5 Discussion

↓111

The following chapter discusses the results and implications of the current study. First, Section 5.1 discusses how the current findings relate to theoretical conceptions regarding the measurement and process of mutual understanding. After this, Sections 5.2 and 5.3 review the outcomes of the main and dyadic effect hypotheses, respectively, and how the findings from the current study relate to the theoretical framework discussed in Chapter 2. Section 5.4 discusses the combined influence of main and dyadic effects in intellectually gifted individuals. After this, limitations of the current dissertation are reviewed in Section 5.5, followed by an overview of strengths in Section 5.6. Finally, general conclusions and recommendations for future research are discussed in Section 5.7.

5.1  The Process of Mutual Understanding

As stated in Chapter 1, MU has not received much attention in psychological research. In social relationship science, a number of related constructs such as empathy, social support, rapport, and intimacy exist. In addition, communication research has identified a number of mechanisms that facilitate the process of MU, such as audience design, perspective taking, and reference to shared experiences. In the following, the present dissertation’s findings regarding the MU process are reviewed critically. A more thorough understanding of the mechanisms behind this process will serve as a background for the subsequent discussion of the findings regarding main and emergent effects in Sections 5.2 and 5.3.

5.1.1  Discriminant Validity of MU

In the introduction, it was contented that MU can be an important first step in the establishment of social support, rapport, and intimacy. However, it was also argued that MU can occur without being followed by social support and feelings of relational harmony, especially in the early stages of relationship formation. Sample 4 provides a good opportunity to test this notion, because it involved two previously unacquainted persons interacting with each other. Although it is not unlikely that these persons would develop a deep sense of understanding of each other’s life domains, it is less probable that they would experience a deep emotional connection to each other during such a limited time.

↓112

Results partially supported the relative independence of MU and emotional closeness. For example, a factor analysis of the post-interaction questionnaire showed that items related to interviewers’ perceived understanding of the interaction partner (i.e., comprehending what is said, providing enough room to explain things) load on a different factor than items that tap into more emotional, rapport-like aspects of communication (i.e., feeling relaxed, experiencing a smooth conversation). This replicates a finding by Emmers-Sommer (2004), who found two factors after factor-analyzing experience sampling reports of social interactions: a Depth factor describing personal and in-depth relationships, and a Smoothness factor describing a relaxed communication free of conflict and breakdowns.

In contrast to the interactions between strangers, evaluations of more established relationships in Samples 1, 2, and 3 showed a lower level of differentiation. When participants assessed their network partners with the help of an ego-centered social relationship instrument, items tapping successful communication and understanding (which are most closely related to MU) loaded on the same general factor as more generalized assessments of closeness, importance, support, and acceptance. From these findings, it may seem that people do not differentiate between MU and broader measures of relationship quality. Note, however, that the high degree of association between these different indicators is exactly what would be predicted by the theoretical framework outlined in Chapter 1. That is, because MU was hypothesized to be a first prerequisite for the establishment of intimacy, MU and more emotional aspects of relationship quality should become increasingly correlated as relationships develop.

5.1.2 Convergent Validity of MU Assessments Across Information Sources

One of the potential critiques of the MU construct is that the phenomenon is too subjective to be of much use as a variable in nomothetic research. If MU is so idiosyncratic that it lacks an “objective” foundation, it may be impossible to test an individual’s claim that he or she is not properly understood. Because Studies 1.II and 4 involved multiple information sources for the level of MU, it is possible to assess the convergent validity of the current operationalizations. The present results indicated a moderate level of agreement. In the first half of the conversation in Study 4, the correlation between interviewer and interviewee reports of MU reached a level of .31, and in the second half, the correlation was .26 (ps = .05). Although this may seem somewhat low, it is comparable to the .37 correlation between self-ratings of rapport found by Bernieri et al. (1996). It is also comparable to the correlation of .23 (p = .01) between the dyadic partners’ composite ratings of relationship quality in Sample 1.II.

↓113

In addition to the agreement between interviewer and interviewee, there was also a moderate level of agreement between participants’ ratings and judgments based on behavioral observations. Specifically, across both interaction halves and interview roles, the average correlation between subjective ratings (understanding as interviewer and interviewee, separately for both interaction partners) and behavioral observations after r-to-z transformation and back-transformation was .30 (ps ≤ .10).47 These levels are comparable to the agreement level of .24 (single coders) and .30 (pooled across N = 45-52 coders) reported by Bernieri et al. (1996). Together, the current results imply that MU is an intersubjective phenomenon that can be reliably inferred from the behavioral stream, though the moderate size of the convergent correlations points to idiosyncratic influences as well.

5.1.3 Chronological Correlates of MU

According to Clark (1992), the initial phase of a conversation is used to establish a „common ground” of shared interpretational knowledge. After this common ground has been set up, it should become easier to understand the utterances of the other person, and the level of MU should go up. Because establishing common ground requires time, MU should be positively related to interaction duration, at least up to a certain point. This expectation could be tested in Sample 4. Specifically, interaction duration was operationalized both as interaction half (as the participants were already somewhat familiar with each other in the second interaction half) and the number of 30-second intervals that had elapsed since the beginning of the interaction. Because of the hierarchical nature of the data, this analysis was carried out with HLM. Results indicated a significant effect of both interaction half (b = .34) and time elapsed (b = .14; both ps = .01)48, whereas the interaction term was not significant (p = .35). This is consistent with the interpretation that as participants in Sample 4 had more time to establish a common ground, they also became better at understanding each other’s utterances.

One of the strongest associations over time found in the current study was the correlation between MU measured across interaction halves. Specifically, the aggregate behavioral judgment correlated .74 across the two halves, and felt understanding as interviewer and interviewee correlated .66 for the person that started the interaction as interviewee (ps = .01). Thus, more than 40% of the variance of MU in the second half of the conversation could be explained by the level of MU in the first half of the conversation. Apparently, participants established a sense of MU that transcended the specific interaction half or interaction role they were involved in. This provides strong evidence that MU is a truly emergent phenomenon.

↓114

Somewhat in contrast, the ratings of the person that started the conversation as interviewer were „only” correlated .27 (p = .05). The difference between these correlations was significant, z = 2.75, p = .01. Apparently, first interviewers had a more differentiated view of the unfolding interaction, whereas first interviewees transferred their impressions of the first interaction half to the second half. As described in Section 4.1.2, this is probably due to the fact that the first interviewer could discriminate between his own (covert) subjective understanding and (overt) communicative behavior, whereas the first interviewee could not. This probably made the judgments of the former person less susceptible to the influence of generalized positivity biases that may have contributed to the high timely stability of the impressions of the second interviewer. This suggests that the participants’ „starting position” shaped their perceptions of the interaction process as the conversation unfolded (this point is further discussed in Section 5.7.2).

5.2 Discussion of Main effects

After discussing the current dissertation’s findings regarding the measurement and process of mutual understanding, the following section reviews the outcomes of the three main effect hypotheses and their implications for the theoretical framework outlined in Chapter 2. Main Effect Hypothesis 1 predicted that fluid intelligence should be related to higher levels of MU. Main Effect Hypothesis 2 predicted that vocabulary is positively related to MU. Finally, Main Effect Hypothesis 3 predicted that openness to experience is positively related to MU. Table 19 shows the results of the tests of the main effect hypotheses, grouped by variable, sample, and assessment method. In the following sections, results for each of these hypotheses are discussed in more detail.

