Observational Learning

Self-observation Can Be Detrimental to Learning

Luc Proteau and Mathieu Andrieux

Département de Kinésiologie, Université de Montréal, 2100 Édouard-Montpetit, Montréal, Canada

1 OBJECTIVES

Observation of a model who is performing a motor

skill improves naïve observers’ learning of that skill

(for a recent review see Ste-Marie et al. 2012).

Research has indicated that action observation and

action production share a common neural network,

which is activated when individuals perform a given

motor task and when they observe others performing

that same motor task (Buccino et al. 2001; Cross et

al. 2009). Recent research has shown that optimal

observational learning occurs with the observation

of both novice and expert models rather than either a

novice or an expert model alone (Andrieux and

Proteau, 2013; Rohbanfard and Proteau, 2011). The

aim of the present study was to determine whether

self-observation or a combination of expert and self-

observation would promote learning better than

observation of an expert model and a “generic”

novice model. Such a scenario could be the case

because self-observation would underline errors that

are specific to oneself, whereas the combination of

expert and self-observation would have the

additional benefit of allowing the learner to

determine what to do to improve his or her

performance.

The task that we chose required that the

participants change the relative timing pattern that

naturally emerged from the task constraints to a new

imposed pattern of relative timing. This is similar to

changing one’s tempo when executing a serve in

tennis or a drive in golf.

2 METHODS

One hundred right-handed university undergraduate

students (55 males and 45 females; mean age = 21.2

years; SD = 1.8 years) participated in the

experiment. The participants had no prior experience

with the task. The participants completed and signed

an individual consent form before participation.

The apparatus was similar to that used by

Rohbanfard and Proteau (2011). The task consisted

of successively hitting four barriers of equal size in a

clockwise motion. The distances between each

barrier were 15, 32, 18, and 29 cm. The participants

were required to complete each of the four segments

of the task in an intermediate time (IT) of exactly

300 ms for a total movement time (TMT) of 1200

ms. All of the participants performed four

experimental phases over a period of three

consecutive days.

On day 1 and before the first experimental phase,

all of the participants received verbal instructions

regarding the TMT and IT goals. The first

experimental phase was a preparatory phase, in

which the participants performed 40 trials with

knowledge of the results (KR) of their TMT but not

their ITs. The participants were filmed during this

first experimental phase. At the end of day 1, the

participants were randomly assigned to one of five

groups: control (C), physical practice (PP), expert

and “generic” novice observation (EGO), expert and

self-observation (ESO), and self-observation (SO).

Day 2 began with a pre-test in which all of the

participants performed 20 trials without knowledge

of the results (KR) of their TMT and ITs. The pre-

test was followed by an acquisition phase. In this

phase, the participants in the PP group physically

practiced the experimental task for 40 trials. The

participants in the EGO group individually watched

a video presentation of two models (an expert and a

“generic” novice model) performing 20 trials each.

The films recorded in the preparatory phase were

edited and used in the acquisition phase of the study

for the ESO and SO groups. The ESO group

observed 20 trials performed by an expert model and

20 randomly chosen trials of their own performance

filmed during the preparatory phase (EGO). For both

the EGO and ESO groups, the model was alternated

every 5 trials (i.e., expert model 1: trials 1–5 and

generic novice or oneself: trials 6–10 and so on).

The participants in the SO group observed the 40

trials of their own performance that were filmed

during the preparatory phase. For the PP, EGO, ESO

Proteau, L. and Andrieux, M..

Observational Learning - Self-observation Can Be Detrimental to Learning.

and SO groups, KR of both the TMT and ITs was

provided in ms after each of the physically

performed (PP group) or observed (EGO, ESO, and

SO groups) trials. The participants in the control

group did not take part in the observation or physical

practice protocol but rather read a provided

magazine for the same duration as the observation

phase for the other groups. All of the participants

completed the third and fourth experimental phases:

10-min and 24-hour retention phases that were

similar in all points to the pre-test. The retention

tests were performed on day 2 and day 3.

For each trial, we computed a root mean square

error (RMSE) of relative timing, which indicates in a

single score how much each participant deviated

from the prescribed relative timing pattern. For each

trial,

RMSE =



∑

Segment 4

Segment1





ITi-target



 (1)

where ITi represents the intermediate time for

segment “i”, and the target represents the goal

movement time for each segment of the task (i.e.,

300 ms).

A preliminary analysis of the individual data

revealed two patterns of results depending on the

initial level of performance of the participants in the

pre-test. To better understand how the initial level of

performance influenced the learning of the new

relative timing pattern, we rank ordered the

participants as a function of their initial performance

and made two subgroups that included the 40

participants that had a “better” initial performance

and the 40 participants that had a “poorer” initial

performance. The data of each subgroup were

individually subjected to an ANOVA comparing

five groups (PP, EGO, ESO, SO and C) × three

phases (pretest, 10-min retention, and 24-hour

retention) × four blocks of trials (1-5, 6-10, 11-15,

and 16-20), with repeated measures for the last two

factors.

