🎾 Calibration In Tennis The Role Of Feedback And Exp¶
Giới Thiệu¶
Calibration In Tennis The Role Of Feedback And Exp — tài liệu 6 trang từ thư viện sách tennis.
Tóm tắt nội dung (trích từ tài liệu gốc): See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239539080 Calibration in tennis: The role of feedback and expertise Article � January 2007 CITATIONS READS 4 193 1 author: Gerard Joseph Fogarty University of Southern Queensland 126 PUBLICATIONS 3,956 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Safety Climate View project Adopting new technology View project All content following this page was uploaded by Gerard Joseph Fogarty on 26 May 2014. The user has requested enhancem
Lưu ý: Nội dung dưới đây được trích xuất tự động từ PDF gốc tiếng Anh, giữ nguyên ngôn ngữ để bảo toàn độ chính xác kỹ thuật.
Nội Dung Gốc (Tiếng Anh)¶
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239539080
Calibration in tennis: The role of feedback and expertise
Article � January 2007
CITATIONS READS
4 193
1 author:
Gerard Joseph Fogarty
University of Southern Queensland
126 PUBLICATIONS 3,956 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Safety Climate View project
Adopting new technology View project
All content following this page was uploaded by Gerard Joseph Fogarty on 26 May 2014.
The user has requested enhancement of the downloaded file.
Calibration in tennis: The role of feedback and expertise
Gerard J. Fogarty (fogarty@usq.edu.au)
Department of Psychology
University of Southern Queensland, Toowoomba QLD 4350 Australia
Anthony Ross (a_ross4@hotmail.com)
Department of Psychology
University of Southern Queensland, Toowoomba QLD 4350 Australia
Abstract subject is well calibrated. A positive bias score
indicates overconfidence, while a negative bias score
People are usually overconfident in their expectations of indicates underconfidence. Research using the
what they can do. The term used to describe the extent calibration paradigm to judge metacognitive bias in
which people are accurate in these self-assessments is the sporting domain normally varies from calibration
"calibration". The present study focused on the effects of testing in the cognitive domain, in that subjects are
feedback and expertise on calibration in the physical task of immediately aware of their result in physical tasks.
serving in tennis, and three cognitive tests related to tennis. Therefore, performance predictions are made before a
Sixty-four male and female tennis players, ranging in block of attempts, rather than after.
ability level from social to professional, took part in the
study. Participants completed a tennis rules test, a tennis Studies in calibration have found that people are
general knowledge test, and a tennis technique test, along generally overconfident when predicting their own
with confidence ratings regarding their answers. They then performance. In the domain of motor skills, Cohen,
completed two trials of two tennis serving tasks, which also Dearnaley, and Hansel (1956) found that when bus
involved estimating their expected performance on each drivers were asked to judge whether they could drive
trial. The results indicated that participants were through a narrow gap, they were generally
overconfident on the general knowledge test, the rules test, overconfident and more experienced drivers were not
and the more difficult serving task, but well calibrated on any better calibrated than less experienced drivers.
the technique test, and the easier serving task. Expertise West and Stanovich (1997) also found
was not strongly related to calibration and feedback was not overconfidence when participants completed a penny
beneficial for the more difficult serving task. The results slide task on a table top, and although calibration
have implications for decision making in tennis players, improved on a second trial, significant
especially in relation to the tendency towards overconfidence remained.
overconfidence on difficult tasks.
Calibration studies of actual physical performance
Introduction in sport are hard to find. Jagacinski, Isaac, and Burke
(1977) tested the ability of college-level and
The ability to monitor past performance and predict professional basketball players to take uncontested
future performance is an important part of day-to-day shots from different positions in the court. Before
life. People use beliefs about their own abilities to they took the shot, both the player and a passive
help judge performance. Because these beliefs often observer predicted if the shot would be made. No
do not match objective performance, they can lead to evidence was found that the players were more
performance judgments that do not relate to real accurate than the observers, and both were
accomplishment (Dunning, Johnson, & Ehrlinger, overconfident in their predictions even when there
2003). The ability to be realistic when rating previous were penalties for poor predictions. McGraw,
performance and making future probability Mellers, and Ritov (2004) measured the confidence
judgments, often referred to as being "well that recreational basketball players felt while making
calibrated", has been shown to have benefits in areas shots and the pleasure they felt with the results of
such as motivation and goal setting (Horgan, 1992). those shots. They also found that most players were
overconfident, and those who were more
In the cognitive domain, where much of the overconfident experienced less enjoyment.