5.2.1  Main Effects of Fluid Intelligence on MU

According to the theoretical background depicted in Chapter 2, fluid intelligence facilitates the use and interpretation of contextual cues in encoding the meaning of verbal messages and a fluent encoding of ideas. Accordingly, Main Effect Hypothesis 1 predicted that fluid intelligence is positively associated with the level of MU. As can be seen in Table 19, this hypothesis was generally confirmed when participants rated both their network partners’ intelligence and the quality of the corresponding relationship. The only exception was Sample 4, where partner-ratings of intelligence were only related to MU when the first interviewer was the target, not the second interviewer. Overall, consistent evidence for a perceived (within-person) association between ratings of partner intelligence and relationship quality was found.

↓115

As can be seen in Table 19, however, the picture for Main Effect Hypothesis 1 was quite different when assessment methods other than partner-ratings were used. In Samples 1 and 2, intelligence self-ratings were negatively associated with the overall MU level when the single item rating scale was used, whereas the intelligence self-concept scale did not produce any significant effects. In addition, psychometrically assessed fluid intelligence in Sample 4 was not significantly associated with MU. Accordingly, Main Effect Hypothesis 1 was not supported when intelligence self-ratings or IQ test results were used.

Table 19 Summary of Results Across Main Effect Hypotheses

Sample

Variable

  

Hypo thesis

Support for Hypothesis

1 +2

Fluid intelligence

PR

Positive

Hm-1

Yes

1.II

Fluid intelligence

PR

Positive/none

Hm-1

Mixed

3

Fluid intelligence

PR

Positive

Hm-1

Yes

4

Fluid intelligence

PR

Positive/none

Hm-1

Mixed

1+2

Fluid intelligence

SR

Negative/none

Hm-1

No

4

Fluid intelligence

SR

None

Hm-1

No

4

Fluid intelligence

TEST

None

Hm-1

No

3

Crystallized intelligence

TEST

None

Hm-2

No

4

Crystallized intelligence

TEST

Positive/none

Hm-2

Mixed

1+2

Openness

SR

None

Hm-3

No

3

Openness

SR

None

Hm-3

No

4

Openness

SR

None

Hm-3

No

Note. PR = Partner-ratings, SR = self-ratings, TEST = psychometric test, Hm = Main Effect Hypothesis

5.2.1.1 Lack of Psychometric and Self-Report Effects

The lack of association between psychometric assessments and self-reports of intelligence and MU seems to contradict previous accounts that intelligence is an important factor in social understanding (Davis & Kraus, 1997; Guilford, 1967). Because fluid intelligence is defined as the ability to solve novel problems, most studies have used artificial problem situations to test the impact of intelligence on social skills. However, social interactions in everyday situations may not involve a great deal of reasoning about novel stimuli. For example, Kellermann and colleagues (Kellermann & Lim, 1990; Kellermann & Palomares, 2004) have repeatedly argued for the existence of schema-like routines that guide the choosing of topics in getting-acquainted conversations. Indeed, according to Kellermann and Lim (1990, p. 1178), conversation is “a relatively simple structure composed of a limited number of scenes, organized into a limited number of subsets, progressed through in a normative manner [...]”Such a pre-structured process may not require much elaborate cognitive processing facilitated by fluid intelligence (see Section 5.7.4.4 for a discussion of alternative situational properties that may produce an association between intelligence and MU).

5.2.1.2 Presence of Perceived Effects

↓116

In contrast to the lack of effects for self-reported and psychometrically assessed fluid intelligence, a consistently positive effect was found for ratings of partner intelligence. When individuals in Studies 1-3 rated their partners as more intelligent, they also perceived the quality of the corresponding relationship as closer and more understanding. In addition, if the first interviewer in Sample 4 was seen by his or her interview partner as more intelligent, the amount of MU (assessed by self-reports, partner-reports, and behavioral observations) was significantly higher. As reviewed above, however, no evidence for an association between MU and psychometric and self-rated intelligence was found. This creates the need to explain the positive association between ratings of partner intelligence and MU. What led participants to perceive a relation between MU and their interaction partner’s intelligence level?

The most likely interpretation for this finding is the presence of a halo effect. That is, because intelligence and MU and other indices of relational quality are highly positive valenced, participants may confuse a favorable impression on one dimension with favorable impressions on the other. Indeed, previous research has provided evidence that intelligence ratings are confounded with feelings of interpersonal liking (Paulhus & Landolt, 2000). This may be due to the fact that people attribute a higher level of intelligence to sympathetic others or perceive intelligent others as more likeable.

Whereas intelligence and relationship quality were positively associated in Samples 1.I, 2, 3, and 4, in Sample 1.II such an association was found only when Mensa members provided both ratings. In contrast, when a network partner provided the ratings of either intelligence or relationship quality, no effect was found. Probably, this difference is due to a method artifact. Recall that Mensa members rated the intelligence and relationship quality for all contact persons, whereas contact persons rated only one person: the Mensa member. It may be that the availability of multiple rating targets made it easier for the Mensa members to differentiate between network partners and thus enhanced validity of both intelligence and relationship quality ratings.

↓117

The interpretation that the null results for the Mensa members’ network partners is due to psychometric problems is backed up by the fact that the network partners provided higher and more restricted ratings of relationship quality than the Mensa members (M = 4.17, SD = . 30 and M = 3.98, SD = . 55, respectively). The differences between the means and standard deviations of the two groups were statistically significant, t(173) = 4.30, p < .01 for the means, and F(173, 173) = 1.89, p < .01, for the SDs. Thus, the lack of association between intelligence ratings and relationship quality for network partners can partly be attributed to a ceiling effect and range restriction in network partners’ ratings of relationship quality.

5.2.2 Main Effects of Crystallized Intelligence on MU

Main Effect Hypothesis 2 predicted that vocabulary size, a facet of crystallized intelligence, is positively associated with MU. This hypothesis was not supported in Sample 3 and partially supported in Sample 4 (see Table 19). In the latter sample, conversations in which the first interviewer had a larger vocabulary were rated as higher in MU. In the following section, the theoretical implications of this positive association are discussed (possible reasons for the lack of effects in the case of Sample 3 and the second interviewer in Sample 4 are explored in Section 5.7.2).

5.2.2.1 Positive Effect of Crystallized Intelligence

As stated above, the fact that the first interviewer’s crystallized intelligence had a positive effect on MU in Sample 4 is consistent with Main Effect Hypothesis 3. According to the theory described in Section 2.2.2, having a larger vocabulary should be positively associated with both encoding and decoding ability. On the encoding level, research has found that having a large number of cognitive constructs is associated with higher communication effectiveness (O’Keefe & Sypher, 1981). On the encoding side, it is well-established that vocabulary knowledge is a crucial first step in verbal comprehension (Cain, Oakhill, & Lemmon, 2004). Because both encoding and decoding effectiveness should lead to higher levels of MU, the current results point to the important role of crystallized intelligence in verbal communication.

