Professional Documents
Culture Documents
Companies try numerous strategies to foster creativity, including restructuring work, selecting people on the basis of their attributes, and behavioral
training; however, these strategies are often unsuccessful (Barron & Harrington, 1981; Farr, 1990;
Hocevar & Bachelor, 1989). Some organizations
form autonomous work teams that are tasked with
identifying and solving ill-defined or poorly structured problems that require creative thought (Cannon-Bowers, Oser, & Flanagan, 1992; Goodman,
Ravlin, & Argote, 1986). Given the increasing use of
teams to foster creativity (Mohrman, Cohen, &Mohrman, 1995; Tesluk, Farr, & Klein, 1997), it is surprising that little is known about the social behavior that determines a group's ability to utilize
individual creative resources effectively.
The primary purpose of this study was to allow
me to look simultaneously at individual and group
creativity, their determinants, and their interrelationships. The conceptual framework that forms
the backbone of this study is afforded by Amabile's
(1983, 1996) componential theory of individual
creativity, which allows for the impact of social
influences on individual creativity. According to
the theory, a product or response is creative when
observers independently agree that it is novel and
appropriate, useful, correct, or valuable to the task
at hand, and when that task is open-ended and
appropriately carried out via discovery rather than
via a predetermined step-by-step procedure.
To date, creativity studies have generally focused
at only one level of analysis at a time (Sternberg &
Lubart, 1999; Woodman, Sawyer, & Griffin, 1993).
316
April
FIGURE 1
Multilevel Latent Variable Model of Team Performance on Tasks Requiring Creativitya
Individual
Differences
Individual
Behaviors
H5
Group-Level
Creativity
TCRP
TCRP
H6
Conscientiousness
H1) .37**
General
Group-Level
Process
(Aggregated)
H7
Agreeableness
Extraversion
Individual-Level
Creativity
04)
^4-
(.04)
TM.65***
H2: .60*** (.04) .38** (.04
CRP
H4
(.03)
Individual
Group
55** (.04)
The path from team creativity-relevant processes to creativity-relevant processes was added with a coefficient of .36*
path coefficients are standardized.
**p < .01
***p< .001
1999). Task motivation can be indicated by behaviors related to the amount and persistence of effort.
Amabile and colleagues (1994), in their study of
artists, found that intrinsically motivated people
showed greater commitment and devoted more
time to task completion. Ruscio and coauthors
(1998) found that behavior related to "involvement
in the task" was associated with intrinsic motivation. Motivated individuals showed deep levels of
involvement in problems by focusing on solving
them, minimizing distractions, and being absorbed
in work (Ruscio et al., 1998: 261).
The domain-relevant-skills component of creativity represents the ability to learn certain types
of domain-specific knowledge (Amabile, 1996). Domain-relevantskills require familiaritywith the domain in question-memory of factual knowledge,
technical proficiency, opinions about various questions in the domain, knowledge of paradigms, performance scripts for solving problems in the domain, and aesthetic criteria (Ruscio et al., 1998).
Domain-relevant skills may be indicated by measures of an individual's depth and breadth of
knowledge related to the problems to be solved by
a team.
Creativity-relevant processes "determine the
flexibility with which cognitive pathways are explored, the attention given to particular aspects of
the task, and the extent to which a particularpath-
(.04). Reported
2002
317
Taggar
Both group and individual outcomes may be affected by the intragroupprocess behaviors of group
members. These behaviors, which I refer to as
"team creativity-relevant processes," may include
(1) inspirational motivation: inspiring group members to elevate their goals (Bass & Avolio, 1994;
Brophy, 1998); (2) organization and coordination:
ized consideration: eliciting and appreciating different ideas, needs, and viewpoints (Bass &Avolio,
1994; Brophy, 1998).
318
April
2002
319
Taggar
through task motivation. In support of this assertion, Collins and Amabile (1999), in their review of
the literature,proposed that behaviors that support
a sense of competence, without undermining a person's sense of self-determination, and attune that
individual to outcome requirements, enhance the
motivational component of creativity. Likewise,
team creativity-relevant processes that encourage
and inspire other group members to link their selfconcepts to the collective interests of a group and
its mission may increase others' intrinsic motivation to work collectively (Shamir, House, &Arthur,
1993). Finally, Collins and Amabile (1999) noted
that external motivators help creative production to
the extent that they generate continued attention to
the task at hand. Consequently, to complete the
individual-level model, I added a link between the
group-level variable team creativity-relevant processes (additively defined) and the task motivation
of individual group members (Figure 1).
