Access to this full-text is provided by SAGE Publications Inc.
Content available from Communication Research
This content is subject to copyright.
https://doi.org/10.1177/0093650221995316
Communication Research
2022, Vol. 49(6) 863 –890
© The Author(s) 2021
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0093650221995316
journals.sagepub.com/home/crx
Article
What People Look at in
Multimodal Online Dating
Profiles: How Pictorial
and Textual Cues Affect
Impression Formation
Tess van der Zanden1, Maria B. J. Mos1,
Alexander P. Schouten1, and Emiel J. Krahmer1
Abstract
This study investigates how online dating profiles, consisting of both pictures and
texts, are visually processed, and how both components affect impression formation.
The attractiveness of the profile picture was varied systematically, and texts either
included language errors or not. By collecting eye tracking and perception data, we
investigated whether picture attractiveness determines attention to the profile text
and if the text plays a secondary role. Eye tracking results revealed that pictures
are more likely to attract initial attention and that more attractive pictures receive
more attention. Texts received attention regardless of the picture’s attractiveness.
Moreover, perception data showed that both the pictorial and textual cues affect
impression formation, but that they affect different dimensions of perceived attraction
differently. Based on our results, a new multimodal information processing model is
proposed, which suggests that pictures and texts are processed independently and
lead to separate assessments of cue attractiveness before impression formation.
Keywords
online dating, eye tracking, impression formation, interpersonal attraction, dating
profiles, profile picture attractiveness, language errors
1Tilburg University, The Netherlands
Corresponding Author:
Tess van der Zanden, Department of Communication and Cognition, Tilburg University, Warandelaan 2,
P.O. Box 90153, Tilburg 5000 LE, The Netherlands.
Email: T.vdrZanden@tilburguniversity.edu
995316CRXXXX10.1177/0093650221995316Communication Researchvan der Zanden et al.
research-article2021
864 Communication Research 49(6)
Cues in both pictures and texts on online dating profiles can affect the impressions
people form of the owner of the profile. For instance, profile owners with attractive
pictures score higher on perceived physical attraction and dating desirability than
those with unattractive pictures (e.g., Fiore et al., 2008; McGloin & Denes, 2018), but
score lower on authenticity (Lo et al., 2013). Regarding dating profile texts, language
errors (Van der Zanden et al., 2020), low self-described ambition (Sritharan et al.,
2010), and providing highly selective positive information (Wotipka & High, 2016)
have been shown to negatively impact perceptions of a profile owner’s attractiveness,
likeability, and dating desirability.
When searching for romantic potential, the single most important determinant is
the other person’s physical attractiveness (e.g., Hitsch et al., 2010; Toma & Hancock,
2010). In multimodal dating profiles, this information can mainly be derived from a
profile picture. Previous research into initial impression formation has shown that
based on pictorial information, people rapidly, intuitively, and effortlessly form
impressions about physical attractiveness and use this information to infer personality
traits (e.g., Dion et al., 1972; Miller, 1970; Olson & Marshuetz, 2005; Willis &
Todorov, 2006). With regards to impression formation in online dating settings, this
implies that information from the profile picture automatically leads to an initial
impression, after which people could either stick to this impression without much
more deliberation or process other available information on the profile more deliber-
ately, such as the profile text. These two ways of processing a multimodal dating pro-
file may be likened to the two processing modes proposed by prevalent dual information
processing theories as elaboration likelihood model (ELM; Petty & Cacioppo, 1986)
and heuristic-systematic model (HSM; Chaiken, 1980, 1987).
A profile picture may thus function as the profile’s gatekeeper, with the profile
owner’s physical attractiveness determining whether there is need for additional infor-
mation processing from the profile text. The ambiguity of picture information is likely
to play a role here: extreme cues with little ambiguity, such as highly attractive or
unattractive pictures, oftentimes result in an immediate and (rather) consistent impres-
sion about this person (e.g., Miller, 1970; Willis & Todorov, 2006). Picture informa-
tion will then often be sufficient for impression formation and there may therefore be
little need for more information from the profile text. On the other hand, ambiguous
pictures, such as those that are neither attractive nor unattractive (i.e., moderately
attractive), may not provide enough information to form an impression. To compen-
sate for this, people can develop this impression by being more attentive to the profile
text and using the cues within this text.
Previous research on the effects of pictorial and textual cues on impression forma-
tion focused primarily on the outcomes of impression formation processes (e.g., Fiore
et al., 2008; Lo et al., 2013; Van der Zanden et al., 2020). While inferences can be
made about how attentive people are to profile pictures and texts based on these stud-
ies, little is known about how attention is actually allocated to these profile compo-
nents to establish these impressions, and how profile cues affect information
processing. An effective way of investigating attention allocation and information pro-
cessing regarding dating profiles is by means of eye tracking. Eye movement behavior
van der Zanden et al. 865
can provide information about which profile component attracts initial and most atten-
tion (e.g., Scott & Hand, 2016; Seidman & Miller, 2013 on Facebook profiles), and
can shed light on how cues in profile components affect information processing.
The main goal of this study is to investigate how pictures and texts on online dating
profiles are visually processed. More specifically, we examine the potential of the
profile picture—and its attractiveness—as the profile’s gatekeeper, by investigating
(a) whether profile picture attractiveness influences text processing, as measured by
frequency and duration of fixations, and (b) whether profile picture attractiveness
influences the extent to which textual cues (here language errors) affect impression
formation, as measured by ratings on perceived attraction. To do so, participants
viewed multimodal dating profiles which comprise a picture that scores high, moder-
ate, or low in attractiveness and a text with or without language errors, and rated the
profile owners in terms of perceived attraction.
Processing and Assessing Profile Cues
Pictures often immediately evoke a spontaneous affective response, especially since
attractiveness cues are highly salient (Olson & Marshuetz, 2005; Willis & Todorov,
2006). With only a few glances at a picture, people can already gather information
ranging from, for example, age, ethnicity, and eye color to physical attractiveness.
Without much effort or awareness, information about physical attractiveness can lead
to fast, intuitive, and unreflective impressions on other dimensions of attractiveness
and personality traits (Locher et al., 1993; Willis & Todorov, 2006). The valence (posi-
tive or negative) of the impression about a person’s physical attractiveness is usually
consistent with impressions regarding other characteristics of that person. For exam-
ple, physically attractive people are also perceived as more likeable, trustworthy, care-
ful and confident, and unattractive people as insensitive, less trustworthy and more
aggressive (Dion et al., 1972; Miller, 1970; Willis & Todorov, 2006).
Picture information is not only easily accessible to form a quick initial impression,
but also carries a great impression formation weight. A visual primacy effect is found
before, emphasizing the importance of profile pictures over profile texts when forming
impressions about Facebook profile owners (D’Angelo et al., 2014; Van der Heide
et al., 2012). Pictures carry more weight in the final assessment (e.g., Fiore et al.,
2008) and are also more likely to attract initial attention (Scott & Hand, 2016; Seidman
& Miller, 2013). In addition to providing important information about profile owners’
looks, pictures are an easy “point of entry” (Scott & Hand, 2016). In particular, picto-
rial information is arguably less densely packed than textual information, which sug-
gests that more information can be decoded per fixation on a picture than on a text
(Rayner et al., 2001). Taken together, this indicates that it is more efficient for people
to first focus on pictorial rather than textual information when being presented with
multimodal dating profiles, and we therefore pose:
H1. When viewing online dating profiles, containing a profile picture and text,
people first fixate on the profile picture rather than on the profile text.
866 Communication Research 49(6)
While profile pictures are likely to attract initial attention regardless of the person’s
beauty, the physical attractiveness of a depicted person presumably determines the
frequency and duration of the fixations on the picture. Both how often and how long
people look at a component can signify which information resources are processed and
receive cognitive attention (Rayner, 1998; Scott & Hand, 2016). Studies have shown
that beauty captures attention: the more attractive the face, the more and longer people
look at it (e.g., Langlois et al., 2000; Leder et al., 2016; Maner et al., 2003; Valuch
et al., 2015). In previous studies, eye tracking has been used as a method to demon-
strate that gazing varied as a function of physical attractiveness. Particularly within the
field of psychology, various eye tracking studies have shown that people tend to look
longer at attractive individuals on pictures (Leder et al., 2010, 2016; Maner et al.,
2003; Mitrovic et al., 2016) and in (offline) dating contexts (Van Straaten et al., 2010),
but also in a social media context where other (non-pictorial) information was avail-
able on profiles (Seidman & Miller, 2013).
