Source
- Barsade (2002), LIT-706, read in full in NOTE-551, from the August 2001 revise-and-resubmit manuscript Wharton deposited in Penn’s repository. The published article may differ in numbers, wording or analyses. Every figure below is the manuscript’s, and page numbers are its pages.
What was actually shown
The spread is causal (Study 1). Barsade, LIT-706, put 94 undergraduates in 29 groups for a 30-minute leaderless, mixed-motive negotiation over a bonus pool. A trained confederate, one male drama student, was randomly assigned to show pleasant or unpleasant mood, with high or low energy. Members’ self-reported pleasant mood rose by .41 with the pleasant confederate and fell by .26 with the unpleasant one (χ²(1) = 9.97, p < .005). Video-coders who saw only the members, and were kept unaware of the conditions, rated members’ mood at the end 2.75 against 2.33 (χ²(1) = 10.30, p < .005) (pp. 26–27). At group level, aggregated self- and coder ratings, coders’ ratings of overall group pleasantness and members’ ratings of the group’s pleasantness all followed the confederate’s valence (Table 3b). This could have come out the other way: two independent measures of a randomised manipulation.
The link to group process is correlational. Hypotheses 4 to 6 regress outcomes on measured contagion, not on the manipulated condition. Individually, a member’s mood change predicted self-rated cooperativeness (b = .89, p < .001) and peer-rated cooperativeness (b = .47, p < .05). At group level (Table 5, N = 26), coder-rated contagion correlated with coder-rated cooperativeness (r = .44, p < .05), with a more even split of the bonus money (r = −.37 with its standard deviation, p < .05), and against coder-rated conflict (r = −.42; −.48 for self-reported contagion) (pp. 28–29). Mood change also predicted self-rated performance (b = .44, p < .001), and only marginally peer-rated performance. The abstract’s “this contagion improved cooperation, decreased conflict” states as an effect what was measured as an association.
Convergence without a confederate (Study 2). In 113 MBA students in 26 randomly formed groups, the intraclass correlation of members’ pleasant mood was .00 before the exercise and .24 after (p. 35). Contagion predicted self-rated cooperation (b = .48, p < .01) and self-rated but not peer-rated performance. Group contagion correlated with members’ ratings of group cooperation (r = .49, p < .01) and conflict (r = −.49, p < .01), but not with the spread of the funds. Barsade offers this as evidence that contagion “is a group phenomenon and not due to the dyadic nature of the relationship between the confederate and the group member” (p. 36). It replicates the convergence and the correlations; it has no manipulation, so it adds nothing causal.
What members noticed (General Discussion, p. 37). In both studies members “did significantly relate how pleasantly they behaved in the experiment to the way other group members acted”, and Barsade herself calls this a “recognition that other group members could influence the degree of pleasantness they felt”. What they did not do was link their mood to their effectiveness: they “indicated absolutely no awareness of the relationship between the mood they were in and their attributions for their degree of effectiveness”, which they put down to the money won and their own abilities. So the unawareness is of the mood’s influence on their judgments, not of the contagion itself, and it rests on attribution items, not a designed test.
What failed. Unpleasant mood did not spread more than pleasant mood (χ² < 1). Energy cut both ways, marginally: high-energy unpleasantness spread more than low-energy (−.44 against −.07, p < .10), but low-energy pleasantness spread more than high-energy (+.66 against +.16, p < .10), the opposite of the prediction (pp. 27–28).
What this does not say
- It does not say pleasant contagion causes cooperation. It is consistent with that and with a third factor driving both a member’s mood and conduct. A test that compares process across the manipulated conditions is not reported.
- It does not say members are unaware that others affect their mood. By Barsade’s own account they partly recognised it. NOTE-551’s C4 and LIT-706’s status note, which say members did not recognise that their mood was caught, state this more strongly than the manuscript does.
- It does not identify the mechanism. Mimicry with afferent feedback (Hatfield, Cacioppo and Rapson) and conscious social comparison are both invoked, and nothing separates them.
- It does not generalise beyond the setting. Ad hoc, competitive student groups; one confederate in all four conditions; Time 2 mood recalled after the exercise for its second half. Ongoing teams may habituate to one another’s moods or grow more sensitive to them; the Study 2 correlation between friendship and contagion (r = .44, p < .05) hints at the second.
- It is about mood, not emotion with an object. That matters for how it stands to the appraisal account, THEORY-055, on which an appraisal of an event’s pertinence to one’s concerns elicits an emotion. Contagion by mimicry and feedback produces mood without an appraisal of any event, but THEORY-055 is a claim about emotions, and Barsade’s object is mood, so the two do not conflict. Her alternative description, that contagion carries “a type of group appraisal of events”, would bring them closer. The feedback route is nearer the bodily-state account of felt affect, THEORY-045; that comparison is the record’s, not Barsade’s.
Connections
- The convergence account, THEORY-126, extends this one. That account defines collective emotion as convergence of affect on one object and treats contagion as one route to it among several, with shared appraisal and group membership. Barsade’s Study 2 ICC is its definition operationalised; her Study 1 is the causal test of its contagion route. Barsade’s groups test the I-mode end only, because they were strangers competing for a bonus.
- It bears on no group subject. Study 2’s “group phenomenon” is a claim about level of analysis, not a claim that the group bears a mood. It is the empirical floor any account of a group’s own emotional state must stand on, and it does not reach that question.
Narrowed and promoted (2026-10-20)
The promotion audit (batch 075) narrowed this entry to the setting Barsade measured and promoted it under ADR-047 v3: one article read in full, NOTE-551, with the title claiming no more than it reports measuring. The read was of the August 2001 manuscript, so the title says the numbers are the manuscript’s; the published ASQ article has not been read.
What the narrowing dropped went to THEORY-826 under ADR-047 v4: that in work groups generally a member’s mood spreads, and that pleasant contagion improves cooperation and reduces conflict. It carries this entry’s former promote_when, which is the test of that wider claim.