Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Learning and Instruction 10 (2000) 293–309

www.elsevier.com/locate/learninstruc

Does motivation affect performance via


persistence?
Regina Vollmeyer *, Falko Rheinberg
Institute of Psychology, University of Potsdam, Postfach 601553, 14415 Potsdam, Germany

Abstract

Our cognitive–motivational process model (Vollmeyer, R. & Rheinberg, F. (1998). Motiv-


ationale Einflüsse auf Erwerb und Anwendung von Wissen in einem computersimulierten Sys-
tem. [Motivational influences on the acquisition and application of knowledge in a simulated
system.] Zeitschrift für Pädagogische Psychologie, 12, 11–23.) assumes that motivational fac-
tors affect performance via mediators. Such a potential mediator is persistence. Fifty-one stu-
dents learnt a complex dynamic system. We measured their initial motivation (mastery confi-
dence, incompetence fear, interest and challenge), then a set of mediating variables during
learning. Performance measures were knowledge acquisition and knowledge application. A
path-analysis showed that initial motivation influenced persistence. However, any possible
relationship between persistence and performance was disrupted because learners with more
knowledge stopped sooner (i.e., were classified as lowly persistent). Thus highly persistent
learners did not have higher mean final performance, despite acquiring more knowledge over
trials.  2000 Elsevier Science Ltd. All rights reserved.

Keywords: Learning; Motivation; Persistence; Problem solving

1. Introduction

Some people perform better than others at acquiring knowledge about an unknown
topic, even when everyone was given the same material. Why do learners differ?
One possible reason is that the better learners start with a good strategy and know
how to find out what is necessary for answering potential questions. Let us call this
a cognitive explanation. In contrast, there could be a motivational explanation: people

* Corresponding author. Tel.: +49-331-977-2854; fax: +49-331-977-2791.


E-mail address: vollmeye@rz.uni-potsdam.de (R. Vollmeyer).

0959-4752/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 9 - 4 7 5 2 ( 9 9 ) 0 0 0 3 1 - 6
294 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

could differ in the nature of their motivation to learn about the topic. Lowly mot-
ivated people may not try hard enough to learn. However, combining these two
explanations yields a third alternative: a cognitive–motivational explanation.
We will start by discussing how motivation could affect learning. Simon (1967,
p. 29) defined motivation as a “goal terminating mechanism, permitting goals to be
processed serially”. As people simultaneously have many goals in their lives, there
has to be a mechanism that determines which goal is activated and has attention
allocated to it. According to Simon, this is the function of motivation. If the goal
to learn about a new topic is activated, the learner tries to reach it and starts learning.
Learning itself, however, is just a cognitive process.
In contrast to Simon, Bandura (1991) combines motivation and cognition into a
cognitive–motivational perspective. He defined motivation as a “multidimensional
phenomenon indexed in terms of selection of pursuits from competing alternatives,
intensity of effort, and persistence of exertion” (p. 158). For learning this means that
motivation not only affects what people learn, but also the intensity and the duration
of the learning activities.
Processes that could be considered either cognitive or motivational are described
in Boekaerts’ (1996a) six-component model of self-regulated learning. This shows
how diverse cognitive and motivation strategies can be. Boekaerts’ model can be
seen as arising from her own empirical work (Boekaerts & Otten, 1993; Seegers &
Boekaerts, 1993), which demonstrated how motivation affected the learning process
in a school context.
Our own research is also based on a cognitive–motivational model. Rheinberg
(1997, p. 14) defined motivation as something that “provides an impetus towards a
goal for all current processes”. In doing so, motivation influences the way people
learn. Vollmeyer and Rheinberg (1998) proposed a cognitive–motivational process
model. The interaction between motivation and cognition is described in more detail
by Schiefele and Rheinberg (1997). They argued that motivation can affect three
aspects of learning: (1) persistence and frequency of learning activities; (2) mode of
performed learning activities; (3) motivational and functional states of the learner
during learning. These three aspects define the framework, within which we looked
for variables that may mediate the effect initial motivation has on performance.
Vollmeyer and Rheinberg supported empirically the assumption that people with
favourable initial motivation have a better mode of learning activities and a more
favourable motivational state, and thus performed better. This result was replicated
(Vollmeyer, Rollett & Rheinberg, 1997a; Rheinberg, 1997). However, the first aspect
of learning, persistence, was not tested in these studies. This was due to an inherent
methodical problem: having participants choose how long or often they wanted to
spend on learning would have added more variance to the measures because the
identical measure would not be exactly equivalent for all learners. For example, the
third time-point may represent the end of the task for one learner, but the middle
for another. In the study reported here, we accepted the consequences of this problem
and tried to investigate in what way persistence mediates the effects initial motivation
has on performance.
The hypothesis that persistence would mediate effects of motivation on perform-
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 295

