Group Dynamics: Theory, Research, and Practice
2000, Vol. 4, No. 4, 307-329
Copyright 2000 by the Educational Publishing Foundation
1O89-269WO0/S5.0O DOI: 10.1O37//1089-2699.4.4.3O7
Cognitive Stimulation and Problem Presentation
in Idea-Generating Groups
Hamit Coskun and Paul B. Paulus
Vincent Brown
University of Texas at Arlington
Clarkson University
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Jeffrey J. Sherwood
Washington State University
One technique that may facilitate group brainstorming is decomposition of the task so
that categories of the problem are considered one at a time rather than simultaneously
(A. R. Dennis, J. S. Valacich, T. Connolly, & B. E. Wynne, 1996). Two studies
examined this possibility for both solitary and interactive brainstorming in which major
categories of a brainstorming problem were presented simultaneously or sequentially.
In the 1st study, participants in the sequential presentation condition generated more
ideas than did those in the simultaneous condition in both the individual and the group
conditions. In the 2nd study, individuals exposed to either a high number or low
number of idea categories demonstrated enhanced performance. Simulations of the data
demonstrated that the results were consistent with an associative memory model of the
idea generation process.
In many contexts, it is important to generate
new ideas. These ideas can be the basis for
innovations in organizations and industry (Paulus, 2000; West, 1997). In an influential book,
Osborn (1957) suggested that brainstorming
could considerably increase the quality and
quantity of ideas produced by group members.
Osborn provided four rules for improving the
effectiveness of brainstorming groups: (a) Criticism is ruled out, (b) freewheeling is welcome,
(c) quantity is wanted, and (d) combination and
improvement are sought. Osborn (1957)
claimed that a group that adopted these rules
could generate twice as many ideas as similar
Hamit Coskun and Paul B. Paulus, Department of Psychology, University of Texas at Arlington; Vincent Brown,
Department of Psychology, Clarkson University; Jeffrey J.
Sherwood, Department of Psychology, Washington State
University.
Hamit Coskun is now at the Department of Psychology,
Abant Izzet Baysal University, Bolu, Turkey. Vincent
Brown is now at the Department of Psychology, University
of Richmond.
The research presented in this article is based partly on
thesis projects by Hamit Coskun and Jeffrey J. Sherwood.
Correspondence concerning this article should be addressed to Paul B. Paulus, Department of Psychology, University of Texas, P.O. Box 19528, Arlington, Texas 760190528. Electronic mail may be sent to paulus@uta.edu.
numbers of individuals working alone (nominal
groups). In contrast to initial expectations,
nearly all recent laboratory studies have found
that group brainstorming leads to the generation
of fewer ideas than comparable numbers of
solitary brainstormers (Diehl & Stroebe, 1987;
Mullen, Johnson, & Solas, 1991).
There have been a number of interpretations
of this productivity loss observed in interactive
brainstorming groups. The most common explanations are evaluation apprehension, free
riding, production blocking, and matching.
Evaluation apprehension results when group
members are concerned that others in the group
will be critical of their suggestions, in spite of
instructions designed to minimize such concerns (Collaros & Anderson, 1969; Diehl &
Stroebe, 1987; Harari & Graham, 1975). Free
riding occurs when individuals reduce their efforts when others in their group are performing
at high levels (Borgatta & Bales, 1953; Diehl &
Stroebe, 1987). Blocking may inhibit the generation of ideas in various ways. Obviously,
only one person at a time can talk effectively in
a group. Individuals may forget ideas while
waiting for others to state theirs or may decide
not to state ideas similar to those of others.
Diehl and Stroebe (1987, 1991) have concluded
307
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
308
COSKUN, PAULUS, BROWN, AND SHERWOOD
that blocking is the major factor responsible for
productivity loss in brainstorming groups.
Paulus and colleagues have suggested that
performance matching plays an important role
in group brainstorming performance (e.g., Paulus & Dzindolet, 1993; Paulus, Dzindolet, Poletes, & Camacho, 1993). Individual group
members may either decrease or increase their
performance to match the performance of others
(Paulus, Larey, Putnam, Leggett, & Roland,
1996). Paulus and Dzindolet (1993) have hypothesized that blocking and matching are in
fact jointly responsible for the group productivity loss (see also Camacho & Paulus, 1995).
Blocking creates an initially low level of group
performance that later becomes normative or
entrained as the group members continue to
match this low performance (Kelly & Karau,
1993). In addition, downward matching may be
more likely than upward matching, in that low
performers in the group may have more influence on the overall performance than other
group members.
Recently research has focused on ways of
overcoming the production loss in groups, such
as the use of facilitators (Offner, Kramer, &
Winter, 1996; Oxley, Dzindolet, & Paulus,
1996; Putnam, Paulus, Dugosh, & Coskun,
1999), special training techniques (Paulus,
Larey, & Dzindolet, 2000), and setting high
goals or standards (Paulus & Dzindolet, 1993).
Cognitive factors also have the potential to improve group brain storming. The studies presented here focus on task decomposition as a
means of facilitating brainstorming. Task decomposition involves breaking a problem or
task down to its component pieces (such as
subcategories) and instructing the group or individual to brainstorm on the components separately. This approach has been advocated by
Osborn (1963) and others (e.g., Armstrong,
Denniston, & Gordon, 1975; Pitz, Sachs, &
Heerboth, 1980; Samson, 1988): "It would have
been better if each of four or five sub-problems
had been brainstormed separately. Although
this would have taken more time, the results
would have been relatively superior" (Osborn,
1963, p. 174).
Dennis, Valacich, Connolly, and Wynne
(1996) have examined the impact of problem
decomposition on electronic brainstorming. In
electronic brainstorming, individuals can generate ideas on one part of the computer screen
while having access to ideas from group members on another part of the screen (Nunamaker,
Briggs, & Mittleman, 1995). Sharing ideas electronically does not lead to the production loss
often observed in conventional groups (Gallupe,
Bastianutti, & Cooper, 1991), This may be in
part because group members can work in parallel and thus do not experience production
blocking (Valacich, Dennis, & Connolly, 1994).
In their study of problem decomposition, Dennis, Valacich, et al. (1996) found that having
groups focus on three components of a problem
one at a time instead of dealing with all three
components simultaneously increased idea generation by more than 60%. In a subsequent
study Dennis, Aronson, Heninger, and Walker
(1996) examined the role of time segmentation
independent of problem structuring. Participants worked in groups of 10 in the electronic
brainstorming paradigm for 30 min. As in the
Dennis, Aronson, et al. (1996) study, participants were presented with a single question or
with three questions. The 30 min were structured either as a single time period or as three
10-min segments. Decomposing the problem
again had a strong positive effect on performance, but time segmentation alone did not.
Although problem decomposition is beneficial in electronic brainstorming, it remains to be
seen whether it facilitates productivity in conventional face-to-face group and individual
brainstorming. So far, the positive effects of
task decomposition have been demonstrated
only in electronic groups using three subcategories. We conducted two experiments to assess
the generality of the decomposition effect and
made predictions based on various interpretations of the source of the decomposition effect.
Our analysis was supplemented by computer
simulations of the model developed by Brown,
Tumeo, Larey, and Paulus (1998), which represents brainstorming in an associative memory
framework. In the first experiment, we examined task decomposition by presenting participants alone or in groups with single words or
short phrases describing categories of solutions
to the brainstorming task via audiotape. In the
sequential condition, the relevant categories
were presented one at a time (1 every 3 min). In
the simultaneous condition, all categories were
presented together at the start of the session.
Participants brainstormed by speaking into microphones. In the second experiment, we exam-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
ined the impact of visually presenting either a
small (2) or a large (10) number of categories in
a simultaneous or sequential fashion. However,
in this case, the categories were presented every 3 min in both the simultaneous and sequential conditions. All of the participants performed
individually and responded in a written format.
Why might task decomposition facilitate
group brainstorming? One possibility is that
decomposition may change the task orientation.
For example, decomposition may increase the
pace of brainstorming because it generally involves dividing the task into smaller time periods. The shorter time periods could encourage
participants to work at a faster pace than longer
time periods (Kelly & Karau, 1993). However,
this did not appear to be a factor with Dennis,
Aronson, et al.'s (1996) electronic brainstorming groups. It seems more likely that task decomposition facilitates the cognitive processing
involved in brainstorming. There are a number
of ways this could occur.
