Inclusion through non-verbal cues:
A moderated mediation model on non-verbal behaviours, warmth perception, race, and the feeling of inclusion
Master Thesis Final version
Name Anita Andrádi
Student number 11703369
Supervisor Brooke Gazdag
EBEC approval number 20220404080426
Word count 11513
Faculty Economics and Business
Programme MSc Business Administration
Specialisation Leadership and Management
Statement of originality
This document is written by Anita Andrádi who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of contents
Abstract ... 1
1. Introduction ... 2
2. Theoretical background ... 4
2.1 Non-verbal behaviour ... 4
2.1.1 Smiling ... 5
2.1.2 Posture ... 7
2.2 Perceived warmth ... 8
2.3 Anticipated feelings of inclusion ... 11
2.4 Match in race ... 14
2.5 Conceptual model ... 17
3. Methods ... 18
3.1 Sample ... 18
3.2 Research design and procedure ... 20
3.2.1 Coding procedure ... 20
3.2.2 Survey procedure ... 21
3.2.3 Measures ... 22
3.2.4 Control variables ... 23
3.3 Data analysis ... 24
4. Results ... 25
4.1 Descriptive statistics ... 26
4.2 Regressions ... 27
4.3 Moderated mediation model ... 29
4.4 Summary of results ... 32
5. Discussion ... 33
5.1 Strengths and limitations ... 37
5.2 Future research ... 38
5.3 Practical implications ... 40
6. Conclusion ... 42
7. References ... 43
8. Appendices ... 55
8.1 Appendix A - Survey introduction page ... 55
8.2 Appendix B - Survey questions ... 55
Diversity and inclusion are important and timely topics in the management literature due to the diversification of the workforce. But increased diversity also comes with stereotyping due to self-identity and self-categorisation. This paper broadens the literature on inclusion by adding non-verbal behavioural cues as potential indicators. In addition, stereotyping literature is also complemented by adding non-verbal elements and similarity in race as a moderator. Therefore, this paper investigates the relationship between positive non-verbal behaviours and the anticipated feelings of inclusion through warmth perception based on the match between perceivers and non-verbal actors’ race. Specifically, I expected a positive relationship between smiling, open posture, and anticipated feelings of inclusion where perceived warmth is a mediator. Furthermore, the match between races was expected to moderate the relationship between non-verbal cues and warmth perception as it becomes stronger when a match occurs.
A sample of 230 respondents provided data for the analysis. Evidence found support for the frequency of smiling leading to higher levels of warmth perception and the positive connection between warmth and inclusion perceptions. Unfortunately, this study provides no significant results for the relationship between the openness of posture and warmth perception, nor for the relationship between positive non-verbal behaviours and anticipated feelings of inclusion, and neither for the moderated mediation model with the match in race. Nevertheless, important managerial practices could be provided based on the insights gained in this research.
Keywords: Non-verbal behaviour, SCM, perceived warmth, anticipated feeling of inclusion, race and ethnicity, moderated mediation
“Diversity is a mix and inclusion is making the mix work.” says Andrés Tapia, the senior client partner and global diversity and inclusion strategist of Korn Ferry, a global organisational consulting company (Korn Ferry, n.d.; Tapia, n.d.). As the workforce of companies continuously diversifies in gender, race, and cultural backgrounds, organisations are putting more effort into efficiently managing it, thus becoming an important part of Human Resource Management (HRM) (Avery, 2011; Buengeler et al., 2018; D. Van Knippenberg et al., 2013).
It can also be seen from the fact, that diversity, equity, and inclusion remain a hot topic in 2022 as well due to becoming an important part of corporate responsibility (Best, 2022). Furthermore, the inclusion of employees can be seen as an outcome of different diversity practices, hence its importance in management studies (Buengeler et al., 2018).
Another important facet of management and leadership is communication. A great amount of research is done on verbal communication, and generally on how to engage with employees, but the realm of non-verbal expressions is also an important part of communication, and it is getting more recognition recently (Bellou & Gkorezis, 2016; Bonaccio et al., 2016).
To date, research on non-verbal behaviours has examined the non-verbal expressions of power, dominance and status and stereotypical emotional expressions of genders (Carney, 2020; J. S.
Smith et al., 2015; Witkower & Tracy, 2019). In this field, more demographic, cross-cultural, or personality-based variables could be further researched as well as the question of the universality of non-verbal behavioural elements (Bonaccio et al., 2016; Carney, 2020; Chung et al., 2020; Van Dijk et al., 2017).
According to the Stereotype Content Model (SCM) people immediately evaluate others based on how competent (intelligent and skilled) and warm (friendly and communal) they perceive them (Van Dijk et al., 2017). Abele et al. (2021) argued that the SCM can be seen as a simple structure and usually its element of competence is more easily detectable compared to
the other component of warmth. Also, although warmth and competence perceptions seem to be universal across people, it is important to mention it has been examined that in-group favouritism exists in these studies and most of them were conducted with Western samples (Cuddy et al., 2008, 2009; Fiske et al., 2002). Hence, it could be interesting to research how warmth interacts with other variables which could essentially lead to a more complex understanding of stereotyping.
When talking about stereotyping and immediate evaluation of others, surface-level characteristics, such as race is an easily observable attribute that not only influences self- categorisation but also how we make a judgement of others (Kammeyer-Mueller et al., 2011).
Today, race is defined as “a group sharing outward physical characteristics and some commonalities of culture and history” (Merriam-Webster, n.d.). Most research on races stems from the ideas of Tajfel et al.’s (1979) Social Identity Theory (SIT), in-group advantage and surface-level similarity regarding how much easier one can decode others’ feelings when their ethnicity matches or how much closer someone may feel to their in-group initially (Elfenbein
& Ambady, 2002; Harris & McGrath, 2012; Kammeyer-Mueller et al., 2011). Therefore, this paper also checks whether the perception of warmth changes if the perceiver’s race matches the displayed person’s race.
Thus, the goal of this thesis is to examine whether the positive relationship between positive non-verbal behaviours (frequent smiling and open posture) and the anticipated feelings of inclusion through warmth perception changes based on the similarity in race. This research provides a number of contributions. First, the literature on perception-based stereotyping would be broadened by tying it to non-verbal elements. Second, it would complement the field of non- verbal behaviour by focusing on the similarity of the race between the perceivers and non- verbal actors as a moderator to see how the understanding of specific non-verbal cues may change. Finally, as diversity is an important topic for businesses in the era of globalisation it
essentially could provide practical implications and knowledge on how different non-verbal expressions could be understood in diverse environments as well as what type of behaviour would be supportive of feelings of inclusion.
In order to answer the research question, the proposed variables, and the conceptual model along with its adherent hypotheses will be further explained in the theoretical background section. After that, the methods section will explain the process of this research in more detail, which will be followed by the results section, giving an interpretation of the analyses done. Then, the discussion section will give a more in-depth explanation of the potential findings of the study, accompanied by theoretical and practical implications. Finally, the limitations of this study and further research ideas will be explained as well as arriving at a conclusion of the paper.
