aexpectations= .98,aexperiences= .77), as were the enjoyment,
strength of bond, liking, and perceived understanding items
(aexpectations= .69,aexperiences= .87). We combined these items to
```
## DEEPER CONVERSATION 7

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
form awkwardness and connectedness scales, respectively. We
then conducted 2 (Conversation: Shallow, Deep) 3 2 (Measure-
ment Phase: Expectations, Experiences) analyses of variance
(ANOVAs) with repeated measures on the second factor and awk-
wardness and connectedness as dependent measures in separate
analyses.
```
## Awkwardness

```
Participants in the deep condition overestimated how awkward
their conversation would feel more than did participants in the shal-
low condition, as indicated by a significant Conversation 3 Measure-
ment Phase interaction effect,F(1, 87) = 71.34,p,.001,g^2 p=.
(see Figure 2). Participants in the shallow condition expected their
conversations to feel less awkward than participants in the deep con-
dition,t(87) = 11.58,p,.001, 95% CIdifference[3.17, 4.48],d=
2.49, and also experienced less awkwardness,t(87) = 5.22,p,.001,
95% CIdifference[.84, 1.87],d= 1.12. As predicted, the significant
interaction indicates that participants in the deep condition (d=2.23)
overestimated how awkward their conversation would be more than
did participants in the shallow condition (d=.55).
```
## Connectedness

```
Participants in both the shallow and deep conditions underesti-
mated how connected they would feel to their partner, as indicated
by a significant main effect of Measurement Phase,F(1, 87) =
98.24,p,.001,g^2 p= .53. The Conversation 3 Measurement
Phase interaction effect was nonsignificant,F(1, 87) = .12,p=
.734,g^2 p= .001, indicating that participants underestimated how
connected they would feel to their partner to similar degrees in the
deep and shallow conditions.
```
## Secondary Analyses

```
We also tested whether participants in the deep condition felt
less lonely overall at the end of their conversations than partici-
pants in the shallow condition by coding the response options on
the perceived loneliness scale (Hardly evervs.Some of the time
vs.Often) numerically from one to three. Unexpectedly, partici-
pants in the deep condition (M= 1.82,SD= .35) reported feeling
```
```
significantly more lonely than participants in the shallow condition
(M= 1.65,SD= .35),t(87) = 2.25,p= .027, 95% CIdifference[.02,
.32],d= .48. We discuss thisfinding further in the study discus-
sion below.
We also tested how our results might be moderated by personal-
ity by regressing expected versus experienced awkwardness and
connection simultaneously over the Big Five personality traits, for
the shallow and deep conditions separately. We did not observe
consistent moderation: The tendency to overestimate the awk-
wardness of a conversation was associated with higher neuroticism
(b=
.26,t(98) =
2.58,p= .011) among participants in the shal-
low condition but was not significantly associated with any Big
Five traits among participants in the deep condition. Underestimat-
ing the sense of connection was associated with higher neuroticism
(b= .22,t(98) = 2.19,p= .031) and lower openness (b=
.21,
t(98) =
2.16,p= .033) among participants in the shallow condi-
tion but was not associated with any Big Five traits among partici-
pants in the deep condition. Ourfindings were not moderated by
extraversion for either awkwardness or connectedness (seeonline
supplemental materialsfor the full analyses). Personality did not
consistently moderate the gap between expected and actual experi-
ence for either shallow or deep conversations.
```
## Discussion

```
The results of this experiment replicate and extend thefindings
from Experiments 1a–1c: Participants’conversations felt less awk-
ward, and led to greater feelings of connectedness, than the partici-
pants expected before engaging in the conversation. Furthermore,
participants who engaged in deep conversations overestimated how
awkward the conversation would be significantlymorethan those
who engaged in shallow conversations. Although deep conversations
did indeed feel more awkward than shallow conversations, the differ-
ence in participants’expectations was roughly four times larger than
the difference that participants actually experienced.
One otherfinding warrants further discussion: In this experiment,
participants in the deep condition reported feeling significantly
lonelier than participants in the shallow condition after their conver-
sations. We did not predict this result, but we are also reluctant to
```
```
Figure 2
Mean Awkwardness and Connectedness Across Conversation Conditions
(Shallow Versus Deep) and Measurement Phase (Expectations Versus Experiences)
in Experiment 2
```
```
Note. Error bars 61 SE.
```
## 8 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
conclude that this effect is a general consequence of deep conversa-
tions rather than an idiosyncratic consequence of the specificques-
tions that participants discussed in this experiment. In particular,
two of the deep questions,“What is one of the more embarrassing
moments in your life?”and“Can you describe a time you cried in
front of another person?,”likely brought to mind memories of nega-
tive social experiences. These questions alone may have driven dif-
ferences in reported loneliness. Because we are primarily interested
in how connected people feel to their conversation partner, rather
than more general feelings of loneliness, we did not include this
measure in the following experiments but instead included meas-
ures that are more closely tied to experienced wellbeing from the
conversation itself. Specifically, we measured participants’feelings
of happiness in Experiments 3–6. We also test a much wider range
of questions in Experiments 3 and 5 to examine whether the effects
observed in Experiment 2 generalize beyond the specificsetof
topics we asked participants to discuss.
```
## Experiment 3:DeeperConversations

```
Experiments 1a–2 indicated that deep conversations between
strangers unfolded better than expected, and that participants are
more likely to overestimate the awkwardness of deep conversa-
tions than shallow conversations. These results suggest that people
might have what they perceive to be better conversations if they
were willing to engage in deeper and more meaningful conversa-
tions in daily life. However, the preceding experiments do not
directly test this potentially important practical conclusion for two
reasons. First, we provided the conversation questions for partici-
pants. People may have more accurate beliefs about their conver-
sations in daily life because they choose which topics to discuss.
People may also be unable to generate deeper conversation topics
on their own that would enhance their connections and wellbeing.
Second, the preceding experiments lack a control condition to
assess whether engaging in deeper conversations than one would
normally engage in leads to better outcomes than expected.
Assessing this is critical for understanding whether miscalibrated
expectations could lead people to be overly shallow in conversa-
tion for their own wellbeing.
To address these limitations, participants in Experiment 3 wrote
the questions that they later discussed with a stranger themselves.
Participantsfirst wrote a series of questions that they would nor-
mally discuss while getting to know someone new, and then wrote
a series of questions that were deeper, involving topics that were
more intimate than they would normally discuss. We then ran-
domly assigned pairs to discuss the“control”questions or the
deeper questions that one of the participants had generated. As in
Experiments 1a–2, we predicted that participants would underesti-
mate how positive these conversations would leave them feeling,
and that participants in the deep condition would overestimate
how awkward the conversation would feel compared with partici-
pants in the control condition. This experiment also enables us to
test whether generating the topics for discussion oneself, versus
discussing topics that another person generated, affects the
expected or actual experience of relatively shallow or deep
conversations.
```
## Method

## Participants

```
We targeted 100 pairs of participants andfinished recruiting
once that target was reached after data exclusions. We achieved
this by recruiting 103 pairs of participants from separate university
and community subject pools (N= 200 individuals after exclu-
sions;Mage = 28.46;SDage= 13.67; 49.00% female; 31.50%
White) to complete the experiment in exchange for $6. We
excluded three of these pairs from analyses because one pair knew
one another beforehand, because one pair began their conversation
before one member had reported expectations, and because one
participant did not write out conversation questions.
```
## Procedure

```
Participants sat in separate rooms and did not interact with one
another before their conversations. Participants were told that they
would develop questions that they might later ask and answer dur-
ing a discussion. Both participantsfirst generatedfivecontrol
questions. Specifically, they were told:
```
```
We would like you to begin by generatingfive questions. These
should be the types of questions that you would naturally ask another
person whilefirst getting to know him or her. Please select questions
that you would actually be willing to ask and answer later in this study,
and these questions should be the types of questions that you would
typically ask whilefirst getting to know somebody.
```
```
The same participants then generatedfivedeepquestions. They
were told:
```
```
Next we would like you to generatefive more questions. This time,
please generatefive questions that are deeper and more intimate than
the types of questions that you would naturally ask another person
whilefirst getting to know him or her. In other words, we would like
you to generate questions that go beyond the surface, beyond small
talk, to probe deeper subject matter that might be more personal or
emotional. For example, you might ask the person about important
experiences they’ve had or activities they’ve enjoyed. You might ask
the person to reveal something important about them. These questions
should require both you and your partner to reveal something about
yourselves that you might not normally reveal in a conversation with a
stranger. These should be topics that you would be more likely, per-
haps, to talk about with a close friend or family member, and they
should dig deeper than the ones you wrote down in the previous set of
questions.
```
```
Pairs were then randomly assigned to either the control condi-
tion or the deep condition. In the control condition, the experi-
menter selected one of the two sets of control questions at random,
whereas in the deep condition, the experimenter selected one of
the two sets of deep questions at random. The participant who
wrote the randomly selected questions then sequenced them in the
order they preferred for the conversation.
Both participants then viewed thefinal set offive discussion
questions on a computer screen. The participant who did not write
thefinal set of questions knew that the questions were written
by the other participant but was not told whether they were view-
ing the control or deep questions. Participants then indicated how
awkward they expected to feel during the conversation, how
```
## DEEPER CONVERSATION 9

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
uncomfortable, how much they expected to enjoy the conversa-
tion, how strong a bond they would feel with their conversation
partner, how much they would like their conversation partner, how
well they would feel they got to know their conversation partner’s
true beliefs, attitudes, preferences, and character, and how happy
they would feel about their conversation on 11-point scales as in
the prior experiments. Participants then predicted their partner’s
experiences on the same measures. The order of the awkwardness
items (awkwardness, discomfort) and connectedness items (enjoy-
ment, strength of bond, liking, and perceived understanding) was
counterbalanced between pairs.
After reporting their expectations, participants entered the same
study room, viewed the discussion questions, and began their con-
versations. Participants discussed each of the five questions
sequentially until they reached their natural conclusions. Whenfin-
ished with their conversation, participants were again separated
into individual rooms and reported their experiences in private.
Participantsfirst reported their own experiences on the same meas-
ures used before the conversation, and then they predicted their
partners’responses on the same measures. Participants then com-
pleted the 10-item personality inventory (Gosling et al., 2003).
Finally, participants reported demographic information and
were thanked and debriefed.
```
## Results

```
The awkwardness items (aexpectations= .91,aexperiences= .86) and
connectedness items (aexpectations= .94,aexperiences= .94) were
highly correlated, and so we combined these items to form awk-
wardness and connectedness scales, respectively.
```
## Manipulation Check

```
To check whether the intimacy manipulation was effective, we
recruited a separate group of participants from MTurk (N= 409)
to rate the intimacy of the control and deep discussion questions.
The additional participants confirmed that the manipulation was
effective: Participants discussed items that were rated as more inti-
mate in the deep condition (M= 6.28,SD= 1.50) than in the con-
trol condition (M= 4.74,SD= 1.67),t(96) =
4.81,p,.001,
95% CIdifference[
2.17,
.90],d=
.97 (seeonline supplemental
materialsfor details).
```
```
We then conducted 2 (Conversation: Control, Deep) 3 2 (Mea-
surement Phase: Expectations, Experiences) ANOVAs with
repeated measures on the second factor and awkwardness, con-
nectedness, and happiness as dependent measures in separate
analyses.
```
## Awkwardness

```
Participants overestimated the awkwardness of their conversa-
tions somewhat more in the deep condition than the control condi-
tion, as evidenced by a marginally significant Conversation 3
Measurement Phase interaction effect,F(1, 98) = 3.74,p= .056,
g^2 p= .04 (seeFigure 3). Planned contrasts indicated that partici-
pants overestimated how awkward their conversation would be
both in the control condition,paired t(49) = 4.13,p,.001, 95%
CIdifference[.57, 1.65],d= .58, and in the deep condition,paired
t(49) = 6.65,p,.001, 95% CIdifference[1.30, 2.42],d= .94. How-
ever, participants in the deep condition also expected their conver-
sations to feel more awkward than participants in the control
condition,t(98) =
2.26,p= .026, 95% CIdifference[
1.82,
.12],
d=
.45, even though experiences of awkwardness did not differ
significantly between conditions,t(98) =
.60,p= .552, 95%
CIdifference[
.95, .51],d=
.12.
```
## Connectedness

```
Participants underestimated how connected they would feel af-
ter speaking with their partner, as indicated by a significant main
effect of Measurement Phase,F(1, 98) = 40.46,p,.001,g^2 p=
.29. The Conversation 3 Measurement Phase interaction effect
was nonsignificant,F(1, 98) = .01,p= .936,g^2 p= .0001. Planned
contrasts indicated that although participants underestimated how
connected they would feel in both the control,paired t(49) =

4.69,p,.001, 95% CIdifference[
1.44,
.57],d=
.66, and
deep conditions,paired t(49) =
4.32,p,.001, 95% CIdifference
[
1.44,
.52],d=
.61, participants in the deep condition did
expect to feel more connected to their partner than did participants
in the control condition,t(98) =
3.12,p= .002, 95% CIdifference
[
1.58,
.35],d=
.62. In this respect, their expectations were
calibrated at above-chance levels because participants in the deep
condition did indeed feel significantly more connected to their
```
```
Figure 3
Mean Awkwardness, Connectedness, and Happiness Across Conversation Conditions (Control
Versus Deep) and Measurement Phase (Expectations Versus Experiences) in Experiment 3
```
```
Note. Error bars 61 SE.
```
## 10 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
partner than did participants in the control condition,t(98) =

2.33,p= .022, 95% CIdifference[
1.74,
.14],d=
.47.
```
## Happiness

```
As with the experience of connection, participants also underesti-
mated how happy they would actually feel about their conversa-
tions, as indicated by a significant main effect of Measurement
Phase,F(1, 98) = 45.07,p,.001,g^2 p= .32. The Conversation 3
Measurement Phase interaction was nonsignificant,F(1, 98) = 1.84,
p=.179,g^2 p= .02. Planned contrasts indicated that participants
underestimated how happy they would feel about the conversation
in both the control,paired t(49) =
3.79,p,.001, 95% CIdifference
[
1.33,
.41],d=
.54, and deep conditions,paired t(49) =

5.70,p,.001, 95% CIdifference[
1.77,
.85],d=
.81. Partici-
pants did not expect to feel differently in the control and deep con-
ditions,t(98) =
.73,p=.465,95%CIdifference[
.85, .39],d=

.15, but participants in the deep condition reported marginally
greater happiness with the conversation when they werefinished,
t(98) =
1.72,p=.088,95%CIdifference[
1.44, .10],d=
.34.
```
## Secondary Analyses

```
The magnitude of miscalibration between expectations and
experiences did not differ significantly between participants who
wrote thefinal set of discussion questions (Writers) and those who
did not (Receivers) on any measures,Fs(1, 98)#2.28,ps$.134,
g^2 p#.02, suggesting that choosing the topics to discuss does not
meaningfully increase the accuracy of people’s expectations about
the outcomes of the conversation.
As in Experiment 2, ourfindings were not consistently moder-
ated in any clear way by personality. Overestimating the conversa-
tion’s awkwardness was not associated with any Big Five traits
among participants in the control condition, but was associated
with lower openness (b= .45,t(94) = 4.34,p,.001) among par-
ticipants in the deep condition. Underestimating connectedness
was not associated with any Big Five traits among participants in
either the control condition or the deep condition. Underestimating
happiness was associated with higher conscientiousness (b= .23,
t(94) = 2.01,p= .048) among participants in the control condition
but was not associated with any Big Five traits among participants
in the deep condition. Because we did not observe consistent mod-
eration of any of our results by personality in either Experiments 2
or 3, we did not collect measures of personality in the following
experiments.
```
## Discussion

