Back to Journal Cover Page


CURRENT RESEARCH IN SOCIAL PSYCHOLOGY


Volume 6, Number 10
Submitted: March 23, 2001
Resubmitted: May 7, 2001
Accepted: May 7, 2001
Publication date: May 8, 2001

GROUPS AND AFFECT: SENTIMENTS, EMOTIONS, AND PERFORMANCE EXPECTATIONS

Robert K. Shelly
Ohio University

Ian M. Handley
Ohio University

Jessica Baer
Ohio University

Stacey Watson
Ohio University

ABSTRACT

Actors form and enact performance expectations based on status advantages brought to a situation or on evaluations of attempted task solutions. Interest in specifying how and under what conditions other social differences result in performance expectations has emerged in recent investigations. We designed our research to detect whether liking manipulated by attitude similarity and difference would affect performance expectations the same way as liking manipulated by reciprocity of gift exchange. The data show that the two different means of manipulating liking produce differences in strength of positive and negative feelings for a fictitious partner, but do not produce different effects on performance expectations. We identify unexpected gender differences in liking. Women report more liking for partners who share similar opinions and hold higher performance expectations for their fictitious partner than do men.

[135]
---------------
[136]

INTRODUCTION

Studies of inequality emergence of in task groups have identified several processes that determine how actors allocate chances to act, respond to influence attempts, and evaluate contributions. This process results in the observable power and prestige order of a group. In some situations group members possess equal initial status and differences emerge through the quality of ideas members provide toward completion of a task. In other situations, some members initially have status advantages over others. These status advantages result in differential patterns of expectations held by group members for the quality of task performances offered by group members. Higher status members are objects of increased performance expectations that increase the relative amount of participation and influence an advantaged individual has on the group. For example, if Person X possesses high status in a group, others in the group will have high performance expectations for this person. These expectations result in Person X receiving opportunities to participate and influence the group’s direction and achievement.

Recent research has investigated the possibility that affective patterns may also affect the emergence and enactment of inequality in task groups. We understand affective patterns as internal manifestations of sentiment and emotion variously distributed by an actor across other members of a task group. Two patterns of investigation have emerged in this work. One focuses on how affective patterns emerge from interaction between unequal group members and impact intragroup behaviors. The other focuses on how affective patterns serve as the foundation of observable inequality. Our research addressed the latter area and examined whether sentiments and emotions have equivalent effects on the formation of performance expectations. We indexed sentiment as patterns of liking and disliking within a group. We indexed emotions as anger and euphoria with a partner.

Sentiments (or liking) may organize group activity in open interaction task groups. Testing this idea, Shelly (1993) manipulated characteristics of a three-person group leader. In the control condition he appointed no leader. In an expertise condition the leader was chosen on the bases of a prior demonstration of superior relevant skill. In sentiment condition groups he told members they had high or low attitude similarities with the leader. This experiment found that, unlike the control condition, in both the expertise condition and sentiment condition, the leader initiated more activity than other group members. This study provides support for the assertion that sentiment structures, like status structures, may influence and organize group activities.

[136]
---------------
[137]

Shelly and Webster (1997) suggest that groups can organize via three structures: status generalization, power and prestige, and sentiment. A person can occupy one position within a structure and one’s position can vary depending on the structure investigated. Two structures can be congruent (an actor occupies the same relative position in all structures) or incongruent (an actor occupies different relative positions depending on the structure). In an experiment, seven conditions based on formal (appointed) position, manipulated sentiment, and status (combinations of these plus a control condition) were studied. Interaction was coded for action opportunities, performance outputs, and positive and negative evaluations. The data support an interpretation that positive sentiment enhances the effects of initial status differences. This study also supports the assertion that sentiment structures influence group interaction.