This bias of gazing at attractive individuals appears to be functional. In relation-
ships, people want to maximize their outputs. In deciding the rewards that people may
derive from others, physical attractiveness is one of the most important factors (Dion
et al., 1972; Walster et al., 1973). Following this line of research, it is expected that on
online dating profiles the attractiveness of a profile picture similarly determines how
often and long people look at the picture. This leads to the following hypothesis:
H2. The more attractive the profile picture on an online dating profile, the more and
longer people fixate on this profile picture.
When being exposed to attractive or unattractive pictures of potential partners on
dating profiles, people form initial impressions about physical attractiveness based on
intuitive reactions (Sritharan et al., 2010). The “what is beautiful is good” stereotype
poses that physical attractiveness fosters positive attributions about other personality
traits that are (socially) desirable (Dion et al., 1972). Considering this stereotype, it is
likely that an initial impression based on physical attractiveness is extrapolated to an
overall impression about the profile owner: a positive impression of a physically
attractive profile owner and a negative impression of a profile owner with an unat-
tractive picture.
In an online dating context, Sritharan et al. (2010) previously examined how intui-
tive heuristic responses to profile pictures affected assessments of other social infor-
mation, in their study, profile text information. They conducted two studies among
female participants who viewed and rated online dating profiles containing an attrac-
tive or unattractive picture and a text with high or low self-described ambition—with
high ambition being more attractive than low ambition. They found that only picture
attractiveness affected spontaneous affective responses, as measured in an affective
priming task, whereas both picture attractiveness and self-described ambition influ-
enced reported scores on likeability. More specifically, text evaluations varied as a
function of spontaneous picture evaluations: for both profiles with high and low self-
described ambition, likeability scores were higher when an attractive picture evoked a
van der Zanden et al. 867
positive spontaneous response than when a negative spontaneous response was evoked
by an unattractive picture (Sritharan et al., 2010). These results indicate that the attrac-
tiveness-related initial impressions influenced likeability impressions based on profile
text attractiveness.
Once an overall initial impression about a profile owner has been formed based
on picture attractiveness, people may or may not proceed to put more effort in pro-
cessing other available profile information, such as the profile text. If the initial
impression about physical attractiveness is strongly valenced, that is: very positive
or very negative, little further information processing may be needed, because phys-
ical attractiveness is such a strong determinant of perceived attraction in (online)
dating. After all, people are willing to date highly attractive others anyways (Walster
et al., 1966), whereas profile owners with unattractive pictures will presumably be
excluded as potential partners based on the picture information that has been evalu-
ated first (Fiore et al., 2008).
This implies that a profile owner’s physical attractiveness, as depicted in the profile
picture, may function as a “gatekeeper” to the rest of the profile: picture attractiveness
can determine to what extent profile texts receive attention in addition to the profile
picture. This not to say that people will not look at the profile text at all when seeing
an attractive or unattractive picture, but that relatively less attention will be paid to the
profile texts. We expect people to process the text more heuristically, as the picture
provides enough information to form an impression of the profile owner’s attractive-
ness, diminishing the need for further information processing (Chaiken, 1980, 1987;
Petty & Cacioppo, 1986). The schematic representation of this picture gatekeeper
model is presented in Figure 1, and here (a) illustrates the extreme case in which the
positive and negative impressions formed based on attractive or unattractive pictures,
respectively, immediately lead to perceptions on all three dimensions of attraction,
with no attention paid to the profile text.1
When a person is moderately attractive, information about physical attractiveness
remains ambiguous in a picture. Therefore, heuristic processing of the profile may not
be possible, as a profile owner can neither be immediately accepted (in the case of an
attractive picture) nor rejected (unattractive picture) as a potential romantic partner. In
such cases, mental shortcuts or stereotypes may not be applicable. People may conse-
quently seek for additional information to form an impression. In the case of multi-
modal online dating profiles, this could mean shifting attention from the profile picture
to the profile text. Therefore, we pose that when a profile picture provides ambiguous
information about physical attractiveness (i.e., when the profile owner is moderately
attractive), more systematic (central) processing will take place, resulting in more
attention to textual information than when the initial impression is unambiguous (i.e.,
when the picture is either attractive or unattractive). In this latter case, more heuristic
(peripheral) processing of the profile will take place, resulting in less profile text atten-
tion. Path (b) in Figure 1 illustrates this alternative route for profiles with moderately
attractive pictures compared to that of profiles with attractive and unattractive pic-
tures. We thus hypothesize:2
868 Communication Research 49(6)
H3. People fixate more often and longer on the profile text when a moderately
attractive picture is shown on the online dating profile than when the profile con-
tains an (a) attractive or (b) unattractive profile picture.
Our picture gatekeeper model, as depicted in Figure 1, is compatible with prevalent
models of dual information processing, such as the elaboration likelihood model
Figure 1. Schematic representation of the picture gatekeeper model.
van der Zanden et al. 869
(ELM; Petty & Cacioppo, 1986) and heuristic-systematic model (HSM; Chaiken,
1980, 1987), in that it proposes an automatic and a more deliberate information pro-
cessing route. Our model has, however, been formulated specifically to describe how
information on multimodal online dating profiles is processed automatically and sys-
tematically, whereas ELM and HSM were introduced in the context of persuasion and
have later been applied to general information processing. Notice, incidentally, that
while information in pictures can be processed more quickly and holistically, and text
processing requires more time and effort (e.g., Barry, 1997; Chaiken & Eagly, 1976),
there is no necessary one-on-one link between pictorial information and automatic
processing on the one hand, and textual information and deliberate processing on the
other hand in theories about dual information processing.
The gatekeeping function of the picture makes it likely that people will not only
be more attentive to textual cues when pictorial cues are ambiguous, but will also
rely on those cues more heavily to accrue an impression of a profile owner with a
moderately attractive picture (Tidwell & Walther, 2002). It is expected that people
extend the initial positive or negative impression for profiles with attractive or unat-
tractive pictures to impressions on other dimensions of attraction, such as about the
person’s social attraction, which indicates people’s desires to spend time with some-
one (McCroskey & McCain, 1974), and romantic attraction, that is, how much peo-
ple feel romantically attracted to someone (Campbell, 1999). This overall impression
of attractiveness is then presumably not amplified by attractive or unattractive tex-
tual cues in a profile. However, if picture information is ambiguous, people may be
more prone to cues that could help them come to an overall impression. Therefore,
cues in texts accompanied by moderately attractive profile pictures may carry more
impression formation weight compared to the same textual cues in profiles with
attractive or unattractive pictures.
To test this in this study, we manipulate language errors in the profile texts.
Language errors have been shown to negatively affect perceptions of attraction, espe-
cially perceived social and romantic attraction (Van der Zanden et al., 2020). Path (c)
in Figure 1 shows how for profiles with moderately attractive pictures, the absence or
presence of language errors may affect perceived attraction, where the presence of
language errors has a negative effect and the absence has a positive effect. This leads
to the following hypothesis:
H4. Language errors in a profile text negatively affect perceptions of profile
owners’ attractiveness more strongly when a profile picture on an online dating
profile is moderately attractive than when a profile picture is attractive or
unattractive.
Method
Before data collection, ethical clearance was obtained in Spring 2019 by the Ethics
Committee (ETC) of the school of our university. On the Open Science Framework,
the research design, hypotheses, and analysis plan were preregistered: osf.io/cbtv2/
registrations.
870 Communication Research 49(6)
Participants
In this study, 57 undergraduate students participated, who earned credits for their par-
ticipation. Due to technical difficulties and calibration problems, data of 48 partici-
pants could be included for analyses. All participants had normal or corrected to
normal vision, were native speakers of Dutch and were between 18 and 27 years old
(M = 21.4 years, SD = 2.34; 67.2% women).
Design and Material
This experiment had a 3 × 2 design, with both profile picture attractiveness (attractive/
moderately attractive/unattractive picture) and language errors (language errors/no
language errors) as within-subject variables. Participants were exposed to all condi-
tions, and were presented with a total of 18 dating profiles, three from each condition.
We thus aimed for a total of 864 cases, that is, 48 (participants) × 18 (dating profiles).
However, due to technical issues 33 observations are missing, resulting in a total of
831 cases. All dating profiles consisted of a picture and text, and whether a participant
saw profiles of males or females was determined by asking the participants what gen-
der they feel most attracted to.
Profile pictures. For both male and female profiles, 18 pictures were selected: six
attractive, six moderately attractive, and six unattractive pictures. Which pictures
belonged to each of the three categories was determined with a pretest, in which 85
participants (Mage = 26.7 years, SDage = 5.47, 68.2% women) rated the physical attrac-
tiveness of 50 depicted individuals of their indicated preferred sex on a ten-point scale.