2. Theoretical background 2.1 Non-verbal behaviour
Humans are social creatures and how one behaves is important since we continuously interpret and make judgements about others and whether they are suitable for interdependent relations (Cottrell et al., 2007). Motions and gestures have been recorded in the 18th century gradually leading to a more formal and detailed examination of this phenomenon through the years (Buck & Knapp, 2006). As mentioned before, researching non-verbal communication has been getting more attention recently (Bellou & Gkorezis, 2016; Bonaccio et al., 2016). Non- verbal behaviour also plays an important role in how attractive employees may find their leaders, as positive hand gestures positively affect attraction perception or how persuasive someone can be because of smiling instead of frowning (Guyer et al., 2019; Talley & Temple, 2015).
Ekman (1957) categorised behaviour in interpersonal situations as verbal, vocal, and non-verbal which is perceived through the visual sense. Later, in their study, Sundaram and
Webster (2000) have put non-verbal behaviours into four categories, which are kinesics, paralanguage, physical appearance, and proxemics. Kinesics includes body movements from posture and gestures to eye contact (Bonaccio et al., 2016). Bellou and Gkorezis (2016) expanded kinesics by adding positive kinesics including friendly expressions, like smiling a lot and being expressive. Paralanguage concerns the way of speaking such as vocal pitch, volume, and pauses, which is similar to Ekman's (1957) vocal category (Sundaram & Webster, 2000).
Physical appearance focuses on how someone presents themselves with the way they dress, but physically attractive appearance also belongs here (Sundaram & Webster, 2000). Finally, proxemics refers to touch and the distance one keeps with others (Bonaccio et al., 2016;
Sundaram & Webster, 2000). Research has found that non-verbal communication is highly important for forming an impression and that it is a very powerful way of communication (Hall
& Schmid Mast, 2007; Lakin, 2006).
Contrarily, Hall and Schmid Mast (2007) came to the conclusion that verbal cues still matter more since non-verbal cues are more ambiguous and less informative, and words tend to be used as the main information source. According to Spence’s (1973) signalling theory, not all information communicated is public, thus usually an information asymmetry emerges between people. Here, a sender must choose how they communicate or signal a piece of information while the receiver decides how to interpret it (Connelly et al., 2011). In the case of non-verbal behaviours specific to this paper, smiling and open posture may signal warmth and inclusivity.
In the case of this specific research, it is expected that smiling and laughing are connected to perceiving someone as friendly (Bellou & Gkorezis, 2016; Frances, 1979). An interesting aspect of smiling is that it can act as a biological sign of cooperation. By smiling one may reveal their “true self” in a sense that they willingly show their potential handicaps
(e.g., one can be judged by the condition of their teeth) which signals more trust while someone who is not smiling may be perceived as someone unwilling to reveal themselves (Scharlemann et al., 2001). In their research, Scharlemann et al. (2001) showed participants pictures of smiling and non-smiling people which were then rated by them based on how they perceived the pictures from being benevolent, and friendly, to negative perceptions such as unpleasant. The research was also complemented with a trust game to measure how much participants trust the smiling and non-smiling characters (Scharlemann et al., 2001). Based on this, they have found evidence that smiling is perceived as a sign of trust and willingness to cooperate (Scharlemann et al., 2001).
Intense smiling also leads to someone being perceived as warm and friendly, although not in all contexts. Min and Hu (2022) found that people working in people-oriented fields (e.g., servers at a restaurant) are perceived as not only warm but also competent individuals when they have broad smiles. However, in different industries, such as in the legal field, the positive effect between broad smiling and warmth perception is significantly lower since in that context it is rather perceived as unprofessional since a different kind of service is needed (Min & Hu, 2022). Thus, although smiling is generally found to be an indicator of friendliness, it does not necessarily stand in all contexts.
Research on smiling has shown that people have the ability to fake smiles, but they are also able to make a distinction between genuine and posed smiles for which eye movement (eyelid dynamics and eye expressions) is an important factor (Calvo et al., 2012; Dibeklioğlu et al., 2012; Scharlemann et al., 2001). Research has also found that people evaluate happy emotions mostly based on the other’s smile (Calvo et al., 2012). This means that although genuine expressions (when eyes and mouth are congruent, thus both are happy) are significantly rated as happy, blended expressions (smile without happy eyes) are also seen immediately as happy due to the salience of mouth movement (Calvo et al., 2012). Positive moods such as
happiness are usually associated with sociability and approachability, while negative moods signal avoidance (Cunningham, 1988). This leads to the assumption that the sole presence of smiling without accompanying “smiling eyes” indicates happiness, which is associated with friendly interactions.
Shifting from facial expressions, findings of Mehrabian's (1968) experiments on posture in communication suggested that relaxed posture, leaning forward, and smaller distance keeping signal a positive attitude towards others, he did not find significant support for open posture leading to the same positive attitude. However, it is important to mention that a high degree of interaction was observed between open posture and other variables (Mehrabian, 1968). On the other hand, newer findings in this area conclude otherwise. Non-verbal behaviours that convey warmth are leaning forward and positioning posture towards the audience, nodding, and relaxed gestures, while tense posture and intrusive gestures lead to perceptions of coldness (Cuddy et al., 2011). Additionally, closed posture signals unfriendliness, and behaviours such as eye contact, gazing, nodding, and open posture are found to be elements of powerful interpersonal interactions (Sundaram & Webster, 2000).
Furthermore, open posture also signals trust, equality, and affection and speakers behaving this way are generally perceived as friendlier (Cuddy et al., 2011). In emotion research, open and expanded posture is also found to be the indicator of love and pride (Sauter, 2017).
High-status individuals are more likely to adopt a more invasive, relaxed, and open posture compared to lower-status people (Burgoon, 1991). Additionally, dominance can be signalled by closed posture (Burgoon, 1991). Also, in his research Burgoon (1991) has found that open posture can parallel touch in the sense that it similarly conveys intimacy, such as trust and affection, but this result is only significant when the non-verbal actor is a male. Contrarily to this, more modern literature has found that dominant individuals tend to display an open and
more upright posture, similarly to high-status individuals (Carney et al., 2005; Peters et al., 2017). Next to that, in their study on how children perceive non-verbal behaviours of robots Peters et al. (2017) have found that open posture evokes the perception of warmth.
Hence, this research specifically focuses on smiling frequency, and open posture as these acts can be most connected to generally perceiving someone warmer or friendlier. In conclusion, this study would examine how these non-verbal behaviours have a positive relationship with the anticipated feelings of inclusion through warmth perception as a mediator, which will be further outlined in the next section. Henceforth, given the positive connotation of the selected behaviours, this collection of non-verbal behaviours – frequent smiling, and open posture – is referred to as “positive non-verbal behaviour” in this paper.
