You are on page 1of 12

INTERPERSONAL RELATIONS AND GROUP PROCESSES

Knowledge of Partners Ability as a Moderator of Group Motivation


Gains: An Exploration of the Kohler Discrepancy Effect

Lawrence A. Messe Guido Hertel


Michigan State University University of Kiel

Norbert L. Kerr, Robert B. Lount, Jr., and Ernest S. Park


This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Michigan State University


This document is copyrighted by the American Psychological Association or one of its allied publishers.

O. Kohler (1926, 1927) found that less able performers tried harder as team members under conjunctive
task demands (Kohler motivation gain effect) and that the greatest gain occurred with moderately
discrepant coworker abilities (Kohler discrepancy effect). Recent investigations have reproduced
Kohlers overall motivation gain but not the discrepancy effect. The present research examined whether
workers foreknowledge of task abilitiespresent in Kohlers research, absent in contemporary studies
moderates the discrepancy effect. Participants worked alone or in 2-person teams under conjunctive task
demands. Experiment 1 manipulated foreknowledge of ability. Experiment 2 manipulated discrepancy:
a (confederate) teammate performed slightly, moderately, or substantially better. Both experiments found
(a) overall motivation gains and (b) discrepancy moderation under foreknowledge conditions. Implica-
tions for understanding group motivation gains are discussed.

About 75 years ago, Otto Kohler provided the very first reports people tend to try less hard when working together than when
of group motivation gains in controlled experimental studies working alone (cf. Hertel, Kerr, & Messe, 2000).
(Kohler, 1926, 1927). Kohlers findings indicated that members of Two features of Kohlers (1926, 1927) motivation gain results
a Berlin rowing club tended to work harder at a physical persis- are particularly noteworthy. First, although this was not stressed by
tence task when performing as part of a (two- or three-person) Kohler explicitly, in his group trials it was virtually impossible for
team than when working individuallya phenomenon that the stronger coworker to continue working at the task after the
has been termed the Kohler motivation gain effect (Hertel, Kerr, weaker coworker had stopped. Thus, Kohlers teams were per-
Scheffler, Geister, & Messe, 2000). This effect stands in stark forming what was later termed a conjunctive task (Steiner, 1972),
contrast to more frequently demonstrated group motivation losses in which a group can do no better than its least capable member.
(e.g., social loafing; Latane, Williams, & Harkins, 1979), whereby As such, the effort increases that Kohler observed on the group
trials were due to motivation gains in the less able coworkers.
Subsequent research suggests that the conjunctive nature of
Kohlers (1926, 1927) task is a crucial feature of this work context,
Lawrence A. Messe, Norbert L. Kerr, Robert B. Lount, Jr., and Ernest
and, as such, it differentiates his effect from social facilitation,
S. Park, Department of Psychology, Michigan State University; Guido
Hertel, Department of Psychology, University of Kiel, Kiel, Germany.
another type of potential motivation gain (Triplett, 1898; Zajonc,
This research was supported in part by National Science Foundation 1965). In social facilitation, the mere presence of others (as coac-
Grant BCS-9974664 to Lawrence A. Messe and Norbert L. Kerr and tors or an audience) can motivate individuals to try harder. Re-
Deutsche Forschungsgemeinschaft Postdoctoral Grant He 2745/1-1/2 to search indicates, however, that the motivation gains that the
Guido Hertel. Parts of the reported research were presented at the meeting weaker coworker on a conjunctive task displays are not due to the
of the Society for Personality and Social Psychology, February 1999, mere presence of others and, thus, are not a product of social
Nashville, Tennessee. We thank Eric Bachman, Leah Brzezinski, Dayle facilitation. For example, Hertel, Kerr, and Messe (2000, Experi-
Jackson, Thomas McDonald, and Dawn Peevers for their help with this
ment 2) found that the less able worker tried significantly harder
research.
under conjunctive task demands than under additive task demands,
Correspondence concerning this article should be addressed to Law-
rence A. Messe, Department of Psychology, Michigan State University, even though exactly the same number of people were present in
East Lansing, Michigan 48824-1117 or to Guido Hertel, University of Kiel, both conditions.
Institut fur Psychologie, Olshausenstrasse 40, D-24 098 Kiel, Germany. It should be noted that task conjunctivityalong with the co-
E-mail: messe@msu.edu or hertel@psychologie.uni-kiel.de workers differential abilitiesalso distinguishes the Kohler effect
Journal of Personality and Social Psychology, 2002, Vol. 82, No. 6, 935946
Copyright 2002 by the American Psychological Association, Inc. 0022-3514/02/$5.00 DOI: 10.1037//0022-3514.82.6.935

935
936 MESSE, HERTEL, KERR, LOUNT, AND PARK

from social compensation (e.g., Williams & Karau, 1991), a third that motivation gains were moderated by how discrepant the
type of potential motivation gains. In contrast to the Kohler mo- coworkers were in ability, as indicated by their respective perfor-
tivation gain effect, in which the weaker coworker shows increased mances on individual trials. As such, the Kohler discrepancy effect
effort under conjunctive task demands, in social compensation, has yet to be reliably replicatedalthough other research has
under additive task demands a more able coworker tries harder in provided some tantalizing hints of its existence (cf. Stroebe et al.,
anticipation of having to make up for the expected poorer perfor- 1996, Experiment 1).
mance of a weaker colleague. The failure of past work to reliably replicate the Kohler discrep-
The second particularly noteworthy feature of Kohlers (1926, ancy effect was disappointing, particularly in light of the pervasive
1927) research is that the magnitude of motivation gains was and robust Kolher motivation gain effect that this research has
moderated by how similar versus discrepant coworkers were in yielded. However, we were optimistic that under appropriate con-
ability, as indicated by how well they performed the task when ditions, the Kohler discrepancy effect could also be demonstrated.
working as individuals. Kohlers qualitative observations (con- This optimism in large part was based on the fact that two theo-
firmed by a subsequent polynomial regression analysis of his data; retical analyses of Kohlers (1926, 1927) results (Hertel, Kerr, &
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Hertel, Kerr, & Messe, 2000) indicated that the motivation gains Messe, 1999, 2000; Stroebe et al., 1996) each have proposed a
This document is copyrighted by the American Psychological Association or one of its allied publishers.