The pictures, which were preselected from free stock image sites (e.g., PXhere, Flickr,
Pexels) and were licensed under creative commons, ranged in expected physical
attractiveness. Any potential confounding picture variables that could affect percep-
tions of picture attractiveness other than the depicted person’s physical attractiveness
were avoided. The set of pictures was homogenous in terms of characteristics and
demographics, and matched with the presumed demographics of the participants (e.g.,
age, ethnicity). All people were frontally depicted with head and shoulders, looked
friendly into the camera, and did not depict characteristics that may repel or attract
participants immediately, such as with regards to clothing style, piercings, or tattoos.
Scores for the 50 male pictures were between 1.58 and 6.77 with an average score
of 3.62 (SD = 1.67) and the 50 female pictures scored between 2.15 and 8.04 with an
average score of 4.96 (SD = 1.67). For both sets of pictures, the six pictures that scored
highest were selected as the attractive pictures (male pictures: M = 6.18, SD = 1.94,
Mdn = 7.00; female pictures: M = 7.46, SD = 1.30, Mdn = 8.00), and the six that scored
lowest were categorized as the unattractive pictures (male pictures: M = 1.83, SD = 1.02,
Mdn = 2.00; female pictures: M = 2.51, SD = 1.58, Mdn = 2.00). The pictures in the
moderately attractive category were those six pictures that scored right above and
below the average score given to the 50 pictures. The six male pictures within this
category scored on average 3.65 (SD = 1.96, Mdn = 3.00) and the six female pictures
van der Zanden et al. 871
4.95 (SD = 1.75, Mdn = 5.00). Planned contrast analyses of the pretest showed that for
both men and women all groups were rated significantly different from each other,
with t’s > 13.00 and p’s < .001. Also in the main experiment, attractive, moderately
attractive, and unattractive pictures differed significantly in the scores given on per-
ceived physical attraction, with for all contrasts t’s > 15.83 and p’s < .001.
Profile texts. Based on existing dating profiles (Van der Zanden et al., 2018), 18
profile texts were constructed, with different contents for each. This content was kept
neutral, meaning that no potentially extreme preferences, interests, or hobbies were
mentioned. Primarily, the texts provided a textual description of who the profile owner
is, what their hobbies/interests are and what kind of partner and relationship (s)he
seeks. The contents of texts supposedly written by male and female profile owners
were identical, with the exception of one or two gender-specific word(s) in each text
(e.g., “man,” “her”). Texts ranged from 60 to 67 words.
For each profile text, a version with and without language errors was created. The
text with errors each contained eight errors, based on observations made in a corpus
analysis on language errors in existing dating profiles (Van der Zanden et al., 2018).
Different types of errors were included in each text (e.g., grammatical, typographical
error). Figure 2 shows two examples of translated versions of the dating profiles used
for the experiment. The profile on the right includes language errors, with the first
three errors, for example, being a spacing error, a typographical error, and an ellipsis,
respectively.
A pretest with 55 participants (Mage = 26.3 years, SDage = 7.71, 63.6% women) was
conducted to check whether the overall text quality of texts with language errors was
rated lower compared to texts without errors. To measure this, for half of the texts
participants answered the question “How do you assess the quality of this text?” on a
scale from 1 to 10. Overall text quality scores were lower for texts with errors
Figure 2. Examples of translated English versions of the original Dutch dating profiles used
in the experiment.
Note. The left profile is a moderately attractive male profile without language errors and with the picture
on the left side, and the right profile presents an attractive female profile with language errors and
with the picture right-sided. Language errors are underlined here, but not in the profiles presented to
participants.
872 Communication Research 49(6)
(M = 4.18, SD = 1.32) than for texts without errors (M = 6.29, SD = 1.29), F(1,
52) = 88.58, p < .001, ηp2 = .630. Based on these pretest findings, some minor adapta-
tions were made to 5 of the 18 texts for the main experiment. This was done to make
the overall perceived text quality scores of all texts with and without errors more
even; parts of two texts were exchanged as the overall score was relatively high for
one text and relatively low for the other. In two other texts, one error was replaced by
a less remarkable one as the error version of this text scored lower compared to other
texts with errors. Finally, the overall text quality score of one error-free text was rela-
tively low, and a sentence was reformulated in such a way it was expected to improve
overall text quality.
Dating profiles. A subset of 108 combinations of pictures and texts were made
as experimental material for both male and female profiles (6 pictures × 18 texts).
All picture-text combinations were unique and randomized to avoid an effect of a
particular picture or text. To control for a potential leftward fixation bias (Leder
et al., 2016), half of the profiles a participant saw had the picture on the profile’s
left side, and the other half depicted it on the right side. The position of the picture
and text was fully crossed across conditions. The profile’s picture and text were of
equal size (See Figure 2). All materials used for this study are available in OSF, at
osf.io/cbtv2/.
Procedure
Participants were seated approximately 70 cm in front of a computer screen
(1,680 × 1,050 pixels) at the eye tracking laboratory of our university. Before the
experiment started, participants were welcomed and provided written informed con-
sent. Then, the participant’s eyes were calibrated and validated using a nine-point pro-
cedure using the SMI RED 2015 device with a sampling rate of 250 Hz. The experiment
was opened in Qualtrics after successful calibration, and participants’ eye movements
on this site were followed. They were then instructed and answered some demographic
questions (i.e., gender, age, sexual preference, relationship status). Participants were
told that they would see around twenty mock-up dating profiles that they had to view
and judge in a natural manner. Before participants were randomly assigned to one of
the available six lists of 18 male or female profiles (based on the indicated sexual
preference), they viewed one test profile. The 18 profiles, from which nine contained
language errors and nine were free of errors, were then presented to them in a random
order. When participants were done viewing a profile, they pressed a key to answer
three impression formation statements about the owner of the profile.3 Once the state-
ments were answered, a cross was displayed on the screen for 3 seconds on which
participants had to focus before the next profile appeared. When all 18 profiles were
viewed and rated, one last post-test profile with language errors was shown, about
which they also had to indicate whether they had noticed language errors in that profile
(86.1% indicated to have noticed the errors). The pre- and post-test profiles used were
the same across lists, and the data from these profiles were not analyzed. In total, the
van der Zanden et al. 873
experiment took approximately 20 minutes, dependent on the duration of the
calibration.
Measures and Analysis
Eye tracking measures. Only fixations that occurred on either the profile picture or
text were included, and a fixation only counted as one if it lasted 40ms or longer (Bar
et al., 2006; Scott & Hand, 2016). As each participant assessed 18 profiles, there were
18 cases for each participant (3.70% missing cases because no fixations occurred on
the profile at all). While viewing was binocular, eye movements from the right eye
were analyzed with an average tracking ratio of 97.2% (SD = 4.67). Five eye tracking
measures were used as dependent variables to test our hypotheses: (a) first fixation
location, that is whether the first fixation occurred on either the profile picture or text
(H1), (b) picture fixation count (abbreviated as PFC), that is the total number of fixa-
tions on the picture (H2), (c) picture fixation duration (PFD), being the total fixation
duration on the picture in milliseconds (H2), (d) fixation count on the text (TFC), that
is the number of fixations on the text (H3), and (e) fixation duration on the text (TFD),
that is the total fixation duration on the text in milliseconds (H3).
Attractiveness perception measures. Impressions of attractiveness were measured
with three items, each covering another dimension of perceived attraction: physical
attraction, social attraction (McCroskey & McCain, 1974), and romantic attraction
(Campbell, 1999). To fit our experiment, the wording of the used items was translated
and slightly adjusted. The items were “I think this person is good-looking” for physi-
cal attraction, “I think this person is nice to spend time with” for social attraction, and
“I feel attracted to this person” for romantic attraction. Each of these items were mea-
sured on a scale from 1 (completely disagree) to 7 (completely agree).4
Analysis. To test whether pictures are more likely to attract initial attention (H1),
a chi-square test was performed, with first fixation location (picture/text) and pic-
ture position (left/right) as binary variables. To test H2, H3, and H4, linear effect
models were conducted in R using the lme4 package (Bates et al. 2014),5 as well
as the lmerTest package to obtain p-values applying the Satterthwaite approxima-
tion (Kuznetsova et al., 2014). Picture attractiveness and language error presence
were included as fixed factors and a random intercept was included for participants.6
Interactions between factors were included. For H2 the dependent variables were the
picture fixation count and duration and for H3 the text fixation count and duration.