```
Consistent with Experiments 1 and 2, participants in Experi-
ment 3 overestimated how awkward and uncomfortable a deep
conversation with a stranger would be, and also underestimated
how connected they would feel to their conversation partner and
how happy they would feel about the conversation afterward.
Moreover, we again observed that participants tended to overesti-
mate how awkward a deep conversation would be compared with
the conversations they would typically have with a stranger.
Although those who were having a deeper-than-normal conversa-
tion with a stranger expected the conversation to feel more awk-
ward and uncomfortable than those who were having a typical
```
```
conversation, the deeper conversations were no more awkward in
reality than the typical conversations.
Thesefindings extend our previous experiments in three impor-
tant ways. First, participants wrote and discussed theirownques-
tions, thereby replicating the primary results of Experiments 1a– 2
across a broad range of ecologically valid conversation topics.
Second, the gaps between expectations and experiences in these
conversations did not differ meaningfully between participants
who chose the discussion topics and those who did not, indicating
that people may overestimate the awkwardness of meaningful con-
versations even when they design the conversations themselves.
Third, this experiment suggests that people might have more posi-
tive conversations with strangers in daily life if they were willing
to dive a little deeper than they normally would.
Finally, these results suggest that miscalibrated expectations, par-
ticularly about the awkwardness and discomfort of a deep conversa-
tion, may create a barrier to having the deeper conversations that
participants in Experiments 1a–1c reported wishing they had more
often. Even though participants in both the control and deep condi-
tions underestimated how connected they would feel to their partner,
participants in the deep condition did anticipate feeling more con-
nected than did participants in the control condition. These partici-
pants seemed to recognize that a deeper conversation would
strengthen their relational bonds with their partner, but mistakenly
expected that it would also come with a cost of having a more awk-
ward and uncomfortable conversation. We designed Experiments
4a–7b to test the mechanism underlying these miscalibrated expecta-
tions, and to examine how they might encourage conversation that is
shallower than participants themselves might consider to be optimal.
```
## Experiments 4a–4b: Underestimating Care

```
We hypothesized that people undervalue deep conversation at
least partly because they underestimate how much others care
about, and are interested in, learning the content that will be shared
in the conversation itself. People have a fundamental need to feel
connected to others (Baumeister & Leary, 1995), a need we are
suggesting that people tend to underestimate in others, especially
in strangers. Assuming that others are somewhat indifferent to the
conversation could lead people to expect that a deep conversation
would be especially awkward and uncomfortable, thereby encour-
aging people to avoid more deep and intimate conversations.
Experiments 1a–1b provided some support for this mechanism,
but only included relatively deep conversations. We conducted a
more comprehensive test in Experiments 4a and 4b by providing
participants with either shallow or deep discussion questions and
then asking them to predict how much they would care about the
content of their own responses in the interaction, and also how
much their partner would care about the participant’s own
responses. Participants also reported their expectations of how
awkward the conversation would feel, how connected they would
feel to their partner, and how happy they would feel about the con-
versation. After speaking, participants reported their actual experi-
ences on the same measures. We predicted that participants would
expect to care about the content of their own responses more than
they would expect others to care. We also predicted that partici-
pants would underestimate how much others would care about the
content of their conversation. Further, we also predicted that
underestimating strangers’ care and concern would mediate
```
## DEEPER CONVERSATION 11

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
participants’tendency to overestimate awkwardness, but underes-
timate connectedness and happiness, during the conversation.
```
## Method (Experiment 4a)

## Participants

```
We recruited 109 pairs of participants during a master’s degree
student orientation event (N= 206 individuals after exclusions;
Mage= 36.83;SDage= 5.48; 29.61% female; 39.32% White). We
excluded six pairs from analyses because two pairs completed
their conversations before reporting their expectations, two pairs
included one participant who reported expectations about the
wrong set of conversation questions, and two pairs included at
least one participant who did not consent at the beginning of the
session to have their data analyzed in published research. In addi-
tion, 14 participants could not be matched with their partner after
the session because either one or three participants reported the
same pair ID. We did not analyze data from these participants.
```
## Procedure

```
The procedure was identical to Experiment 2 with two excep-
tions. First, we modified the discussion questions and reduced the
total number of questions fromfive to four (due to time constraints
in the experimental session). Participants in the shallow condition
received the following questions:
```
1. What do you think about the weather today?
2. What do you think of the city of Chicago?
3. How did you celebrate last Halloween?
4. How often do you get your hair cut? Where do you go?
    Have you ever had a really bad haircut experience?

```
Participants in the deep condition received the following
questions:
```
1. For what in your life do you feel most grateful? Tell your
    partner about it.
2. If a crystal ball could tell you the truth about yourself,
    your life, your future, or anything else, what would you
    want to know?
3. If you were going to become a close friend with your
    partner, please share what would be important for him or
    her to know.
4. Can you describe a time you cried in front of another
    person?

```
Second, we modified the dependent measures. Before meeting
their conversation partners, participants reported how much they
thought they would care about and feel concerned or interested in
their own responses to the questions (0 =not at all;10=quite a
bit) and also how much they thought their partner would care
about and feel concerned or interested in the participant’s
responses (0 =not at all;10=quite a bit). These items allow us to
test whether participants expect to care more about the content of
```
```
the conversation than their partner, and if so, whether this self/
other caring gap would be especially pronounced in the deep con-
dition. Participants then reported how awkward they would feel
during the discussion, how strong of a bond they would feel with
their partner, how much they would like their partner, and how
happy they would feel about the conversation with their partner on
11-point scales as in the prior experiments. After reporting their
expectations but before speaking, participants indicated how often
they wished they would engage in small talk with strangers (
5=
much less often than I do now;0=neither less nor more often
than I do now;5=much more often than I do now) and how often
they wished they would engage in deep conversations with strang-
ers (
5=much less often than I do now;0=neither less nor more
often than I do now;5=much more often than I do now). Partici-
pants then engaged in the conversations and reported their experi-
ences on the same measures used for measuring their expectations.
Finally, participants reported demographic information and
were thanked and debriefed.
```
## Results (Experiment 4a)

```
The strength-of-bond and liking measures were highly corre-
lated in both expectations (a= .84) and experiences (a= .68) and
so we combined these items to form a connectedness scale.
```
## Care Measures

```
As predicted, participants expected to care more about their
own responses than their partner would,F(1, 101) = 32.90,p,
.001,g^2 p= .25. Further, this self/other caring gap was significantly
larger in the deep condition than in the shallow condition: A 2
(Conversation: Shallow, Deep) 3 2 (Target: Self, Partner)
ANOVA on participants’expectations yielded a significant inter-
action,F(1, 101) = 6.00,p= .016,g^2 p= .06. Although participants
in the shallow condition expected to care significantly more about
their responses to the questions (M= 5.69,SD= 1.58) than their
partner would (M= 5.11,SD= 1.56),paired t(51) =
2.60,p=
.012, 95% CIdifference[
1.04,
.13],d=
.36, this gap was signif-
icantly larger among participants in the deep condition (Ms = 7.
vs. 6.07, respectively;SDs = 1.24 vs. 1.41),paired t(50) =
5.26,
p,.001, 95% CIdifference[
2.02,
.90],d=
.74.
Perhaps more important, comparisons with actual experience
indicated that participants systematically underestimated how
much they would actually perceive their partner to care about their
responses. A 2 (Conversation: Shallow, Deep) 3 2 (Measurement
Phase: Expectations, Experiences) ANOVA on the partner care
measure with repeated measures on the second factor produced a
significant main effect of Measurement Phase,F(1, 101) = 93.22,
p,.001,g^2 p= .48, such that participants underestimated partner
care. The Conversation 3 Measurement Phase interaction effect
was nonsignificant,F(1, 101) = .29,p= .591,g^2 p= .003, indicating
that participants underestimated the recipients’care similarly in
the deep and shallow conditions. Consistent with our hypotheses,
participants also underestimated their partner’s care significantly
more than they underestimated their own,F(1, 101) = 61.70,p,
.001,g^2 p= .38.
We then conducted a series of 2 (Conversation: Shallow,
Deep) 3 2 (Measurement Phase: Expectations, Experiences)
```
## 12 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
ANOVAs with repeated measures on the second factor, separately
for the awkwardness, connectedness, and happiness measures.
```
## Awkwardness

```
Participants in the deep condition overestimated how awkward
their conversation would feel significantly more than did partici-
pants in the shallow condition, as indicated by a significant Con-
versation 3 Measurement Phase interaction,F(1, 101) = 4.30,p=
.041,g^2 p= .04. Although participants overestimated how awkward
the conversation would be in the shallow condition,paired t(51) =
3.76,p,.001, 95% CIdifference[.51, 1.70],d= .52, participants
overestimated feelings of awkwardness significantly more in the
deep condition,paired t(50) = 6.44,p,.001, 95% CIdifference
[1.37, 2.61],d= .90 (seeFigure 4, top panel).
```
## Connectedness

```
Participants underestimated how connected they would feel
across conditions, as evidenced by a significant main effect of
Measurement Phase,F(1, 101) = 144.32,p,.001,g^2 p= .59.
Unexpectedly, the Conversation 3 Measurement Phase interaction
effect was also significant,F(1, 101) = 11.76,p,.001,g^2 p= .10:
Although participants underestimated how connected they would
feel in the deep condition,paired t(50) =
6.85,p,.001, 95%
CIdifference[
1.49,
.81],d=
.96, they did so significantly more
in the shallow condition,paired t(51) =
9.94,p,.001, 95%
CIdifference[
2.49,
1.65],d=
1.38. Participants in the deep
condition expected to feel more connected than did participants in
the shallow condition,t(101) = 5.43,p,.001, 95% CIdifference
[.89, 1.91],d= 1.07, and actually did feel more connected in the
```
```
deep condition than in the shallow condition,t(101) = 2.46,p=
.016, 95% CIdifference[.09, .86],d= .48. The significant interaction
indicates that these differences in expectation were larger (d=
1.07) than the differences in experience (d= .48).
```
## Happiness

```
Participants underestimated happiness across conditions, as evi-
denced by a significant main effect of Measurement Phase,
F(1, 101) = 130.18,p,.001,g^2 p= .56. Unexpectedly, the Conver-
sation 3 Measurement Phase interaction effect was also significant
for happiness,F(1, 101) = 13.18,p,.001,g^2 p= .12: Although par-
ticipants significantly underestimated how happy they would feel
about the conversation in the deep condition,paired t(50) =
6.09,
p,.001, 95% CIdifference[
1.36,
.68],d=
.85, they did so sig-
nificantly more in the shallow condition,paired t(51) =
9.80,p,
.001, 95% CIdifference[
2.37,
1.57],d=
1.36. Participants in the
deep condition expected to feel happier about their conversation
than did participants in the shallow condition,t(101) = 3.55,p,
.001, 95% CIdifference[.42, 1.47],d= .70, but experienced happiness
with the conversation did not vary by condition,t(101) =
.05,p=
.962, 95% CIdifference[
.43, .41],d=
.01.
```
## Mediational Analyses

```
We found only limited evidence that underestimating a partner’s
care could explain why participants overestimated how awkward
and uncomfortable their interaction would be, while also underes-
timating how connected and happy they would feel following the
conversation. We performed a series of within-pairs mediational
analyses using Measurement Phase (Expectation, Experience) as
```
```
Figure 4
Mean Awkwardness, Connectedness, and Happiness Across Conversation Conditions (Shallow
Versus Deep) and Measurement Phase (Expectations Versus Experiences) in Experiments 4a and 4b
```
```
Note. Error bars 61 SE.
```
## DEEPER CONVERSATION 13

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
the independent variable, partner care as the mediating variable,
and awkwardness, connectedness, and happiness as dependent var-
iables in separate analyses. In the shallow condition, the indirect
effects were nonsignificant for awkwardness (b= .19, 95% CI
[
.26, .64]), nonsignificant for connectedness (b=
.44, 95%
CI [
1.02, .03]), and significant for happiness (b=
.43, 95% CI
[
.96,
.03]). In the deep condition, the indirect effects were non-
significant for awkwardness (b= .14, 95% CI [
.32, .68]), signifi-
cant for connectedness (b=
.47, 95% CI [
.89,
.17]), and
nonsignificant for happiness (b=
.33, 95% CI [
.65, .06]).
Thus, Experiment 4a provides some evidence that misunderstand-
ing how much others will care explains why deep conversations
unfold more positively than people expect, but it does not provide
strong evidence of this.
```
## Desire for Shallow and Deep Conversation

```
Participants reported somewhat stronger desires to engage in
deep conversation than shallow conversation with strangers. Par-
ticipants reported that they wanted to engage in more small talk
with strangers than they typically do (M= 1.12,SD= 2.35),
t(205) = 6.82,p,.001, 95% CI [.79, 1.44],d= .47, and also that
they wanted to engage in more deep conversation with strangers
than they typically do (M= 1.48,SD= 2.12),t(205) = 9.99,p,
.001, 95% CI [1.18, 1.77],d= .70. Participants’preference to
engage in more deep conversation was marginally stronger than
their preference to engage in more small talk,paired t(205) =

1.90,p= .059, 95% CIdifference[
.73, .01],d=
.13. These
results are important because they indicate that our participants
would generally prefer interacting with others more, rather than
less, suggesting that any reluctance to engage in deeper conversa-
tion than one normally would does not stem from participants’
own disinterest in doing so. These preference ratings instead sug-
gest a psychological barrier that may keep people from engaging
as deeply with others as much as they might prefer.
```
## Method and Results (Experiment 4b)

```
We conducted a replication experiment with only minor
changes to the procedure (N= 101 pairs; seeonline supplemental
materialsfor the full method and results). Experiment 4b largely
replicated the results of Experiment 4a: Participants overestimated
the awkwardness of their conversations marginally more in the
deep condition than the shallow condition,F(1, 99) = 3.69,p=
.058,g^2 p= .04, but underestimated their connectedness signifi-
cantly more in the shallow condition than the deep condition,F(1,
99) = 17.96,p,.001,g^2 p= .15, and underestimated their happi-
ness marginally more in the shallow condition,F(1, 99) = 3.21,
p= .076,g^2 p= .03 (seeFigure 4, bottom panel).
Experiment 4b also found somewhat stronger evidence that
underestimating a partner’s care could explain why people under-
value the positive outcomes of their conversations. We conducted
mediational analyses with Measurement Phase (Expectation, Ex-
perience) as the independent variable, partner care as the media-
ting variable, and each of the primary measures as dependent
variables in separate analyses. In the shallow condition, the indi-
rect effects were nonsignificant for awkwardness (b= .35, 95% CI
[
.31, .99]), but significant for connectedness (b=
.54, 95% CI
[
1.02,
.19]), and happiness (b=
.36, 95% CI [
.63,
.17]).
```
```
In the deep condition, the indirect effects were likewise nonsignifi-
cant for awkwardness, (b= .37, 95% CI [
.25, .92]), but signifi-
cant for connectedness (b=
.53, 95% CI [
.93,
.24]), and
happiness (b=
.39, 95% CI [
.89,
.06]).
```
## Discussion (Experiments 4a–4b)

```
Participants underestimated how positive they would feel in
conversation with a stranger, expecting to feel more awkward and
less connected and happy than they actually did. Consistent with
the prior experiments, participants overestimated how awkward
the conversation would feel to a greater extent in the deep condi-
tions than in the shallow conditions. Participants underestimated
how much others would care about the content of the conversation,
but we found only mixed evidence that underestimating strangers’
care helped to explain why participants underestimated the overall
positivity of their conversations.
These experiments also revealed an unexpected pattern of
results that we did not observe in Experiments 2 and 3. Although
participants in the shallow condition expected to feel substantially
less connected to one another than participants in the deep condi-
tion, participants in both conditions felt strongly connected after
speaking. As a result, participants in the shallow condition under-
estimated connectednessmorethan participants in the deep condi-
tion. Anecdotal reports during the session debriefing indicated that
participants in the shallow condition often turned what we
intended to be relatively superficial conversations into deeper and
more meaningful conversations than they had anticipated, meaning
that our experimental manipulations were not as strong as
intended. Although evaluations of intimacy differed significantly
between the deep (M= 7.18,SD= 1.61) and shallow conditions
(M= 5.92,SD= 1.84) in Experiment 4b,t(64) = 2.95,p= .004,
95% CIdifference[.41, 2.11],d= .73, intimacy ratings even in the
shallow condition exceeded the midpoint of the scale,one-sample
t(32) = 2.88,p= .007, 95% CI [5.27, 6.58],d= .50. Instead of
being limited to the questions that we provided, which would have
guided people’s expectations, actual conversations seemed to drift
in a more meaningful direction than people may have anticipated
based on the questions alone. We think this is a potentially inter-
esting and unexpected consequence in some conversations, which
might help to explain why conversations with strangers can be sur-
prisingly positive (Epley & Schroeder, 2014). We will discuss this
possibility in greater detail in the General Discussion.
```
## Experiment 5: Manipulating Perceived Care via

