ArticlePDF Available

Abstract and Figures

Would we think more negatively of a person who caused rather than produced an outcome or who is described as utterly rather than totally unconventional? While these word choices may appear to be trivial, cause and utterly occur more frequently in a negative context in natural language use than produced or totally, even though these words do not have an explicit valenced meaning. Words that are primarily used in a valenced context are said to have semantic prosody. Five studies show that semantically-prosodic descriptors affect the impressions formed of others. These effects occur even in situations where perceivers are likely to be skeptical of messages, and they impact behavioral intentions toward targets. An utterly changed person was perceived as less warm and competent than a totally changed person (Study 1), and people held more negative impressions of an utterly rather than totally unconventional boss (Study 2). People had stronger intentions to vote for a political candidate who produced budget changes over one who caused them (Study 3) and preferred a bank that lends money (a word with positive semantic prosody) over a bank that loans money (Study 4). Finally, participants had more (less) romantic interest in potential dating partners with Tinder profiles that utilized words with positive (negative) semantic prosody (Study 5). We conclude that semantically prosodic descriptors that lack a clear positive or negative meaning still lead people to infer the valence of what is to come, which colors the impressions they form of others.
Content may be subject to copyright.
Running head: SEMANTIC PROSODY AND IMPRESSION FORMATION 1
How seemingly innocuous words can bias judgment: Semantic prosody and impression
formation
David J. Hausera and Norbert Schwarzb
a Corresponding author; University of Southern California, USC Dornsife Mind & Society
Center, 205 VPD, 635 Downey Way, Los Angeles, CA 90089-3331. davidhau@usc.edu; 717-
965-7816
b University of Southern California, USC Dornsife Mind & Society Center, 205 VPD, 635
Downey Way, Los Angeles, CA 90089-3331; norbert.schwarz@usc.edu
This is a preprint of a paper to appear in the Journal of Experimental Social Psychology
Word count: 7,489
Author notes:
We wish to thank Isabel Saville and Jenna Manske for braving the world of Tinder in order to
help us develop the materials for Study 5 and the PR lab for their helpful comments. This
research did not receive any specific grant from funding agencies in the public, commercial, or
not-for-profit sectors.
SEMANTIC PROSODY AND IMPRESSION FORMATION 2
Abstract
Would we think more negatively of a person who caused rather than produced an outcome or
who is described as utterly rather than totally unconventional? While these word choices may
appear to be trivial, cause and utterly occur more frequently in a negative context in natural
language use than produced or totally, even though these words do not have an explicit valenced
meaning. Words that are primarily used in a valenced context are said to have semantic prosody.
Five studies show that semantically-prosodic descriptors affect the impressions formed of others.
These effects occur even in situations where perceivers are likely to be skeptical of messages,
and they impact behavioral intentions toward targets. An utterly changed person was perceived
as less warm and competent than a totally changed person (Study 1), and people held more
negative impressions of an utterly rather than totally unconventional boss (Study 2). People had
stronger intentions to vote for a political candidate who produced budget changes over one who
caused them (Study 3) and preferred a bank that lends money (a word with positive semantic
prosody) over a bank that loans money (Study 4). Finally, participants had more (less) romantic
interest in potential dating partners with Tinder profiles that utilized words with positive
(negative) semantic prosody (Study 5). We conclude that semantically prosodic descriptors that
lack a clear positive or negative meaning still lead people to infer the valence of what is to come,
which colors the impressions they form of others.
Keywords: semantic prosody, impression formation, social cognition, judgment, language
<246/250 words>
SEMANTIC PROSODY AND IMPRESSION FORMATION 3
How seemingly innocuous words can bias judgment: Semantic prosody and impression
formation
If we read that Susan is friendly and that Bob is aggressive, we will form a more positive
impression of Susan than of Bob. Person descriptors such as these have a clear positive or
negative valence and are seen as similar to other words of similar valence (Bradley & Lang,
1999, Warriner, Kuperman, & Brysbaert, 2013; Osgood, Suci, & Tannenbaum, 1957). Such
person descriptors guide interpretation and influence evaluative impressions of others (Higgins,
Rholes, & Jones, 1977; Srull & Wyer, 1979; Kelley, 1973).
However, if we read that Susan lends something and Bob causes an outcome, would we
similarly arrive at a more positive impression of Susan than of Bob? These seemingly innocuous
words have no clear valence in themselves, independent of what is being lent or caused. Do they
nevertheless have the power to shift our impressions of these actors? This is what the current
research assesses.
Language and person impressions
Social cognition has long examined how descriptors impact person perception. Generally,
positive/negative person descriptors result in positive/negative impressions (D’Andrade, 1974;
Kelley, 1973; Fiske & Neuberg, 1990; for a review of descriptors in stereotyping, see Hamilton
& Sherman, 1989). For instance, Asch (1946) found that inserting the positive trait warm (vs the
negative trait cold) into a personality description resulted in more positive person impressions
and more positive interpretations of other traits. Person impressions are also affected when
person descriptors are rendered accessible in prior unrelated tasks (Higgins, Rholes, & Jones,
1977). For instance, prior activation of the positive trait assertive (vs the negative trait
aggressive) in a sentence-unscrambling task created more positive impressions of a character
named Donald who performs many ambiguous behaviors (such as refusing to pay his rent until
his landlord repaints his apartment; Srull & Wyer, 1979). The meaning behind person descriptors
impacts the impressions we form. In such cases, the descriptor has an explicitly valenced
meaning that is widely shared among language users and usually is part of the lexical entry for
the term.
However, word meaning has social components that go beyond what finds entry into the
lexicon. The meaning of words and constructions is learned from interactions with other
language speakers and is socially-constructed from how that word is used (Ellis, O’Donnell, &
Romer, 2015; Kilgarriff, 1997; Hoey, 2005). Words that serve as person descriptors may
therefore have meanings that are not entirely obvious, and these seemingly innocuous words may
subtly influence impressions of other persons.
Semantic prosody
While many character traits are clearly positive or negative, words such as lend or cause
have affective attributes that are less clear and explicit. These words have semantic prosody,
which describes that they occur in a valenced context in natural language use, even though the
words themselves do not have an inherent valenced meaning (Sinclair, 1991). For instance, while
most people see cause as neutral in valence, the things that are caused within everyday discourse
SEMANTIC PROSODY AND IMPRESSION FORMATION 4
are overwhelmingly negative. The most frequent noun collocates within four words to the right
of cause are death, problems, damage, pain, cancer, trouble, concern, disease, effect, harm
(Stubbs, 1995; Davies, 2008). And while most people see lend as something neutral, the most
common things that are lent are positive in valence. The most frequent noun collocates within
four words to the right of lend are money, support, hand, credence, credibility, name, air,
legitimacy, institutions, voice (Davies, 2008). While comprehensive lists of words with semantic
prosody are missing, some words with semantic prosody have been documented in numerous
languages (Tognini-Bonelli, 2001; Xiao & McEnery, 2006).
A word’s context is a key part of its representation. Context creates meaning for concepts
that goes beyond lexical definitions and even explicit knowledge (Casasanto & Lupyan, 2015;
Elman, 2011; Hoey, 2005). Hence, a word’s co-occurrence with a predominantly positive or
negative context may foster subtle associative meaning. This may serve adaptive purposes, such
as assisting with reading comprehension by helping people predict the valence of adjacent
concepts (Hoey, 2005). But it may also color such words with subtle affective tones (Sinclair,
1991; Louw, 1993), which may, in turn, guide global evaluations of other persons.
While semantic prosody may seem obvious to those concerned with how words are
typically used (such as learners and teachers of second languages, Xiao & McEnery, 2006), it
may seem less obvious to casual language users (Sinclair, 1991; Stubbs, 1995). Empirical
evidence of the effects of semantic prosody on inferences is mostly lacking, but initial evidence
suggests that semantic prosody can guide evaluative inferences (Hauser & Schwarz, 2016).
Although participants see some words with semantic prosody (such as cause) as being neutral in
valence and synonymous with non-semantically prosodic words (like produce), they nevertheless
infer that caused outcomes are more negative than produced outcomes over a variety of domains.
For example, participants are more likely to infer that “endocrination of abdominal lipid tissue,”
a fictional medical outcome, is negative when it is caused rather than produced (Hauser &
Schwarz, 2016, Studies 1a and 1b). Because words with semantic prosody lead people to infer
the valence of what is to come, these words color ambiguous concepts with evaluative meaning.
Semantic prosody and person impressions
While prior research has shown that semantic prosody affects evaluative judgments, the
extent of such effects is relatively unknown. To our knowledge, our prior work (Hauser &
Schwarz, 2016) is the first investigation into the effects of semantic prosody on judgments. This
prior work established that semantically-prosodic words can influence the evaluation of
ambiguous events, but left many questions regarding the underlying processes and the robustness
of the phenomenon unanswered.