The mean scores on all three dimensions of perceived attraction were the dependent
variables for H4. The data underlying this article are available in OSF, at osf.io/cbtv2/.
Results
In this section, we first report on the tests of the preregistered hypotheses on both the
eye tracking and perception data, followed by an exploratory follow-up analysis sec-
tion that zooms in more specifically on how the dating profiles are processed.
874 Communication Research 49(6)
Independent of picture or text condition, participants fixated on average 10.97 sec-
onds (SD = 6.70) on the dating profiles, from which 2.23 seconds (20.3%, SD = 2.75)
were on the picture and 8.74 seconds (79.7%, SD = 4.99) on the text. Within this period
of time, people made on average 59.8 (SD = 32.3) fixations per profile, from which 9.5
(15.9%, SD = 11.6) were on the picture and 50.3 (84.1%, SD = 25.4) on the text.
Main Results
Eye tracking data
First fixation location. In accordance with H1, the chi-square test with first fixation
location and picture position as the variables was significant, χ2(1) = 11.29, p < .001.
People were in general more likely to fixate initially on the picture than on the text,
and were more likely to initially fixate on the picture when it occurred on the profile’s
left side than on the profile’s right side. Table 1 provides the percentage scores. On
average, first fixations lasted 164.98 milliseconds (SD = 154.67), with the first fixa-
tions on pictures being longer (M = 194.0, SD = 190.5) than those on texts (M = 129.9,
SD = 79.7). This is a significant difference according to Welch’s t-test, t(662.1) = 6.58,
p < .001, d = 0.44.7
Picture fixation count (PFC) and duration (PFD). The total PFC and PFD in all three pic-
ture attractiveness conditions are presented in Table 2. Picture attractiveness affected
PFC, F(2, 783.9) = 9.21, p < .001, ηp
2 = .008, and PFD, F(2, 784.0) = 10.99, p < .001,
ηp
2 = .009. Planned contrast analyses for both PFC and PFD showed that people tend
to fixate more and longer on attractive than on moderately attractive and unattractive
Table 1. Percentages of First Fixations on the Profile Picture or Text Posited on the
Profile’s Left or Right Side.
Picture position left (%) Picture position right (%) Total (%)
First fixation on picture 62.4 50.6 56.5
First fixation on text 37.6 49.4 43.5
Table 2. Mean Scores (SD) for Fixation Count and Duration on Profile Pictures for All
Three Picture Attractiveness Conditions.
Picture attractiveness
Attractive
(n = 278)
Moderately attractive
(n = 281)
Unattractive
(n = 273)
Picture fixation count 10.88 (12.01)a9.14 (12.03)b8.51 (10.51)b
Picture fixation duration (in sec) 2.55 (2.93)a2.22 (2.77)b1.91 (2.46)c
Note. Different superscripts in rows indicate significant differences between the levels of the picture
attractiveness condition.
van der Zanden et al. 875
pictures (all t’s > 2.66, p’s < .008, d’s > 0.210). Furthermore, while fixation durations
were longer on moderately attractive than on unattractive pictures, t(783.95) = 2.03,
p = .043, d = 0.118, participants did not look more at moderately attractive than at unat-
tractive pictures, t(783.90) = 0.76, p = .443. Robust support was thus found for Hypoth-
esis 2: people looked more often and longer at attractive than at moderately attractive
and unattractive pictures, and looked longer, but not more often, at moderately attrac-
tive than at unattractive pictures.8
Text condition did not affect the PFC and PFD and there were no interaction effects
(with all F’s < 2.71 and p’s > .100).
Text fixation count (TFC) and duration (TFD). The number of fixations and the total
fixation duration on the profile text did not differ depending on the attractiveness of
the picture (TFC: F(2, 784.1) = 0.945, p = .389, TFD: F(2, 784.1) = 0.643, p = .526).
This means that people do not pay more attention to the profile text when a profile con-
tains a moderately attractive picture than when it contains an attractive or unattractive
picture. Thus, H3 is not confirmed.
While picture attractiveness did not affect text attention, language errors did.
Participants fixated more and longer on texts with language errors than on those with-
out, with both F’s > 4.16, p’s < .042, ηp
2’s > .002. There were no interaction effects of
picture attractiveness and language errors on text attention (both F’s < 0.604,
p’s > .547). The mean scores of text attention are presented in Table 3.
Perception data
Picture attractiveness. Results showed main effects of picture attractive-
ness on perceived physical, F(2, 784.5) = 705.2, p < .001, ηp
2 = .545, social, F(2,
785.0) = 37.91, p < .001, ηp
2 = .072, and romantic attraction, F(2, 784.6) = 334.3,
p < .001, ηp
2 = .392. For all three attractiveness dimensions, profile owners with
attractive pictures received significantly higher scores than those with moderately
attractive pictures, and unattractive pictures got significantly lower scores still (with
all t’s > 1.98 and p’s < .048).
Table 3. Mean Scores (SD) for Fixation Count and Duration on Profile Texts in the Picture
Attractiveness and Language Errors Condition.
Language error
presence
Picture attractiveness
Attractive
Moderately
attractive Unattractive
Text fixation count No errors 49.47 (22.23)a49.22 (30.42)a48.79 (23.55)a
Errors 53.06 (26.00)b51.12 (26.45)b49.91 (23.13)b
Text fixation duration (in sec) No errors 8.38 (4.37)a8.38 (5.56)a8.35 (4.73)a
Errors 9.36 (5.17)b9.16 (5.17)b8.83 (4.85)b
Note. Different superscripts in columns indicate significant differences between the language error conditions within
each picture attractiveness condition.
876 Communication Research 49(6)
Language errors. Profile owners with language errors in their texts scored lower
on perceived social, F(1, 784.9) = 20.26, p < .001, ηp
2 = .020, and romantic attraction,
F(1, 784.5) = 24.27, p < .001, ηp
2 = .023, but not on perceived physical attraction, F(1,
784.4) = 2.03, p = .154. Table 4 presents for both conditions the mean scores on all
three dimensions of attraction.
Interaction. For perceived physical attraction a significant interaction was found,
F(2, 784.7) = 4.66, p = .010, ηp
2 = .007, but there was none for social nor for romantic
attraction, F(2, 784.4) = 0.217, p = .805 and F(2, 784.9) = 0.485, p = .616, respectively.
Simple effects analyses for physical attraction showed that when a profile contained a
moderately attractive picture, the given scores on physical attraction to profile owners
with and without language errors in their texts differ significantly with lower scores for
profiles with (M = 3.40, SD = 1.55) than for those without errors (M = 3.75, SD = 1.53),
F(1, 784.5) = 7.54, p = .006, while this is not the case for profiles with attractive or
unattractive pictures (with F’s < 2.29, and p’s > .131). This means H4 is partially sup-
ported: the expected interaction effect of picture attractiveness and language errors
was only found for perceived physical attraction.9
Exploratory Analyses
To further examine how people processed the multimodal dating profiles, we per-
formed additional exploratory analyses. First, we examined how often participants
switched between the picture and text during the profile viewing process, then looked
into the proportion of fixations on pictures and texts over the course of the process, and
finally, we zoomed in on the different profile processing strategies that could have
been used, and investigated whether the used processing strategy affected impression
formation differently.
Table 4. Mean Scores (SD) for Both Conditions on All Three Dimensions of Perceived
Attraction.
Language error
presence
Picture attractiveness
Attractive
Moderately
attractive Unattractive
Perceived physical attraction No errors 5.22 (1.48)a3.75 (1.53)a1.91 (1.09)a
Errors 5.41 (1.14)a3.40 (1.55)b1.76 (0.82)a
Perceived social attraction No errors 4.86 (1.24)a4.59 (1.16)a4.00 (1.37)a
Errors 4.51 (1.14)b4.18 (1.24)b3.71 (1.25)b
Perceived romantic attraction No errors 4.46 (1.62)a3.02 (1.50)a1.76 (1.01)a
Errors 3.98 (1.47)b2.57 (1.42)b1.47 (0.66)b
Note. Given scores on all three dependent variables differ significantly between levels of the picture
attractiveness condition. Different superscripts in columns indicate significant differences between the
two levels in the language error condition.
van der Zanden et al. 877
General viewing patterns. On average, participants switched 3.12 times (SD = 2.04)
between the picture and text on a profile, which is relatively low considering the aver-
age of sixty fixations on each profile in total. This suggests that the average profile
viewing process consists of four episodes of (in most cases a larger number of) fixa-
tions on the profile components, from which two are on the picture and two on the text.
