JOURNAL OF EXPERIMENTAL SOCIAL PSYCHOLOGY 10, 371-3&1 (1974)
The Ringelmann Effect: Studies of Group Size
and Group Performance1
ALAN G. INGHAM
University of Washington
AND
GEORGE LEVINGER,~ JAMES GRAVES, AND VAUGHN PECKHAM
University of Massachusetts, Amherst
Ringelmann’s classic finding-that the addition of co-workers in
a rope-pulling task leads to a linear decrement in the individual
group member’s average performance-was reexamined experimentally.
Study I attempted to replicate the effect, using groups
of subjects ranging in size from 1 to 6. Performance dropped
signticantly as group size was increased from one individual to
two or to three, but the addition of a fourth, fifth, or sixth member
produced insignificant additional decrements; thus, the effect
was not linear but curvilinear. Study II was designed to examine
possible sources of performance loss, separating the factors of
“coordination” and “motivation” loss (Steiner, 1972). The possibility
of intermember incoordination was eliminated, while motivation
loss remained free to vary: Each experimental subject
pulled alone, and in “groups” where he believed there were from
one to five other members. Once again, individual performance
declined significantly with the addition of the first and second
perceived co-worker, but then leveled off for perceived groupsizes
three to six. Some implications are discussed.
In 1927, the German industrial psychologist, Walther Moede, reported
an interesting finding obtained by his student Ringelmann. According
to Moede, Ringelmann had observed that when groups of co-workers
pulled on a rope, their collective group performance was inferior to
‘The research was facilitated, in part, from funds from the Graduate Research
Council of the University of Massachusetts and from Grant HD-04319 from the
U. S. Public Health Service to the second author. We are indebted to Benjamin
Ricci for his generosity in making available his laboratory and equipment, and to
Ivan Steiner for his valuable advice.
‘Requests for reprints should be addressed to George Levinger, Departmentof
Psychology, University of Massachusetts, Amherst, MA 01002.
371
Copyright @ 1974 by Academic Press, Inc.
All rights of reproduction in any form reserved,
372 INGHAM ET AL.
the sum of their individual performances. As groups increased in size
in the tug-of-war (rope pulling) task, the discrepancy between a group’s
actual performance and its potential performance increased progressively.
If we assume that one person individually performed at 100% of
his ability, members of Ringelmann’s dyads performed at 93% of their
potential average ability, members of his triads at 85% of their potential
average ability, and members of his groups of eight at a mere 49% of
their average ability ( Moede, 1927, p. ZOO). Apart from Moede’s summary
description, we have been unable to find any additional information
about the Ringelmann study despite its widespread citation in
American literature on group performance (Dashiell, 1935; Davis, 1969;
Steiner, 1966, 1972; Zajonc, 1966).
In the context of recent theoretical developments (Steiner, 1972),
Ringelmann’s study has assumed renewed significance. Steiner suggested
that productivity is dependent upon three factors: task demands, individual
and group resources, and the process by which resources are
mobilized. Thus, a knowledge of task demands and of group members’
resources indicates a group’s potential productivity. The discrepancy
between a group’s actual productivity and its potential productivity,
therefore, may be attributed to faulty social process. Since Ringelmann’s
task is unitary, additive, and maximizing (Steiner, 1972), it permits a
linear projection of a group’s potential productivity. And, since task
performance relies solely upon the individual members’ resources, ropepull
data are not complicated by extraneous factors as are data from
other unitary additive tasks such as rowing (in which performance is
affected by shell size, crew weight, and water displacement). In additive
tasks like Ringelmann’s, the separate members’ efforts are combined
through social coordination. As group size increases, the number of
“coordination links” and thus the opportunity for faulty social process
increases; hence, the greater becomes the discrepancy between actual
and potential group performance. The Ringelmann study, then, demonstrated
dramatically how the efficiency of the member may decrease
purely with the enlargement of group size-even though the total group
output increases. Finally, Ringelmann’s task has significance for methodological
reasons: it is a situation that strongly involves subjects without
recourse to deception and it yields ratio measures of individual and
group output,
Despite those interesting features, the Ringelmann rope-pull paradigm
has not been used in later research on group performance. Faulty social
process, due to losses in either coordination or motivation, has been investigated
primarily in the context of intellective problem solving (Davis,
1969). Because of that, and also because Moede’s original report of
GROUP SIZE AND GROUP PERFORMANCE 373
Ringelmann’s finding was scanty-giving no more information than that
mentioned in our first paragraph-it remained unclear as to whether
the effect was replicable, how it might be affected by differences in
procedure, in subject selection, or in other sorts of group conditions.