Hypothesis 8. Team creativity-relevant processes, as measured by specific behaviorswithin
a team context (aggregated),representa contextual variable that will be positively associated
with greaterindividual task motivation.
METHODS
Participants, Procedures, and Tasks
Of a
In week 11 of the study, the critical incident technique (CIT;Flanagan,1954) was used to gather specific examples of (in)effective behaviors displayed
by group members. CIT is a useful initial step in
developing performanceassessment tools (Latham&
320
April
2002
321
Taggar
TABLE 1
Behavioral Observation Scale Measures of the Components of Creativitya
Dimension
Creativity Component
Sample Items
Team commitment
Task motivation
Preparation
Creativity-relevantprocesses
Team creativity-relevantprocesses
Team citizenship
Performancemanagement
Effective communication
Involving others
Providing feedback
Reaction to conflict
Addresses conflict
Averts conflict
a Items
322
conducted a LISREL 8 maximum-likelihood confirmatory factor analysis (CFA; Joreskog & S6rbom,
1993) to determine whether the item groupings developed by the judges adequately fitted the data.
At the individual level, Hypotheses 1-3 and 8
were tested by a two-stage maximum-likelihood
LISREL analysis (Anderson & Gerbing, 1988). In
stage 1, the measurement model was fitted to the
data. Next, a measurement model specifying perfect correlation among all latent variables was assessed as a means to evaluate overall discriminability. Alternative nested models, which combined
theoretically independent constructs, were then
contrasted with the original model. Model structure linkages were examined in stage 2 of the analysis. The dependent variable was the average rating
of individual creativity as assessed by peers.
Some variables that have been found to affect
creativity were controlled for prior to analyses;
these were age, gender, and group size. The control
variables were modeled into the causal structure by
freeing paths to individual creativity and to team
creativity-relevant processes. Controlling for these
was expected to mitigate some pregroup differences that were created by failing to randomly assign participants to groups.
I tested Hypotheses 6 to 7 (stating that extraversion, conscientiousness, and agreeableness are positively associated with intragroup process behavior) using simultaneous regression analysis, after
first controlling for group size.
At the group level, to test Hypothesis 4, I performed
a four-step hierarchical regression. Control variables
(age, gender, and group size) were entered in step 1.
The groups' creativity scores were entered in step 2,
and collective team creativity-relevant processes, as
measured by the behavioral observation scales, were
entered in step 3. The aggregated interaction of individual creativity and team creativity-relevant processes was entered in step 4. Statistical significance at
each step was assessed by the change in the F-statistic
associated with the incremental increase in variance
in group creativity accounted for by the variable entered at that step. The dependent variable here was
average group creativity on the 13 open-ended tasks,
as assessed by the independent judge.
RESULTS
Individual-Level
Analysis
behaviors.
CFA was
Performance-relevant
used to assess the fit of the 14 behavioral observation scale dimensions to observed behaviors. This
analysis revealed adequate fit (root-mean-square error of approximation [RMSEA] = .06, goodness-of-
April
p < .001).
1 Factor
loadings and a scree plot of eigenvalues from
an oblique exploratoryfactoranalysis supported the confirmatoryresults. This analysis is available from the author upon request.
2
Means, standard deviations, and correlations among
indicator variables are available from the author upon
request, as are indicator variable loadings on latent
factors.
2002
323
Taggar
method variance is largely responsible for covariation-here, the covariation of the behavioral observation scale and creativity measures-a factor analysis should yield a single factor. LISREL8 was used
to conduct confirmatory factor analyses. A 15factor model (14 behaviors and individual creativity) was tested first. Fit indexes suggested that the
15-factor model fitted reasonably well (RMSE =
.07, GFI = .95, CFI = .96, and NFI = .94). In comparison, a 1-factor model did not fit the data well
(RMSE = .11, GFI = .52, CFI = .54, and NFI = .56);
the 15-factor model fitted the data significantly better than did the 1-factor model (ACFI = .42; AX2 =
203.85, p < .001). Taken together, these results
suggest that common method variance does not
pose a serious threat to interpreting the present
findings.
Zero-order correlations, descriptive statistics,
and reliability coefficients for latent constructs are
in Table 2. Although conscientiousness correlated
significantly with individual creativity (r = .19,
p < .001), it achieved a stronger correlation with
task motivation (r = .35, p < .001). Similarly, although general cognitive ability correlated significantly with individual creativity (r = .26, p < .001),
it achieved a stronger correlation with domainrelevant skills (r = .57, p < .001). Openness to
experience had a correlation of .17 (p < .001) with
individual creativity; however, the correlation with
creativity-relevant processes was .34 (p < .001).