calibration research has been conducted, simple
techniques are used to assess calibration. Participants Fogarty and Else (2005) used the calibration
usually answer knowledge-related questions and then paradigm to measure metacognitive bias in 54 male
indicate how confident they are that their answer was golfers ranging in age from 13 to 75. Golfers were
correct in percentage terms (Lichtenstein & Fischoff, required to complete a putting task and a chipping
1977). When all questions are completed, the bias task after first estimating how well they would
score can be obtained by subtracting the percentage perform on each of the tasks. Each exercise was
of correct responses from the average confidence repeated once. Results indicated that golfers tended to
rating. If the proportion of correct responses be reasonably well calibrated on the putting task but
corresponds with the average confidence rating, the
148
slightly overconfident on the chipping task. calibration. In the domain of general knowledge,
Participants were also overconfident on the golf rules Pulford and Colman (1997) found that feedback is
test, which is consistent with other cognitive only effective in improving calibration for hard
calibration research. questions. Kruger and Dunning (1999) found that in
comparison to their more competent peers,
This research was extended by Graham (2006) incompetent subjects were less able to use feedback
when he studied 137 junior golfers who gave to adjust calibration. Fischer and Budescu (2005)
estimates of their ability on putting, chipping, and found that when testing categorical decision making,
pitching tasks before completing the physical tests. learning depends on the type of feedback given.
Two experiments were conducted. Experiment 1,
which required players to putt and chip to the shortest The ability of athletes to learn from feedback
target, revealed good calibration whereas experiment when completing physical tasks is critical for success
2, which required players to chip and pitch at a more in sport. Fogarty and Else (2005) found improvement
difficult target, revealed overconfidence. in calibration of putting and chipping in golf when
using only two trial blocks. Graham (2006), also
Summarising these findings, it appears that working in the sport of golf, found that players who
overconfidence generally exists in the cognitive were initially poorly calibrated used feedback from
domain, while in physical tasks and sport, the limited earlier trials to become better calibrated. On the basis
research suggests that people vary from good of these findings, it was therefore hypothesised (H3)
calibration to overconfidence. In relation to the that calibration would improve on a tennis serving
present study of metacognition in the sport of tennis, task where feedback is immediate and complete.
it was therefore hypothesised (H1) that
overconfidence would be displayed on a test of tennis Method
rules, a test of tennis technique, a test of tennis
general knowledge, and on two tennis serving tasks. Participants
The findings relating to expertise are less clear, Sixty-four tennis players ranging in age from 14
but researchers have generally found that expertise to 48 years (M = 20.63, SD = 6.97) were recruited
does lead to better calibration. Keren (1987) found through personal contact in Cairns, the Sunshine
that expert bridge players were well calibrated when Coast, Brisbane, and the Gold Coast to take part in
predicting the chances of a final contract being calibration tests. Players were selected based on
reached whereas amateurs were overconfident. variation in expertise and gender. There were 41 male
Horgan (1992) found that better chess players were and 23 female players. Participants included current
well calibrated, whereas players with lower ratings and former professional players, social adult players,
were overconfident. Toward (1997) split 24 female and tournament standard junior players. Current and
undergraduate basketball players into expert and former professional players were defined as experts
novice groups where classification was based on how (n = 25), and juniors and social players were defined
many seasons of competitive basketball members had as non-experts (n = 39) for the expertise analyses.
played. He tested the relationship between action and Participants were also ranked according to expertise
cognition in the basketball foul shot, and found that by a representative from Tennis Queensland as a
experts monitored and predicted outcomes better than cross reference for these groupings.
novices. Against this trend, Fogarty and Else (2005)
found no evidence to suggest that lower handicap Instruments
golfers were better calibrated on chipping and putting
than high handicap golfers. Despite this last finding, Test of tennis rules. Fifteen multiple-choice questions
the weight of evidence suggests that calibration and were designed to test calibration in knowledge of
expertise are associated. It was therefore tennis rules (e.g., What is the ruling if during doubles
hypothesised (H2) that expert tennis players would a player receives out of turn?). Participants were
display better calibration on a serving task than non- asked to circle the correct answer. They were then
experts. asked to indicate how confident they were that their
answer was correct by selecting a confidence rating
The final variable examined in this study was for each question in percentage terms (25%, 50%,
feedback, where again the available evidence 75%, or 100%). Three scores were attained from this
suggests an effect provided certain conditions are test: Tennis Rules Confidence Rating, Tennis Rules
met. Keren (1987) suggested that the accuracy of Correct Answers (converted to a percentage), and
calibration depends on the similarity of the mental Tennis Rules Bias Score, the Bias score being the
processes necessary for repeated probability difference between predicted and obtained scores,
assessments. When task items are similar and where positive scores suggest overconfidence and
sufficient practice has occurred, he argued that it is negative scores suggest underconfidence.