We were led, therefore, to an interest in replicating and extending the
Ringelmann effect, with a view to its eventual usefulness for testing
more general models of group performance.
OVERVIEW OF RESEARCH
Two pilot studies to replicate Ringelmann’s work were done first, one
with four six-man groups of college students and a second with seven
six-member groups of young boys. These pilot studies revealed that
our procedure was satisfactory and they yielded some preliminary results
showing the existence of noticeable decrements after the addition
of a second and of a third co-worker; further additions resulted in only
minor performance decrements. Subsequently, the main replication
(Study I) was undertaken with 17 six-man groups of college students.
In these studies, the only experimental variation was the size of the
pulling group-varying between one person pulling alone and all six
group members pulling together. The questions to which we sought
answers were: (1) Would there be a linear decrease in the amount
pulled per person as group size increases? (2) If the results should
not parallel those of Ringelmann, what would be the best description
of the functional relationship between member output and group
size?
The output losses found in the replication research left us interested
in the reasons for the decrement in group member performance. Therefore,
Study II was designed to examine two alternative explanations
of the source of process loss: coordination versus motivation loss. Details
are reported below.
STUDY I: AN ATTEMPT TO REPLICATE THE RINGELMANN EFFECT
Moede’s (1927) report gave no information about Ringelmann’s procedure
for determining pulling performance. A number of factors may
not have been controlIed at that time, in&ding subjects’ slippage while
pulling, their angle of pull and their effort of holding up the rope. Nor
do we know about the reliability of measurement of pulling performance.
Our apparatus was designed, therefore, to minimize slippage and angular
variation, and to maximize reliability of measurement.
Apparatus
A woodenframeworkwasbuilt with six pulling stations at 3-ft intervals (Fig. I).
The total length was 27 ft (8.24 m). It wouId have been preferable to have eight
374 INGHAM ET AL.
FIG. 1. Rope-pull apparatus.
pulling stations, as in the Ringelmann study, but it was not feasible to house an
apparatus longer than the present one. The rope was placed at a height of 3 ft
3 in. (1 m), the standard height of a doorknob: it was attached by an eye-bolt to
a Baldwin-Lima-Hamilton strain gauge (BLH Load Cell U 361) at the front
of the framework.
1. Reduction of slippage. To ensure that the subjects’ feet would not slip,
footrests were placed across the floor of the framework. Each puller’s front foot
could, therefore, be planted securely on a footrest angled at 45”. To permit a firm
grip on the rope and to control where different subjects pulled, the rope was
covered with athletic tape at regular intervals.
2. Support of rope. The framework contained a manila rope, 1% in. (2.86 cm)
thick. To control the pulling angle, as well as the weight that different-sized groups
had to support, a gear-toothed crank at the rear of the framework held the rope
rather taut (Fig. 1). With the rope maintained at a constant tension, subjects did
not have to divert effort into lifting its weight, but could concentrate entirely on
their pull.
3. Reduction of measureement error. Pulling scores were recorded electronically.
Strain transmitted along the rope caused an electrical signal to be emitted by the
BLH strain gauge. This signal was transmitted via an Accudata-195 Gauge Control
Unit to the Honeywell Visicorder (Model 1508 ) recording oscillograph. Contained
within the recording apparatus was a Honeywell galvanometer (Model M 40 350 A
was used in Study I; a more sensitive model, M 24 350, was used in Study II).
GROUP SIZE AND GROUP PERFORMANCE 375
Variations in pull transmitted from the BLH strain gauge produced differential
deflections on the mirror surface of the galvanom’eter, and were recorded in millimeter
deviations upon an oscillograph. In order to transform these millimeter deviations
into an index of “poundage pulled,” the strain gauge was calibrated up to
loo0 lb of pull.
Subjects
The subjects in Study I were 102 male students from the introductory psychology
course at the University of Massachusetts, who received extra credit for their participation
in the experiment. They were scheduled to arrive at the laboratory in
groups of six. Upon his arrival each subject was given an identification tag containing
a letter from A to F.
