1. External reward facilitates visual perceptual learning over a night’s sleep
Aaron V Berard, Tyler Barnes-Diana, Jose Nañez, Yuka Sasaki, Takeo Watanabe
Laboratory for Cognitive and Perceptual Learning, Department of Cognitive, Linguistic, and Psychological Sciences, Brown University
Script:
“Hi. Thank you for the introduction.”
“In this talk, I hope to demonstrate to you our findings on the interaction of sleep and reward in visual perceptual learning”

2. Visual Perceptual Learning
Visual perceptual learning (VPL) is defined as a long-term performance improvement on a visual task as a result of visual experience
Plasticity in the visual system
Both sleep and reward play significant roles in VPL
Karni et al. (1994); Mednick et al. (2003); Yotsumoto et al. (2009); Seitz et al. (2009); Arsenault et al. (2013); Kim et al. (2015)
Script:
1. “Just to begin, Visual Perceptual Learning (VPL – as I will be referring to it) is defined as a long-term performance improvement on a visual task as a result of visual experience”
2. “VPL is regarded as a manifestation of plasticity in the visual system and brain “
3. “Research has shown that both sleep and reward play significant roles in VPL”

3. Sleep and VPL
Both NREM and REM sleep may play significant roles in the consolidation of VPL
Karni et al. (1994); Stickgold (2001); Yotsumoto et al. (2009)
Even a short 90-minute nap has been shown to consolidate VPL
Mednick et al. (2003)
Script:
“In the past 20 years, sleep’s effect on VPL has been critically studied.”
“It has been found that both NREM and REM sleep play significant roles in the consolidation of VPL.”
“That is to say, without sleep after visual training, VPL is abolished and performance improvement is not seen.”
“The benefits from sleep are not specific to an overnight’s rest however, and even a short 90-minute nap during daytime has been shown to consolidate VPL.”

4. Reward and VPL
In addition to sleep, reward has been found to enhance VPL
Seitz et al. (2009); Arsenault et al. (2013); Kim et al. (2015)
Script:
“In addition to sleep’s effect, research has also shown that reward given during training has been found to enhance VPL.”
“Together these findings suggests that sleep and reward are significant factors in facilitating VPL.”

5. The Problem of Sleep and Reward
In our studies that have shown the effect of reward on VPL, subjects slept after training
Seitz, Kim, & Watanabe (2009); Kim et al., (2015)
These studies have not
clarified how sleep and
reward interact on VPL
Script:
“However, to our knowledge, the relationship between sleep and reward and how they influence VPL is largely not understood.”
“In our lab’s studies that have shown the effect of reward on VPL, all subjects have slept after rewarded training.”
“This poses a problem. In these reward studies, we cannot ignore the possibility that sleep may have influenced the primary findings.”
“It is quite possible that the effect from reward may be dependent on sleep, or that there may be entirely independent processes.”
“For example, in the study conducted by Seitz and colleagues in 2009, the experimental design consisted of daily training sessions lasting over a week long.”
“Subjects were trained on a subliminal Gabor orientation paired with water reward. The results showed improved performance on the orientation trained over several days. Here, in this graph, the red line represents the accuracy of orientation discrimination during the final testing session, whereas the blue line represents the accuracy during the initial testing session conducted before any training.”
“This study, along with others, include multiple days of training, which leaves the open question of whether sleep is influencing the performance improvement.”

6. Predictions
Prediction 1: The effects of sleep and reward on VPL are independent
Prediction 2: The effects of sleep and reward on VPL are interactive
Script:
“Thus, the purpose of the present study is to dissociate the effects of sleep and reward on VPL.”
“For the present study, we proposed two predictions.”
“First, sleep may not necessary for reward to be effective on VPL.”
“This would mean that the effects of sleep and reward on VPL should be independent and show no interaction.”
“Second, sleep may be interacting with reward on VPL.”
“This would propose the novel theory that suggests that the effect of reward on VPL training is present ONLY after a period of sleep and not during a period of wakefulness.”

7. Aim of the Present Study
In order to test which prediction is correct, we employed an overnight vs. daytime VPL comparison, with and without reward
Script:
“In order to test which of these predictions is correct, we employed an overnight vs. daytime VPL comparison, with and without reward.”