One possibility is that decomposition into
multiple components or categories may decrease the tendency of groups to converge on a
small number of dominant categories (Bouchard, Drauden, & Barsaloux, 1974; Lamm &
Trommsdorff, 1973; Larey & Paulus, 1999;
Taylor, Berry, & Block, 1958; Vroom, Grant, &
Cotton, 1969). Groups may tap only a small
number of the available possibilities (Connolly,
Routhieaux, & Schneider, 1993; Gettys, Pliske,
Manning, & Casey, 1987). Task decomposition
may provide a structure that motivates and enables groups to more fully tap ideas from a
broad range of categories (Larey & Paulus,
1999). If task decomposition helps increase the
focus on a wide range of categories, it may be
especially beneficial for group brainstorming
because interactive groups tend to converge on
a smaller number of categories than nominal
groups.
Second, task decomposition may reduce the
cognitive load resulting from the demands of
group interaction. Being in a group requires
individuals to monitor a number of situational
factors, such as turn taking and assessing the
knowledge and contributions of others. Exposure to a high number of diverse stimuli in the
form of ideas may also contribute to the cognitive processing load. The limited capacity of
short-term memory may prevent group brainstormers from optimally concentrating on idea
309
generation while processing the additional information associated with group interaction. As
a result, brainstormers may reduce the attentional resources allocated to processing the
ideas from others as well as to their own internal
idea-generation process. Task decomposition
may reduce this overload by organizing the
stimuli into meaningful subsets (Nagasundaram
& Dennis, 1993), enabling group members to
more easily focus on one piece of the problem at
a time. Decomposition into categories has
proven effective in overcoming collaborative
inhibition in group recall (B. H. Basden, Basden, Bryner, & Thomas, 1997). If sequential
presentation aids in reducing information overload in groups, groups might again show more
benefit of problem decomposition than
individuals.
Third, focusing on a single aspect of a problem may allow for greater stimulation of related
ideas. Because a major determinant of responsiveness to stimuli is the degree of similarity to
one's present cognitive state (Mednick, 1962),
exposure to other group members' ideas should
be most beneficial if the ideas share some degree of similarity with concepts or categories
that are currently being considered. From this
perspective, cognitive facilitation takes place
when one brainstormer's idea serves to activate
related ideas in the mind of his or her listeners.
These newly activated lines of thought serve, in
turn, when communicated to the group, to activate still more ideas in the minds of the group
members. The effects of this type of spreading
activation, or semantic priming, are well studied by memory researchers (e.g., Neely, 1991).*
Having groups or individuals focus on one aspect of the problem at a time is one way to
facilitate this type of relatedness. It is possible
that individual brainstormers might benefit even
more than group brainstormers, because in a
group the probability is higher that one member
might produce an idea from a different category, thus disrupting the group's focus on the
current category. On the other hand, assigning a
1
Spreading activation usually refers to the process of one
concept activating another related concept within the mind
of a person. We use the term here to refer to such cognitive
facilitation between people as well.
COSKUN, PAULUS, BROWN, AND SHERWOOD
310
Table 1
Matrix of Category Associations Used to Simulate Experiment 1 and Experiment 2
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Next category
category
1
2
3
4
5
6
7
8
9
10
11
Null
1
2
3
4
5
6
7
8
9
10
11
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
.01
.01
.01
.01
.01
.90
0
0
0
0
0
0
0
0
0
0
0
.01
.01
.01
.01
.01
.90
.01
.01
.01
.01
.01
Note. Entries in the matrix are probabilities of generating an idea from the next category given an idea in the current
category as brainstorming begins. Diagonal entries represent the probability of generating a new idea from the same
category. Off-diagonal entries represent the probability of generating an idea from a different category. The last column,
labeled the null category, represents the probability of coming up with no new idea during the current time interval. As a
brainstorming session progresses, the probabilities in the null category increase.
specific category may reduce the occurrence of
such stray ideas.2
Thus, there are a number of reasons to expect
that task decomposition will facilitate idea generation in both individuals and groups. Although there is a basis for expecting that task
decomposition should be more beneficial for
groups than individuals, the distractions associated with group interaction may conspire to
reduce the benefits of decomposition for groups.
A Model of Cognitive Processing in
Interactive Brainstorming
Many of the potential advantages of group
brainstorming and task organization being discussed here can be described using the associative memory framework from cognitive psychology. Concepts and ideas in the mind of an
individual are assumed to be organized in an
associative structure called a semantic network
(Collins & Loftus, 1975). Similar or related
concepts are assumed to be more strongly connected to each other than less similar concepts
(e.g., "tree" is more strongly connected to
"leaf" than it is to "grass" or "automobile").
Concepts can be activated through external
means, by presenting written or spoken primes,
or through internal means, when the ideas that
one is considering activate other ideas, which
then become the focus of one's attention. When
a particular concept is activated, the concepts to
which it is connected also become active to a
greater or lesser degree, depending on the
strength of the connections. Short-term memory
is often defined as the set of currently active
concepts (e.g., Cowan, 1995).
Brown et al. (1998); Paulus, Brown, and Ortega (1999); and Sherwood, Dugosh, Brown,
and Paulus (2000) have developed a computersimulation model of the cognitive factors in
group brainstorming based on the associative
memory framework. The basic representation
used by the model is a matrix of associations
between categories of ideas: Each entry in the
matrix represents the probability of generating
an idea from a particular category, given that
the brainstormer is currently generating ideas
from another (or the same) category. Table 1
illustrates the matrix used in the simulations
reported here.
The matrix format conveniently captures
many of the features of individuals' idea representation and organization relevant to group
brainstorming. The sum of the row probabilities
represents the total knowledge a brainstormer
has about a particular category: the fluency
(Mednick, 1962). The last column in the matrix
2
On the other hand, unique or seamingly unrelated ideas
may be the more creative ones.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
(called the null category) is the complement of
the fluency and represents the probability of
coming up with no idea. As a brainstorming
session progresses and ideas are used up, the
probabilities accumulate in the null category to
represent the fact that fluency effectively decreases during a session. The sum of the offdiagonal column probabilities represents the
probability of considering ideas from a particular category during the current session. This is
called the accessibility of the category. Lowaccessible categories will contain ideas that an
individual may not think of on her or his own
but may be primed to consider by other group
members. The presence of low-accessible but
high-fluency categories is one factor that should
make brainstorming with other people quite
useful. For example, an individual seeking
counseling might not be immediately aware of
all of the underlying factors that motivated him
or her to seek counseling but may be led to
consider more of them after listening to other
people, perhaps a therapist or individuals in a
group therapy session.
The matrix representation also captures differences in individual cognitive styles, with
convergent thinkers having relatively high within-category transition probabilities, and therefore being likely to continue with the current
category, and divergent thinkers being relatively more likely to move between categories.
(Thus, the matrix depicted in Table 1 represents
a relatively convergent knowledge structure.)
Also, many of the other possible similarities and
differences between individual brainstormers'
knowledge about a task or problem can be represented in the matrix model by varying the
category association strengths in each brainstormer's matrix.
Effective cognitive facilitation in groups is
assumed to depend on the two related processes
of attention and short-term memory. Clearly,
the amount of attention that group members pay
to each other as they are brainstorming will
determine the effectiveness of the group interaction. Similarly, the extent to which the ideas
that other members have presented remain active in the mind of an individual group member
will determine how much influence these ideas
have on that individual. The amount of attention
and short-term memory capacity available during a group brainstorming session varies from
individual to individual and depends on the
311
social and cognitive demands of the situation.
The more cognitive resources devoted to other
factors, the less will be available to concentrate
on the task of brainstorming itself. In our model,
attention and short-term memory are represented as weights between 0 and 1.3 Attention is
represented as the probability of an individual
group member switching from considering her
or his own category of ideas to the category
presented by the current speaker in the group.
For example, an attention weight of 0 means
that the listener never switches to the speaker's
category, that is, the listener is not paying attention at all. An attention weight of 1 means
that the listener always switches to the speaker's
category. This is probably unrealistic, and a
value of 0.5 is used in the simulations reported
here. The short-term memory weight multiplies
the most recently generated category and functions as a decay rate. A short-term memory
weight of 0 means that only the current category
is considered. A short-term memory weight of 1
means that all previously generated categories
are equally likely to be the source of the next
idea. This type of perfect memory is, of course,
unrealistic. The role of attention and short-term
memory in brainstorming is examined in detail
in the simulations of the two experiments and in
the General Discussion.
A simulated group brainstorming session begins with a category being chosen randomly
from each brainstormer's matrix, with the most
accessible categories being most likely to be
chosen and the least accessible categories being
least likely to be chosen. Once a category is
activated, an idea is assumed to be generated
from that category according the probabilities in
the row of the matrix representing that category.