Design
A treatment-by-subjects or repeated measures design (Winer, 1962, pp. 105-124)
was used in this study. To negate any residual effects, due to either fatigue or
practice, the pulling order was randomized prior to the testing of each group.
Since the task induced fatigue, the design tried to ensure: (1) that all subjects
pulled an equal number of times; (2) that each subject had approximately the
same rest interval between pulls; and (3) that the number of pulls per subject
was at a minimum. Subjects were assigned to the various group sizes on the basis
of a partial permutation (see Table 1). Thus, in group size 6, all subjects were
involved in the task; in sizes 5 and 4, each subject rested once while the others
pulled; in size 3, each person pulled thrice and rested thrice; in group size 2, each
subject rested twice while others pulled; in the individual condition, subjects took
turns at pulling while all others rested. Tracing any subject through the illustrated
treatment sequence reveals that the total number of pulls per subject was 1.3
and the total number of rest intervals was 12.
TO minimize the effects of pulling position, the subjects were randomly assigned
to their tug-of-war stations in each treatment. Since each subject had a letter idenTABLE
1
SUBJECT COMBINATIONS USED FOR .~LL GROUPS IN THESE STUDIES
Number of individuals in pulling unit
6 5 4 3 2 1
ABCDEF BCDEF Aa CDEF AB CDF ABE CD ABEF C ABDEF
ACDEF B ABEFCD CDE ABF EFABCD I) ABCEF
ABDEF C ABCD EF BEF ACD AB CDEF E ABCDF
ABCEF D AEF BCD F ABCDE
ABCDF E ABD CEF A BCDEF
ABCDE F ABC DEF B ACDEF
a Subjects with identification letters in italics are resting.
Note: Mean individual scores were obtained by dividing pull scores in each condition
by the number of individuals pulling. Thus, for each group of subjects, six such scores
were computed: The mean pull for sizes 1,2,3,4,5, and 6. These six scores were the data
items used for the analysis of variance.
376 INGHAM ET AL.
tification tag and each pulling station was numbered, the assignment of subjects to
stations was accomplished by the experimenter calling out a series of letter-number
combinations-for instance: “A goes to 5, B goes to 1, . . . .” In this example,
subject A would position himself at pulling station 5, B at station 1, and SO forth,
until all co-workers had been assigned.
A model first demonstrated the standard tug-of-war stance. The following instructions
were then given:
“This apparatus is designed to test your pulling power. You will be tested
alone and in groups of various sizes. This is not a competition and no results
will be given you until the required number of pulls are completed. You will
be given three commands:
( 1) ‘Take the Strain. During this phase, you will have time to secure a
grip on the rope and adjust your foot position. One foot must be placed on the
number in the center of the footrest and your other foot on the baseboard;
(2) ‘Pull! During this phase you are required to pull as hard as you can
for a period not exceeding 6 sec. Pull along the line of the rope. Do not pull
down or to the side, but attempt to remain in the required stance. (A second
demonstration of the correct and incorrect methods of pulling was given by
the model. )
(3) ‘Rest’-You may return to the standingpositionand recoverfrom your
exertions.”
At this point, the experimental commands were given to the model in the way they
would be given during the experiment.This was the final demonstration that subjects
received.
Results
Figure 2 shows the percentage of loss in mean individual pulling performance
across all conditions of group size. The assumed near 100%
efficiency in the alone condition was equivalent to 130 lb/ (59 kg).”
A marked drop in average member efficiency resulted from the addition
of one and of two co-workers: 9 and 18%, respectively. Further
additions of co-workers, however, produced only small additional decrements.
While three persons’ performance was reduced to 82%of their
average individual efficiency, a group of six pulled at an average of
78%hardly much less.
TO examine the significance of these effects, a treatment-by-subjects
analysis of variance was performed. Athough the treatment orders (size
‘The data are reported in percentage-of-performance-lost in order to facilitate
comparison between our two studies and that of Ringelmann. A second reason for
reporting percentages is that we are more confident of the within-study than the
between-study comparisons of actual poundage pulled. After Study I, we discovered
a malfunction in the voltmeter used to standardize the signal span;it had
consistently reduced the sensitivity of tbe gauge control unit by about 23%.
GROUP SIZE AND GROUP PERFORMANCE 377
; ;I
‘5
oA\kp,,/I
“, “X..,
e
....
%fm
‘“..,
*..,
cl %.‘...