2.2 Perceived warmth
People automatically generate stereotypes which are defined as the “mental representations of social groups” (Lepore & Brown, 1997; Van Knippenberg & Dijksterhuis, 2000, p. 107). According to the Stereotype Content Model, people evaluate others instantly based on two dimensions, which are warmth (including friendliness, good nature, kindness, empathy, trustworthiness, etc.) and competence (including intelligence, skill, efficacy, power, etc.) (Cuddy et al., 2008, 2009; Fiske et al., 2002; Van Dijk et al., 2017). Further components of the warmth dimension defined by Cuddy et al., (2008) are generosity, helpfulness, tolerance, fairness, and sincerity. Sociability and morality are also sub-dimensions of warmth (Kervyn et al., 2015). Research also states that warmth is closely related to communion (Cuddy et al., 2008). Groups of people who are usually categorised in the high warmth level are the elderly, disabled, housewives (when competence is seen as low), and middle-class, Whites, Christians, students, and generally in-group members or close allies (when competence is also seen as high) (Cuddy et al., 2008, 2009; Fiske et al., 2002; Lee & Fiske, 2006).
Although according to Abele et al. (2021), competence is the more easily detectable aspect, literature on SCM regards warmth as the primary factor (Cuddy et al., 2008). The reason why the dimension of warmth tends to be more fundamental is because based on this, one can decide whether others are “friends or foes”, or whether the intention of others is to help or hurt (Fiske, 2018, p. 67; van Dijk et al., 2017). The underlying explanation for this is that while competence predicts passive behaviours (either convenient cooperation when high or neglect when low), warmth facilitates active behaviour (either helping when high or attacking when low) making the latter a more fundamental characteristic (Cuddy et al., 2008). Also, judgements on warmth are generated more quickly compared to competence judgements (Cuddy et al., 2011). Additionally, since humans are social beings, they seek suitable people for their interdependent relations, thus constantly evaluating based on trustworthiness and cooperation is an important factor (Cottrell et al., 2007).
In their study, Cottrell et al. (2007) looked at what characteristics people find important for ideal members of different social groups ranging from the professional field (work and project-related) to sports, study teams and personal life (friends and family). They concluded that trustworthiness is the most valued trait in most situations while cooperativeness is less ambiguous (Cottrell et al., 2007). They also found evidence that out of the two types of cooperativeness, communal orientation is important for situations when emotional support is needed (e.g., friends) while exchange orientation is needed in a more functional setting (e.g., professional life) (Cottrell et al., 2007). This means that although warmth is closely related to perceived willingness to cooperate (Cuddy et al., 2008) it has different interpretations depending on the situation.
Similarly, in the political field, which is closer to exchange orientation, competence- based traits are emphasised as the more important factor (Laustsen & Bor, 2017). On the other hand, other evidence of experiments done in the US and UK show that warmth traits of
candidates are actually more important for how people evaluate them, translating to vote choice (Laustsen & Bor, 2017). Warmth perception is usually an important part of service jobs. As mentioned before, broad smiles of servers signal warmth (Min & Hu, 2022), but it was also found that in the same industry weight and gender of servers also influence perceived warmth such that heavy women are evaluated higher in warmth compared to less heavy women or men (N. A. Smith et al., 2016). This means that next to kinesics, physical appearance also affects warmth perception. Furthermore, in medical studies, it was found that empathetic non-verbal behaviour (such as smiling, open posture, eye contact, and touch) of physicians leads to participants perceiving them warm (Kraft-Todd et al., 2017).
In leadership research, Tjosvold (1984) found in an experiment that subordinates put in a condition where their leader is conveying warmth are more likely to find the leader helpful, be satisfied with their relationship, and would work again with that leader. Furthermore, a new transformational leadership style also requires social skills and warmth and leads to higher follower satisfaction, performance, and empowerment (Cuddy et al., 2011; Den Hartog &
Koopman, 2001; Kark et al., 2003). Additionally, in organisations due to warmth stereotypes, those groups who are usually categorised as warm (e.g., women) may be disproportionally hired for jobs where sociability is more important (Cuddy et al., 2011). Also, when evaluating others in a professional setting people may think that someone who is stereotypically seen as cold (e.g., Asians) but acts warmly is being manipulative, which evokes negative feelings (Cuddy et al., 2011).
To conclude, the perception of warmth is connected to whether someone is perceived to have friendly intentions and has significance in various disciplines. Not only is it a trait made up of several positive characteristics, but it also can lead to positive outcomes as well. Some non-verbal cues such as smiling and open posture are also connected to whether someone is perceived as approachable, cooperative, friendly, or trustworthy which are all elements of the
warmth dimension. Consequentially, individuals wanting to convey these friendly intentions may also use these behaviours as signals (Connelly et al., 2011). Thus, the display of positive non-verbal cues can translate to perceiving others as warm which leads to the following hypotheses:
Hypothesis 1a: Positive non-verbal behaviour (smiling) is positively related to warmth perception.
Hypothesis 1b: Positive non-verbal behaviour (open posture) is positively related to warmth perception.
2.3 Anticipated feelings of inclusion
People naturally have the tendency to desire stable connections with others which can be traced back to an evolutionary need for survival (Krahmer et al., 2008). Additionally, an experiment showed that the mood of people significantly drops when feeling excluded while it raises when being included (Krahmer et al., 2008). A level of inclusion in groups is also needed to remain healthy in a psychological and physical sense (De Waal-Andrews & Van Beest, 2012). Chung et al. (2020) found that inclusion depends on two factors, which are belongingness and uniqueness. This was based on Shore et al’s (2011) model. Uniqueness refers to being allowed to freely express individuality (Chung et al., 2020). Belongingness here means that people are motivated to form groups and maintain strong interpersonal relationships and for this, positive interactions are needed (Chung et al., 2020). From an evolutionary point of view, collaboration or communion is an important facet of social interactions, especially because of a natural need to belong (Abele et al., 2021; Wagoner & Hogg, 2016). Thus, fulfilment of belongingness can be connected to warmth perception (as warmth also emphasises interpersonal relations and friendly nature) and essentially the anticipated feelings of inclusion as well.
According to Wagoner and Hogg (2016), warmth indicates acceptance and inclusion.
They also found that in high-uncertainty conditions, people seek inclusion, while in a low- uncertainty situation, the need for inclusion is lower, thus a need for competence and status prevails (Wagoner & Hogg, 2016). In their research, De Waal-Andrews and Van Beest (2012) further strengthened the argument that exclusion has detrimental effects on people’s needs and argued that the way people receive inclusion also matters. That is when individuals successfully claim inclusion does not satisfy them as much compared to when inclusion is granted to them (De Waal-Andrews & Van Beest, 2012). Additionally, they found that inclusion needs to be accompanied by interpersonal warmth perception in order to have a strong, positive effect on well-being and satisfy belongingness (De Waal-Andrews & Van Beest, 2012).