These ideas represent those generated internally
by each group member. There will be time
intervals, especially later in the session, when
no internal ideas are generated by some (or even
all) of the brainstormers. Next, a speaker is
chosen from among the group members who
have generated an idea. The model has the
flexibility to incorporate a number of the factors
that determine a given individual's likelihood of
3
Fora description of the details of the model's representation of attention, see Brown et al. (1998). For the details
of the implementation of short-term memory, see Sherwood, Dugosh, Brown, and Paulus (2000).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
312
COSKUN, PAULUS, BROWN, AND SHERWOOD
stating her or his ideas during a group brainstorming session, including individual differences in sensitivity to blocking (which is perhaps a function of the personality trait of introversion-extraversion) and the tendency to
match the performance of others (Brown &
Paulus, 1996; Sherwood, 1998). In the simulations reported here, all group members were
assumed to have equal probabilities of stating
their ideas. The spoken idea becomes available
to each individual brainstormer as the source of
their next idea. A group member who is attending to the speaker uses that idea to continue
brainstorming. A group member who happens
not to attend to the speaker in a given time
interval continues generating ideas on the basis
of the current contents of his or her short-term
memory. Whether a person attends to the speaker's idea during a given time interval is randomly determined by the attention weight. Each
new idea generated by a brainstormer is added
to short-term memory, after the current contents
of short-term memory are decreased by being
multiplied by the short-term memory weight.
The updated contents of short-term memory
determine the category of the next internally
generated idea. In addition to categories being
selected from the contents of short-term memory or determined by attending to the current
speaker, categories can be externally primed in
the model by activating the desired category (or
categories) at various intervals throughout the
simulated session. This feature of the model
allows us to simulate the effects of task decomposition, which was implemented in the following experiments by presenting participants with
category labels at the appropriate times during
the brainstorming session.
Experiment 1
In the first experiment, we examined the effects of task decomposition by comparing sequential to simultaneous idea presentation in
both nominal and interactive groups. In both the
simultaneous and sequential conditions, the participants were presented with 10 categories of a
problem via audiotape. In the simultaneous condition, all 10 categories were presented at the
start of the session. In the sequential condition,
the categories were presented 1 every 3 min.
The number of nonrepetitive ideas generated
was the main dependent variable. We examined
the time course of idea generation over the
30-min brainstorming period (ten 3-min intervals) and the extent to which participants focused on a few categories of the problem or
many categories during each 3-min interval.
We predicted that nominal groups would outperform interactive groups and that sequential
problem presentation would enhance performance for both. If the benefits of sequential
problem presentation were due largely to pacing, then the benefits would be similar for nominal and interactive groups. If sequential problem presentation reduced convergent thinking,
interactive groups might show more benefit because they were more prone to convergent
thinking. Convergent thinking would be reduced if groups were led to change their focus
to each new category as it was presented: This
would be evident in the analysis of category use
over the ten 3-min time periods. If associative
memory organization factors underlay the positive effects of sequential presentation, beneficial effects might be more pronounced with
nominal groups, because interacting participants were more likely to be exposed to ideas
from others that did not fit within the category
that was being considered. Associative memory
factors would also lead participants to focus
their idea generation primarily on the presented
category. Such a pattern was not predicted from
a strict pacing perspective because this would
produce a general increase in idea generation
regardless of the particular categories used to
set the pace.
Method
Participants. In exchange for experimental
credit, 156 students enrolled in introductory
psychology classes participated in the study.
Each treatment condition involved 13 groups,
each with 3 participants. Most studies have indicated that production losses occur for groups
of size 3 or larger but not for dyads (Diehl &
Stroebe, 1987). One of the groups in the simultaneous condition was excluded from the data
analysis because its members did not follow the
instructions.
Procedure. After all participants had signed
the informed-consent form, the experimenter
gave them a second page, which included detailed instructions about both the brainstorming
procedure and the brainstorming problem. The
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
experimenter read the instructions aloud to the
participants as they followed along.
We used The University Problem, which requires students to generate suggestions for improving their university. Previous research has
shown that the responses to the University Problem consist of categories such as classes, campus, activities, and parking (Larey, 1994). Each
of these categories has subcategories. For example, the category of parking problems consists of subcategories such as tickets and towing.
In the present study, categories related to the
University Problem were selected on the basis
of the number of their subcategories, as determined from the Larey data set. Each selected
category had 15 or more subcategories. The
categories of the problem were presented to
brainstormers in decreasing order of number of
subcategories: classes, campus life, building,
student life, academic policies, activities, instruction, parking, dorms/housing, and athletic
teams.
In each presentation condition, all participants were instructed by means of a standardized audiotape. In the sequential-problempresentation condition, the tape instructed the
participants to begin generating ideas on a new
category every 3 min. In the simultaneous-problem-presentation condition, the tape instructed
the participants to generate ideas on the 10
categories of the University Problem by presenting all of the categories at the start of the
session. The participants in the simultaneouspresentation condition also were provided a
sheet that included 10 categories of the problem. The participants in both the simultaneous
and sequential conditions were instructed to
generate ideas on any aspect of me problem that
came to mind, whether or not these ideas fit the
categories.
In the nominal condition, participants were
recruited in groups of 3, but each participant
was led to a separate room that contained a desk
and two tape recorders. One of the tapes was
used for the presentation of the relevant problem and the other for recording the ideas generated by each individual. In the interactive
condition, all participants remained in the experimental room seated around a circular table
that contained one tape recorder with a microphone and a second recorder for the presentation of the relevant categories. In this way, the
participants' ideas were recorded on a common
313
tape. In both conditions, after answering questions regarding the brainstorming problem, the
experimenter started both tapes at the same time
and left the room. At the end of 30 min, the
speaker on the tapes instructed participants to
stop generating ideas. At that point, the experimenter returned and stopped the tapes.
At the end of the brainstorming session, participants were also asked to rate the number and
quality of ideas they had generated on a 9-point
scale ranging from very few or very low to very
many or very high. As a way to assess whether
the participants were attending to the tape, all
participants were asked to list the categories (10
categories of the University Problem) to which
they were exposed during the brainstorming
session.
Results
Experimental data. Each individual's and
group's audiotape was transcribed, and each
transcript was coded by one rater for unique
ideas. In the case of the solitary performers, the
rater obtained a raw group total by aggregating
the 3 members' individual totals to form a nominal group total in accord with the procedures of
past studies. Then the rater checked the transcripts for repetitive ideas and calculated a new
group total for nonrepetitive ideas. The nonrepetitive group total was used as a productivity
index for comparing the general performance
levels of nominal and interactive groups. As a
check on the first rater, a second rater repeated
the coding for 25% of the transcripts. The interrater reliability (Cronbach's alpha) for the
raw group total was .94. The interrater reliability for the nonrepetitive group total was .96. In
addition, the participants' use of categories
within each 3-min time period was coded separately by two independent raters, who assigned
each idea to 1 of the 10 categories or other
categories of the University Problem. The interrater reliability for this measure was .94.
Figure 1 shows the effects of presentation and
group type on idea-generation performance
over the ten 3-min time intervals. A 2 (presentation: sequential vs. simultaneous) X 2 (group
type: interactive vs. nominal) X 10 (time interval) analysis of variance (ANOVA), with time
interval as a within-subject factor, yielded significant main effects of presentation and group
type, F(U 44) - 29.24, p < .0001, and F(l,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
314
COSKUN, PAULUS, BROWN, AND SHERWOOD
12
15
18
21
24
30
Time Interval (minutes)
- 9 - Nominal/Sequential
-G-lnteractive/Sequential
• No m i nal/S im u Itaneous
-Interactive/Simultaneous
Figure 1. The number of nonrepetitive ideas generated over time in Experiment 1 as a
function of group type and presentation type.
44) = 89.80, p < .0001, respectively. The participants in the sequential problem presentation
condition (M = 114.11, SD = 66.33) generated
more ideas than those in the simultaneous condition (Af - 72.24, SD = 40.77), and the nominal groups (Af = 137.00, SD = 49.00) generated significantly more ideas than interactive
groups (M = 48.44, SD = 22.39). There was
also a significant Presentation X Group Type
interaction, F(l, 44) = 5.46, p < .02, with the
difference between the simultaneous- and sequential-presentation conditions being greater
for nominal brainstormers than for interactive
brains tormers.
A main effect of time interval on idea-generation performance, F(9, 396) = 15.93, p <
.0001, reflected the fact that idea-generation
performance was higher in the early intervals
than in the later intervals. There was a significant Time Interval X Problem Presentation interaction, F(9, 396) = 3.40,/> < .0005, indicating that the difference in performance due to
problem presentation was greater in the later
intervals than the earlier ones. This was because
idea generation declined more over time with
simultaneous presentation than it did with sequential presentation. A Time Interval X Group
Type interaction, F(3, 396) = 1.96, p < .04,
reflects the fact that the difference between
nominal and interactive groups was larger in the
early time intervals than in the later ones.
A similar analysis was performed for the number of categories from which ideas were generated
within each 3-min time period (see Figure 2).