feasible to develop procedures that can lead to
accurate predictions. Keren also suggested that
immediate, relevant feedback is imperative for good
149
Test of tennis general knowledge. Fifteen multiple- participant was provided with an information sheet
choice questions were designed to test calibration in detailing the study. Parental consent was obtained for
knowledge of tennis general knowledge (e.g., Which each participant under 18 years. Informed consent
year did Pat Cash win Wimbledon?). The same was held for those participants over 18 years.
procedure was followed as for the test of tennis rules Participants were offered the chance to indicate their
giving a further three measures: General Knowledge desire to receive a copy of the results of the study via
Confidence Rating, General Knowledge Correct the consent form. Immediately prior to each task,
Answers (converted to a percentage), and General participants were given a verbal description of the
Knowledge Bias Score. task, told approximately how long it would take to
complete the task, and that results would not be
Test of tennis technique. Fifteen multiple-choice shared with anyone outside the experimental team.
questions were designed to test calibration in Participants completed the tests, the questionnaire,
knowledge of tennis technique (e.g., Which grip and the tasks in the order they appear above. Average
would most advanced players use for a smash?). The testing time was 45 minutes.
same procedure was followed as for the previous tests
giving a further three measures: Technique Results
Confidence Rating, Technique Correct Answers
(converted to a percentage), and Technique Bias Six cases were identified as having missing values.
Score. These cases did not perform the serving tasks. This
resulted in 64 cases completing the cognitive tasks
Serving task 1. Participants were required to hit 10 and 58 cases completing the serving tasks.
first serves on a tennis court into a target area that Examination of z-scores calculated from the
measured one-quarter of the service box. Participants skewness and kurtosis statistics indicated that the
were allowed five warm up serves before the variables were normally distributed.
instructions were explained. They were then asked to
estimate how many first serves out of the 10 they Hypothesis 1 stated that there would be general
could hit into the target area. Instructions emphasized overconfidence displayed on the tennis rules test, the
that the first serves were to be hit like they would in a tennis general knowledge test, the tennis technique
real match, and the estimate was to be a realistic test, and the two serving tasks. The hypothesis was
estimate of their actual score and not what they would tested by running a repeated measures ANOVA for
`like' to score. Participants then completed the 10 the cognitive tasks. This resulted in a 2 � 3
serves. Three scores were attained from this task: (calibration: confidence rating/percentage correct;
Serving Estimate 1 (converted to a percentage), task: rules/general knowledge/technique) within-
Serving Score 1 (converted to a percentage), and subjects design. The analysis revealed a significant
Serving Bias 1, the Bias score being the difference interaction between calibration and task [Wilks'
between estimated and obtained scores, where Lambda F(2,126) = 34.28, p < .05], indicating that
positive scores suggest overconfidence and negative the amount of participant overconfidence depended
scores suggest underconfidence. on the type of cognitive task (See Figure 1).
Serving task 2. Participants were required to hit 10 Paired sample t-tests showed that on the rules test,
first serves on a tennis court into a target area that participants were overconfident, t(63) = 9.86, p < .05.
measured one-eighth of the service box. The same On the general knowledge test, participants were
procedure was followed as for Serving Task 1 except overconfident, t(63) = 7.39, p < .05. On the technique
that participants were not allowed warm up serves. test, participants were well calibrated, t(63) = 1.70, p
Three outcome measures were attained: Serving > .05.
Estimate 2 (converted to a percentage), Serving Score
2 (converted to a percentage), and Serving Bias 2. 80 rules
Serving task 1 retest. Serving Task 1 was repeated 75 genknow
immediately after the completion of Serving Task 2,
giving a further three measures: Serving Estimate 3, marginal mean % tech
Serving Score 3, and Serving Bias 3. 70
Serving task 2 retest. Serving Task 2 was also 65
completed a second time, giving Serving Estimate 4,
Serving Score 4, and Serving Bias 4. 60
Procedure 55
Ethics approval was attained from the University of 50 %correct
Southern Queensland. Data were collected at regional conf rating
tennis associations and various tennis centres across
Brisbane. Prior to the experimental procedure, each calibration
Figure 1. Overall calibration for cognitive tasks.