ance is supported by two fundamental assumptions. First, motivation affects persist-


ence: if sufficient learning time is available then highly motivated learners spend
more time, or more often interact, with the learning task than lowly motivated lear-
ners (see Atkinson & Raynor, 1974; Carroll, 1963). The second assumption is that
persistence and frequency affect performance: the more time people spend learning
a task, or the more often they interact with it, the better they perform (Bloom, 1976;
Carroll 1963, 1985). Further support for this assumption is that the longer students
spend studying a certain domain, the higher their academic achievement tends to be
(Fisher, 1996; Helmke & Schrader, 1996; Volet, 1997). Taking both assumptions
together, persistence should mediate the effect of motivation on performance.

2. Motivational factors and learning

Thus far, we have discussed motivation in the abstract. However, to make predic-
tions we must be more specific. When a learner approaches a learning task, the
literature suggests that several motivational factors are relevant and can be measured:

1. Learners can vary in their certainty that they will succeed in understanding the
task. This factor we will call mastery confidence (similar concepts have been
proposed: e.g., subjective probability of success (Atkinson, 1957) and self-efficacy
(Bandura, 1977);
2. Learners can differ in their anxiety about failing in the task. This factor we will
call incompetence fear (a similar concept is Atkinson’s fear of failure, however,
for him this concept is measured as a trait whereas we assess a person’s actual
state);
3. Learners may vary in the degree to which they perceive this task as a challenge
(e.g., Czikszentmihalyl, 1975). Perceiving the task as a challenge requires a per-
sonal achievement standard for this task against which to test task performance;
4. The task may or may not evoke the learner’s interest (Krapp, 1992; Schiefele,
1991).

All these motivational factors are said to affect learning; however, it has not been
clearly shown how. In accordance with our cognitive–motivational process model
(Vollmeyer & Rheinberg, 1998), we investigated the assumption that these four
motivational factors influence learning via the following three mediators: (1) the
motivational state during learning, (2) the systematicity of learners’ strategies, and
(3) the number of learning rounds the learners deliberately choose to spend improv-
ing their knowledge (persistence).

3. Biology-lab: a complex system

For this investigation we needed a learning task which was difficult enough to be
challenging and to allow the possibility of participants failing to learn the task. Also
296 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

it needed to last long enough to allow us to study learning as a process — and


not only the outcome. Such a task is the dynamic system biology-lab presented on
a computer.
This system was similar to that used in Vollmeyer, Burns and Holyoak (1996)
and Vollmeyer, Rollett and Rhienberg (1998). It is a computer-driven system that
was constructed with the shell DYNAMIS (Funke, 1991). In a cover story, the parti-
cipants were told that they were in a biology lab in which there is a tank with three
water-quality factors (oxygenation, chlorine and temperature). These quality factors
were the output variables of this system, affected by three input variables (salt, lime
and carbon). The structure of the system is illustrated in Fig. 1. One output is rela-
tively simple to manipulate because it is influenced by only one input
(lime→oxygenation). The other two outputs are more complex, because each is
influenced by two factors. One output (chlorine) is affected by two inputs, and the
other (temperature) is affected by a decay factor (marked with a circle connected to
the output) in addition to a single input variable. The decay factor was implemented
by subtracting a percentage of the output’s previous value on each trial. Decay is a
dynamic aspect of the system, because it yields state changes even if there is no
input (i.e., all inputs are set to zero). The system is therefore complex in that it
involves multiple input variables that must be manipulated to control multiple out-
put variables.
To explore the system, participants in Vollmeyer et al. (1996) were given a learn-
ing phase and an application phase. In the learning phase, participants entered num-
bers on each trial for every input variable, then observed the changes in the output
variables. The changes in the outputs should have been used to figure out which
input affected which output and how strong the impact was. After entering numbers
on six trials (which defined one round) participants were asked to indicate what they
knew about the system. In the application phase, participants had to apply their
knowledge about the system’s structure in order to reach a certain target amount for
each output variable.
Persistence in this task can be assessed in term of how many rounds participants
explored the system and how many rounds they used trying to reach the final goals.

Fig. 1. Biology-lab system.