How sentiment organizes behavior is not clear based on these limited studies of open interaction. Two models address the role of sentiment as a status organizer within groups: the constituent model and the translation (or mediation) model. They differ in how they create affective differences within the context of expectations states theory and its behavioral models (Wagner and Berger 1993). The constituent model portrays status and sentiment as independently leading to the formation of expectation states, which then influence behavior. Studies by Lovaglia and Houser (1996) and Shelly (2001) show that affective structures can have independent effects on expectations and behavior. The translation model suggests that only status, and its correlates, lead to the formation of expectation states and that sentiment patterns mediate behaviors resulting from expectations. Driskell and Webster (1997) obtained evidence consistent with this model. Interestingly, support for both models has been found in similar experimental paradigms. In the following, the methodologies and results of experiments supporting both models will be discussed.

Lovaglia and Houser (1996) report several experiments investigating the effects of status and sentiment on expectations and behavior (i.e., influence). In their second study, they manipulated participants to be angry or happy with a partner. In the happy condition, the partner reciprocated in a gift exchange whereas in the angry condition, the partner did not reciprocate a gift exchange. They created six conditions in this experiment. In four, status and liking were manipulated, in the remaining two, they manipulated only status. Participants then engaged in a controlled interaction task with the partner in which they were to decide which of two pictures contained more of the color black. The participants saw their partner’s response (initial choice) and then were able to change their response to match their partner’s initial choice if they wished (final choice). The partner disagreed with the participant in 20 of the 25 initial choices. Participants then filled out questionnaires including items assessing the partner's ability and influence. The results of this experiment show that a participant’s liking for the partner influenced their perceptions of ability (e.g., intelligence) and behavior.

[137]
---------------
[138]

Support for the constituent model is provided by the results of three vignette studies reported by Shelly (2001). In the first study, Shelly found that when a person was described as liked, participants perceived this person as having more positive personal qualities and positively evaluated capacities whereas the opposite was true for a person described as disliked. Participants seem to have higher performance expectations for a person when they are described as liked as opposed to disliked. In the second study, informing participants that the person in the vignette was liked (or disliked) manipulated liking and have (or have not) worked well with this person in the past. Direct measures of performance expectations were assessed (e.g., "this person will do better in class"). Results of this study show that liked people elicit more positive performance expectations than disliked people. In a final study, Shelly asked participants to choose an advisor to work with who differed in sentiment (like, neutral, dislike) and combinations of status attributes (high vs. low ACT scores and class rank). Shelly created twelve variations of status and sentiment combination. The results of this study show that participants chose to work with the liked advisor more than the neutral advisor, who, in turn, was chosen more often than the disliked advisor. The more highly relevant the status information was to the task, the more respondents used the information to choose an advisor. In combination, these three studies show that people find liked individuals to have better qualities and capacities, have higher performance expectations for, and choose more often to work with liked individuals whatever their status.

Driskell and Webster (1997) had participants engage in a contrast sensitivity task in which the partner’s initial choice disagreed with the participant 20 out of 23 trials. Before this occurred, liking or disliking was manipulated for some conditions by informing participants that their partner either agreed (or disagreed) with them on most items in an earlier attitude questionnaire. Besides the manipulation of liking, some conditions varied the status of the partner. Two of the six conditions varied only status, two varied only sentiment, and two put participants in the situation of low status with a liked or disliked partner. The results of this experiments found that only status influenced performance expectations. Only the status conditions or the status and sentiment conditions produced differences in yielding to influence attempts by the partner. In the sentiment only condition, no effects for performance expectations or behavior was observed. The only time sentiment affected any of the main dependent variables was when it interacted with status to effect the behavioral measure. Thus, this experiment supports the translation model in that sentiments did not affect performance expectations but affect behavior when combined with status.