↓118

Results from Study 4 are consistent with the dual function of crystallized intelligence in facilitating both encoding and decoding processes. That is, when the first interviewer had a larger vocabulary, his or her interview partner not only reported a higher level of felt understanding as interviewee (r = .30, p = .01), but also as interviewer in the second interaction half (r = .26, p = .03). Of course, it is possible that both correlations merely reflect a generalized sense of satisfaction with the conversation, regardless of the conversation half and the ordering of interviewer positions. Indeed, recall that the correlation between the second interviewer’s felt understanding as interviewer and interviewee was .66, which was highly significant (p < .01). This strong association was not mediated by the first interviewer’s vocabulary level, as controlling for this factor did not affect the correlation (partial r = .64, p < .01). Accordingly, it seems possible that the second interviewer’s higher sense of MU was already firmly established in the first half of the interaction as a result of the superior interviewing skills of the first interviewer and not further enhanced by the latter’s superior encoding skills when acting as interviewee.

In the current dissertation, no association was found between psychometrically assessed fluid intelligence and MU, whereas there was a significant relation between the first interviewer’s crystallized intelligence and MU in Sample 4. It should be noted that previous research has mostly ignored this distinction when reporting associations between mental ability and social skills (e.g., Davis & Kraus, 1997). In spite of strong empirical associations between fluid and crystallized intelligence (Carroll, 1993), there exist plausible reasons why they may be differentially associated with the ability to understand the utterances of other people. A link to theories regarding the decoding of nonverbal displays can be helpful here. An important concept in this field are so-called nonverbal decoding rules, which are defined by Buck (1983, p. 217) as “cultural rules or expectations about the attention to, and interpretation of, nonverbal displays”. As cultural rules seem at least as important in the decoding of verbal messages (in terms of language, conventional expressions, etc.), such “crystallized” decoding rules may be more important in verbal communication than “fluid” reasoning about the meaning of novel communicative utterances.

Scattered empirical evidence is consistent with the notion that skills that facilitate interpersonal understanding are more dependent on crystallized intelligence than on fluid intelligence. For example, MacCann, Roberts, Matthews, and Zeidner (2004) reported that a performance measure of emotional intelligence is correlated with crystallized but not with fluid intelligence (also see Ciarrochi, Chan, and Caputi, 2000). Davis and Kraus (1997) reported an average correlation of .23 between observed empathic accuracy and „intellectual functioning“ (a category not further specified but probably including both crystallized and fluid intelligence). By comparison, a slightly higher average correlation of .27 was reported between empathy and “cognitive complexity”, a construct usually operationalized as the number of verbal dimensions used to describe stimuli and thus akin to crystallized intelligence. Because this evidence is highly indirect, however, more studies are needed to address the hypothesis that crystallized and fluid intelligence are differentially related to MU.

5.2.3 Main Effects of Openness to Experience on MU

↓119

According to Main Effect Hypothesis 4, openness to experience should be positively associated with MU because open individuals are better able to cope with the unstructured, ambiguous process of tailoring their messages to the knowledge background of their interaction partners. Results across all samples overwhelmingly disconfirmed this hypothesis: In no case was openness to experience significantly associated with the level of MU, regardless of the instrument used to measure openness (BFI vs. NEO-FFI), the type of social relationships that were assessed (strangers vs. well-established social relationships), or the sample that was used (Mensa members vs. university alumni/students). In the following, an explanation for this lack of association is offered.

5.2.3.1 Explanation for Inconsistency with Previous Literature

One of the empirical foundations of Main Effect Hypothesis 4 was a study by Richter and Kruglanski (1999), who reported that descriptions of abstract figures by more open participants are more likely to be recognized by “naïve” individuals. Specifically, their descriptions are more detailed and contain less idiosyncratic references, presumably making them easier for others to understand. These results provide indirect evidence that open people are more effective actors in the process of audience design. In contrast, the current dissertation found no significant main effects of openness on MU. What may have caused this discrepancy?

The most likely explanation seems to lie in differences in experimental setting between the two studies. Specifically, participants in Richter and Kruglanski’s (1999) experiment only read other people’s written descriptions and thus did not interact directly with each other. This made it impossible for them to rely on dynamic discourse cues, such as back-tracking responses, facial feedback (e.g., a puzzling look), and clarification questions. Such a task likely makes the audience design process a lot tougher, but it may not necessarily be representative of more everyday-like conversations. Also, the focus on abstract figures in their study may have been quite untypical of naturalistic conversations. As described above, everyday conversations may not be directed so much on novel, complex topics but instead proceed in a fairly predictable manner (Kellermann & Lim, 1990). As indicated by the results of the present dissertation, the ability of open individuals to tolerate ambiguity may therefore not be such a crucial asset in more everyday interactions.

5.3 Discussion of Dyadic Effects

↓120

Moving from the level of individuals to the level of dyads, one of the overarching hypotheses of the current study was that between-person differences in intelligence and dispositional valuations limit the degree of overlap in life experiences and the meaning attached to these experiences. Accordingly, dyadic personality differences should be negatively associated with the level of MU. Specifically, it was predicted that differences in fluid and crystallized intelligence, openness to experience, and values would be related to decreases in MU. In the following section, the results regarding these hypotheses are summarized and discussed.

5.3.1  Status of Dyadic Effect Hypotheses

5.3.1.1 Dyadic Effects of Fluid Intelligence on MU

Table 20 summarizes the results of the tests of the dyadic effect hypotheses. As can be seen, results were generally not supportive of Dyadic Effect Hypothesis 1. In line with predictions, (absolute) rated intelligence differences in Sample 2 were negatively associated with relationship quality. However, additional analyses showed that this effect was solely due to the fact that participants regarded relationships with people they judged to be less intelligent as lower in quality; a similar deleterious effect for relationships with more intelligent partners was not found. In addition, negative dyadic effects of perceived intelligence differences in Sample 4 were neither found for rated intelligence nor for psychometric intelligence.

Table 20 Summary of Results Across Dyadic Effect Hypotheses

Sample

Between-person difference

  

Hypo thesis

Support for hypothesis

2

Fluid intelligence

SR

(Negative) a

Hd-1

Mixed

4

Fluid intelligence

SR

None

Hd-1

No

4

Fluid intelligence

TEST

None

Hd-1

No

4

Crystallized intelligence

TEST

None

Hd-2

No

1.II

Openness

SR

None

Hd-3

No

4

Interests

SR

None

Hd-4

No

4

Values

SR

None

Hd-4

No

Note. PR = partner-ratings, TEST = psychometric test, SR = self-ratings
a Relationships with more intelligence partners were not rated as lower in quality

5.3.1.2 Dyadic Effects of Crystallized Intelligence on MU

↓121

According to the theory outlined in Section 1.2, people face difficulties in if they want to communicate words or facts that are not shared by their interaction partners. Accordingly, Dyadic Effect Hypothesis 2 predicted that interpersonal differences in crystallized intelligence are associated with decrements in the level of MU. This hypothesis was tested in Sample 4, which included an assessment of the knowledge level of participants involved in a dyadic interaction. However, no effect of absolute knowledge differences was found, which is inconsistent with Dyadic Effect Hypothesis 2.