were maximal when there was a moderate discrepancy (i.e., a ratio single mechanism to explain both Kohler effectsspecifically, (a)
of about 0.7) in the capabilities of the two dyad members. In why the weaker coworker performing under conjunctive task de-
contrast, weaker motivation gains apparently occurred when group mands would show overall motivation gains, and (b) why these
members were either nearly equal or very discrepant in their increases in effort would be greatest when there is a moderate
capabilitiesa phenomenon that has been termed the Kohler dis- discrepancy in coworkers ability. It is important to note, however,
crepancy effect (Hertel, Kerr, Scheffler, et al., 2000). that each conceptualization has proposed a different unifying ex-
Kohlers (1926, 1927) research had long gone unnoticed, until it planatory mechanism.
was rediscovered by Witte (1989). Since then, there have been a Hertel, Kerr, and Messe (2000) have proposed an explanatory
number of attempts to replicate and explain his effects. Although framework for understanding both Kohler effects that is consistent
some of these attempts were not very encouraging with regard to with the well-established Instrumentality Value models of work
the replicability and generalizability of Kohlers findings (Stroebe, performance effects (e.g., Karau & Williams, 2001; Stroebe &
Diehl, & Abakoumkin, 1996), other, more recent work has been Frey, 1982; Vroom, 1964). In this perspective, for instance, a
more promising (e.g., Hertel, Kerr, & Messe, 2000; Hertel, Kerr, weaker workers subjective sense of how important her or his
Scheffler, et al., 2000; Lount, Messe, & Kerr, 2000), particularly performance is to the groups successa feeling that is likely to
with regard to the Kohler motivation gain effect. vary as a function of the work context could be a critical factor
In these latter studies, a paradigm was used that incorporated in explaining both Kohler effects. It is plausible that, when work-
most of the basic features of Kohlers (1926, 1927) original pro- ing as an individual, a less capable person could easily sense that
cedure but was more efficient and afforded less risk of pain or her or his performance is not very important, because other, more
injury to participants than did Kohlers task. In this new procedure, able members could (and probably would) continue to contribute
participants held their arms extended above a trip-alarm device for to the groups success after she or he had stopped. In contrast,
as long as they felt they could without experiencing undue distress when working as part of a group on a conjunctive task, a less
or risking injury. The researchers made this task more difficult capable person would sense that, as the weakest link, her or his
either by having the workers hold a metal bar or by attaching a performance is very important, as it is clearly the sole determinant
weighted band to their wrist. In the individual condition, the task of how well the group would do. And the realization of these
ended whenever the participant lowered his or her arm far enough contingencies should motivate the less able person to try harder
to trigger the alarm. In the group condition, the task was over when when working as a group member under conjunctive task demands
either of two coworkers did so; thereby, conjunctive task demands than when working as an individual (i.e., the Kohler motivation
were created that were equivalent to those in Kohlers paradigm. gain effect).
As noted elsewhere (Hertel, Kerr, & Messe, 2000), this new task However, it also could be that this overall group motivation gain
provides several advantages over Kohlers (1926, 1927) proce- is moderated by differences in the weaker coworkers perception
dure. Perhaps of most importance, successful performance of the of how instrumental any given increase in effort is likely to be,
revised task requires minimal (if not zero) intra- or interindividual relative to the maximum outcome that potentially could be
coordination even less than for Kohlers taskand, thus, one achieved. When the discrepancy in ability is very small, she or he
can reasonably assume that effort and output are monotonically
could readily perceive that only a little extra effort is necessary to
related.
promote the maximum possible group success, given the slightly
Using this paradigm in five studies (Hertel, Kerr, & Messe,
stronger coworkers presumed capability. When the discrepancy is
2000, Experiments 1 and 2; Hertel, Kerr, Scheffler, et al., 2000,
very large, the much less able coworker could plausibly perceive
Experiments 1 and 2; Lount et al., 2000), we have examined
that any increase in effort that she or he could realistically muster
peoples performance under both individual and conjunctive task
would result in a group score that is only a fraction of what could
demands (i.e., in which the less able participants score determined
have been achieved had the much stronger partner been paired
how well the team did). Across all these studies, we have consis-
tently found significant motivation gains for the weaker coworker,
with average performance increases in group compared with indi- 1
Note as well that Hertel, Deter, and Konradt (2001) generated similar
vidual trials ranging between 10% and 50%.1 However, unlike motivation gains when team members worked conjunctively on a cognitive
Kohlers (1926, 1927) results, none of these experiments found task through a computer network.
KNOWLEDGE OF ABILITY AND MOTIVATION GAINS 937

with someone more closely matched in ability. In either circum- from our procedures an aspect of Kohlers research context that
stance (i.e., if the discrepancy in coworkers abilities is either may be necessary for generating discrepancy moderation: partici-
slight or very large), the motivation to try harder should be pants well-established knowledge of one anothers task ability.
reduced. In contrast, when the discrepancy in ability is moderate, Recall that Kohlers (1926, 1927) participants were members of
the somewhat weaker team member could very likely perceive that a rowing club. As such, they trained together several times a week
a concerted but achievable increase in effort would produce an (cf. Kohler, 1926), competed together against other clubs, and so
outcome that is reasonably close to the maximum possible, given forth. Over the course of sharing these experiences, it is highly
the partners ability level. Therefore, moderate ability discrepancy likely that each of these individuals came to have a firm sense of
would promote the greatest motivation gain (the Kohler discrep- his own and his fellow club members capacity to persist at
ancy effect).2 physical activity. In contrast, the participants in our earlier re-
Stroebe, Diehl, and Abakoumkin (1996) have proposed a dif- search were strangers to each other. Moreover, they never were
ferent mechanism to explain both Kohler effects: the extent to given feedback about either their own or their fellow participants
which the weaker coworker uses the more able partners perfor- past task performance, and they worked as members of a team with
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

mance as a goal-setting or goal-comparison reference (cf. Harkins a specific other person for only one or two trials. Thus, they could
This document is copyrighted by the American Psychological Association or one of its allied publishers.

& Petty, 1982; Locke & Latham, 1990). In this framework, work- have acquired, at most, only a vague sense of relative ability.
ers are assumed to be somewhat uncertain about what constitutes We suggest that definite knowledge of ability differences may
appropriate effort and, thus, are inclined to use how well their be necessary for discrepancy moderation to occur. It is difficult to
partner performs as a basis for gauging or refining what their own see how ability differences could moderate motivation gains with-
efforts should be. In Kohlers (1926, 1927) context as well as most out a firm understanding in less able participants that their own
other types of group performance settings, such a mechanism capacity to perform the task is about equal to, moderately lower
should lead to increased effort in the weaker coworker (i.e., the than, or much lower than their teammates. At most, the brief
Kohler motivation gain effect) as a consequence of the higher group experiences that participants in our past studies (e.g., Hertel,
performance goal that upward social comparison to the better Kerr, & Messe, 2000; Lount et al., 2000) shared with a teammate
coworker generates.3
could only provide the less able worker with a general sense that
However, the degree of discrepancy in the abilities of the
his or her coworker was better, not a knowledge of how much
stronger and weaker coworker should affect goal comparison and,
better. If so, then the failure of this past work to find that ability
thus, moderate the magnitude of the latters motivation gain (the
discrepancy moderated overall motivation gains is reasonable,
Kohler discrepancy effect). A small difference in ability is likely to
given that none of these investigations provided participants with
produce only a small increase in the height of the goal and, thus,
such specific ability information. In contrast, as noted above, it is
generate only a small motivation gain. Both the change in goal and
likely that the rowing club members who were the participants in
the resulting increase in motivation gain should be markedly
Kohlers (1926, 1927) research were well aware of each others
greater when the difference in ability is moderate. In contrast,
physical ability, and, on the basis of his results, it appears that their
consistent with the idea that unrealistic goals can undermine mo-
tivation (Hinsz, 1995), when the discrepancy in abilities is large, task performance could have been affected by this knowledge.
the weaker coworker is likely to reject the much better partners Within this framework, the present work investigates the im-
performance as a viable standard of comparison and, thus, increase portance of knowledge of ones own and ones coworkers ability
effort to a much smaller extent. for moderating overall motivation gains. Experiment 1 manipu-
Although both explanations are reasonably plausible, it is clear lated knowledge of participants own and their coworkers past
that any utility they may have as bases for understanding the task performance but allowed relative ability to vary naturally;
Kohler discrepancy effect is moot if the effect itself cannot be Experiment 2 systematically manipulated this discrepancy. For
reliably produced. To do so, then, is the goal of the present both studies, we predicted that (a) overall, participants would
research. perform better when paired conjunctively with a stronger partner
We reasoned that if the Kohler discrepancy effect is, in fact, a than when working alone, which would thereby replicate past
real phenomenon, then the failure to replicate it in our past at- demonstrations of the Kohler motivation gain effect (Hypothesis
tempts to study motivation gains (Hertel, Kerr, & Messe, 2000; 1), and (b) when participants had specific knowledge of their
Hertel, Kerr, Scheffler, et al., 2000; Lount et al., 2000) was likely relative abilities, such motivation gains would be moderated by
due to the specific procedures that we used. It is likely that these relative ability differences, with the greatest gain occurring when
procedures lacked one or more conditions that are necessary for capability discrepancy was moderate, versus small or large (Hy-
motivation gains to be moderated by the magnitude of the discrep- pothesis 2).
ancy in the coworkers abilities. Thus, in the two studies presented
in this article, we sought to identify conditions, absent from our 2
This example focuses on one process from an Instrumentality Value
earlier work, whose incorporation into the experimental context
perspective that could underlie the Kohler effects. Elsewhere (e.g., Hertel,
would produce the pattern of discrepancy moderation that Kohler
Kerr, & Messe, 1999, pp. 330 331), we discuss, within the same theoret-
(1926, 1927) first observed. ical perspective, some other processes that could also be involved.
In our earlier work, we developed a feasible method for studying 3
Some research (Hertel, Kerr, & Messe, 2000) has yielded findings that
motivation gains under conjunctive task demands that retained are inconsistent with this goal-comparison explanation of the Kohler mo-
many of the features of Kohlers (1926, 1927) original paradigm tivation gain effect. However, as we discuss below, it remains a plausible
(see Hertel, Kerr, & Messe, 2000, pp. 585586, for a detailed explanation of the Kohler discrepancy effect, as its utility in this regard has
summary of these features). However, in doing so, we omitted yet to be examined empirically.
938 MESSE, HERTEL, KERR, LOUNT, AND PARK