[138]
---------------
[139]

At first glance, deciding which model is correct in specifying how affect and status organize expectations and are translated to behavior is difficult. Lovaglia and Houser (1996) and Driskell and Webster (1997) employ similar methodologies but obtain contradictory results. In an attempt to provide some evidence for the superiority of one model over the other, Fisek and Berger (1998) analyzed the data from both studies. The results of their goodness-of-fit analyses suggest that both models fit the present data well. They discovered that, although the translation model seems to fit the data better, statistically speaking this model does not fit the data better than the constituent model. Nonetheless, they assert that the translation model currently has the most support and should be temporarily adopted until new evidence comes in. We think this conclusion is premature for at least two reasons. First, given both models fit the data well and that both models did not differ significantly in their fit of the data, the authors’ conclusion that the translation model is likely the better choice seems tenuous. More likely to us is the possibility that, given the small amount of data on the topic, any slight differences in fit are due to chance. Second, we feel that, before sociological researchers attempt to unravel the inconsistencies in the fledgling research area of affect within expectation states theory, researchers should examine methodological differences that may have contributed to these incongruent findings. Exploring methodological differences between experiments that may have led to differences in results seems much more beneficial.

We attempt to find methodological differences between the studies supporting the two models and investigate whether or not these may be responsible for the differences in results. We suggest that the reasons for these contradictions may be due to the different sentiment and emotion manipulations used in Driskell and Webster’s (1997), Lovaglia and Houser’s (1996), and Shelly’s (2001) studies. To manipulate emotion, Lovaglia and Houser (1996) used a gift exchange (reciprocity vs. no-reciprocity). Shelly (2001) told participants that they liked the partner and worked well or poorly with the partner in the past. Driskell and Webster (1997) told participants they either agreed or disagreed with their partner on an attitude questionnaire. Thus, the affect manipulated by each investigator may have been qualitatively different in constituent model studies versus the translation model study. Shelly (1993) suggested that, sentiment is an affective state in which the object is a person, and liking and disliking is a low-strength positive or negative inclination toward an actor. Using these definitions, we suggest the sentiment manipulations in the studies that supported the constituent model were more affective or emotional than the "agreement"-only manipulation employed in the translation model study. We suggest the "agreement" sentiment manipulations in the studies supporting the translation model likely produced low-strength liking whereas studies supporting the constituent model used high-strength manipulations (e.g., gift exchanges).

[139]
---------------
[140]

Not only was the manipulation of affect qualitatively different in the studies supporting each model, but perhaps the task in the Driskell and Webster experiment undid the sentiment manipulation. These researchers manipulated liking by telling participants they agreed or disagreed with their partner on many items of a prior questionnaire. Thus, participants only basis for sentiment was their extent of agreement with the partner on this questionnaire. After this manipulation, participants engaged in a task in which their partner disagreed with them in 20 out of 23 trials. It is possible that the basis for liking in this study, the experimental task attenuated agreement. In this study, the status manipulation produced a greater difference between liking and disliking on the affect measure than did the actual sentiment manipulation! This interpretation is consistent with Ridgeway and Johnson’s (1990) explanation of the effects of status processes on sentiments.

Our study investigates the possibility that more emotional manipulations of sentiment and agreement/disagreement manipulations of sentiment may differ in significant ways and, as a result, differentially affect performance expectations. In this study, liking or disliking was manipulated by presenting participants with a situation in which they agreed or disagreed with their partner on attitude items or by having participants imagine a situation in which their partner either did, or did not, reciprocate a gift. We expected that, as in Driskell and Webster’s study, sentiments based on agreement or disagreement would not affect performance expectations. Consistent with Lovaglia and Houser’s experiment, positive affect was expected to increase, and negative affect was expected to decrease participants’ performance expectation for their partner when emotionally charged reciprocity/no-reciprocity manipulations were employed. Such a finding would replicate previous research, and offer an explanation for why performance expectations are sometimes affected, and sometimes not affected, by affect manipulations. That is, perhaps the type and emotional properties inherent in the manipulations of affect may influence performance expectations differently.