5.3.1.3 Dyadic Effects Openness to Experience on MU

According to Dyadic Effect Hypothesis 3, open and closed individuals have highly different “thinking cultures”, which should produce a negative association between between-person differences in openness and their level of MU. However, in Samples 1.II and 4, where this hypothesis was tested, no negative dyadic effects of openness were found. Accordingly, Dyadic Effect Hypothesis 3 was not confirmed.

5.3.1.4 Dyadic Effects Interests and Values on MU

Finally, Dyadic Effect Hypothesis 4 predicted there would be a negative association between interpersonal differences in interests and values on the one hand, and MU on the other. This hypothesis was tested by correlating the Euclidean distance between interaction partners’ interests and values profiles with their level of MU. Inconsistent with Dyadic Effect Hypothesis 4, however, between-person differences in interests or values were not associated with decreases in MU.

5.3.2 Implications of the Lack of Dyadic Effects

↓122

As becomes clear from the above overview, the basic assumption behind the dyadic effect hypotheses was not confirmed in the current study. Of course, there are several alternative explanations for the lack of univariate results, such as methodological weaknesses that limit the generalizability of the current findings (see Section 5.5). Together, however, the pattern of null results suggests that, in contrast to the theory cited in the second chapter, people are actually able to bridge personality differences in their mutual communication. In the following sections, the implications of this conclusion are discussed in more detail.

5.3.2.1 Bridging Personality Differences

As stated above, participants in the current study were able to bridge between-person differences in personality in establishing mutually understanding relationships. According to Byrne’s (1971) similarity paradigm, people have an active preference for interactions with like-minded others. The fact that the current study found similarity in intelligence, openness to experience, interests, and interests/values to be unimportant in shaping MU appears to contradict this notion.

In interpreting the discrepancy between the current results and the similarity paradigm, it should be noted that the similarity paradigm has relied almost exclusively on evidence from the so-called bogus-stranger method. In this method, people are given a fake personality profile that either resembles their own personality profile or is very much different. Results from these studies have overwhelmingly shown that people are more attracted to others with similar personality profiles (Byrne, 1997; Sunnafrank, 1992).

↓123

Although the experimental results obtained with the bogus-stranger method belong to the classical canon of social psychology, their implications for more everyday interactions can be questioned. In a series of well-designed though ill-recognized49 studies, Sunnafrank (1983; 1984; 1992) showed that this effect is only present when people are not allowed to communicate with the unknown person (which is always the case when this is a bogus person). When participants engage in a brief getting-acquainted interaction after being exposed to each other’s personality profiles, similarity no longer exerts an effect.

According to Sunnafrank (1992), the failure to replicate results found with the bogus-stranger method in real-life communication settings is due to its artificial character. That is, each person is assumed to strive for a stable, predictable, and controllable environment. This goal is hypothesized to be threatened when people are to discuss controversial topics with a person they have not met before (i.e., the bogus stranger). For example, imagine a pro-life person who is told he or she is about to meet an unknown pro-choice person. This makes the activation of stereotypes about people of “the other side” more likely (e.g., “a leftist hippy”). In addition, the question may come up whether such a stranger will respect the opinion of the participant. When two persons with very different opinions actually meet, however, these questions are typically resolved rather quickly. After all, except for a few radical persons, most people are able to remain respectful and polite when discussing controversial topics with others.

Of course, some situations are less supporting of such a friendly, polite conversation where every person gets to have its say. In competitive, task-oriented situations, a different interpersonal dynamic may exist. In his 1985 article on intelligence and group influence, Simonton himself acknowledged the possibility that dyadic intelligence differences may only be related to interpersonal understanding in situations where a group needs to find a novel solution to a problem. Such a situation may be typical of many domains (e.g., politics, science). However, communication in emotion-focused groups may be much less affected by this limit on complexity. Thus, the current results do not contradict the Simonton (1985) model but rather questions its applicability to everyday social interactions between acquaintances and strangers.

5.4 Differences Between Gifted and Comparison Individuals

↓124

A corollary of the hypothesis that dyadic intelligence differences impair MU is that people who are extremely intelligent should face difficulties in their communication with others. To test this notion, university alumni and intellectually gifted members of Mensa were compared in terms of their level of social adjustment. Because university alumni are also above-average in intelligence, this comparison provided a stringent test of the existence of adjustment problems for gifted individuals. In line with the Extreme Group Hypothesis, Mensa members reported less satisfying and understanding social relationships. In contrast, no differences were found in terms of more generalized adjustment features, such as general self-esteem and neuroticism. These social differences of the Mensa members were limited to specific relationship categories, most importantly family members and partners. Accordingly, generalized response tendencies cannot be invoked to explain away the differences between the gifted and comparison sample.

According to the theoretical background of the current study, dyadic intelligence differences should be associated with communication decrements. This prediction was not confirmed in the current data. As such, the fact that Mensa members experience more social difficulties than a comparison group of university alumni represents a theoretical challenge. That is, in the current dissertation, only the predicted outcome (adjustment difficulties for gifted people), but not the predicted mechanism (communication difficulties between partners with very different intelligence levels) could be confirmed. Because of this, an alternative explanation for the differences between Mensa members and university alumni must be offered.

5.4.1  Self-Selection

The most likely explanation for the apparent social difficulties of the Mensa members is self-selection. That is, it is possible that gifted individuals who experience social problems are more likely to join Mensa. If this is true, Mensa members’ social problems may not be representative of gifted individuals in general. Some evidence is consistent with this explanation. Recall that the only descriptive studies of gifted individuals that relied on unselected samples are those of Terman in the US (1925; Terman & Oden, 1959) and Rost (2000) in Germany. The gifted individuals in the Terman study were socially well-adjusted and did not report an elevated frequency of mental-health problems. Somewhat in contrast, Rost’s (2000) gifted individuals50 had a somewhat more negative self-view of peer popularity and a lower frequency of meeting friends. Note, however, that this does not necessarily represent a subjective problem for gifted individuals. Indeed, his gifted children may have set other life priorities that led them to voluntarily invest more energy in non-social domains (e.g., by reading more books). Consistent with this argument, they did not have a lower self-concept of peer relationships.

↓125

Compared to the general lack of social problems reported for unselected gifted individuals, some studies examining members of Mensa have found evidence for adjustment difficulties. For example, Taft (1971) studied 244 Australian Mensa members. Consistent with the current finding that Mensa members report less satisfying relationships with family members, one third of his sample described their family life as unhappy, which according to the author is double the figure in comparison groups. Furthermore, Bögels, de Mey, and Derksen (1996) found that a substantial subgroup of Mensa members (22.6%) report a host of psychological problems as evidenced by their responses on the Minnesota Multiphasic Personality Inventory. Anecdotal evidence from the current study is consistent with this evidence. Specifically, one Mensa member emailed the current author to point out that many members discover their intellectual giftedness in psychotherapy. This may have led to an oversampling of individuals with social problems.