R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 297

4. Ecological validity of the learning situation

The decision to study the effects of motivational factors on learning in a laboratory


situation has the consequence that we might buy internal validity at the cost of
external validity (Campbell, 1957). That is, by removing the influence of unmeasured
variables through having participants work individually on a computer, we might
lose some of the effects that challenge or interest usually have in a classroom setting.
Therefore, the results of this study were only valid for situations in which self-
regulated learners individually acquire competence with a computer, as Boekaerts
(1996b) remarked. However, the results of such a study are by no means irrelevant,
as self-regulated learning with computers is becoming more and more important
(Issing & Klimsa, 1997; Vosniadou, De Corte, Glaser & Mandl, 1996). For this
emerging type of learning situation we wanted to study how motivational factors
affect learning and performance.

5. Method

5.1. Participants

Fifty-one psychology students at the University of Potsdam participated in the


study and they received course credit for every hour they stayed (84 percent received
course credit for two hours; 16 percent stayed longer and received three or four
hours credits). To the participants it was clear that their performance on the biology-
lab system was not relevant for their studies.

6. Measures

6.1. Initial motivation

After reading the instructions on how to learn about the biology-lab, participants
had to indicate their current motivation. Thus, they filled out the Questionnaire on
Current Motivation (QCM, Vollmeyer & Rheinberg, 1998) to measure the four fac-
tors of challenge, mastery confidence, incompetence fear, and interest (e.g., items
see Table 1).

7. Mediating variables

7.1. Strategy systematicity

Each of the six trials during a learning round was coded for systematicity. We
used three values: (1) low systematicity, all input variables were varied simul-
taneously (e.g., salt 10, lime 20, carbon 50); (2) medium systematicity, a system-
298 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

Table 1
Example items for the motivational factors

Factor score

Mastery confidence
I think everyone could do this task 0.73
I can’t wait to start 0.67
I think I am up to the difficulty of the task 0.65
Challenge
This task is a real challenge for me 0.76
If I can do this task, I will feel proud of myself 0.75
I’m excited about how well I will perform in this task 0.70
Incompetence fear
I’m a little bit worried 0.72
I feel paralysed by the demands of the task 0.72
I’m afraid I will make a fool out of myself 0.71
Interest
After having read the instruction, the task seems to be very interesting 0.75
I like riddles and puzzles 0.75
I would work on this task even in my free time 0.75

aticity was recognisable (e.g., two variables were varied, one was set constant: salt
10, lime 10, carbon 0. For one variable there is a positive number, for two a negative:
salt 10, lime ⫺10, carbon ⫺10); (3) high systematicity, only one or no input variable
was varied (e.g., salt 10, lime 0, carbon 0). So a value of three indicated a systematic
strategy for a trial, a value of one was an unsystematic strategy. A sample of 180
trials was coded by two raters, and a high inter-rater reliability was found, ␬=0.94
(Cohen, 1960).
As in Vollmeyer et al. (1996), participants were instructed to use a highly system-
atic strategy. Consequently, the strategy systematicity measure was not normally
distributed (Round 1: Kolmogorov-Smirnov[K-S]-Z=2.01, Round 2: K-S-Z=2.09).
Furthermore, although a good strategy was explained to participants, not all of them
took advantage of it, as is shown by the standard deviations for these measures
(Round 1: m=2.64, s.d.=0.61, Round 2: m=2.61, s.d.=0.59). Over the rounds parti-
cipants changed their strategy (Cronbach’s aRound 1,2, individual last Round=0.61). There-
fore, it appeared that even when instructed on a good strategy, some participants
explored the system differently.

7.2. Motivational state

At the end of every learning round, participants answered three questions on a


seven-point scale, which were averaged together. One item measured their positive
valence (“The task is fun”), and two items measured their self-efficacy or probability
of success (“I’m sure I will find the correct solution”, “It’s clear to me how to
continue”). The first self-efficacy item is similar to those used in studies by Bandura
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 299

and Wood (1989) and Schoppek (1997). The three items were homogenous
(Cronbach’s a=0.80).

7.3. Persistence

In the literature, persistence is often measured as time on task (in minutes). How-
ever, for the biology-lab system, time per round seems to lack reliability. Using the
data from Vollmeyer et al. (1998), we analysed time spent on each round. Across
the three learning rounds time on round was very inconsistent (Cronbach’s a about
0.30) and did not correlate with the performance measure on that round. For example,
spending more time on Round 2 did not lead to more knowledge about the system
after Round 2. An explanation for this result could be that learners can have different
reasons to spend more or less time on a round (e.g., a learner who wants to discover
the system’s structure right now, as opposed to a learner who is too tired and starts
daydreaming). Furthermore, these reasons may even vary over time. As time on task
was not reliable, we chose a more robust measure, namely the number of rounds that
a learner spent on finding the system’s structure (learning rounds) and the number of
rounds the learner spent on reaching the goal states (application rounds).