METHOD

Participants
One hundred-seventy introductory sociology students volunteered to participate in this study. We told participants that they would receive no benefits nor incur any penalty for their participation or its lack. In total, one hundred five women and sixty-five men participated.

[140]
---------------
[141]

Procedures
For this study, one or two researchers entered the beginning of two introductory sociology classes. After reading brief instructions, they distributed a two-page packet that contained all materials and instructions for this study. Participants received one of four possible packets, each with a different manipulation of affect (i.e., an emotional- or agreement-based manipulation of liking or disliking). Participants read vignettes that manipulated sentiment toward a fictitious partner. We told them to imagine the partner was in a separate room and therefore unavailable for direct contact, the same age, and the same sex as the participant. The manipulations were presented as agreement/disagreement with the partner on a series of attitude items, or manipulation of emotions toward the partner by reciprocated or unreciprocated gift exchange between them. We call these either an agreement or emotion manipulation of affect, respectively.

In the positive agreement manipulation, we told participants to imagine that they had already completed the first part of the experiment. They and their partner separately completed an opinion questionnaire that asked questions about their likes, dislikes, opinions and values. We told the participant that on this questionnaire their partner agreed with them on 17 of the 20 possible answers. Because the participant shares similar attitudes, the participant should come to like the partner more (see Byrne 1971). For the negative agreement manipulation, we told the participant that their partner disagreed with him or her on 17 of the possible 20 answers. Because of the attitude dissimilarity, we expected the participant to dislike their partner.

In the positive emotion manipulation of sentiment, we told participants to imagine that they had already completed the first part of the experiment with their partner. Both the participant and their partner had earned a few dollars for their participation by doing so. We asked participants to imagine that the experimenter suggested that the participant buy a candy bar for his or her partner as a pleasing gesture. In this vignette the participant agrees and gives a candy bar to the partner through the experimenter. The fictitious partner is thankful and responds in kind by giving the participant a candy bar and a can of soda. We expect the participant to like the partner through this gift-giving exchange. In the negative emotion manipulation of sentiment, the participant buys a candy bar for their partner. However, the partner responds rudely, by saying "Let’s just get this stupid thing over with" and they reciprocated no gift. We intended this manipulation to cause the participant to dislike their imagined partner. After the participants read the emotion or agreement manipulations of sentiment (liking or disliking), they were asked all a series of three questions designed to assess liking for the partner.

[141]
---------------
[142]

The second part of the study was the same for all participants. We asked participants to imagine that they were engaged in a contrast sensitivity task identical to that used by Lovaglia and Houser (1996). This task required participants to imagine they were still in a separate room from their partner and that both the participant and their partner view a series of slides on a computer screen in front of them. Each slide contained two black-and-white figures, one on top of the other. The participant was told that he or she had been asked to choose the figure that contained more black than white area. We told the participant to imagine that he or she and their partner made their initial choice of which of the two black-and-white checkerboard slides contained the most black area. Both the participant and the partner then saw each other’s initial choice. After seeing the other’s initial choice, both made a final choice of which figure had the most black area. We then tell the participant that this task consisted of 23 trials, and that on 20 of these 23 trials, the partner disagreed with the participant on their initial choice. After imagining this task, participants were asked to answer a series of questions by circling their responses on 9-point scales. Finally we asked the participant to indicate his or her gender. This questionnaire was intended to measure the participants’ expectations for their partner’s future performance.

Hypotheses
We expected emotion-based manipulations to have more impact on performance expectations than attitude-based manipulations. We also expected that positive manipulations of affect will produce higher liking scores for the fictitious partner than negative manipulations. We examined the data to decide if there are interactions between type of manipulation and positive versus negative manipulations on liking for the partner or performance expectations. These expectations are stated as explicit hypotheses below.

Hypothesis I: If we manipulate sentiments with agreements on attitude items to produce low strength feelings of liking, then actors will not form, and act upon, performance expectations for others.