Participants processed fewer categories as time
progressed, F(9, 396) = 27.86, p < .0001. There
was also a Presentation X Time Interval interaction, F(9, 396) = 8.01, p < .0001, reflecting the
facts that the participants in the simultaneouspresentation condition generated ideas from more
categories in each time interval than those in the
sequential-presentation condition and that this category usage declined more over time in the two
simultaneous conditions than it did in the two
sequential conditions. Those in the sequentialpresentation condition tended to focus on about
one category on average within the 3-min time
intervals. There was also a significant effect of
problem presentation on the recall of the presented
categories, F(l, 47) - 27.98, p < .0001. Participants in the sequential-problem-presentation
condition (Af = 6.81, SD = 1.32) recalled significantly more categories than did those in the
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
12
15
18
21
315
27
30
Time Interval (minutes)
- B - Nominal/Sequential
-^-Interactive/Sequential
- Nominal/Simultaneous
- Interactive/Simultaneous
Figure 2. Category usage per 3-minute interval as a function of group type and presentation
type in Experiment 1.
simultaneous-problem-presentation condition
(M = 4.72, SD = 1.44).
An analysis of ratings of the number of ideas
revealed a significant effect of the group type on
perceived performance, F(l, 152) = 8.17,/? <
.005. Using a 9-point scale, participants in the
interactive groups (M ^ 5.24, SD = 2.01) reported that they generated more ideas than those
in the nominal groups (M = 4.29, SD = 1.60).
There was no significant difference in rating of
quality, although there was a trend for the participants in the interactive condition to report
generating higher quality ideas than those in the
nominal condition.
Model simulations of Experiment I. The
patterns of results obtained were consistent
with an associative memory perspective. To
provide further support for this perspective,
the conditions of the experiment were simulated using a model based on associative
memory assumptions (Brown et al., 1998).
Simulations of Experiment 1 activated all of
the relevant categories in immediate succession at the beginning of the session to represent the simultaneous-presentation condition.
In the case of orally presented external
primes, there can be no true simultaneous
condition, because spoken words must be presented in sequence to be properly recognized.
In the simulations of the sequential condition,
the 10 primes were presented with much more
time between them (every seven time cycles).
The short-term memory weight was set to 0.2
for the simulations of Experiment 1. Each
point in the figures represents the average
of 5,000 tests.
Figure 3 shows that in keeping with the
data from Experiment 1 (see Figure 1), the
model simulations generated both a strong
effect of presentation type and a Group
Type X Presentation Type interaction. The
model clearly predicts that successive presentation will improve the productivity of brainstorming individuals and groups. Figure 3
shows that although in the simultaneous conditions, participants started the session at a
slightly higher level of productivity than
those in the sequential conditions, productivity declined more rapidly in the simultaneous
conditions than in the sequential conditions,
producing a long-term advantage for the se-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
316
COSKUN, PAULUS, BROWN, AND SHERWOOD
2
3
4
5
Time Interval (simulated)
•
Nominal/Sequential
—©—Interactive/Sequential
•—Nominal/Simultaneous
Nominal/No Priming
•—Interactive/Simultaneous - - - -Interactive/No Priming
Figure 3. Model predictions for number of ideas generated over time for the four conditions of
Experiment 1. The dashed lines represent the performance of hypothetical no-priming conditions.
quential conditions.4 This is because the sequential conditions continued to get a boost as
new category primes were presented throughout the session, whereas the primes were used
up at the start of the session in the simultaneous conditions. Successive presentation
helped groups to overcome two related tendencies: (a) the tendency for rate of idea
generation to decrease, often rapidly, from the
beginning to the end of a brainstorming session, and (D) groups' tendency to converge
and generate ideas from only a subset of the
possible problem categories. With nearsimultaneous presentation, only the most recently presented categories in the sequence
(or some other select few) remained strongly
active in memory. Because of short-term
memory decay, these primes were not as
readily available later in the session. With
sequential presentation, each primed category
was likely to be active in short-term memory
until it was replaced somewhat later by the
next external prime.5
It is possible that overwhelming the brainstormers with simultaneous primes right at the
start of the task could actually reduce performance to a level below what it would have been
if no primes had been presented. The hypothetical no-priming conditions generated by the
model are shown as the dashed lines in Figure 3.
4
The advantage in the simulations for the simultaneous
presentation condition during the first time bin was the result of
the continuous priming that occurred during the starting phase
of the brainstorming session. This early advantage was not as
readily apparent in the simulations of interactive groups (see
the bottom lines of Figure 3), because most of the extra ideas
were lost due to blocking: There was not time for everyone to
state their ideas even if a high number was initially generated
in response to the simultaneous priming. For simplicity's sake,
the simulations assumed that brainstormers always attended
perfectly to the primes throughout the session (the attention
weight for the primes was set to 1.0), In reality, participants
probably did not attend perfectly to the primes, and in particular, attention to a sequence of primes presented rapidly at the
start of a session was likely to be somewhat poor: Participants
simply might not have the time to hear a prime, process it,
activate relevant ideas, and switch their focus to the next
primed category. Using a lower attention weight during the
initial simultaneous primes, such as the value of 0.5 (the value
with which group members were assumed to attend each
other), reduced this spurious advantage.
5
The simulations did not take into account the fact that
participants in the simultaneous presentation conditions
were given a printed list of the categories, which they kept
throughout the session. If the list had helped participants to
any significant extent, performance in the simultaneous
conditions would have been much closer to performance in
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
2
3
317
4
5
Time Interval (simulated)
—B—Nominal/Sequential
—&—Interactive/Sequential
-•—Nominal/Simultaneous
Nominal/No Priming
-•—Interactive/Simultaneous - - - -Interactive/No Priming
Figure 4. Model predictions for number of categories used each time interval for the four
conditions of Experiment 1. The dashed lines represent the performance of hypothetical
no-priming conditions.
The long dashes represent a nominal no-priming condition; the short dashes represent an
interactive no-priming condition. Both of these
lines fall roughly between the simulated data for
the sequential and simultaneous conditions. In
fact, there is evidence from the study of memory that improper cuing can be detrimental. The
surprising result, known as the "part-list cuing"
effect, occurs when participants in a free-recall
task are given some of the items on the remembered list as cues to recall. Such cues often
actually serve to decrease recall performance.
This is believed to occur because the seemingly
friendly prompts actually disrupt the participant's own internal organization of the material
(D. R. Basden, Basden, & Galloway, 1977;
Raaijmakers & Shiffrin, 1981). B. H. Basden et
the sequential conditions. Furthermore, the fact that participants in the simultaneous conditions recalled fewer of the
presented categories than participants in the sequential conditions, despite having die categories available to diem
continuously throughout the session, strongly suggests that
they were not as attentive to the printed list as they might
have been.
al. (1997) have suggested that this may be the
primary factor responsible for inferior group
performance in experiments on collective recall.
It is not surprising then that a rapidly presented
sequence of primes could actually serve to confuse brainstormers.
Figure 4 illustrates the prediction that simultaneous priming for nominal groups will lead to
the highest level of category usage of the four
experimental conditions. Simulated nominal
groups tended to generate high category-usage
scores, owing to the fact that there was no
interaction between individuals in a nominal
group and so there was little reason that their
brainstorming will be synchronized. There was
also the tendency for simulated-simultaneouspresentation conditions to generate higher category-usage scores than simulated-sequentialpresentation conditions. This was because in the
early time intervals the large number of simultaneous primes activated a wide variety of ideas
without providing a specific focus to the task
and during later time intervals the primes were
less available in short-term memory, giving rise
to performance that was closer to that observed
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
318
COSKUN, PAULUS, BROWN, AND SHERWOOD
when no primes were presented. The hypothetical no-priming conditions are again shown as
dashed lines in Figure 4. These two factors led
the nominal simultaneous condition to always
have the highest category usage, but exactly
where the other three conditions fell with respect to each other depended on a number of
other factors. Extensive simulations demonstrated that the frequency of sequential priming,
the convergent or divergent styles of the brainstormers involved, attention, and short-term
memory all had effects on category usage. Presenting primes more frequently caused brainstormers in the sequential condition to focus
more on the primed category, whereas too much
time between the primes allowed the brainstormers to wander, and category usage per time
interval increased. When more divergent brainstormers were simulated, category usage over
time increased, reflecting the less strict category
boundaries of a divergent thinker. Thus, category usage in the simultaneous and sequential
conditions became more equal (divergent thinkers just did not stick with a sequentially primed
category as much as convergent thinkers). The
effects of attention and short-term memory on
idea generation are further examined in the
General Discussion.