Further research has shown that perceived warmth is more important than perceived competence in the case of choosing collaborative partners. The reason for that is that people tend to categorise people as warm who are ingroup members or just not their competitors (Van Dijk et al., 2017). Also, due to a shift towards more ethical and transformational leadership, warmth perception has become more important in a leadership aspect as well (Van Dijk et al., 2017). Since the perception of warmth is for distinguishing whether someone is a friend or not, or whether they are likely to contribute (Van Dijk et al., 2017), it can lead to the assumption that when a leader is perceived higher in warmth, followers may more likely feel inclusion.
According to Nishii (2013), to create an inclusive work climate, equal measures need to be complemented by a change in interaction patterns. So, to integrate all aspects of inclusion, their framework covers three dimensions of inclusivity, which are “fairly implemented employment practices” (creating a level playing field), “integration of differences” (openness and representation of self), and “inclusion in decision making” (democratic decision-making processes) (Nishii, 2013, p. 1756-1757). The aspect of recognising and being open to own values aligns with Shore et al.'s (2011) uniqueness element of inclusion. Thus, in order to feel
included literature emphasises the importance of promoting equality and being welcoming but avoiding forced conformity as people also need the feeling of being themselves for the feeling of inclusion.
Hence, the feeling of inclusion is mostly based on several types of interactions. Next to that, the feeling of uniqueness and belongingness should be achieved to also feel included. So, since this paper focuses on the connection between warmth and inclusion, it can be argued that positive interpersonal relations such as communion and being able to collaborate do not only go hand-in-hand with how warm we perceive others but it is also needed for feeling part of a group, thus feeling included. It is important to mention that this paper focuses on the anticipated feelings of inclusion since participants of this survey-based study just imagine what they would experience if they worked at the organisation instead of being an employee. Therefore, the second hypothesis is the following.
Hypothesis 2: Warmth perception is positively related to the perceiver’s anticipated feelings of inclusion.
For warmth perception to work as a mediator, its indirect effect on anticipated feelings of inclusion needs to be compared to the direct effect between positive non-verbal behaviour and inclusion. As mentioned before, due to the diversifying workforce organisations pay more attention to managing it and the inclusion of employees gains more recognition in managerial research as well (Avery, 2011; Buengeler et al., 2018; Van Knippenberg et al., 2013).
According to research on inclusive work climate, it is not enough to have a level playing field and increase representation as norms emphasising openness, voice, and active integration are also crucial (Ashikali et al., 2021; Nishii, 2013). Thus, inclusive leadership is also gaining more importance as it is required for the proper appreciation of diverse members, the creation of diversity management, and the creation of inclusivity (Ashikali et al., 2021; Ashikali &
Inclusive leaders are characterised by showing support, providing resources and information, and encouraging contributions (Carmeli et al., 2010; Randel et al., 2018). The feeling of inclusion leads to organisational commitment, higher levels of creativity, group identification, empowerment, better job performance, and decreased turnover (Ashikali &
Groeneveld, 2015; Carmeli et al., 2010; Randel et al., 2018). For group identification, it is important to feel inclusion for which perceived warmth acts as a cue (Wagoner & Hogg, 2016).
Therefore, since perceived inclusion is mainly created and facilitated by communication, which consists of both verbal and non-verbal elements, it is predicted that displaying positive non- verbal behavioural cues lead to the anticipated feeling of inclusion. Thus, the following hypothesis is predicted.
Hypothesis 3a: Positive non-verbal behaviour (smiling) is positively related to the anticipated feeling of inclusion.
Hypothesis 3b: Positive non-verbal behaviour (open posture) is positively related to the anticipated feeling of inclusion.
2.4 Match in race
Managerial and psychological research has already found that race, gender, and ethnicity affect people’s attitudes through stereotypes, and influence how we socialise (Kammeyer-Mueller et al., 2011; Livingston et al., 2017). The initial categorisation of others mostly depends on perceived similarity or dissimilarity (Harrison et al., 1998). Literature often differentiates between two types of similarities, which are surface-level (including characteristics that are reflected through physical features such as age, race/ethnicity, and gender) and deep-level characteristics (referring to traits that are not observable but need to be learned through communication such as attributes, values, skills) (Harrison et al., 1998;
Kammeyer-Mueller et al., 2011).
According to literature, how comfortable people feel around each other facilitates their relationship building which can explain why people are more likely to orientate towards demographically similar groups (Barsade et al., 2000; Kammeyer-Mueller et al., 2011).
Kammeyer-Mueller et al. (2011) found that in a work setting, the newcomers who perceive more surface-level similarity with colleagues are more proactive in socialisation such as building relationships and seeking information. Contrarily, perceived similarity in terms of ethnicity was not significant in their study (Kammeyer-Mueller et al., 2011). On the other hand, Tsui et al. (2002) found that subordinates' similarity to supervisors on demographic attributes (including race, gender, age, etc.) facilitates interpersonal attraction and leads to better supervisory ratings on subordinates’ extra-role behaviour (Tsui et al., 2002).
Research also showed that in situations of leader-member exchange (LMX), the surface- level similarity is an important factor (Pichler et al., 2019). As mentioned before, deep-level similarity is getting more recognition, but it is often connected to surface-level similarity (Barsade et al., 2000; Kammeyer-Mueller et al., 2011). Pichler et al. (2019) found that when leaders and members share the same nationality it will be positively connected to the leader’s perceived value similarity due to Tsui et al.'s (1992) notion that physical characteristics are used as information about personality. This essentially leads to a better exchange between the parties and increased performance of employees (Pichler et al., 2019). This leads to the assumption that similarity in surface-level characteristics such as race positively affects socialisation and initial liking (e.g., giving better performance reviews). Thus, as perceived warmth also covers interpersonal interactions and sympathy, similarity in race may also lead to a change in warmth perception as it becomes stronger when there is a match.
This can be explained through the Social Identity Theory which states that people tend to identify with other members whom they perceive as similar to themselves, who then are categorised as in-group, while others become part of out-group (Tajfel et al., 1979).
Additionally, a social comparison exists between in and out-groups, called positive distinction where individuals have a more positive outlook on in-group members (Tajfel et al., 1979). This categorisation also gives people a sense of belonging to the social world (Hogg, 2001). This can be further supported by research done in the context of virtual intelligence. Bergmann et al.
(2012) tested people’s first impressions regarding warmth and competence on robot and human- like virtual agents. They have found that although robot-like agents are also perceived as warmer, this decreases over time while in the case of human-like agents the level of perceived warmth remains steady over time (Bergmann et al., 2012).
Furthermore, individual status is found to be influencing non-verbal behaviour (Dovidio et al., 2006). As societies usually have a structure of hierarchy, the group identity of people can translate to a level of status (e.g., in the US, “Whites” have had traditionally greater social power and status compared to “Blacks”) (Dovidio et al., 2006). Thus, Dovidio et al. (2006) hypothesised that members of oppressed groups are more attentive to the social environment which leads to being better at understanding non-verbal cues. This is supported by the findings of Elfenbein and Ambady (2002) stating that emotions are better recognised by people who share the same ethnic, national, or regional background due to having an in-group advantage.