An empirical indication that the social problems of the Mensa members are caused by a sampling bias and not by their higher levels of intelligence per se can be obtained by comparing Mensa members with different levels of intelligence. If it is indeed the case that extremely high intelligence levels are associated with adjustment problems, then “extremely gifted” Mensa members should be worse off than “moderately gifted” members. As stated in Section 3.3.2.6, 76 Mensa members reported the result of the psychometric intelligence test they had used to become a member of Mensa. Within this subsample, 37 individuals were assigned to a „moderately gifted” group (IQ ≤ 135, mean IQ = 132.1, SD = 1.3), whereas the remaining 39 individuals were assigned to the „extremely gifted” group (IQ > 135, mean IQ = 138.9, SD = 3.7). This extreme group membership was used as a between-subjects variable in a subsequent ANOVA predicting neuroticism, social and general self-concept, and feelings of loneliness.

Results indicated no significant differences between the two giftedness levels (all ps ≥ .18). In addition, a supplementary HLM analysis of group differences in the parameters of the ego-centered network did not indicate any differences in the quality of relationships with parents, other family members, partners, or friends (ps > 0.26). Thus, no evidence was found that moderately and extremely gifted individuals differ in their level of social adaptation. These results are inconsistent with the notion that as people’s intelligence level becomes progressively more extreme, it is more difficult for them to establish mutually understanding relationships with peers.

5.5 Limitations

↓126

A number of factors limit the generalizability of the conclusions of the current study. Specifically, these factors include the use of subjective assessments to operationalize MU, anomalies in the assessment of intelligence, the use of biased samples, and the correlational nature of the evidence. In the following, these limitations are discussed in more detail.

5.5.1  Subjective Assessment of MU

One of the possible limitations of the current dissertation is the use of subjective impressions to assess MU. In Studies 1-3, MU was assessed by asking participants to evaluate the degree to which they are able to communicate effectively with their network partners and felt understood by them. In Study 4, participants rated the degree to which they understood their interaction partners as well as the degree to which they themselves felt understood during the conversation. In addition, these self-ratings were supplemented by trained judges’ ratings of MU. Yet even in the case of the behavioral observations, judgments were based on the degree of MU as indicated by these interviewee’s reactions and the raters’ own sense of what constitutes an appropriate interviewer reaction.

As stated previously, subjective ratings of interaction quality have been recommended by Bernieri and Rosenthal (1991) as being valid and cost-effective. Nevertheless, the emphasis on subjective impressions may come with some disadvantages when studying a concept like MU. That is, MU requires both partners to grasp the meaning of each other’s thoughts and feelings. In principle, this understanding is a covert, intra-psychic act. That is, it is possible that a person understands the meaning of his or her interaction partner’s utterance without expressing this understanding. Equally possible, a person may overtly express understanding without really knowing what the other person is talking about. When the goal of the interaction is to achieve a smooth conversation, both interaction partners might be quite content with such a state of appearance and thus not bother to verify the expressed understanding. Accordingly, both persons may feel understood without “true” MU.

↓127

One indication that person’s sense of reciprocal understanding may be only a proxy for “true” MU is the moderate convergent validity of the different MU indicators. Specifically, the correlation between different rating sources (first interviewer, second interviewer, behavioral observations) in Study 4 was only about .30. In addition, the correlation between the assessments of relationship quality of Mensa members and their network partners in Study 1.II was .23. Although these convergent validities were all statistically significant, they point to sizable idiosyncratic influences. For example, it may be that some people dispositionally feel more understood by their interaction partners than others (Sarason, Shearin, & Pierce, 1987). In addition, it may be that people are not able to differentiate the feeling of being understood from other intimacy-related feelings of validation and care (Reis, 1990). Finally, it may be that MU is asymmetric between interaction partners, in that an individual understands his or her interaction partner but this understanding is not reciprocated. In any case, such issues can only be addressed with the help of more objective assessment methods (see Section 5.7.4.2 for an example).

5.5.2 Reliability and Validity of Intelligence Assessment

In the current dissertation, a wide range of intelligence assessment procedures were used. In spite of this diversity, some of these procedures suffer from measurement problems. Most importantly, this concerns the intelligence ratings. As stated previously, intelligence self-ratings can be biased by self-serving tendencies (Paulhus Lysy, & Yik, 1998). This is especially worrisome when calculating indices that depend on the difference between intelligence self-ratings and partner-ratings. One indication that this difference score may have been biased is that fact that participants in Sample 2 rated only 4.3% of all network partners as more intelligent than themselves. Of course, this is an extremely low value.

In evaluating these figures, two points need to be kept in mind. First, it may be that participants were actually correct in their impression that they have an above-average IQ (Taylor, Lerner, Sherman, Sage, & McDowell, 2003). After all, most of them were university alumni with high level of academic achievement and probably had above-average intelligence. Second, it may that the tendency to oneself as more intelligent than others is part of a pervasive phenomenon called the above-average effect (Kruger, 1999). Of course, if this tendency to judge one’s own level of skills as higher than other people is universal across participants, it would have been possible to mathematically adjust self-ratings and partner-ratings according to a certain constant. However, as long as no psychometrical assessments of intelligence are available, the magnitude of self-enhancement processes cannot be determined.

↓128

Like self-ratings, partner-ratings of intelligence are also ridden with methodological problems. Most importantly, they are affected by stereotypes and halo effects. To statistically control the effect of stereotypes, the present dissertation partialed out age and gender effects from the intelligence ratings. However, it was not possible to control for the presence of halo effects. This is especially problematic as the current study used intelligence ratings to predict self-ratings of another highly positively valenced construct: relationship quality. As stated previously, it is not unlikely that the associations found between partner-ratings of intelligence and relationship quality are a result of a tendency to attribute positive qualities to well-liked people, a preference for people with positive attributes, or both.

A final problem regarding both intelligence self- and partner-ratings in the present study is the use of the single-item scale. First of all, the use of single items precludes the calculation of internal consistency needed for an empirical assessment of the degree of attenuation in the relation between intelligence ratings and MU. Second, the single-item scale turned out to be confounded by group membership in Sample 1. Specifically, Mensa members in this sample were explicitly aware of their maximum intelligence level and thus always rated themselves with the highest possible value of the scale (i.e., 20). By comparison, most university alumni probably did not know their IQ, so they may not have been so inclined to place very high bets on their own intelligence. This confounding of group membership and intelligence rating makes the interpretation of results more difficult (e.g., as evidenced by the divergence of results based on the single-item scale compared to the intelligence self-concept scale51).

Although psychometric intelligence tests are not affected by the kinds of problems described above, the measures used in the current study might have suffered from some additional problems. For example, the figural intelligence test in Sample 4 had an alpha level of only .65, which is below the .70 that is considered acceptable. The other two intelligence tests had adequate reliabilities, but they were implemented online, which may have introduced additional sources of bias. Especially the vocabulary test may have been affected by this, since it requires participants to pick the correct word out of a list of five possibilities. It cannot be ruled out that internet users used search engines or online dictionaries to look up the correct word. Accordingly, it may be that the results of this test were not only affected by participants’ vocabulary level, but also by motivational factors.