150
A repeated measures ANOVA was then run on the little evidence of improved calibration was present,
serving tasks. This resulted in a 2 � 2 � 2 (calibration: with 36.21% improving, 34.48% becoming worse,
serving estimates/serving scores; task: 1, 2; trial: 1, 2) and 29.31% staying the same between test and retest.
within-subjects design. There were no significant
interactions. Inspection of the main effects revealed a Discussion
significant effect for calibration [Wilks' Lambda
F(1,57) = 18.33, p < .05] indicating that participants Hypothesis 1 was partially supported. Results of
were overconfident across tasks. the rules test and the general knowledge test were in
accord with previous research that has generally
Hypothesis 2 stated that experts would be better found overconfidence on cognitive tests (Lichtenstein
calibrated than non-experts. To test this hypothesis, & Fischhoff, 1977; West & Stanovich, 1997).
participants were ranked in order of expertise then However, participants were well calibrated on the
split into expert (n = 25) and non-expert (n = 39) technique test which was surprising. Participants
groups. An expert was defined as any current or scored approximately 10% more correct on the
former professional player and a non-expert was technique test than the other cognitive tests, but not
defined as any junior or social player. Repeated only did participants know more on the technique
measures ANOVAs were run on each of the serving test, they also knew more about how much they
tasks. Results indicated that there was no effect for knew. It is possible that this was due to the emphasis
expertise on the first serving task but that an effect that is placed on technical elements when learning
was present in the more difficult second serving task tennis. An average player of tournament standard
[Wilks' Lambda F(1,56) = 5.1, p < .05]. Figure 2 would have been bombarded with technical
shows the nature of this effect on Trial 2 of Serving information for many years through coaching, and
Task 2. should generally have a good understanding of how
much they know in this area. However, little
marginal mean % 65 emphasis is placed on the rules and the history of the
expert game when learning to play, therefore making these
subjects susceptible to the same biases that some
60 researchers have argued cause overconfidence in
non-expert people who have limited knowledge on cognitive
tasks (Kruger & Dunning, 1999).
55
On the serving tasks, players were well calibrated
50 on the easier task (task 1), but overconfident on the
more difficult task (task 2 and the retest of task 2).
45 Fogarty and Else (2005) and Graham (2006) also
found that golfers were well calibrated on easier tasks
40 (putting) and overconfident on more difficult tasks
(chipping and pitching). This pattern is common in
35 the cognitive field where it is known as the
`calibration difficulty-effect' (Keren, 1991). It
30 appears that whatever causes this difficulty-effect is
common to both physical tasks and cognitive tasks.
25 serv score
serv est Findings relating to expertise were equivocal,
with an effect emerging only for Serving Task 2. The
calibration fact that experts were better calibrated than non-
experts on the more difficult task suggests that they
Figure 2. Experts versus non-experts. may be more familiar with aiming at a smaller target
area when serving than non-experts. Fogarty and Else
Hypothesis 3 stated that feedback would improve (2005) also failed to find an effect for expertise in
calibration on the serving tasks. To test this their study of golfers. Perhaps expertise actually
hypothesis, an underconfident person was defined as interferes with calibration by making experts overly
someone whose serving estimate was lower than confident in their judgments, especially in the case of
his/her serving score. An overconfident person was easier tasks,
defined as someone whose serving estimate was
higher than his/her serving score. Those who were The outcomes were also equivocal in relation to
underconfident on task 1 improved calibration (M = feedback. In the case of the first serving task,
10.95) on the retest, t(20) = 4.26, p < .05. Those who participants who were mis-calibrated on Trial 1
were overconfident on task 1 improved calibration (M tended to improve on Trial 2, but this trend was not
= 18.97) on the retest, t(28) = 4.56, p < .05. Those evident on the more difficult second serving task.