An interesting addition to that is that the advantage decreases when different groups are more exposed to each other (Elfenbein & Ambady, 2002). Additionally, they found that members of majority groups have more difficulties in decoding the emotions of minorities compared to the other way around (Elfenbein & Ambady, 2002).
Lastly, as mentioned before, in SCM there are some groups of people based on their ethnicities as well who are stereotypically categorised on some level of warmth or competence.
For instance, Asians are put in low warmth and high competence, Hispanics are usually low competence and mid-level warmth, Black professionals are high competence and warmth as well as Whites in general (Cuddy et al., 2008, 2009; Fiske et al., 2002; Lee & Fiske, 2006). On
the other hand, this specific matrix of categorisation is most prevalent in Western cultures, where most of these studies were done (Cuddy et al., 2008). Nevertheless, there is an overall in-group favouritism examined in SCM (Cuddy et al., 2009; Fiske et al., 2002) on which this thesis focuses. Contrarily, due to the aforementioned stereotypes, it may also happen that the race of the non-verbal actor overpowers the perceptions such that White and Black actors are perceived as warmer especially compared to Asians regardless of having the same ethnicity as the perceiver.
Thus, due to the effects of the social identity theory, surface-level similarity, and especially in-group advantages and favouritism, the match between the race of perceivers and non-verbal actors is expected to have a moderating effect on how warm they may perceive the positive non-verbal cues displayed, as in members of the same group are more likely to perceive the other warmer. Therefore, the hypotheses are the following.
H4a: Perceiver’s race moderates the relationship between smiling and warmth perception as it becomes stronger when the race of the perceiver and non-verbal actor matches.
H4b: Perceiver’s race moderates the relationship between open posture and warmth perception as it becomes stronger when the race of the perceiver and non-verbal actor matches.
2.5 Conceptual model
The conceptual model addressed by this paper is outlined in Figure 1. The first hypothesis concerns the proposed positive relationship between certain non-verbal behaviours and the perception of warmth. Then, it will be tested whether there is a significant positive relationship between perceived warmth and anticipated feelings of inclusion. For perceived warmth to work as a mediator, the direct relationship between non-verbal cues and anticipated feelings of inclusion will also be tested. Finally, the fourth hypothesis investigates the moderating effect of perceiver’s race between non-verbal behaviour and warmth perception,
since according to literature surface-level similarity between people and in-group favouritism may affect how they categorise and socialise with each other.
This section outlines the methodology of the study. First, the sample is described along with its accompanying demographic data. Second, the research design and procedure are explained, by outlining the two main phases of the study namely, the video coding and surveying. Third, the various measurements and control variables are described that are relevant for this study. Finally, the data analysis part provides an explanation of how the results are going to be reached.
This study did not have a specific population of interest due to focusing on perceptions of others. For this quantitative study a non-probability, convenience sampling method was used.
Participants were found by sharing the link to the survey with students’ friends and relatives asking them to share the survey with further acquaintances as well. This means that based on the thesis project group’s composition, mostly Dutch and Hungarian people participated. The link to the survey was also shared on various social media platforms (e.g., LinkedIn, Instagram,
WhatsApp groups) to reach a greater audience. In order to participate, respondents first had to agree to answer questions honestly.
The final sample consisted of 308 participants. After the initial cleaning and the removal of the missing values, the final dataset consisted of 230 full responses yielding a response rate of 74.67%. Since not every participant filled in all the demographic questions, there are some differences in the total number of answers in the sample overview (Table 1). The sample consists of 93 male (40.4%), and 127 female (55.2%) respondents. Most participants (47%) indicated Caucasian/White ethnicity, 4.8% Asian, 3.5% Hispanic, 1.3% African American/Black, and 39.6% indicated Other. Most of the respondents (33%) were in their twenties, and the majority of them (40.9%) indicated to be college graduates, thus having a bachelor’s degree.
Demography Frequency Percentage
Gender (221) Male 93 40.43%
Female 127 55.22%
Other 1 0.43%
Age groups (217) 10-19 14 6.09%
20-29 76 33.04%
30-39 30 13.04%
40-49 38 16.52%
50-59 38 16.52%
60-69 21 9.13%
Ethnicity (221) African American 3 1.30%
Hispanic 8 3.48%
Asian 11 4.78%
Caucasian 108 46.96%
Other 91 39.57%
Demography Frequency Percentage
Education (221) Some high school or less 10 4.35%
High school graduate or equivalent
Trade, technical, or vocational training
Some college credit, no degree
College graduate 94 40.87%
Some postgraduate work 10 4.35%
Postgraduate degree 54 23.48%
Leadership position (219)
Yes 69 30.00%
No 150 65.22%
Note: This table categorises the 230 respondents according to their demographic data, namely gender, age, ethnicity, education level, and whether they are in a leadership position.
3.2 Research design and procedure
Since this study focuses on non-verbal cues, perceptions of people shown in videos are examined. For this first, the videos used had to be coded by the researchers which will be outlined in the coding procedure section. Then, the survey procedure is described.
3.2.1 Coding procedure
Videos can act as primary data sources for observational purposes which aligns well with this study’s aim of examining non-verbal behaviour (Baecker et al., 2007). So, in Phase 1 of the research, members of the thesis project group coded some videos in order to determine what kind of non-verbal cues were done by the recorded individuals and at what frequency. In total, the group analysed 300 video clips. The videos were gathered from US-based companies showing speakers of various gender and ethnicity in managerial roles. The six members of the thesis project group were divided into pairs and were assigned 100 videos to analyse with their partners to ensure a level of triangulation. This way due to more researchers examining the same videos and discussing them, reproducibility and intercoder reliability becomes higher (Halperin & Heath, 2017).
The videos presented by the supervisor were created in a way that three 10-second segments were cut together of entire speeches of CEOs since such a segment is enough to capture perceptions of individuals (Ambady et al., 2000). Additionally, the sound of the 30- second videos was excluded to be able to focus more on the non-verbal cues of the displayed people. The coding of the non-verbal behaviours was done with Qualtrics survey tool with quantified measures of the cues. In total, 64 non-verbal signs were included in the coding survey (ranging from gazing and scanning to arm and body movements) which were also provided by the supervisor, based on literature in this field (Carney, 2020). By coding non-verbal behaviour with quantitative means, it enabled statistical testing.
3.2.2 Survey procedure
In Phase 2, we conducted quantitative research by using a cross-sectional survey design.
Qualtrics survey tool was used for carrying out the study. The survey was written in English to be able to reach as many respondents as possible. The questionnaire started with a disclaimer granting anonymity and an indication of the approximate time it would take to finish filling out (See Appendix A). The survey continued with a randomly assigned video clip of a corporate leader displaying some non-verbal cues. After watching the videos participants were asked to answer the questions concerning the video.