For instance, some might wonder whether semantically-prosodic terms influence
judgment because of norms of everyday discourse, described by Grice (1975). A maxim of
relevance stipulates that communicators provide only information that is relevant to the
conversation and listeners interpret utterances accordingly. In research settings, this maxim can
render minor aspects of question wording or scale design “relevant” in unanticipated ways (for
reviews, see Schwarz, 1994, 1996). In our prior studies, we provided participants with sentences
SEMANTIC PROSODY AND IMPRESSION FORMATION 5
that did or did not contain semantically-prosodic words and asked them to make evaluative
judgments about elements of the sentence. This renders the particular elements highly salient and
participants may have attended to, and reacted to, semantically-prosodic words because the norm
of relevance dictates that they do so. If so, semantically-prosodic words may have little influence
when they are less focal.
Person impressions offer an avenue for addressing this issue. People are often hesitant to
make judgments about others in experimental contexts when they feel they lack diagnostic
information (Yzerbyt, Schadron, Leyens, & Rocher, 1994). Because semantically-prosodic
words, such as the word cause, lack clear valenced meanings (Sinclair, 1991; Stubbs, 1995), they
should not affect evaluation unless they bring valenced information to mind. If such words do
have subtle social influence, they should color evaluations of other persons and potentially even
affect behavioral intentions relevant to person impression.
Five studies suggest that semantically prosodic descriptors have subtle social influence.
Pilot studies and previous research identify words with semantic prosody that are seen as neutral
in valence and similar to synonyms with no semantic prosody. Five experiments show that
impressions of other persons and entities are more negative (positive) when they are described by
words with negative (positive) semantic prosody. These effects on global evaluative impressions
appear for generic others (Study 1), managers (Study 2), political candidates (Study 3), brands
(Study 4), and even potential dating partners on Tinder (Study 5). Semantically prosodic
descriptors also affect behavioral intentions (Studies 3 and 5) and affect impressions of persons
in situations where perceivers are likely skeptical of the information (Studies 4 and 5). Overall,
semantic prosody exhibits generalized effects across different words (nine stimuli words in
total), parts of speech (adverbs in Studies 1 & 2 and verbs in Studies 3, 4, and 5), and valence
(positive and negative semantic prosody). Semantically prosodic descriptors that lack a clear
positive or negative meaning still lead people to infer the valence of what is to come, which
colors the impressions they form of others.
Methodological overview
All study protocols were determined to be exempt from review by the University of
Michigan Institutional Review Board (IRB). Respondents provided informed consent at the start
of each study and were debriefed upon completion. We report all measures, manipulations, and
exclusions, and studies that were run in this line of research. Sample sizes were determined prior
to data collection as detailed in each study. Data and materials for all studies can be found at
https://osf.io/f2atn. Study 1
In Study 1, we investigate whether a semantically-prosodic adverb can affect global
evaluative impressions of a target person. Participants read a description of an unknown man that
described him as being utterly or totally changed. Pilot testing established that participants see
the adverbs utterly and totally as being synonymous. However, the adverb utterly has negative
semantic prosody and predominantly appears in negative contexts; its most common collocates
are helpless, useless, unable, forgotten, failed, ruined, destroyed, changed, different, pleasant,
SEMANTIC PROSODY AND IMPRESSION FORMATION 6
clear (Davies, 2008). In contrast, the adverb totally is equally likely to appear in a positive or
negative context and has no semantic prosody (Partington, 2004). If semantically-prosodic
descriptors are relevant to person perception, participants should form a more negative
impression of the target person when he is described as being utterly (vs totally) changed.
Method
Participants. Based on effect sizes found in prior research (Hauser & Schwarz, 2016),
sample size was determined to exceed 80% power for a slightly larger than “small” effect size (d
= .25). Five hundred and eighty workers from Amazon Mechanical Turk (MTurk) were directed
to our survey in exchange for 20 cents. Workers were required to be Americans who had not
participated in our prior studies on semantic prosody with a 90% HIT acceptance rate and 100
prior completed HITs. However, 17 workers (3%) exited the survey before being randomly
assigned to an adverb condition. This left us with a final sample of 563 participants (284 female;
age range 18 73).
Pretest and materials. A separate group of eighty-one participants were randomly
assigned to read the utterly (n = 41) or totally (n = 40) version of a sentence: “When Shelly first
visited the market in Hong Kong, it was utterly (totally) unlike anything she had ever
experienced before.” They then defined utterly (totally) in a free response box. Finally, they
indicated which of six words (order randomized) could be substituted for the word utterly
(totally) and still mean the same thing (i.e., which terms were synonymous): entirely, completely,
totally (utterly), partially, almost, miserably.
Definitions were run through a sentiment analysis tool (Miner et al., 2012); Porter
stemmed words (Porter, 2001) within each definition were matched to Porter stemmed entries of
the word norming database of Warriner et al (2013) and the corresponding valence rating for
each word was retrieved (1 = very negative, 9 = very positive). We then computed the average
valence of the words within each definition. There was no difference in the average valence of
definitions for utterly (M = 5.90, SD = 0.94) vs totally (M = 6.19, SD = 0.83), F(1, 69) = 1.89, p
= .173
1
. Further, most participants indicated that utterly and totally were quasi-synonymous;
86% indicated that the words could be exchanged in a sentence and still have the sentence mean
the same thing. Thus, the words are similar in meaning and valence.
Finally, both utterly and totally have semantic preference for describing transformations
(Partington, 2004). The mutual information scores (Church & Hanks, 1990) of “changed” with
both utterly (3.29) and totally (3.35) suggest it is a common collocate of each adverb (COCA,
Davies, 2008). Thus, each adverb is similarly frequent and fluent in the context of this sentence.
Procedure. In a study ostensibly on decision making, participants were randomly
assigned to read either the utterly (n = 288) or totally (n = 275) version of a sentence: “As his
siblings discovered, Daniel was a(n) utterly (totally) changed man when he returned.”
Participants then rated the extent to which sixteen trait adjectives (order randomized) described
Daniel (1 = not at all, 5 = extremely). Consistent with Fiske, Cuddy, Glick, and Xu (2002), half
1
Ten participants provided definitions that did not match any words in Warriner et al. (2009).
SEMANTIC PROSODY AND IMPRESSION FORMATION 7
of the trait adjectives corresponded to the warmth domain (warm, good natured, sincere, friendly,
well-intentioned, trustworthy) while the other half corresponded to the competence domain
(competent, confident, intelligent, capable, efficient, skillful). Finally, participants judged the
valence of Daniel’s change (“to what extent was Daniel’s change a good or bad thing”; 1 = very
negative, 4 = neutral, 7 = very positive).
Results and discussion
Following Fiske et al. (2002), we averaged the ratings of the warmth adjectives (α = .96)
and competence adjectives (α = .94) to create composite warmth and competence indices.
Because the ratings of the warmth and competence indices and the valence rating of Daniel’s
change were made on different scales, we standardized each of these dependent measures to z-
scores.
We conducted a 2 (adverb: utterly, totally) x 3 (rating: warmth, competence, change
valence) mixed analysis of variance on participants’ ratings (higher scores = more positive). As
predicted, semantic prosody affected participants’ judgments of Daniel’s traits, F(1, 561) = 7.54,
p = .006, r = .12, 95% CI [.06, .35], for the main effect of adverb. Ratings of Daniel’s warmth,
competence, and change were more negative when he was described as having utterly changed
(M = -.10, SE = .05) than when he was described as having totally changed (M = .11, SE = .05),
95% CI [.06, .35]. The main effect of rating was not significant, F < 1.
There was also a marginally significant interaction of adverb and rating, F(2, 1122) =
2.73, p = .066, ηp2 = .01. As shown in Figure 1, the effect of semantic prosody was strongest on
change valence ratings, F(1, 561) = 12.04, p = .001, r = .15, 95% CI [.13, .45] for the simple
effect. In comparison, the adverb moderately affected trait impressions of Daniel’s warmth, F(1,
561) = 4.98, p = .026, r = .09, 95% CI [.02, .35] for the simple effect. Adverb also had a
marginally significant effect on trait impressions of Daniel’s competence, F(1, 561) = 2.73, p =
.099, r = .07, 95% CI [-.02, .31] for the simple effect. This smaller effect of semantic prosody on
evaluations of competence (compared to warmth) is in line with evidence that changes in warmth
are more predictive of perceived identity change (Strohminger & Nichols, 2015). Throughout,
the effect of adverb followed the predicted pattern: more negative impressions of Daniel’s
change, warmth, and competence when he was described as utterly rather than totally changed.