8. Dependent variables

Two dependent variables measured learning.

8.1. Structure score (knowledge acquisition)

After each round of the learning phase, participants had to complete a diagram.
They used this diagram to indicate their knowledge about the system’s structure by
drawing for each relationship they noticed: a link, a direction (plus or minus), and
a weight between an input and an output. To indicate that an output had a decay,
participants could write a weight into the empty circle attached to each output.
A structure score derived from the links, directions, and weights was calculated
(see Vollmeyer et al., 1996). This score varied between 3.0 (best value) and a theor-
etical minimum of 1.8. (Because a correction for guessing was applied, a value was
negative if a participant incorrectly guessed too much.) Participants had as many
structure scores as they had rounds that they chose to spend exploring the system.
For some analyses, only the individual’s last round was used.

8.2. Goal achievement (knowledge application)

Goal achievement in reaching the goal state during the application phase was
computed as the sum of the absolute differences between the target and the obtained
number for each of the three output variables. As this measure produced a skewed
distribution, the variance was corrected by applying a logarithmic transformation
(ln). Goal achievement was computed for each of the six trials that comprised each
300 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

round in the application phase, in order to determine how well participants were
able to approach the target goal. As there was no difference in performance between
trials, the mean error for the six trials was used. Because participants could choose
the number of rounds spent reaching the goal, they could have several goal achieve-
ment scores.
Note that in Vollmeyer et al. (1996), this measure was referred to as solution
error, instead of the term goal achievement. However, this meant that high scores
were indicators of poor performance. So that all performance measures would be in
the same direction, we subtracted all these scores from the arbitrary constant seven,
a linear transformation that has no effect on the correlations that we will report,
except to reverse their sign.

9. Procedure

Participants were shown the biology-lab system and told that they could use as
many rounds as they liked for both learning about the system and reaching the final
goal states. While working with the system they were allowed to take as many notes
as they wanted and could use a calculator.
Vollmeyer et al. (1996) showed that a good strategy for learning about the system
was to vary only one input variable at a time. This strategy was explained to all
participants in order to reduce the variance of their performance. After the instruc-
tions they filled out the QCM. Then they began exploring the system. After every
round they filled out the structure diagram and answered the motivational state ques-
tionnaire. When the participants said they had learnt enough, they were presented
with the goal states. The application phase was finished whenever the participants
wanted. Depending on the participant, the entire experiment took between one and
four hours. As the study was presented as lasting two hours (credit was for two
hours), most of the participants stayed between one and two hours. Only seven asked
for a second appointment within the next week, for which they received extra credit.
As 86 percent of the participants received the same reward, we assumed no credit
effects on motivation.

10. Results

10.1. Questionnaire on current motivation (QCM)

The means of the initial motivation (QCM) show that this task was expected to
be challenging, m=4.95, s.d.=1.75, and the other motivational factors were rated near
the middle of the scale (interest: m=3.75, s.d.=1.09; mastery confidence: m=4.14,
s.d.=1.09; incompetence fear: m=3.77, s.d.=1.13). All means refer to a seven-point
scale (1 to 7). The QCM factors’ intercorrelations can be seen in Table 2. It was
evident that for this task the constructs were not independent from each other. The
highest correlation was between incompetence fear and mastery confidence, a finding
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 301

Table 2
Correlations of the motivational factors and the dependent variables (n=51)

Mastery Incompetence Challenge Interest


confidence fear

Incompetence fear ⫺0.57b


Challenge ⫺0.06 0.32a
Interest 0.48b ⫺0.32a 0.06
Motivational state (last round) 0.25 ⫺0.28 ⫺0.04 0.39b
Structure score (last round) ⫺0.10 0.14 ⫺0.20 0.21
Goal achievement (last round) ⫺0.01 0.25 0.07 0.12
Learning rounds 0.32a ⫺0.14 ⫺0.03 0.36b
Application rounds ⫺0.19 0.30a 0.12 0.10

a
p⬍0.05.
b
p⬍0.001.

that argued for combining them into a latent variable for a structural equation analy-
sis.