Although the model was not specifically designed to simulate participants' memory performance, a somewhat informal analysis of the
model output supported the expectation that recall of the presented categories would be better
in the sequential condition than in the simultaneous condition. One factor affecting memory
performance was the amount of time the to-beremembered material was studied or rehearsed,
regardless of whether participants knew that a
memory test was to be given. A rough measure
of rehearsal of the categories in the model was
the amount of time a category had been active in
working memory, taking into account the
degree of activation (short-term memory
strength). This measure was obtained simply by
summing the short-term memory strengths of
each category over all of the time intervals in a
simulated session. The simulations showed that
the total short-term memory strength (rehearsal
time) with sequential presentation was greater
than that with simultaneous presentation for
about 73% of the categories for nominal groups
and for about 64% of the categories for inter-
active groups. This would lead to better recall of
sequentially presented categories in both cases.
Discussion
The sequential-presentation condition produced
a higher rate of idea generation than did the simultaneous-presentation condition. This effect
was obtained for both the nominal and interactive
conditions. The results for the interactive condition are consistent with those of Dennis, Aronson,
et al. (1996) and Dennis, Valacich, et al. (1996)
and indicate that the benefits of problem decomposition are not limited to the electronic ideageneration format. The finding of an effect for
individual performance suggests that the decomposition effect for idea generation is also not limited to interactive settings. In fact, nominal groups
benefited more than interactive groups from task
decomposition. Participants in the sequential condition generally focused their idea generation on
the category presented. This led to a very consistent number of ideas being generated from each of
the categories.
Priming the brainstormers by explicitly presenting them with categories of solutions to the
problem makes it possible for all relevant categories to be considered. The cognitive model
simulations suggest that presenting the primes
sequentially throughout the session helps to
compensate for the limitations of short-term
memory, which might be particularly noticeable
under highly attention-demanding conditions.
Simultaneous (i.e., rapid sequential) presentation may overwhelm the participants and prevent them from focusing their attention adequately on each of the primes presented. In fact,
the results of simulations suggest that rapid
sequential presentation may lead to performance below that occurring when no primes are
presented. The stronger effect of presentation
mode for nominal groups indicates that the
priming benefits of sequential presentation are
more easily induced when individuals are performing alone. Once again, the inhibitory factors occurring in groups seem hard to overcome.
Experiment 2
In Experiment 1, we used 10 subcategories, but
the decomposition studies by Dennis, Valacich, et
al. (1996) and Dennis, Aronson, et al. (1996) used
only 3 subcategories. The theoretical rationales
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
discussed previously did not specify any definite
boundary conditions regarding the number of categories. Increasing the number of categories, up to
a point, would be beneficial in providing cues for
the idea-generation process. Such a stimuli-enhancement view was also advocated by Nagasundaram and Dennis (1993), suggesting that
the greater the variety of stimuli presented, the
more ideas generated. With each new category, a
potentially new set of related ideas can be stimulated (Brown et al., 1998; Paulus et al., 2000).
However, if the number of categories becomes too
large, there may be too little time allotted for each
category in a typical, brief experimental session
(30 min or less). In this case, there would be little
opportunity for related idea generation and thus
little benefit of decomposition. Conversely, if the
number of categories is too small, the beneficial
effect of priming would also be less evident (see
also Nagasundaram & Dennis, 1993). A proper
test to examine such predictions requires a paradigm that includes a control or no-priming condition in addition to a low or high number of categories. The studies by Dennis and his colleagues
did not include a control or no-priming condition as a baseline to the other conditions in their
research. Furthermore, although Dennis, Aronson, et al. (1996) found that the three 10-min
segmentations did not have any effect on ideageneration performance, it is possible that
smaller time segmentations (e.g., 3-min time
segmentations) may have strong impact and
motivational or cognitive stimulation value on
performance (Kelly & Karau, 1993).
In Experiment 2, we examined the impact of
presenting either a small or a large number of
categories in a simultaneous or sequential fashion. In this experiment, there were two independent variables: the type of presentation (simultaneous or sequential) and the number of categories (2 or 10). Differences in timing or time
constraints were controlled by having 10 task
segments, each of which lasted 3 min, in all four
conditions. In contrast to Experiment 1, in Experiment 2, participants in both the simultaneous
and sequential conditions were exposed to category primes every 3 min, and the writing format was used instead of an oral one. There was
also a control condition in which the problem
was presented to the participants without information about the categories or task segmentation.
We predicted that a greater number of categories would have positive effects on idea generation
319
because the participants would have sufficient
time (3 min) available for each category. However, presenting both the simultaneous and sequential category information every 3 min might
decrease the difference between the two presentation methods because the categories would be
reactivated in short-term memory each time the
cues were re-presented. However, if the written
simultaneous presentation produced some degree
of task overload or the use of suboptimal ideageneration strategies, the sequential presentation
would still lead to superior performance.
Method
Participants. Eighty-five students enrolled
in introductory psychology classes participated
in the study in exchange for experimental credit.
Seventeen brainstormers were assigned to each
treatment condition: either a low- or high-category condition and to either a simultaneous- or
sequential-presentation condition, and 17 were
assigned to the no-primary control condition.
Procedure. When participants arrived for the
experiment, they were handed a packet that consisted of the informed-consent form, which provided general information about the nature of research on the 1 st page, detailed instructions about
both the brainstorming procedure and the problem
on the 2nd page, and the categories appropriate to
the participant's experimental condition from the
1st page to the 10th page. The same problem used
in Experiment 1 was used: Students were asked to
generate ideas on how to improve their university.
The same categories of the problem and order of
presentation used in Experiment 1 were used in
Experiment 2.
After the presentation of the brainstorming
procedure and the University Problem, participants were instructed (a) to write down their
ideas, (b) to number each idea before they wrote
it down, and (c) that they did not need to make
complete sentences when writing the ideas.
They were also instructed to turn to a new page
every 3 min and to continue generating ideas on
this page. This feature equated all conditions in
terms of number of distinct task segments.
In the low-category condition, participants
were given the two most popular categories
(classes and campus life), whereas in the highcategory condition, participants were given the
same 10 categories of the University Problem
used in Experiment 1. In the sequential-pre sen-
COSKUN, PAULUS, BROWN, AND SHERWOOD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
320
tation condition, the relevant categories of the
problem were presented one at a time on each
new page, whereas in the simultaneous-presentation condition, categories were presented at
the same time on each new page. In the sequential-low-category condition, the first 5 pages
mentioned the first category, and the second 5
pages mentioned the second category. In the
sequential-high-category condition, each page
presented a new category. On the other hand, in
the simultaneous-low-category condition, each
page mentioned each of the two categories. In
the simultaneous-high-category condition, each
page mentioned all 10 categories.
After finishing instructions and answering
questions regarding the brainstorming problem,
the experimenter instructed all participants to turn
to page 1, read the instructions, and begin generating ideas. As soon as the experimenter finished
the instructions, the experimenter started a stopwatch and allowed participants to brainstorm until 3 min had passed. When 3 min had passed, the
experimenter instructed the participants to stop
brainstorming. Then the experimenter instructed
them to turn to the next page, read the instructions,
and continue generating ideas. Again, the experimenter started a stopwatch, and allowed them to
6
9
12
15
brainstorm until 3 min passed. This procedure
remained the same until the end of the 10th problem page. All participants were instructed to generate ideas on any aspect of the problem that came
to mind, whether or not these ideas fit the categories. At the end of the brainstorming session, participants were also asked to rate the number and
quality of ideas generated on a 9-point scale ranging from very few or very low to very many or very
high.
Results
Experimental data. As in Experiment 1,
each participant's transcript was coded separately
by one rater for unique ideas. The rater checked
each participant's transcripts for repetitive ideas
and calculated a new individual total for nonrepetitive ideas. A second rater repeated coding for
25% of the transcripts. The nonrepetitive ideas
were used as a productivity index for comparing
the performance levels of individual brainstormers. The interrater reliability for tfie raw total was
.98, that for the nonrepetitive total was .96, and
that for assigning ideas to categories was .96.
Figure 5 shows the idea-generation performance for each 3-min time interval for individ-
18
21
24
27
Time Interval (minutes)
- High/Sequential
—•— High/Simultaneous —6— Low/Sequential
-Low/Simultaneous
No Priming
Figure 5. The number of nonrepetitive ideas generated over time in Experiment 2 as a
function of number of primes and presentation type.
30
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
uals in the four experimental conditions and the
no-prime control condition. The effects were
examined by means of a 2 (presentation: sequential vs. simultaneous) X 2 (the number of
categories: low [2] vs. high [10] categories) X
Time Intervals (10) ANOVA, with time interval
as a within-subject factor. This analysis indicated a main effect of number of categories on
idea-generation performance, F(l, 64) = 13.53,
p < .0005, with more ideas generated with a
high number of categories. However, there was
no main effect of presentation, F(l, 64) = 1.75,
p > .19, nor Presentation X Number of Categories interaction, F(l, 64) = .07, p > .79. The
main effect of time interval was significant, F(9,
576) = 11.40, p < .0001, reflecting decreased
idea generation over time. There was a significant three-way Time Intervals X Presentation X
Number of Categories interaction, F(9,
576) = 2.22, p < .02, reflecting the fact that the
differences in idea-generation performance became evident in the later time periods for the
high-category condition.