↓129

Finally, a further problem affecting the psychometric measures is the lack of significant correlations between the figural, numeric, and vocabulary tests. Thus, the current study could not replicate the general intelligence factor that is so well-established in psychometric research. Note, however, that the figural and vocabulary measures showed a clear restriction of range as evidenced by a SD almost half of the value reported for the norm population (see Table 4). In addition, it should be noted that Sample 4 evidenced a high degree of profile specialization with regard to intelligence, with more mathematically-oriented students having higher values in numerical intelligence and language-oriented students having larger vocabularies. Nevertheless, the failure to replicate a general intelligence factor gives rise to caution in interpreting the current results.

5.5.3 Sampling Bias

Another major limitation of the conclusions of the current study is the possibility of sampling bias. As stated above, the reliance on university students may have led to restrictions in range in intelligence in Study 4. In addition, selection bias may have been exacerbated even more by differences in the invitation letter sent to potential participants in Study 2. Specifically, it is possible that most participants who took part in the study did so because they felt they received an invitation because of their high intelligence level (see Section 3.1.3.2). Finally, a source of bias may also have resulted from indirect hints regarding the study’s topic and hypotheses in the recruitment material. Specifically, descriptions of the project contained cues that the study was interested in the effect of personality similarity on social relationship quality. This may have led to an oversampling of participants with highly salient or idiosyncratic experiences in this domain.

Arguably, the most serious source of sampling bias in the current dissertation is the inclusion of Mensa members as representatives of gifted individuals. Because not every gifted individual in Germany is a member of this organization, Mensa members may not be typical of the general population of gifted individuals. As stated earlier (Section 4.4.3), it could be that the Mensa membership represents some kind of compensation for certain social problems. For example, it may be that gifted people who do not feel valued by their family members join Mensa to bolster their general sense of self-esteem. After all, Mensa is highly elitist in the sense that all members have an extremely high intelligence, a trait that is very positively valued in present-day society. In any case, the extent to which Mensa members are unrepresentative of gifted individuals in general limits the generalizability of the current results.

5.5.4 Lack of Experimental Data

↓130

A final limitation of the current dissertation is the lack of controlled, experimental data.52 Such data does not permit the drawing of causal inferences. For example, in Sample 4 it was found that vocabulary level is positively associated with MU. In the theoretical framework of Chapter 2, this association is explained by the fact that having a large vocabulary facilitates the encoding and decoding of thoughts and feelings, leading to higher levels of MU. However, it could also be that the positive association between both factors is based on a third factor. One possibility is that individuals from privileged socioeconomic backgrounds acquire better language skills during the course of socialization and are also better equipped to deal with the complexities of social life. If this is true, then the correlation between vocabulary and MU has nothing to do with actual communication ability but is a spurious result of a shared association with “social capital” (Bourdieu, 1986). In this and all other cases of cross-sectional results, statements about causal mechanisms must remain tentative at best.

5.6 Strengths

In spite of a number of limitations, the current dissertation also has a number of strengths. Specifically, these include the study of naturalistic situations, the inclusion of multiple assessment methods, the use of dyadic designs, and the modeling of the hierarchical (nested) data structure with advanced statistical tools. In the following, each of these strengths is discussed in more detail.

5.6.1  Use of Multiple Assessment Methods

A further strength of the current study is the use of a broad array of assessment methods. This focus on inclusion was present in the broad range of instruments to assess key constructs. For example, intelligence ratings in Samples 1 and 2 were available as a single-item and a self-concept scale. In addition, these ratings were supplemented by the results of psychometric tests of fluid and crystallized intelligence in Samples 3 and 4. Social adjustment was operationalized by the use of self-report scales tapping into individuals’ perceived self-worth regarding certain social relationship categories (e.g., with parents, peers) and an ego-centered relationship instrument. Finally, the current dissertation used two different scales (NEO-FFI and BFI) to assess openness to experience and the other Big Five factors.

↓131

In their historical review of seventy years of research on personality and close relationships, Cooper and Sheldon (2002) have lamented the pervasive reliance on self-report measures as a common methodological weakness.53 Because people may be unable or unwilling to accurately report certain intra-psychic phenomena (e.g., whether they understood their interaction partner; Nisbett & Wilson, 1977), self-reports may be inaccurate, especially when studying complex patterns of social interactions (Cooper & Sheldon, p. 785). For this reason, the current research supplemented self-report data with behavioral observations of the level of MU in their conversation. Although both self-reports and observations may contain valuable information and none is necessarily “more objective” than the other, it is clear that their combined use greatly increases the confidence that the current findings are not a result of measurement artifacts.

5.6.2 Use of Dyadic Design

A second requirement identified by Cooper and Sheldon (2002) for research on personality and social relationships is the use of dyadic designs. In fact, it could even be claimed that social relationships cannot be studied at all when the focus is on single individuals. From such a design, it is only possible to draw conclusions about individuals’ construction of relational processes, not about these processes themselves. From research on social support (e.g., Sarason et al., 1987), it is well-known that such perceptions sometimes have little in common with individuals’ actual social relationships.

Through the use of a dyadic design, it is also possible to discover possible dissociations of effects across dyadic partners. In the current dissertation, Samples 1.II and 4 included data from both members of dyads. As results showed, the correlation between two interaction partners’ assessments of relationship quality was not very high. Indeed, results of Study 1.II showed that the association between intelligence ratings and perceived relationship quality differs between Mensa members and their network partners. In addition, Study 4 found that only the personality of the first interviewer influences MU, not the personality of the second interviewer. This dissociation across dyadic partners could not have been uncovered without the inclusion of both dyadic informants.

5.6.3 Modeling Nested Data

↓132

A further strength of the current dissertation is the analysis of the ego-centered social network data with the use of hierarchical linear modeling (HLM). At present, the overwhelming majority of studies have used generalized assessments of social support (e.g., self-concept of relationships with peers). However, the use of such generalized measures can be criticized because people tend to dispositionally under- or overestimate the degree of support they perceive from others (Sarason et al., 1987). In order to anchor evaluations to specific relationships, the use of ego-centered networks has been recommended as an alternative (e.g., Neyer, 1997). Because of the greater specificity of this method, the influence of generalized response tendencies can be reduced.

Although the use of an ego-centered technique is to be preferred for methodological reasons, the statistical analysis of such networks has so far proved tedious. This technique allows networks participants to generate networks of varying size and composition. For example, one participant may list 3 network partners including a mother, a father, and one friend, whereas another participant may list 24 contact persons including 13 co-workers. The most common solution thus far has been to aggregate information across certain categories that (almost) all networks have in common (e.g., support by mother, number of friends). However, this procedure ignores a great deal of useable information, such as relationships with categories of persons that are less common (e.g., neighbors, club members). More fundamental, findings about processes within a social network (i.e., at the level of individual relationships) may not necessarily generalize to comparisons across social networks (i.e., at the participant-specific level; Molenaar, 2004).