Experiment 1 of the teams of their sex turned out to be greater than the average
performance of teams of the other sex.4
For the reasons discussed above, when participants were aware Teams were randomly assigned to either the individual or the conjunc-
of their relative capabilities, we expected an overall motivation tive group conditions. In the individual condition, participants were told
gain by the weaker worker that would be moderated by the that on Trial 1, 2 members, selected at random, would be performing in
discrepancy in coworkers abilities. More specifically, in this separate rooms, while their remaining teammates rested (also in separate
circumstance, we predicted that, as Kohler (1926, 1927) found, the rooms). Then, on Trial 2, the performer and rester roles would be switched,
highest motivation gains would occur under conditions of moder- and so on, until the end of the session. A performers score would be the
total number of seconds between the start of the trial and when your arm
ate capability discrepancy. In contrast, we predicted that, as in past
hits a flexiglas [i.e., fiberglass] bar. And, we will sum up all the scores
studies (e.g., Hertel, Kerr, & Messe, 2000), no such moderation of on all the trials today to come up with a total team score.
motivation gains would occur when the (weaker) participant had In the conjunctive group condition, the experimenter explained the same
no direct knowledge of dyad members relative capabilities. Trial 1 and Trial 2 procedure as in the individual condition. He then went
on to state that on Trial 3, the pair that had worked separately in Trial 1
Method would be put in the same room to perform together as a dyad, whose score
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

would be the total number of seconds between the start of the trial and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Participants. Eighty-four female undergraduate students at Michigan when either persons arm hits a flexiglas bar. That is, when either persons
State University participated in this study in partial fulfillment of a course arm hits her flexiglas bar, the trial is over for the dyad. Similarly, the pair
requirement. We recruited only women to take part because (a) there were who had performed separately in Trial 2 would then work together in a
considerably more women in the potential participant pool and (b) our past dyad in Trial 4. This pattern of alteration of individual and dyad trials
work (Hertel, Kerr, & Messe, 2000, Experiment 1), in which female and would be repeated until the session was over. As in the individual condi-
male participants were examined in same-sex teams, yielded no evidence tion, participants were informed that the total team score would be the sum
of sex effects. Two male experimenters conducted all sessions. of all the trial scores.
Design. The experiment used a 2 (task condition: individual control vs. The trials were conducted according to a fixed schedule (see Table 1)
conjunctive dyad) 2 (performance feedback: no vs. yes) 2 (performing that was the same for every session. Which experimenter worked in which
arm: dominant vs. nondominant) 2 (trials: first vs. second with a given room was alternated across experimental sessions. Participants were as-
arm) design, with the two last factors being within-subject variables. Four signed randomly to the letter roles of the schedule. Order of performing
participants took part in a session. Each session was assigned randomly to arm was counterbalanced across all conditions; in the conjunctive condi-
one of the four between-subject conditions. As a result, there were 16 tion, order of individual and group trials was held constant. Half the
participants in the individual task/no performance feedback condition, 16 participants performed their initial and second trials with their dominant
in the individual task/performance feedback condition, 28 in the conjunc- arm, the other half with their nondominant arm (see Table 1). The third and
tive dyad/no performance feedback condition, and 24 in the conjunctive fourth performance trials for each participant repeated the conditions of the
dyad/performance feedback condition. The dependent variable was the first two trials, but the participant now performed the task with her other
difference between task performance in the dyadic and the individual trials arm (again, see Table 1).
for a given arm, measured in terms of the time that a participant persisted In the feedback conditions, just before a pair of participants worked on
in holding the bar above the trip rod. the task for the second time, the experimenter announced to both how long
Procedure. Prior to each session, an experimenter made sure that none (in seconds) each, performing in separate rooms, had persisted at working
of the participants had any disabling arm, shoulder, or back injuries. on the previous trial as well as the discrepancy between the two scores.
Participants were also asked to put their watches and any other wrist Such feedback was also given at the equivalent point in the second
jewelry in their pocket or purse until the end of the session. The experi- sequence of trials, when the pair worked at the task with their other arm.
menter next distributed name tags printed with the identifying letters AD Thus, in the group conditions, previous individual performance informa-
to participants while he explained to them that they would be a 4-woman tion was provided to both members of a participant pair just before they
team. Participants wrote their name on their tag and fixed it on their blouse. began working together on the (now conjunctive) task. In the individual
We took these steps to maximize feelings of group identification (cf. conditions, each received this feedback (about her own and a teammates
Hertel, Kerr, & Messe, 2000; see also Gaertner, Mann, Murrell, & Dovidio, previous performance) just before she once again, in a separate room,
1989). performed the task alone. In contrast, no performance information of any
Participants were informed that their task would be to hold a metal bar kind was given to participants in the no-feedback conditionsthereby
above a trip bar for as long as they felt comfortable in doing so. This task reproducing the procedure that we followed in past studies.
was cast as a means of comparing the persistence of men and women at As noted above, 2 participants performed their Trial 1 task while their
physical work. Therefore, the average performance of 4-person male teams teammates waited to take their turn, each in a separate room. Then, the
ostensibly would be compared with the average performance of 4-person original pair of performers waited (and rested) while their teammates now
female teams. Participants were also told that to adjust for differences in performed Trial 2. This alternation of one pair performing then resting while
mean body strength, different weights, pretested for equivalence, were their teammate pair performed was repeated until every participant had worked
being used for male and female teams so that the task would be equally at the task four times (twice with each arm), as illustrated in Table 1.
difficult for the average man and woman. They were further told that they
would be performing this task over a series of trials (the exact number was Results
not disclosed).
We offered financial incentives to perform well by telling participants Performance in the individual condition. First, we analyzed
that all teams (including theirs) would earn a point for every second each the performance times in the individual conditions to ascertain and
team member held her or his bar up, and, for one lucky team chosen at later adjust for fatigue effects in later trials. Moreover, we explored
random, the point total would be converted to money at a rate of 5 per
point (up to a maximum of $160) to be distributed equally to the team
4
members. Moreover, to further increase interest in doing welland to One team was, in fact, chosen at random, and team members were
heighten the salience of the cover storywe told participants that a 15% actually paid according to this scheme. To be fair, as there were no male
bonus would be added to this teams earnings if the average performance teams, we also included the 15% bonus in the chosen teams pay.
KNOWLEDGE OF ABILITY AND MOTIVATION GAINS 939