A planned one-factor ANOVA comparing
the number of ideas generated in the low-category, high-category, and control conditions was
significant, F(2, 36) = U.96,p< .0001. Tukey
9
12
15
321
tests showed that there was a significant difference (a = .01) between the mean of participants
in the high-category conditions (M = 48.92,
SD = 15.32) and the mean of participants in the
low-category conditions (M — 30.77,
SD = 10.96), as well as the mean of participants
in the high-category conditions and the mean of
those in the control condition (M = 25.85,
SD = 11.26). There was no significant difference between the mean of participants in the
low-category conditions (Af = 30.77,
SD = 10.96) and the mean of those in the
control condition (M = 25.85, SD = 11.26; see
Figure 5).
Figure 6 shows the number of categories processed within each 3-minute time period for the
high-category conditions. The processing of number of categories was assessed for the highcategory conditions similar to the category usage
analyses performed in Experiment 1. A 2 (presentation: sequential vs. simultaneous) X 10 (time
interval) ANOVA yielded a main effect of presentation, f(l, 32) = 85.25,;? < .0001, indicating
that participants in the simultaneous-presentation
condition did process more categories than those
in the sequential-presentation condition during
each time interval. The main effect of time inter-
18
21
27
Time Interval (minutes)
-0-High/Sequential -+-High/Simultaneous
Figure 6. Category usage per 3-min interval as a function of presentation type for the high
number of categories condition in Experiment 2.
30
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
322
COSKUN, PAULUS, BROWN, AND SHERWOOD
val on category-processing performance was significant, F(9, 288) - 12.62, p < .0001, reflecting
a reduction in number of categories used over
time. The interaction effect between time interval
and presentation also was significant, F(9T
288) = 4.68, p < .0003, indicating that the gap in
number of categories processed between the simultaneous and sequential conditions became
smaller over time. The perceptions of individual
brainstormers regarding the number of ideas and
quality of ideas they generated showed no significant effects for either problem presentation or
number of categories.
Model simulations of Experiment 2. Simulations of visually presented primes, as used in
Experiment 2, are potentially more complex
than modeling the auditory primes used in Experiment 1. Simultaneous presentation of written primes is truly all at once, and different
individual brainstormers will probably read the
items differently depending on a number of
factors, such as attentional capacity, short-term
memory capacity, knowledge of the task, and
interest. The model is flexible enough to accommodate a number of individual variations on
these parameters. We present the results of a
simulation in which each brainstormer is as-
2
sumed to read a subset of words from the longer
list and enter them simultaneously into shortterm memory. Categories were activated at
equal intervals in both the sequential- and simultaneous-presentation conditions. In the sequential conditions, categories were activated
one at a time, every seven time cycles. Tn the
high-category-sequential condition, a different
category was activated every seven cycles; in
the low-category-sequential condition, 1 category was activated five times, and then a 2nd
category was activated five times, as was done
in the experiment. In the high-category-simultaneous condition, a random number of up to 7
categories, randomly chosen from the 10 available, were activated simultaneously, also every
seven time cycles. In the low-category-simultaneous condition, 1 or 2 of the 2 available
categories were randomly activated every seven
time cycles. Each point in the figures represents
the average of 15,000 tests. Short-term memory
weight was set to 0.8 in the Experiment 2
simulations.
As can be seen in Figure 7, the simulations
predicted that this method of presentation
would cause presentation type to have a much
smaller effect on overall productivity than it did
3
4
5
Time Interval (simulated)
-High/Sequential
—•—High/Simultaneous —A—Low/Sequential
-Low/Simultaneous —*— No Priming
Figure 7. Model predictions for the number of ideas generated over time for the five
conditions of Experiment 2.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
for the procedure used in Experiment 1. Both
simultaneous and sequential priming seemed to
be more effective than the control condition in
which no primes were presented, especially
when a high number of categories were primed.
Because the primes were re-presented at the
same intervals in both presentation conditions,
both the simultaneous and sequential conditions
provided a boost throughout the session. However, the effectiveness of the simultaneouspriming condition depended on the number of
items scanned and the short-term memory
weight. Much smaller values of either parameter caused the simultaneous condition to become less effective than the sequential condition, because the brainstormer was less able to
process and retain the information provided by
the cues. The simulations nicely reproduced the
crossover of the high-sequential and highsimultaneous graphs, because idea generation
declined much less rapidly in the sequential
than in the simultaneous condition, even though
overall output for the session was nearly equal
for the two. Note that the model predicted a
"bump" in the idea-generation graph for the
sequential-low-category condition. This corre-
2
323
sponded to the change in categories presented to
the brainstormer halfway through the session.
This slight increase in idea generation did not
appear to be visible in the empirical data, although the graph did seem to flatten out at that
point.
Figure 8 shows the simulations of category
usage, which appear to nicely match the empirical data of Figure 6. The high-sequential condition maintained a much more even sampling
of categories throughout the session than did the
high-simultaneous condition. Category usage
declined over time in this case; because the
same categories were presented each interval,
some of them were eventually sampled from
more than others and thus were not as readily
available as the session progressed.
Discussion
The main rinding of Experiment 2 is that
presenting many categories of a problem enhances the generation of ideas for that problem.
This is consistent with the notion that category
labels provide explicit stimuli for a range of
related ideas and with Osborn's (1957, 1963)
3
4
Time Interval (simulated)
-^-High/Sequential -•-High/Simultaneous
Figure 8. Model predictions for number of categories used each time interval for the
high-category-simultaneous- and sequential-presentation conditions of Experiment 2.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
324
COSKUN, PAULUS, BROWN, AND SHERWOOD
suggestion that the brainstorming problem
should be very specific. The effect of number of
categories also is consistent with both the Nagasundaram and Dennis (1993) and the Paulus
et al. (2000) models of cognitive stimulation in
groups. The greater the number of explicit stimuli (categories) encountered during the brainstorming session, the greater the number of
ideas generated.
However, there was no difference in performance due to type of problem presentation (simultaneous vs. sequential). The fact that in both
of the problem-presentation conditions the task
was broken into 3-min intervals may be critical.
This feature controlled for differences in time
restrictions or pacing. Although 10-min segments
do not seem to have much impact on brainstorming performance (Dennis, Aronson, et al., 1996),
smaller time segments may have more impact
(Kelly & Karau, 1993) and may have eliminated
the potential differences due to problem structuring. In addition, because the categories were
re-presented at the start of each 3-min segment
in both presentation conditions, both sets of
participants were reminded of task-relevant categories instead of having to keep them in shortterm memory throughout the session.
General Discussion
Two experiments found that presenting
groups and individuals with categorical information during performance of a brainstorming
task increased the generation of ideas. In Experiment 1, when the category primes were presented either all at once at the start of the
session (the simultaneous condition), or one at a
time at 3-min intervals throughout the session
(the sequential condition), sequential presentation led to the best performance for both interactive and nominal groups. Nominal groups
outperformed interactive groups in both presentation conditions, and the effect of presentation
type was stronger with nominals. Thus, although categorical priming seems to be an effective aid to brainstorming for both groups and
individuals, it does not enable groups to overcome the productivity gap. Categorical priming
may be more effective for individual performers
because in group brainstorming, individuals
may be exposed to ideas by other group members that do not fall into the category being
primed. In Experiment 2, category primes were
presented to individual brainstormers either simultaneously every 3 min during the brainstorming session or one category at a time at
3-min intervals, as in Experiment 1. In this case,
there was little difference between the two
methods of presenting the category primes.
However, presenting primes to 10 categories
did lead to greater idea generation than presenting primes to just 2 categories. These results
suggest that the frequency of explicit category
priming during a session may be a more important determinant of performance than how many
primes are given at each presentation, at least
for the fairly straightforward type of idea-generation task used in the present experiments.
The brainstorming model of Brown et al.
(1998) accounted qualitatively very well for the
results of both experiments. The model suggests
that brainstormer's short-term memory capacity
plays a role in determining the effectiveness of
externally presented primes. Presenting primes
every so often throughout the session helps remind brainstormer's of the categories of ideas
available to them. Figure 9 illustrates the role of
short-term memory by showing how the performance of individual brainstormers changes as a
function of short-term memory weight for the
conditions of Experiment 1. The short-term
memory parameter has little effect on performance in the sequential condition, presumably
because the presentation of primes throughout
the session is effective at activating the relevant
categories of ideas in the (simulated) brainstorming participants. On the other hand, increasing the strength of short-term memory improves performance in the simultaneous conditions. When brainstormers are able to keep in
mind the categories that are presented to them at
the start of the session, they are less likely to
need to be reminded of those categories
throughout the rest of the session. However, the
performance gap between the sequential and the
simultaneous conditions is not eliminated until
the short-term memory weight approaches the
unrealistic value of one, representing perfect
memory.