These findings provide first evidence that semantically-prosodic adverbs can affect
person perception. Although the terms utterly and totally are near-synonyms, utterly tends to
occur in negative contexts, and when used to describe a person, it conveys negative information
about him/her. A seemingly innocuous choice of adverb can shift the perceived meaning of an
utterance.
SEMANTIC PROSODY AND IMPRESSION FORMATION 8
Figure 1. Mean impression ratings by adverb. Bars denote +/- 1 standard error of the mean.
Study 2
Study 2 provides a conceptual replication of the observed influence of utterly vs. totally
and examines whether these semantically-prosodic adverbs can affect person impressions under
conditions that provide more contextual information. Specifically, participants were asked to
imagine they had a tech job in an office, and their boss was recently retired. A new boss was
hired, whose management style was described by a co-worker as being utterly (or totally)
unconventional. We predict that participants will form a more negative impression of their future
boss when he is described as being utterly (vs totally) unconventional.
Method
Participants. Based on the effect size found in Study 1, sample size was determined to
exceed 80% power for a slightly larger than “small” effect size (d = .22). Seven hundred and two
workers from MTurk were directed to our survey in exchange for 30 cents. Workers were
required to be Americans who had not participated in our prior studies on semantic prosody with
a 90% HIT acceptance rate and 100 prior completed HITs. Twenty-three workers (3%) dropped
out of the survey upon reading the informed consent information on the first page and therefore
had no data. Another fifteen people (2%) had missing data due to dropping out before
completing the survey, but they were not overrepresented in either wording condition (7 totally
dropouts vs 8 utterly dropouts). Finally, we excluded the data of thirteen participants (2%) who
took over 12 minutes to complete the survey, which was 2.5 standard deviations beyond the
mean survey duration. Again, these exclusions were not overrepresented in either wording
condition (7 totally exclusions vs 6 utterly exclusions). This left us with a final sample of 651
participants (320 female, age range 18 73).
Procedure. Participants were recruited for a survey on impressions of management. They
first read the following scenario, wording manipulation (nutterly = 315, ntotally = 336; randomly
assigned) in brackets:
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
Warmth Competence Positive Change
Rating
totally
utterly
SEMANTIC PROSODY AND IMPRESSION FORMATION 9
Imagine that you have a tech job in an office. Your boss recently retired, and your
company just hired someone to take his place (i.e., to be your new boss). You
haven’t met him yet, but you’ve overheard people talking about him. One of your
co-workers who already knows him says that his management style is totally
[utterly] unconventional.
Both utterly and totally are equally infrequent in describing the adjective unconventional; each
appear only twice before the adjective in COCA (Davies, 2008).
Participants then responded to seven-point impression scales indicating how they thought
they would feel toward the new boss (1 = very negatively, 7 = very positively), how their co-
worker feels toward the new boss (1 = very negatively, 7 = very positively), how well their boss
will fit in with the company (1 = poor fit, 7 = great fit), and how much they would like working
with the new boss (1 = dislike very much, 7 = like very much).
Results and discussion
Participants responses to the impression questions were averaged to form an overall
impression index (α = .86) with higher numbers indicating more positive impressions. As
predicted, semantically-prosodic descriptors affected impressions of the new boss. The boss was
seen more negatively when he was described as utterly unconventional (M = 3.90, SD = 1.04)
than when he was described as totally unconventional (M = 4.08, SD = 0.97), F(1, 649) = 4.80, p
= .029, r = .09, 95% CI [.018, .328] for the main effect of wording. Therefore, semantically-
prosodic descriptors affect person impressions even when more contextualized scenarios are
presented. Compared to Study 1, the observed size of the effect is somewhat smaller, which is
compatible with the common observation that the impact of a given piece of information
decreases with the amount of other information considered (for reviews, see Bless, Schwarz, &
Wänke, 2003; Wyer, 1974). This suggests that semantically-prosodic descriptors may have
smaller effects the more other information relevant to the target is available.
Study 3
Going beyond semantically prosodic adverbs as the independent and trait judgments as
the dependent variable, Study 3 assesses the influence of semantically-prosodic verbs as
descriptors of a person’s actions on perceivers’ impressions and behavioral intentions.
Participants read about candidates in a hypothetical election. The incumbent candidate was
described as having introduced regulations that caused (or produced) extreme changes to the
budget in his past term. Prior research has established that people see the words cause and
produce as being synonymous and similar in valence (Hauser & Schwarz, 2016). However,
cause has negative semantic prosody while produce has no semantic prosody (Stubbs, 1995). If
semantic prosody guides behaviors toward others, participants should have more negative
impressions and less favorable voting intentions for a candidate who is described as having
caused (vs produced) changes.
Method
Participants. Based on the effect sizes for similar studies using the same semantically
prosodic words (Hauser & Schwarz, 2016), sample size was determined to approximate 80%
SEMANTIC PROSODY AND IMPRESSION FORMATION 10
power for an effect size of d = .60. One hundred and one workers from MTurk were directed to
our survey. Workers were required to be Americans with a 95% HIT acceptance rate and 500
prior completed HITs. We excluded the data of one participant who took over 26 minutes to
complete the task, which was over nine standard deviations beyond the mean duration (less than
a minute). This left us with a final sample of 101 participants (49 female, age range 18 71).
Pretest and materials. Hauser and Schwarz (Pilot Study, 2016) pretested cause and
produce in a similar design as the pretest in Study 1. Forty participants defined either word and
selected its synonyms. The definitions of the words were similar in valence (p > .254) and 95%
of participants identified the words as being synonyms. Thus, the words are similar in meaning
and valence.
Additionally, cause and produce appear in constructions with the noun changes to a
similar extent. The mutual information scores (Church & Hanks, 1990) of changes with both
cause (2.26) and produce (2.70) suggest it is an equally common collocate of each verb (COCA,
Davies, 2008). Thus, each verb is similarly frequent and fluent in the context of this sentence.
Procedure. After completing a prior survey in which they defined and identified
synonyms for words, participants were directed to our survey ostensibly on inferences. They
were randomly assigned to read one of two passages about a political election, which solely
differed in the use of caused (n = 49) vs. produced (n = 52) in describing one of the candidates:
“In his past term, Governor Steve Williams introduced regulations that caused
(produced) extreme changes to the budget that were felt by many. He is the incumbent
candidate running against local politician, Joshua Bayer.”
Participants then reported whether they inferred that the governor made the situation
better or worse (forced choice) and who they would vote for in this election: Joshua Bayer or
Steve Williams (forced choice).
Results and discussion
Responses to the two questions were averaged to form a favorability index, with higher
numbers being more favorable to the incumbent (α = .85). As expected, the semantic prosody of
the verb affected impressions of the candidates. Participants were more likely to infer that the
incumbent candidate made the situation worse and were less likely to vote for him when he
caused changes (M = 1.11, SD = .29) than when he produced changes (M = 1.26, SD = .41), t(99)
= 2.05, p = .043, r = .20, 95% CI [.01, .29]. Study 4
In studies 1 to 3, the person descriptions were provided by the experimenters. As seen in
previous research (Schwarz, 1994, 1996), contributions of the researcher/experimenter are seen
as particularly relevant to the task the researcher presents, making it possible that participants
attended more to the semantically-prosodic descriptors and their potential implications than
might otherwise be the case. People are less likely to take utterances at face value when the
utterance is potentially self-serving, casting doubt on whether the speaker is a cooperative
Gricean communicator. Using an advertising context, Study 4 tests whether semantically
prosodic descriptors might guide impressions even in these potentially adverse conditions.
SEMANTIC PROSODY AND IMPRESSION FORMATION 11
Participants were recruited for a survey on consumer decisions and evaluated a bank after
reading its slogan: “Bayview Bank – we lend (loan) more than money.” Prior research has
established that people see the words lend and loan as being synonymous and similar in valence
(Hauser & Schwarz, 2016). However, lend has positive semantic prosody while loan has no
semantic prosody (Ellis & Frey, 2009). If semantic prosody has social influence even in
potentially adverse conditions where people are skeptical of the target’s self-description,
participants should have more positive impressions of the bank when it lends (vs loans) more
than money.
Method
Participants. Based on the effect size found in Study 1, sample size was determined to
approximate 80% power for a slightly larger than “small” effect size (d = .22). Six hundred and
ten workers from MTurk were directed to our survey in exchange for 20 cents. Workers were
required to be Americans who had not participated in our prior studies on semantic prosody with
a 95% HIT acceptance rate and 500 prior completed HITs. Three participants dropped out of the
survey upon reading the informed consent information. This left us with a final sample of 607
participants (275 female, age range 19 78).