## Relationships

```
Experiments 1a–4b suggest that people expect deeper conversa-
tions to create stronger connections than shallow conversations,
but also that deeper conversations will be relatively awkward and
uncomfortable. These experiments also indicate that these expecta-
tions are somewhat miscalibrated, such that people underestimate
how connected they will feel to their partners and overestimate
how awkward and uncomfortable their deep conversations will be.
However, Experiments 4a and 4b provide only mixed evidence
that miscalibrated expectations about the consequences of deep
conversation stem from underestimating others’care and interest
in learning more intimate information about oneself. Perhaps more
important, correlational analyses of mediation do not test the
```
## 14 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
causal influence of a mediator, which only an experimental
approach to testing mediation can do. We designed Experiment 5
to provide a causal test of the importance of a partner’s perceived
care and interest by manipulating it directly through the strength
of an existing social relationship. Specifically, we asked partici-
pants to have a conversation with either a distant stranger or with a
close other who would already be known to care about one’s
responses to intimate conversation topics, including a friend, fam-
ily member, or romantic partner. We predicted that participants
would be less likely to underestimate a close other’s interest com-
pared with a distant stranger’s interest, and would also be more
calibrated when predicting feelings of awkwardness, enjoyment,
and happiness before speaking with a close other compared with a
distant stranger.
We also designed Experiment 5 to address a potential methodo-
logical concern. In the previous experiments, participants did not
see their partner before reporting their expectations, and so it is
possible that they imagined meeting an atypical stranger who
would be even less interested in the conversation than the stranger
they were later paired with. We theorize instead that people under-
estimate how much even atypicalstranger will care about the con-
tent of a conversation. To address this concern, participants in
Experiment 5 were able to see their partner in the stranger condi-
tion at the start of the study session. We still predicted, consistent
with the previous experiments, that participants in the“stranger”
condition would underestimate their partner’s interest and under-
estimate how positively their conversations would unfold.
```
## Method

## Participants

```
We targeted 200 pairs of participants andfinished recruiting
once that target was reached after data exclusions. We achieved
this by recruiting 204 pairs of participants from several public
parks (N= 400 individuals after exclusions;Mage= 35.27;SDage=
16.21; 61.50% female; 61.50% White) to complete the study in
exchange for a $5 gift card. We excluded four of these pairs from
analyses because both participants in one pair answered their
phones during the conversation, and because participants in three
pairs discussed the dependent measures while responding to those
measures. Among 200 participants in the“close other”condition,
31.50% reported that they were friends, 24.50% reported that they
were spouses, 21.50% reported that they were dating, 17.00%
reported that they were family members, 3.00% reported that they
were acquaintances, 2.00% reported that they were colleagues,
and .50% did not report the nature of their relationship.
```
## Procedure

```
Experimenters recruited either pairs of distant strangers who
had never met one another or pairs of close friends, family, or part-
ners who were visiting the park together. Participants in both the
close and distant conditions saw their conversation partner at the
beginning of the experiment before receiving instructions. After
providing informed consent, participants in the close conditions
reported how close or connected they currently felt to the other
person (0 =not close at all;10=extremely close).
In both conditions, participants were separated and were then told
by different experimenters that they would develop a set of discussion
```
```
questions that they might later ask and answer with the other partici-
pant. Following a similar procedure to that used in Experiment 3, par-
ticipantsfirst generated two control questions, and then were asked to
generate two deeper questions. The instructions in the distant stranger
were nearly identical to those used in Experiment 3, while the instruc-
tions in the close other condition asked participants to write questions
that they typically discuss with this person (control) or that were
deeper than they normally discuss with this person (see theonline
supplemental materialsfor complete instructions).
The procedure was then largely identical to that of Experiment
3, except that we included slightly different dependent variables
given that participants in the close condition were already ac-
quainted and connected with each other. In particular, participants
first reported how much they would care about and feel concerned
or interested in their own responses, and how much their partner
would care about and feel concerned or interested in the partici-
pant’s responses, on separate scales. Participants then reported
how awkward they would feel during the discussion, how much
they would enjoy the conversation, and how happy they would
feel about the conversation, on 11-point scales as in the prior
experiments. The awkwardness and enjoyment items were coun-
terbalanced between pairs. As a manipulation check, participants
then viewed the control questions and the deep questions that they
had written earlier—regardless of whether their own questions or
their partner’s questions were selected for the discussion—and
separately rated the intimacy of each pair of questions (0 =not
intimate at all;10=extremely intimate).
Participants then viewed the discussion questions and began
their conversations as in Experiment 3, and then reported their
experiences on the same measures used to measure their expecta-
tions before the conversation. As a second manipulation check,
participants also reported how intimate their conversation was (0 =
not at all intimate;10=extremely intimate).
Finally, participants reported demographic information and
received their compensation.
```
## Results

## Manipulation Checks

```
The intimacy manipulation was effective. Participants reported
that their questions were less intimate in the control conditions
than in the deep conditions,F(1, 196) = 214.30,p,.001,g^2 p=
.52, and this pattern did not differ depending on whether they were
distant strangers or close others,F(1, 196) = .61,p= .436,g^2 p=
.003. After the conversation, participants likewise reported having
less intimate conversations in the control conditions than in the
deep conditions,F(1, 196) = 63.05,p,.001,g^2 p= .24, again
regardless of whether they were distant strangers or close others,
F(1, 196) = .20,p= .658,g^2 p= .001. Participants also reported
having deeper conversations with close others than with distant
strangers,F(1, 196) = 10.48,p= .001,g^2 p= .05.
```
## Care Measures

```
As anticipated, participants expected to care about their own
responses more than their partner would across experimental con-
ditions,F(1, 196) = 6.73,p= .010,g^2 p= .03. Furthermore, this
self/other caring gap was significantly larger in the deep conditions
```
## DEEPER CONVERSATION 15

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
than in the control conditions, as we observed in Experiment 4.
Specifically, a 2 (Conversation: Control, Deep) 3 2 (Relationship:
Distant, Close) 3 2 (Target: Self, Partner) ANOVA on partici-
pants’expectations yielded a significant Conversation 3 Target
interaction,F(1, 196) = 8.86,p= .003,g^2 p= .04. Participants in
the deep conditions expected to care more about their responses to
the questions than their partner would,F(1, 98) = 19.11,p,.001,
g^2 p= .16, but participants in the control conditions did not,F(1,
98) = .06,p= .805,g^2 p= .001.
More important for the unique contribution of this experiment,
pairs in the distant conditions underestimated each other’s care
more than did pairs in the close conditions. A 2 (Conversation:
Control, Deep) 3 2 (Relationship: Distant, Close) 3 2 (Measure-
ment Phase: Expectations, Experiences) ANOVA on the partner
care measure with repeated measures on the third factor produced
a significant Relationship 3 Measurement Phase interaction effect,
F(1, 196) = 55.84,p,.001,g^2 p= .22. Although participants who
spoke with close others underestimated how much their partner
would care about their responses,F(1, 98) = 4.13,p= .045,g^2 p=
.04, those who spoke with distant strangers did so significantly
more,F(1, 98)= 156.61,p,.001,g^2 p= .62.
We next report a series of 2 (Conversation: Control, Deep) 32
(Relationship: Distant, Close) 3 2 (Measurement Phase: Expecta-
tions, Experiences) ANOVAs with repeated measures on the third
factor, separately for the awkwardness, enjoyment, and happiness
measures.
```
## Awkwardness

```
Participants in the deep conditions overestimated how awkward
their conversations would be more than did participants in the control
conditions, as revealed by a significant Conversation 3 Measurement
Phase interaction effect,F(1, 196) = 4.94,p= .027,g^2 p=.02.We
also observed a significant Relationship 3 Measurement Phase inter-
action effect,F(1, 196) = 11.05,p= .001,g^2 p=.05(seeFigure 5):
Although participants in the close conditions overestimated how
awkward and uncomfortable their conversations would feel,F(1,
98) = 19.36,p,.001,g^2 p= .16, participants in the distant conditions
did so significantly more,F(1, 98) = 86.34,p,.001,g^2 p= .47.
Analyses of simple effects within this interaction indicated that
among distant strangers, participants in the deep condition
expected that their conversations would feel marginally more awk-
ward than those in the control condition,t(98) =
1.88,p= .062,
95% CIdifference[
1.46, .04],d=
.38, but did not feel signifi-
cantly different in experience (Ms = 1.49 vs. 1.56, respectively;
SDs = 1.50 vs. 1.52),t(98) = .23,p= .817, 95% CIdifference[
.53,
.67],d= .05. Among close others, participants in the deep condi-
tion likewise expected their conversations to feel more awkward
than did those in the control condition,t(98) =
2.58,p= .011,
95% CIdifference [
1.68, 
.22], d = 
.52, but felt only
marginally more awkward in actual experience (Ms = 1.88 vs.
1.20, respectively;SDs = 2.10 vs. 1.52),t(98) =
1.86,p= .066,
95% CIdifference[
1.41, .05],d=
.37. Once again, deep conver-
sations were not as awkward and uncomfortable as expected.
```
```
Figure 5
Mean Awkwardness, Enjoyment, and Happiness Across Conversation Conditions (Control Versus
Deep), Relationship (Distant vs. Close) and Measurement Phase (Expectations Versus Experiences) in
Experiment 5
```
```
Note. Error bars 61 SE.
```
## 16 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


## Enjoyment

```
As predicted, participants underestimated how much they would
enjoy their conversations across conditions, as indicated by a sig-
nificant main effect of Measurement Phase,F(1, 196) = 135.31,
p,.001,g^2 p= .41. More important, we also observed a significant
Relationship 3 Measurement Phase interaction effect,F(1, 196) =
29.58,p,.001,g^2 p= .13: Although participants in the close con-
ditions underestimated how much they would enjoy their conver-
sations,F(1, 98) = 21.45,p,.001,g^2 p= .18, participants in the
distant conditions underestimated their enjoyment significantly
more,F(1, 98) = 131.76,p,.001,g^2 p= .57. Neither the main
effect of Conversation, nor interactions with Conversation, were
significant,Fs(1, 196)#2.15,ps$.144,g^2 ps#.01, indicating
that participants expected and experienced similar enjoyment in
the control and deep conditions.
```
## Happiness

```
Participants underestimated how happy they would feel about
their conversations across conditions, as indicated by a significant
main effect of Measurement Phase,F(1, 196) = 188.47,p,.001,
g^2 p= .49. More important, we also observed a significant Relation-
ship 3 Measurement Phase interaction effect,F(1, 196) = 29.96,
p,.001,g^2 p= .13: Although participants in the close conditions
underestimated how happy they would feel about their conversa-
tions,F(1, 98) = 34.31,p,.001,g^2 p= .26, participants in the dis-
tant conditions did so significantly more,F(1, 98) = 183.11,p,
.001,g^2 p= .65. Neither the main effect of Conversation, nor inter-
actions with Conversation, were significant,Fs(1, 196)#3.28,
ps$.072,g^2 ps#.02, indicating that participants expected and
experienced similar levels of happiness in the control and deep
conditions.
```
## Mediational Analyses

```
We predicted that participants would overestimate awkward-
ness, and underestimate their enjoyment and happiness, because
they would underestimate how much their partner would care
about one’s responses. We did not observe support for these
hypotheses in the control conditions, but observed stronger support
in the deep conditions.
Specifically, we performed a series of within-pairs mediational
analyses with Measurement Phase (Expectation, Experience) as
the independent variable and partner care as the mediating vari-
able. Among participants in the control condition who spoke with
distant strangers, the indirect effects were nonsignificant for awk-
wardness (b= .25, 95% CI [
.25, .82]), nonsignificant for enjoy-
ment (b=
.30, 95% CI [
.76, .19]), and nonsignificant for
happiness (b=
.31, 95% CI [
.63, .05]). Among participants in
the control condition who spoke with close others, the indirect
effects were nonsignificant for awkwardness (b= .02, 95% CI
[
.03, .24]), nonsignificant for enjoyment (b = 
.07, 95%
CI [
.41, .09]), and nonsignificant for happiness (b=
.06, 95%
CI [
.29, .11]).
Among participants in the deep condition who spoke with dis-
tant strangers, the indirect effects were significant for awkward-
ness (b= .88, 95% CI [.36, 1.44]), nonsignificant for enjoyment
```
```
(b=
.28, 95% CI [
1.13, .51]), and significant for happiness
(b=
.46, 95% CI [
.98,
.12]). Among participants in the deep
condition who spoke with close others, the indirect effects were
nonsignificant for awkwardness (b=
.004, 95% CI [
.25, .19]),
significant for enjoyment (b= 
.21,95% CI [
.50,
.03]), and
significant for happiness (b=
.14, 95% CI [
.36,
.02]). People
may underestimate the positivity of deep conversations in part
because their conversation partners are more caring and interested
than people anticipate.
Furthermore, underestimating others’care helped to explain
why participants who spoke with distant strangers were more
likely to underestimate enjoyment and happiness than those who
spoke with close others. We performed a series of between-pairs
mediational analyses with Relationship (Close, Distant) as the in-
dependent variable, underestimation of partner care as the media-
ting variable, and the magnitude of miscalibration for each of the
primary measures as dependent variables in separate analyses,
using the PROCESS macro (Hayes, 2013). Among participants in
the control conditions, the indirect effects were nonsignificant for
awkwardness (b= .27, 95% CI [
.04, .69]), significant for enjoy-
ment (b=
.57, 95% CI [
.97,
.16]), and significant for happi-
ness (b=
.54, 95% CI [
.88,
.24]). Among participants in the
deep conditions, the indirect effects were also nonsignificant for
awkwardness (b= .34, 95% CI [
.004, .78]), significant for enjoy-
ment (b=
.41, 95% CI [
.91,
.003]), and significant for happi-
ness (b=
.37, 95% CI [
.71,
.14]). People’s expectations of
conversations with close others may be more calibrated because
they more correctly recognize how much close others will care
about the content of their conversation.
```
## Secondary Analyses

```
Consistent with Experiment 3, the primary results did not differ
meaningfully between Writers (who wrote thefinal set of discus-
sion questions) and Receivers (who did not). A series of 2 (Con-
versation: Control, Deep) 3 2 (Relationship: Distant, Close) 32
(Measurement Phase: Expectations, Experiences) 3 2 (Role:
Writer, Receiver) ANOVAs on awkwardness, enjoyment, and hap-
piness with repeated measures on the third and fourth factors
yielded nonsignificant Measurement Phase 3 Role interaction
effects for awkwardness,F(1, 196) = 3.50,p= .063,g^2 p= .02,
enjoyment,F(1, 196) = 1.94,p= .166,g^2 p= .01, and happiness,
F(1, 196) = .31,p= .578,g^2 p= .002. People seem to underestimate
the positive outcomes of their conversations even when they gen-
erate the topics themselves.
```
## Discussion

```
Experiment 5 reveals that people are more calibrated anticipat-
ing how much close friends, family members, and partners will
care about their conversation. As a result, people are also more
calibrated predicting how awkward, enjoyable, and happy they
will feel in the conversation with close others compared with dis-
tant strangers. As we observed in prior experiments, people over-
estimate how awkward deep conversations with strangers will be
more than typical conversations, but underestimate how enjoyable
and happy they will feel about both typical and deeper conversa-
tions. We did not measure connectedness in Experiment 5, which
sometimes varied by conversation content in prior experiments,
```
## DEEPER CONVERSATION 17