The number of switches did not differ depending on the picture or text condition (all
conditions ranged between 3.02 and 3.22 switches).
For a general view on the processing order of the two components on the profile,
we divided each profile that was processed into 20 bins, each representing 5% of the
total number of profile fixations, and then aggregated that data across participants and
stimuli. The proportion of fixations on the picture and text was then calculated for each
consecutive bin.10 Figure 3 above shows that during the first 5% of the fixations on the
profile, picture processing is most likely to take place. Then, the proportion of text
fixations increases to almost full text attention. Toward the end of the viewing process
(the last 15–20%), there is again an increase in picture fixations, which could indicate
that fixations are directed to both components for integrative processing. Across all six
conditions, the viewing patterns were almost identical, with only minimal differences
in the proportion scores at the beginning and end of the viewing processes. Note that,
overall, the proportion of text fixations is much higher than that of picture fixations,
which is not surprising considering the relatively high number of fixations needed to
read a text compared to processing a picture. Figure 4 gives an example of the order of
fixations on the picture and text on a dating profile, corresponding with the general
viewing pattern described.
Different viewing strategies and perceived attraction. There are a number of potential
strategies participants could use to process the dating profiles. For each profile pro-
cess, we categorized whether a multi-switching strategy, a first-picture-then-text strat-
egy, or a first-text-then-picture strategy was used. The unit of analysis was each
individual profile that was processed, resulting in a total of 831 cases.
Figure 3. Plot of the proportion of fixations on the profile pictures and texts within bins
of 5%.
878 Communication Research 49(6)
The following criteria were used for categorization: cases with no switches at all
(n = 22; 2.6%) were not categorized because one of the two components did not receive
any attention. All cases with one to six switches were categorized as either first-pic-
ture-then-text or first-text-then-picture processing. Cases in which at least two of the
first three fixations were on the picture were categorized as picture-first (50.8%) and
cases with two or more of the first three fixations on the text as text-first (40.7%).
Furthermore, a case was categorized as multi-switching processing when seven or
more switches occurred on the profile (n = 47; 5.6%). This threshold of seven was
determined following previous eye tracking research (e.g., Bucher & Schumacher,
2006; Rayner et al., 2001) and empirically, by inspection of the data. We observed that
some people first looked briefly at both the picture and text (which could indicate
orientating), then attended longer to both components (deeper processing), and then
briefly fixated on both components again (merging). As such, seven switches would
indicate more switches than if people would look at both the picture and text in each
of the three phases.
To examine whether processing strategy led to different effects on perceived
attraction, we ran additional linear mixed effect models with processing strategy as
added fixed factor compared to the earlier analyses. The picture- and text-first strate-
gies were the two processing strategies that were considered in this factor. Only for
perceived romantic attraction a main effect of processing strategy was found, as well
as an interaction effect of processing strategy and text condition. In general, lower
romantic attraction scores were given when the text was processed first (M = 2.86,
SD = 1.71) than when the picture was processed first (M = 3.04, SD = 1.78), F(1,
754.8) = 6.21, p = .013. The interaction effect showed that processing strategy did
affect perceived romantic attraction for profiles without errors but not for profiles
Figure 4. Examples of a scan path of a profile where the picture (moderately attractive) is
processed before the text (without language errors).
Note. Circles indicate fixations, with larger circles indicating longer fixations. Consecutive numbers
indicate the viewing order.
van der Zanden et al. 879
with errors, F(1, 726.1) = 4.28, p = .039. For profiles without language errors, scores
on perceived romantic attraction were significantly higher when the picture was pro-
cessed first (M = 3.33, SD = 1.84) than when the text was processed first (M = 2.70,
SD = 1.66), F(1, 761.8) = 10.4, p = .001, while for profiles with errors there were no
differences between romantic attraction scores between picture-first (M = 2.74,
SD = 1.66) and text-first views (M = 2.53, SD = 1.52), F(1, 758.7) = 0.258, p = .612. No
main or two-way interaction effects of used processing strategy with picture and text
condition were found for physical and social attraction (all F’s < 1.91, p’s > .127).
Moreover, a significant three-way interaction effect of used processing strat-
egy, picture condition, and text condition on perceived physical attraction
(F(2, 722.1) = 5.02, p = .007) revealed that processing strategy only affected perceived
physical attraction for profiles with an attractive picture and without errors: for these
profiles, physical attraction scores were higher when the picture was processed first
(M = 4.70, SD = 1.57) than when the text was processed first (M = 4.04, SD = 1.59),
F(2, 735.8) = 8.70, p = .003. This suggests that when the picture is processed first,
attractive pictures have a stronger positive effect. In all other conditions, perceptions
of physical attraction were not affected by the order in which the profile components
were processed (with all other F’s < 2.96 and p’s > .086).
Overall, the exploratory analyses suggest a clear pattern in the viewing behavior by
participants: they look at both the picture and the text, often by focusing on the picture
first, then on the text. Switching repeatedly between the two components is rare.
Moreover, processing strategies remain relatively unaffected by the picture and text
manipulations, suggesting that the viewing patterns are robust. The used processing
strategy seems to have some influence on how (strong) the effects of the manipulations
on impressions formation scores are.
General Discussion
This study investigated how online dating profiles that contain pictures varying in
attractiveness and texts with or without language errors affect profile processing and
impressions of profile owner attractiveness. One goal of this study was to test the pic-
ture gatekeeper model, which proposes that the attractiveness of a profile picture is
key in the impression formation process. It was therefore assumed that pictures on
dating profiles would receive first attention and that the picture’s attractiveness would
determine how much attention people would pay to the profile text. To investigate this,
we collected both eye tracking and perception data.
Eye tracking results revealed that when people are presented with multimodal dat-
ing profiles containing a picture and a text, pictures are more likely to attract initial
attention (which is consistent with for example Scott & Hand, 2016, and Seidman &
Miller, 2013). This confirms H1 and is in line with the picture gatekeeper model.
Results also supported H2: the more attractive the picture, the more frequent and lon-
ger people look at the picture (in accordance with for example Leder et al., 2016, and
Valuch et al., 2015). In addition, people fixated more and longer on profile texts with
language errors than on those without errors (see also Rayner, 1998).
880 Communication Research 49(6)
While our results highlight the importance of the profile picture, they also indicate
that a picture does not necessarily function as gatekeeper to the rest of the profile.
Regardless of the picture’s attractiveness, the profile text attention was around 9 sec-
onds and 50 fixations, that is, around 80% of the total profile attention. Inconsistent
with H3, there was not more attention for the profile text when a picture was moder-
ately attractive than when it was attractive or unattractive. More specifically, texts on
profiles with attractive and unattractive pictures received more attention than was
originally expected.
In line with H4 and the picture gatekeeper model, language errors had a negative
effect on perceived physical attraction when a profile included a moderately attractive
picture but not when it contained an attractive or unattractive picture. This could indi-
cate that when picture information is not (yet) sufficient to form an impression about
physical attraction, textual cues carry greater weight. However, inconsistent with H4
and the picture gatekeeper model, these interaction effects of picture attractiveness and
language errors were not found for perceived social and romantic attraction. This sug-
gests that both picture and text attractiveness influence impressions about social and
romantic attraction, irrespective of the attractiveness of the profile component in the
other modality. People may thus use cues about picture and text attractiveness rela-
tively independently to form separate impressions about social and romantic attrac-
tion. These differential results imply that it is not just the heuristic of “what is beautiful
is good” that leads to impressions about attractiveness.
To get a better view on how multimodal dating profiles were processed, we con-
ducted exploratory analyses. General viewing patterns revealed that people are most
likely to process the picture prior to the text, with the first 5% of the profile fixations
mostly being on the picture. The absence of differences in general viewing patterns
and number of modality switches across picture and text conditions suggests that pro-
file cues have little impact on the general profile viewing process. At the same time,
both profile cues do affect impression formation scores, but there are few interaction
effects between the two types of cues. This seems to indicate that profile processing
and impression formation occur in two relatively independent stages. This is corrobo-
rated by the finding that what profile component was processed first had little effect on
further impression formation.
Implications
Our study has several theoretical implications. First, finding no interaction effect of
picture attractiveness on text attention reveals that people did not look longer at texts
of profiles with moderately attractive pictures than at profiles with attractive or unat-
tractive pictures. It seems that even when a profile picture was attractive or unattract-
ive, the initial impression based on pictorial information did not withhold people from
further text processing. Participants dedicated significant attention to all profile texts
which suggests they processed profile texts deliberately, irrespective of the picture’s
attractiveness. In fact, our results may even imply that neither the text nor the picture
was processed (solely) heuristically, because both the profile’s picture and text received
van der Zanden et al. 881
a considerable amount of attention, and cues in both components affected perceptions
of attraction.