Creativity-relevant processes within the team context had the strongest correlation with individual
creativity (r = .62, p < .001). Hence, standard individual differences appear to relate to specific behaviors more strongly than to the overall rated creativity of an individual.
Table 3 reveals that the proposed measurement
model adequately reproduced the correlation matrix (RMSEA = .05, GFI = .94, CFI = .92, and NFI =
.92). In contrast, the single-factor model used to
assess overall discriminability (model 2) poorly accounted for the sample data (RMSEA = .18, GFI =
.63, CFI = .60, NFI = .61, and ACFI = .32). These
results supported the multidimensionality of the
proposed model being tested. Sequential chisquare difference tests showed that constraining
equality between pairs of highly correlated constructs resulted in worse fit. Moreover, decreases in
CFI indicated a material reduction in model fit for
each of the constrained models. Therefore, the
eight-factor measurement model was retained.
Structural model (stage 2). Behavioral observation scale ratings and assessments of creativity
were based on averaging peer assessments; therefore, it was necessary to determine whether peers
agreed in their ratings. Agreement was estimated by
the average interrater agreement statistic and intraclass correlations. These estimates suggested adequate agreement between peers (behavioral: lowest
average rWG= .74 [range = .73-.85], lowest ICC [2,
1] = .31; creativity:
average
TABLE 2
Zero-Order Correlations among Latent Factors of the Individual-Level Modela
Variable
Mean
s.d.
1. Neuroticism
2. Extraversion
3. Openness to experience
4. Agreeableness
5. Conscientiousness
6. General cognitive ability
7. Task motivationb
8. Domain-relevant skillsb
9. Creativity-relevant processesb
10. Individual team creativity-relevant processes
11. Individual creativity
12. Group team creativity-relevant processes
79.10
109.40
110.60
124.30
123.10
21.75
8.17
46.13
16.11
30.33
12.81
29.87
21.20
18.40
17.30
15.80
17.60
7.60
0.70
5.66
1.90
2.70
1.96
1.85
1
(.84)
-.06
-.09
-.15**
-.22***
-.18***
-.02
-.02
-.09
-.09
-.10*
-.05
(.79)
.02
.08
-.15**
.04
-.06
.01
.14**
.27***
.13**
.04
(.72)
.06
.16***
.12*
.02
.15**
.34***
.10
.17***
.01
(.79)
-.10
-.17***
.11*
.06
.07
.26***
.08
.07
(.82)
-.03
.35***
.23***
.27***
.26***
.19***
.11*
(.89)
.07
.57***
.42***
.17***
.26***
.03
(.71)
.12*
.32***
.31***
.41***
.37***
an = 480, for all variables except group-level team creativity-relevant processes, where n = 94. Alpha coefficients of reliabilities are display
b Behavioral observation scale measure.
* p < .05
2002
325
Taggar
TABLE3
Fit Indexes for the Measurement and Structural Modelsa
2 (dfl
Test
RMSEA
GFI
CFI
NFI
2,189.07 (594)
7,694.49 (599)
.05
.18
.94
.63
.92
.60
.92
.61
2,437.06 (599)
.09
.87
.86
.85
2,402.82 (599)
.09
.86
.85
.85
2,373.13 (599)
.09
.87
.86
.85
Measurement model
1. Proposed model
2. Single-factor model
Structural model
6. Hypothesized model
7. Modified model
.07
.05
2,719.29 (754)
2,227.34 (753)
.86
.94
.75
.94
A2 b
df
ACFI
5,505.42***
34
.32
247.99**
.06
213.75**
.07
184.06**
.06
481.95***
.21
.73
.93
a RMSEA =
root-mean-square error of approximation; GFI = goodness-of-fit index; CFI = comparative fit index; NFI = normed fit index.
n = 480.
b
The top four statistics show the difference between the indicated model and the proposed model (model 1); the fifth (last in the
column) AX2 shows the difference between the hypothesized model and the modified model.
** p < .01
**p
< .001
Group-Level Analysis
Zero-order correlations and descriptive statistics
examined at the group level are in Table 4. As
indicated in Table 4, aggregated individual creativity is significantly correlated with both group creativity and team creativity-relevant processes.
