Professional Documents
Culture Documents
Effectiveness
Author(s): John E. Mathieu, Scott I. Tannenbaum and Eduardo Salas
Source: The Academy of Management Journal , Oct., 1992, Vol. 35, No. 4 (Oct., 1992),
pp. 828-847
Published by: Academy of Management
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to
The Academy of Management Journal
JOHN E. MATHIEU
Pennsylvania State University
SCOTT I. TANNENBAUM
State University of New York at Albany
EDUARDO SALAS
Naval Training Systems Center
The views expressed herein are ours and do not reflect the official positions of the
izations we are affiliated with. We thank Irwin Goldstein, Janis Cannon-Bowers, and t
ymous reviewers for their helpful comments on an earlier version of this article. We
Ron Landis and Susan Johns for their help in the data collection and preparation phase
work.
828
interrelationships (e.g., Alliger & Janak, 1989; Tannenbaum & Yukl, 1992).
Accordingly, the purpose of the present study was threefold. First, using a
valence-instrumentality-expectancy framework, we developed a model that
links individual and situational variables to individuals' training motiva-
tion. Second, we related training motivation to several outcome measures
commonly employed to evaluate training effectiveness. In so doing, we pro-
posed a different relationship among the outcome variables than has previ-
ously been considered. Finally, we tested the proposed relationships with
data gathered from participants in a proofreading training program using
"latent structural modeling" techniques (Jorekog & S6rbom, 1989).
FIGURE 1
Hypothesized Structural Model
1. Career
1.lannin + 8. Reaction to
PlnP7 1 Training
2. Job + P7 2 9 8
Involvement ~ 7. Training + p 9 7 + P10 9
+ lv7 3 Motivation . 9. Learning + 10. Posttest
3. Assignment P9 8.7
4. Situational7 4
Constraints Training Motivation
by Reactions
+ P9 5
5. Education
+ P10 6
6. Pretest 10
hypothesized relationships
which trainees learn a progr
training outcome measures.
however, we consider individ
tivation.
Previous research has found support for the influence of several ind
vidual variables on valence-instrumentality-expectancy cognitions. For e
ample, Lawler and Suttle (1973) obtained significant correlations betwe
individual role perceptions and ability measures and various valence-
instrumentality-expectancy components and composites. James, Hartman,
Stebbins, and Jones (1977) found support for the influence of several dimen-
sions of psychological climate on instrumentality and valence ratings and, to
a lesser degree, expectancy ratings. In a simulated organizational study,
Jorgenson, Dunnette, and Pritchard (1973) found a significant correlation
between manipulated effort-outcome probabilities and subjects' instrumen-
tality ratings. Pritchard, DeLeo, and Von Bergen (1976) manipulated behav-
ior-reward contingencies in a series of field experiments. Their findings
illustrated significant effects for the manipulations on instrumentality and
valence ratings, but not on expectancy ratings. In short, many investigations
have obtained significant correlations between individual and situational
variables and valence-instrumentality-expectancy cognitions in a variety of
settings.
simply because they are not prepared to learn. On the other hand, a negative
reaction to a program may turn off even motivated trainees, reducing their
attention, lowering their receptivity to new ideas, and inhibiting learning.
Thus, we anticipated the following ordinal interaction (P9 8.7): training
motivation will exhibit a steeper positive slope with learning for individuals
who react positively to a program than for those who react less positively
(Pedhazur, 1982). In agreement with Alliger and Janak (1989), we did not
anticipate a significant linear relationship between trainees' reactions to a
program and their learning scores (P9 8), although we did hypothesize that
learning would relate positively to performance improvement (P10 9).
METHODS
Participants
Employees from a large state university in the Northeast part
training sessions offered by the institution's human resource d
each of 140 participants attending one of seven 8-hour series. The
worked primarily in clerical and administrative assistant pos
mailed surveys containing pretraining measures to all individuals
for the program one week before they were scheduled to beg
were asked to complete the surveys and to bring them to their firs
At the first session, trainees received a general overview of the p
were administered a pretest behavior measure. They then started
ing. At the completion of the program, participants completed
survey, a learning measure, and a posttest behavior measure, in th
Unfortunately, 20 participants failed to complete the pretraining s
another 14 individuals had to be dropped because they missed o
sessions as a result of scheduling difficulties. Thus, complete
available for 106 people (76 percent of the original group). Sup
analyses of the data that were available for people who were dr
cated that they did not differ significantly from those for wh
complete data. Women composed 97 percent of the study group
average age was approximately 27 years.