Hypothesis II: If we manipulate sentiments with emotionally charged acts, such as norm violations, then actors will form performance expectations and act upon performance expectations for others.

RESULTS

We coded items when necessary so that higher numbers reflect more positive evaluations of the imagined partner. Therefore, higher ratings show more liking or higher performance expectations.

[142]
---------------
[143]

Sentiment Manipulation Check
To assure the manipulations of sentiment produced their intended effects of liking or disliking for the imagined partner, we constructed three items assessing liking ("If you were to participate in this experiment again would you prefer your; ‘partner’ or a ‘stranger’?", "I would like to socialize with my partner in a different context; ‘not at all’ to ‘a lot’," and "How much would you say you like your partner; ‘not at all’ to ‘a lot’"). Because these items seem to index liking reliably (alpha = .714), they were averaged and entered a 2 (liking: like vs. dislike) by 2 (affect manipulation type: agreement vs. emotional) by 2 (gender: female vs. male) between-subjects analysis of variance (ANOVA). Table 1 contains the means and standard deviations for this analysis.

Table 1. Liking index mean score by liking manipulation, affect manipulation type, and gender.

Like Dislike
Emotion Agree Emotion Agree
Men Women Men Women Men Women Men Women
6.10
(1.66)
n = 13
6.81
(1.23)
n = 27
5.52
(1.64)
n = 22
6.58
(1.25)
n = 22
3.33
(1.62)
n = 12
2.93
(1.27)
n = 28
3.52
(1.33)
n = 18
4.89
(1.12)
n = 28

Agree versus emotion manipulation: 5.18 (1.68, n = 90), 4.82 (2.25, n = 80).
Gender difference for men versus women: 4.68 (1.92, n = 65), 5.22 (1.98, n = 105).

The ANOVA revealed a significant main effect of liking. Participants rated their partner more likable when they received a liking condition (M = 6.30) than if they received a disliking condition (M = 3.75; F [1, 170] = 139.72, p < .0001). The expected, and significant Liking by Affect Manipulation Type interaction qualify this result. This interaction shows that the emotional manipulation of affect produced more extreme liking (M = 6.58) or disliking (M = 3.05) for the imagined partner than did the agreement manipulation of sentiment (Ms = 6.05 and 4.36 for liking and disliking, respectively; F [1, 170] = 11.59, p < .001).

[143]
---------------
[144]

We employed individual comparisons to assure that liking manipulations produced more liking than disliking manipulations for both the emotion and agreement sentiment manipulation types. This was done because the main effect of liking may have been an artifact of the significant Liking by Affect Manipulation Type interaction. As expected, the manipulation of liking verses disliking via the agreement manipulation produced significantly different amounts of liking with t (88) = 5.51, p < .0001. Additionally, the manipulation of liking verses disliking via the emotional manipulation produced significantly different amounts of liking with t (78) = 11.35, p < .0001. Thus, the significant main effect was not simply due to a large effect for the emotional manipulation and a small effect for the agreement manipulation; both manipulations produce significant differences in liking. The interaction shows that the emotional manipulation of affect produces a stronger effect of liking than does the attitude similarity manipulation.

A significant main effect of gender showed that, overall, women liked their imagined partners (M = 5.22) more than men liked their imagined partners (M = 4.68; F [1, 170] = 9.84, p = .002). A significant interaction between gender and Affect Manipulation Type, F (1, 170) = 5.92, p = .016 was also found. Women and men seem to like their imagined partner comparably when affect is manipulated emotionally (Ms = 4.84 and 4.77 for women and men, respectively), whereas women (M = 5.63) like their imagined partner more than men (M = 4.62) when affect is manipulated by attitude agreement. We did not expect these gender effects. We do not pursue this point here, but it may have important implications for future tests of how we relate sentiments and emotions to performance expectations and behavior. No other main effects or interactions attained significance.

