Fig 1.
Flow diagram of experiment of participant performing reach-to-touch task with virtual hand under training feedback that is prescribed as either ‘positive’ or ‘negative’.
Performance and agency are assessed with each reach-to-touch trial.
Fig 2.
Experimental set-up elements–A) Subject body and arm position at start of each trial, B) Motion capture marker cluster affixed by Velcro on back of hand, C) Subject head with Oculus virtual reality headset and noise-cancelling headphones.
Fig 3.
Example trials of message feedback provided upon touch of virtual hand on highlighted (yellow) target.
A) ‘GOOD’ message interpreted as positive feedback. B) ‘BAD’ message interpreted as negative feedback. Positive feedback training sessions initially prescribed a ‘GOOD’ for 80% of trials. Punishment feedback training sessions initially prescribed a ‘GOOD’ message for 20% of trials. All other trials provided ‘BAD’ message.
Fig 4.
Virtual Reality environment–A) Subject hand and virtual hand initial position with virtual timer before subject’s movement, B) Subject hand and virtual hand reaching for the targeting circle against speed tracker, C) Positive/Negative visual feedback provided with virtual hand make contact with target circle.
Fig 5.
Experimental protocol included five major testing blocks: (1) Screen participants for sufficient ability to estimate time-intervals, (2) Allow subjects to practice virtual reaching and calibrate internal reference of 1 second time-interval, (3) Establish baseline performance for virtual reach-to-touch task, (4) Provide training that conditions subject to either reward or punishment feedback, (5) Observe short-term retention effects of positive or negative conditioning on performance of reach-to-touch for subsequent comparison to baseline.
Fig 6.
Typical performance and agency observed during reach-to-touch task.
TOP: Mean accumulation of path length over time for reaching trajectory during baseline. Path length accumulation plotted against trial time (0 to 8 seconds). All pathlength accumulation data normalized (divided) by the minimum ‘final’ pathlength for each respective trial. This minimum ‘final’ pathlength is the linear distance between initial hand position and center of target for that trial. BOTTOM: Example slope fits for agency and performance data during 75-trial training session for one subject receiving disproportionate positive feedback.
Fig 7.
Mean ‘GOOD’ message rate metrics shown for disproportionate positive versus negative feedback training sessions.
LEFT—Net rate of ‘GOOD’ message during disproportionate positive versus negative feedback. Participants initially prescribed to receive ‘GOOD’ message 80% and 20% of trials during disproportionate positive and negative feedback, respectively. Deviations from initially expected rates due to select trial overrides generated by outlier performance. RIGHT–The net override rate of ‘GOOD’ message trials converted to ‘BAD’ message due to outlier poor performance shown for both feedback groups.
Table 1.
Comparing positive feedback rate, including override, during positive feedback (PF) and negative feedback (NF) training.
Fig 8.
Agency measured as compression in perceived time-interval between action (target touch) and proceeding sensory consequence (auditory beep).
LEFT–Mean agency prior to training (baseline), MIDDLE–Per trial change in agency during training with positive versus negative feedback, RIGHT–Mean change in agency from baseline following training with positive versus negative.
Table 2.
A. Training slope in agency and performance metrics for positive feedback (PF) versus negative feedback (NF) training.
B. Mean change in agency and performance metrics from baseline after PF versus NF training.
Fig 9.
Performance metric of path length in hand trajectory measured during reach-to-touch task.
LEFT–Mean path length prior to training (baseline), MIDDLE–Per trial change in path length during training with positive versus negative feedback, RIGHT–Mean change in path length from baseline following training with positive versus negative feedback. all y-axis data in ‘cm’ after normalized path length data re-multiplied by 30 cm, the minimum path length to the center target.
Fig 10.
Performance metric of mean time to contact touch a target with the virtual hand during reach-to-touch task.
The target touch time was 3000 msec. LEFT–Mean contact timing prior to training (baseline), MIDDLE–Per trial change in contact timing during training with reward versus punishment, RIGHT–Mean change in contact timing from baseline following training with positive versus negative feedback.
Fig 11.
Performance metric of contact accuracy measured as distance error between location of contact touch of hand to center of target during reach-to-touch task.
LEFT–Mean contact error prior to training (baseline), MIDDLE–Per trial change in contact error during training with positive versus negative feedback, RIGHT–Mean change in contact error from baseline following training with positive versus negative feedback.
Fig 12.
Average change in agency after either reward or punishment training is plotted against scores from MHLC (multi-dimensional) health locus of control survey for internality (I), chance (C), and power of others (P).
Table 3.
MHLC scores and correlation to post-training change in agency for positive feedback (PF) and negative feedback (NF).