Then, in the questionnaires, all variables of the participating research students were included in order to maximise the number of respondents. Although, this method means that the final survey is relatively long. This can be an issue since long surveys usually discourage participants from filling them out (Edwards, 2002). Thus, the decision was made to shorten the number of questions as much as possible by using single-item measures and adapting shorter versions of measures (e.g., in case of inclusion which will be outlined in the next section). The final version of the survey consisted of 27 questions starting with questions on the different
variables and finishing with questions on the participants’ demographic information (See Appendix B).
Positive non-verbal behaviour has been measured by coding videos based on the displayed non-verbal cues of the actors. The non-verbal cues selected in this paper to focus on are frequency of smiling and open posture. Smiling was evaluated on 3 dimensions, which are
‘no smiling’, ‘a little’, or ‘a lot’. Posture was evaluated on four dimensions, which are
‘withdrawn’, ‘neutral’, ‘open’, or ‘hard to see’. These non-verbal cues were then summed to create more statistical measures for the analysis.
Perceived warmth was measured using the warmth-scale of four warmth traits based on Fiske et al.'s (2002) work. Items are named understanding, warm, good-natured, and sincere.
The Cronbach’s alpha of the variable is 0.82. An example question is “Based on the video clip that you watched, did you think the person is good natured?”. The answers were indicated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).
Anticipated feelings of inclusion were measured by adapting 6 items out of 15 from Nishii's (2013) statements on inclusion. Two out of each of the three dimensions were chosen in order to cover all main elements of the measure. The Cronbach’s alpha of this variable is 0.93. An example sentence is “Top management probably exercises the belief that problem- solving is improved with input from different roles, ranks and functions is considered”. The answers were indicated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).
Match in race is combined from two ethnicity measures. First, the ethnicity of the respondent is a nominal variable, and it was measured in the demography section with a question on the participant’s ethnicity. Participants were able to choose from the following items: African American, Hispanic, Asian, Caucasian, Other (in case ‘Other’ is chosen there
was some space left to specify in text form). The values of this measure ranged from 1 to 5.
Second, the ethnicity of the individual displayed in the video was provided in the dataset by the supervisor. This was a bivariate measure with 0 = Non-white, 1 = White.
3.2.4 Control variables
To rule out certain alternative effects on the hypotheses, both the participants’ and the displayed individuals’ gender were chosen as control variables. The reason for this lies in the literature on gender stereotypes regarding non-verbal behaviours. According to the findings of Johnson et al. (2008), female leaders are mostly associated with sensitivity while male leaders with tyranny and strength. Additionally, based on the SCM, females are often stereotyped as warmer compared to males (Cuddy et al., 2008, 2009; Lee & Fiske, 2006). This may lead to females shown in the videos being categorised as warmer. Johnson et al. (2008) also showed that feminine participants expect leaders to be more sensitive while masculine individuals expect leaders to be stronger and tyrannical. Although in this research participants were not told that the people in the videos are in leadership positions, participants’ gender may still interfere because of their potential initial expectations. Furthermore, since females usually have to live up to expectations of their warm and communal role, when they are less expressive in this aspect, they get more negatively judged compared to males (Brescoll, 2016).
Male and female differences not only exist in how their non-verbal behaviour is perceived but also in their abilities to decode these cues. Usually, women tend to outperform men in how accurately they perceive non-verbal and emotional behaviour (Hall & Gunnery, 2013; Rosip & Hall, 2004). This could be because women are more expressive in their non- verbal behaviour, for example by smiling and nodding more (Hall & Gunnery, 2013), which may lead to them being able to recognise these cues more easily. Another explanation could be that women are taught to be more sensitive toward non-verbal cues from an early age (Schmid et al., 2011). Thus, female perceivers of this study may have better accuracy in decoding the
nonverbal behavioural cues seen in the videos compared to males. Also, due to gender stereotypes, the displayed individuals in the videos may be evaluated as warmer by the perceivers solely based on their gender.
3.3 Data analysis
IBM’s Statistical Package for the Social Sciences (SPSS) software version 25 was used for carrying out the statistical analyses. First, Likert-scale items from the measurement of warmth were recoded accordingly since values 3 and 4 were interchanged initially. Then new variables were created for the relevant non-verbal behavioural cues which are the frequency of smiling and the type of posture. Thus, the coded data for smiling was merged so that ‘no smiling’, ‘little smiling’ and ‘lot smiling’ became values of 0, 1, and 2 subsequently. Similarly, in the case of posture, ‘withdrawn/closed’, ‘neutral’, and ‘open’ became 0, 1, and 2, while instances coded as ‘hard to see’ became missing values due to not providing information on actual posture. The variables were also kept separately by becoming dummy variables. This method allows checking the effect of the independent variables in different forms to see whether it makes a difference in the relationships with other variables.
Next, the moderating variable had to be created. A new variable was created based on whether there was a match between the participants’ and the displayed individual’s race. Thus, here ‘match’ (1) means the participant was either Caucasian and their video had a ‘White’
individual, or the participant was Asian, African American, or Hispanic with a video displaying a ‘Non-white’ individual. A ‘non-match’ (0) is when this does not happen. Unfortunately, as it can be seen from Table 1, 39.6% of the respondents answered with ‘Other’, where their textual reply indicated European, Dutch, or Hungarian. In the Netherlands, 25.4% of people are foreign-born or have at least 1 foreign-born parent (CBS, n.d.). In Europe, statistics mostly concern minorities, or nationalities and have been done in a “colour-blind” approach for a long
time due to anti-discrimination policies (Simon, 2012). Thus, I decided to have these answers as missing variables since respondents’ race would be difficult to infer correctly.
Then preliminary analysis was done including the testing of the normality, homoscedasticity, and linearity assumptions. Skewness and kurtosis coefficients showed evidence for normal distribution except for the new, merged variable, ‘Smiling Frequency’
where the value of skewness was a little above two. Thus, this variable was transformed by taking its square root which decreases right-sided skewness (Field, 2018). According to the Central Limit Theorem, as the sample size gets larger, the less would normality matter because sampling distribution becomes normal regardless of population (Field, 2018), which would be the case in this study due to having a final sample size of 230.
For testing the hypotheses, correlation and regression analyses were run for each relationship in order to test whether the relationships exist separately (Field, 2018). Then, Hayes’ PROCESS macro extension tool was used to check the model as a whole. More precisely, model 7 of the program was used (moderated mediation with moderated a-path from the independent variable to the mediator) to test all hypotheses at the same time (Hayes, 2017).
This method also allows investigating whether there is a difference in results when comparing regression analyses with PROCESS.
The findings obtained from the statistical analyses are described in this section. First, the descriptive statistics can be seen to give an outline of the results and whether there is initial support for the hypotheses based on correlations between variables. Then, the hypotheses are tested and discussed based on the results of the regression analyses and the PROCESS model analysis. Finally, the results are shortly summarized at the end of the chapter.