Performance Expectations
Seven items taken from Lovaglia and Houser (1996) and Driskell and Webster (1997) were used to measure participants’ performance expectation for their imagined partner. We created an index using all seven items (which assessed the intelligence, competence, persuasiveness, influence, goodness, ability, and capability of the participants’ imagined partner). This index had an acceptable reliability coefficient (alpha = .706) and was therefore used as our measure of performance expectation. We entered the index in a 2 (liking: like vs. dislike) by 2 (affect manipulation type: agreement vs. emotional) by 2 (gender: female vs. male) between-subjects ANOVA. As in the liking analysis, we expected a significant main effect of liking and a significant Liking by Affect Manipulation Type interaction. That is, we anticipated that, overall, participants would have higher performance expectations for a liked partner than a disliked partner. We also expected the effect of liking on the index to be more pronounced and significant when we manipulated affect emotionally as opposed to through attitude agreement. The means are presented in Table 2.

[144]
---------------
[145]

Table 2. Expectation index mean score by liking manipulation, affect manipulation type, and gender.

Like Dislike
Emotion Agree Emotion Agree
Men Women Men Women Men Women Men Women
4.70
(1.51)
n = 13
5.19
(.87)
n = 27
4.99
(.87)
n = 22
5.36
(1.03)
n = 22
4.11
(1.21)
n = 12
4.56
(.93)
n = 28
4.13
(1.22)
n = 18
4.65
(.74)
n = 28

Positive versus negative manipulation means: 5.11 (1.04, n = 84), 4.44 (1.00, n = 86).
Male versus female: 4.53 (1.22, n = 65), 4.91 (.94, n = 105).

The main effect of liking on the expectation index was significant, F (1, 170) = 18.39, p < .0001. Participants had higher performance expectations for liked imagined partners (M = 5.11) than disliked ones (M = 4.44). Our main hypothesis that linked Affect Manipulation type to performance expectations, was not supported as the Liking by Affect Manipulation Type interaction failed to reach significance (F < 1). Thus, the present study suggests that, consistent with the constituent model, liking leads to higher performance expectations than disliking whatever the nature of the affect manipulation.

Besides the main effect of liking, this ANOVA also revealed a main effect of gender. Apparently women had higher performance expectations for their imagined partners (M = 4.91) than did men (M = 4.53; F [1, 170] = 7.86, p = .006). No other main effects or interactions attained significance in this analysis. We call attention to this result because of its implications for future tests of links between sentiments, emotions, and performance expectations.

[145]
---------------
[146]

Mediation Analysis

Figure 1.

Note. Bold lines represent significant pathways. The path coefficients are standardized Beta weights from the regression analyses.

As the above analysis shows, the liking manipulation affected performance expectations such that participants had higher performance expectations for those they liked than those they disliked. Following the procedures of Kenny (1979), a path analysis was done to decide whether the liking manipulation directly affected performance expectations or the effect of the liking manipulation on performance expectation was mediated through participants’ liking for their imagined partner. Path coefficients were estimated using a regression analysis and are presented in Figure 1. Paths from the liking manipulation to liking and from liking to performance expectation were both significant (p’s < .001) whereas the path from liking manipulation to performance expectation was not significant (p = .42).

[146]
---------------
[147]

DISCUSSION

Two frameworks seek to explain and predict the effect of affect processes on performance expectations and behavior. The translation model suggests that only status influences expectation states and that sentiments and emotions mediate the behavior resulting from those expectations. The constituent model suggests that status plus sentiments and emotions result in performance expectations and organize behavior. Support for these two models has resulted from experiments that have manipulated sentiments in very different ways. Studies supporting the translation model manipulate sentiments by telling participants they agree or disagree with their partner on an attitude questionnaire. Studies supporting the constituent model manipulate sentiments in more emotionally laden ways by means of gift reciprocity or imagined productive/unproductive interactions and liking/disliking.