.c_ **...
aI
g 40-
“....,
.‘...
?
.....
:
- Study I (N = 17 groups)
“X.,
%.
m--a Study II IN=36 indlvidualsl
“‘.....
: so- Q---~O Rlngelman”
“o
LY
G
a
60
I I , t J
1 2 3 4 5 6 7 8
Sire of pulling unlf
FIG. 2. Mean individual pull scores according to group size.
of pulling unit) had been randomized to negate the influence of practice
and fatigue, the randomization was not systematic and, hence, the
data were not amenable to a Latin square analysis which would have
yielded further information. The treatment-by-subjects design, however,
provides a more conservative estimate of variability due to experimental
conditions than does the Latin square in that the variance due to order
effects is lumped together with the variance due to error. A summary
of the analysis is presented in Table 2; it indicates that at least one
treatment mean differed significantly from the others (F = 26.36, p <
.Ol). To find out which of the six means actually differed significantly,
a Duncan’s Multiple Range Comparison was performed. This analysis
TABLE 2
SUMMARY TABLE OF ANALYSIS OF VARIANCE AND TREND ANALYSIS: STUDY I
Source df MS F
Subjects 16 1113.242
Treatments 5 1389.528 26.36**
Linear 1 5800.725 110.06**
Quadratic 1 1038.149 19.70**
Cubic 1 0.174
Remainder 2 108.591
Error 80 52 705
** p < .Ol.
378 INGHAM ET At.
revealed that the means for treatment sizes of 1 and 2 differed significantly
from each other, and from every other treatment mean (p < .Ol) :
The mean scores of pulling unit sizes 3 through 6, however, did not
significantly differ from each other.
The results of the multiple-range test were confirmed by a trend analysis
in that both significant linear ( F1 in = 110.66; p < .Ol) and quadratic
(Fpuad = 19.70; p < .Ol ) trends were obtained. Thus, the Ftingelmann
effect was only partially replicated. The present findings confirmed the
results of our two pilot studies.
STUDY II: MOTIVATION AS A SOURCE OF PROCESS LOSS
Neither the pilot nor main replication research could reveal the source
of decrement in group members’ performance. On an additive task, the
addition of new members to a group increases its resources and, therefore,
its potential performance. But such additions also increase the
possibility of process loss due to either lack of coordination and/or
declining motivation (Steiner, 1972). Thus, we asked whether, as group
size increased, the output loss was mainly due to increasing problems
of synchronizing the members’ efforts or whether such losses resulted
from reductions in personal exertion.
To separate these two distinct sources of process loss, the following
study was designed: Only the individual’s perception of group size was
varied, while keeping constant the fact that he was really pulling alone.
Thus, motivation losses due to perceived size remained free to vary,
while coordination losses due to actual group size were eliminated. Individual
subjects were to be recruited for the rope-pulling experiment
as before, but the remaining “subjects” were to be experimenter
confederates.
Subjects
Forty-one male students were recruited from the introductory psychology course
at the University of Massachusetts. Five subjects did not adequately follow the
instruction, “pull without jerking the rope”; data from only 36 subjects, therefore,
were retained for analysis.
The five confederates who served as pieudosubjects were trained merely to give
the impression of pulling SO that the naive subject would be measured while he
perceived himself to be functioning in groups of various sizes. To create the image
that they were also “pulling,” the pseudosubjects were trained to produce the
. .
appropnate kmesthetrc feedback necessary to convince the naive subject. The naive
subject was always assigned to pulling station 1, and the pseudosubject who stood
immediately behind him created a sway in the rope. Pretests had shown that a
sway of only I or 2 in. was enough to create the impression that others were also
GROUP SIZE AND GROUP PERFORMANCE 379
pulling, without affecting the strain on the rope measured by the load cell. Coupled
with occasional sounds of physical exertion, this sway convinced all naive subjects
of the reality of their co-workers’ “efforts.”
A treatment-by-subjects design was again used. As in Study I, the treatment
orders were randomized prior to each experimental session to negate any systematic
residual effects of one treatment upon another.
To maintain the credibility of the pseudosubjects and also of the instructions,
the subjects (naive and pseudo) were assigned to groups according to the partial
permutation used in Study I (Table I). To further disguise the facade, all subjects
(naive and pseudo) were asked to put on blindfolds when they were assigned
to a pulling station; the pretext was that we were testing effects of “visual feedback”
on group performance. Whenever the naive subject was assigned to pull,
the pseudosubjects refrained from pulling as previously discussed. However, when
the naive subject had a rest period and was allowed to remove his blindfold, the
pseudosubjects taking their turn on the rope exhibited maximal exertions.