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
because we directly manipulated social connection across condi-
tions in this experiment.
Because people expect close others to be relatively more caring
and interested than distant strangers, our theory also predicted that
people would prefer to have deeper conversations with close
others. The data from Experiment 5 allow us to begin testing this
hypothesis because participants in the control conditions discussed
questions that they would typically discuss with their conversation
partner (close others) or that they typically discuss while getting to
know somebody new (distant strangers). In these control condi-
tions, participants who spoke with close others (M= 3.70,SD=
2.65) reported that their conversation topics were more intimate
than those who spoke with distant strangers (M= 2.50,SD=
2.40),t(98) = 2.37,p= .020, 95% CIdifference[.20, 2.20],d= .47.
After the conversation, participants who spoke with close others
(M= 5.03,SD= 2.46) also reported marginally more intimate con-
versations than those who spoke with distant strangers (M= 4.21,
SD= 2.22),t(98) = 1.75,p= .083, 95% CIdifference[
.11, 1.75],
d= .35. People may engage in deeper conversations with others
who they perceive to be caring and interested. Underestimating
distant strangers’care may create a barrier to having deeper and
more intimate conversations.
Of course, Experiment 5 was not designed to test whether people
prefer to have deeper conversations with close others than distant
strangers. We conducted a supplementary experiment on MTurk (N
= 109) in which participants imagined visiting the lab for a conver-
sation with either a close friend or a distant stranger. Participants
viewed a list of 20 pretested discussion questions (seeAppendix)
and selected thefive questions they preferred to discuss with the
other person. Participants in the“friend”condition (M=3.10,SD=
1.51) chose a greater number of deep questions than participants in
the“stranger”condition (M=1.79,SD= 1.32),t(107) = 4.82,p,
.001, 95% CI [.77, 1.84],d= .92, and likewise selected questions
higher in average intimacy (M=5.04,SD= 1.24) than did partici-
pants in the stranger condition, (M=3.97,SD=1.10),t(107) =
4.76,p,.001, 95% CIdifference[.62, 1.51],d=.91.Furthermore,
participants who imagined speaking with a close friend reported
significantly greater interest in discussing the deep questions (M=
7.00,SD=1.64),t(107) = 4.28,p,.001, 95% CIdifference[.83,
2.26],d= .82, and expected the other person to care more about
one’s own responses to the deep questions (M=7.11,SD= 1.42),t
(107) = 3.67,p,.001, 95% CIdifference[.56, 1.88],d=.70,com-
pared with participants who imagined speaking with a distant
stranger (Ms = 5.45 & 5.89, respectively;SDs = 2.08 & 1.98; see
online supplemental materialsfor the full method and results).
Thesefindings provide further evidence that people refrain from
having deep and intimate conversations when they are concerned
that another person will be uncaring and indifferent toward the con-
versation, and suggest that underestimating strangers’care may cre-
ate a psychological barrier to having deeper conversations with
distant others. Experiments 7a-7b will test whether removing this
barrier encourages people to choose deeper conversation topics.
```
## Experiments 6a–6b: Engaging in ShallowandDeep

## Conversations

```
Our data thus far reveal that participants overestimate the awk-
wardness of deep conversations more than shallow ones, and
expect a wide range of conversations to lead to weaker
```
```
connections and less happiness than they do. However, two limita-
tions of the previous designs have hindered our ability to test for
differences in experiences between shallow and deep conversa-
tions. We address both limitations in the following experiments.
First, participants have rated their experiences near the ceilings of
our connectedness scales: Participants tend to feel highly con-
nected to their conversation partner after a shallowordeep conver-
sation. We address this limitation in Experiments 6a and 6b by
instructing participants to engage in a shallow conversation with
one stranger and a deep conversation with another stranger, ena-
bling clearer comparisons of their connections to two different
partners. We hypothesized that participants would report feeling
more connected to their deep conversation partner.
Second, in the prior experiments, we included both positive
measures of partner care, connectedness, and happiness, and nega-
tive measures of awkwardness. People perceive their conversation
partner to be more caring during deep conversations, and at times
feel happier and more connected to the person, but they have also
felt more awkward during deep conversations in some experi-
ments. This mix of positive and negative experiences does not
indicate whether participants themselves prefer their overall expe-
riences in deep conversation relative to shallow conversation, nor
whether engaging in these conversations would affect their prefer-
ence for future shallow versus deep conversations. To address
these limitations, participants in Experiment 6b engaged in both
types of conversations and then indicated which conversation they
preferred. We hypothesized that participants would tend to prefer
their deep conversations.
```
## Method (Experiment 6a)

## Participants

```
We recruited 30 groups of four participants (N= 120 individu-
als;Mage= 28.33;SDage= 11.89; 65.00% female; 39.17% White).
Among these, 10 groups were recruited during an event advertised
as“Dinner and a Study Night”in exchange for $10 and dinner,
and 20 groups were recruited through the university’s Virtual Lab
using the Zoom video conferencing software, in exchange for an
$8 digital gift card, due to the coronavirus disease 2019
(COVID-19) pandemic. Two groups were excluded from analyses:
In one group, the participants engaged in their second conversation
before reporting expectations, and in the other, Internet connection
problems prevented the participants fromfinishing their conversa-
tions (seeonline supplemental materialsfor the analyses including
these groups). In addition, another six groups were removed from
analyses and were not analyzed: In four groups, one participant
typed in the wrong condition code and reported expectations about
the wrong topics before each conversation, and in two groups, one
participant typed in the wrong group ID and could not be matched
in the data with their group members.
```
## Procedure

```
Participants were randomly assigned to one of four roles (A, B,
C, or D) to determine whom they would speak with during each
round of conversation. Participants were then informed that they
were about to speak with another person who they had never met
before. They then read the three conversation questions. In the shal-
low-first counterbalance, participants read the following questions:
```
## 18 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


1. How is your day going so far?
2. What do you think about the weather today?
3. How often do you come here?^4

```
In the deep-first counterbalance, participants instead read the
following questions:
```
1. For what in your life do you feel most grateful? Please
    tell me about it.
2. If you could undo one mistake you have made in your
    life, what would it be and why would you undo it?
3. Can you tell me about one of the last times you cried in
    front of another person?

```
Participants then reported their expectations about the conversa-
tion, including how much they would care about and feel con-
cerned or interested in their own responses, how much their
partner would care about and feel concerned or interested in one’s
responses, how awkward they would feel while speaking with the
other person, how strong of a bond they would experience with
the other person, how much they would like the other person, and
how happy they would feel after the discussion, on 11-point scales
as in the prior experiments.
Participants were then paired with another group member
(A with D; B with C) for the discussion. Because anecdotal evi-
dence suggested that participants in the prior experiments dis-
cussed content outside the scope of the conversation questions,
potentially weakening the intimacy manipulations, the experi-
menter instructed participants not to drift off-topic while discus-
sing the questions. After their conversations, participants reported
their actual experiences on the same measures described above.
Participants also reported how intimate the conversation was (0 =
light conversation and small talk;10=meaningful conversation
and deep talk) and how negative or positive the topics they dis-
cussed were (
5=very negative;0=equally positive and nega-
tive;5=very positive).
Participants were then informed that they would discuss a new
set of conversation questions with a different person who they had
not met before. Participants in the shallow-first counterbalance
now read the deep questions, while those in the deep-first counter-
balance now read the shallow questions. Participants reported their
expectations, engaged in their conversations (A with C; B with D),
and reported their experiences on the same measures described
above. Afterfinishing the second conversation, participants addi-
tionally reported which person they felt more connected to (
5=I
feel much more connected to the FIRST person;0=I feel equally
connected to BOTH people;5=I feel much more connected to the
SECOND person), then indicated how intimate their typical con-
versations with strangers are versus how intimate they wished
these conversations would be on separate scales. Participants
recruited to the Virtual Lab then indicated whether they had trou-
ble seeing or hearing the other participants during their conversa-
tions over Zoom.
Finally, participants completed demographic items and were
thanked and debriefed.
```
## Method (Experiment 6b)

## Participants

```
We recruited 30 groups of four participants (N= 120 individu-
als;Mage= 26.01;SDage= 9.15; 61.67% female; 25.83% White) in
exchange for a $7 digital gift card through the university’s
Virtual Lab.
```
## Procedure

```
The procedure was identical to Experiment 6a with several
exceptions. First, after providing informed consent, participants
immediately read both sets of conversation questions and then
indicated which conversation they thought they would prefer
(
5=will strongly prefer Conversation A;0=will equally prefer
both conversations;5=will strongly prefer Conversation B).
Second, we removed the remaining expectation items before each
conversation (care, awkwardness, connectedness, and happiness).
We did this because participants in our prior experiments may
have anchored their experienced awkwardness ratings on their ear-
lier expectations, potentially exaggerating differences in actual
awkwardness between shallow and deep conversations. Third, we
included three additional experience items after each conversation
in which participants indicated their interest in being friends with
the person they spoke with: specifically, how interested they
would be in speaking with the other person again outside the lab
(0 =not at all;10=very), how much they would have to talk
about with this person outside the lab (0 =nothing at all;10=
quite a bit), and how strong a friend they would become upon get-
ting to know this person outside the lab (0 =weak, like a stranger;
10 =strong, like a close friend).
Finally, after reporting their experiences for both conversations,
participants completed two items about their preferences for shal-
low versus deep conversation. Theyfirst thought about the two
conversations they just engaged in and indicated which conversa-
tion they preferred (
5=strongly preferred the FIRST conversa-
tion;0=equally preferred both conversations;5=strongly
preferred the SECOND conversation). They then imagined partici-
pating in a future experiment in which they would discuss the
same two sets of questions with two different strangers who they
had never met before. Participants reread both sets of questions
and indicated which future conversation they thought they would
prefer (
5=will strongly prefer Conversation A;0=will
equally prefer both conversations;5=will strongly prefer Conver-
sation B).
```
## Results (Experiment 6a)

```
The strength-of-bond and liking items were highly correlated
for both the shallow (aexpectations= .77,aexperiences= .72) and deep
(aexpectations= .80,aexperiences= .68) conversations, and so we col-
lapsed these items to form a connectedness scale. The results
described below were not qualified by significant interactions with
study location (dinner event vs. Virtual Lab) and so we pooled the
data across both locations.
```
(^4) After moving the study to our“Virtual Lab”during the COVID-
pandemic, we changed the third conversation question to,“How often do
you participate in research studies?”

## DEEPER CONVERSATION 19

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


## Manipulation Check

```
The intimacy manipulation was effective: Participants reported
that their deep conversations (M= 6.37,SD= 2.19) were signifi-
cantly more intimate than their shallow conversations (M= 3.08,
SD= 2.64),paired t(119) =
11.84,p,.001, 95% CIdifference
[
3.84,
2.74],d=
1.35.
```
## Care Measures

```
Replicating the prior experiments, participants expected to care
more about their own responses than their partner would,b=1.51,
SE=.14,t(298.19) = 10.49,p,.001, 95% CI [1.23, 1.79]. Further,
this self or other caring gap was significantly larger for deep con-
versations than shallow conversations. A mixed linear model with
fixed-effects terms for Conversation (Shallow, Deep), Target (Own
Care, Partner Care), and their interaction, and random intercepts for
the participant and the partner, yielded a significant Conversation 3
Target interaction,b=.97,SE=.29,t(298.19) = 3.36,p,.001,
95% CI [.40, 1.53]: Although participants expected to care signifi-
cantly more about their responses during shallow conversations
(M=5.05,SD= 2.61) than their partner would (M=4.03,SD=
2.39),b=1.03,SE=.20,t(298.19) = 5.04,p,.001, 95% CI [.63,
1.42], this gap was significantly larger for deep conversations (Ms=
7.23 vs. 5.24, respectively;SDs = 2.04 vs. 2.21),b=1.99,SE=.20,
t(298.19) = 9.79,p,.001, 95% CI [1.59, 2.39].
Perhaps more important, comparisons with actual experience
indicated that participants systematically underestimated how
much they would actually perceive their partner to care about their
responses. A mixed linear model withfixed-effects terms for Con-
versation (Shallow, Deep), Measurement Phase (Expectations,
Experiences), and their interaction, and random intercepts for the
participant and the partner, yielded a significant effect of Measure-
ment Phase,b= 1.77,SE= .16,t(357.00) = 11.25,p,.001, 95%
CI [1.46, 2.07], such that participants underestimated partner care.
The Conversation 3 Measurement Phase interaction effect was
nonsignificant,b=
.38,SE= .31,t(357.00) =
1.22,p= .223,
95% CI [
1.00, .23], indicating that participants underestimated
the recipients’care similarly for the deep and shallow conversa-
tions. Consistent with our hypotheses, participants also underesti-
mated their partners’ care significantly more than they
underestimated their own,b= 1.42,SE= .21,t(773.31) = 6.61,p
,.001, 95% CI [1.00, 1.84]. We then constructed a series of
mixed linear models withfixed-effects terms for Conversation,
```
```
Measurement Phase, and their interaction, and random intercepts
for the participant and the partner, separately for the awkwardness,
connectedness, and happiness measures.
```
## Awkwardness

```
Participants again overestimated the awkwardness of deep
conversations significantly more than shallow ones, as indicated
by a significant Conversation 3 Measurement Phase interaction,
b= .94,SE= .38,t(302.11) = 2.48,p= .014, 95% CI [.20, 1.69].
Although participants overestimated how awkward their shallow
conversations would be,b=
.93,SE= .27,t(302.11) =
3.44,
p,.001, 95% CI [
1.45,
.40], they did so significantly more
for deep conversations,b=
1.87,SE= .27,t(302.11) =
6.94,
p,.001, 95% CI [
2.39,
1.34] (seeFigure 6).
```
## Connectedness

```
As predicted, participants underestimated how connected they
would feel to both partners, as shown by a significant effect of
Measurement Phase,b= 1.02,SE= .18,t(292.99) = 5.58,p,
.001, 95% CI [.66, 1.38]. The Conversation 3 Measurement Phase
interaction was nonsignificant,b=
.16,SE= .26,t(292.99) =

.61,p= .541, 95% CI [
.66, .35]: Participants expected to feel
more connected to their deep conversation partner,b= .98,SE=
.18,t(292.99) = 5.38,p,.001, 95% CI [.63, 1.34], and felt signif-
icantly more connected to this person after speaking,b= 1.14,
SE= .18,t(292.99) = 6.24,p,.001, 95% CI [.78, 1.50].
We also analyzed the item in which participants directly com-
pared the strength of their connection to their two conversation
partners. Participants reported significantly stronger connections
to their deep conversation partner than their shallow conversation
partner (M= 1.62,SD= 3.00),one-sample t(119) = 5.90,p,
.001, 95% CI [1.07, 2.16],d= .54.
```
## Happiness

```
As predicted, participants underestimated how happy they
would feel after both conversations, as indicated by a significant
effect of Measurement Phase,b= 1.17,SE= .17,t(296.00) = 7.03,
p,.001, 95% CI [.84, 1.49]. Unexpectedly, we also found a sig-
nificant Conversation 3 Measurement Phase interaction effect,
b=
.62,SE= .23,t(296.00) =
2.63,p= .009, 95% CI [
1.08,