There are at least two important factors that
could influence an individual brainstormer's
ability to keep in mind the information presented during an idea-generating session. One
factor, which was discussed earlier, is the attention demands or degree of distraction present
during the task. A second factor is the degree of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
325
15
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Short Term Memory Weight
- O - Nominal/Sequential
- O - Interactive/Sequential
- Nominal/Simultaneous
- Interactive/Simultaneous
Figure 9. Effect of shoit-tenn memory weight on performance in model simulations of
Experiment 1.
experience the participants have had with the
particular task or problem (i.e., their level of
expertise). Experts who are highly familiar with
a particular domain of knowledge or performance have usually developed a domain-specific organization of memory that allows them
to keep in mind greater amounts of information,
usually in the form of larger "chunks," than
individuals who are less experienced in that
domain (e.g., Ericsson & Charness, 1994).
Another parameter of the model that will be
affected by the attention demands of a particular
brainstorming situation is the attention weight.
In interactive settings, the more attention group
members must devote to factors such as turntaking or evaluation concerns, the less attention
they have to allocate to the ideas being presented by other group members. In the model,
the attention weight represents the degree to
which an individual is likely to use the ideas
being presented by others. Figure 10 shows how
the performance of individuals in interactive
groups in the sequential- and simultaneouspriming conditions of Experiment 1 changes as
a function of the amount of attention allocated
to other group members' ideas. As was seen
with the short-term memory weight, increased
attention improves performance in the simulta-
neous condition, but not in the sequential condition. Apparently, the simulations suggest, if
group members are all attending closely to externally presented primes, attending to each
other is unnecessary because the primes have
already focused the group on a particular category. The sequential primes will be most effective in this way, when they are presented alone
to provide a single focus and when they are
sufficiently closely spaced to prevent brainstormers from wandering too much into other
categories before the next prime is presented.
Although we have interpreted the results of
the two experiments in terms of cognitive processes, we also considered the potential influence of other mediating factors. Sequential task
presentation may increase the pace of idea generation because of the task segmentation involved in this condition. This factor could explain the fact that differences in performance
were obtained for task presentation in Experiment 1 (only the sequential task involved task
segmentation) but not in Experiment 2 (both the
sequential and simultaneous conditions involved task segmentation). However, this perspective would not predict that sequential presentation would have a more positive effect for
nominal groups than interactive groups. Dennis,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
326
COSKUN, PAULUS, BROWN, AND SHERWOOD
17
0.4
0.6
0.6
0.7
0.B
0.9
Attention Weight
- O Interactive/Sequential - • - Interactive/Simultaneous
Figure 10. Effect of attention weight on performance in model simulations of Experiment 1.
Aronsoti, et al. (1996) also did not find support for the role of a pacing factor. However,
given the evidence for the importance of pacing factors in productivity (Kelly & Karau,
1993), it may be that both pacing and cognitive factors play a role in the effects of task
segmentation.
Although our analysis of task decomposition
has focused on brainstorming tasks, our findings
have implications for any context in which
groups share information or ideas. This could
include teams engaged in knowledge work
(Paulus, 2000), collaborative learning groups
(Paulus & Paulus, 1997), and therapy groups
involved in sharing personal information (Paulus, 2000). For example, in therapy groups,
more effective accessing of feelings and relevant information may occur when the group is
led to consider specific subaspects of a problem
in sequence rather than focusing on all aspects
of the problem as a whole.
Although the model used to simulate the experiments presented here is not designed to
model all the intricacies of interactions in complex social situations like therapy groups, it can
be used to support the possibility that task decomposition may be an effective means of fa-
cilitating group therapy discussions. In standard
idea-generation situations, particularly those in
the laboratory, the goal is generally just to produce as many ideas as possible; in fact, number
of ideas produced is the standard dependent
variable in the experimental brainstorming literature. Sheer quantity of verbal output is not
necessarily the goal of group therapy. Although
it is important that group members share experiences and ideas with each other, it is likely
that those who share less or not at all may still
benefit from the experience. What an individual
gets out of a group therapy session, much like
what one gets out of a class discussion or any
other interaction where ideas and information
are exchanged, depends on what thoughts and
feelings the discussion generates in her or him,
even if these thoughts and feelings are not actually expressed at the time. The model allows
us to examine (hypothetically) the number of
ideas generated by an individual during interactive brainslorming over and above the number
of ideas he or she actually presented to the rest
of the group. Figure 11 shows the hypothetical
number of unspoken ideas generated by each
individual in the interactive conditions of Ex-
327
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
1
2
3
4
5
Time Interval (simulated)
—O— Interactive/Sequential — # - Interactive/Simultaneous
Interactlv«/No Priming
Figure 11. Hypothetical number of unspoken ideas generated over time under the interactive
conditions of Experiment 1. The dashed line represents the performance of a hypothetical
no-priming condition.
periment I. 6 Although the generation of ideas
decreases over time for all conditions, brainstormers in the sequential condition seem to
produce more unspoken ideas. The tendency to
elicit more relevant thoughts and feelings would
seem to be an advantage (although quite speculative at this point) of the sequential organization produced by problem decomposition. The
present research suggests that group therapy
sessions may be most effective in tapping the
full range of thoughts and feelings of individuals when the group is encouraged to work
through the topic in a sequential fashion. Although, to the extent that therapy sessions and
brainstorming sessions share some common
processes, the fact that nominal groups still outperform interacting groups in all of the experiments and simulations presented here suggests
that individual therapy sessions may still be the
most effective in facilitating full expression of
feelings and ideas.
We believe that experimental research and
simulations are very useful in developing and
testing theoretical models of group dynamics.
To what extent can our and other findings from
studies on laboratory group brainstorming be
generalized to real-life settings? We believe that
there is considerable potential for such application, but much additional research will be required to assess this potential, and applications
may require significant fine-tuning (Paulus,
1996). In contrast, Ford (1999) has suggested
that this type of laboratory research may have
very limited applicability. He suggests that
studies need to be done in more realistic contexts with a longer time frame. We strongly
endorse this suggestion, and in fact, a number of
such programs of research are ongoing (e.g.,
Dunbar, 1995; West, 1997). However, realism
is no guarantee of generalizability from one
context to another (Anderson & Bushman,
1997). For example, the research by Sutton and
Hargadon (1996) that is highlighted by Ford as
a model to emulate is basically an impressionistic study with no quantitative data on group
process or comparison groups. Although these
6
The model generates output for the total number of
ideas generated by each group member in addition to the
number of ideas spoken by each member. This is because
the model first generates ideas for each individual and then
determines which individual is to be the speaker for a given
time interval. Figure 11, depicting unspoken ideas, is the
difference between the two.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
328
COSKUN, PAULUS, BROWN, AND SHERWOOD
types of field studies provide interesting descriptions of real-world groups, they do not
allow for clear theoretical conclusions. Moreover, the potential generalizability of laboratory
research may be much greater than that suggested by the skeptics. Several reviews have
demonstrated that findings from artificial laboratory studies in a number of empirical domains
mesh quite nicely with those of field studies in
more realistic contexts (Anderson & Bushman,
1997; Anderson, Lindsay, & Bushman, 1999).
References
Anderson, C. A., & Bushman, B. J. (1997). External
validity of "trivial" experiments: The case of laboratory aggression. Review of General Psychology, 1, 19-41.
Anderson, C. A., Lindsay, J. J., & Bushman, B. J.
(1999). Research in the psychological laboratory:
Truth of triviality? Current Directions in Psychological Science, 8, 3-9.
Armstrong, J. S., Denniston, W. B., & Gordon, M. M.
(1975). The use of the decomposition principle in
making judgments. Organizational Behavior and
Human Performance, 14, 257-263.
Basden, B. H., Basden, D. R., Bryner, S., & Thomas,
R. L. III. (1997). A comparison of group and
individual remembering: Does group participation
disrupt retrieval? Journal of Experimental Psychology: Learning, Memory, and Cognition, 23,
1176-1189.
Basden, D. R., Basden, B. H., & Galloway, B. C.
(1977). Inhibition with part-list cuing: Some tests
of the item strength hypothesis. Journal of Experimental Psychology: Human Learning and Memory, 3, 100-108.
Borgatta, E. F., & Bales, R. F. (1953). interaction of
individuals in re-constituted groups. Sociometry, 16, 302-320.