Table 1 mance on the second trial across arms and feedback conditions as
Trial Schedule of Experiment 1 an index of fatigue. We then corrected for fatigue in the conjunc-
tive task conditions by multiplying the second trial scores of
Individual condition Dyad condition participants in these conditions by this value (1.197).5
Trial and arm Room 1 Room 2 Room 1 Room 2 Overall motivation gains. Consistent with past work (e.g.,
Hertel, Kerr, & Messe, 2000), we calculated the conjunctive dif-
1. dom. A B A B ference score (CDS) as an index of group motivation effects (gains
2. n.d. D C D C or losses) in the conjunctive dyad conditions. This score is the
3. dom. B A A, B
4. n.d. C D C, D difference (in seconds) between a dyads fatigue-corrected perfor-
5. n.d. A B B A mance score and the lower of the 2 dyad members individual
6. dom. C D C D performance scores. For example, suppose Participant A, when
7. n.d. B A A, B working alone, had persisted at the task for 180 s, whereas Par-
8. dom. D C C, D
ticipant B had persisted for 264 s. Moreover, their (fatigue-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Note: Letters AD stand for the 4 participants in each session. Nonper- corrected) performance score when working together as a dyad
This document is copyrighted by the American Psychological Association or one of its allied publishers.

forming persons in each trial were waiting in an additional room. dom. under conjunctive task demands was 225 s. This dyads CDS
dominant arm; n.d. nondominant arm. would be 45 s (i.e., 225 180).6 In the individual control condi-
tions, we were able to define a pseudodyad CDS score that was just
whether feedback of participants own and others performance the difference between the lower of the two corrected scores of the
scores per se affected the performance times. Also, we tested individuals on their second testing and the lower score of those
whether the kind of performing arm (dominant or nondominant) or same two individuals on their initial testing. These pseudodyad
the order of performing arms during the four trials had effects on CDS scores in the individual-control conditions provide an empir-
participants performance. Recall that participants performed with ical baseline against which to compare performance in the con-
the same arm on their first two performance trials, with a rest junctive conditions. Trials with the dominant and the nondominant
period between them (Trial Block 1), and then performed the two arm were treated as independent cases, as in 11 out of 41 dyads the
following trials with the other arm, again with a rest period in weaker person was not the same across the trial blocks.7
between (Trial Block 2). A 2 (individual vs. conjunctive task) 2 (performance feed-
The change in task performance across trials (i.e., second per- back) ANOVA of CDS scores yielded a significant main effect for
formance trial time first performance trial time) was examined the first factor, F(1, 79) 10.43, p .002.8 Performance on the
with a 2 (performance feedback) 2 (order of performing (fatigue-corrected) second trial for each arm was significantly
arms) 2 (performing arm) 2 (repetition: first vs. second higher when participants worked conjunctively in a team (M
performance trial per block) analysis of variance (ANOVA), with 36 s, SD 51.9) than in the individual baseline conditions
repeated measures on the last two factors. This analysis yielded
two significant main effects: for repetition, F(1, 28) 29.27, p
.001; and for arm, F(1, 28) 10.61, p .01. The significant 5
This value is very similar to those found and used as fatigue-correction
repetition effect reflected fatigue. Performance times on the first factors in earlier research that used the same experimental task (e.g.,
trial in each block (M 199.2 s, SD 45.0) were significantly Hertel, Kerr, & Messe, 2000; Hertel, Kerr, Scheffler, et al., 2000).
longer than on the second (M 166.4 s, SD 39.4). The effect 6
It is possible that there were a few instances in which, when the
of the arm indicated that participants performed better with their participants were working together, the dyad member who had the better
dominant arm (M 199.2 s, SD 33.4) than their nondominant individual performance score quit first and, thus, was the one responsible
arm (M 174.8 s, SD 40.3). However, the Repetition Arm for ending the group trial. To the extent that this may have happened,
however, it would have caused the CDS to underestimate the previously
interaction was not significant (F 1), which suggests that the
weaker coworkers willingness to persist (past the point at which her
fatigue effect was similar across arms. Also, no significant main coworker quit). Therefore, at worst, this index can be considered an
effect for the order factor or the Repetition Order interaction estimate of the lower bound of the actual Kohler motivation gain effect.
occurred (Fs 1). This lack of moderation of fatigue effects by 7
Note that separate analyses for each trial block yielded equivalent
performing arm or order of performing arms is consistent with results. See the Results and Discussion section of Experiment 2 for a more
similar results of earlier research that used the same task (e.g., detailed presentation of these findings.
Hertel, Kerr, & Messe, 2000; Hertel, Kerr, Scheffler, et al., 2000). 8
A first screening of the discrepancy data revealed one extreme outlier
The ANOVA also yielded a marginally significant main effect case in the first trial block of the group condition with feedback. In this
for performance feedback, F(1, 28) 3.01, p .10, which case, the discrepancy score (329 s) was more than three standard deviations
suggests that feedback may have improved performance. Scores (SD 78.0) above average (M 72.0 s). Because the discrepancy of the
were somewhat higher (M 193.1, SD 36.4) when participants participants in this dyad was much smaller in the second trial block with
were told (within each block) their own and their fellow team- the other arm (183 s), this might suggest that, in the first trial, the weaker
mates earlier performance times just prior to beginning their participant was somehow distracted by unsystematic influences. Thus, we
excluded this case from further analyses. (Omission of this case did not
second trial than when no feedback was given (M 172.4,
change the significance of the general Kohler motivation gain effect.) No
SD 30.0). However, the Feedback Repetition interaction was such outlier was observed in the other group or individual conditions. (The
not significant (F 0.1), which suggests similar fatigue effects in outlier status of the one omitted case was even clearer when the analysis
both experimental conditions. was computed not within each experimental condition but across all four
On the basis of these results, for the individual-condition data, conditions, resulting in a mean discrepancy of 58.0 s and a standard
we computed the ratio of performance on the first trial to perfor- deviation of 54.0.)
940 MESSE, HERTEL, KERR, LOUNT, AND PARK

(M 0.0 s, SD 38.8). Thus, as predicted (Hypothesis 1), we effects is .29, with a 95% confidence interval of .15 r .66.
replicated the Kohler motivation gain effect. Thus, the observed correlation in this condition appears to fall well
Feedback per se (i.e., whether participants received information within the limits of sampling error. Moreover, polynomial regres-
relevant to their relative abilities, without taking into account the sion analysis indicated that higher order, nonlinear components did
magnitude of this discrepancy) did not affect overall motivation not explain more variance (quadratic model: R2 .31; adjusted
gain, as neither its main effect nor the interaction with task demand R2 .20; the latter adjusted for sample size and number of
(individual vs. conjunctive) even approached significance (both predictors) than did the simple linear model (R2 .29; adjusted
Fs 1). In the conjunctive condition in which participants re- R2 .24). These results indicate that in this condition, there was
ceived feedback about their own and their partners capability, the little, if any, actual association of any kind between magnitude of
weaker participants performance increased about 24.8% over their ability discrepancy and (pseudo-) motivation gaina reasonable
efforts during the individual trials. When participants did not result, given that these participants always worked alone in sepa-
receive feedback, the increase of the weaker persons performance rate locales and never received any feedback on their own or their
was about 22.1%. However, these findings do not address the teammates performance.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

impact on group motivation gains of knowing specific discrepan- A very similar pattern occurred in the individual control with
This document is copyrighted by the American Psychological Association or one of its allied publishers.