The significance of this manipulation difference becomes evident when one considers the fact that Driskell and Webster (1997) and Lovaglia and Houser (1996) have conducted similar experiments yet obtained different results. We suggest that the reason for divergent findings may be that these studies manipulated sentiment in different ways. At least two differences between these manipulations are noteworthy. First, it seems likely that, although the manipulation is successful, the agreement/disagreement manipulation of sentiment is not as powerful as the reciprocity/no-reciprocity emotional manipulation. In fact, our experiment showed that this was the case. Although all liking manipulations produced more liking than the disliking manipulations, the significant Liking by Sentiment Manipulation Type interaction for our liking index shows that the emotional manipulation of affect was a stronger manipulation. Second, it seems possible that the reciprocity/no-reciprocity manipulation not only produces more extreme liking or disliking, but it also influences other emotional reactions to the target. That is, not only may a participant like a partner more if they reciprocate a gift, but they may feel the kindling of friendship, which could affect performance expectations far more than mere liking alone.

We hypothesized that the agreement/disagreement manipulation of sentiment would not influence performance expectations whereas the reciprocity/no reciprocity manipulation of emotions would. We expected those who liked their imagined partner would form higher performance expectations than those who disliked their imagined partner because of an emotional manipulation. However, if we manipulated sentiment through attitudinal agreement or disagreement, we expected no difference in performance expectations. This predicted interaction was not significant in our study. Interestingly, whatever the type of sentiment manipulation, those who liked their imagined partner had higher performance expectations than those who did not. Thus, affective reactions influenced performance expectations in all of the conditions we studied. The translation model would not predict this. The translation model suggests that sentiments only influence behavior, not performance expectations. The constituent model, which suggests sentiments and status can influence performance expectations independently, could account for these results.

[147]
---------------
[148]

One assertion of the constituent model is that liking directly influences performance expectations. Examining of the results of the ANOVAs, one could argue that the manipulation of liking had direct and independent effects on both liking and performance expectations. We conducted a mediational analysis to decide if liking affected performance expectation or if the differences observed in performance expectations resulted directly from the liking manipulations. The results of this analysis revealed that the liking manipulations affected liking which, in turn, affected performance expectation. No direct effect of liking manipulation on performance expectation was found when liking for the imagined partner was in the analysis. In other words, these results show that the effects of the liking manipulation on performance expectation were indirect and mediated through participants’ liking for their imagined partner. The only variable that directly affected performance expectation was liking. Because the manipulation of liking and the ratings of liking were made before we ever mentioned the contrast sensitivity task and before participants reported their performance expectations for the imagined partner, it is unlikely that performance expectations influenced liking. Due to the order in which the stimuli and ratings were presented, liking seems to have influenced performance expectations.

More elaborate experimental procedures could consolidate this new area of research. For the time being, some confusion remains. The conceptualization of our study was very similar to both the studies that supported both models. However, our results are more consistent with the constituent model. Future experimenters may employ ideas from our investigation and create robust tests in the influence experiment setting. Such studies could help clarify this area of investigation.

Future research could also explore other differences between the studies supporting the translation and constituent models. For example, Driskell and Webster used a military population in their second study whereas research supporting the constituent model used student populations. It seems plausible that for a military population, status is a highly salient feature of any interaction. Therefore, because of heightened sensitivity to status compared to students, the impact of status manipulations in the Driskell and Webster study far exceeded the status manipulations in the other studies. As a result, sentiment manipulations may not have had nearly as large an effect as they may for other populations.

Our study found gender differences on both of our dependent variables. First, we found that women like their imagined partners more than men when sentiment is manipulated using attitude agreement or disagreement. However, women and men do not differ when they experienced the emotional manipulation. Apparently, agreement may be a more important factor for women than for men in interaction. Additionally, we found that women had higher ratings on the expectation index than men. For whatever reason, women seem more willing to have positive expectations about their partner’s performance. Perhaps women give their partners the benefit of the doubt. Perhaps men are more competitive and feel their partner will not do as well as they themselves will. This discussion is entirely speculative. The results suggest that gender could become a theoretically interesting variable in the study of sentiments in the expectation states theoretical research program. Another interpretation is that women paid more attention to our manipulations than did men. This suggestion was made by Murray Webster in a personal communication. Identifying which of these processes led to our findings is worthy of further study.