who were underconfident on task 2 did not improve Fogarty and Else (2005) and Graham (2006) both
calibration on the retest, t(13) = 1.24, p > .05. Those
who were overconfident on task 2 improved
calibration (M = 13.44) on the retest, t(31) = 3.70, p >
.05. To test this hypothesis further, a count was taken
to determine whether individuals' calibration
improved, declined, or stayed the same on the second
trial. For task 1, 48.28% improved, 24.14% became
worse, and 27.59% stayed the same between test and
retest, indicating support for hypothesis 3. On task 2,
151
found that poor calibration was more likely to occur basketball shooting. Journal of Motor Behavior,
on difficult performance tasks in golf, and that 9(3), 225-234.
feedback did not have much effect on this mis- Keren, G. (1987). Facing Uncertainty in the Game of
calibration. It appears that whatever causes the Bridge: A Calibration Study. Organisational
`calibration difficulty-effect' in sporting tasks also Behavior and Human Decision Processes, 39, 98-
contributes to making it more robust and resistant to 114.
change. Keren, G. (1991). Calibration and probability
judgements: Conceptual and methodological
The findings of the current study should be treated issues. Acta Psychologia, 77, 217-273.
with some caution. It is the first time calibration Kruger, J., & Dunning, D. (1999). Unskilled and
procedures have been applied to the sport of tennis, unaware of it: How difficulties in recognizing
and to the best of our knowledge this is only the third one's own incompetence lead to inflated self-
time this technique has been used in any sport. The estimates. Journal of Personality and Social
major limitation of our methodology is that Psychology, 77(6), 1121-1134.
conditions under which tennis players serve in Lichtenstein, S., & Fischoff, B. (1977). Do those who
competitive matches were not replicated in this study. know also know more about how much they
Although the physical task of serving was the same as know? Organisational Behaviour and Human
would occur in a match, there was no one returning Performance, 20, 159-183.
the serve and participants were not required to finish McGraw, A. P., Mellers, B. A., & Ritov, I. (2004).
the point. Calibration is important in all walks of life, The affective cost of overconfidence. Journal of
including sport, and future research should be Behavioral Decision Making, 17(4), 281-295.
directed at methodological as well as theoretical Pulford, B. D., & Colman, A. M. (1997).
issues. Overconfidence: Feedback and item difficulty
effects. Personality and Individual Differences,
References 23(1), 125-133.
Toward, J. I. (1997). Metacognitive knowledge and
Cohen, J., Dearnaley, E. J., & Hansel, C. E. M. skilled sport performance. Dissertation Abstracts
(1956). Risk and Hazard: Influence of training on International, 58(6), 2136A.
the performance of bus drivers. Operational Tversky, A. & Kahneman, D. (1974). Judgment under
Research Quarterly, 7, 67-82. uncertainty: Heuristics and biases. Science, 185,
1124-1131.
Dunning, D., Johnson, K., & Ehrlinger, J. (2003). West, R. F., & Stanovich, K. E. (1997). The domain
Why people fail to recognize their own specificity and generality of overconfidence:
incompetence. Current Directions in Individual differences in performance estimation
Psychological Science, 12(3), 83-87. bias. Psychonomic Bulletin and Review, 4 (3),
387-392.
Ehrlinger, J., & Dunning, D. (2003). How Chronic
Self-Views Influence (and Potentially Mislead) NB. The correct APA citation for this paper is:
Estimates of Performance. Journal of Personality
and Social Psychology, 84(1), 5-17. Fogarty, G., & Ross, A. (2007). Calibration in tennis: The role of
feedback and expertise. In K Moore (Ed.), Proceedings of the
Fischer, I., & Budescu, D. V. (2005). When do those 2007 Conference of the Australian Psychological Society,
who know more also know more about how much pp.148-152. Brisbane, Australia, 25-29 September, 2007.
they know? The development of confidence and
performance in categorical decision tasks.
Organisational Behavior and Human Decision
Processes, 98(1), 39-53.
Fogarty, G., & Else, D. (2005). Performance
calibration in sport: Implications for self-
confidence and metacognitive biases.
International Journal of Sport and Exercise
Psychology, 3(1), 41-57.
Graham, C. J. (2006). Calibration in sport: The role
of feedback, age and gender. Unpublished
manuscript, University of Southern Queensland,
Toowoomba, Australia.
Horgan, D. D. (1992). Children and chess expertise:
The role of calibration. Psychological Research,
54, 44-50.
Jagacinski, R. J., Isaac, P. D., & Burke, M. W. (1977).
Application of signal detection theory to
perceptual-motor skills: Decision processes in
152
View publication stats