Training Program
The training program was designed to enhance employees' proo
skills in terms of both quality and quantity. The program was con
a series of four 2-hour modules presented one per week. In eac
particular topic was presented on videotape; examples were pag
techniques, grammar, and punctuation rules. Structured pract
followed. Participants then received feedback concerning how
mastered each topic or skill.
Pretraining Survey Measures
The survey mailed prior to the first training session assess
pants' perceptions of their work environments, training-related m
1 Some controversy exists concerning the use of categorical variables in LISREL analyses.
Although research is ongoing on this matter, it appears that the standard errors for dichotomous
variables may be inflated in LISREL analyses, which reduces the likelihood of finding signifi-
cant results (Bollen, 1989). Therefore, our results concerning the influence of assignment are
likely to be conservative estimates. Further, we reconstituted our covariance matrix using
PRELIS, a preprocessor for LISREL also developed by Joreskog and Sorbom (1988). Among other
features, PRELIS can calculate polyserial correlations that estimate the "true" underlying rela-
tionships between differently scaled variables such as dichotomous, ordinal, and interval vari-
ables. In brief, we reestimated all our models using the revised covariance matrix produced
through PRELIS, but the results did not differ in any respect from those reported.
Reactions to Training
The second path in the model demanded by our study design related to
the performance scores. To account for initial differences in trainees' skills,
we controlled for their pretest behavior scores when examining the influ-
ence of other factors on their posttest behavior scores (Figure 1, P10 6). The
hypothesized model also includes both mediated and moderated hypothe-
sized relationships. Although structural models are generally limited to lin-
ear relationships, Baron and Kenny (1986) and Cohen and Cohen (1983)
have discussed how moderated relationships may also be incorporated. We
tested the model using LISREL VII (Joreskog & Sorbom, 1989). The relatively
small study group precluded the simultaneous estimation of both the mea-
surement and structural models. However, we used the scale reliabilities
noted above to fix the relationships between the observed scale scores and
their corresponding latent constructs, as well as the error variances for each
variable. This approach is also consistent with a method Busemeyer and
Jones (1983) advocated for testing moderated relationships in latent struc-
tural models, which is described further below. Joreskog and Sorbom (1989:
171-187) outlined how the reliabilities of observed scores can be used to
estimate latent variables in a structural model tested using a covariance
matrix. Specifically, the path from a latent variable to its corresponding
observed variable (lambda) is equal to the square root of the reliability of the
observed score. In addition, the associated amount of random error variance
(theta) is equal to one minus the reliability of the observed score times the
variance of the observed score.
Trainees' education levels and assignment methods were each asse
with single items, so reliability scores are not available. Nevertheless,
reasonable to assume that single-item measures are less than perfectly re
able. We employed a strategy recommended by Anderson and Gerb
(1988), who suggested that a conservative approach is to assume a reliabili
of .90 for single-item measures. Finally, we standardized both training m
tivation and reactions so that we could apply a formula advanced by Bohr
stedt and Marwell (1978) to calculate the reliability of the interaction ter
Their formula takes into account the reliabilities of both variables used to
form a product term and the correlation between the two latent variables; f
training motivation and reactions, it yielded a value of .77.
Four indexes were used to evaluate the fit of the structural models: the
chi-square test, the goodness-of-fit index (GFI), the root-mean-squared resid
ual (RMSR), and a Tucker-Lewis (1973) incremental fit index (T-L). Chi-
square is sensitive to sample size and in the present case has relatively lo
power to detect a misspecified model. Therefore, the statistic should be
interpreted cautiously. The goodness-of-fit index indicates the relative
amounts of the covariances among the latent variables that are collectively
accounted for by a hypothesized model. The root-mean-squared residual
a measure of the average of the fitted residuals; when a covariance metric i
analyzed, it must be interpreted relative to the magnitude of the variables'
variances and covariances.
The Tucker-Lewis index compares the fit and degrees of freedom
RESULTS
5. Job
involvement -.209 .046 .135 .061 .751 .171 .088 -
6. Situational
constraints .027 .027 -.183 -.212 -.214 .850d -.170
7. Training
motivation .005 .082 .095 .170 .108 .196 .882d
8. Reaction to
training .181 -.084 .088 .338 - .129 .014 .505 .