.16]: Participants expected to feel equally happy after their shal-
low and deep conversations,b=
.07,SE= .17,t(296.00) =
.40,
```
```
Figure 6
Mean Awkwardness, Connectedness, and Happiness Across Conversation Conditions (Shallow
Versus Deep) and Measurement Phase (Expectations Versus Experiences) in Experiment 6a
```
```
Note. Error bars 61 SE.
```
## 20 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
p= .688, 95% CI [
.39, .26], yet felt significantly happier after
their deep conversations,b= .55,SE= .17,t(296.00) = 3.31,p=
.001, 95% CI [.23, .87].
```
## Mediational Analyses

```
We found some evidence that underestimating the other per-
son’s care explained why participants expected their conversations
to unfold less positively than they did. For the shallow conversa-
tions, partner care significantly mediated the effect of Measure-
ment Phase (expectations vs. experiences) on connectedness,b=

.69, 95% CI [
1.17,
.32], and happiness,b=
.37, 95% CI
[
.77,
.05], but not awkwardness,b= .07, 95% CI [
.45, .52].
For the deep conversations, partner care significantly mediated the
effect of Measurement Phase on awkwardness,b= .73, 95% CI
[.05, 1.50], and connectedness,b=
1.02, 95% CI [
1.65,
.52],
but not happiness,b=
.30, 95% CI [
.73, .22].
```
## Desired Conversation Intimacy

```
Replicating the previous experiments, participants reported that
their typical conversations with strangers (M= 3.55,SD= 2.33)
were significantly less intimate than they would ideally prefer
(M= 5.45,SD= 2.52),paired t(119) =
9.03,p,.001, 95%
CIdifference[
2.32,
1.48],d=
.78.
```
## Results (Experiment 6b)

```
The strength-of-bond and liking items were highly correlated in
both the shallow (a= .77) and deep (a= .78) conditions, as were
the three friendship interest items (ashallow= .92,adeep= .91). We
collapsed these items to form connectedness and friendship inter-
est scales, respectively.
```
## Manipulation Check

```
The intimacy manipulation was effective: Participants reported
that their deep conversations (M= 6.78,SD= 2.06) were signifi-
cantly more intimate than their shallow conversations (M= 2.19,
SD= 2.17),paired t(119) =
16.22,p,.001, 95% CIdifference
[
5.15,
4.03],d=
2.17.
```
## Preference Measures

```
Before engaging in the conversations, participants expected to
prefer the shallow conversation relative to the deep conversation
(M=
1.56,SD= 3.25),one-sample t(119) =
5.26,p,.001,
95% CI [
2.15,
.97],d=
.48. After the conversations, how-
ever, participants reported preferring the deep conversation (M=
.88,SD= 3.26),one-sample t(119) = 2.94,p= .004, 95% CI [.29,
1.46],d= .27. As a result, participants significantly underesti-
mated the extent to which they would prefer the deep conversa-
tion,paired t(119) =
6.77,p,.001, 95% CIdifference[
3.14,

1.72],d=
.75.
After engaging in both conversations, participants also reported
which questions they expected to prefer discussing in a future
study session. Participants did not show a statistically significant
preference for one conversation type over another (M= .43,SD=
3.40),one-sample t(119) = 1.39,p= .166, 95% CI [
.18, 1.05],
d= .13. This means that participants’reported preferences shifted
significantly after having their conversations: Whereas participants
expected to prefer the shallow conversation at the start of the study
```
```
session, their predicted preferences shifted significantly in the
direction of the deep conversation questions after actually engag-
ing in both conversations,paired t(119) =
6.16,p,.001, 95%
CIdifference[
2.63,
1.35],d=
.60. This result suggests partici-
pants learned something during their conversations that they then
used to update their expectations. Indeed, a post hoc analysis indi-
cates that participants’reported preferences before their conversa-
tions were only modestly correlated with their experienced
preferences after their conversations,r= .27,t(118) = 3.03,p=
.003, 95% CI [.09, .43]. In contrast, their experienced preferences
were very strongly correlated with their preferences for a future
conversation,r= .73,t(118) = 11.77,p,.001, 95% CI [.64, .81].
These two correlations differ significantly from each other,Z=

6.13,p,.001. Participants learned that the relatively deep con-
versation was more positive than the relatively shallow conversa-
tion, and used their direct experience to update their expectations
for the future. Misunderstanding the outcomes of relatively deep
conversation could create a barrier to engaging in them more often
in everyday life, thereby keeping people from having the very
experiences that would allow them to learn that their expectations
could be miscalibrated (see alsoEpley & Schroeder, 2014, Experi-
ment 4).
```
## Experience Measures

```
Participants preferred the deep conversation because their expe-
riences were more positive overall during the deep conversation
than the shallow conversation. We constructed mixed linear
models withfixed-effects terms for Conversation (Shallow, Deep)
and random-intercept terms for the participant and the partner,
separately for each measure. Participants cared significantly more
about their own responses during the deep conversation than the
shallow conversation,b=
1.84,SE= .29,t(235.00) =
6.43,
p,.001, 95% CI [
2.40,
1.28], and perceived the other person
to care more about one’s own responses during the deep conversa-
tion,b=
1.37,SE= .28,t(238.00) =
4.90,p,.001, 95% CI
[
1.91,
.82]. Participants reported that their deep conversations
felt only marginally more awkward than their shallow conversa-
tions,b=
.63,SE= .35,t(232.81) =
1.82,p= .070, 95% CI
[
1.31, .05], but that they felt significantly more connected to
their partner,b=
1.10,SE= .25,t(238.00) =
4.33,p,.001,
95% CI [
1.59,
.60], and more interested in being friends with
the person,b=
.94,SE= .27,t(238.00) =
3.42,p,.001, 95%
CI [
1.48,
.40],after the deep conversation. Happiness did not
differ significantly between the deep and shallow conversations,
b=
.28,SE= .21,t(238.00) =
1.34,p= .182, 95% CI [
.70,
.13] (seeFigure 7).
```
## Desired Conversation Intimacy

```
Replicating the previous experiments, participants reported that
their typical conversations with strangers (M= 2.79,SD= 2.26)
were significantly less intimate than they would ideally prefer
(M= 5.54,SD= 2.65),paired t(119) =
11.11,p,.001, 95%
CIdifference[
3.24,
2.26],d=
1.11.
```
## Discussion

```
Consistent with Experiments 1a–5, participants again generally
underestimated how positive their conversations would be, and also
```
## DEEPER CONVERSATION 21

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
overestimated how awkward deep conversations would be relative
to shallow conversations. In contrast to some of the preceding
experiments, participants also felt significantly more connected to
their deep conversation partner than their shallow conversation
partner, suggesting that people establish stronger connections
through deep conversation—yet may be better able to report the rel-
ative strength of this connection when they compare two conversa-
tions directly.
More important, Experiment 6b provided direct measures of par-
ticipants’preferences for relatively deep versus shallow conversa-
tions. Participants expected to prefer the shallow conversation more
than the deep conversation before having the conversations, but
consistent with our hypotheses, they reported actually preferring
the deep conversation after having both of them. Not only does hav-
ing a deep conversation with another person seem to be a surpris-
ingly positive experience, it seems to bemorepositive than having
a shallow conversation. As a result of their direct experience, partic-
ipants also appeared to learn that they would prefer a relatively
deep conversation in the future more than their initial expectations
suggested. This result highlights how miscalibrated expectations
could create a barrier to engaging in deeper and more meaningful
conversations, thereby encouraging shallower conversations than
might be optimal for one’s own relationships.
Experiments 7a and 7b test thefinal step in our theorizing. Spe-
cifically, we predict that directly manipulating the psychological
barrier created by miscalibrated expectations would encourage
people to have deeper conversations than they might otherwise.
Because the preceding experiments suggest that people undervalue
deep conversations at least partly because they underestimate how
much others will care, and be interested in, the conversation’s con-
tent, we manipulated how much participants would expect their
conversation partner to care and then measured the intimacy of
questions they were interested in discussing with their partner. We
did so in Experiment 7a by manipulating whether people imagined
talking with someone they knew to be very caring or a person they
knew to be indifferent, and in Experiment 7b by either informing
or not informing participants that their beliefs about others’care
tend to be systematically miscalibrated. We expected participants
to select deeper conversational topics to discuss when they
```
```
imagined talking with a relatively caring person in Experiment 7a,
and when they learned that people tend to underestimate how
much others care in conversation (vs. overestimate, or no informa-
tion). If underestimating the extent to which others care about the
content of conversation creates a barrier to deeper conversations,
then removing it either by imagining conversation with a caring
other or by calibrating participants’expectations should encourage
deeper conversation. Note that this would occur only if partici-
pants were personally interested in having deeper conversations
more often in daily life, as participants in Experiments 1a, 1b, 1c,
4a, 4b, 6a, and 6b reported.
```
## Experiments 7a–7b: Anticipated Indifference as a

## Barrier to Deep Talk

## Pretest

```
Before conducting the experiments, we pretested 20 discussion
questions for perceived intimacy, some of which were used in
Experiments 1a
1c, 2, 4a-4b, and 6a-b (seeonline supplemental
materialsfor full details). The four perceived intimacy items were
highly correlated (a= .96) and so we combined them to form an
intimacy scale. We designated the 10 questions with the highest
average intimacy ratings as deep questions and the 10 questions
with the lowest average intimacy ratings as shallow questions (see
Appendix).
```
## Method

## Participants

```
We recruited 100 participants from MTurk (N= 93 after exclu-
sions;Mage= 35.30;SDage= 9.99; 36.56% female; 78.49% White)
to complete Experiment 7a in exchange for $1.75. We excluded
seven participants from analyses because they failed the attention
check described below.
We recruited 160 participants from MTurk (N= 130 after exclu-
sions;Mage = 35.92;SDage= 12.36; 46.15% female; 75.38%
White) to complete Experiment 7b in exchange for $1.20. We
```
```
Figure 7
Mean Awkwardness, Connectedness, and Happiness Experiences Across Shallow and Deep
Conversations in Experiment 6b
```
```
Note. Error bars 61 SE.
```
## 22 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
excluded 30 of these participants because they failed one or more
attention of the checks described below.
```
## Procedure

```
Participants in Experiment 7a imagined visiting a“social inter-
action”research lab where they would speak with another study
participant they had never met before in a discussion called the
“sharing game.”Participants were then randomly assigned to the
caring or uncaring condition. Participants in the“caring”condition
read,“Before the session begins, suppose you see the other person
in the waiting room. You've seen this person around and you have
the impression that this person is very sociable, caring, and consid-
erate of others. Although you don't know this person, you feel
pretty confident in your judgment.”Participants in the“uncaring”
condition read,“Before the session begins, suppose you see the
other person in the waiting room. You've seen this person around
and you have the impression that this person is rather indifferent
toward others and isn't very caring or considerate. Although you
don't know this person, you feel pretty confident in your
judgment.”
Participants in both the caring and uncaring conditions then
read the 20 possible discussion questions and selectedfive they
would most like to ask and answer while speaking with the other
person. The questions were presented in random order and not la-
beled in any way. After selectingfive questions, participants then
viewed each of the 20 questions separately and for each question
reported how much they wanted to discuss the question with the
other person (0 =not at all;10=quite a bit), how much they
would care about their own response to the question (0 =not at
all;10=quite a bit), and how much they believed the other person
would care about the participant’s response to the question (0 =
not at all;10=quite a bit). After evaluating the questions, partici-
pants completed an attention check by reporting whether the other
participant was described as sociable, caring, and considerate or as
indifferent and not very caring or considerate. Finally, participants
reported demographic information and received payment.
The procedure for Experiment 7b was identical to Experiment
7a with three exceptions. First, we designed manipulations that ei-
ther did or did not inform participants that people’s expectations
tend to be systematically miscalibrated. Participants in theunder-
estimationcondition read,“In these experiments, wefind that peo-
ple tend to UNDERESTIMATE how much strangers will care
about each other’s responses to these questions. That is, strangers
tend to be MORE concerned and interested in each other’s
responses than people expect.”Participants in theoverestimation
condition read,“In these experiments, wefind that people tend to
OVERESTIMATE how much strangers will care about each
other’s responses to these questions. That is, strangers tend to be
LESS concerned and interested in each other’s responses than peo-
ple expect.”Participants in thecontrolcondition were not told
anything about the extent to which others’expectations tend to be
calibrated. Second, after selectingfive out of the 20 questions, par-
ticipants in this experiment reported how interested they would be
in asking and answering each of the 20 questions with the other
person (0 =not at all;10=extremely), but did not predict the
degree of care for either themselves or their conversation partner.
Third, we tailored the attention checks to the current procedure by
asking participants to report whether they imagined speaking with
```
```
a friend or stranger, and to report what they had been told about
the results of our previous research experiments.
```
## Results

```
We preregistered analyses of the number of deep and shallow
questions participants selected across conditions in both Experi-
ments 7a and 7b, but also report analyses of the average intimacy
level of the questions selected.
As predicted, participants in the caring condition of Experiment
7a (M=3.09,SD= 1.44) selected significantly more deep questions
than did participants in the uncaring condition (M= 2.02,SD=
1.73),t(91) = 3.23,p= .002, 95% CIdifference[.41, 1.72],d=.67.
Participants in the caring condition also selected questions higher in
average intimacy (M=4.96,SD= 1.22) than did participants in the
uncaring condition, (M=4.24,SD=1.48),t(91) = 2.53,p=.013,
95% CIdifference[.15, 1.27],d= .53. These differences in choice
could arise either because participants in the caring condition have
a stronger desire to discuss deep questions or because they have a
weaker desire to discuss shallow ones, compared with participants
in the uncaring condition. We computed each participant’smean
desire to discuss the 10 shallow questions and the 10 deep ques-
tions, separately. Consistent with our predictions, participants in the
caring condition were particularly interested in discussing deep
questions. A 2 (Partner: Caring, Uncaring) 3 2 (Question Type:
Shallow, Deep) ANOVA with repeated measures on the second fac-
tor yielded a significant Partner 3 Question Type interaction,
F(1, 91) = 9.52,p=.003,g^2 p= .09 (seeFigure 8): Participants in
the caring and uncaring conditions did not differ in their desire to
discuss shallow questions,t(91) =
.70,p=.486,95%CIdifference
[
1.07, .51],d=
.14, but those in the caring condition were sig-
nificantly more interested in discussing deep questions,t(91) =
3.30,p= .001, 95% CIdifference[.54, 2.16],d= .68. Underestimating
others’care may create a barrier to having deeper and more inti-
mate conversations.
In Experiment 7b, a one-way ANOVA indicated that the num-
ber of deep questions selected varied significantly across care con-
ditions,F(2, 127) = 3.71,p= .027,g^2 p = .06. As predicted,
participants in the underestimation condition (M= 2.41,SD=
1.12) selected significantly more deep questions than participants
in the overestimation condition (M= 1.69,SD= 1.26),t(127) =

2.68,p= .008, 95% CIdifference[
1.26,
.19],d=
.57. The
number of deep questions selected in the control condition fell in
between (M= 2.15,SD= 1.44), differing neither from the underes-
timation condition,t(127) =
.95,p= .344, 95% CIdifference[
.83,
.29],d=
.21, nor the overestimation condition,t(127) =
1.69,p
= .094, 95% CIdifference[
1.00, .08],d=
.36. Likewise, the mean
intimacy of selected questions varied by condition,F(2, 127) =
3.40,p= .036,g^2 p= .05, with participants in the underestimation
condition (M= 4.53,SD= 1.02) selecting questions of higher av-
erage intimacy than participants in the overestimation condition
(M= 3.92,SD= 1.10),t(127) =
2.60,p= .010, 95% CIdifference
[
1.06,
.14],d=
.55. The mean intimacy of selected questions
in the control condition fell in between (M= 4.16,SD= 1.14), and
did not differ significantly from either the underestimation condi-
tion,t(127) =
1.51,p= .133, 95% CIdifference[
.84, .11],d=

.33, or the overestimation condition,t(127) = 1.03,p= .304,
95% CIdifference[
.22, .70],d= .22.
```
## DEEPER CONVERSATION 23

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
As with Experiment 7a, we tested whether these differences
arose from differences in interest in discussing deep questions, shal-
low questions, or possibly both. We did so by comparing partici-
pants’reported desire to discuss the 10 deep and 10 shallow
questions, on average, between the underestimation and overestima-
tion conditions. A 2 (Care: Underestimation, Overestimation) 32
(Question Type: Shallow, Deep) ANOVA with repeated measures
on the second factor and mean interest as the dependent measure
yielded a significant main effect of Care,F(1, 87) = 5.89,p=.017,
g^2 p= .06, indicating that participants in the underestimation condi-
tion were more interested in discussing the questions overall, and a
significant main effect of Question Type,F(1, 87) = 4.62,p=.034,
g^2 p= .05, indicating that participants in both conditions were more
interested in discussing deep versus shallow questions. The Care 3
Question Type interaction effect was nonsignificant,F(1, 87) = .69,
p=.410,g^2 p= .01. However, planned contrasts provide some evi-
dence that participants in the underestimation condition may have
been somewhat more interested in discussing deep questions: Par-
ticipants in the underestimation condition reported only marginally
greater interest in discussing shallow questions than participants in
```
```
the overestimation condition,t(127) =
1.68,p=.095,95%CIdiffer-
ence[
1.65, .13],d=
.36, but reported significantly greater inter-
est in discussing deep questions,t(127) =
2.68,p= .008, 95%
CIdifference[
2.04,
.31],d=
.57 (seeFigure 9).
```
## Discussion