Procedure
The apparatus and procedure were similar to that of the earlier work. Since
more sensitivity was needed to measure the smaher, pureIy individual, performance,
a more sensitive galvanometer (Honeywell Model M 24 350) was substituted for
the one used previously.
The use of the blindfolds required explanation. Thus, the following orientation
statement was included in the subject’s initial instructions:
“In this experiment, you will be asked to pull on a rope as individuals and as
members of groups varying in size. You are instructed to pull as hard as you possibly
can without jerking on the rope. Your eyes will be covered by a blindfold
before you are given the command to ‘take the strain,’ and they must remain
covered until the end of each pull. The reason we are using blindfolds is to compare
the effect of ‘no visual feedback’ upon the pulling strength of individuals with
data compiled earlier when ‘visual feedback’ was allowed. You will be given three
commands . . .”
After the completion of the experiment, subjects were debriefed as they had
been in the earlier studies. It is notable that in no single imtance had any naive
subject become suspicious of the proceedings. During the Fostexperimental explanation
session, subjects often laughed with great surprise upon hearing how they were
led to believe that other group members were pulling behind them.
After each subject’s particiuation, both he and the pseudosubjects left the laboratory
through an exit door. The pseudosubjects filtered back individually to take
part in the next subject’s pulls. If the next naive subject had already arrived, pseudopartners
would come in with naive questions such as “Is this the room for the
psychology experiment?” Occasionally, a confederate would arrive late and apo’ogize
for delaying the experiment.
Results
Figure 2 shows the mean pull scores for individuals who pulled in
what they believed were groups of varying size. The assumed 100%
efficient performance in the alone condition was equivalent to a mean
380 INGHAM ET AL.
TABLE 3
SUMMARY TABLE OF ANALYSIS OF VARIANCE AND TREND ANALYSIS: STUDY II
Source df MS F
Subjects 35 5140.364
Treatments 5 2574.390 11.51**
Linear 1 6848.826 30.61**
Quadratic 1 4096.889 18.31**
Cubic 1 282.668 1.26
Remainder 2 821.784
Error 175 223.742
**p < .Ol.
puII of 131 pounds (59.47 kg). 0 rice again, these data indicate a clear
drop in performance, corresponding to the addition of one or two perceived
co-workers. However, after the addition of the second co-worker,
there was no further decrease in the group’s pulling scores.While three
persons pulled at 85%of their average individual efficiency, a group of
six still pulled at an average of 86%(Fig. 2).
A Treatment X Subjects design was again employed in the analysis
of the data. The data points for this analysis were the mean individual
performance scores for each “group” size. A summary of the analysis
of variance is presented in Table 3; it shows that at least one cell
treatment mean differed significantly from the others (F = 11.51;
p < .Ol).
To find out which of the six treatment means actually differed significantly,
Duncan’s Multiple Range Comparison was performed; it revealed
that the mean for the “size 1” treatment differed significantly
from all other means. Furthermore, the “size 2” mean differed significantly
from those for
TI Tz Td Ts Ta TS
group sizes1, 3, and 5 (but not from those of sizes4 and 6). The results
of the multiple range test were confirmed by a trend analysis in that
both significant linear (Flin = 30.61; p < .Ol) and quadratic trends
(Fwaa = 18.31; p < .Ol) were again found. Both the multiple range
and trend anaIyses support the interpretation that subjects lost some
motivation as soon as they perceived that even one “co-worker” was
added.
DISCUSSION
The findings of both studies indicate that individual performance on
the rope-pull task declined substantially when subjects either were or
GROUP SIZE AND GROUP PERFORMANCE 381
believed that they were members of a group of co-workers. To that
extent, these results offer a partial confirmation of the Ringelmann effect4
In these studies (and also in our two pilot studies) the average
individual’s output declined E-188 from his performance in the alone
condition to his performance in the three-man group. Beyond group
size 3, however, the addition of further co-workers failed to produce
continued performance decrements as appeared to occur in Ringelmann’s
initial research. (Only in the second pilot study, where our
subjects were young boys, was there an almost linear decrement with
the addition of co-workers beyond group size 3.)