Pretest and materials. Hauser and Schwarz (Pilot Study, 2016) pretested lend and loan
in a similar design as the pretest in Study 1. Forty participants defined either word and selected
its synonyms. The definitions of the words were similar in valence (p > .254) and 100% of
participants identified the words as being synonyms. Thus, the words are similar in meaning and
valence.
Additionally, neither lend and loan are strong collocates of the word more in natural
language. Neither have mutual information scores (Church & Hanks, 1990) above 3 in COCA
(Davies, 2008). On the other hand, both lend (MI = 5.72) and loan (MI = 6.28) are strong
collocates of the word money in COCA (Davies, 2008). Thus, each verb is similarly infrequent
alongside more but are similarly frequent alongside money.
Procedure. Participants were recruited for a survey on consumer decisions. They were
randomly assigned to read one of two slogans for Bayview Bank, which differed only in the use
of lend (n = 294) vs. loan (n = 313): “Bayview Bank – we lend (loan) more than money.”
Participants indicated how much they liked the bank’s slogan (1 = dislike very much, 2 =
dislike a little, 3 = neutral, 4 = like a little, 5 = like very much) and rated the extent to which
eight warmth-related trait adjectives (warm, good natured, sincere, friendly, well-intentioned,
trustworthy; order randomized) described how Bayview Bank acts towards its customers (1 = not
at all, 5 = extremely). These 8 ratings were averaged to form an index of warmth impressions (α
= .94).
Results and discussion
As expected, semantic prosody guided evaluations of the bank. Participants liked the lend
slogan more (M = 3.21, SD = 1.00) than the loan slogan (M = 2.99, SD = 1.10), t(605) = 2.62, p
= .009, r = .11, 95% CI [.06, .39]. They also thought the bank’s customer service was warmer
when the bank lends money (M = 3.52, SD = 0.80) than when the bank loans money (M = 3.35,
SEMANTIC PROSODY AND IMPRESSION FORMATION 12
SD = 0.97), t(605) = 2.47, p = .014, r = .10, 95% CI [.04, .32]. Thus, the expected influence of
semantically-prosodic descriptors can even be observed under potentially adverse conditions
when the target provides the descriptors and has a clear benefit from crafting a positive
impression. Study 5
Study 5 extends these findings to another social domain known for strategic self-
presentations, namely online dating, and includes a larger number of targets and descriptors. We
asked hetereosexual female participants to read sixteen ostensible Tinder dating profiles of
hetereosexual men and to report their romantic interest. Each of the 16 profiles contained a
semantically-prosodic word (drawn from a set of four positive and four negative descriptors) or a
matched non-semantically prosodic synonym. For example, one profile read “People say I tend
to escalate (heighten) any situation.” If semantic prosody has subtle social influence that persists
even when perceivers are skeptical of the self-serving nature of information, participants should
have more romantic interest when the profile contains words with positive (vs no) semantic
prosody and less romantic interest when the profile contains words with negative (vs no)
semantic prosody.
Method
Participants. Based on the effect sizes of within-subject studies with multiple
semantically prosodic terms (Hauser & Schwarz, 2016), sample size was determined to exceed
80% power for an effect size of d = .30. One hundred and five workers from MTurk were
directed to our survey in exchange for 60 cents. Using TurkPrime’s MTurk Toolkit Panel feature
(Litman, Robinson, & Abberbock, 2016), workers were required to be American females, who
had not participated in our prior studies on semantic prosody, and had a 95% HIT acceptance
rate and 100 prior completed HITs. Four participants (4%) dropped out of the survey before
completing all questions. This left us with a final sample of 101 participants (age range 21 66).
We also requested that MTurkers only participate in our study if they were hetereosexual
females; while these conditions were not met for four participants, we retained all cases in the
current analyses to minimize exclusions; the same patterns hold if such exclusions are made.
Pretest and materials. Eight semantically-prosodic words were identified for use in the
study. Six out of eight words (cause, encounter, commit, attain, restore, lend) were established as
having semantic prosody in prior literatures (Hauser & Schwarz, 2016; Ellis & Frey, 2009).
Hauser and Schwarz (Pilot Study, 2016) pretested these six words and their non-semantically
prosodic synonyms (produce, happen upon, engage in, get, bring back, loan) in a similar design
as the pretest in Study 1. Forty participants defined a word and selected its synonyms. The
definitions of the words were similar in valence (ps > .254) and the majority of participants
identified the semantically prosodic words as being synonymous with their non-semantically
prosodic synonyms (cause-produce: 95%; encounter-happen upon: 87.5%; commit-engage in:
90%; attain-get: 90%; restore-bring back: 100%; lend-loan: 100%). Thus, the words are similar
in meaning and valence.
SEMANTIC PROSODY AND IMPRESSION FORMATION 13
Additionally, the words in each pair were similarly infrequent in the context of their
profile from analyses of collocations in COCA (Davies, 2008). Neither cause nor produce have
mutual information scores above three within four words to the right of the word work; the same
is true for encounter-happen upon for the word issues, for commit-engage in for the word what,
for attain-get for the word what, and for lend-loan for the word help. Neither restore nor bring
back have mutual information scores above three within four words to the left of the word things.
Thus, the verbs in each pair are similarly infrequent in the context of their profile.
The remaining two semantically-prosodic words were selected from word norming data.
Warriner et al (2013) contains valence norms for words while Snefjella and Kuperman (2016)
contains the average valence of context for words. These data were combined, words with
relatively neutral valence were selected, and escalate and vibes were identified as words with
extreme negative and positive context valence, respectively. We identified non-semantically
prosodic synonyms for these semantically-prosodic words from thesauri: heighten and feelings,
respectively.
We pretested these two word pairs in a similar design as the pretest in Study 1. Forty-one
participants were given eight sentences (order randomized), each containing an underlined word.
Participants defined the underlined word and selected its synonyms. One sentence pretested the
escalate-heighten pair (word in parentheses randomly assigned): “Emotions were escalated
(heightened) on that last day. Participants identified whether increased, intensified, heightened
(escalated), lowered, concealed, and anticipated were synonyms (order randomized). Another
sentence pretested the vibes-feelings pair: Matt enjoyed the good vibes (feelings) at the music
festival. Participants identified whether spirit, passion, feelings (vibes), physicality, food, and
conversations were synonyms. The remaining sentences pretested other potential words with
semantic prosody which weren’t utilized in the current study. Importantly, participants saw the
words in each pair as being synonymous, as 97.5% of people selected heighten as a synonym for
escalate, and 97.5% of people selected feelings as a synonym for vibes. Thus, participants view
the words as having a similar meaning. Further, the words in each pair were similarly infrequent
in the context of their profile from analyses of collocations in COCA (Davies, 2008). Neither
escalate nor heighten occur more than ten times before the word situation, and neither vibes nor
feelings have mutual information scores above three within four words to the right of you. Thus,
the verbs in each pair are similarly infrequent in the context of their profile.
Procedure. Women were recruited for a survey on the evaluation of statements from
heterosexual male Tinder dating profiles. Profile statements contained either a word with
semantic prosody (positive = vibes, attain, restore, and lend; negative = escalate, cause,
encounter, commit) or a matched non-semantically prosodic synonym (positive control =
feelings, get, bring back, loan; negative control = heighten, produce, happen upon, engage in,
respectively). All manipulations were within-subjects and the order of the resulting 16 profiles
was randomized. All profile statements are shown in Table 1.
Participants read each profile statement and rated the profile holder on seven-point scales:
to what extent the man in the profile seemed friendly (1 = not at all friendly, 7 = very friendly),
SEMANTIC PROSODY AND IMPRESSION FORMATION 14
the likelihood they would “swipe right” on him if he looked attractive in his profile picture (i.e.,
“like” him on Tinder and agree to communicate with him; 1 = very unlikely, 7 = very likely),
and the likelihood they would go out on a date with him if he asked (1 = very unlikely; 7 = very
likely). These ratings were averaged to form a romantic interest index; Cronbach’s α for this
index varied across the profiles, with the lowest α = .85.
Results and discussion
A 2 (semantic prosody: present, absent) x 2 (valence of the semantically prosodic term’s
usual collocates: positive, negative) within-subjects ANOVA was conducted on the romantic
interest index. As shown in Table 1, semantically-prosodic words affected romantic interest in
the dating profiles. Participants reported less romantic interest for profiles containing a negative
semantically-prosodic word vs a matched non-semantically prosodic synonym, t(100) = 8.86, p <
.001, d = .88, 95% CI [0.47, 0.75] for the simple effect of semantic prosody. Participants also
reported more romantic interest for profiles containing a positive semantically-prosodic word vs
a matched non-semantically prosodic synonym, t(100) = 9.56, p < .001, d = .95, 95% CI [0.48,
0.73] for the simple effect of semantic prosody. This is reflected in a significant interaction of the
presence vs. absence semantic prosody and the valence of the semantically prosodic terms’ usual
collocates, F(1, 100) = 144.49, p < .001, ηp2 = .59.