By comparison, hierarchical linear modeling approaches allow one to analyze every single network partner/social relationship (van Duijn, van Busschbach, & Snijders, 1999). Technically, this is done by specifying a participant-specific intercept representing the base-line expectation of support in a social network. Moreover, it is possible to include Level 1 characteristics that may influence the relationship (e.g., the gender of the network partner) or dummy variables that allow for a fine-grained distinction between certain relational categories.54 Finally, HLM uses information on the participant-specific Level 2 to predict each relationship-specific parameter. This methodological innovation holds great promise for future research.

5.7 Conclusions and Recommendations for Future Research

↓133

The current dissertation gives rise to a number of conclusions and recommendations for future research. First, the current dissertation provides arguments for the usefulness of MU as a construct in personality and social relationships research. Second, the presence of limited main effects of vocabulary on emerging social relationships has theoretical implications for understanding dynamic transactions between persons and their environment. Third, it seems that people are able to bridge between-person differences in personality. Finally, some recommendations for future research are formulated. In the following, each of these points is elaborated in more detail.

5.7.1  Usefulness of MU Construct

First of all, the current dissertation points to MU as a potentially important construct in social relationship research. The data from Study 4 demonstrate that MU is only moderately related to more generalized indicators of interpersonal liking, such as rapport. Results from Studies 1-3 were consistent with theoretical accounts that have postulated a high degree of convergence between MU and more diffuse, emotional qualities in well-established relationships (Reis, 1990). The lack of such strong convergence in newly establishing relationships supports the notion that MU is more independent from generalized relationship quality in newly-emerging relationships. It would be interesting to see whether MU and relationship quality become increasingly correlated as relationships mature.

A promising way to investigate the mechanisms by which MU affects the establishment of more intimate relationships is to consider its effect on outcome expectancies. According to Sunnafrank, Ramirez, and Metts (2004), people getting acquainted have perceptions regarding the “outcome value” they can expect from the other person. When an individual expects highly rewarding future interactions with an interaction partner, he or she will strive to increase interaction intensity. Because felt understanding is a highly rewarding state in social relationships (Fehr, 2004), a high degree of MU should be an important predictor of relationship continuation and quality.

↓134

In spite of the potential theoretical relevance of MU, its determinants remain somewhat elusive. That is, only a few of the hypothesized main and emergent effects of personality were confirmed in the current study, and these effects were often restricted to certain assessment methods (self-ratings vs. partner-ratings) or persons (e.g., interviewer 1 vs. interviewer 2). Of course, this does not exclude the possibility of an association between MU and social and personality factors not included in the present study. For example, it could be that people feel most understood when they encounter a highly acquiescent partner who agrees with everything they say. Moreover, it is possible that dyadic partners’ congruence of goals increases their sense of MU. Finally, it is possible that MU is maximized when two people are similar in terms of the personality constructs that are most central to their identities. Future research is needed to substantiate these possibilities.

5.7.2 Limited Main Effects in Study 4

In Study 4 found main effects of intelligence were limited to the person who played the role of interviewer in the first interaction half. Because this phenomenon occurred for both vocabulary test results and intelligence ratings by the interaction partner, it is less likely that this pattern was produced by chance. Rather, the dissociation between the effects found for the first and second interviewer may be a result of different behavioral opportunities associated with the experimental context. Specifically, it is possible that the first interviewer had more opportunities to “channel” the conversation by asking questions and reacting to the utterances of the interviewee. By comparison, the possibilities of the first interviewer’s interaction partner to influence the communication were more limited. Of course, he or she could choose the topic of the conversation, but because of the structured nature of the situation, the establishment of MU during this interaction half was probably more dependent on the first interviewer.

The above-described differential opportunities for the first and second interviewers’ personality to influence the quality of their interaction calls to mind Henry Murray’s (1938) concept of “press”, which he defined as the directional force of an environment, object, or person. As the current results suggest, the “press of personality” (Thorne, 1987) may vary according to the dynamic, temporal, and social role properties of the environment. Apparently, people who get acquainted develop relatively stable perceptions of their emerging relationship rather quickly (Sunnafrank et al., 2004), which limits the time frame for personality traits to influence this process. After this initial phase, relationships may come to depend more on the (negotiated) interaction history between both dyadic partners (Clark and Marshall, 1992) that is not captured by the additive effect of their personalities.

5.7.3 Bridging Between-Person Differences

↓135

The final main conclusion of the current dissertation is that people are able to bridge between-person differences in intelligence and dispositional valuations. To speak with John Donne (1573-1631), no man is an island; people can reach out and understand each other’s thoughts and feelings even though they have different personalities. In a sense, this represents an upbeat message. After all, if between-person differences in personality were to pose a serious threat to interpersonal communication, society would eventually transform into a conglomeration of relatively homogeneous subgroups of like-minded others (e.g., as predicted by Herrnstein & Murray, 1994, for intelligence). The fact that people seem able to overcome such differences, at least in the contexts of well-established relationships and interactions between strangers, provides indirect evidence against this position.

At first sight, the conclusion that dyadic differences did not seem to affect people’s level of mutual understanding runs against common sense. After all, folk wisdom has it that “birds of a feather flock together”, and couples break up because “they are just too different”. In the context of romantic relationships, work by researchers such as Murray et al. (2002) has led to some intriguing conclusions regarding this common wisdom. That is, when people think they are very much different from their partners, they tend to report less satisfying and stable relationships. Researchers have also shown, however, that this perception is independent of actual between-person differences, which are unrelated to relationship quality. Consistent with this, the current study provides evidence that people are able to establish mutually understanding relationships even when they have very different personalities.

Simonton (1985, p. 536) acknowledged that it may be possible for “a truly intelligent individual [… to] avoid ‘talking over the heads’ of fellow group members [and thus] get around the comprehensibility limitation […]” However, he hypothesized that highly intelligent people will not usually lower the complexity of their utterances, as “it is probably more difficult in small problem-solving and social groups to accomplish such simplification without sounding insincere or condescending.” In contrast to Simonton’s assumption, the current findings suggest that people place a high importance on establishing a sense of MU with their interaction partners, even if this comes at the price of lowering the complexity of their utterances.

5.7.4 Recommendations

↓136

From the current dissertation, a number of recommendations for future research can be derived. Specially, it is argued that future research would benefit from the use of more representative samples, a variation of task demands, the inclusion of objective tests of MU, and the online assessment of interpersonal perceptions. In the following, each of these recommendations is discussed in more detail.

5.7.4.1 More Representative Samples

First of all, it is imperative that future studies are conducted with more representative samples. For example, it would be highly informative to compare an unselected gifted sample with a matched comparison group of average intelligence. Only this way, can it be ruled out that differences between both groups are the result of sample selection bias. Another possible effect of sample bias may stem from the use of university students in Study 4, who can be expected to have a restricted range of above-average intelligence scores. Because range restriction attenuates the correlation between two variables, the inclusion of a more diverse sample may have uncovered significant associations between mutual understanding and between-person differences in personality.