```
Underestimating the extent to which others care about the con-
tent of deep conversation can create miscalibrated expectations
about the outcomes of these conversations, and also create a bar-
rier to engaging in these conversations in thefirst place. Experi-
ments 7a and 7b suggest that manipulating this barrier directly
may influence people’s preferences for deeper conversations with
strangers. Participants in Experiment 7a chose deeper questions
when they expected their conversation partner to be caring rather
than indifferent. Participants in Experiment 7b likewise chose
deeper questions when they were told that people tend to underes-
timate strangers’care than when they were instead told that people
tend to overestimate strangers’care. It is important to note that
although the control condition in Experiment 7b was different
from neither the overestimation nor underestimation condition, it
```
```
Figure 9
Mean Interest Across Care (Overestimation Versus Underestimation Versus
Control) and Question Type (Shallow Versus Deep) in Experiment 7b
```
```
Note. Error bars 61 SE.
```
```
Figure 8
Mean Desire to Discuss Across Partner (Uncaring Versus Caring) and Question
Type (Shallow Versus Deep) in Experiment 7a
```
```
Note. Error bars 61 SE.
```
## 24 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
was at least descriptively more similar to the underestimation con-
dition. We are not sure how to interpret this result because we did
not obtain a manipulation check to test how much this particular
experimental manipulation actually affected participants’expecta-
tions. It could be that it is simply easier to convince people that
others care less than expected about a deep conversation than to
convince them that others care more. These particular results con-
firm that manipulating perceptions of others’care can affect pref-
erences for deep talk, but they cannot at this point confirm whether
weakening or strengthening that barrier is likely to have a larger
impact on people’s social preferences.
```
## General Discussion

```
Human beings are a deeply social species, whose wellbeing and
health are tightly tied to the quality of their social relationships
(Diener & Seligman, 2002; Holt-Lunstad, 2018; Sun et al., 2020).
It is not surprising that most people have strong desires to
strengthen and maintain meaningful relationships (Baumeister &
Leary, 1995; Ryan & Deci, 2000), actively combatting feelings of
loneliness and isolation by seeking social connection with others
(Maner et al., 2007). Disclosing meaningful and intimate informa-
tion with others in conversation is one way to strengthen social
relationships (Aron et al., 1997; Collins & Miller, 1994), and so it
is also not surprising that participants in our experiments reported
wanting to have more deep and meaningful conversations with
others more often in their daily lives. If people want deeper con-
versations with others, then why aren’t they having them?
Our data suggest that underestimating others’deeply social na-
ture—assuming that others will be more indifferent and uncaring
in conversation than they actually are—could help to explain why
conversations in daily life are shallower than people might prefer.
Our participants consistently expected their conversations to be
more awkward, and to lead to weaker connections and less happi-
ness, than they actually did regardless of whether participants gen-
erated the topics of conversation (Experiments 3 and 5) or not
(Experiments 1a, 1b, 1c, 2, 4a, 4b, 6a, and 6b). Participants also
consistently overestimated how awkward deep conversations
would be more so than shallow conversations (Experiments 2–6a),
but were more calibrated for conversations with close others
whose care and interest were more clearly known (Experiment 5).
Finally, participants chose deeper questions when they expected
their conversation partner to be relatively more caring (Experi-
ments 7a and 7b), providing the clearest evidence that miscali-
brated inferences about others’ sociality may act as a
psychological barrier to deeper conversations.
```
## Limitations

```
Although participants’tendency to underestimate the positive
outcomes of deep conversation was robust across important meth-
odological variations, there are at least three potential limitations
on the generalizability of our results.
Ourfirst limitation on generalizability is that all of our experi-
ments manipulated intimacy in experimental settings in which
both people in the conversation knew they were going to have a
conversation as part of an experiment. This is obviously not the
case in naturalistic conversations in daily life in which people are
likely to be more uncertain about whether another person is likely
```
```
to reciprocate deeper and more intimate disclosures, and also must
infer why another person is asking relatively deeper or shallower
conversation topics. Asking another person to share the last time
they cried in front of someone else might be received much differ-
ently when you raise the question to the person standing next to
you in the supermarket line than it was in our experiments. Of
course, any differences between our experimental procedures and
more naturalistic circumstances are also likely to affect people’s
expectations of the interaction in a way that could still maintain
the critical gap we observed between the expected and actual out-
comes of more naturalistic conversations. The deeper question one
might consider asking in the supermarket line might still yield a
better conversation than expected.
In fact, we think people might be even more likely to underesti-
mate the positive outcomes of deep conversations in naturalistic
settings outside the lab. Previous research in both the United States
and in the United Kingdom indicates that people may underesti-
mate strangers’willingness to engage in conversation in naturalis-
ticfield experiments on trains, busses, and cabs, which can partly
then explain people’s tendency to underestimate how much they
would enjoy having a conversation with a stranger (Epley &
Schroeder, 2014; Schroeder et al., in press; see alsoSandstrom &
Boothby, 2021). Our experimental procedures eliminate this con-
cern by instructing participants to speak with another study partici-
pant who they know will engage them in conversation. Our
experiments may provide a more conservative test of whether peo-
ple underestimate the value of deep conversations compared with
more naturally occurring conversations. In addition, starting a
meaningful conversation signals one’s trust and interest in the lis-
tener. This signal may be more influential when the listener cannot
attribute the depth of conversation to an experimental instruction.
In daily life, listeners may be even more concerned about intimate
details that another person freely shares with them. Nevertheless,
it could also be that listeners form more negative inferences about
another person’s interest or intent when they raise deeper conver-
sation topics in daily life, such as inferring that another person is
interested in a romantic relationship or is potentially trying to
manipulate or take advantage of a listener in some way. It is possi-
ble these inferences are not anticipated by those trying to start a
conversation, in which case the gap between expected and actual
outcomes of deeper conversations might vary across contexts.
Future research could examine the generalizability of our results
in more naturalistic contexts by manipulating whether both partici-
pants are aware of the conversation instructions or not. To the
extent that both participants reveal meaningful information about
themselves during the conversation, our theory predicts that deep
conversations should lead to similar, if not even stronger, connec-
tions when one person is unaware of the conversation instructions.
The generalizability of our effects could also be tested using diary
methods in which people are asked to engage in relatively deeper
or shallower conversations over the course of a day and then report
their experiences at the end of the day, or simply by asking people
to engage in a deeper conversation when they have a natural op-
portunity to do so in daily life. Existing research using this meth-
odology has found that people report more positive affect when
asked to act extraverted than when asked to act introverted (Marg-
olis & Lyubomirsky, 2020), when asked to engage in conversation
on a train or bus compared with remaining silent (Epley &
Schroeder, 2014; Schroeder et al., in press), or simply when asked
```
## DEEPER CONVERSATION 25

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
to try to make a connection to a coffee shop barista compared with
being as efficient as possible (Sandstrom & Dunn, 2014a; see also
Gunaydin et al., 2021). These existing results, showing positive
affect resulting from simply engaging with others in naturalistic
settings, give us reason to believe that engaging more deeply in
these same settings would enhance positive affect as well.
A second potential limitation on generalizability is that our
experiments investigated only short-term outcomes of a single rel-
atively shallow and/or deep conversation. There could be longer-
term consequences, either more positive or negative, of these con-
versations that we did not measure. For instance, participants in
Experiment 2 felt somewhat lonelier after having a relatively deep
conversation than after a shallower conversation. Although we
only measured loneliness in one experiment, and hence do not
know how robust this result is, it could be that a single conversa-
tion highlights dissatisfying features of a person’s other relation-
ships, such as other relationships being shallower than one might
desire, that affect a person’s wellbeing in a more complicated fash-
ion beyond the single interaction we examined. It could also be
that people feel worse over time after a very satisfying conversa-
tion with a stranger after recognizing that they are not likely to
have another conversation with that person again. Again,field
experiments involving diary methods would be an appropriate
way of examining longer-term outcomes of having relatively more
or less deep conversations in everyday life.
Afinal potential limitation on the generalizability of our results
comes from conducting all of our experiments within in a rela-
tively narrow time period within the United States. Although our
samples are relatively diverse, including university students, mas-
ter’s degree students from around the world, and corporate
employees and executives in the United States, it is reasonable to
wonder how our effects may—or may not—vary across cultures.
Cultures may vary in their openness to strangers, with individual-
istic cultures generally being more accepting and open to strangers
and collectivistic cultures prioritizing ingroup relationships. If cul-
tural differences exist in the willingness to engage in deep and
meaningful conversations with strangers, then they could arise ei-
ther from cultural differences in expectations or from cultural dif-
ferences in actual experiences. Given the fundamental importance
of social relationships across humans’evolutionary history (Dun-
bar, 1998; von Hippel, 2018), we would predict that expectations
about others’care and interest are more likely to vary across cul-
tures than the actual experience of deep and meaningful conversa-
tion. Cultures could vary in the relative frequency of deep versus
shallow conversations because they vary in their social expecta-
tions rather than in their social experiences. Note that the social
expectations we have identified are the very kind that would create
social avoidance and hence keep people from learning about the
actual consequences of relatively shallow versus deep conversa-
tions. In Experiment 6b, participants reported a stronger prefer-
ence for deep conversation after having both a shallow and a deep
conversation, preferences that our data suggest would be more
closely aligned with optimizing people’s subjective experience
and the quality of their relationships. Similarly, individuals within
a given culture might vary in their social expectations, such that
those who are highly socially avoidant (Strachman & Gable,
2006 ), high in rejection sensitivity (Pietrzak et al., 2005), or inse-
curely attached (Keelan et al., 1998) might also be especially pes-
simistic about deeper conversations, and hence avoid them more
```
```
often in daily life. Miscalibrated expectations may come partly
from avoiding the experiences that would otherwise calibrate a
person’s expectations.
```
## Unanswered Questions

```
We believe our current results make important theoretical con-
tributions to the literatures on both self-disclosure and wellbeing.
Decisions to open up to another person or remain more guarded
are based in part on how people expect that their interaction part-
ner will respond (Afifi& Steuber, 2009; Kardas et al., 2021;
Omarzu, 2000; Ruan et al., 2020), yet psychologists have primar-
ily examined the causes (Altman & Taylor, 1973; Berg & Clark,
1986 ; Cline, 1989) and consequences (Collins & Miller, 1994;
Kelly & McKillop, 1996) of self-disclosure separately from one
another. Our article highlights the importance of combining these
two streams of research. Understanding how miscalibrated infer-
ences about others’social interest and care guide decisions about
the content of conversation can help to explain why people may be
missing out on opportunities to establish stronger connections with
others.
We believe our research also highlights the importance of meas-
uring the accuracy of the causes underlying self-disclosure, in this
case the calibration between expectations of others’reactions in
conversation and their actual reactions. Doing so opens up new
questions for theoretical inquiry, especially whether people dis-
closeenoughdeep and intimate content in their interactions to
maximize the quality of relationships. Experiments 1a–5 do not
answer this question definitively. Although participants tended to
underestimate how positive deep conversation would be overall,
the deep conversations were indeed somewhat more awkward than
the shallow conversations in actual experience. Instead of guessing
about how people themselves might integrate these unique aspects
of their experience into an overall preference for conversations,
we enabled participants to compare these two conversations
against each other directly in Experiments 6a and 6b by instructing
them to have both a deep and a shallow conversation. These
experiments provided the clearest evidence that people uniquely
underestimate the positive outcomes of deep conversation, that
deep conversations indeed create more positive experiences than
shallow conversations overall, and that people report preferring
the deeper conversation afterward to a greater extent than they
expected beforehand. These results suggest that miscalibrated
expectations may leave people being overly reluctant to discuss
deep and intimate content in their everyday conversations with
strangers.
Although participants significantly preferred relatively deeper
conversations to shallower conversations in Experiments 6a and
6b, it is worth noting that participants across our experiments also
tended to report positive experiences after relatively shallow con-
versations, and tended to report more positive (and slightly less
negative) experiences in these conversations than they anticipated
as well. These results are consistent with prior research suggesting
a more general phenomenon in which conversations with strangers
tend to be surprisingly pleasant (Epley & Schroeder, 2014; Sand-
strom & Boothby, 2021). Indeed, we observed somewhat inconsis-
tent differences in experienced connectedness, happiness, and
enjoyment across the experiments in which participants experi-
enced only a relatively deep or shallow conversation. For instance,
```
## 26 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
participants in the deep condition felt significantly more connected
to their conversation partner than did participants in the shallow or
control condition in Experiments 3 (p= .022,d= .47) and 4a (p=
.016,d= .48), but did not differ significantly in Experiments 2
(p= .135,d= .32) and 4b (p= .970,d=
.01). Participants in the
deep condition also felt marginally happier about their conversa-
tion than did participants in the shallow or control condition in
Experiment 3 (p= .088,d= .34), but did not report feeling signifi-
cantly happier in Experiments 4a (p= .962,d=
.01), 4b (p=
.309,d=
.20), or in Experiment 5 among participants who spoke
with either distant strangers (p= .190,d= .26) or close others (p=
.160,d= .28). These somewhat inconsistent results are surprising
because previous literature suggests that intimate self-disclosure
should lead to more positive conversations than impersonal disclo-
sures (Aron et al., 1997; Collins & Miller, 1994), and our manipu-
lations were designed precisely to encourage more intimate
disclosures in the deep conditions.
Inconsistency in the connection value of deep conversation may
come partly from the unique nature of our experimental exchanges
compared with the existing literature. Many prior experiments
have more carefully constrained the breadth and depth of self-dis-
closure through impression-formation tasks in which participants
read about or observe others’interactions (e.g.,Berg & Archer,
1980 ; Bradac et al., 1978; Chaikin & Derlega, 1974) or by using
written essays and self-descriptions (Brewere & Mittelman, 1980;
Rubin, 1975). Even procedures involving live, spoken interactions
have often required that participants deliver uninterrupted verbal
self-descriptions (Chaikin et al., 1975; Jones & Archer, 1976)or
engage in structured question-answer sequences (Sprecher et al.,
2013 ). Each of these methodologies carefully controls the amount
of self-disclosure and thereby tests the impact of self-disclosure on
liking and connectedness. However, doing so also reduces ecologi-
cal validity by constraining the naturalflow of back-and-forth con-
versation. In contrast, participants in Experiments 1a–5 engaged in
back-and-forth conversations and were not explicitly forbidden
from discussing content beyond the scope of the discussion ques-
tions. Participants in Experiments 6a and 6b were encouraged not
to stray from the questions we asked them to discuss, but their
conversations were not monitored or constrained beyond that
instruction. Overall, our experiments may better approximate con-
versation in everyday life, but they also sacrifice the tight experi-
mental control that would come from more scripted or one-sided
exchanges.
Ratherthan call into question existing literature on self-disclo-
sure, we believe that ourfindings may suggest an important mod-
erator to explain when, and why, intimacy manipulations influence
positive outcomes such as liking and connectedness (Collins &
Miller, 1994). Intimacy manipulations may have greater impact in
static exchanges such as impression-formation tasks, written
exchanges, and structured question-answer sequences than in
dynamic exchanges involving back-and-forth conversation.
Indeed, the small number of prior experiments that have manipu-
lated intimacy in less structured interactions have produced mixed
findings. Some experiments have produced differences in experi-
enced connectedness (Aron et al., 1997; Shearer, 2017) while
others have produced mixed evidence across multiple experiments
(Kashdan et al., 2011) or no differences (Besst, 2016; Kashdan &
Roberts, 2006). Because conversation naturally becomes more
intimate as people continue talking (Taylor, 1968), even relatively
```
```
superficial questions may promote liking and connectedness by
acting as a gateway to somewhat deeper conversation, thereby pro-
moting relatively positive outcomes regardless of the initial depth
of the conversation.
As already mentioned, anecdotal reports from the live debrief-
ing of Experiment 4b support this possibility, with participants
describing how surprisingly deep and meaningful their ostensibly
“shallow”conversations became as they unfolded. Some results
from Experiment 5 are at least consistent with this possibility of
increasing intimacy even in what we intended to be relatively shal-
low conversations. In the control conditions, participants who
spoke with distant strangers perceived their conversations (M=
4.21,SD= 2.22) to be significantly more intimate than the ques-
tions they discussed (M= 2.50,SD= 2.40),paired t(49) =
4.98,
p,.001, 95% CIdifference[
2.40,
1.02],d=
.70, and those
who spoke with close others likewise perceived their conversa-
tions (M= 5.03,SD= 2.46) to be significantly more intimate than
the questions they discussed (M= 3.70,SD= 2.65),paired t(49) =

3.69,p,.001, 95% CIdifference[
2.05,
.61],d=
.52. In con-
trast, these patterns weakened or reversed in the deep conditions,
such that participants who spoke with distant strangers perceived
their conversations (M= 6.41,SD= 1.67) to belessintimate than
the questions they discussed (M= 7.04,SD= 1.58),paired t(49) =
2.09,p= .042, 95% CIdifference[.02, 1.24],d= .30, and those who
spoke with close others likewise perceived their conversations
(M= 7.49,SD= 1.86) to be no more intimate than the questions
they discussed (M = 7.78,SD= 1.43),paired t(49) = 1.22,
p= .230, 95% CIdifference[
.19, .77],d= .17.
These results raise two interesting topics for future research.
First, relatively shallow questions could act as a gateway to deeper
conversation, potentially helping to explain why our deep conver-
sation conditions did not consistently create significantly stronger
experiences of connectedness, happiness, and enjoyment in our
experiments than our comparison conditions. Second, our experi-
ments attempted to hold the depth of conversation topics constant
across conditions, but self-disclosure may naturally change over
the course of naturalistic conversations. People may expect that
conversations are better when they become increasingly intimate
over the course of the conversation, and hence start conversations
more shallowly than they hope to end them (Aron et al., 1997). It
would be interesting and important to understand how people’s
expectations about changes in intimacy across conversations are
aligned with their actual experiences.
If unscripted conversations do indeed tend to change over the
course of the conversation in ways that people do not fully antici-
pate, then this could also help explain one other unexpected result:
In Experiments 4a and 4b, participants in the shallow condition
expected to feel significantly less connected to each other than par-
ticipants in the deep condition, yet both groups actually felt
strongly connected after speaking. As a result, participants in the
shallow condition underestimated how connected they would feel
morethan did participants in the deep condition. Natural conversa-
tion may transform relatively shallow topics into relatively deeper
content as the conversation continues. If people overlook this
property of conversation and instead focus myopically on the
topics that they plan to discuss—in a similar way that people
expect their emotional reactions to major life events to be deter-
mined largely by the event in question and not by other future
events or by their ability to make sense of the event (Gilbert et al.,
```
## DEEPER CONVERSATION 27