cies of varying magnitude, which is the focus of Hypothesis 2. We


feedback condition. Here, too, the linear correlation was moder-
now turn to analyses that explore this issue.
ately high, r(14) .47, but well within the estimated limits of
Relationship between member ability discrepancy and motiva-
chance. And, once again, higher order effects did not explain more
tion gain. We calculated magnitude of discrepancy by subtract-
variance (quadratic model: R2 .22; adjusted R2 .10) than did
ing, for a particular arm, the worse of the 2 dyad members
the simple linear model (R2 .21; adjusted R2 .16). Thus, it
first-trial performance score from the other, better first-trial score.
appears that for participants who worked alone, specific knowl-
Recall that on these trials (two per pair; see Table 1) participants
always worked as individuals in separate rooms. To explore edge of their own and a teammates abilities did not moderate task
whether workers discrepancy in ability, as reflected by these performance in the subsequent (individual) work trial.
scores, moderated the extent of motivation gains in the conjunctive A similar analysis of the scores from the conjunctive group
conditions, we computed within each condition sets of linear and without feedback condition once again yielded a positive apparent
nonlinear correlations between the ability discrepancy index and correlation between discrepancy and the motivation gain index,
the CDS measure. although this time the coefficient was somewhat lower, r(26)
In the individual control without feedback condition, the linear .20. Exploring higher order solutions through polynomial regres-
correlation between discrepancy and CDS appeared to be substan- sion analysis again provided no better fit (quadratic model: R2
tial, r(14) .54. However, as Hertel, Kerr, and Messe (2000) .04; adjusted R2 .03) than the simple linear model (R2 .04;
showed, regression-to-the-mean effects alone tend to produce an adjusted R2 .00). (See Figure 1, which shows the best fit
apparent positive correlation between the discrepancy in dyad regression line for this data set.) Thus, similar to the pattern in the
members individual performance scores and their subsequent individual control conditions, there was no indication of a reliable
dyad score. Moreover, a previous Monte Carlo simulation based linear or curvilinear moderation of motivation gains by discrep-
on similar experimental conditions (Hertel, Kerr, & Messe, 2000) ancy in coworkers abilities. These results mirror the findings of
estimated that the mean correlation due to regression-to-the-mean our earlier work (Hertel, Kerr, & Messe, 2000), as they also fail to

Figure 1. Motivation gain scores (fatigue corrected) in no-knowledge condition as a function of coworkers
first-trial ability discrepancy.
KNOWLEDGE OF ABILITY AND MOTIVATION GAINS 941
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 2. Motivation gain scores (fatigue corrected) in knowledge condition as a function of coworkers
first-trial ability discrepancy.

replicate the connection between discrepancy and motivation gain the-mean artifact) that our past work found; knowledge would
that Kohler (1926, 1927) had observed. yield maximum increases in the weaker members effort when
However, as predicted by Hypothesis 2, evidence of this mod- there was a moderate discrepancy, the pattern that Kohler (1926,
eration occurred for the conjunctive group condition when ability 1927) had found. The results of Experiment 1 support this predic-
feedback was given. Although this condition did not generate the tion and, in doing so, identify a necessary condition for generating
typical, apparently robust linear correlation between discrepancy the Kohler discrepancy moderation effect.
and the motivation gain index, r(22) .09, n 0.23, ns (R2 .00; Experiment 1 also replicates our past demonstrations of
adjusted R2 .04), a quadratic solution improved the amount of Kohlers (1926, 1927) overall motivation gain effect (e.g., Hertel,
explained variance considerably (R2 .20; adjusted R2 .12; no Deter, & Konradt, 2001; Hertel, Kerr, & Messe, 2000; Hertel,
further improvement by a cubic solution), generating significant Kerr, Scheffler, et al., 2000; Lount et al., 2000). Once again, across
beta weights for both the linear and the quadratic component (B differences in relative ability, it was generally the case that the
1.54, p 0.04, and B 1.51, p 0.04, respectively).9 As weaker person tried harder when teamed with a more capable
shown in Figure 2, the curvilinear regression function showed the coworker than when performing as an individual. Therefore, the
inverted-U pattern that our hypothesis predicted. These results results of Experiment 1 basically reproduce the pattern of perfor-
suggest that knowledge of the relative capability within a group is mance that Kohler had foundan overall motivation gain that was
indeed a precondition for the Kohler discrepancy effect to occur. maximal when the discrepancy in coworkers relative ability was
moderate but only when participants had foreknowledge of each
Discussion others task ability. In contrast, in the absence of such foreknowl-
edge, Experiment 1 findings reproduce the seemingly linear rela-
The primary goal of this study is to explore the possibility that tionship between discrepancy and motivation gains that our past
past lack of success in replicating Kohlers (1926, 1927) investigations (e.g., Hertel, Kerr, & Messe, 2000) consistently
inverted-U pattern of group motivation gains was due to research generated.
participants not possessing some vital informationspecific and It is important to note that in both this experiment and Kohlers
reliable knowledge about each others performance ability be- (1926, 1927) original investigations (as well as in other studies;
fore they worked together. It is almost certain that Kohlers par- e.g., Hertel, Kerr, & Messe, 2000), discrepancy in ability had to be
ticipants (members of the same rowing club who exercised, prac- somewhat confounded by absolute ability. Given that chance de-
ticed, and competed together) had such knowledge. In contrast, in termines who is paired with whom, it is much more probable that
our past work, participants were never given any feedback about the weaker member of a dyad will be markedly low in absolute
anyones (their own, their coworkers, their teammates) individ- ability when ability discrepancy is large than when it is small. In
ual performance, so they had no valid way of even estimating other words, the less able the weaker member is, the larger the
relative abilities before a dyad trial began. difference in the relative ability of the dyad members is likely to
On the basis of this difference, we predicted that participants be. It is possible, then, that this natural connection between abso-
knowledge of their own and coworkers ability would moderate lute and relative ability was responsible for at least a part of the
the pattern of groups motivation gains: lack of knowledge would
yield the apparent linear relationship between ability discrepancy
9
and the weaker members effort (attributable to a regression-to- Centering the predictor variable yielded very similar findings.
942 MESSE, HERTEL, KERR, LOUNT, AND PARK

discrepancy moderation effect that Kohler and we (in the ability would be greatest when there appeared to be a moderate difference
feedback condition) found. in the coworkers relative ability, thereby replicating the Kohler
For instance, it could be that participants who are truly low in discrepancy effect.
ability cannot increase their actual performance much, even if
highly motivated to do so. If so, then the apparent lower motiva-
tion gain seen in high-discrepancy dyads is a consequence of the Method
greater frequency with which low-ability participants fell into this Participants. Sixty-three female undergraduates at Michigan State
condition. In any event, to determine whether the Kohler discrep- University participated in the study. As in Experiment 1, by taking part in
ancy moderation effect occurs independently of the weaker dyad this research they received credit toward fulfillment of a research require-
members absolute ability, we conducted Experiment 2, in which ment in their introductory psychology course. At least 3 weeks prior to
we experimentally manipulated (apparent) relative ability indepen- taking part in the study, participants had completed an apparently uncon-
dently of participants absolute ability. nected interests and skills survey on which one item asked them to rate
Manipulating ability discrepancy in this second study also al- themselves (using a 5-point scale) on how physically strong they were.
Overview and design. Participants were randomly assigned to either an
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

lows us to address another issue that Experiment 1 left unresolved.