10.2. Persistence

Participants chose to spend about five rounds (m=5.20, s.d.=2.56 on exploring the
system and about four rounds (m=3.72, s.d.=2.50) on reaching the goal states in the
application phase.
Having an experimental design in which participants chose the amount of learning
rounds made it impossible to run a repeated-measures analysis over all rounds
because of missing values. Therefore, our analysis comparing rounds can only be
descriptive. First, we will describe the analysis of Fig. 2, which shows after what
round participants terminated their learning phase. Second, the same analyses are
given for Fig. 3, which shows when the application phase was finished.
Our hypothesis was that initial motivation affected learning via persistence. Thus,
we expected a positive relationship between initial motivation and persistence (i.e.,
number of learning rounds). Table 2 shows that the QCM-factors mastery confidence
and interest did indeed correlate with learning rounds. We also expected a positive
relationship between learning rounds and the structure score gained in the individ-
ual’s last round, however, the correlation was not significant, r=⫺0.11, p=0.46. As
this result seems quite surprising and contradicts widely accepted assumptions about
practice and learning, we explored whether participants who took more rounds were
still learning. For this analysis we formed four groups: (1) participants who finished
in Rounds 3 or 4 of the learning phase (n=27); (2) participants who finished in
Rounds 5 or 6 (n=11); (3) participants who finished in Rounds 7 to 9 (n=9); and (4)
participants who finished in Rounds 10 to 15 of the learning phase (n=4). The mean
structure scores during the learning phase for the four groups are shown in Fig. 2.
At first glance, all groups appeared to have acquired knowledge over the rounds.
However those who finish earlier had a higher structure score from the beginning.
302 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

Fig. 2. Structure scores for participants over rounds depending on the round in which they finished the
learning phase.

For three groups (the 10 to 15 group was excluded as it consisted of only four
participants), we tested whether their structure scores over rounds formed a statisti-
cally significant linear trend. This was only true for participants finishing in Rounds
7 to 9, F(1, 8)=9.59, p=0.010. For the other two groups, the linear trend was not
significant (participants finishing Rounds 3 or 4, F[1, 26]=2.60, p=0.12, participants
finishing Rounds 5 or 6, F[1, 10]=1.28, p=0.29). Therefore, it appears that parti-
cipants who stopped early had quickly gained most of the knowledge they needed;
whereas participants who finished later acquired their knowledge more slowly, but
ended up with similar amounts of knowledge to that acquired by the faster parti-
cipants. An analysis of variance with the individual’s last learning round as depen-
dent variable and the learning rounds as independent variable (finishing Rounds 3
or 4, finishing Rounds 5 or 6, finishing Rounds 7 to 9) yielded a not significant
effect, F(2,44)=0.19, p=0.83.
Similar to the learning rounds, the (negative) correlation between the individual
last goal achievement and the number of application rounds was not significant,
r(51)=⫺0.11, p=0.45. Therefore, in analogy to the learning rounds, we formed four
groups: (1) participants who finished in Round 1 in the application phase (n=17);
(2) participants who finished in Rounds 2 or 3 (n=12); (3) participants who finished
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 303

Fig. 3. Goal achievement for participants over rounds depending on the round in which they finished
the application phase.

in Rounds 4 or 5 (n=11); and (4) participants who finished in Rounds 6 to 12 (n=11).


The same pattern, as that found for the learning rounds can be seen for the application
rounds (see Fig. 3. However, for the application phase all tests for a linear trend
were statistically significant (participants finishing in Rounds 2 or 3, F[1, 11]=16.23,
p=0.002, participants finishing in Rounds 4 or 5, F=[1, 10]=9.52, p=0.012; parti-
cipants finishing in Rounds 6 to 12, F[1, 10]=8.97, p=0.013). This shows that persist-
ence helped those who were not so accurate in reaching the goal states at the begin-
ning.
Could persistence be a personal trait as it is, for example, in Kuhl’s (1984) action
control theory? In that case, participants who spent more rounds for learning should
have also spent more rounds in reaching the goal states. However, this was not true:
participants who chose to use many learning rounds did not necessarily use many
rounds reaching the goal states, r(51)=⫺0.06, p=0.69.

10.3. Influence of initial motivation on mediating and dependent variables

Before running a path analysis (using EQS; Bentler, 1995) we looked at the corre-
lations between initial motivation and the mediating and dependent variables (see
Table 2). As participants could choose how many rounds they wanted to spend learn-
304 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

ing, we used their structure score and motivational state for the last round. Similarly,
goal achievement was measured for the last application round. The correlations
showed again that challenge was not a good predictor for learning on this task,
however, it could be a moderator (Vollmeyer, Rollett & Rheinberg, 1997b). Rather
than going into the details of each correlation, a path analysis is a better way to
analyse interdependencies.