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
1998 ; Wilson et al., 2000)—then they could underestimate posi-
tive outcomes more for shallow questions than for deep questions
due to exaggerated differences in their expectations. This hypothe-
sis predicts that people should underestimate the positive out-
comes of conversation even more fortrivialtopics that initially
require virtually no self-disclosure. In contrast, people should have
more accurate expectations of more highly structured question-an-
swer sequences that constrain the natural dynamics of back-and-
forth conversation and keep conversations from becoming increas-
ingly intimate. We believe this increasing intimacy hypothesis
about unconstrained conversation warrants further investigation
because it may help to explain a range offindings indicating that
people underestimate how much they will enjoy connecting with
distant strangers (Dunn et al., 2007; Epley & Schroeder, 2014;
Mallett et al., 2008).
Ourfindings also raise questions about other novel mechanisms
guiding social behavior that may lead people to mismanage their
relationships. One question emerges from the different patterns of
results obtained on our different outcome measures: In most of our
experiments, participants underestimated their connectedness and
happiness, and their partner’s care, equally for deep and shallow
conversations yet overestimated their awkwardness more for deep
conversations. Moreover, we found stronger evidence that under-
estimating others’ care explained participants’ miscalibrated
expectations of connectedness and happiness than awkwardness.
One explanation for these divergent findings concerns the
expected ambivalence of deep talk: Participants generally
expected deep conversations to create more positive experiences
of connectedness and happinessandmore negative experiences of
awkwardness than superficial conversations. Prior research sug-
gests that as an event approaches in time, people may focus more
narrowly on its perceived risks relative to its perceived reward
(Baumeister et al., 2001; Brown, 1948; Epstein & Fenz, 1962;
Miller, 1944). This heightened sensitivity to immediate risks could
cause people to expect immediate deep conversations to be partic-
ularly awkward, creating a reluctance to discuss deep topics in the
midst of an ongoing conversation—even if they recognize in the
abstract that they would prefer to have deeper conversations more
often in their daily lives. Such a mechanism could lead to excess
avoidance in a wide range of social situations because almost any
decision to engage with others can lead to either positive outcomes
of social connection and happiness or negative outcomes of awk-
wardness stemming from others’indifference or outright hostility.
Future research could test this explanation by manipulating
whether a deep conversation will take place immediately or in the
distant future and measuring whether participants expect immedi-
ate deep conversations to be more negative than distant ones.
Another possible explanation is that multiple classes of mecha-
nisms contributed to the patterns of miscalibrated expectations
observed in our experiments: Apart from expectingothersto be
less caring than they were, participants may have failed to appreci-
ate how comfortable and positivetheywould personally feel about
disclosing self-relevant information to another person. Self-disclo-
sure may feel less awkward than people expect, independent of
feedback received from the listener. If so, thisfinding would
bridge the literatures on social judgment and affective forecasting,
which have been studied separately but could operate jointly to
create unnecessary barriers to self-disclosure (see alsoLeunissen
et al., 2014).
```
## Broader Implications

```
We document that people undervalue the positive consequences
of conversations, especially deep conversations, at least partly
because they underestimate how much strangers in conversation
will be interested in the content of the conversation, and care about
the intimate information being shared. We believe this is part of
broader tendency to underestimate others’sociality, thereby creat-
ing a wide variety of psychological barriers to social engagement.
People may underestimate how interested strangers on trains,
buses, and cabs are in talking, thereby creating reluctance to start a
conversation and increase one’s wellbeing (Epley & Schroeder,
2014 ). Similarly, people underestimate how positive others will
feel following an expression of gratitude (Kumar & Epley, 2018),
a random act of kindness (Kumar & Epley, 2021), a compliment
(Boothby & Bohns, 2021; Zhao & Epley, in press), a trusting dis-
closure of a negative secret (Kardas et al., 2021), and even a con-
structive confrontation in an established relationship (Dungan &
Epley, 2021). People expect others to be more indifferent to these
prosocial acts than others actually are, just as we found for deep
conversation. There are several possible explanations for these
effects. One is that people tend to dehumanize others by assuming
that others have weaker mental capacities, including capacities for
empathy and compassion, than they do themselves (Haslam et al.,
2005 ; Waytz et al., 2014). Another is that egocentric biases in
judgment may lead people to undervalue the importance of
expressing warmth in others’evaluations. Actors in a situation
tend to focus relatively more on competency-related aspects of
their actions—how well one might express intimate thoughts in a
conversation or in writing—while observers or recipients of these
actions tend to be focused on warmth-related aspects of the same
actions (Fiske et al., 2007; Wojciszke, 1994). If expectations of
social interactions are guided by a heightened attention tohow
they are connecting to others—such as by focusing on the content
of relatively shallow or deep conversation topics—but experiences
are guided more heavily by the warmth conveyed in those interac-
tions—such as the trust conveyed by revealing intimate details in a
deep conversation—then people could systematically underesti-
mate the power of social motivations in others. We believe that
understanding these deeper mechanisms is a critical topic for
future research because it could explain a wide variety of seem-
ingly related phenomena in social life.
Finally, we believe ourfindings have important practical impli-
cations. Social life affords many opportunities to engage, and to
engage more deeply, with others around us. Our participants’
expectations about deeper conversations were not woefully mis-
guided, but they were reliably miscalibrated in a way that could
keep people from engaging a little more deeply with others in their
daily lives. These miscalibrated expectations could matter for two
reasons. First, by creating a barrier to engaging with others, these
expectations could be self-fulfilling because they keep people
from obtaining data that could otherwise calibrate their expecta-
tions. If you think that a deep conversation is likely to be espe-
cially awkward, then you are unlikely to give yourself the chance
tofind out that you might be a little bit wrong. Only by engaging
with others do people accurately understand the consequences of
doing so. Second, strengthening social relationships is critical for
wellbeing, meaning that a reluctance to engage more deeply with
others may leave people being less social than would be optimal
```
## 28 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
for their own wellbeing. Being willing to dig a little deeper than
one might normally go in conversation brings the opportunity to
create a stronger sense of connection with others, especially with
strangers. The world that Schopenhauer described of prickly por-
cupines existing a moderate distance from each other describes a
social life thatcould be, not the one thathas to be.
```
## References

```
Adolphs, R., Tranel, D., & Damasio, A. R. (1998). The human amygdala
in social judgment.Nature, 393 , 470–474. https://doi.org/10.1038/30982
Afifi, T., & Steuber, K. (2009). The revelation risk model (RRM): Factors
that predict the revelation of secrets and the strategies used to reveal
them.Communication Monographs, 76 (2), 144–176.https://doi.org/10
.1080/03637750902828412
Altman, I., & Taylor, D. A. (1973).Social penetration: The development
of interpersonal relationships. Holt, Rinehart & Winston.
Ames, D., Maissen, L. B., & Brockner, J. (2012). The role of listening in
interpersonal influence.Journal of Research in Personality, 46 (3),
345 – 349. https://doi.org/10.1016/j.jrp.2012.01.010
Aron, A., Melinat, E., Aron, E. N., Vallone, R. D., & Bator, R. J. (1997).
The experimental generation of interpersonal closeness: A procedure
and some preliminary findings.Personality and Social Psychology Bul-
letin, 23 (4), 363–377. https://doi.org/10.1177/0146167297234003
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001).
Bad is stronger than good.Review of General Psychology, 5 (4),
323 – 370. https://doi.org/10.1037/1089-2680.5.4.323
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for
interpersonal attachments as a fundamental human motivation.Psycho-
logical Bulletin, 117 (3), 497–529.https://doi.org/10.1037/0033-2909
.117.3.497
Becker, G. S. (1993). Nobel lecture: The economic way of looking at
behavior.Journal of Political Economy, 101 (3), 385–409.https://doi
.org/10.1086/261880
Berenson, K. R., Gyurak, A., Ayduk, O., Downey, G., Garner, M. J.,
Mogg, K., Bradley, B. P., & Pine, D. S. (2009). Rejection sensitivity
and disruption of attention by social threat cues.Journal of Research in
Personality, 43 (6), 1064–1072.https://doi.org/10.1016/j.jrp.2009.07.007
Berg, J. H., & Archer, R. L. (1980). Disclosure or concern: A second look
at liking for the norm breaker.Journal of Personality, 48 (2), 245–257.
https://doi.org/10.1111/j.1467-6494.1980.tb00831.x
Berg, J. H., & Clark, M. S. (1986). Differences in social exchange between
intimate and other relationships: Gradually evolving or quickly appa-
rent? In V. J. Derlega & B. A. Winstead (Eds.),Friendship and social
interaction(pp. 101–128).Springer.https://doi.org/10.1007/978-1-4612
-4880-4_6
Besst, S. (2016).The level of relational self-construal moderates the rela-
tionship between disclosure and well-being(Unpublished honors thesis).
College of Saint Benedict and Saint John’s University, Collegeville,
MN.
Boothby, E. J., & Bohns, V. K. (2021). 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,
47 (5), 826–840. https://doi.org/10.1177/0146167220949003
Boothby, E. J., Clark, M. S., & Bargh, J. A. (2017). The invisibility cloak
illusion: People (incorrectly) believe they observe others more than
others observe them.Journal of Personality and Social Psychology,
112 (4), 589–606. https://doi.org/10.1037/pspi0000082
Bradac, J. J., Hosman, L. A., & Tardy, C. H. (1978). Reciprocal disclosures
and language intensity: Attributional consequences. Communication
Monographs, 45 (1), 1–17. https://doi.org/10.1080/03637757809375947
```
```
Brewere, M. B., & Mittelman, J. (1980). Effects of normative control of
self-disclosure on reciprocity.Journal of Personality, 48 (1), 89–102.
https://doi.org/10.1111/j.1467-6494.1980.tb00968.x
Brown, J. S. (1948). Gradients of approach and avoidance responses and
their relation to level of motivation.Journal of Comparative and Physio-
logical Psychology, 41 (6), 450–465. https://doi.org/10.1037/h0055463
Bruk, A., Scholl, S. G., & Bless, H. (2018). Beautiful mess effect: Self-
other differences in evaluation of showing vulnerability.Journal of Per-
sonality and Social Psychology, 115 (2), 192–205.https://doi.org/10
.1037/pspa0000120
Chaikin, A. L., & Derlega, V. J. (1974). Variables affecting the appropri-
ateness of self-disclosure.Journal of Consulting and Clinical Psychol-
ogy, 42 (4), 588–593. https://doi.org/10.1037/h0036614
Chaikin, A. L., Derlega, V. J., Bayma, B., & Shaw, J. (1975). Neuroticism
and disclosure reciprocity.Journal of Consulting and Clinical Psychol-
ogy, 43 (1), 13–19. https://doi.org/10.1037/h0076294
Cislak, A., & Wojciszke, B. (2008). Agency and communion are inferred
from actions serving interests of self or others.European Journal of
Social Psychology, 38 (7), 1103–1110.https://doi.org/10.1002/ejsp.554
Cline, R. J. W. (1989). The politics of intimacy: Costs and benefits deter-
mining disclosure intimacy in male-female dyads.Journal of Social
and Personal Relationships, 6 (1), 5–20. https://doi.org/10.1177/
026540758900600101
Collins, N. L., & Miller, L. C. (1994). Self-disclosure and liking: A meta-
analytic review.Psychological Bulletin, 116 (3), 457–475.https://doi
.org/10.1037/0033-2909.116.3.457
Cuddy, A. J., Fiske, S. T., & Glick, P. (2007). The BIAS map: Behaviors
from intergroup affect and stereotypes.Journal of Personality and
Social Psychology, 92 (4), 631–648.https://doi.org/10.1037/0022-3514
.92.4.631
Derlega, V. J., Winstead, B. A., & Greene, K. (2001). Self-disclosure and
starting a close relationship. In S. Sprecher, A. Wenzel, & J. Harvey
(Eds.),Handbook of relationship beginnings(pp. 153–174). Routledge
Handbooks Online.
Diener, E., & Seligman, M. E. (2002). Very happy people.Psychological
Science, 13 (1), 81–84. https://doi.org/10.1111/1467-9280.00415
Dunbar, R. I. (1998). The social brain hypothesis.Evolutionary Anthropol-
ogy, 6 (5), 178–190.https://doi.org/10.1002/(SICI)1520-6505(1998)6:
5 ,178::AID-EVAN5.3.0.CO;2-8
Dungan, J. A., & Epley, N. (2021).Surprisingly good talk: Misunderstand-
ing others creates a barrier to constructive confrontation[Unpublished
manuscript]. University of Chicago.
Dunn, E. W., Aknin, L. B., & Norton, M. I. (2008). Spending money on
others promotes happiness.Science, 319 (5870), 1687–1688.https://doi
.org/10.1126/science.1150952
Dunn, E. W., Biesanz, J. C., Human, L. J., & Finn, S. (2007). Misunder-
standing the affective consequences of everyday social interactions: The
hidden benefits of putting one’s best face forward.Journal of Personal-
ity and Social Psychology, 92 (6), 990–1005.https://doi.org/10.1037/
0022-3514.92.6.990
Epley, N., & Schroeder, J. (2014). Mistakenly seeking solitude.Journal of
Experimental Psychology: General, 143 , 1980–1999.https://doi.org/10
.1037/a0037323
Epley, N., & Waytz, A. (2010). Mind perception. In S. T. Fiske, D. T.
Gilbert, & G. Lindzey (Eds.),The handbook of social psychology
(5th ed., pp. 498–541). Wiley.https://doi.org/10.1002/9780470561119
.socpsy001014
Epstein, S., & Fenz, W. D. (1962). Theory and experiment on the measure-
ment of approach-avoidance conflict.The Journal of Abnormal and
Social Psychology, 64 (2), 97–112. https://doi.org/10.1037/h0046187
Fishman, M., & Gardner, W. (2017).Stranger danger? Removing barriers
to social connection through meaningful conversation(Unpublished
honors thesis). Northwestern University, Chicago, IL.
```
## DEEPER CONVERSATION 29

