Professional Documents
Culture Documents
Since the mid-1980s, the main effects of dispositional and contextual factors
have been examined on a variety of affective, cognitive, and skill-based learn-
ing outcomes in training contexts (Colquitt, LePine, & Noe, 2000; Salas &
Cannon-Bowers, 2001). Most of this empirical research emphasized either the
main effects of dispositional or contextual factors. Researchers have strongly
argued that future research should study the interactive effects of dispositional
and situational factors on training outcomes (Murtha, Kanfer, & Ackerman,
1996).
Several researchers have turned their attention to the relationships between
dispositional goal orientation and different training outcomes (Button,
Mathieu, & Zajac, 1996; Ford, Smith, Weissbein, Gully, & Salas, 1998). This
article highlights the current state of this literature. Goal orientation is a broad
⫹ ⫹ ⫹
⫹
⫹ ⫹
People who are directed to think of ability as a fixed entity focus on eval-
uative concerns about personal capability, which they believe demonstrate
intellectual capacity (Elliott & Dweck, 1988). Mistakes represent personal and
social evaluative threats, which imply an absence of personal control (Bandura,
1991). When threatened, people who subscribe to an entity view may adopt
avoidant strategies for coping (Thompson, 1981). Subscribers to the entity
view are likely to experience low self-efficacy beliefs because they tend to dwell
on their deficiencies and perceive environmental demands as more pressing
than they are (Wood & Bandura, 1989). Based on this rationale, we make the
following hypothesis:
Method
Sample. The sample contained ninety-six clerical workers at a Midwest-
ern university. The participants were 84 percent female, and the average age
was 41.9 years (SD ⫽ 10.1). Over 95 percent had never used this software
prior to the training. One hundred three participants enrolled in and attended
the classes. Ninety-six completed the training, representing a response rate of
93.2 percent.
Training and Procedure. A free four-hour introductory course about
Microsoft Access 97 software was offered to clerical workers. The personnel
office at the university provided a complete mailing list of clerical employees.
The advertisement, mailed to a random sample of those employees, offered
participants a choice of several four-hour sessions, from which they were asked
to select one based on scheduling convenience. The advertisement also indi-
cated that the training was being offered as part of a research project and out-
lined the main topics to be covered during training: creating a database,
creating a report, using the query feature, and converting into Microsoft Access
97 from other database software.
Classes were held in a computer classroom with twenty computer work-
stations. We conducted ten identical four-hour sessions to accommodate
employees’ work schedules. The two of us taught the classes, following the same
written script and procedures for administering surveys and tests. We each
taught the same number of participants. Multivariate analysis of variance did
not reveal a significant multivariate effect of instructor or session for the demo-
graphic variables or baseline variables discussed below. T tests revealed that
none of the outcome variables in this research differed by instructor.
Participants completed several scales and training activities in the same
sequence. Prior to the start of the training, participants signed informed-consent
422 Martocchio, Hertenstein
forms and completed several scales: cognitive ability, dispositional goal orien-
tation (both learning and performance), and pretraining self-efficacy. Then they
were provided two and one-half hours of lectured instruction during which
they practiced the concepts and facts. After the instruction, midtraining self-
efficacy was assessed followed by a test of declarative knowledge and a manip-
ulation check. Fourth, trainees completed hands-on exercises at their own pace.
Finally, posttraining self-efficacy was assessed, and a second manipulation check
was administered.
Goal Orientation Inductions. The inductions were embedded in the topic
outlines that the instructor asked the participants to read carefully over the
course of the training and in the instructions for the exercises. The inductions
were adapted from past research (Martocchio, 1994; Wood & Bandura, 1994)
and were designed to induce performance- or learning-oriented responses.
The participants were given folders at the beginning of the session that
contained class materials, including the test instruments, outlines of each
topic, and exercises. Participants received a packet with either learning or
performance orientation inductions on a random basis in each class. Neither
instructor knew which induction the participants received. Equal numbers
of both inductions were prepared for each session by an individual not
involved in the research project, and the instructors did not view the con-
tent of the packets. Each instructor followed a detailed script for delivering
training that excluded learning- or performance-oriented response patterns.
Furthermore, each instructor followed the same procedure for answering
participants’ questions by answering the question exclusively based on fact
without remarks that could induce a learning- or performance-oriented
response pattern.
The following is a performance goal induction: “As you go through the
Microsoft Access 97 training, you will benefit most by using features of the pro-
gram. You will have more than one opportunity to perform 5 hands-on exer-
cises. Focus on performing these exercises well by minimizing the mistakes
you make. The fewer mistakes you make, the better you can use the Microsoft
Access 97 program. So, try to keep mistakes to a minimum.”