10.4. The cognitive-motivational process model

Our research question was: does initial motivation affect learning via certain
mediators? In Vollmeyer and Rheinberg (1998) and Vollmeyer et al. (1997a) strategy
systematicity and motivational state were found to be significant mediators. In this
study, we added persistence on the task as a mediator. An EQS-analysis (see Fig.
4) tested whether our empirical data fit to our theoretical assumptions. As in our
previous studies, we had three measuring points for the learning phase (Round 1,
Round 2, and the individual’s last round). In the application phase, 25 percent of
the participants had already finished in Round 1. Therefore, for the EQS-model we
used only the goal achievement for a participant’s last round. Fitting the theoretically
derived model to the data gave a high model fit, CFI=1.00, X2=(71)=67.02, p⬎0.05.
In this study, participants were instructed on how to use a systematic strategy. As
Fig. 4 shows, if they used this strategy, it led to better learning and hence to better
solutions. We called this path a cognitive path (systematic strategy→more knowl-
edge→more accurate goal achievement). However, we were especially interested in
motivational effects. As in Vollmeyer and Rheinberg (1998) interest fits worse as a
predictor for the learning process than the latent motivation variable that consisted
of mastery confidence and incompetence fear. However, the correlations in Table 2
show that interest did affect learning. Incompetence fear and mastery confidence

Fig. 4. Path analysis for the cognitive–motivational process model.


R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 305

strongly affected how participants felt during learning: if they were more confident
at the beginning of the task, then during the learning phase they had more fun, and
were more certain of success (as measured by motivational state). The motivational
state affected strategy systematicity and also knowledge about the system’s structure
(i.e., structure score): the more motivated the participants were, the more they stayed
with the systematic strategy and acquired more knowledge.
On the basis of other studies (Vollmeyer et al., 1997a Study 1), we expected a
direct effect of motivational state on goal achievement which, however, did not
occur. A possible explanation for this is that we changed the procedure by allowing
participants to choose the number of rounds for learning and application. As Fig. 3
shows, this option allowed learners to compensate poor initial goal achievement via
persistence. Thus, the goal achievement measure in this study cannot be directly
compared to that in our former studies.
In the current study, two new mediating variables were included to measure per-
sistence: learning rounds and application rounds. The best predictors for learning
rounds were mastery confidence and structure score: participants with confidence in
their success tried for longer to discover the system’s structure, and those with more
knowledge in Round 1 stopped learning about the system earlier. For application
rounds, the knowledge in Round 1 was again a good predictor, but the motivational
factor that was predictive was incompetence fear instead. Participants who had more
knowledge in Round 1 of the learning phase spent fewer rounds trying to reach the
goal states, and participants with incompetence fear tried for longer to reach the goal
states. This result demonstrated that initial motivation did influence how many rounds
people worked on that task. However, inter-individual differences in learning or
application rounds did not correspond with better results: there were no significant
paths (see Fig. 4) from the number of learning or application rounds to the corre-
sponding performance measures (structure score or goal achievement).

11. Discussion

In this study, we had two aims: (1) to replicate the cognitive-motivational process
model under the condition that participants could choose the number of rounds spent
exploring the system, as well as the number of rounds in which they tried reaching
the goal states; (2) to explore whether persistence is a possible mediator for motiv-
ational effects on performance.

11.1. A cognitive-motivational model

In the introduction we asked the question: what function does motivation have
for learning? This very general question was made more specific by conceptualising
four motivational factors. Of these factors, we found mastery confidence and incom-
petence fear were good predictors for the learning process. In addition to initial
motivation, we measured participants’ dynamic motivational state during the learning
process. This experimental design allowed us to study how motivational and cogni-
306 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

tive variables interact. With a path analysis we demonstrated that, apart from the
cognitive path (systematic strategy → more knowledge → more accurately reaching
the goal states), the motivational state affected knowledge acquisition and encouraged
participants to stay with the instructed systematic strategy. This result replicates earl-
ier findings (Vollmeyer & Rheinberg, 1998 and Vollmeyer et al., 1998) and thus
adds strong support to the claim that a cognitive–motivational model for learning
can best explain learning outcomes.