Bouchard, T. J., Drauden, G., & Barsaloux, J. (1974).
A comparison of individual, subgroup, and total
group methods of problem solving. Journal of
Applied Psychology, 59, 226-227.
Brown, V., & Paulus, P. B. (1996). A simple model
of social factors in group brainstorming. Small
Group Research, 27, 91-114.
Brown, V., Tumeo, M., Larey, T. S., & Paulus, P. B.
(1998). Modeling cognitive interactions during
group brainstorming. Small Group Research, 29,
495-526.
Camacho, L. M , & Paulus, P. B. (1995). The role of
social anxiousness in group brainstorming. Journal of Personality and Social Psychology, 68,
1071-1080.
Collaros, P. A., & Anderson, L. R. (1969). Effect of
perceived expertness upon creativity of members
of brainstorming groups. Journal of Applied Psychology, 53, 159-163.
Collins, A. M., & Loftus, E. F. (1975). A spreadingactivation theory of semantic processing. Psychological Review, 82, 407-428.
Connolly, T., Routhieaux, R. L., & Schneider, S. K.
(1993). On the effectiveness of group brainstorming: Test of one underlying cognitive mechanism.
Small Group Research, 24, 490-503.
Cowan, N. (1995). Attention and memory: An integrated framework. New York: Oxford University
Press.
Dennis, A., Aronson, I., Heninger, B., & Walker, E.
(1996). Task and time decomposition in electronic
brainstorming. Proceedings of Hawaii International Conference on Systems Science, 3, 51-59.
Dennis, A. R., Valacich, J. S., Connolly, T., &
Wynne, B. E. (1996). Process structuring in electronic brainstorming. Information Systems Research, 7, 268-277.
Diehl, M., & Stroebe, W. (1987). Productivity loss in
brainstorming groups: Toward the solution of a
riddle. Journal of Personality and Social Psychology, 53, 497-509.
Diehl, M., & Stroebe, W. (1991). Productivity loss in
idea generating groups: Tracking down the blocking effect. Journal of Personality and Social Psychology, 61, 392-403.
Dunbar, K. (1995). How scientists really reason: Scientific reasoning in real-world laboratories. In R. J.
Sternberg & I. E. Davidson (Eds.), The nature of
insight (pp. 365-395). Cambridge, MA: MTT Press.
Ericsson, K. A., & Charness, N. (1994). Expert performance: Its structure and acquisition. American
Psychologist, 49, 725-747.
Ford, C. M. (1999). Thinking big about small groups
in the real world: Comment on Craig and Kelly
(1999). Group Dynamics: Theory, Research, and
Practice, 3, 257-262.
Gallupe, R. B., Bastianutti, L. M., & Cooper, W. H.
(1991). Unblocking brainstorms. Journal of Applied Psychology, 76, 137-142.
Gettys, C. F., Pliske, R. M., Manning, C , & Casey,
J. T. (1987). An evaluation of human act generation performance. Organizational Behavior and
Human Decision Processes, 39, 23—31.
Harari, O., & Graham, W. K. (1975). Task and task
consequences as factors in individual and group
brain storm ing. Journal of Social Psychology, 95,
61-65.
Kelly, J. R., & Karau, S. J. (1993). Entrainment of
creativity in small groups. Small Group Research, 24, 179-198.
Lamm, H., & Trommsdorff, G. (1973). Group versus
individual performance on tasks requiring ideational proficiency (brainstorming): A review. European Journal of Social Psychology, 3, 361-387.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
STIMULATION AND PRESENTATION
Larey, T. S. (1994), Convergent and divergent thinking> group composition, and creativity in brainstorming groups. Unpublished doctoral dissertation, University of Texas at Arlington.
Larey, T. S., & Paulus, P. B. (1999). Group preference and convergent tendencies in groups: A content analysis of group brainstorming performance.
Creativity Research Journal, 12, 175-184.
Mednick, S. A. (1962). The associative basis of creative process. Psychological Review, 69, 220-232.
Mullen, B., Johnson, C , & Salas, E. (1991). Productivity loss in brainstorming groups: A meta-analytic integration. Basic and Applied Social Psychology, 12, 3-24.
Nagasundaram, M., & Dennis, A. R. (1993). When a
group is a not group: The cognitive foundation of
group idea generation. Small Group Research, 24,
463-489.
Neely, J. H. (1991). Semantic priming effects in
visual word recognition: A selective review of
current findings and theories. In D. Besner &
G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264-350). Hillsdale, NJ: Erlbaum.
Nunamaker, J. R Jr., Briggs, R. O., & Mittleman, D. D.
(1995). Electronic meeting systems: Ten years of
lessons learned. In D. Coleman & R. Khanna (Eds.),
Groupware: Technology and applications (pp. 149193). Upper Saddle River, NJ: Prentice Hall.
Offner, A. K., Kramer, T. J., & Winter, J. P. (1996).
The effects of facilitation, recording, and pauses
on group brainstorming. Small Group Research, 27, 283-298.
Osborn, A. F. (1957). Applied imagination: Principles and procedures of creative problem-solving.
New York: Scribner's.
Osbom, A. F. (1963). Applied imagination: Principles and procedures of creative problem-solving
(2nd ed.). New York: Scribner's.
Oxley, N. L., Dzindolet, M. T., & Paulus, P. B.
(1996). The effects of facilitators on the performance of brainstorming groups. Journal of Social
Behavior and Personality, 11, 633-646.
Paulus, P. B. (1996). Making groups effective: Easier
written than done. Contemporary Psychology, 41,
612-613.
Paulus, P. B. (2000). Groups, teams, and creativity:
The creative potential of idea generating groups.
Applied Psychology: An International Journal, 49,
237-262. '
Paulus, P. B., Brown, V., & Ortega, A. H. (1999).
Group creativity. In R. E. Purser & A. Montuori
(Eds.), Social creativity (Vol. 2, pp. 151-176).
Cresskill, NJ: Hampton.
Paulus, P. B., & Dzindolet, M. T. (1993). Social influence processes in group brainstorming. Journal of
Personality and Social Psychology, 64, 575-586.
View publication stats
329
Paulus, P. B., Dzindolet, M. T., Poletes, G., & Camacho, L. M. (1993). Perception of performance in
group brainstorming: The illusion of group productivity. Personality and Social Psychology Bulletin, 19, 78-79.
Paulus, P. B., Larey, T. S., & Dzindolet, M. T. (2000).
Creativity in groups and teams. In M. Turner (Ed.),
Groups at work: Advances in theory and research (pp.
237-262). Mahwah, NJ: Erlbaum.
Paulus, P. B., Larey, T. S., Putman, V. L., Leggett,
K. L., & Roland, E. J. (1996). Social influence
processes in computer brainstorming. Basic and
Applied Social Psychology, 18, 3-14.
Paulus, P. B., & Paulus, L. E. (1997). Implications of
research on group brainstorming in gifted education. Roeper Review, 19, 225-229.
Pitz, G. F., Sachs, N. J., & Heerboth, J. (1980).
Procedures for eliciting choices in the analysis of
individual decision making. Organizational Behavior and Human Performance, 26, 396-408.
Putnam, V. L., Paulus, P. B., Dugosh, K. L., &
Coskun, H. (1999). Effects offacilitators and task
instructions on productivity in brainstorming
groups. Unpublished manuscript, University of
Texas at Arlington.
Raaijmakers, J. G., & Shiffrin, R. M. (1981). Search
of associative memory. Psychological Review, 88,
93-134.
Samson, D. (1988). Managerial decision analysis.
Homewood, IL: Irwin.
Sherwood, J. (1998). Modeling cognitive facilitation
in brainstorming groups. Unpublished master's
thesis, University of Texas at Arlington.
Sherwood, J., Dugosh, L,, Brown, V., & Paulus, P. (2000).
Predicting semantic priming effects in group brainstorming: A cognitive modeling approach Unpublished
manuscript, Washington State University.
Sutton, R. I., & Hargadon, A. (1996). Brainstorming
groups in context: Effectiveness in a product design firm. Administrative Science Quarterly, 41,
685-718.
Taylor, D. W., Berry, P. C , & Block, C. H. (1958).
Does group participation when using brainstorming facilitate or inhibit creative thinking? Administrative Science Quarterly, 3, 23-47.
Valacich, J. S., Dennis, A. R., & Connolly, T. (1994).
Idea generation in computer-based groups: A new
ending to an old story. Organizational Behavior
and Human Decision Processes, 57, 448-467.
Vroom, V. H., Grant, L. D., & Cotton, T. S. (1969).
The consequence of social interaction in group
problem-solving. Organizational Behavior and
Human Performance, 4, 77-95.
West, M. A. (1997). Developing creativity in organizations. Leicester, England: British Psychological Society.