Limits...
Effect of reinforcement history on hand choice in an unconstrained reaching task.

Stoloff RH, Taylor JA, Xu J, Ridderikhoff A, Ivry RB - Front Neurosci (2011)

Bottom Line: We modeled the shift in hand use using a Q-learning model of reinforcement learning.The results provided a good fit of the data and indicate that the effects of increasing and decreasing the rate of positive reinforcement are additive.These experiments emphasize the role of decision processes for effector selection, and may point to a novel approach for physical rehabilitation based on intrinsic reinforcement.

View Article: PubMed Central - PubMed

Affiliation: UCSF Joint Graduate Group in Bioengineering, University of California Berkeley Berkeley, CA, USA.

ABSTRACT
Choosing which hand to use for an action is one of the most frequent decisions people make in everyday behavior. We developed a simple reaching task in which we vary the lateral position of a target and the participant is free to reach to it with either the right or left hand. While people exhibit a strong preference to use the hand ipsilateral to the target, there is a region of uncertainty within which hand choice varies across trials. We manipulated the reinforcement rates for the two hands, either by increasing the likelihood that a reach with the non-dominant hand would successfully intersect the target or decreasing the likelihood that a reach with the dominant hand would be successful. While participants had minimal awareness of these manipulations, we observed an increase in the use of the non-dominant hand for targets presented in the region of uncertainty. We modeled the shift in hand use using a Q-learning model of reinforcement learning. The results provided a good fit of the data and indicate that the effects of increasing and decreasing the rate of positive reinforcement are additive. These experiments emphasize the role of decision processes for effector selection, and may point to a novel approach for physical rehabilitation based on intrinsic reinforcement.

No MeSH data available.


(A) A computer monitor projected stimuli onto a mirror, creating the impression that the stimuli were in the same plane as the participant's hands. The robotic device restricted movement to this plane. (B) Stimuli appeared in one of seven locations in Experiment 1 (shown here) and one of nine locations in Experiment 2. While the visible size of the targets remained constant, a staircase algorithm adjusted the radius of a virtual target region that was used to achieve a specified reward rate.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3066466&req=5

Figure 1: (A) A computer monitor projected stimuli onto a mirror, creating the impression that the stimuli were in the same plane as the participant's hands. The robotic device restricted movement to this plane. (B) Stimuli appeared in one of seven locations in Experiment 1 (shown here) and one of nine locations in Experiment 2. While the visible size of the targets remained constant, a staircase algorithm adjusted the radius of a virtual target region that was used to achieve a specified reward rate.

Mentions: The experiment was performed in a virtual environment that interfaced with a 3-D robotic manipulandum (PHANToM 1.5 System, SensAble Technologies). A mirrored projection system was used to display the visual stimuli (Figure 1). The participants’ task was to reach through a target that appeared at one of seven locations along a semicircular array. The participant held a robotic manipulandum in each hand and moved this device to reach through the target location. Movements were confined to the horizontal plane.


Effect of reinforcement history on hand choice in an unconstrained reaching task.

Stoloff RH, Taylor JA, Xu J, Ridderikhoff A, Ivry RB - Front Neurosci (2011)

(A) A computer monitor projected stimuli onto a mirror, creating the impression that the stimuli were in the same plane as the participant's hands. The robotic device restricted movement to this plane. (B) Stimuli appeared in one of seven locations in Experiment 1 (shown here) and one of nine locations in Experiment 2. While the visible size of the targets remained constant, a staircase algorithm adjusted the radius of a virtual target region that was used to achieve a specified reward rate.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3066466&req=5

Figure 1: (A) A computer monitor projected stimuli onto a mirror, creating the impression that the stimuli were in the same plane as the participant's hands. The robotic device restricted movement to this plane. (B) Stimuli appeared in one of seven locations in Experiment 1 (shown here) and one of nine locations in Experiment 2. While the visible size of the targets remained constant, a staircase algorithm adjusted the radius of a virtual target region that was used to achieve a specified reward rate.
Mentions: The experiment was performed in a virtual environment that interfaced with a 3-D robotic manipulandum (PHANToM 1.5 System, SensAble Technologies). A mirrored projection system was used to display the visual stimuli (Figure 1). The participants’ task was to reach through a target that appeared at one of seven locations along a semicircular array. The participant held a robotic manipulandum in each hand and moved this device to reach through the target location. Movements were confined to the horizontal plane.

Bottom Line: We modeled the shift in hand use using a Q-learning model of reinforcement learning.The results provided a good fit of the data and indicate that the effects of increasing and decreasing the rate of positive reinforcement are additive.These experiments emphasize the role of decision processes for effector selection, and may point to a novel approach for physical rehabilitation based on intrinsic reinforcement.

View Article: PubMed Central - PubMed

Affiliation: UCSF Joint Graduate Group in Bioengineering, University of California Berkeley Berkeley, CA, USA.

ABSTRACT
Choosing which hand to use for an action is one of the most frequent decisions people make in everyday behavior. We developed a simple reaching task in which we vary the lateral position of a target and the participant is free to reach to it with either the right or left hand. While people exhibit a strong preference to use the hand ipsilateral to the target, there is a region of uncertainty within which hand choice varies across trials. We manipulated the reinforcement rates for the two hands, either by increasing the likelihood that a reach with the non-dominant hand would successfully intersect the target or decreasing the likelihood that a reach with the dominant hand would be successful. While participants had minimal awareness of these manipulations, we observed an increase in the use of the non-dominant hand for targets presented in the region of uncertainty. We modeled the shift in hand use using a Q-learning model of reinforcement learning. The results provided a good fit of the data and indicate that the effects of increasing and decreasing the rate of positive reinforcement are additive. These experiments emphasize the role of decision processes for effector selection, and may point to a novel approach for physical rehabilitation based on intrinsic reinforcement.

No MeSH data available.