11.2. Motivation and persistence

The second aim of the study was to examine the role of persistence in learning.
Our examination of persistence yielded results that were surprising at a first glance:
motivation had an impact on persistence, however, highly persistent learners had no
better performance than less persistent learners. The latter result seems to run against
the plausible assumption that time on task and exercise affects performance. This
contradiction, however, holds only for the analysis of inter-individual differences. If
we analyse intra-individual gains in knowledge acquisition and application instead,
Figs. 2 and 3 showed that there was an effect of persistence on learning: those
learners who, at the beginning, had little knowledge about the system’s structure
used more rounds to explore the system. However, in subsequent rounds they
improved and nearly reached the performance level of those learners who knew more
about the system’s structure right from the first round. Thus, the slower learners
compensated for their initial knowledge deficits in the first two rounds with persist-
ence, which led them to use a greater number of learning rounds. This was especially
true for learners who were high on mastery confidence (see Fig. 4, path from mastery
confidence to learning rounds). More learning rounds could compensate for the defi-
cits in the first two learning rounds, although it did not lead to a better performance
than that of the learners who finished earlier. Therefore, the impact of learning rounds
on performance did not provide a significant path in the EQS-analysis.
The same interpretation applies to persistence in the application phase. Again,
learners with less knowledge in the first round of the learning phase compensated
for this deficit with higher persistence in reaching the goal states. Motivation was
important for producing persistence, in particular, high incompetence fear led lear-
ners to use many application rounds. In order to try to understand why the latter
relationship arose, we examined the correlations between the individual items of the
incompetence fear scale and the number of application rounds. Items concerning
social evaluation (e.g., “I’m afraid I will make a fool out of myself”) were especially
responsible for this relationship. Obviously, this kind of anxiety could cause learners
to persist in striving for a given specific goal. This is not a desirable incentive for
learning, of course, but it seemed to work in our situation, in which adult learners
were convinced that an expert (e.g., experimenter) would evaluate their performance.
Regardless of the quality of this incentive, we know from participants’ persistence
that those who used many rounds reached the same level of performance as those
who finished in early rounds.
Can these results be transferred to everyday learning situations, for example, learn-
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 307

ing at school or university? As the analysis of initial motivation showed, participants


regarded our task as very challenging. This often is not true for tasks like homework.
However, if a task is not challenging, people lose critical incentives for learning and
it becomes hard to believe that students will be motivated enough to work on the
task until they are compensated for their initial knowledge deficits. Therefore, we
limit the scope of our results to tasks that have a high challenge for learners. With
less challenging tasks, other activities might become too attractive, especially if mas-
tering the task requires high cognitive effort.
Another limitation of the scope of our results arises from the fact that the parti-
cipants’ performance had no influence on their studies. Under conditions in which
performance is more important, content specific interest might have a bigger effect
than we found in our study. In everyday situations, the correlation between interest
and performance is about r⬇0.30 (Krapp, 1992) which is higher than the correlations
in our study (interest with structure score r=0.21, interest with goal achievement
r=0.12). The argument that effects are always smaller in an experimental setting
should not hold for this study because the situation should be irrelevant in an analysis
of content-specific interest. The content is the critical issue, both in everyday life
and the experimental setting.

Acknowledgements

We would like to thank Bruce Burns and two anonymous reviewers for comments
on this paper. This research was supported by DFG Grant No. 514/5 to Regina
Vollmeyer and Falko Rheinberg. A preliminary report of this experiment was
presented at the nineteenth annual Conference of the Cognitive Science Society,
1997 (August).

References

Atkinson, J. W. (1957). Motivational determinants of risk-taking behavior. Psychological Review, 64,


359–372.
Atkinson, J. W., & Raynor, J. O. (1974). Motivation and achievement. Washington, DC: Winston.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavior change. Psychological Review,
84, 191–215.
Bandura, A. (1991). Human agency: The rhetoric and the reality. American Psychologist, 46, 157–162.
Bandura, A., & Wood, R. (1989). Effect of perceived controllability and performance standards on self-
regulation of complex decision-making. Journal of Personality and Social Psychology, 56, 808–814.
Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software, Inc.
Bloom, B. S. (1976). Human characteristics and school learning. New York: McGraw-Hill.
Boekaerts, M. (1996a). Self-regulated learning at the junction of cognition and motivation. European
Psychologist, 1, 100–112.
Boekaerts, M. (1996b). Personality and the psychology of learning. European Journal of Personality, 10,
377–404.
Boekaerts, M., & Otten, R. (1993). Handlungskontrolle und Kernanstrengung im Schulunterricht [Action
control and learning related effort in the classrom]. Zeitschrift für Pädagogische Psychologie, 7,
109–116.
308 R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309