SEMANTIC PROSODY AND IMPRESSION FORMATION 15
Table 1. Mean impression index for each dating profile by word in Study 5
Mean impression (SD)
Negative semantic prosody (vs matched
synonyms)
“People say I tend to ____ any situation.”
Escalate
2.00 (1.16)
Heighten
3.02 (1.49)
“I have ____ some work for my
girlfriends in the past.”
Caused
2.09 (1.19)
Produced
2.96 (1.49)
“My relationships tend to ____ issues.”
Encounter
2.12 (1.17)
happen upon
2.36 (1.30)
“I ____ what I do.”
Commit
4.54 (1.36)
engage in
4.85 (1.13)
Overall
Negative semantic prosody
2.69 (0.87)
Matched synonym
3.30 (0.91)
Positive semantic prosody (vs matched
synonyms)
“Hope to get some good ____ with you.”
Vibes
4.21 (1.53)
Feelings
3.95 (1.56)
“My friends say I ____ what I want.”
Attain
3.46 (1.47)
Get
2.88 (1.50)
“I like finding old things to ____.”
Restore
5.37 (1.12)
bring back
4.68 (1.34)
“I can ____ my help.”
Lend
4.99 (1.23)
Loan
4.10 (1.49)
Overall
Positive semantic prosody
4.51 (0.92)
Matched synonym
3.91 (3.91)
Note. * p < .05 | ** p < .01 | *** p < .001
As shown in Table 1, this effect held for each of the semantically-prosodic descriptors,
although the effect size varied. This provides further evidence of semantic prosody’s effect on
person impressions even when perceivers are aware of a target’s persuasive intent. Dating profile
holders on Tinder are undoubtedly attempting to craft favorable impressions of themselves in
SEMANTIC PROSODY AND IMPRESSION FORMATION 16
order to generate romantic interest. Nonetheless, semantic prosody guided perceiver’s romantic
impressions of these profiles, suggesting these words have subtle social influence. Finally, the
observation that semantic prosody not only influenced impressions of the target, but also
perceivers’ intention to “swipe right” (i.e., to select the profile as a person they would be willing
to chat with) converges with the observation of behavioral intention effects in electing a political
candidate in Study 3. Summary analysis
We conducted a meta-analysis of the current studies to assess the strength of the evidence
of semantic prosody on person impressions. The effect within each study was transformed into a
standardized mean difference for the predicted effect of semantic prosody, with positive numbers
representing effects in the predicted direction. In Study 4, the predicted effect was computed as
an average of the slogan impression and warmth perception of the bank, and Study 5 was
included twice, once per predicted effect of positive vs. negative semantic prosody. Using a
random effects model in the R metafor package (Viechtbauer, 2010), we estimated the
standardized mean effect of semantic prosody to be .36 (SE = 0.09, z = 4.02, p < .0001, 95% CI
[0.19, 0.54], see Figure 2).
Figure 2. Forest plot of standardized mean effect of semantic prosody for each study and random
effects model for meta-analytic effect at the bottom. Dot size represents sample size and bars
represent 95% confidence intervals.
General Discussion
As social animals, we constantly form impressions of the many others we meet and hear
about. The bulk of impression formation research has focused on descriptions with clearly
valenced implications (for a review, see Schneider, Hastorf, & Ellsworth, 1979). Going beyond
valenced descriptions, previous research has also shown how grammatical aspects can affect
inferences about the stability of another person’s traits (Hart & Albarracin, 2011; Fausey &
Matlock, 2011; Semin & Fiedler, 1991). However, little is known about how word choice
SEMANTIC PROSODY AND IMPRESSION FORMATION 17
between seemingly synonymous words can guide positive or negative impressions of others. The
five studies reported here consistently demonstrated that semantically prosodic descriptors that
lack a clear positive or negative meaning still lead people to infer the valence of what is to come,
which colors the impressions they form of others.. This effect occurs not only for evaluations of
generically-described others (Study 1), but also for evaluations of managers (Study 2), political
candidates (Study 3), brands (Study 4), and potential romantic partners (Study 5). The findings
converge on the conclusion that semantic prosody can exact profound effects on impressions of
others in consequential domains of everyday life.
The findings suggest that semantically-prosodic descriptors can exert a subtle social
influence (for a review of other factors that subtly affect social inferences, see Uleman, Saribay,
& Gonzalez, 2008). While people are hesitant to offer judgments about persons when they feel
that diagnostic information is lacking (Yzerbyt et al., 1994), semantic prosody guided judgments
under conditions devoid of contextual information (Study 1) as well as under conditions typical
for person perception experiments (Studies 2, 3, and 5). Semantically-prosodic descriptors also
affected behavioral intentions to vote for political candidates (Study 3) and intentions to initiate
interaction with potential romantic partners (Study 5). Finally, semantic prosody influenced
person impressions even under conditions that can draw the informational value of the
description into question, namely, conditions where person descriptors were provided by targets
who have much to gain from portraying themselves in a positive light. Perceivers formed more
favorable impressions of brands (Study 4) and potential romantic partners (Study 5) when they
described themselves using descriptors with positive rather than negative or no semantic
prosody.
The observed effects have implications for many areas of behavioral science, including
social influence, decision making, and social science measurement. Previous research found that
information is more persuasive (Clark & Javiland, 1977; Loftus, 1979), less likely to be
identified as misleading (Loftus, 1975) and more likely to result in false memories (Loftus, 1975;
Strack & Bless, 1994; Fiedler, Walther, Armbruster, Fay, & Naumann, 1996), when it is
introduced in the form of presuppositions rather than focal claims. This reflects that focal claims
are more likely to attract attention and deliberative evaluation (Petty & Cacioppo, 1986) as well
as psychological reactance (Brehm, 1966), resulting in boomerang effects in social influence
(Silvia, 2006). From this perspective, semantic prosody offers a particularly promising avenue
for crafting influential communications. Due to the usually negative collocates of cause, stating
that something causes X entails the presupposition that X is negative. Given people’s limited
insight into the operation of semantic prosody, such wordings will rarely attract scrutiny, making
them a potentially powerful tool of influence. We would similarly expect that semantically
prosodic words can bias decisions when used in the description of choice alternatives or the
framing of the decision task. By the same token, semantically prosodic words may exert an
unintended influence on respondents’ interpretation of questions in public opinion surveys,
potentially biasing the resulting inferences.
SEMANTIC PROSODY AND IMPRESSION FORMATION 18
The current findings also illustrate the nuances of word meaning. While we all have
strong intuitions about what words mean, meaning is not as explicit and static as once believed
(c.f. Evans, 2009). Like most cognitive processes, inferring what a word means is a situated
process, and inferences of meaning are constructed from past encounters of the word and its
context (Hoey, 2005; Kilgarriff, 1997), grammatical constructions (Elman, 2011; Ellis,
O’Donnell, & Romer, 2015), and situational inputs (McKoon & Ratcliff, 1992). Maki and
Buchanan (2008) classify word knowledge as having three separate components: semantic
(shared features among words), associative (words that other words bring to mind), and
collocational (co-occurrence of words in natural language). The effect of semantic prosody
appears most similar to collocational word knowledge but with one major caveat. The
semantically prosodic words in these studies did not collocate any more or less with the other
specific word stimuli than the non-semantically prosodic synonyms. However, they do collocate
with a generalized valence, which seems to be expected when they are encountered. Future
research could explore whether this is best conceptualized as an extension of collocational
associations or as a novel component of word knowledge.
The results also carry implications for common natural language processing tools in
psychology. Words can convey a range of meanings given the history of their usage and the
context in which they appear. LIWC (Pennebaker, Boyd, Jordan, & Blackburn, 2015) classifies
the verb cause as being a cognitive verb one that implies thinking about underlying structure.
However, we find here that cause often promotes negative impressions of the person who causes
an outcome. While the context in which the word cause is used differs between our studies and
the life-event descriptions on which LIWC classifications are based, the example highlights the
context-sensitive nature of word meaning.
Semantic prosody effects also emphasize the situated and constructive nature of social
cognition (Schwarz, 2007; Smith & Semin, 2004). The semantically-prosodic terms used in the
current research are unlikely to be found in the person schemas and stereotypes that figure
prominently in social cognition research (Hamilton & Sherman, 1994). The terms are simple
verbs, adverbs, and nouns that often appear to be agnostic about expressing positivity or
negativity. However, their meaning incorporates information about the typical context
surrounding the word (Hoey, 2005) and people draw on the contexts in which they previously
encountered the word in making sense of current utterances. As a result, we judge words by the
company they keep -- and we judge others by the company the words in their descriptions keep.