4.1 Descriptive statistics
The means, standard deviations and inter-correlations of the main variables can be seen in Table 2. The Pearson correlation shows a significant positive correlation (r = 0.72, p < 0.01) between perceived warmth and anticipated feelings of inclusion. This demonstrates a co- occurrence between the variables which is an indication of initial support for hypothesis 2 (Field, 2018). The remaining main variables have insignificantly low correlations, indicating that there might be little or no relationship between them. This can also forecast similarly insignificant results in the regression and PROCESS analyses.
The control variables yielded some interesting results in the correlation table. The gender of the participant showed no significant correlation between the other variables; thus, it was not used as a control in further analyses. On the other hand, the gender of the displayed individual illustrates a significant negative correlation (r = -0.19, p < 0.01) with the non-verbal behaviour of smiling meaning that women are more likely to smile compared to men. Although this connection is not in the scope of this research, the variable remained as a control for the upcoming analyses because it may influence the outcomes.
Means, standard deviations, and correlation coefficients
Variable M SD 1 2 3 4 5 6 7
1. Smiling 0.18 0.40 1
2. Posture 1.02 0.34 0.07 1
3. Warmth 4.04 1.19 0.10 -0.03 1 (0.80)
4. Inclusion 3.93 1.04 0.10 0.04 0.72** 1 (0.91)
5. Ethnicity match
0.65 0.48 -0.09 -0.01 0.03 0.02 1 6. Gender
1.58 0.5 0.07 -0.08 0.10 -0.05 -0.05 1 7. Gender
0.93 0.25 -0.19** -0.03 -0.05 -0.11 0.03 -0.13 1 Note. N=230 **p<0.01 (two-tailed)
Since the data was collected through a survey, I tested the reliability of the variables that were measured with items. Cronbach’s alpha is the most popular measure of reliability in social and organisational sciences thus, it was used in this case as well (Bonett & Wright, 2015). The results for this can also be seen in Table 2 in brackets and italics. The 4-item measure for warmth yielded an alpha of 0.80 which is in the range of acceptable values. The 6 items chosen to measure inclusion had an alpha of 0.91 which is excellent (Bonett & Wright, 2015). Thus, it can be concluded that reliable data was provided by the survey.
The next step of the analysis is checking the direct relationships between the independent variables of smiling and posture and the dependent variable of anticipated feelings of inclusion. According to hypotheses 3a and 3b, positive non-verbal behaviours of smiling and open posture are positively related to anticipated feelings of inclusion. The results of the analysis can be seen in Table 3.1. First, the control variable gender of the displayed person was tested. The result was not statistically significant (p = 0.11) and explained only 1.2% of the variance. Then I added the corresponding non-verbal behaviours in the next step, and the total variance explained by the model decreased to 0.6%. Furthermore, the beta coefficient of smiling (0.18) and posture (0.11) were positive but very low. Neither smiling (p = 0.32), nor posture (p
= 0.59) had a significant effect on inclusion. Thus, the linear regression does not support hypotheses 3a and 3b. Since the presence of the control variable changed the model’s R2 value it can be assumed that the gender of the displayed individual has a stronger effect on the anticipated feelings of inclusion than the examined non-verbal cues.
Linear regression results for the effects of smiling and posture on anticipated feelings of inclusion
change B SE β t p
Model 1 0.109 0.012 0.012 4.32 0.26 16.82 0.00
Gender video -0.43 0.27 -0.11 -1.60 0.11
Model 2 0.135 0.018 0.006 4.12 0.34 11.95 0.00
Gender video -0.37 0.27 -0.09 -1.36 0.18
Smiling 0.18 0.18 0.07 1.00 0.32
Posture 0.11 0.21 0.04 0.54 0.59
Note. Model 1 is the first step of analysis with the control variable, Model 2 is the second step with the independent variables added where **p<0.01
After that, the non-merged versions of smiling and posture were also tested to see whether the presence of one of their permutations would have better significance in the model or change the explanatory value. The results for that can be found in Table 3.2. All versions were checked separately, then simultaneously by adding one more each time. In the case of posture, no positive change emerged for ‘open’, ‘neutral’, or ‘closed’ even when paired with smiling. In the case of smiling, adding both ‘little’ and ‘lot’ resulted in a little increase in the explained variance, from 1.1% to 1.8%. While this result still does not support hypotheses 3a and 3b, it can be inferred that frequency of smiling has more effect on feelings of inclusion than open posture.
Linear regression results for the effects of a little and lot smiling on anticipated feelings of inclusion
change B SE β t p
Model 1 0.105 0.011 0.011 4.32 0.26 16.72 0.00
Gender video -0.42 0.27 -0.11 -1.57 0.12
Model 2 0.171 0.029 0.018 4.23 0.27 15.66 0.00
Gender video -0.37 0.27 -0.09 -1.35 0.18
Smile little 0.33 0.19 0.12 1.76 0.08
Smile lot -0.45 0.52 -0.06 -0.86 0.39
Note. Model 1 is the first step of analysis with the control variable, Model 2 is the second step with the independent variables added where **p<0.01
4.3 Moderated mediation model
After executing Hayes’s PROCESS Model 7, the results for that can be seen in Tables 4.1 and 4.2 separately. The analyses focus on whether hypotheses 1, 2, and 4 are supported.
First, the non-verbal behaviour of smiling frequency was tested, and then the non-verbal behaviour of posture was set as the independent variable. For running the extension, the categorical nature of smiling and posture was indicated. Additionally, the gender of the displayed individual remained a control for the whole analysis.
The first part of Table 4.1 exhibits the results of the regression with the mediator, perceived warmth as the dependent variable, the independent variable smiling (little or a lot), the moderator match of races, and the interaction between non-verbal behaviours and matching race. The variable smiling a lot had a significant effect on perceived warmth (B = 3.23, p = 0.01). Since a little smiling had less significant effects (B = 0.49, p = 0.26), it can be inferred that a higher frequency of smiling leads to a higher perception of warmth which supports hypothesis 1a. It turned out that there is no moderating mediation effect as the interactions were not significant (p = 0.97, 0.19). Thus, hypothesis 4a is not supported. The second part of the
table shows the results from the regression of the non-verbal behaviours and perceived warmth on anticipated feelings of inclusion. Warmth perception shows a significant positive effect on inclusion (B = 0.58, p = 0.00) which confirms hypothesis 2.