[148]
---------------
[149]

The question of which model, the translation model or the constituent model, is correct is still undecided. Unfortunately, we were unable to show that the reason for the findings that support one or the other model simply resulted from the types of affect manipulation employed. We maintain, however, that we should pursue this idea (or others based on differing methodologies used in this area of research) farther with more rigorous experimental procedures. Perhaps the vignette methodology we employed was not powerful enough to test our hypotheses adequately. For now, we offer a perspective supported by Fisek and Berger’s analysis: perhaps both models are correct. That is, perhaps sentiment leads to formation of performance expectations and mediate the effects of performance expectations on behavior. Depending on the situation, sentiment may influence only performance expectations, only behavior, or both. Several psychological models find that a variable may or may not affect one or another construct depending on the situation. A research program examining differences in initial conditions might provide the answer to whether the translation or constituent model is appropriate for explaining the role of sentiment in expectation states processes.

REFERENCES

Byrne, Donn (1971). The Attraction Paradigm. New York. Academic Press.

Driskell, James E. and Murray Webster, Jr. (1997). "Status and Sentiment in Task Groups." In Jacek Szmatka, John Skvoretz, and Joseph Berger (editors) Status, Network, and Structure: Theory Construction and Theory Development. Stanford University Press.

Fisek, M. Hamit and Joseph Berger (1998). "Sentiment and Task Performance Expectations." Pp. 23-40 Advances in Group Processes, Volume 15 edited by John Skvoretz and Jacek Szmatka. Stamford, CT. JAI Press, Inc.

Kenny, D. A. (1979). Correlation and Causality. New York: Wiley - Interscience.

Lovaglia, Michael J. and Jeffrey Houser (1996). "Emotional reactions and Status in Groups." American Sociological Review. 61: 867-883.

Ridgeway, Cecilia and Catherine Johnson (1990). "What Is the Relationship Between Socioemotional Behavior and Status in Task Groups?" American Journal of Sociology. 95:1189-1212.

[149]
---------------
[150]

Shelly, Robert K. (2001). "How Performance Expectations Arise from Sentiments." Social Psychology Quarterly.

_____________ (1993). "How Sentiments Organize Interaction." Pp. 113-132 in Advances in Group Processes, Volume 10 edited by Edward J.Lawler, Barry Markovsky, Karen Heimer, and Jodi O’Brien. Greenwich, CT JAI Press, Inc.

Shelly, Robert K. and Murray Webster, Jr. (1997). "Compatible Social Processes: Some New Theoretical Principles and a Test." Sociological Perspectives. 40: 81-108.

Wagner, David G. and Joseph Berger (1993). "Status Characteristics Theory: The Growth of a Program." Pp. 23-63 in Theoretical Research Programs: Studies in the Growth of Theory, edited Joseph Berger and Morris Zelditch Jr. Stanford, CA: Stanford University Press.

AUTHOR BIOGRAPHIES

Robert K. Shelly is Professor of Sociology at Ohio University. His research has focused on the emergence of inequality in task groups. This work examines situations in which actors are initially equal to one another and situations where some actors have advantages over others. His email address is shelly@ohio.edu.

Ian M. Handley is a doctoral candidate in the Department of Psychology at Ohio University. His research interests include the effects of mood on information processing and persuasion. His email address is ih198289@oak.cats.ohiou.edu.

Jessica Baer is a senior at Ohio University. Her email address is jb230796@ohio.edu.

Stacey Watson is a senior at Ohio University. Her email address is sw431297@ohio.edu.

[150]
---------------
[151]

Back to Journal Cover Page