SEMANTIC PROSODY AND IMPRESSION FORMATION 19
Open Practices
All data and materials for the studies can be found at http://osf.io/f2atn.
SEMANTIC PROSODY AND IMPRESSION FORMATION 20
References
Asch, S. E. (1946). Forming impressions of personality. The Journal of Abnormal and Social
Psychology, 41(3), 258-290.
Bless, H., Schwarz, N., & Wänke, M. (2003). The size of context effects in social judgment. In J.
P. Forgas, K. D. Williams, & W. von Hippel (Eds.), Social judgments: Implicit and
explicit processes (pp. 180-197). Cambridge, UK: Cambridge University Press.
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words (ANEW): Instruction
manual and affective ratings (pp. 1-45). Technical report C-1, the center for research in
psychophysiology, University of Florida.
Brehm, J. W. (1966). A theory of psychological reactance. New York: Academic Press.
Casasanto, D., & Lupyan, G. (2015). All concepts are ad hoc concepts. In E. Margolis & S.
Lawrence (Eds.), The conceptual mind: New directions in the study of concepts (pp. 543
566). Cambridge, MA: MIT Press.
Church, K. W., & Hanks, P. (1990). Word association norms, mutual information, and
lexicography. Computational Linguistics, 16, 22-29.
Clark, H. H., & Haviland, S. E. (1977). Comprehension and the given-new contract. In R. O.
Freedle (Ed.), Discourse processes Advances in research and theory (Vol. 1). Norwood,
N.J.: Ablex.
D'Andrade, R. L. (1974). Memory and the assessment of behavior. In H. M. Blalock Jr. (Ed.),
Measurement in the social sciences (pp 159-186). Chicago: Aldine.
Davies, M. (2008). The Corpus of Contemporary American English: 450 million words, 1990
present. Retrieved from http://corpus.byu.edu/coca/
Ellis, N. C., & Frey, E. (2009). The psycholinguistic reality of collocation and semantic prosody.
In R. Corrigan, E. Moravcsik, H. Ouali, & K. Wheatley (Eds.), Formulaic language Vol 2
(pp 473-497). Philadelphia: John Benjamins.
Ellis, N. C., O’Donnell, M., & Römer, U. (2015). Usage-based language learning. The Handbook
of Language Emergence, 163-180.
Elman, J. L. (2011). Lexical knowledge without a lexicon? The Mental Lexicon, 6, 133.
http://dx.doi.org/10.1075/ml.6.1.01elm
Evans, V. (2009). How Words Mean. Oxford: Oxford University Press.
Fausey, C. M., & Matlock, T. (2011). Can grammar win elections?. Political Psychology, 32(4),
563-574.
Fiedler, K., Walther, E., Armbruster, T., Fay, D., & Naumann, U. (1996). Do You Really Know
what you have seen? Intrusion errors and presuppositions effects on constructive
memory. Journal of Experimental Social Psychology, 32(5), 484-511.
Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation, from category-
based to individuating processes: Influences of information and motivation on attention
and interpretation. Advances in Experimental Social Psychology, 23, 1-74.
SEMANTIC PROSODY AND IMPRESSION FORMATION 21
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.
Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and
semantics: Vol. 3. Speech acts (pp. 41-58). New York: Academic Press.
Hamilton, D. L. & Sherman, S. J. (1989). Illusory Correlations: Implications for Stereotype
Theory and Research. In D. Bar-Tal, C. Graumann, A. Kruglanski, & W. Stroebe, (Eds.),
Stereotyping and Prejudice (pp. 59-82). Springer: New York.
Hamilton, D. L. & Sherman, S. J. (1994). Stereotypes. In R. S. Wyer & T. K. Srull (Eds.),
Handbook of social cognition (pp. 1-68). Hillsdale, NJ: Erlbaum.
Hart, W., & Albarracín, D. (2011). Learning about what others were doing: Verb aspect and
attributions of mundane and criminal intent for past actions. Psychological
Science, 22(2), 261-266.
Hauser, D. J. & Schwarz, N. (2016). Semantic prosody and judgment. Journal of Experimental
Psychology: General, 145, 882-896.
Higgins, E. T., Rholes, W. S., & Jones, C. R. (1977). Category accessibility and impression
formation. Journal of Experimental Social Psychology, 13(2), 141-154.
Hoey, M. P. (2005). Lexical priming: A new theory of words and language. London, UK:
Routledge. http://dx.doi.org/10.4324/9780203327630
Kelley, H. H. (1973). The processes of causal attribution. American Psychologist, 28(2), 107-128
Kilgarriff, A. (1997). I don’t believe in word senses. Computers and the Humanities, 31(2), 91-
113.
Litman, L., Robinson, J., & Abberbock, T. (2016). TurkPrime.com: A versatile crowdsourcing
data acquisition platform for the behavioral sciences. Behavior Research Methods, 1-10.
Loftus, E. F. (1975). Leading Questions and the Eyewitness Report. Cognitive Psychology, 7(4),
560-72.
Loftus, E. F. (1979). The malleability of human memory: Information introduced after we view
an incident can transform memory. American Scientist, 67(3), 312-320.
Louw, B. (1993). Irony in the text or insincerity in the writer? The diagnostic potential of
semantic prosodies. In M. Baker, G. Francis, & E. Tognini-Bonelli (Eds.), Text and
technology: In honour of John Sinclair (pp. 157176). Philadelphia, PA: John Benjamins.
Maki, W. S., & Buchanan, E. (2008). Latent structure in measures of associative, semantic, and
thematic knowledge. Psychonomic Bulletin & Review, 15(3), 598-603.
McKoon, G., & Ratcliff, R. (1992). Spreading activation versus compound cue accounts of
priming: Mediated priming revisited. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 18(6), 1155.
Miner, G., Edler, J., Hill, T., Nisbet, R., Delen, D., & Fast, A. (2012). Practical text mining and
statistical analysis for non-structured text data applications. Waltham: Academic Press.
Osgood, C., Suci, G., & Tannenbaum, P. (1957). The Measurement of Meaning. Urbana, IL:
University of Illinois Press.
SEMANTIC PROSODY AND IMPRESSION FORMATION 22
Partington, A. (2004). “Utterly content in each other's company": Semantic prosody and
semantic preference. International Journal of Corpus Linguistics, 9, 131-156.
Pennebaker, J.W., Boyd, R.L., Jordan, K., & Blackburn, K. (2015). The development and
psychometric properties of LIWC2015. Austin, TX: University of Texas at Austin.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances
in Experimental Social Psychology, 19, 123-205.
Porter, M. F. (2001). Snowball: A language for stemming algorithms.
Schneider, D. J., Hastorf, A. H., & Ellsworth, P. (1979). Person perception. Random House.
Schwarz, N. (1994). Judgment in a social context: Biases, shortcomings, and the logic of
conversation. Advances in Experimental Social Psychology, 26, 123-162.
Schwarz, N. (1996). Cognition and communication: Judgmental biases, research methods and
the logic of conversation. Hillsdale, NJ: Erlbaum.
Schwarz, N. (2007). Attitude construction: Evaluation in context. Social Cognition, 25, 638656.
Semin, G. R., & Fiedler, K. (1991). The linguistic category model, its bases, applications and
range. European Review of Social Psychology, 2(1), 1-30.
Silvia, P. J. (2006). Reactance and the dynamics of disagreement: Multiple paths from threatened
freedom to resistance to persuasion. European Journal of Social Psychology, 36(5), 673-
685.
Sinclair, J. (1991). Corpus, concordance, collocation. Oxford, UK: Oxford University Press.
Smith, E. R., & Semin, G. R. (2004). Socially situated cognition: Cognition in its social context.
Advances in Experimental Social Psychology, 36, 53117.
Snefjella, B., & Kuperman, V. (2016). It’s all in the delivery: Effects of context valence, arousal,
and concreteness on visual word processing. Cognition, 156, 135-146
Srull, T. K., & Wyer, R. S. (1979). The role of category accessibility in the interpretation of
information about persons: Some determinants and implications. Journal of Personality
and Social psychology, 37(10), 1660-1672.
Strack, F., & Bless, H. (1994). Memory for nonoccurrences: Metacognitive and presuppositional
strategies. Journal of Memory and Language, 33(2), 203.
Strohminger, N., & Nichols, S. (2015). Neurodegeneration and identity. Psychological Science,
26(9), 1469-1479.
Stubbs, M. (1995). Collocations and semantic profiles: On the cause of the trouble with
quantitative studies. Functions of Language, 2(1), 23-55.