The present departure from the continued linear decrement observed
by Ringelmann may simply be due to the tighter controls and improved
apparatus design of the present studies. For one thing, the experimental
controls placed on the subjects-such as when to pull, and how long to
spend pulling-may have reduced incoordination from lack of leadership.
Furthermore, incoordination problems due to differing angles of
pull, differential weight of rope per puller, and slippage of feet or hands
were minimized in our apparatus. As a result, purely mechanical sources
of process loss were probably less than those of Ringelmann. Our conjectures
about differences are tentative, however, because of the inadequate
report of Ringelmann’s experimental design in Moede’s (1927)
article. Certainly, the results of Study I (Ringelmann Replication) do
not allow us to specify any group size where the further addition of
co-workers will fail to increase total output. Such a point of diminishing
returns was suggested by Riihlmann (1900) who, in observations of the
additive task of rowing, noted that the eight is the largest boat in rowing
since larger boats are inefficient. Ruhlmann cited the example of a
canoe with 24 paddlers which travelled around Lake Wannsee in Berlin
no faster than another with a crew of eight. A point of diminishing
returns is also suggested by an imaginary extension of Ringelmann’s
7% decrement to a hypothetical group size 15, where the addition of a
sixteenth member would presumably add nothing more to the group’s
output (Zajonc, 1966, p. 102-103)) or where 50 co-workers would supposedly
reach zero effectiveness ( Dashiell, 1935, p. 1113).
‘It is worth noting that the effect is not necessarily confined to the performance
of humans. Thus, Koehler (1927) reviews data from two different studies where
performance decrements were found among horses, when pulling teams were increased
in size. And J. H. Sudd found that groups of ants pulled their prey with
measurably less average force than did individual ants: “For individual ants the
weight of prey was about 85 milligrams . . . two ants operated at about 94 per
cent efficiency, and three and four at 77 per cent efficiency” (cited in Zajonc,
1972, p. 16).
382 INGHAM ET AL.
Although the results of Study I established an inverse relationship
between group size and individual performance, they gave no clue about
the possible sources of performance loss. It remained unclear whether
group members pulled less hard because of incoordination or because
of losses in motivation. Study II, then, aimed to separate these determinants
of process loss; the possibility of coordination loss was eliminated
so that motivation effects could be isolated. Although minor procedural
differences negate a strict comparison of the two studies, Study
II again demonstrated a curvilinear relationship between group size and
individual performance. So that while Steiner (1972) attributed Ringelmann’s
results mainly to increasing difficulty in intermember coordination,
Study II shows that declining motivation itself has a deleterious
effect on productivity.
What influences would account for the motivation loss apparent in
Study II? One possible explanation can be derived from the absence of
feedback. There was neither kinesthetic feedback associated with overcoming
an object’s inertia-since the rope was stationary on an immobile
framework-nor comparative feedback-since the experimenters did not
inform subjects of the pull recorded by the load cell on successive trials.
It is possible, then, that the absence of cognitive feedback reduced performance
levels from that which could have been achieved under feedback
conditions (see Zajonc & Brickman, 1969).
A second explanation for motivation decrements would be that subjects
felt less responsibility when pulling with others than when pulling
alone. The task permitted the individual to “hide in the crowd” as the
group size increased (Davis, 1969, p. 72). Either unintentionally or
intentionally, subjects may have reduced their effort when they felt less
personally responsible for the performance (see also Latane & Darley,
1970). Should the subjects have been cognizant of the anonymity provided
by the addition of co-workers, they might have engaged in “impression
management” (Goffman, 1959). That is, when it was possible
for the experimenter to estimate individual scores, a maximal effort
appeared necessary. Under group conditions, however, an individual’s
separate pull was unascertainable, and subjects may have given convincing
impressions of straining hard while actually pulling with less
than their maximum strength.
The present results, then, tend to confirm the generality of the Ringelmann
phenomenon-increases in group size are inversely related to
individual performance. Our results also illustrate Steiner’s (1966, 1972)
assertion that both coordination and motivation are determinants of
process km in group performance. And, they have interesting implications
for other research on group performance. For example, Zajonc and
GROUP SIZE AND GROUP PERFORMANCE 383
Brickman ( 1969) in their analysis of expectancy (an emphasis on cognitive
process) and aspiration (implied motivational forces), note that “it
has not been illustrated that setting a goal has in and of itself motivational
consequences.” In our initial study, subjects had no knowledge
of results and were thus denied the possibility of goal setting behavior.
