IEEE Robotics & Automation Magazine - June 2019 - 19

leg-movement frequency was higher for all of the reward conditions than it was for the baseline reading. No significant difference was present in infant motion between the reward phases.
Imitation Results
Imitation occurs when infants replicate observed human
behavior (or, in this work's case, the behavior of humanoid
robots). Operationally, imitation is often defined as an
increase in the performance of some behavior after observing it. Thus, similarly to the contingency learning results,
we defined infant imitation of the NAO robot as the infant
performing the knee-extension ball-kicking motion at a
rate of 1.5 times or greater in one or more of the reward
phases than during the initial baseline. Our trained video
coder's counts of ball-kicking motions throughout the
study procedure yielded the information needed to perform this analysis. We used the infant's baseline as a basis
for comparison against later ball-kicking frequencies.
As shown in Table 3, nine of the 12 participating infants
exhibited imitation behaviors during at least one reward
phase. One infant (TD2) imitated the robot during all of
the reward phases. For the imitators who did not consistently imitate the robot across all of the reward phases, the
temporal aspects of their imitation may prove helpful for
designing follow-up studies. Three infants (TD4, TD5, and
TD9) failed to imitate the robot during the first reward
phase that they experienced but imitated the robot during
the remainder of the reward phases. Two infants (TD6 and
TD8) imitated the robot during only the second reward
phase that they experienced. TD7 imitated the robot until
the final reward phase, TD12 imitated the robot during

only the first reward phase, and TD13 imitated the robot
during only the final reward condition experienced.
Thus, we found that it is most common for infants to experience one or two reward phases before they initiate (or demonstrate) imitation of the robot kicking the ball. Some infants
stopped imitating the
robot during later phases.
These tendencies may
In general, parents thought
indicate that infants take
longer to learn to imitate
the infants were happy,
the robot than to learn the
contingent rewards disexcited, successful, and
cussed previously. Because
of this mix of tendencies,
engaged during their
there was no significant
difference between overall
interactions with the robot.
ball kicks during the baseline and subsequent phases (Figure 5). Our video
data suggest that some infants kicked the ball deliberately. The
perceived effort expressed by some participants may have led
to fatigue or boredom at sufficient levels to halt imitation of
the robot during later reward phases. Overall, these initial
results demonstrate that some young infants may imitate a
humanoid robot, but follow-up research is needed. Some
infants did not kick the ball at all during the initial baseline, so
they needed to perform only a few ball kicks during the
reward conditions to be classified as imitators. Future work
will focus on investigating which aspects of action the infants
strive to imitate and exploring the tendencies of infants to imitate other types of motions or behaviors.

Table 3. The frequency of infant ball-kicking motion during study phases.
Movement Reward
Activity
(Ball Kicks/Min)

Movement and
Lights Reward
Activity
(Ball Kicks/Min)

Movement and
Sound Reward
Activity
(Ball Kicks/Min)

Extinction
(Ball Kicks/Min)

Infant

Baseline
(Ball Kicks/Min)

Imitation
Threshold
(Ball Kicks/Min)

TD2

4.12

2.625

TD3

0.5

0.75

0.375

TD4

1.12

2.25

TD5

1.12

2.25

TD6

0.5

0.75

1.5

2.5

TD7

4.87

10.5

11.25

TD8

2.5

3.75

0.75

3.37

3.75

2.5

TD9

0.75

0.375

TD10

4.87

4.5

4.12

2.5

TD11

7.5

11.25

4.12

0.75

0.37

TD12

0.5

0.75

0.37

0.75

TD13

2.25

3.37

1.5

3.5

Note: The third column shows the imitation threshold, i.e., the ball-kicking rate required to conclude that the infant imitated the behavior. Shaded
headings match the plot color coding. Gray boxes indicate phases during which imitation occurred.

JUNE 2019

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2019