Tognini-Bonelli, E. (2001). Corpus linguistics at work. Philadelphia, PA: John Benjamins.
http://dx.doi.org/10.1075/scl.6
Uleman, J. S., Adil Saribay, S., & Gonzalez, C. M. (2008). Spontaneous inferences, implicit
impressions, and implicit theories. Annual Review of Psychology, 59, 329-360.
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. Journal of
Statistical Software, 36(3), 1-48.
SEMANTIC PROSODY AND IMPRESSION FORMATION 23
Warriner, A. B., Kuperman, V., & Brysbaert, M. (2013). Norms of valence, arousal, and
dominance for 13,915 English lemmas. Behavior Research Methods, 45(4), 1191-1207.
Wyer, R. S. (1974). Cognitive organization and change. Hillsdale, NJ: Erlbaum.
Xiao, R., & McEnery, T. (2006). Collocation, semantic prosody, and near synonymy: A cross-
linguistic perspective. Applied Linguistics, 27, 103129.
http://dx.doi.org/10.1093/applin/ami045
Yzerbyt, V. Y., Schadron, G., Leyens, J. P., & Rocher, S. (1994). Social judgeability: The impact
of meta-informational cues on the use of stereotypes. Journal of Personality and Social
Psychology, 66(1), 48-55.
... Specific words may also retain the valence of historical periods because of the other words that often surrounded them in conversation. Indeed, psycholinguistic work demonstrates that words absorb the valence of other words that frequently follow them in natural language (i.e., semantic prosody; Hauser & Schwarz, 2018;Sinclair, 1991). For example, in common conversation, the word cause is most frequently followed by negatively valenced words such as death, problems, and damage (https://www.english-corpora.org/coca/), ...
... whereas the synonym produced is frequently followed by more neutral terms. Consequently, people perceive a person who caused changes more negatively than a person who produced changes (Hauser & Schwarz, 2018). For example, a fictitious candidate who was labeled as "causing" budget changes received fewer reelection votes than one labeled as "producing" budget changes (Hauser & Schwarz, 2018). ...
... Consequently, people perceive a person who caused changes more negatively than a person who produced changes (Hauser & Schwarz, 2018). For example, a fictitious candidate who was labeled as "causing" budget changes received fewer reelection votes than one labeled as "producing" budget changes (Hauser & Schwarz, 2018). ...
Article
Full-text available
History can inconspicuously repeat itself through words and language. We explored the association between the “Black” and “African American” racial labels and the ideologies of the historical movements within which they gained prominence (Civil Rights and Black Power, respectively). Two content analyses and two preregistered experimental studies ( N = 1,204 White American adults) show that the associations between “Black” and “bias and discrimination” and between “African American” and “civil rights and equality” are evident in images, op-eds, and perceptions of organizations. Google Images search results for “Black people” evoke more racially victimized imagery than search results for “African American people” (Study 1), and op-eds that use the Black label contain more bias and discrimination content than those that use the African American label (Study 2). Finally, White Americans infer the ideologies of organizations by the racial label within the organization’s name (Studies 3 and 4). Consequently, these inferences guide the degree to which Whites support the organization financially.
... Where does the emotional meaning of these words come from? As laid out in more detail elsewhere (Van Berkum, 2019), it presumably reflects experiences that pair the word with emotions triggered by other things in a sufficiently reliable way, such as other emotion-laden words that it typically collocates with (e.g., Hauser and Schwarz, 2018;Snefjella et al., 2020), the typical social intention or emotional stance of the speaker using this word, the typical impact on listeners of this word, or the nature of the things referred to by this word. Because of these correlations, the release of that stored emotional meaning as part of lexical retrieval can be seen as a quick-and-dirty context-free prediction of what the word might in the end mean in this particular context. ...
Article
Full-text available
Verbal insults go against a universal moral imperative not to inflict harm on others, and they also pose a threat to one's face or reputation. As such, these “verbal slaps in the face” provide a unique opportunity to explore the interface between language and emotion. We used electroencephalography (EEG) and skin conductance (SC) recordings to compare the short-term impact of verbal insults such as “Linda is an idiot” or “Paula is horrible” to that of more positive evaluations (e.g., “Linda is an angel”, “Paula is impressive”) and neutral factual descriptions (e.g., “Linda is a student”), examining how responses to these various speech acts adapt as a function of massive repetition. By using either the participant's or somebody else's name, we also explored how statement impact depended on who was being evaluated. Multilevel ERP analysis with three predesignated latency ranges revealed an early insult effect in P2 amplitude that was highly robust over repetition and also did not depend on who the insult was about. This P2 effect points to a very rapid and stable capture of emotional attention, plausibly triggered by the retrieval of evaluative word meaning from long-term memory. Insults also elicited a larger late positive potential (LPP), again regardless of who the insult was about, but this later effect did not withstand repetition. Skin conductance responses showed that insults did not lead to more arousal than compliments did. In all, our findings suggest that in a standard psycholinguistic comprehension experiment without real interaction between speakers, insults deliver lexical “mini-slaps in the face,” such that the strongly negative evaluative words involved (e.g., “idiot”) automatically grab attention during lexical retrieval, regardless of how often that retrieval occurs.
... Thus, the media's objective, role, focus, and quality of a society can be measured (Mukhamediev et al., 2020). Finding in semantic prosody's study reveals that prosodic descriptors affect the impressions formed of others in five different circumstances for produced versus cause and utterly versus totally (Hauser & Schwarz, 2018). Corpus also defines ideology on Arabic political discourse related to Arab Israeli conflict (Sayaheen & Malkawi, 2019), which is possibly the result of one's recognition of word based on his/her inner psycholinguistics' element (Ellis, Frey, & Jalkanen, 2009) as a response toward generalized 'schema' constructed by contexts appearing in every discourse on corpora or product of delusionality, dogmatism, and religion fundamentalism, which results in wrong belief (Bronstein et al., 2019). ...
Article
This research discussed Indonesian hoax on transportation, spread using chain messages over social media, a major hoax subject. Setting social structure as an element to consider in hoax discourse gave different perspectives on why hoax news took place. Contrasting identified online hoax discourse to corpus-derived prosody could be interpreted using a sociopragmatic approach by Lin and Chung (2016). This research tried to look at how discourse on hoax correlated to real social issues among Indonesian people. Digital forensic was done by tracking discourse used as data from hoax news collection on turnbackhoax.com during 2017-2019. Discourse was later generated into clusters of subject idea of ‘tilang’, ‘jalan tol’, public land transport and infrastructure, ‘kecelakaan pesawat’, ‘lowongan kerja’, ‘China’, insecurity toward government and ‘kecelakaan kapal’. The subject idea was later consulted to Indonesian Web Corpus (IWAC). Through the socio-pragmatic approach, prosody derived by contrasting process portrayed hoax on transportation as a reflection of unsolved current social problems among Indonesian community including distrust toward the government, unawareness on transport safety, law and regulation, economic status, and trend. The finding of this research is recommended to take action toward those problems to eliminate hoax spread in the transportation sector.
... A word acquires additional meaning, which is not fixed in dictionaries, in its context. Contextual meaning can influence or guide our evaluative inferencing, thus creating a positive or negative semantic prosody [20]. ...
Article
Full-text available
The Master’s Degree Programme “Digital technologies in philology. Computer linguistics” at Southern Federal University, Russia, Rostov-on-Don, offers innovative multi-disciplinary courses based on the ideology of Digital Humanities. It combines knowledge from various sciences which complement each other in digital research. Cognitive linguistics is interdisciplinary in its essence as it makes use of different sciences, such as philosophy, linguistics, psychology, neuroscience, anthropology, mathematical statistics and others in order to explain the workings of our mind. This humanitarian science widely implements in its linguistic inquiry technologies, methods and tools from other related spheres. All these make it important to incorporate the academic course of cognitive linguistics into the Master’s Degree Programme. To better benefit the Programme the course is divided into two parts, theoretical and practical. During the course the students get knowledge of basic assumptions in cognitive linguistics, which help to set their minds in ‘cognitive linguistics’ thinking. Further course-based experience enables students to carry out linguistic research applying modern methods of corpus analysis, cognitive analysis, introspection and other big data tools. The practical part is based on a definite algorithm of steps, which help to conduct multidisciplinary analysis and easy to follow in the students’ independent research.
... The current research findings allow for a new opportunity to reconsider how and at what level different aspects of word knowledge are taught and assessed. Undoubtedly, semantic prosody needs to be explicitly taught in the process of developing learners' competence of aspects of word knowledge (Hauser & Schwarz, 2018;McGee, 2012;Nation, 2013;Omidian & Siyanova-Chanturia, 2020). Opportunities for exposure to such word knowledge information can be created at any level, despite the fact that, as the results of this study suggest, semantic prosody recognition and production may not go hand in hand. ...