Consequently, the main theoretical implication of this study is that our results are
not consistent with the picture gatekeeper model. Therefore, based on the patterns we
observe in our data, we put forward an alternative model of how pictures and texts on
multimodal dating profiles are processed and how the two profile components affect
impression formation. We dub this model the multimodal information processing
(MIP) model (see Figure 5). Based on the results of this study, we suggest that to
develop impressions about a dating profile owner, people take stock of the information
provided in each of the two modalities. They seem to do this consecutively, often start-
ing with the picture modality. This MIP model could be an important contributor to
theories about (multimodal) information processing and online impression formation
on three levels, which leads to three discernable stages in the model: the processing
stage, the cue assessment stage, and the impression formation stage.
In the first stage, people separately process the profile’s picture and text, in which
the picture is likely to be processed first. This accords with both H1 and the general
viewing pattern of the exploratory analyses. Picture and text processing seem to occur
as two relatively isolated events—with little switching between the components (see
also Carroll et al., 1992; Rayner et al., 2001). This is potentially in line with extant
research indicating distinct picture and text processing (e.g., Barry, 1997; Paivio,
1990; Powell et al., 2019). Paivio (1990) proposed earlier in his dual coding theory
that pictorial and textual information is processed in different subsystems in working
memory, resulting in a parallel construction of two separate mental models. Prior to
the construction of these separate mental models, no interplay is assumed to take place
(Eitel et al., 2013).
In the second stage, people assess the different pictorial and textual cues that are
available on the profile. In the specific case of this study, this results in separate main
effects of the picture (attractiveness) and text (language error presence) manipulations.
At this stage, perceptions are attributed to profile cues, such as about the attractiveness
of the pictorial and textual cues. As no interaction effects of picture and text attractive-
ness were obtained on the processing of these two components—suggesting relatively
independent picture and text processing—it is most likely that pictorial and textual
cues also lead to separate assessments about the attractiveness of these cues.
In the third and final stage, perceptions that are developed about picture and text
attractiveness are used to form impressions about profile owners and their attractive-
ness. Our data suggest that separate impressions are formed for each dimension of
perceived attraction instead of one overall impression. This finding deviates from pre-
vious research that assumes that people favor to form one general impression rather
than evaluating someone on each attribute to form different impressions (cf. Kahneman,
2011; Willis & Todorov, 2006). We found that picture attractiveness primarily affects
perceived physical attraction, and to a lesser extent social and romantic attraction,
while text attractiveness affects social and romantic but not physical attraction. The
finding that pictorial and textual cues differ in impact depending on the specific aspect
of attraction that is assessed draws on previous findings of Van der Zanden et al.
882 Communication Research 49(6)
(2020), who posed that interpersonal attraction may not always be considered a unidi-
mensional construct. The arrows in the model in Figure 5 represent the effects of pic-
ture and text attractiveness on each dimension of attraction, with the size of the effects
being indicated by the arrows’ thickness.
Figure 5. Schematic representation of the processing flow in the multimodal information
processing (MIP) model.
van der Zanden et al. 883
This study also yields practical implications for online daters. Given our results,
online daters are recommended to invest time and effort in both the picture and text of
their profile. Profile pictures and texts are not only both likely to receive attention, but
online daters should also take into account that both pictorial and textual profile cues
affect perceived attraction: attractive pictures positively affect perceived attraction,
but regardless of the picture, language errors have a negative effect. The latter finding
highlights that online daters should try to avoid language errors in their profiles.
Another reason why daters should consider both profile components is because people
seem to use pictorial information primarily for impressions about physical attraction,
and to a lesser extent social and romantic attraction, whereas language errors most
heavily impact impressions of social attraction, which is mostly concerned with per-
ceptions of a profile owner’s personality. This implies that specific impressions are
formed on the basis of different pieces of information on the profile.
Finally, our results reveal practical insights for designers of online dating platforms.
Our finding that pictures are more likely to capture initial attention when they occur
on the left rather than the profile’s right side may suggest that the way dating platforms
organize profiles could influence (to some extent) how their members view profiles of
others. Most dating platforms seem to anticipate on the relevance of picture attractive-
ness by placing the picture on the position of the profile where it appears most promi-
nent and almost certainly receives initial attention, that is, either at the top of the
profile (e.g., on applications like Tinder) or on the profile’s left side (e.g., on web-
based sites like eHarmony). There are however dating platforms that want to place less
emphasis on initial picture attention and more emphasis on profile texts or other (tex-
tual) information in the initial impression formation phase. For these platforms, it
could be relevant to experiment with the position of the profile picture and text as to
enhance initial text attention.
Directions for Future Research
Our data seem to fit the MIP model better than the picture gatekeeper model. However,
the independent processing of cues in different modalities and the extent to which they
lead to separate assessments of attraction should be investigated further, for example
by comparing cue assessments when presenting a one-modality profile with a profile
that includes that and another component (i.e., two modalities). Moreover, future
research could attempt to further disentangle the second and third stage of the MIP
model; for instance by collecting assessments regarding the picture, the text, and the
profile owner’s attractiveness separately, as based on our data we cannot test whether
cue attractiveness assessments occur prior to the impressions formed about the person
“behind” the profile. New (preregistered) studies should thus be conducted to further
examine the model’s viability and the different stages within the model.
The separate, independent effects of the profile cues on the different dimensions of
perceived attraction raise the question how and when the specific aspects of impression
formation eventually integrate into one final decision. In the case of online dating, it
could be that one eventually integrates the different attractiveness impressions—formed
884 Communication Research 49(6)
on the basis of different profile cues—to come to a final decision about whether to
pursue contact with the profile owner. While in our study the measure of perceived
romantic attraction may have been an indication of romantic interest, this was not a
statement that inquires actual interest in sending a message to or going on a date with
the profile owner. To investigate the relative weight of the different impressions, it
would be interesting for future work to also test when people integrate the impressions
to one decision, for example by providing them the option to like or dislike a profile
after assessing it.
The profile setup that was used for this study matched with how daters on various
large profile-based sites (e.g., Relatieplanet, eHarmony) encounter a profile for the
first time, that is, with a single picture and a short self-description. On such sites, dat-
ers often decide on the basis of this information whether they want to find out more
about the profile owner, which they can do by clicking on the profile. Usually, this
involves not only more demographic and written answers to questions, but also more
pictures. A next step would be to investigate whether comparable viewing patterns
would occur when people view the full profile, that is, by looking at all pictures prior
to processing other textual information.
The immediate and simultaneous availability of the picture and text on the profiles
in this study, may have influenced participants’ perceived importance of both profile
components for impression formation. This may have resulted in more attention to the
profile and profile cues than may have been given in a real-life dating setting, where
immediately accepting or discarding a profile might occur. Especially on photo-based
dating platforms, such as Tinder, pictures are primarily used for impression formation
and deciding whether to swipe left (no interest) or right (interest). Although in these
contexts picture attractiveness may be used as the profile’s gatekeeper, our MIP model
may still hold as people may make the decision on an incomplete impression based
only on picture attractiveness and the associated physical attraction perception, result-
ing in little or no consideration of other impression formation dimensions (e.g., social
attraction). This, however, should be tested in a future experiment.
The aim of this study was to gain insight in the impression formation process by
collecting both eye tracking data and perception data. While viewing dating profiles,
eye movement data can reveal in a precise and objective manner (unconscious) cogni-
tive processes and preferences of people, such as what parts of a profile have been
attended to, when, how long and often, and in what order. The data can however not
tell us why people did—or did not—attend to these different components (Holsanova,
2014). Based on our results, it is for example difficult to ascertain participants’ ratio-
nale to look at all profile texts. It could be that picture attractiveness did not immedi-
ately lead to an overall impression which made participants shift their attention to
profile texts, or participants formed overall impressions based on picture attractive-
ness but tried to confirm these initial impressions by paying attention to profile texts.
Interviews or thinking-aloud methods in which participants try to verbalize the ratio-
nales behind their profile viewing behavior could be a way to further extend our under-
standing of the online impression formation process.
van der Zanden et al. 885
To conclude, our study is the first that used eye tracking to investigate people’s
impression formation processes while looking at multimodal online dating profiles.
Our results seem to indicate that picture attractiveness and language errors mostly
separately affect picture and text processing, which could result in distinct impression
formation processes. In general, this research shows that most profile attention is
devoted to texts but that pictures have the strongest impact on impression formation.