Thus, the discrepancy between potential and actual group performance
may have been due purely to incoordination. However, if Ringelmann’s
subjects knew the results of their efforts, then such knowledge of results
may have contributed to decreased motivation. That is, since increases
in group size reduced actual performance from what was potentially
expected, Ringelmann’s subjects may have shifted their expectations to
bring them closer to their feedback-namely, downward. Such reasoning
might account for the difference between the results of our study
and those obtained by Ringelmann. Thus, the question remains begging:
If the knowledge of results were manipulated to induce subjects to set
higher goals, would the discrepancy between actual and potential performance
be reduced? That is, if the experimenter gave feedback of
favorable pull scores, would subjects exert themselves to greater performance
levels? The implication for future research is to ask “what
kinds of expectancies are optimal for performance in the long run as
well as the short run, with differing degrees of risk and failure. . . .”
(Zajonc & Brickman, 1969, p. 155).
Our research suggests a second avenue for inquiry. Since the groups
used in our experiments were quasi-groups, incoordination problems
would probably be greater than in established groups who are used to
working together. One might ask whether the discrepancy between
actual and potential productivity would be reduced if our subjects had
been team members on the same rowing crews. Inspection of the times
of recent Olympiads ( 1952-1964) reveals that coxed fours were, on the
average, only 13% faster than coxed pairs, and eights were only 23%
faster than pairs. In the 1972 Olympiad, the winning time for double
sculls was only 4% faster than for single-man boats, while the fastest
coxed eight rowed only 6% faster than the fastest coxed four. However,
as previously noted, such times are not directly comparable because
of variations in shell size, crew weight and water displacement. What
would occur, though, if such crews after having had many practice
sessions for coordinating their efforts, were placed on the rope-pulling
apparatus where extraneous factors such as those identsed above were
negated? What are the effects of practice, learning, and incentive on
the discrepancy between actual and potential performance? Furthermore,
regarding the results of our second study, it may be asked whether
members of established teams-compared to members of quasi-groups-
384 INGHAM ET AL.
would be significantly less likely to reduce their personal efforts under
conditions of responsibility diffusion. These and other questions await
future research.
REFERENCES
DASHIELL, J. F. Experimental studies of the influence of social situations on the
behavior of individual human adults. In C. Mm&son (Ed.), A handbook
of social psychology. Worcester, MA: Clark University Press, 1935, Pp. 1097-
1158.
DAVIS, J. H. Group performance. Reading, MA: Addison-Wesley, 1969.
GOFFMAN, E. The presentation of self in everyday life. Garden City, NY: Doubleday
& Co., 1959, Pp. 1-16.
K~HLER, 0. Uber den Gruppenwirkungsgrad der men&lichen K&perabeit und die
Bedingung optimaler Kollektivkraftreakton. Industm’elle Psychotechnik, 1927,
4, 209-226.
LATAN& B., & DARLEY, J. M. The unresponsive bystander: Why doesn’t he help?
New York: Appleton-Century-Crofts, 1970.
MOEDE, W. Die Richtlinien der Leistungs-Psychologie. IndustrieUe Psychotechnik,
1927, 4 193-207.
R~~HLMANN. Allgemeine Maschinenlehre, Berlin, 1900. Cited in Kohler, 0. tfber den
Gruppenwirkungsgrad der menschlichen Korperarbeit und die Bedingung
optimaler Kollektivkraftreaktion. Inclustrielle Psychotechnik, 1927, 4, 209-226.
STE~NNER, I. D. Models for inferring relationships between group size and potential
group productivity. Behavioral Science, 1966, 11, 273-233.
STEINER, I. D. Group process and productivity. New York: Academic Press, 1972.
WZNER, B. J. Statistical principles in experimental design. New York: McGraw-Hill,
1962.
ZAJONC, R. B. Social psychology: An experimental approach. Belmont, CA: Brooks/
Cole, 1966.
ZAJONC, R. B. Animal social behavior. Morristown, NJ: General Learning Press,
1972.
ZAJONC, R. B., & BRICICMAN, P. Expectancy and feedback as independent factors
in task performance. Jownul of Personality and Social Psychology, 1969, 11,
148-156.
(Received August 3, 1972)