Campbell, D. T. (1957). Factors relevant to the validity of experiments in social settings. Psychological
Bulletin, 54, 297–311.
Carroll, J. B. (1963). A model of school learning. Teachers College Record, 64, 723–733.
Carroll, J. B. (1985). The model of school learning: Progress of an idea. In C. W. Fisher, & D. C. Berliner,
Perspectives on instructional time (pp. 29–58). New York: Longman.
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measure-
ment, 20, 37–46.
Czikszentmihalyi, M. (1975). Beyond boredom and anxiety. San Francisco, CA: Jossey-Bass.
Fisher, C. W. (1996). Academic learning time. In E. De Corte, & F. E. Weinert, International encyclopedia
of developmental and instructional psychology (pp. 675–694). Oxford: Elsevier.
Funke, J. (1991). Solving complex problems: Exploration and control of complex systems. In R. J.
Sternberg, & P. A. Frensch, Complex problem solving: Principles and mechanisms (pp. 195–222).
Hillsdale, NJ: Erlbaum.
Helmke, A., & Schrader, F. W. (1996). Kognitive und motivationale Bedingungen des Studierverhaltens:
Zur Rolle der Lernzeit [Cognitive and motivational conditions of study behaviour: The role of time
on learning]. In J. Lompscher, & H. Mandl, Lehr-und Lernprobleme im Studium [Teaching and learn-
ing problems at university] (pp. 39–53). Bern: Huber.
Issing, L. J., & Klimsa, P. (1997). Information und Lernen mit Multimedia [Information and learning
with multimedia]. (2nd ed). Weinheim: Psychologie Verlags Union.
Krapp, A. (1992). Das Interessenkonstrukt. Bestimmungsmerkmale der Interessenhandlung und des indivi-
duellen Interesses aus der Sicht einer Person-Gegenstands-Konzeption [The interest construct]. In A.
Krapp, & M. Prenzel, Interesse, Lernen, Leistung [Interest, learning, performance] (pp. 297–329).
Münster: Aschendorff.
Kuhl, J. (1984). Volitional aspects of achievement motivation and learned helplessness: Toward a compre-
hensive theory of action control. Progress in Experimental Personality Research, 13, 100–164.
Rheinberg, F. (1997). Motivation. (2nd ed.). Stuttgart: Kohlhammer.
Schiefele, U. (1991). Interest, learning, and motivation. Educational Psychologist, 26, 299–323.
Schiefele, U., & Rheinberg, F. (1997). Motivation and knowledge acquisition: Searching for mediating
processes. In P. R. Pintrich, & L. Maehr, Advances in motivation and achievement (pp. 251–301).
Greenwich, CT: JAI Press.
Schoppek, W. (1997). Wissen bei der Steuerung dynamischer Systeme – ein prozeβorientierter For-
schungsansatz [Knowledge in the control of dynamic systems – A process oriented approach]. Zeit-
schrift für Psychologie, 205, 269–295.
Seegers, G., & Boekaerts, M. (1993). Task motivation and mathematics achievements in actual task
situations. Learning and Instruction, 3, 133–150.
Simon, H. A. (1967). Motivational and emotional controls of cognition. Psychological Review, 74, 29–39.
Volet, S. E. (1997). Cognitive and affective variables in academic learning: The significance of direction
and effort in students’ goals. Learning and Instruction, 7, 235–254.
Vollmeyer, R., Burns, B. D., & Holyoak, K. J. (1996). The impact of goal specificity on strategy use
and the acquisition of problem structure. Cognitive Science, 20, 75–100.
Vollmeyer, R., & Rheinberg, F. (1998). Motivationale Einflüsse auf Erwerb und Anwendung von Wissen
in einem computersimullerten System [Motivational influences on the acquisition and application of
knowledge in a simulated system.]. Zietschrift für Pädagogische Psychologie, 12, 11–23.
Vollmeyer, R., Rollett, W., & Rheinberg, F. (1997a). How motivation affects learning. In M. G. Shafto, &
P. Langley, Proceedings of the nineteenth annual conference of the Cognitive Science Society (pp.
796–801). Hillsdale, NJ: Erlbaum.
Vollmeyer, R., Rollett, W., & Rheinberg, F. (1997b). Herausforderung, Erfolgszuversichtlichkeit und
Miβerfolgsbefürchtung beim Wissenserwerb in komplexen Systemen [Challenge, mastery confidence
and incompetence fear when learning a complex system]. In H.-P. Langfeldt, 6. Tagung der Fach-
gruppe Pädagogische Psychologie in der Deutschen Gesellschaft für Psychologie, Frankfurt am Main
(pp. 160). Landau: Verlag Empirische Pädagogik.
Vollmeyer, R., Rollett, W., & Rheinberg, F. (1998). Motivation and learning in a complex system. In P.
Nenniger, R. S. Jäger, A. Frey, & M. Wosnitza, Advances in motivation (pp. 53–67). Landau: Verlag
Empirische Pädagogik.
R. Vollmeyer, F. Rheinberg / Learning and Instruction 10 (2000) 293–309 309

Vosniadou, S., De Corte, E., Glaser, R., & Mandl, H. (1996). International perspectives on the design
of technology-supported learning environments. Mahwah, NJ: Erlbaum.

You might also like