Even though picture attractiveness is highly determining, it does not seem to affect
text processing and the resulting effects on impression formation. Thus, pictures and
texts are likely to be processed relatively independently and lead to separate impres-
sions on different aspects of perceived attraction.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship,
and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this
article.
ORCID iD
Tess van der Zanden https://orcid.org/0000-0002-4392-2617
Notes
1. Alternatively, one could also argue that people are interested to learn more about attractive
individuals (Garcia et al., 1991; Langlois et al., 2000), which would lead to the prediction
that people would pay closer attention to texts of profile owners with attractive pictures
compared to less attractive pictures, as a way to confirm the positive initial impression and
increase reward feelings. However, even though this is ultimately an empirical question,
we believe this is in fact unlikely to happen in the context of online dating considering the
decisive role of physical and picture attractiveness in this setting (Fiore et al., 2008) com-
bined with people’s tendency to minimize (processing) efforts (e.g., Kahneman, 2011).
2. H3 presented here in the manuscript deviates in two ways from what we stated in our
preregistration. First, the category that was called average attractive in the preregistration
is called moderately attractive in the manuscript. Second, we said that we would look into
the relative attention to the profile text as opposed to the fixation count and duration in
absolute numbers. Before collecting the data, we took into account potential differences
between relative and absolute measures, but during the analyses we found that results were
actually very similar for both measures. We decided to primarily report the results of abso-
lute picture and text attention to retain consistency in the measures reported on to test the
eye tracking related hypotheses. Full analyses of all measures can be found at osf.io/cbtv2/.
3. Participants were asked to answer two other impression formation statements regarding
perceived intelligence (based on Leach et al., 2007) and attributional confidence (based
on Clatterbuck, 1979). Only two main effects of picture attractiveness and language errors
on perceived intelligence were found, with both F’s > 22.91 and p’s < .001. Profile owners
886 Communication Research 49(6)
with an unattractive picture were rated as significantly less intelligent than profile owners
with attractive and moderately attractive pictures, while profile owners with attractive and
moderately attractive pictures were rated as similarly intelligent. Moreover, profile owners
with language errors in their profiles scored lower on perceived intelligence than profile
owners without errors. No main or interaction effects of picture and text attractiveness
were found on attributional confidence (with all F’s < 1.11 and p’s > .33). As our hypoth-
eses did not focus on perceived intelligence and attributional confidence, results on these
variables were not integrated in the manuscript.
4. By not clustering the three dimensions of attraction, we deviate from what is stated in the
preregistration. Our reason to look into the dimensions separately is that different dimen-
sions of perceived attraction may not always been seen as one unidimensional construct, in
particular, when rating dating profiles with language errors (Van der Zanden et al., 2020).
To get a better insight into whether and how different profile components may be used to
form impressions on different dimensions of attraction, we decided to not cluster them. The
Cronbach’s alpha of the three items was .82.
5. In our preregistration, it was stated that we would perform MANOVA’s to test our hypothe-
ses. Upon data collection, we recognized this test would not be optimal to analyze our data,
as we could then not control for individual differences (Valuch et al., 2015). We therefore
decided to deviate from what was posed in the preregistration and employed linear mixed
effect models.
6. For H2, H3, and H4, the same models with random by-participant slopes for picture and
text condition were included in an additional model. Results with and without these slopes
were similar, and we therefore reported those without. Also adding picture position or pic-
ture or text version did not alter the obtained findings and did not lead to better performing
models of H2, H3, and H4. Therefore, the results reported are those with only participants
as random intercept, and picture and text condition as fixed factors.
7. Previous eye tracking studies differ in what is counted as a fixation. Previously, fixations
of 40 (e.g., Scott & Hand, 2016) and 100 milliseconds (e.g., Leder et al., 2016) have for
example been taken as a cut-off point. In this study, each fixation that lasted 40ms or longer
counted as a fixation. From all first fixations on the picture or text, 33.2% was between 50
and 99.9 ms. The same chi-square tests with only first fixations of at least 100 ms included
provided similar results as what is reported now with first fixations of 40 ms or longer,
χ2 (1) = 11.23, p < .001 (30.5% on the left-sided picture, 24.9% on the right-sided picture,
19.3% on the left-sided text, 25.3% on the right-sided text).
8. Results reported on the effects of picture attractiveness on PFC and PFD are those in which
only picture attractiveness was added as a factor in the model. Similar results were found
when the language errors condition was added to the model as additional factor.
9. Interaction effects of picture attractiveness and gender were found on perceived physical
and romantic attraction, with both F’s > 4.10 and p’s < .024. These showed that men and
women differ more in their ratings given to profiles with attractive and moderately attrac-
tive pictures than in their ratings to profiles with unattractive pictures. This seems to indi-
cate that pictures that are more attractive have a stronger positive influence on men’s than
on women’s ratings.
10. In this manuscript, we present the results of the proportion of fixations on the picture and
text with bins of 5% for each processed profile. We decided to go for bins of 5% as almost
each profile received at least 20 fixations (n = 769) and this figure could thus provide the
most detailed viewing pattern. We found comparable viewing patterns when we used bins
of 10% or 20%.
van der Zanden et al. 887
References
Bar, M., Neta, M., & Linz, H. (2006). Very first impressions. Emotion, 6, 269–278. https://doi.
org/10.1037/1528-3542.6.2.269
Barry, A. M. (1997). Visual intelligence: Perception, image, and manipulation in visual com-
munication. State University of New York Press.
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2014). lme4: Linear mixed-effects models
using Eigen and S4 (R Package Version 1.1-6). http://cran.r-project.org/package=lme4
Bucher, H. J., & Schumacher, P. (2006). The relevance of attention for selecting news content.
An eye-tracking study on attention patterns in the reception of print and online media.
Communications, 31, 347–368. https://doi.org/10.1515/COMMUN.2006.022
Campbell, W. K. (1999). Narcissism and romantic attraction. Journal of Personality and Social
Psychology, 77, 1254–1270. https://doi.org/0022-3514/99/53.00
Carroll, P. J., Young, J. R., & Guertin, M. S. (1992). Visual analysis of cartoons: A view from
the far side. In K. Rayner (Ed.), Eye movements and visual cognition: Scene perception and
reading (pp. 444–461). Springer.
Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source
versus message cues in persuasion. Journal of Personality and Social Psychology, 39, 752–
766. https://doi.org/10.1037/0022-3514.39.5.752
Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna, J. M. Olson, & C. P.
Herman (Eds.), Social influence: The Ontario symposium (Vol. 5, pp. 3–39). Lawrence
Erlbaum Associates.
Chaiken, S., & Eagly, A. H. (1976). Communication modality as a determinant of message per-
suasiveness and message comprehensibility. Journal of Personality and Social Psychology,
34, 605–614. https://doi.org/10.1037/0022-3514.34.4.605
Clatterbuck, G. W. (1979). Attributional confidence and uncertainty in initial interaction.
Human Communication Research, 5, 147–157. https://doi.org/10.1111/j.1468-2958.1979.
tb00630.x
D’Angelo, J., Zhang, C., Eickhoff, J., & Moreno, M. (2014). Facebook influence among incom-
ing college freshmen: Sticky cues and alcohol. Bulletin of Science, Technology & Society,
34, 13–20. https://doi.org/10.1177/0270467614538002
Dion, K., Berscheid, E., & Walster, E. (1972). What is beautiful is good. Journal of Personality
and Social Psychology, 24, 285–290. https://doi.org/10.1037/h0033731
Eitel, A., Scheiter, K., & Schüler, A. (2013). How inspecting a picture affects processing of
text in multimedia learning. Applied Cognitive Psychology, 27, 451–461. https://doi.
org/10.1002/acp.2922
Fiore, A. T., Taylor, L. S., Mendelsohn, G. A., & Hearst, M. (2008, April 5–10). Assessing
attractiveness in online dating profiles [Conference session]. SIGCHI Conference on
Human Factors in Computing Systems, Florence, Italy.
Garcia, S., Stinson, L., Ickes, W., Bissonnette, V., & Briggs, S. R. (1991). Shyness and physi-
cal attractiveness in mixed-sex dyads. Journal of Personality and Social Psychology, 61,
35–49. https://doi.org/10.1037/0022-3514.61.1.35
Hitsch, G. J., Hortaçsu, A., & Ariely, D. (2010). What makes you click?—Mate preferences in
online dating. Quantitative Marketing and Economics, 8, 393–427. https://doi.org/10.1007/
s11129-010-9088-6
Holsanova, J. (2014). 20. Reception of multimodality: Applying eye tracking methodology in
multimodal research. In C. Jewitt (Ed.), Routledge handbook of multimodal analysis (2nd
ed., pp. 285–296). Routledge.