Note, however, that the inclusion of more representative samples comes with a number of additional methodological challenges. Because society is segmented according to intelligence, people with different intelligence levels may also have quite different habitats, which may act to limit the exchange of shared experiences (Duck, 1994). For example, if a university professor is not able to communicate satisfyingly with a construction worker, this may be due to the former person’s more sophisticated vocabulary. However, it may also be that the two persons live such different lives that they have “nothing to share”. Thus, future studies need to deal carefully with the problem of separating the effects of “pure” intelligence from more lifestyle-related aspects of personality.

5.7.4.2 Objective Tests of MU

↓137

According to Jones and Guerrero (2001, p. 571), normal conversations are characterized by a minimal level of person-centeredness. That is, there exist strong social pressures to act in a civilized manner and react with a modicum of interest and respect to other people’s utterances. Such “polite” expressions of understanding may obscure an inability to understand deeper layers of meaning and hinder the progression of the MU process. Accordingly, future studies should include more precise, performance-based indicators of MU that transcend the immediate reactions of both dyadic partners to the interaction (see also Section 5.5.1 for the limitations of relying on subjective assessments of MU).

A starting point for the development of more objective tests of MU could be Ickes’s (1993) mind-reading paradigm. This paradigm works with videotaped interactions between two people. In Ickes (1993) original method, the tape is shown to each interaction partner separately, who have to indicate when they remember having had a significant thought and describe the content of that thought. In a second step, the other person is shown the video, which is paused at the moments where the interaction partner had previously stated thinking about something. Empathic accuracy is operationalized as the degree to which participants are able to predict what their interaction partners were thinking about at that time.

An interesting approach would be to adapt the method developed by Ickes (1993) to conform to the dynamic features of the MU phenomenon. For example, participants could be made to view a tape of their dyadic interaction, which is stopped after each speaking turn. MU could then be operationalized as the degree to which the listener’s interpretation of the meaning of the utterance corresponds to the intended meaning by the speaker. Alternatively, when the focus on such short, verbatim interaction turns does not produce enough diverging interpretations, listeners could be asked to provide a short summary of the subjective, psychological meaning of their interaction partners’ narrative. In this case, the measure of MU would be the degree to which the speaker can identify with this summary.

5.7.4.3 Online Measurement of Perception Changes

↓138

A final recommendation for future research concerns the use of more dynamic designs to measure perceptions of MU and dyadic differences across time. In the current Study 4, half of the participants were told that they were either very similar or dissimilar to their interaction partners. Unexpectedly, intelligence ratings provided after the 20-minute conversation failed to show a significant effect of this manipulation: Participants who were told they were very different did not perceive more differences than participants who were told they were highly similar. Provided that participants believed the performance feedback (based on their pretest results) at the time it was given, these first impressions seem to have given way to an alternative view based on cues from the ongoing conversation. Possibly, they engaged in a process of identity negotiation (Swann, 1987) during the conversation, eliminating the effect of the experimental manipulation.

Similar to perceptions of dyadic differences, the perceived level of MU might go up or down as well, depending on experiences made within the conversation. In the current study, there was an increase in observed MU across time, but it is not known whether this increase was paralleled by an increase in subjective feelings of understanding and being understood. According to Sunnafrank et al. (2004), people quickly develop expectations regarding the predicted outcome value of certain relationships. These expectations supposedly guide the selection of topics that are discussed during the conversation. If people want to increase relational contact, they will choose a topic they think will be of interest their interaction partner. If this is the case, both dyadic partners may stick to this topic instead of broadening their conversation (i.e., as Duck, 1994, predicts). In such a case, perceptions of MU may be highly stable across the conversation. However, when a chosen topic receives a more mixed reception by the interaction partner, the feeling of being understood may momentarily drop and only become reestablished once a “secure” alternative topic is found.

A way to test this notion is to assess social perceptions of both dyadic partners’ personalities and MU at multiple moments of their interaction history. Ideally, this would involve pre-interaction assessments of interpersonal sympathy and expected MU on the basis of personality profiles, followed by similar assessments after participants are shown a picture of each other. When these individuals eventually meet, regular assessments may then capture dynamic changes in their perceptions of themselves, their interaction partners, and the emerging relationship between them. These changes could then be contrasted with pretest measures or modeled with dynamic statistical methods. Note, however, that such designs may be difficult to implement because of constraints in participants’ motivation and cognitive capacity. In addition, they also go hand in hand with a danger of disclosing the research hypothesis by asking repeated and/or obvious questions (Sunnafrank, 1992).

5.7.4.4 Variation of Task Demands

↓139

Of course, the conclusions of the current dissertation may only apply to a limited context. After all, there are many more personality variables than intelligence, openness to experience, interests, and values. In addition, different relationship contexts other than the ones studied here exist. For example, it may be that dyadic similarity is more important when people engage in conversations about idiosyncratic or unusual phenomena or when they have to rely on less than optimal communication media (e.g., telephone calls in long-distance relationships). Accordingly, future studies should try to vary communicative task demands when they want to assess the impact of between-person personality differences.

One of the most promising ways to vary task demands may be to increase the level of intellectual demands of the communicative situation. This follows directly from Simonton’s (1985) model, which states that in competitive, task-focused groups, between-person differences in the intelligence of group members limit the ability of the more intelligent members to make themselves understood. Although the current study failed to replicate this notion in more emotion-focused situations, this does not mean that the premise of the Simonton model is wrong. Future studies could explicitly test this option by creating conditions in which participants are motivated to display their highest possible level of intellectual sophistication (e.g., by awarding rewards to more complicated messages) while interacting with another person. It is predicted that this will produce the conditions where between-person differences impair the level of MU.


Footnotes and Endnotes

47  It could be objected that this level of agreement was partially inflated because the four most extreme self-ratings were used to „anchor” the behavioral observations. However, excluding these cases still produced an average agreement of .21, which is significant, t(64) = 1.72, p = 0.05.

48  When the interaction half was ignored and the number of 30-seconds intervals was counted from the beginning of the first to the end of the second interaction half, the beta coefficient specifying the association between time elapsed and MU increased to .23 (p = .01).

49  A quick search of the PsycINFO database uncovered that Sunnafrank’s series of 7 empirical articles and book chapters on this topic are only cited 10 times by other authors.

50  Note that Rost (2000) included both an intellectually gifted and a highly achieving sample. The highly achieving sample reported a lower self-concept of peer relationships compared to controls. Although the highly achieving sample also had above-average intelligence, however, they cannot be considered as gifted according to the conventional IQ > 130 criterion.

51  Note, however, that the single-item measure and the self-concept scale to rate intelligence were significantly correlated in the alumni sample, r = .53, p < .01. In samples that are not explicitly aware of their IQ, both measures would probably have produced equivalent results.

52  Because the experimental manipulation in Study 4 apparently failed, only naturalistic observations and self-reports were drawn from this source. Therefore, results based on these data are not experimental in the strict sense of the word.

53  In fact, no less than 77% of all studies exclusively rely on self-reports, whereas only 6% of all studies include behavioral observations.

54  According to J. Nezlek (personal communication, 14 April, 2005), the inclusion of Level 2 variables to predict the beta weights of these relationship-category dummies faces some limitations if certain participants fail to list network partners for a category.



© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 4.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated:
29.11.2006