Article
Full-text available
This study examines low-intermediate to advanced ESL learners’ awareness of semantic prosody. Semantic prosody is a word/unit association that expresses an attitudinal meaning (described as favourable, neutral, unfavourable), which emerges from the word’s typical lexical environment. Sensitivity to and awareness of such associations is an indicator of the learners’ depth of vocabulary knowledge. The study investigates ESL learners’ implicit semantic prosody knowledge via high-frequency verbs with known semantic prosody tendencies. 123 ESL students, whose proficiency ranged from low-intermediate to advanced, took an elicitation and recognition test on 13 of these verbs. The test assessed the students’ performance on productive and receptive semantic prosody. The results of this study evidence the learners’ difficulty to produce semantic prosody in comparison to their ability to observe and recognise it. They confirm that semantic prosody awareness appears to need time to develop significantly, and such development does not always correlate with the learners’ proficiency level. Implications for incorporating semantic prosody in vocabulary teaching and developing learners’ pragmatic competence are provided.
Article
Surveys, commonly employed in the social and behavioral sciences, are practical tools that can be used to assess the needs and attitudes of a given population. If not implemented in thoughtful ways, however, surveys can be inefficient or even harmful. With surveys often informing critical policy decisions, survey administrators must make careful methodological choices in order to obtain meaningful results and make sound decisions. In this paper, we review the social scientific literature on survey administration to aid policymakers, practitioners, and other survey administrators in understanding their position, identifying their participants, and establishing a plan for their surveys. We provide an overarching framework for survey design, guided by the idea that surveys are conversations between administrators and participants, in order to help creators of surveys make better decisions and engage more effectively with the communities they serve.
Article
Full-text available
We like people and objects more when they are described in positive than in negative terms. But even seemingly-neutral words can elicit positive or negative responses. This is the case for words that predominantly occur alongside positive (negative) words in natural language. Despite lacking positivity/negativity when evaluated in isolation, such semantically prosodic words activate the evaluative associations of their usual company, which can color judgment in unrelated domains. For example, people are more likely to infer that “endocrination” (a fictional medical outcome) is negative when it is “caused” (a word with negative semantic prosody) rather than “produced” (a synonymous word without semantic prosody). We review what is known about the influence of semantically prosodic words and highlight their importance for judgment and decision making.
Article
Full-text available
In everyday language, concepts appear alongside (i.e., collocate with) related concepts. Societal biases often emerge in these collocations; e.g., female (vs. male) names collocate with art- (vs. science-) related concepts, and African American (vs. White American) names collocate with negative (vs. positive) concepts. It is unknown whether such collocations merely reflect societal biases or contribute to them. Concepts that are themselves neutral in valence but nevertheless collocate with valenced concepts provide a unique opportunity to address this question. For example, when asked, most people evaluate the concept “cause” as neutral, but “cause” is frequently followed by negative concepts (e.g., death, pain, and trouble). We use such semantically prosodic concepts to test the influence of collocation on the emergence of implicit bias: do neutral concepts that frequently collocate with valenced concepts have corresponding implicit bias? In evaluative priming tasks, participants evaluated positive/negative nouns (Study 1) or pictures (Study 2) after seeing verb primes that were (a) strongly valenced (e.g., hate and comfort), (b) neutral in valence but collocated with valenced concepts in corpora (e.g., ease and gain), or (c) neutral in valence and not collocated with valenced concepts in corpora (e.g., reply and describe). Throughout, neutral primes with positive (negative) collocates facilitated the evaluation of positive (negative) targets much like strongly valenced primes, whereas neutral primes without valenced collocates did not. That neutral concepts with valenced collocates parallel the influence of valenced concepts suggests that their collocations in natural language may be sufficient for fostering implicit bias. Societal implications of the causal embedding hypothesis are discussed.
Article
This paper presents a corpus-driven Sinclairian analysis of five high-frequency Slovene verbs covering the lexical paradigm ‘to express orally’ in combination with their premodifying adverbs of manner. One of the main goals of the paper is to establish how frequent the phenomenon of semantic prosody actually is among high-frequency lexical items (here, adv-v pairs). A methodology aiming to provide an answer to this question has been proposed featuring the top-down approach (i.e., in order of decreasing frequency of occurrence). It involves setting up the widest possible parameters of searching for so-called ‘extended units of meaning’ and their semantic prosody amongst the most frequent lexical patterns in a language. A total of twenty-six adv-v pairs have been examined. Results indicate a strong correlation between the frequency of multi-word lexical items and their tendency to develop semantic prosodies: high-frequency collocations are thus more likely to have semantic prosodies compared to their lower-frequency counterparts. Overall, results also corroborate the trend of semantic prosody to be found with mainly negative meanings and to a lesser extent in neutral meanings, while no positive semantic prosody has been determined in this study.
Article
At present, we often encounter the use of the words 'suspect' and 'patient' in various discussions of the COVID-19 pandemic. In this case, the writer is eager to identify the frequency of using the words 'suspect' and 'patient' in the COVID-19 corpus and to explain the significant collocates based on the frequency score whether they are in a positive or negative sense. The two words are selected because they are identical synonyms which frequently used during the COVID-19 pandemic. This study applied the qualitative method with the COVID-19 data instrument through a sketch engine and used a corpus approach. The analysis results are how the word tokens from the synonym 'suspect' and 'patient' are described in the context of COVID-19 data. The word 'suspect' in the COVID-19 corpus shows collocation with the results of 14,505 (51.66 per million) and the word 'patient' with the number 518,618 (1,847.18 per million). The word 'patient' is more widely used in scientific research on the topic of COVID-19 than the word 'suspect'. Based on the semantic prosody, the words that follow the word 'suspect' are collocating negative tendencies. Conversely, words that follow the word 'patient' have collocates with a positive tendency
Article
Full-text available
Prior research has examined how distributional properties of contexts (number of unique contexts or their informativeness) influence the effort of word recognition. These properties do not directly interrogate the semantic properties of contexts. We evaluated the influence of average concreteness, valence (positivity) and arousal of the contexts in which a word occurs on response times in the lexical decision task, age of acquisition of the word, and word recognition memory performance. Using large corpora and norming mega-studies we quantified semantics of contexts for thousands of words and demonstrated that contextual factors were predictive of lexical representation and processing above and beyond the influence shown by concreteness, valence and arousal of the word itself. Our findings indicate that lexical representations are influenced not only by how diverse the word's contexts are, but also by the embodied experiences they elicit.
Article
Full-text available
In recent years, Mechanical Turk (MTurk) has revolutionized social science by providing a way to collect behavioral data with unprecedented speed and efficiency. However, MTurk was not intended to be a research tool, and many common research tasks are difficult and time-consuming to implement as a result. TurkPrime was designed as a research platform that integrates with MTurk and supports tasks that are common to the social and behavioral sciences. Like MTurk, TurkPrime is an Internet-based platform that runs on any browser and does not require any downloads or installation. Tasks that can be implemented with TurkPrime include: excluding participants on the basis of previous participation, longitudinal studies, making changes to a study while it is running, automating the approval process, increasing the speed of data collection, sending bulk e-mails and bonuses, enhancing communication with participants, monitoring dropout and engagement rates, providing enhanced sampling options, and many others. This article describes how TurkPrime saves time and resources, improves data quality, and allows researchers to design and implement studies that were previously very difficult or impossible to carry out on MTurk. TurkPrime is designed as a research tool whose aim is to improve the quality of the crowdsourcing data collection process. Various features have been and continue to be implemented on the basis of feedback from the research community. TurkPrime is a free research platform.
Article
Full-text available
Some words tend to co-occur exclusively with a positive or negative context in natural language use even though such valence patterns are not dictated by definitions or are part of the words’ core meaning. These words contain semantic prosody, a subtle valenced meaning derived from co-occurrence in language. As language and thought are heavily intertwined, we hypothesized that semantic prosody can affect evaluative inferences about related ambiguous concepts. Participants inferred that an ambiguous medical outcome was more negative when it was caused, a verb with negative semantic prosody, than when it was produced, a synonymous verb with no semantic prosody (Studies 1a, 1b). Participants completed sentence fragments in a manner consistent with semantic prosody (Study 2), and semantic prosody affected various other judgments in line with evaluative inferences (estimates of an event’s likelihood in Study 3). Finally, semantic prosody elicited both positive and negative evaluations of outcomes across a large set of semantically prosodic verbs (Study 4). Thus, semantic prosody can exert a strong influence on evaluative judgment.
Chapter
assesses what [has been] learned about some of [the] issues [surrounding stereotypes] from social psychological research, and particularly from research guided by a social cognition approach cognitive processes in stereotype formation / stereotypes as cognitive structures / stereotyping and information processing / affect, cognition, and stereotyping / stereotype change (PsycINFO Database Record (c) 2012 APA, all rights reserved)