9. Learning .272 .276 .022 -.009 -.300 .293 .140 -
10. Posttest .578 .033 .006 .171 -.172 .080 .060
11. Training
motivation
by reactions .216 .105 -.080 .278 -.170 .112 -.140
Means 18.54 2.31 4.81 1.81 2.68 3.01 0.00 0
Standard deviations 6.07 1.05 1.12 0.39 1.10 0.78 1.00
TABLE 2
Summary of Fit Index Resultsa
Model Fit Indexes
Root-Mean-
Tucker- Goodness- Squared
Modelsb df X2 Lewis of-Fit Residual
1. Hypothesized,
with interaction 31 62.33* .513 .913 0.379
2. Hypothesized,
without interaction 32 66.17* .486 .909 0.435
3. Final revised,
with interaction 31 32.02 .984 .950 0.440
4. Final revised,
without interaction 32 35.98 .940 .945 0.511
5. Null structural 41 126.17* na .830 3.074
aN = 106.
paths to the model if they (1) were consistent with the theoretical foundation
advanced earlier and (2) yielded significant results. We emphasize that th
revision phase was exploratory and simply yields one of several potenti
models that could account for the relations among these data (cf. Stelzl,
1986).
The revised model, presented in Figure 3, contains an additional posi
tive path from assignment method to reactions to training and a direct path
from reactions to training to posttest performance. In addition, after the fou
hypothesized antecedents of trainees' motivation had been dropped, situa
tional constraints exhibited a marginally significant (p < .10) negative path
that we reintroduced, given its hypothesized role in the model and the
relatively low power of these analyses (James, Mulaik, & Brett, 1982). A
expected, the revised model fit the data quite well, both compared to the null
structural model [X2 difference (10) = 94.15, p < .01] and in general [X2(31)
= 32.02, n.s; T-L = .984; GFI = .950; RMSR = .440]. Further, the revised
model that included the interaction term represented a significantly better fit
[X2 difference (1) = 3.96, p < .05] than the revised model without the in-
teraction term.
DISCUSSION
FIGURE 2
Results of the Hypothesized Modela
.934
-.144
.236*
Situational
Constraints Tra aining Motivation
by Reactions
-.7 - ---- -- - ---
Education .231
.541*
Pretest
aN - 106.
* p < .05
FIGURE 3
Results of the Revised Modela
702
.960
.838 .627
.273*
Assignment ' Reaction to
Training
j .462/ \.188*
.244*/
Training Motivation
by Reactions
Education .228*
Pretest .517*
aN = 106.
t p < .10
* p < .05
reading does not represent the most enriched portion of the participan
jobs. Proofreading simply may not be energizing to a highly involved e
ployee, although other types of training may be. Thus, upon reflection, pe
haps individuals who were highly involved in their jobs or who had greater
career plans did not see the program as instrumental in obtaining valu
outcomes.
straints is likely to stifle the learning process itself. There has recently been
a movement toward establishing continuous learning environments in many
organizations (Rosow & Zager, 1988). The debilitating negative cycle we
described would directly interfere with attempts to implement a continuous
learning philosophy.
As for the second aim of our study, results of the revised model support
the role of reactions to training as a moderator of the relationship between
training motivation and learning. Furthermore, we found training reactions
to play a complex role linking individual and situational influences to dif-
ferent criteria measures. Specifically, although reactions to training moder-
ated the relationship between training motivation and learning, they also
mediated the influences of training motivation and assignment method on
posttest performance scores. Both high training motivation and self-
assignment led to positive reactions to the program, which in turn led to high
scores on the posttest. Thus, reactions to training simultaneously operated as
a mediator of some relationships and a moderator of others (cf. Baron &
Kenny, 1986). This is the first study to hypothesize, test, and confirm such
a complex role for training reactions in explaining training effectiveness.
These findings provide one explanation for Alliger and Janak's (1989)
observation that reaction criteria measures generally exhibit low correlations
with learning measures. Had we examined only linear relationships, we
would have reached the conclusion that reactions have no significant effects
on learning. However, our findings suggest that participants' reactions to the
program played a multifaceted role in linking individual and situational
characteristics to other training effectiveness measures. The implications of
this finding are that reactions are important for training effectiveness, but
not in and of themselves. The best results were observed when trainees were
both motivated to do well and reacted positively to the program. If trainees
were not first motivated, or if attention had been directed solely at making
training enjoyable at the expense of developing skills, less than optima
results would have been obtained.