3 RESULTS

For the participants who had a “better” initial

performance, (Figure 1, top panel) the ANOVA

revealed a significant group × phase interaction (F

[8, 70] = 4.06, p = 0.001). The breakdown of this

interaction did not reveal any difference in RMSE

proceeding from the pre-test to both the 10-min and

the 24-hour retention tests for the C, EGO, and ESO

groups (F [2, 34] = 0.38, 1.20, and 1.10, p > 0.25,

respectively). However, although we noted a

significant decrease in RMSE for the PP group

proceeding from the pre-test to either retention tests

(F [2, 34] = 5.27, p = 0.01), we noted a significant

increase in RMSE for the SO group (F [2, 34] =

6.12, p = 0.005). For the participants who had a

“poorer” initial performance (Figure 1, bottom

panel), the ANOVA revealed a significant group ×

phase interaction (F [8, 70] = 4.67, p < 0.001). The

breakdown of this interaction did not reveal any

difference in RMSE proceeding from the pre-test to

both the 10-min and the 24-hour retention tests for

the C group (F [2, 34] < 1). However, for the EGO,

ESO, SO and PP groups, there was a significant

decrease in RMSE proceeding from the pre-test to

either retention tests (F [2, 34] = 8.71, 4.67, 24.16,

and 3.62, p < 0.05, respectively).

4 DISCUSSION

The live or video observation (Rohbanfard and

Proteau, 2012) of a model practicing a motor skill

favors the learning of that skill by the observers.

One goal of our laboratory is to determine the

conditions of observation that would optimize

learning.

The results of the present study confirm previous

findings indicating that one can learn a new relative

timing pattern through observation (Andrieux and

Proteau, 2013; Rohbanfard and Proteau, 2011).

However, although physical practice resulted in a

significant reduction in the RMSE of relative timing

regardless of the initial level of performance, this

was not the case for the observation groups. In this

regard, our results indicate that if physical practice is

not possible (e.g., because of lack of material or

injury) or not advisable (e.g., when there is an

element of danger), observation is a powerful

learning tool with novices whose performance

largely departs from the desired relative timing

pattern. Our results also suggest that mixed

observation of either oneself or a generic novice

model combined with that of an expert model

provides better learning than self-observation. We

suggest that the comparison of expert and novice

performance in a mixed observation protocol helps

the observer to both detect his or her errors and to

develop a good representation of what to do.

Figure 1: The root mean square error of relative timing as

a function of the initial performance, the experimental

phases and the experimental groups.

However, when a novice’s initial performance is

relatively good, our results indicate that self-

observation could be detrimental to learning a new

relative timing pattern. We suggest that this could be

the case because self-observation (a) does not

underline the technical aspect on which to focus

and/or (b) encourages the learner to try to correct

errors that are beyond his or her actual level of

performance (or to perform maladaptive corrections,

as previously termed by Schmidt and Bjork [1992]).

A mixed observation protocol apparently alleviates

these problems, which should encourage the

practitioner to use an EGO or an ESO protocol

rather than only self-observation.

In conclusion, observation is a powerful learning

tool that is available to anyone with a minimal

equipment requirement. Self-observation does not

appear to be optimal for the learning of new relative

timing patterns and could even be detrimental in

some cases. Therefore, it appears that a mixed

protocol of observation, which allows one to

compare and contrast the performance of a novice to

that of an expert, should be favored.

ACKNOWLEDGEMENTS

This research was supported by a Discovery Grant

provided by the Natural Sciences and Engineering

Research Council of Canada.

REFERENCES

Andrieux, M., Proteau, L., 2013. Observation learning of a

motor task: who and when? Experimental Brain

Research, vol. 229, pp. 125-137.

Buccino, G., Binkofski, F., Fink, G.R., et al., 2001. Action

observation activates premotor and parietal areas in a

somatotopic manner: an fMRI study. European

Journal of Neuroscience, vol. 13, pp. 400-404.

Cross, E.S., Kraemer, D.J.M., Hamilton, A.F.D., Kelley,

W.M., Grafton, S.T., 2009. Sensitivity of the action

observation network to physical and observational

learning. Cerebral Cortex, vol. 19, pp. 315-326.

Rohbanfard, H., Proteau, L., 2011. Learning through

observation: a combination of expert and novice

models favors learning. Experimental Brain Research,

vol. 215, pp. 183-197.

Rohbanfard, H., Proteau, L., 2012. Live vs. video

presentation techniques in the observational learning

of motor skills. Trends in Neuroscience and

Education, vol. 2, pp. 27-32.

Schmidt, R.A., Bjork, R.A., 1992. New conceptualizations

of practice: common principle in three paradigms

suggest new concepts for training. Psychological

Science, vol. 3, pp. 207-217.

Ste-Marie, D.M., Law, B., Rymal, A.M., Jenny, O., Hall,

C., McCullagh, P., 2012. Observation interventions

for motor skill learning and performance: an applied

model for the use of observation. International

Review of Sport and Exercise Psychology, vol. 5, pp.

145-176.

40#

50#

60#

70#

80#

90#

100#

110#

Pre$test'

10$min'

24$hour'

40#

50#

60#

70#

80#

90#

100#

110#

CTRL# PP# ESO# EGO# SO#

Pre$test'

10$min'

24$hour'

Pre-test

10-min

reten on

24-hour

reten on

Root

mean

square

error

rela ve

ming

(m s)

Experimental

groups

Be er

ini al

performance

Poorer

ini al

performance