8. Design and Procedure 4 independent groups were tested (n=48 in total)
Sleep group with reward (n=12)
Sleep group without reward (n=12)
Wake group with reward (n=12)
One subject excluded due to daytime napping
Wake group without reward (n=12)
Script:
“Our design consisted of 4 independent groups with a total of 48 subjects.”
“In order to assess the effect from sleep, we trained and tested two sleep groups, where one group received reward during training and the other group did not.”
“Along with these two groups, two wake groups were trained and tested.”
“One wake group received reward during training, whereas the other did not.”
“All groups contained 12 subjects. The only exception is the wake group with reward, where one subject collected had to be excluded due to an unexpected period of daytime napping, which could have confounded the results.”
“These types of situations are important to consider, since during traditional sleep experimentation, very specific regulation must be upheld in order to ensure we are measuring the effect from sleep accurately.”

9. Training reward Test reward Training no reward Test no reward
Reward & Sleep
No Reward & Sleep
Training reward
Test reward
Training no reward
Test no reward
Sleep
Sleep
Reward & No Sleep
No Reward & No Sleep
Script:
1. “Our manipulations effectively created a 2x2 design with sleep and reward as factors impacting the process of VPL.”
2. “The left 2 boxes represent the 2 groups that received reward during VPL training and testing.”
3. “The right 2boxes represent the 2 groups that did not receive reward during VPL training and testing.”
4. The two boxes on top represent the subject groups that were allowed to sleep after training, whereas the bottom 2 boxes represent the wake groups.”
4. “Training and testing sessions lasted roughly 1 hour and were equal in design and specifications for all subjects.”
5. “The only experimental differences that took place were the presence of sleep and the presence of reward between the groups.”
6. “A typical experimental schedule for a subject in the sleep groups would consist of a training session at either 9 or 10 o’clock at night. After roughly an hour of training, the subject would be instructed to go home for a night’s rest. The subject would then return to the lab approximately 12 hours later at 9 or 10 in the morning, where they would perform one hour of testing.”
7. “Subjects in the wake groups were treated in a similar manner, except their first session would be at either 9 or 10 in the morning. After completing an hour of training, subjects would go about their day as normal, abstaining from any additional practice or periods of napping. Subjects in these wake groups would then return to the lab at 9 or 10 at night to complete their remaining testing session.”
8. “All subjects reported regular sleeping habits and had not been trained on VPL tasks in the past.”
Training reward
Test reward
Training no reward
Test no reward
Wake
Wake

10. Method Subjects were trained on a texture discrimination task (TDT)
1. “ L or T ? ”
2. “ H or V ? ”
Script:
“For our experiment, subjects were trained on a texture discrimination task, or TDT for short.”
“Each subject was instructed to sit in a dimly lit room and fixate on a computer display.”
“Stimulus presentation consisted of a field of vertical white lines with two embedded targets.”
“The first target was located at the fixation point, and displayed letters either L or T for the subject to discriminate.”
“This first target was used to ensure fixation throughout the experiment.”
“The second target was an array of oblique lines arranged either vertically or horizontally located in the lower left quadrant of the screen. All subjects were trained and tested in the same visual quadrant to account for retinotopic specificity.”
“Subjects had to report whether they saw a vertical or horizontal array in their peripheral vision.”
“Following the stimulus presentation, a brief mask was shown and then a blank screen where the subjects gave their responses to the two targets.”
“Subjects who were in the reward conditions received reward upon correct discrimination of the horizontal or vertical task although everyone in the experiment received noise feedback for correct or incorrect responses of both targets.”
“The time between the presentation of the stimulus and mask is defined as stimulus onset asynchrony (or SOA for short) and was varied in length to establish task difficulty.”
“The shorter the length of the SOA, the less time the stimulus remained in the screen, the more difficult the task became.”
Fixation
Stimulus
Blank
Mask
Response
SOA
Karni et al. (1991); Yotsumoto et al. (2009)