The current results also indicate that trainees reacted more positively
the program studied here if they decided to participate rather than b
assigned to training. The implications of this result must be tempered, h
ever, by Baldwin and colleagues' (1991) findings that trainees who are ask
their preferences for training but do not receive their top preference re
lower training motivation and learning than do people who were not offe
a choice at all. Thus, practitioners should ensure that employees receive t
training they need to be successful in their jobs while minimizing potent
reactance stemming from forcing individuals into programs. Perhaps ma
employees formally part of a training needs analysis helps achieve su
delicate balance. This area is a ripe one for future work.
The current study is not without limitations. First, our study group
somewhat small, reducing the extent to which we could obtain signifi
relationships. The group size also prevented us from simultaneously testin
REFERENCES
Alliger, J. M., & Janak, E. A. 1989. Kirkpatrick's levels of training criteria: Thir
Personnel Psychology, 42: 331-341.
Anderson, J. C., & Gerbing, D. W. 1988. Structural equation modeling in practice
recommended two-step approach. Psychological Bulletin, 103: 411-423.
Baldwin, T. T., & Ford, J. K. 1988. Transfer of training: A review and direct
research. Personnel Psychology, 41: 63-105.
Baldwin, T. T., Magjuka, R. J., & Loher, B. T. 1991. The perils of participation: E
of training on trainee motivation and learning. Personnel Psychology, 44:
Baron, R. E., & Kenny, D. A. 1986. The moderator-mediator variable distinctio
chological research: Conceptual, strategic, and statistical considerations. J
sonality and Social Psychology, 51: 1173-1182.
Blau, G. J. 1985. A multiple study investigation of the dimensionality of job
Journal of Vocational Behavior, 27: 19-36.
Bohrnstedt, G. W., & Marwell, G. 1978. The reliability of products of two rando
K. F. Schuessler (Ed.), Sociological methodology: 254-273. San Francisco:
Bollen, K. A. 1989. Structural equations with latent variables. New York: Wil
Busemeyer, J. R., & Jones, L. E. 1983. Analysis of multiplicative combination
causal variables are measured with error. Psychological Bulletin, 93: 549-5
Cammann, C., Fichman, M., Jenkins, D., & Klesh, J. 1979. The Michigan Org
sessment Questionnaire. Unpublished manuscript, University of Michigan,
Campbell, J. P. 1988. Training design for performance improvement. In J
R. J. Campbell (Eds.), Productivity in organizations: 177-216. San Francisc
Campbell, J. P. 1989. An agenda for theory and research. In I. L. Goldstein (Ed
development in organizations: 469-486. San Francisco: Jossey-Bass.
Cascio, W. 1989. Using utility analysis to assess training outcomes. In I. L.
Training and development in organizations: 63-88. San Francisco: Josse
Clark, C. S. 1990. Social processes in work groups: A model of the effect o
credibility, and goal linkage on training success. Unpublished doctoral di
partment of Management, University of Tennessee, Knoxville.
Clement, R. W. 1982. Testing the hierarchy theory of training evaluation: An ex
trainee reactions. Public Personnel Management Journal, 11: 176-184.
Cohen, J., & Cohen, P. 1983. Applied multiple regression/correlation analysis
ioral sciences (2d ed.). Hillsdale, NJ: Erlbaum.
Farr, J. L., & Middlebrooks, C. 1990. Enhancing motivation to participate in pro
opment. In S. L. Willis & S. S. Dubin (Eds.), Maintaining professional co
213. San Francisco: Jossey-Bass.
Gist, M. E. 1987. Self-efficacy: Implications for organizational behavior and
management. Academy of Management Review, 12: 472-485.
Goldstein, L. L. 1986. Training in organizations: Needs assessment, developm
uation (2d ed.). Monterey, CA: Brooks/Cole.
Goldstein, L. L. 1991. Training in work organizations. In M. D. Dunnette & L.
Handbook of industrial & organizational psychology (2d ed.), vol 2: 507-62
CA: Consulting Psychologists Press.
Gould, S. 1979. Characteristics of career planners in upwardly mobile occupati
Management Journal, 22: 539-550.
Joreskog, K. G., & Sorbom, D. 1989. LISREL VII user's reference guide. Mooresville, IN: Sci-
entific Software.