The following is an example of a learning goal induction: “Learning how to
use Microsoft Access 97 is developed through persistence and hard work. The
old saying ‘Practice makes perfect’ holds true for computer skills. You will prob-
ably make mistakes. That’s normal. People who learn how to use Microsoft
Access 97 do not begin with faultless performance. Again, it is important to
remember that the more practice you have, the more capable you will become.”
Measures
General Cognitive Ability. Cognitive ability was assessed with the re-
vised form of the Wonderlic Personnel Test, Form 1 (Wonderlic & Associates,
1992). The Wonderlic is a twelve-minute timed test with fifty items arranged
Learning Orientation and Goal Orientation Context 423
Data Analysis
We conducted multivariate analysis of covariance as a preliminary analysis
prior to testing model fit and the statistical significance of paths because of
multiple related dependent measures of baseline and posttraining outcomes
(Cohen & Cohen, 1983). That is, we wanted to be sure that the relationship
between the set of independent variables (learning orientation, pretraining self-
efficacy) and the set of dependent variables (posttraining self-efficacy, learn-
ing) was not the result of a chance occurrence. A statistically significant
MANCOVA analysis indicates that there is a multivariate effect of the treatment
and covariate variables on the outcome variables.
Path analysis using maximum likelihood estimation (Analysis of
Moment Structures, Arbuckle, 1994; Hox, 1995) tested the hypothesized
model. Path analysis allowed us to test the effects of the independent vari-
ables on the dependent variables in one analysis rather than testing parts of
it in separate analysis. This analysis allowed us to make stronger conclu-
sions about the operation of the theory. Specifically, we examined the over-
all fit of the hypothesized model in Figure 1 to the data as well as the
parameters for the paths.
424 Martocchio, Hertenstein
The input data were a correlation matrix, which we corrected for atten-
uation using scale reliability scores (Ghiselli et al., 1981); that is, we
removed the effects of any measurement error. The hypothesized interaction
term between learning orientation and goal orientation context was also cor-
rected for attenuation with a formula developed by Busemeyer and Jones
(1983).
Consistent with an hierarchical approach to path modeling (Cohen &
Cohen, 1983), we first tested the hypothesized model containing main effects
only (that is, we excluded the interaction between learning orientation and
goal orientation context). Then we tested the model with the interaction term
(we subsequently included the interaction between learning orientation and
goal orientation context). This term was added to the path model as an exoge-
nous variable (that is, as a variable that serves as an independent variable only),
without correlation to either of the main effect variables. Following these tests,
we examined theoretically plausible alternative models to determine whether
these fit the data better than the hypothesized model.
Seven fit indices guided our judgment regarding whether the hypothesized
model was the best theoretical fit to the data: 2兾degrees-of-freedom (df) ratio,
goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), non-
normed fit index (NNFI), comparative fit index (CFI), root-mean-square error
of approximation (RMSEA), and likelihood ratio (LR) test of ␦2 (Bollen,
1995). 2兾df ratios less than two-to-one are suggestive of adequate model fit
(Bentler & Chou, 1987). GFI, CFI, and AGFI indicate better fit as they
approach 1, and a score of .9 or better is generally considered to be indicative
of good fit. The NNFI, also known as the Tucker-Lewis index (TLI), indicates
better fits as it approaches 1; unlike other indices, it is not limited to the range
of 0 to 1 (Bentler & Bonett, 1980). The RMSEA measures the average fitted
residual, and a score less than .1, and preferably less than .08, is desired
(Bollen, 1995).
There is some concern that smaller samples (particularly, n ⫽ 200) may
lead to biased results in maximum likelihood estimation (Boomsma, 1987).
Bentler and Chou (1987) have suggested that minimum sample size depends
on the model complexity. Specifically, the recommended ratio of sample to esti-
mated parameter is 5-to-1 or higher for reliable estimates. The ratio in this
study approaches this recommendation closely at 4.6-to-1. To provide greater
confidence in the results based on maximum likelihood estimation, we com-
pare those results with the results based on the ordinary least squares (OLS)
estimation method, which is less dependent on sample size.
Results
Zero-order correlations for the observed study variables are presented in Table 1.
The table also displays the means and standard deviations overall as well as by
condition.