individual control condition or one of three dyad conditions in which
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Recall that, as in past studies (e.g., Hertel, Kerr, & Messe, 2000), participants worked together with a (seemingly) very slightly, moderately,
we constructed the CDS index of motivation gain in Experiment 1 or much stronger partner. Individual control participants performed the
by subtracting the lower of the two individual performance scores weighted raised-arm persistence task alone, in separate rooms, for four
from the dyad score. As we noted above (see Footnote 6), this trials, alternating between their dominant and nondominant arms. Partici-
procedure, at worst, underestimates the actual Kohler motivation pants in the dyad conditions also first performed the task individually, once
gain effect to the extent that at least some dyad scores were a for each arm; then they were paired with a (seemingly) stronger coworker
consequence of the previously more capable coworker (rather than to perform conjunctively for two trials, again alternating arms. The arm
her previously weaker partner) quitting first. But, given that they that participants in both conditions used first (and then again on the third
would lead to underestimation of actual motivation gain in the trial) versus the one they used for the second and fourth trials was
systematically varied. All participants rested for the same amount of time
weaker coworker, such possible dyad performance outcomes also
between each performance trial.
had potential relevance for interpreting the effects of ability dis- The manipulations reflect an experimental design with the following
crepancy that we found in Experiment 1, particularly if the mag- factors: 4 (work condition: individual control [n 13]; conjunctivesmall
nitude of the underestimation differed as a function of discrepancy. discrepancy in ability [n 16]; conjunctivemoderate discrepancy [n
For instance, if such events occurred more frequently when 18]; conjunctivelarge discrepancy [n 16]) 2 (arm use order: domi-
individual ability scores were either very similar or very dissimilar, nant first vs. nondominant first) 2 (same-arm trial set: Trials 1 and 3 vs.
compared with when they were moderately discrepant, then actual Trials 2 and 4; a repeated measure).
motivation gains would be underestimated and thus appear to be Procedure. The procedure was essentially the same as that used in
lower, in precisely those conditions in Experiment 1 (and Kohler, Experiment 1, with the following notable exceptions:
1926, 1927) that yielded lower motivation gain scores. Although 1. Instead of having participants hold a weighted bar, we attached a
commercial weight bracelet basically a broad strip of vinyl cloth with
such a scenario is highly unlikely for those dyads in which there
numerous pockets (for housing metal weight rods) and a velcro fas-
was a large discrepancy in individual ability scores, it possibly tenerto their appropriate wrist. For actual participants, the bracelet,
could have occurred with some frequency when dyad members including the metal rods attached to it, weighed 0.8 kg. For the confeder-
were more evenly matched. In Experiment 2, however, we used a ates (three female undergraduate assistants who had been carefully trained
procedure that ensured that all participants had a lower individual to enact the role of the coworker in the dyad conditions), the metal weights
performance score than their dyad partner and that they always had been replaced with much lighter pseudoweightssegments of wooden
were the person who was responsible for ending the dyad trial (by dowels of the same length that had been painted black to closely resemble
quitting first). As such, in this study, any observed performance the actual metal rods. As such, their bracelet was light enough to permit
increases should reflect, with reasonable accuracy, actual motiva- them to last as long as necessary and, thus, perform appropriately for each
tion gains in the weaker coworker. discrepancy condition.
2. As noted above, all participants performed individually in both
Trial 1 and Trial 2. Those assigned to the individual control conditions
Experiment 2 merely repeated this sequence (with an appropriate rest period in between).
Before Trials 3 and 4, they were given performance feedback for the earlier
In this study, we first had all participants engage in the weighted trial with the same arm for both themselves and a teammate in another
raised-arm persistence task as individuals. To obtain a potential room; their own score was veridical, whereas their teammates score was
productivity baseline that reflected the effects of fatigue on per- always about 1.4 times better than theirs had been.10
formance, we then had some (individual control) participants do so For participants in the dyad conditions, a fellow teammate (actually a
again. We paired others with a confederate who had been carefully confederate) was brought to their room after Trial 2, so that they could then
trained to appear very slightly better, moderately better, or much work together on Trials 3 and 4. At this point, these participants were given
better at the task. We also provided participants with false feed- conjunctive task instructions, as in Experiment 1, and then informed how
back about individual performances, prior to a dyad trial, that was
consistent with the (apparently) stronger coworkers subsequent 10
To minimize the number of participants that were necessary to con-
actions. As in Experiment 1, we predicted that (a) across dyad duct this study, we decided to provide all the individual controls with the
conditions, participants would work harder as (weaker) coworkers same moderate-discrepancy feedback. We felt justified in doing so by the
than as individuals, thereby once again replicating the Kohler fact that variations in discrepancy information did not affect performance
overall motivation gain effect; and (b) these motivation gains in the individual control condition of Experiment 1.
KNOWLEDGE OF ABILITY AND MOTIVATION GAINS 943

they and their coworker each had performed on Trial 1 (the individual trial The ANOVA also yielded a significant Work Condition Trial
with the same arm as the one they were about to use for Trial 3). Set interaction, F(3, 55) 3.21, p .03. However, parallel
Information about the actual participants Trial 1 score was veridical, planned comparisons within each trial set (see Winer, Brown, &
whereas the coworkers score was systematically varied. In the small Michels, 1991, pp. 529 531) showed that there were reliable
discrepancy condition, the feedback indicated that the coworker lasted a
overall motivation gains in each set, t(110) 2.16 and 3.97, ps
few seconds longer, and the experimenter commented that the 2 of them
.02, for the first (Trial 3 Trial 1) and second (Trial 4 Trial 2)
had performed just about the same. In the moderate discrepancy condition,
the dyad was informed of the coworkers Trial 1 time, such that it was trial sets, respectively, but this effect was somewhat stronger for
always about 1.4 times longer than the actual participants, and the exper- the latter.
imenter explicitly noted that the coworker had done moderately better. In Hypothesis 2: Ability discrepancy moderation. Given that par-
the large discrepancy condition, the coworkers Trial 1 score was stated to ticipants in the conjunctive conditions were given feedback reflect-
be a time that worked out to be just over twice as long as the actual ing three distinct differences in self coworker abilities, we ex-
participants, and the experimenter remarked that the coworker had done pected to replicate the results of both Experiment 1 and Kohler
substantially better. This same procedure was repeated before Trial 4, with (1926, 1927) by finding the largest motivation gain under moder-
dyad members now being given specific feedback on Trial 2 performances.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ate discrepancy. The data support this hypothesis in that inspection


3. During the conjunctive task trials, the coworker (confederate) per-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of the average (fatigue-corrected) motivation gain for each dis-


formed in a manner that was congruent with the feedback about her
crepancy condition revealed the expected inverted-U pattern
individual trial performance that the dyad had been given. In the small
discrepancy condition, she basically mimicked the actions of the actual (Ms 49.6 s, 65.0 s, 44.8 s; SDs 48.5, 50.7, 45.1, for the small,
participantshowing similar signs of fatigue (e.g., lowering her arm a bit moderate, and large discrepancy conditions, respectively), but only
and then raising it back up with difficulty; approaching the trip rod at the weakly so, as the planned comparison was just marginally signif-
same pace and with the same signs of exertion)with a delay of less than icant t(55) 1.36, p .09.
a second. Of course, in this (as well as the other two discrepancy condi- However, given the significant Condition Trial Set interaction
tions) the confederate always took care to lower her hand past the trip wire noted above, it was appropriate to test Hypothesis 2 separately
only after the actual participant had already done so. Therefore, in every within each trial set. These contrasts revealed that the moderation
case, the actual participants performance determined the dyad score. In the effect occurred for the second set, t(110) 2.12, p .02, but not
moderate discrepancy condition, the confederate displayed somewhat
the first, t(110) 0.30, p .40. Figure 3, which presents moti-
fewer signs of fatigue, and, when doing so, she delayed showing them for
vation gain scores separately for each trial set, confirms that the
about 10 15 s after the actual participant had. In the large discrepancy
condition, the confederate displayed few signs of fatigue or stress during first set showed no evidence of a discrepancy effect, thereby
the trial, basically conveying the sense that she could persist substantially
longer than the participant could.