Jorgenson, D. O., Dunnette, M. D., & Pritchard, R. D. 1973. Effects of the manipulation of
performance-reward contingency on behavior in a simulated work setting. Journal of Ap-
plied Psychology, 57: 271-280.
Kirkpatrick, D. L. 1976. Evaluation of training. In R. L. Craig (Ed.), Training and developmen
handbook (2d ed.). New York: McGraw-Hill.
Latham, G. P. 1988. Human resource training and development. In M. R. Rosenzweig & L. W
Porter (Eds.), Annual review of psychology, vol. 39: 545-582. Palo Alto, CA: Annual
Reviews.
Latham, G. P., Wexley, K. N., & Pursell, E. D. 1975. Training managers to minimize rating
in the observation of behavior. Journal of Applied Psychology, 60: 550-555.
Lawler, E. E. 1981. Pay and organizational development. Reading, MA: Addison-Wes
Lawler, E. E., & Suttle, J. L. 1973. Expectancy theory and job behavior. Organizational
ior and Human Performance, 9: 482-503.
Lefkowitz, J. 1972. Evaluation of a supervisory training program for police sergeants. Pe
Psychology, 25: 95-106.
Lodahl, T., & Kejner, M. 1965. The definition and measurement of job involvement. Jou
Applied Psychology, 49: 24-33.
Mathieu, J. E., & Farr, J. L. 1991. Further evidence for the discriminant validity of meas
organizational commitment, job involvement, and job satisfaction. Journal of Ap
Psychology, 76: 127-133.
Noe, R. A. 1986. Trainees' attributes and attitudes: Neglected influences on training eff
ness. Academy of Management Review, 11: 736-749.
Noe, R. A., & Schmitt, N. 1986. The influence of trainees' attitudes on training effecti
Test of a model. Personnel Psychology, 39: 497-523.
Pedhazur, E. J. 1982. Multiple regression in behavioral research. New York: Holt, Rine
Winston.
Peters, L. H., & O'Connor, E. J. 1980. Situational constraints and work outcomes: The influence
of a frequently overlooked construct. Academy of Management Review, 5: 391-398.
Peters, L. H., O'Connor, E. J., & Eulberg, J. R. 1985. Situational constraints: Sources, con
quences, and future considerations. In G. Ferris & K. Rowland (Eds.), Research in perso
nel and human resources management, vol. 3: 79-114. Greenwich, CT: JAI Press.
Phillips, J. S., & Freedman, S. M. 1984. Situational performance constraints and task characte
istics: Their relationship to motivation and satisfaction. Journal of Management, 10: 3
331.
Pritchard, R. D., DeLeo, P. J., & Von Bergen, C. W. 1976. A field experimental test of expec
Rails, R. S., & Klein, K. J. 1991. Trainee cognitive ability and motivation: Effects on
training performance. Paper presented at the 6th Annual Conference of Industr
Organizational Psychology, St. Louis.
Rosow, J. M., & Zager, R. 1988. Training-The competitive edge. San Francisco: Josse
Ryman, D. H., & Biersner, R. J. 1975. Attitudes predictive of diving success. Personnel
ogy, 28: 465-481.
Sobel, M., & Bohrnstedt, G. W. 1985. Use of null models in evaluating the fit of covariance
structure models. In N. B. Tuma (Ed.), Sociological methodology: 290-312. San Francisco:
Jossey-Bass.
Stelzl, I. 1986. Changing a causal hypothesis without changing the fit: Some rules for generating
equivalent path models. Multivariate Behavioral Research, 21: 309-331.
Tannenbaum, S. I., Mathieu, J. E., Salas, E., & Cannon-Bowers, J. A. 1991. Meeting trainees'
expectations: The influence of training fulfillment on the development of commitment,
self-efficacy, and motivation. Journal of Applied Psychology, 76: 759-769.
Tannenbaum, S. I., & Yukl, G. 1992. Training and development in work organizations. Annual
Review of Psychology, 43: 399-441.
Tucker, L. R., & Lewis, C. 1973. A reliability coefficient for maximum likelihood factor analysis.
Psychometrika, 38: 1-10.
Vroom, V. H. 1964. Work and motivation. New York: Wiley.
Williams, T. C., Thayer, P. W., & Pond, S. B. 1991. Test of a model of motivational influences
on reactions to training and learning. Paper presented at the 6th annual Conference of
Industrial and Organizational Psychology, St. Louis.