888 Communication Research 49(6)
Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus, and Giroux.
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2014). ImerTest: Tests for random
and fixed effects for linear mixed effect models (Imer Objects of lme4 Package). R Package
Version 2.0-6. http://cran.r-project.org/package=lmerTest
Langlois, J. H., Kalakanis, L., Rubenstein, A. J., Larson, A., Hallam, M., & Smoot, M. (2000).
Maxims or myths of beauty? A meta-analytic and theoretical review. Psychological
Bulletin, 126, 390–423. https://doi.org/10.1037//0033-2909.126.3.390
Leach, C. W., Ellemers, N., & Barreto, M. (2007). Group virtue: The importance of morality (vs.
competence and sociability) in the positive evaluation of in-groups. Journal of Personality
and Social Psychology, 93, 234–249. https://doi.org/10.1037/0022-3514.93.2.234
Leder, H., Mitrovic, A., & Goller, J. (2016). How beauty determines gaze! Facial attractiveness
and gaze duration in images of real world scenes. i-Perception, 7(4), 2041669516664355.
Leder, H., Tinio, P. P., Fuchs, I. M., & Bohrn, I. (2010). When attractiveness demands longer
looks: The effects of situation and gender. Quarterly Journal of Experimental Psychology,
63, 1858–1871.
Lo, S. K., Hsieh, A. Y., & Chiu, Y. P. (2013). Contradictory deceptive behavior in online dating.
Computers in Human Behavior, 29, 1755–1762. https://doi.org/10.1016/j.chb.2013.02.010
Locher, P., Unger, R., Sociedade, P., & Wahl, J. (1993). At first glance: Accessibility of
the physical attractiveness stereotype. Sex Roles, 28, 729–743. https://doi.org/10.1007/
BF00289990
Maner, J. K., Kenrick, D. T., Becker, D. V., Delton, A. W., Hofer, B., Wilbur, C. J., & Neuberg,
S. L. (2003). Sexually selective cognition: Beauty captures the mind of the beholder. Journal
of Personality and Social Psychology, 85, 1107–1120. https://doi.org/10.1037/0022-
3514.85.6.1107
McCroskey, J. C., & McCain, T. A. (1974). The measurement of interpersonal attraction. Speech
Monograph, 41, 261–266. https://doi.org/10.1080/03637757409375845
McGloin, R., & Denes, A. (2018). Too hot to trust: Examining the relationship between attrac-
tiveness, trustworthiness, and desire to date in online dating. New Media & Society, 20,
919–936. https://doi.org/10.1177/1461444816675440
Miller, A. G. (1970). Role of physical attractiveness in impression formation. Psychonomic
Science, 19, 241–243. https://doi.org/10.3758/BF03328797
Mitrovic, A., Tinio, P. P., & Leder, H. (2016). Consequences of beauty: Effects of rater
sex and sexual orientation on the visual exploration and evaluation of attractiveness in
real world scenes. Frontiers in Human Neuroscience, 10, 122. https://doi.org/10.3389/
fnhum.2016.00122
Olson, I. R., & Marshuetz, C. (2005). Facial attractiveness is appraised at a glance. Emotion, 5,
186–201. https://doi.org/10.1037/1528-3542.5.4.498
Paivio, A. (1990). Mental representations: A dual coding approach (Vol. 9). Oxford University
Press.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In
L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 19, pp. 123–205).
Academic Press.
Powell, T. E., Boomgaarden, H. G., De Swert, K., & de Vreese, C. H. (2019). Framing fast and
slow: A dual processing account of multimodal framing effects. Media Psychology, 22,
572–600. https://doi.org/10.1080/15213269.2018.1476891
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124(3), 372–422. https://doi.org/10.1037/0033-2909.124.3.372
van der Zanden et al. 889
Rayner, K., Rotello, C. M., Stewart, A. J., Keir, J., & Duffy, S. A. (2001). Integrating text and
pictorial information: Eye movements when looking at print advertisements. Journal of
Experimental Psychology: Applied, 7, 219–226. https://doi.org/10.1037/1076-898X.7.3.219
Scott, G. G., & Hand, C. J. (2016). Motivation determines Facebook viewing strategy: An eye
movement analysis. Computers in Human Behavior, 56, 267–280. https://doi.org/10.1016/j.
chb.2015.11.029
Seidman, G., & Miller, O. S. (2013). Effects of gender and physical attractiveness on visual
attention to Facebook profiles. Cyberpsychology, Behavior, and Social Networking, 16,
20–24. https://doi.org/10.1089/cyber.2012.0305
Sritharan, R., Heilpern, K., Wilbur, C. J., & Gawronski, B. (2010). I think I like you: Spontaneous
and deliberate evaluations of potential romantic partners in an online dating context.
European Journal of Social Psychology, 40, 1062–1077. https://doi.org/10.1002/ejsp.703
Tidwell, L. C., & Walther, J. B. (2002). Computer-mediated communication effects on disclo-
sure, impressions, and interpersonal evaluations: Getting to know one another a bit at a time.
Human Communication Research, 28, 317–348. https://doi.org/10.1111/j.1468-2958.2002.
tb00811.x
Toma, C. L., & Hancock, J. T. (2010). Looks and lies: The role of physical attractiveness in
online dating self-presentation and deception. Communication Research, 37, 335–351.
https://doi.org/10.1177/0093650209356437
Valuch, C., Pflüger, L. S., Wallner, B., Laeng, B., & Ansorge, U. (2015). Using eye tracking to
test for individual differences in attention to attractive faces. Frontiers in Psychology, 6, 42.
https://doi.org/10.3389/fpsyg.2015.00042
Van der Heide, B., D’Angelo, J. D., & Schumaker, E. M. (2012). The effects of verbal ver-
sus photographic self-presentation on impression formation in Facebook. Journal of
Communication, 62, 98–116. https://doi.org/10.1111/j.1460-2466.2011.01617.x
Van der Zanden, T., Mos, M., & Schouten, A. (2018). Taalaccommodatie in online dating-
profielen: Effecten van opleidingsniveau en type datingsite op taalgebruik [Language
accommodation in online dating profiles: Effects of education level and type of dating
site on language use]. Tijdschrift voor Taalbeheersing, 40, 83–106. https://doi.org/10.5117/
TVT2018.1.zand
Van der Zanden, T., Schouten, A. P., Mos, M. B., & Krahmer, E. J. (2020). Impression forma-
tion on online dating sites: Effects of language errors in profile texts on perceptions of
profile owners’ attractiveness. Journal of Social and Personal Relationships, 37, 758–778.
https://doi.org/10.1177/0265407519878787
Van Straaten, I., Holland, R. W., Finkenauer, C., Hollenstein, T., & Engels, R. C. (2010).
Gazing behavior during mixed-sex interactions: Sex and attractiveness effects. Archives of
Sexual Behavior, 39, 1055–1062. https://doi.org/10.1007/s10508-009-9482-x
Walster, E., Aronson, V., Abrahams, D., & Rottman, L. (1966). Importance of physical attrac-
tiveness in dating behavior. Journal of Personality and Social Psychology, 4(5), 508–516.
Walster, E., Berscheid, E., & Walster, G. W. (1973). New directions in equity research. Journal
of Personality and Social Psychology, 25, 151–176. https://doi.org/10.1037/h0033967
Willis, J., & Todorov, A. (2006). First impressions: Making up your mind after a 100-ms
exposure to a face. Psychological Science, 17, 592–598. https://doi.org/10.1111/j.1467-
9280.2006.01750.x
Wotipka, C. D., & High, A. C. (2016). An idealized self or the real me? Predicting attraction
to online dating profiles using selective self-presentation and warranting. Communication
Monographs, 83, 281–302. https://doi.org/10.1080/03637751.2016.1198041
890 Communication Research 49(6)
Author Biographies
Tess van der Zanden is a PhD candidate at Tilburg University at the Department of
Communication and Cognition. Her research focuses on language use in online dating.
Maria B. J. Mos is an assistant professor at Tilburg University at the Department of
Communication and Cognition.
Alexander P. Schouten is an assistant professor of Business Communication and Digital Media
at Tilburg University at the Department of Communication and Cognition.
Emiel J. Krahmer is a full professor of Communication, Cognition and Computation at Tilburg
University and Head of the Department Communication and Cognition.
Available via license: CC BY 4.0
Content may be subject to copyright.