Table 1. Zero-Order Correlations and Descriptive Statistics for the Study Variables
Variable M SD 1 2 3 4 5 6 7
Pretraining measures
1. Cognitive ability 23.81 5.54 1.00
2. Pretraining self-efficacy 66.96 8.69 .15 1.00
3. Learning orientation 47.27 4.51 ⫺.03 .43 1.00
Midtraining measures
4. Goal orientation context .53 .50 .22 .09 .17 1.00
(0 ⫽ performance; 1 ⫽ learning)
5. Midtraining self-efficacy 68.97 10.02 .16 .73 .37 .05 1.00
6. Declarative knowledge 6.71 1.74 .32 .18 ⫺.01 .15 .25 1.00
Posttraining measures
7. Posttraining self-efficacy 67.55 11.48 .21 .69 .30 .00 .88 .19 1.00
Context
Performance M 22.18 66.13 45.95 — 68.51 6.56 67.38
SD 5.01 9.43 4.16 — 10.93 1.79 12.06
Learning M 25.29 68.00 47.93 — 69.41 6.93 67.83
SD 5.97 8.44 4.73 — 10.19 1.68 11.53
Note: Total N ⫽ 96, Performance N ⫽ 45, Learning N ⫽ 51; .20 ⬍ r ⬍ .24, p ⬍ .05; r ⬎ .24, p ⬍ .01 (two-tailed).
426 Martocchio, Hertenstein
.47*
Cognitive Declarative
Ability Knowledge
.32* .15*
.20* ⫺.19*
.52* ⫺.04
*p ⬍ .01.
self-efficacy are positive and statistically significant. Together, these facts and
rationale are supportive of these hypothesized effects.
Hypothesis 7 predicted an interactive effect of learning orientation and
goal orientation context on posttraining self-efficacy. Specifically, we expected
significantly higher posttraining self-efficacy ratings where learning orientation
and goal orientation context are most similar (within the learning context, self-
efficacy will be significantly greater for people rating high in learning orienta-
tion than low in learning orientation) than in situations where learning
orientation and goal orientation context are most dissimilar (within the
performance context, self-efficacy ratings will not be significantly different
based on learning orientation).
The path coefficient for the interaction term was statistically significant. A
view of the interaction plot in Figure 3 shows strong support for this hypothe-
sis. In the learning context, the difference between posttraining self-efficacy for
high and low learning orientation (based on the median split) was statistically
significant and in the predicted direction (high learning orientation, M ⫽ 76.60,
SD ⫽ 8.04; low learning orientation, M ⫽ 69.89, SD ⫽ 12.94, t ⫽ 1.72,
p ⬍ .05). In the performance context, the difference between posttraining self-
efficacy was not statistically significant (high learning orientation, M ⫽ 64.47,
SD ⫽ 11.52; low learning orientation, M ⫽ 65.92, SD ⫽ 9.67, t ⫽ ⫺.51, ns).
For high learning-oriented trainees, posttraining self-efficacy was higher in the
learning context than in the performance (learning context, M ⫽ 76.60, SD ⫽
8.04; performance context, M ⫽ 64.47, SD ⫽ 11.52, t ⫽ ⫺3.67, p < .01). For
low learning-oriented people, there was no significant difference in self-efficacy
72
70
68
66
64
62
High Learning Orientation
60 Low Learning Orientation
58
Performance Learning
Note: Units represent the sum of the responses to each of the self-efficacy questions. Larger numbers
correspond with higher levels of self-efficacy.
430 Martocchio, Hertenstein
Discussion
This research extends our knowledge of the role of learning orientation and
goal orientation context in training settings by examining the interactive effect
of these two variables on affective and cognitive learning outcomes. While past
research studied main effects of dispositional goal orientation or the corre-
sponding context (Ford et al., 1998; Wood & Bandura, 1989), this investiga-
tion examined the moderating role of goal orientation context in the
relationship between learning orientation and posttraining self-efficacy.
The results endorsed past research findings that dispositional learning orien-
tation is positively related to learning outcomes, but it did not replicate previ-
ous research findings that goal orientation context affects learning outcomes.
Given our emphasis on the moderating role of goal orientation context in this
study, the lack of support for a goal orientation context main effect does not
represent a problem. Interpreting statistically significant main effects of
the constituent variables is relatively meaningless in the face of a statistically
significant interaction effect (Cohen & Cohen, 1983). The statistically signifi-
cant and predicted effect of the interaction informs our knowledge of the mod-
erating role of goal orientation context.
Support for the interactive effect is consistent with Murtha et al.’s
situational-dispositional hypothesis (1996) that situations may interact with
dispositional factors to modify the strength of behavioral or cognitive responses
to dispositional influences. Murtha et al. (1996) argued that the strength of the
dispositional–outcome relationship will be stronger when the disposition and
situation match than when the disposition and situation match less well. Our
results showed that the relationship between learning orientation and post-
training self-efficacy was statistically significant in the learning context but not
in the performance context.
Learning Orientation and Goal Orientation Context 431
References
Ackerman, P. L. (1987). Individual differences in skill learning: An integration of psychometric
and information processing perspectives. Psychological Bulletin, 102, 3–27.
Anderson, J. R. (1985). Cognitive psychology and its implications. New York: Freeman.
Learning Orientation and Goal Orientation Context 433