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
Fiske, S. T., Cuddy, A. J., & Glick, P. (2007). Universal dimensions of
social cognition: Warmth and competence.Trends in Cognitive Scien-
ces, 11 (2), 77–83. https://doi.org/10.1016/j.tics.2006.11.005
Fiske, S. T., Cuddy, A. J., Glick, P., & Xu, J. (2002). A model of (often
mixed) stereotype content: Competence and warmth respectively follow
from perceived status and competition.Journal of Personality and
Social Psychology, 82 (6), 878–902.https://doi.org/10.1037/0022-3514
.82.6.878
Gilbert, D. T., & Malone, P. S. (1995). The correspondence bias.Psycholog-
ical Bulletin, 117 (1), 21–38. https://doi.org/10.1037/0033-2909.117.1.21
Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley,
T. P. (1998). Immune neglect: A source of durability bias in affective
forecasting. Journal of Personality and Social Psychology, 75 (3),
617 – 638. https://doi.org/10.1037/0022-3514.75.3.617
Gosling, S. D., Rentfrow, P. J., & Swann, W. B., Jr. (2003). A very brief mea-
sure of the Big-Five personality domains.Journal of Research in Personal-
ity, 37 (6), 504–528. https://doi.org/10.1016/S0092-6566(03)00046-1
Gunaydin, G., Oztekin, H., Karabulut, D. H., & Salman-Engin, S. (2021).
Minimal social interactions with strangers predict greater subjective
well-being.Journal of Happiness Studies, 22 , 1839–1853.https://doi
.org/10.1007/s10902-020-00298-6
Hart, E., VanEpps, E., & Schweitzer, M. (2021). The (better than expected)
consequences of asking sensitive questions.Organizational Behavior
and Human Decision Processes, 162 , 136–154.https://doi.org/10.1016/j
.obhdp.2020.10.014
Haslam, N., Bain, P., Douge, L., Lee, M., & Bastian, B. (2005). More
human than you: Attributing humanness to self and others.Journal of
Personality and Social Psychology, 89 (6), 937–950.https://doi.org/10
.1037/0022-3514.89.6.937
Hayes, A. F. (2013).Introduction to mediation, moderation, and condi-
tional process analysis: A regression-based approach. Guilford Press
Publications.
Holt-Lunstad, J. (2018). Why social relationships are important for physi-
cal health: A systems approach to understanding and modifying risk and
protection.Annual Review of Psychology, 69 , 437–458.https://doi.org/
10.1146/annurev-psych-122216-011902
Hughes, M. E., Waite, L. J., Hawkley, L. C., & Cacioppo, J. T. (2004). A
shortscale for measuring loneliness in large surveys: Results from two
population-based studies.Research on Aging, 26 (6), 655–672.https://
doi.org/10.1177/0164027504268574
John, L. K., Barasz, K., & Norton, M. I. (2016). Hiding personal informa-
tion reveals the worst.Proceedings of the National Academy of Sciences
of the United States of America, 113 (4), 954–959.https://doi.org/10
.1073/pnas.1516868113
Jones, E. E. (1979). The rocky road from acts to dispositions.American
Psychologist, 34 (2), 107–117.https://doi.org/10.1037/0003-066X.34.2
.107
Jones, E. E., & Archer, R. L. (1976). Are there special effects of personal-
istic self-disclosure?Journal of Experimental Social Psychology, 12 (2),
180 – 193. https://doi.org/10.1016/0022-1031(76)90069-X
Kardas, M., Kumar, A., & Epley, N. (2021).Let it go: How exaggerating
the reputational costs of revealing secrets discourages transparency in
relationships. Manuscript submitted for publication.
Kashdan, T. B., McKnight, P. E., Fincham, F. D., & Rose, P. (2011).
When curiosity breeds intimacy: Taking advantage of intimacy
opportunities and transforming boring conversations. Journal of
Personality, 79 (6), 1369–1402.https://doi.org/10.1111/j.1467-6494
.2010.00697.x
Kashdan, T. B., & Roberts, J. E. (2006). Affective outcomes in superficial
and intimate interactions: Roles of social anxiety and curiosity.Journal
of Research in Personality, 40 (2), 140–167.https://doi.org/10.1016/j.jrp
.2004.10.005
Keelan, J. P. R., Dion, K. K., & Dion, K. L. (1998). Attachment style and
relationship satisfaction: Test of a self-disclosure explanation.Canadian
```
```
Journal of Behavioural Science, 30 (1), 24–35. https://doi.org/10.1037/
h0087055
Kelly, A. E., & McKillop, K. J. (1996). Consequences of revealing perso-
nal secrets.Psychological Bulletin, 120 (3), 450–465.https://doi.org/10
.1037/0033-2909.120.3.450
Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006).Dyadic data analysis.
Guilford Press.
Kluger, A. N., & Zaidel, K. (2013). Are listeners perceived as leaders?
International Journal of Listening, 27 (2), 73–84. https:// doi.org/10
.1080/10904018.2013.754283
Kluger, A. N., Malloy, T. E., Pery, S., Itzchakov, G., Castro, D. R., Lipetz,
L., Sela, Y., Turjeman-Levi, Y., Lehmann, M., New, M., & Borut, L.
(2021). Dyadic listening in teams: Social relations model.Applied Psy-
chology, 70 (3), 1045–1099.https://doi.org/10.1111/apps.12263
Kumar, A., & Epley, N. (2018). Undervaluing gratitude: Expressers
misunderstand the consequences of showing appreciation.Psycho-
logical Science, 29 (9), 1423–1435.https://doi.org/10.1177/0956797618
772506
Kumar, A., & Epley, N. (2021).A little good goes an unexpectedly long
way: Underestimating the positive impact of kindness on recipients.
Manuscript in preparation.
Kumar, A., & Epley, N. (2021). It’s surprisingly nice to hear you: Misun-
derstanding the impact of communication media can lead to suboptimal
choices of how to connect with others.Journal of Experimental Psychol-
ogy: General, 150 (3), 595–607. https://doi.org/10.1037/xge0000962
Leary, M. R. (1983). A brief version of the Fear of Negative Evaluation
Scale.Personality and Social Psychology Bulletin, 9 (3), 371–375.
https://doi.org/10.1177/0146167283093007
Leunissen, J. M., De Cremer, D., van Dijke, M., & Folmer, C. P. R.
(2014). Forecasting errors in the averseness of apologizing.Social
Justice Research, 27 (3), 322–339.https://doi.org/10.1007/s11211-014
-0216-4
Levine, E. E., & Cohen, T. R. (2018). You can handle the truth: Mispre-
dicting the consequences of honest communication.Journal of Experi-
mental Psychology: General, 147 (9), 1400–1429.https://doi.org/10
.1037/xge0000488
Mallett, R. K., Wilson, T. D., & Gilbert, D. T. (2008). Expect the unex-
pected: Failure to anticipate similarities leads to an intergroup forecast-
ing error.Journal of Personality and Social Psychology, 94 (2),
265 – 277. https://doi.org/10.1037/0022-3514.94.2.94.2.265
Maner, J. K., DeWall, C. N., Baumeister, R. F., & Schaller, M. (2007).
Does social exclusion motivate interpersonal reconnection? Resolving
the“porcupine problem.Journal of Personality and Social Psychology,
92 (1), 42–55. https://doi.org/10.1037/0022-3514.92.1.42
Margolis, S., & Lyubomirsky, S. (2020). Experimental manipulation of
extraverted and introverted behavior and its effects on well-being.Jour-
nal of Experimental Psychology: General, 149 (4), 719 – 731.https://doi
.org/10.1037/xge0000668
Mehl, M. R., Vazire, S., Holleran, S. E., & Clark, C. S. (2010). Eavesdrop-
ping on happiness: Well-being is related to having less small talk and
more substantive conversations.Psychological Science, 21 (4), 539–541.
https://doi.org/10.1177/0956797610362675
Milek, A., Butler, E. A., Tackman, A. M., Kaplan, D. M., Raison, C. L.,
Sbarra, D. A., Vazire, S., & Mehl, M. R. (2018). Eavesdropping on hap-
piness”revisited: A pooled, multisample replication of the association
between life satisfaction and observed daily conversation quantity and
quality.Psychological Science, 29 (9), 1451–1462.https://doi.org/10
.1177/0956797618774252
Miller, N. E. (1944). Experimental studies of conflict. In J. McV.
Hunt (Ed.), Personality and the behavior disorders (pp.
431 – 465). Ronald.
Montoya, A. K., & Hayes, A. F. (2017). Two-condition within-participant
statistical mediation analysis: A path-analytic framework.Psychological
Methods, 22 (1), 6–27. https://doi.org/10.1037/met0000086
```
## 30 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


```
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know:
Verbal reports on mental processes.Psychological Review, 84 (3),
231 – 259. https://doi.org/10.1037/0033-295X.84.3.231
Omarzu, J. (2000). A disclosure decision model: Determining how and
when individuals will self-disclose.Personality and Social Psychology
Review, 4 (2), 174–185.https://doi.org/10.1207/S15327957PSPR0402
_05
Pennebaker, J. W. (1997). Writing about emotional experiences as a thera-
peutic process.Psychological Science, 8 (3), 162–166.https://doi.org/10
.1111/j.1467-9280.1997.tb00403.x
Pietrzak, J., Downey, G., & Ayduk, O. (2005). Rejection sensitivity as an
interpersonal vulnerability. In M. Baldwin (Ed.),Interpersonal cognition
(pp. 62–84). Guilford Press.
Ruan, Y., Reis, H. T., Clark, M. S., Hirsch, J. L., & Bink, B. D. (2020).
Can I tell you how I feel? Perceived partner responsiveness encourages
emotional expression.Emotion, 20 (3), 329–342.https://doi.org/10.1037/
emo0000650
Rubin, Z. (1975). Disclosing oneself to a stranger: Reciprocity and its lim-
its.Journal of Experimental Social Psychology, 11 (3), 233–260.https://
doi.org/10.1016/S0022-1031(75)80025-4
Russell, A. M. T., & Fiske, S. T. (2008). It’s all relative: Competition and
status drive interpersonal perception.European Journal of Social Psy-
chology, 38 (7), 1193–1201.https://doi.org/10.1002/ejsp.539
Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facil-
itation of intrinsic motivation, social development, and well-being.
American Psychologist, 55 (1), 68–78. https://doi.org/10.1037/0003
-066X.55.1.68
Sandstrom, G. M., & Boothby, E. J. (2021). Why do people avoid talking
to strangers? A mini meta-analysis of predicted fears and actual experi-
ences talking to a stranger.Self and Identity, 20 (1), 47–71. https://doi
.org/10.1080/15298868.2020.1816568
Sandstrom, G. M., & Dunn, E. W. (2014a). Is efficiency overrated? Mini-
mal social interactions lead to belonging and positive affect.Social Psy-
chological and Personality Science, 5 (4), 437–442.https://doi.org/10
.1177/1948550613502990
Sandstrom, G. M., & Dunn, E. W. (2014b). Social interactions and
well-being: The surprising power of weak ties.Personality and
Social Psychology Bulletin, 40 (7), 910–922.https://doi.org/10.1177/
0146167214529799
Savitsky, K., Epley, N., & Gilovich, T. (2001). Do others judge us as
harshly as we think? Overestimating the impact of our failures, short-
comings, and mishaps.Journal of Personality and Social Psychology,
81 (1), 44–56. https://doi.org/10.1037/0022-3514.81.1.44
Schopenhauer, A. (1964).The pessimist’s handbook(T. B. Saunders,
Trans.). University of Nebraska Press. (Original work published
1851)
Schroeder, J., Lyons, D., & Epley, N. (in press). Hello, stranger?: Pleasant
conversations are preceded by fears of starting one.Journal of Experi-
mental Psychology: General.
Shearer, C. (2017).More than just talking: The role of self-disclosure in
the fast friends procedure(Unpublished honors thesis). Seton Hall Uni-
versity, South Orange, NJ.
```
```
Slepian, M. L., & Moulton-Tetlock, E. (2019). Confiding secrets and well-
being.Social Psychological and Personality Science, 10 (4), 472–484.
https://doi.org/10.1177/1948550618765069
Sprecher, S., Treger, S., Wondra, J. D., Hilaire, N., & Wallpe, K. (2013).
Taking turns: Reciprocal self-disclosure promotes liking in initial inter-
actions.Journal of Experimental Social Psychology, 49 (5), 860–866.
https://doi.org/10.1016/j.jesp.2013.03.017
Strachman, A., & Gable, S. L. (2006). What you want (and do not want)
affects what you see (and do not see): Avoidance social goals and social
events.Personality and Social Psychology Bulletin, 32 (11), 1446–1458.
https://doi.org/10.1177/0146167206291007
Sun, J., Harris, K., & Vazire, S. (2020). Is well-being associated with the quan-
tity and quality of social interactions?Journal of Personality and Social Psy-
chology, 119 (6), 1478–1496.https://doi.org/10.1037/pspp0000272
Taylor, D. A. (1968). The development of interpersonal relationships:
Social penetration processes.The Journal of Social Psychology, 75 (1),
79 – 90. https://doi.org/10.1080/00224545.1968.9712476
Todorov, A. (2008). Evaluating faces on trustworthiness: An extension of
systems for recognition of emotions signaling approach/avoidance
behaviors.Annals of the New York Academy of Sciences, 1124 (1),
208 – 224. https://doi.org/10.1196/annals.1440.012
Todorov, A., Pakrashi, M., & Oosterhof, N. N. (2009). Evaluating faces on
trustworthiness after minimal time exposure.Social Cognition, 27 (6),
813 – 833. https://doi.org/10.1521/soco.2009.27.6.813
von Hippel, W. (2018).The social leap. Harper Collins.
Waytz, A., Schroeder, J., & Epley, N. (2014). The lesser minds problem.
In P. G. Bain, J. Vaes, & J.-P. Leyens (Eds.),Humanness and dehuman-
ization(pp. 49–67). Psychology Press.
Wesselmann, E. D., Cardoso, F. D., Slater, S., & Williams, K. D. (2012).
To be looked at as though air: Civil attention matters.Psychological Sci-
ence, 23 (2), 166–168. https://doi.org/10.1177/0956797611427921
Wilson, T. D., Wheatley, T., Meyers, J. M., Gilbert, D. T., & Axsom, D.
(2000). Focalism: A source of durability bias in affective forecasting.
Journal of Personality and Social Psychology, 78 (5), 821–836.https://
doi.org/10.1037/0022-3514.78.5.821
Wojciszke, B. (1994). Multiple meanings of behavior: Construing actions
in terms of competence or morality.Journal of Personality and Social
Psychology, 67 (2), 222–232. https://doi.org/10.1037/0022-3514.67.2
.222
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. https://doi.org/10.1177/
01461672982412001
Zelenski, J. M., Whelan, D. C., Nealis, L. J., Besner, C. M., Santoro,
M. S., & Wynn, J. E. (2013). Personality and affective forecasting: Trait
introverts underpredict the hedonic benefits of acting extraverted.Jour-
nal of Personality and Social Psychology, 104 (6), 1092–1108.https://
doi.org/10.1037/a0032281
Zhao, X., & Epley, N. (in press). Insufficiently complimentary?: Underes-
timating the positive impact of compliments creates a barrier to express-
ing them.Journal of Personality and Social Psychology.
```
```
(Appendix follows)
```
## DEEPER CONVERSATION 31

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.


## Appendix

## Intimacy Ratings by Question

```
Index Question Mintimacy(SDintimacy)
S01 What do you think about the weather today? 1.21 (2.11)
S02 How often do you come here? 1.97 (2.06)
S03 How did you celebrate last Halloween? 2.14 (2.09)
S04 How often do you get your hair cut? Where do you go? Have you ever had a really bad haircut experience? 2.39 (2.13)
S05 What is the best TV show you've seen in the last month? Tell your partner about it. 2.68 (2.31)
S06 When was the last time you walked for more than an hour? Describe where you went and what you saw. 2.69 (2.19)
S07 Do you like to get up early or stay up late? Why? 2.89 (1.93)
S08 Do you have anything planned for later today? When are you going to do it? 3.22 (2.01)
S09 Can you describe a conversation you had with another person earlier today? 3.30 (2.14)
S10 What’s your daily routine like? 4.40 (1.85)
D01 What would constitute a“perfect”day for you? 4.75 (2.38)
D02 Where is somewhere you've visited that you felt really had an impact on who you are today? 5.24 (1.94)
D03 If you were going to become a close friend with the other participant, please share what would be important for him or her to
know.
```
```
6.35 (1.99)
```
```
D04 If a crystal ball could tell you the truth about yourself, your life, the future, or anything else, what would you want to know? 6.38 (2.12)
D05 For what in your life do you feel most grateful? Tell the other participant about it. 6.47 (1.96)
D06 Is there something you've dreamed of doing for a long time? Why haven't you done it? 6.50 (2.10)
D07 What is one of the more embarrassing moments in your life? 6.83 (1.88)
D08 What is one of your most meaningful memories? Why is it meaningful for you? 7.29 (2.02)
D09 Can you describe a time you cried in front of another person? 7.39 (2.16)
D10 If you could undo one mistake you have made in your life, what would it be and why would you undo it? 7.82 (1.98)
Note. Intimacy ratings by question. Numbers outside parentheses denote means. Numbers inside parentheses denote standard deviations. S01 through
S10 denote shallow questions in ascending order of intimacy. D01 through D10 denote deep questions in ascending order of intimacy.
```
```
Received November 19, 2020
Revision received April 29, 2021
Accepted May 4, 2021n
```
## 32 KARDAS, KUMAR, AND EPLEY

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.