Moderated mediation results of PROCESS Model 7 (with IV: smiling)
Perceived warmth (M) Confidence interval
Coefficient SE t p LL UL
Constant 3.90 0.40 9.68 0.00 3.10 4.70
Smile little 0.49 0.43 1.13 0.26 -0.37 1.35
Smile lot 3.23 1.18 2.74 0.01** 0.89 5.56
Race match 0.18 0.24 0.74 0.46 -0.30 0.65
Smile little x Race
match -0.02 0.57 -0.04 0.97 -1.14 1.10
Smile lot x Race match -2.18 1.66 -1.31 0.19 -5.46 1.11
Gender video -0.13 0.39 -0.33 0.74 -0.89 0.64
Note. R2=0.08, F=1.84, p=0.10, *p<0.05, **p<0.01
Anticipated inclusion (Y) Confidence interval
Coefficient SE t p LL UL
Constant 1.61 0.32 5.05 0.00 0.98 2.24
Smile little 0.17 0.17 0.96 0.34 -0.18 0.51
Smile lot -0.21 0.52 -0.41 0.69 -1.25 0.82
Warmth perception 0.58 0.05 10.62 0.00** 0.47 0.69
Gender video -0.15 0.24 -0.64 0.52 -0.62 0.32
Note. R2=0.50, F=31.01, p=0.00, *p<0.05, **p<0.01
Similar to the previous section, the results of Table 4.2 show the regression with the mediator, perceived warmth as the outcome variable, the independent variable posture (neutral and open respectively), the moderator match of races between participants and videos, and the interaction between non-verbal behaviours and matching race. Results of both types of posture show no significance (p = 0.78, p = 0.95) thus, hypothesis 1b is not supported. No moderating mediation can be seen in the case of interaction with posture as well (p = 0.60, p = 0.63), so
non-verbal behaviours and perceived warmth on the anticipated feelings of inclusion, where only warmth shows a significant effect on inclusion (B = 0.59, p = 0.00).
Moderated mediation results of PROCESS Model 7 (with IV: posture)
Perceived warmth (M) Confidence interval
Coefficient SE t p LL UL
Constant 3.94 0.88 4.47 0.00 2.19 5.68
Posture neutral 0.25 0.90 0.27 0.78 -1.53 2.02
Posture open -0.06 1.07 -0.06 0.95 -2.18 2.05
Race match 0.56 1.07 0.53 0.60 -1.56 2.68
Posture neutral x Race
match -0.57 1.10 -0.52 0.60 -2.75 1.60
Posture open x Race match -0.63 1.31 -0.48 0.63 -3.23 1.96
Gender video -0.12 0.41 -0.30 0.76 -0.93 0.69
Note. R2=0.01, F=0.22, p=0.97, *p<0.05, **p<0.01
Anticipated inclusion (Y) Confidence interval
Coefficient SE t p LL UL
Constant 1.74 0.42 4.20 0.00 0.92 2.56
Posture neutral -0.11 0.30 -0.39 0.70 -0.70 0.47
Posture open 0.09 0.36 0.26 0.80 -0.62 0.80
Warmth perception 0.59 0.05 10.97 0.00** 0.48 0.70
Gender video -0.20 0.23 -0.85 0.40 -0.66 0.26
Note. R2=0.51, F=30.50, p=0.00, *p<0.05, **p<0.01
Finally, since the PROCESS models indicated a promising relation between perceived warmth and anticipated feelings of inclusion, I ran a separate linear regression analysis to check this effect separately (Table 5). Results here show a significant positive relationship (β = 0.72, p = 0.00) as well as it explains 52.3% of the variance. Thus, as expected through the analyses, hypothesis 2 is supported.
Linear regression results for the effect of warmth perception on anticipated feelings of inclusion
change B SE β t p
Model 1 0.105 0.011 0.066 4.32 0.26 16.72 0.00
Gender video -0.42 0.27 -0.11 -1.57 0.12
Model 2 0.726 0.527 0.523 1.67 0.25 6.75 0.00
Gender video -0.28 0.19 -0.07 -1.49 0.14
Warmth 0.62 0.04 0.72 15.49 0.00**
4.4 Summary of results
Based on the statistical analyses of the previous section it can be said that the relation between the frequency of smiling and perceived warmth is significant as hypothesis 1a was supported. On the other hand, I did not find support for posture openness and therefore, cannot conclude that posture openness has an effect on perceived warmth. In line with hypothesis 2, I found support for the positive relationship between perceived warmth and the anticipated feelings of inclusion. The analysis did not show support for hypotheses 3a and 3b. This indicates that neither smiling frequency nor openness of posture is related to anticipated feelings of inclusion. This also means that warmth cannot be seen as a mediator due to not having a direct relationship in the first place. Finally, I did not find moderated mediation during the analyses so both hypotheses 4a and 4b remain unsupported. In Figure 2, two conceptual models, one for each non-verbal behaviour visualise the coefficients and their levels of significance.
Conceptual models including statistical results where *p<0.05, **p<0.01 (two-tailed)
The goal of this paper was to examine whether a positive relationship exists between positive non-verbal behaviours and the anticipated feelings of inclusion through the perception of warmth and whether it changes when there is a match between the race of the perceiver and the non-verbal actor.
This study provided partial evidence on the positive relationship between positive non- verbal behaviours and warmth perception (H1a and H1b). The frequency of smiling turned out to be a better indicator of this relationship than the openness of posture. There can be several explanations for this outcome. It can be said that the findings of this paper align with Scharlemann et al.'s (2001) and Bellou and Gkorezis's (2016) studies of smiling being related to cooperation, trust, and friendly intentions which are elements of someone perceived as warm.
Interestingly this conclusion goes against a finding stating that smiling does not always lead to positive judgements, such that extensive smiling can be seen as unprofessional in less people- oriented jobs (Min & Hu, 2022). The reason for that can be that the respondents of the survey were not told beforehand that they would see CEOs in the videos. So, it can be inferred that if the context is not entirely clear for the perceiver, smiling stays connected to positive perceptions. This could be further backed up by the findings that due to the salience of mouth movements, individuals immediately evaluate others as happy and sociable (Calvo et al., 2012;
Cunningham, 1988). Similarly, warmth is also quickly evaluated by people (Cuddy et al., 2011), thus the salience of both smiling and warmth perception can be an explanation for the results of this study.
The non-findings in the case of posture openness are closer to the conclusion of Mehrabian (1968) stating that there is more interaction between different non-verbal behaviours leading to perceptions instead of open posture in itself being salient. Other research also refers to more types of non-verbal behaviours such as nodding, eye contact, body orientation, etc.
(Sundaram & Webster, 2000) which can add to the assumption that posture can be an element of an interaction effect. In their research Cuddy et al. (2011) rather emphasised leaning forward and positioning in direction to the audience as positives and tenseness as cold. Thus, even though tense body posture was found to be perceived as cold it does not necessarily mean that open posture is automatically warm. Furthermore, open posture is related to a lot of other perceptions and feelings other than warmth, like pride and love which can have an effect on the non-findings as well (Sauter, 2017). Finally, literature also says that individuals of high status or with higher levels of dominance tend to adopt a more open posture (Carney et al., 2005;
Peters et al., 2017). Hence, this can indicate a shift in how individuals perceive open posture and connect it with dominance which is more related to the competence branch of SCM (involving competence and status) (Cuddy et al., 2008).