Results and Discussion


The second trial first trial (i.e., Trial 3 Trial 1; Trial 4
Trial 2) difference in how long each participant persisted at the
task for each arm (i.e., for each trial set) was used as the index of
performance change (loss or gain). The hypotheses were tested
with planned comparisons conducted within the framework of an
overall 4 (work condition: individual, dyadsmall ability discrep-
ancy, dyadmoderate ability discrepancy, dyadlarge ability dis-
crepancy) 2 (arm use order) 2 (trial set; a repeated measure)
ANOVA of these scores. Specifically, the predicted overall moti-
vation gain was tested with an orthogonal contrast that compared
the individual condition with the three dyad conditions, combined.
The predicted curvilinear discrepancy effect was tested with a
contrast between the moderate-discrepancy condition and the
small- and large-discrepancy conditions, combined.
Hypothesis 1: Overall motivation gain. Replicating past find-
ings, performance change scores (across the two trial sets) showed
a substantial decrease for participants in the individual control
condition (M 30.88 s per set, SD 52.20) and a substantial
increase for participants in the three conjunctive conditions
(M 22.76 s per set, SD 48.08). The planned contrast revealed
that, as predicted, the difference in the two conditions was highly
significant, t(55) 3.34, p .001, reflecting the same pattern of
overall motivation gain that Kohlers (1926, 1927) rowing club
members generated 75 years ago and that we consistently have
found in our investigations (e.g., Hertel et al., 2001; Hertel, Kerr,
& Messe, 2000; Hertel, Kerr, Scheffler, et al., 2000; Lount et al., Figure 3. Motivation gain scores (fatigue corrected) in each trial set as a
2000). function of ability-discrepancy conditions.
944 MESSE, HERTEL, KERR, LOUNT, AND PARK

mirroring the findings of past work (e.g., Hertel, Kerr, & Messe, experience that Trial 3 provided. And it was not until Trial 4 that
2000; Lount et al., 2000), in which participants were not sure of this now-confirmed ability difference affected their level of effort.
relative ability differences when they performed as dyad members. In contrast, it seems reasonable to assume that in Experiment 1, in
In contrast, Figure 3 also shows a clear Kohler discrepancy effect which no confederates were used, appearance cues were likely to
for the second trial set. be more congruent with the (completely veridical) performance
Given these results, we went back to the performance data of feedback provided to participants. If, in fact, the feedback was
Experiment 1 to explore whether the first and the second trial more immediately credible in Experiment 1, it would have been
blocks also produced differences in discrepancy moderation. How- more likely to affect performance on the initial dyad trial. And,
ever, curvilinear regression analysis of each block separately re- given that the subsequent experience was likely to further confirm
vealed basically equivalent effects: a marginally significant qua- their beliefs, it is reasonable that the Kohler discrepancy effect
dratic solution (B 1.65, p 0.06, and B 1.53, p 0.07 for would then persist in the second dyad trial as well, as the results
the linear and the quadratic component; R2 .30, adjusted R2 indicate.
.13) occurred in the first trial block, and a very similar pattern Taken together, the results of Experiments 1 and 2 provide
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

occurred for the second trial block (i.e., a significant quadratic converging evidence that for conjunctive tasks, participants rela-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

solution, B 2.09, p 0.05, and B 2.03, p 0.05, for the tive ability can moderate motivation gains in the weaker worker; as
linear and the quadratic component; R2 .29, adjusted R2 .13). Kohlers (1926, 1927) data first suggested, increases in effort are
Thus, the two studies apparently differed in terms of the immedi- greatest when ability differences are moderate and well known.
acy with which the Kohler discrepancy effect emerged. Note that this is essentially a group composition effect: Different
Finally, with regard to the Kohler discrepancy effect found in combinations of participant attributes generated different group
Experiment 2, it is important to emphasize that because partici- performance outcomes. Most past investigations of homogeneous
pants were randomly assigned to work with a partner who appar- versus heterogeneous group composition (cf. Forsyth, 1999, pp.
ently was slightly, moderately, or much stronger than they, dis- 277280) have examined how the mix of qualitatively different
crepancy in ability could not have been systematically related to attributesfor example, personality (Bond & Shiu, 1997), ethnic-
participants own ability level. Moreover, their subjective, self- ity (McLeod, Lobel, & Cox, 1996), and gender (e.g., Kerr &
rated level of physical strength (assessed well before they took part MacCoun, 1984; Lount et al., 2000)affect performance. In con-
in this study) support this assumption, as analyses revealed that trast, the present research demonstrates that quantitative differ-
scores on this measure (a) did not significantly vary across the ences in the same attribute (task ability) can also have a predictable
experimental conditions (all Fs 1) and (b) was not significantly impact on how well group members function.
related to motivation gains, r(56) .11, p .40. These find- Moreover, most past explorations of the impact of member
ings suggest that, to the extent that self-ratings accurately reflect diversity on performance have addressed group composition ef-
actual strength, it is unlikely that absolute level of abilitywhich fects from a rather simple, linear perspective, basically posing such
covaries with the magnitude of naturally occurring ability questions as, Do heterogeneous groups outperform homogeneous
discrepancies contributes to the Kohler effects. groups? In contrast, as we discuss in more detail elsewhere (Hertel
et al., 1999), the Kohler discrepancy effect is nonlinear, as weaker
General Discussion participants do best under moderate levels of ability diversity,
compared with working in teams that are either much more ho-
Experiment 2 provides convincing evidence that it is the differ- mogeneous or much more heterogenous. Our results suggest, then,
ence in relative ability rather than absolute ability per se that that the impact of diversity on group functioning may be a far more
underlies the Kohler discrepancy effect. Moreover, manipulating complex question than is commonly presumed. Thus, it is likely
specific levels of ability differencesrather than allowing them to that future work on this issue will prove most fruitful if it explores
occur naturally, as Experiment 1 and Kohler (1926, 1927) had how much diversity in groups is optimal under various task de-
generated more visually clear-cut evidence of how relative ability mands rather than focusing on whether diversity linearly improves
can moderate motivation gains than inspection of scatter plots (or detracts from) performance.
typically can (compare, e.g., Figures 2 and 3). However, it is also Another implication of the current research is that there are
likely that the difference in the manner in which the two experi- circumstances in which the performance effort of less capable
ments examined relative abilitymanipulation versus natural oc- workers possibly could be increased if they are misinformed about
currenceplayed a role in determining when the Kohler discrep- relative abilities. Our results suggest, for example, that less able
ancy effect emerged. workers in a situation where one or more of their colleagues has
All three of the confederates used in Experiment 2 had rather substantially greater ability would likely try harder on conjunctive
average body builds. And, as participants were randomly assigned tasks if they were led to believe this discrepancy was less severe
to discrepancy conditions, it seems very likely that in several than it actually was. Similarly, it is likely that when group mem-
instances (e.g., whenever either a petite or a relatively muscular bers are rather homogeneous in ability, motivation to perform well
participant was paired with a supposedly equal-ability partner, or on conjunctive tasks would be increased if workers erroneously
when a relatively muscular participant was paired with a suppos- perceived that one or more of their coworkers was moderately
edly much stronger partner), the Trial 1 feedback was not highly better than they.
credible to participants, until they saw for themselves the level at The present research provides reasonable evidence that under
which their coworker could perform. If so, participants may well certain conditionswhich seem likely to occur rather frequently in
have needed some explicit validation of the verbal feedback they natural settingsthe discrepancy moderation effect that Kohler
had been given, which they then obtained through the first-hand (1926, 1927) first reported is a reproducible phenomenon. Thus, it
KNOWLEDGE OF ABILITY AND MOTIVATION GAINS 945