Hypothesis 4 suggests that group creativity is an
interactive function of aggregated individual creativity and the amount of team creativity-relevant
processes. After controlling for age, gender, and
group size (none of which contributed significantly
to explaining variance in group creativity), I found
significant main effects for aggregated individual
DISCUSSION
TABLE4
Zero-Order Correlations and Descriptive
Statistics for Additively Aggregated Predictor
and Criterion Measures at the Group Levela
Variable
1. Group creativity
2. Aggregated individual
creativity
3. Team creativity-relevant
processes
Mean
s.d.
14.74
2.72
1.42
0.28
.56***
29.87
1.93
.41***
an = 94.
*** p < .001, two-tailed tests
.63***
creativity (AR2 = .28, p < .001) and team creativityrelevant processes (AR2 = .07, p < .01) through
hierarchical regression analysis. There was also a
statistically significant interaction between aggregated individual creativity and team creativityrelevant processes (AR2 = .05, p < .01). It is evident
that groups with creative members and high levels
of creativity-relevant behaviors yielded high group
creativity. A low incidence of team creativity-relevant processes neutralized the effect of a group
high in creativity.3 Similarly, a group low in creativity neutralized the effects of high levels of team
creativity-relevant processes.
3 The
findings reportedin this work arebased on all 13
tasks; however, a task-by-task analysis reveals that the
findings presented here do generalize over the last 10 of
the 13 tasks. That is, team creativity-relevantprocesses
did not play a significant moderating role in the first
three tasks.
326
ways to standard individual differences. A contribution of this study is to show that although it is
necessary for a group to contain members who are
creative, team creativity-relevant processes that
emerge as part of group interaction are also important. Indeed, without this latter type of behavior,
the benefits of putting together a group of highly
creative individuals are neutralized. In effective
groups, members engage in creativity-supporting
behavior, establishing the right sort of social environment for each other. These behaviors are indicated by eight behavioral observation scale measures, which yield new insights into exactly how
group members can support each other's creativity.
Very little prior research has attempted to so extensively specify behaviors in intact groups working
on open-ended tasks.
Individual Differences
Through incorporating standard individual differences into Amabile's (1983, 1996) componential
model, I hoped that this study would aid in pulling
together relevant personality research and somewhat validate the use of personality as a predictor
of behavior associated with group creativity. This
study goes beyond previous studies in that it utilized the unifying five-factor model, which reduces
terminological confusion and makes personality
testing useful in organizational contexts (Hogan,
1991). By no means do I claim that use of this
model is the most appropriate level at which to
measure individual differences; rather, it seems an
appropriate starting point. Effort should be made to
show that newly developed measures of individual
attributes improve upon the predictiveness of the
now commonly accepted five-factor-model taxonomy, and also that of general cognitive ability, in a
variety of situations.
By incorporating personality and general cognitive ability into the model, I hoped to show that
their associated behaviors are the most appropriate
criterion for validation. One may obtain a consistently modest relationship between standard individual differences and individual creativity while
at the same time obtaining strong relationships
between individual differences and their relevant
behaviors. For instance, it is more appropriate to
validate general cognitive ability against domainrelevant skills than against overall creativity ratings
because global ratings are impacted by an individual's task motivation or by the social environment.
Feist's (1999) review of personality research
in the creativity literature indicates that facets of
to experience
fantasy(especially
openness
oriented imagination) and conscientiousness (espe-
April
2002
327
Taggar
328
Academy of ManagementJournal
April
2002
329
Taggar
Hogan, R. T. 1991. Personality and personality measurement. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology
(2nd ed.), vol. 2: 873-919. Palo Alto, CA: Consulting
Psychologists Press.
330
Academy of ManagementJournal
APPENDIX
April
Exercise 1
1. Use the concepts of stereotyping and halo to explain the
contrast between the Golden Boys and the Audit Drones.
2. Are there any aspects to the organization of work at
BH&A that could lead to perceptual problems in performance appraisal?
3. Suppose that you were appointed to a newly created
position at BH&A, Manager of Diversity Assurance.
What would you do to better manage diversity at the
firm?
Exercise 2
1. Which plan (discussed in lectures and the textbook)
does the Levi payment scheme most closely resemble?
2. Is this plan likely to be effective? Why or why not?
Please be specific and substantiate your arguments in
reference to class lectures &/or textbook.
3. If you were the administrator of this payment scheme
what might you do differently to maximize employee
motivation toward the corporate objective? Be creative
and give a full answer.
AAA
Simon Taggar (staggar@morgan.ucs.mun.ca) is an assistant professor of human resources management at York
University. He received his Ph.D. from McMaster University. His research interests include team composition,
team creativity, team leadership, and collective efficacy.
M0