seems appropriate for future work to begin to focus on identifying when and why the group context sometimes stimulates people to
the socialpsychological mechanisms that underlie it. Although, give their best efforts.
as we discuss in the introduction to this article, a case can be made
for a number of processes, we believe that perceived instrumen-
References
tality is particularly likely to be involved in generating the Kohler
discrepancy effect. Bond, M. H., & Shiu, W. Y. (1997). The relationship between a groups
Our past research (e.g., Hertel, Kerr, & Messe, 2000, Experi- personality resources and the two dimensions of its group processes.
ment 2; Lount et al., 2000) suggests that the overall Kohler Small Group Research, 28, 194 217.
motivation gain effect is more likely to be due to a sense in less Forsyth, D. R. (1999). Group dynamics (3rd ed.). Belmont, CA: Brooks/
able participants that their efforts are particularly crucial to team Cole Wadsworth.
success than to some goal-comparison mechanism, as Stroebe et Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J. F. (1989). Reducing
group bias: The benefits of recategorization. Journal of Personality and
al. (1996) proposed. For example, we found significant motivation
Social Psychology, 57, 239 249.
gains when task demands were conjunctive but not when they were Harkins, S. G., & Petty, R. E. (1982). Effects of task difficulty and task
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

additive (where one teammate can continue to perform after the uniqueness on social loafing. Journal of Personality and Social Psychol-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

other has stopped; see Hertel, Kerr, & Messe, 2000, Experiment 2). ogy, 43, 1214 1229.
However, within the goal-comparison perspective, one would ex- Hertel, G., Deter, C., & Konradt, U. (2001). Motivation gains in computer-
pect equivalent motivation gains across conditions of task de- mediated work groups. Manuscript submitted for publication.
mands, as it always was the case that coworkers could observe Hertel, G., Kerr, N. L., & Messe, L. A. (1999). Revisiting the Kohler effect:
each other perform and, thus, use this information as a basis for Does diversity enhance motivation and performance in groups? Psy-
chologische Beitrage, 41, 320 337.
setting their own performance goals. Moreover, past work has
Hertel, G., Kerr, N. L., & Messe, L. A. (2000). Motivation gains in
yielded evidence from responses to a brief posttrial questionnaire performance groups: Paradigmatic and theoretical developments on the
that weaker coworkers are aware that their efforts matter more Kohler effect. Journal of Personality and Social Psychology, 79, 580
under conjunctive task demands (see Hertel, Kerr, & Messe, 2000; 601.
Lount et al., 2000). Hertel, G., Kerr, N. L., Scheffler, M., Geister, S., & Messe, L. A. (2000).
As we explained above, perceived instrumentality differences Exploring the Kohler motivation gain effect: Impression management
can also potentially underlie the Kohler discrepancy effect. For and spontaneous goal setting. Zeitschrift fur Sozialpsychologie, 31, 204
instance, for this conjunctive task, how important weaker cowork- 220.
Hinsz, V. B. (1995). Goal setting by groups performing an additive task: A
ers perceive their efforts to be and, thus, how hard they try could
comparison with individual goal setting. Journal of Applied Social
be a function of their considering how much their increased effort
Psychology, 25, 965990.
would help their group, compared with how much it could achieve Karau, S. J., & Williams, K. D. (2001). Understanding motivation in
given their more capable partners potential level of performance. groups: The collective effort model. In M. E. Turner (Ed.), Groups at
However, as we also summarized above, it also is plausible that work: Theory and research. Applied social research (pp. 113141).
goal comparison processes underlie this effect, as Stroebe et al. Mahwah, NJ: Erlbaum.
(1996) have proposed, and that other mechanisms (e.g., implicit Kerr, N. L., & MacCoun, R. (1984). Sex composition of groups and
competition; cf. Lount et al., 2000) may be involved as well. member motivation II: Effects of relative member ability. Basic and
Thus, future work needs to explore more directly the validity of Applied Social Psychology, 5, 255271.
Kohler, O. (1926). Kraftleistungen bei Einzel- und Gruppenabeit [Physical
an array of possible explanations for why the weaker coworker on
performance in individual and group situations]. Industrielle Psycho-
a conjunctive task tries the hardest when there is a moderate technik, 3, 274 282.
discrepancy in abilities. Research also needs to be conducted to Kohler, O. (1927). U ber den Gruppenwirkungsgrad der menschlichen
explore other possible processes that could be involved in produc- Korperarbeit und die Bedingung optimaler Kollektivkraftreaktion [On
ing both the Kohler motivation gain effect and the Kohler dis- group efficiency of physical labor and the conditions of optimal collec-
crpeancy effectas well as other types of group motivation gains tive performance]. Industrielle Psychotechnik, 4, 209 226.
(e.g., social compensation; Williams & Karau, 1991) and group Latane, B., Williams, K., & Harkins, S. (1979). Many hands make light the
diversity effects more generally. For example, there is evidence work: The causes and consequences of social loafing. Journal of Per-
sonality and Social Psychology, 37, 822 832.
(e.g., Lount et al., 2000; Stroebe et al., 1996) that suggests that
Locke, E. A., & Latham, G. P. (1990). A theory of goal-setting and task
self-presentation concerns may, under certain conditions, motivate performance. Englewood Cliffs, NJ: Prentice Hall.
people to try harder when working collectively. It appears, then, Lount, R. B. Jr., Messe, L. A., & Kerr, N. L. (2000). Trying harder for
that it would be fruitful to explore the extent to which self- different reasons: Conjunctivity and sex differences as bases for moti-
presentation also generates the types of diversity moderation ef- vation gains in performing groups. Zeitschrift fur Sozialpsychologie, 31,
fects on group performance that the current work and past studies 221230.
(cf. Wood, 1987) have demonstrated. McLeod, P. L., Lobel, S. A., & Cox, T. H. (1996). Ethnic diversity and
From this discussion, it is apparent that extensive empirical creativity in small groups. Small Group Research, 27, 248 264.
Steiner, I. D. (1972). Group process and productivity. New York: Aca-
work remains to be performed before a comprehensive understand-
demic Press.
ing of group motivation gains can be achieved. Taken together,
Stroebe, W., Diehl, M., & Abakoumkin, G. (1996). Social compensation
however, research to date has provided a promising start toward and the Kohler effect: Toward a theoretical explanation of motivation
identifying the major mediators and moderators of this important gains in group productivity. In E. Witte & J. Davis (Eds.), Understand-
group phenomenon. Given this foundation, it is likely that subse- ing group behavior: Consensual action by small groups (Vol. 2, pp
quent investigations will soon yield a more integrative picture of 37 65). Mahwah, NJ: Erlbaum.
946 MESSE, HERTEL, KERR, LOUNT, AND PARK

Stroebe, W., & Frey, B. (1982). Self-interest and collective action: The Witte, E. H. (1989). Kohler rediscovered: The anti-Ringelmann effect.
economics and psychology of public goods. British Journal of Social European Journal of Social Psychology, 19, 147154.
Psychology, 21, 121137. Wood, W. (1987). Meta-analytic review of sex differences in group per-
Triplett, N. (1898). The dynamogenic factors in pacemaking and compe- formance. Psychological Bulletin, 102, 5371.
tition. American Journal of Psychology, 9, 507533. Zajonc, R. (1965). Social facilitation. Science, 149, 269 274.
Vroom, V. H. (1964). Work and motivation. New York: Wiley.
Williams, K. D., & Karau, S. J. (1991). Social loafing and social compen-
sation: The effects of expectations of co-worker performance. Journal of Received June 13, 2001
Personality and Social Psychology, 61, 570 581. Revision received November 16, 2001
Winer, B. J., Brown, D. R., & Michels, K. M. (1991). Statistical principles
Accepted November 29, 2001
in experimental design (3rd ed.). New York: McGraw-Hill.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

You might also like