11. Method
Subjects in the reward groups received water as a reward upon correct discrimination response
“In accordance with our lab’s previous experiments, a water delivery system was used as a reward for the subjects upon correct task response.”
“Water was used as a reward instead of cash payment, since water is defined as a primary reward, and should thus illicit a stronger response.”
3. “The subjects in the reward conditions were deprived of food and water for 5 hours prior to the training and testing sessions to ensure that the water acted as a sufficient reward.”
Seitz et al. (2009); Kim et al. (2015)

12. TDT Threshold
Results calculated to find the SOA at the 75% performance threshold
Plotted with a logistic function
75%
Script:
“In total, 7 different SOA’s were used during the experiment, each ranging from 180ms to 60ms, as shown on the example graph to your right.”
“SOA presentation became more difficult as subjects progressed through the training. The result was something like this, where subjects performed with roughly % accuracy at the higher SOA’s, but then dropped off as they advanced through the session and approached more difficult SOA’s such as 80 or 60ms.
“In order to find the optimal performance threshold defined for the session, a logistic function was fitted to the data. The point on the curve that corresponded to the SOA in which 75% accuracy was achieved represented the subject’s optimal performance known as the performance threshold.

13. Results: Training Session
No significant group difference in threshold for the training sessions
Reward
No Reward
Threshold SOA (ms)
Script:
“Now for our results. First, we examined the training session only, to determine if there was any initial group difference in performance, given that circadian factors may have been at play due to the different time of day subjects in the wake group compared to the sleep group began their training session”
“A 2x2 ANOVA was conducted to reveal no significant difference between the groups.”
“This suggests that there was no circadian influence or initial threshold difference for the training sessions between the subjects used in our study.”

14. Results: Performance Improvement
2-way (sleep vs. no sleep) x (reward vs. no reward) ANOVA
Significant interaction between sleep and reward (p<0.05)
****
***
Reward +
No Reward
Script:
“Next, in order to assess performance improvement, we compared threshold change between the training and testing sessions by subtracting final from initial thresholds and then dividing by then initial threshold. We multiplied this value to obtain our threshold performance improvement in percentage as seen here.”
“A 2x2 ANOVA was applied to examine the effect sleep and reward on VPL.”
“We found a significant main effect of sleep and a significant main effect of reward, as well as a significant interaction between sleep and reward factors.”
“Additionally, a significant simple main effect of sleep was observed for the reward groups, along with a significant simple main effect of reward on the sleep groups.”
“A marginally significant simple main effect of sleep was also found for the no reward groups.”
‘We noted no significant difference in the wake groups between reward and no-reward conditions.”
A significant main effect of sleep (p<0.001)
A significant main effect of reward (p<0.05)
No significant difference between reward and no-reward conditions if (wake group – no sig difference between bars)
Star:
1. Main effect of sleep
2. Main effect of reward
3. Interaction,
4. post-hoc simple main effect test
5. Then show stars and *’s
----- Meeting Notes (5/19/15 17:50) -----
the
Improvement (%)
Performance
****p<0.001
***p<0.005
+p = 0.06

15. Summary of Results
There was a significant interaction between the effects of sleep and reward on VPL
Without sleep, there was no significant effect of reward
With sleep, there was a significant effect of reward
Script:
“Our results demonstrated a significant interaction between the effects of sleep and reward on VPL.”
“Post-hoc analysis indicated that without sleep, there was no significant effect of reward and that with sleep, there was a significant effect of reward.”
3. “In summary, in order for the to be an effect of reward on VPL, we found that sleep was needed after initial training.”

16. Discussion
Our results suggest that reward if effective on VPL only after subjects sleep
No significant difference in performance of the training sessions suggest that the effect of reward is not shown before sleep
There are two possibilities:
The reward system effective on VPL is triggered by sleep
The reward system effective on VPL is inhibited during wakefulness
Script:
“Given our findings, it is important to speculate on possible mechanisms that could explain what we observed.”
“Since there was no significant difference in performance during the training sessions across our groups, we are left with two possibilities explaining the performance increase across the two experimental sessions.”
“First, it is possible that the reward system effective on VPL is triggered by sleep.”
“This would suggest that mechanisms during the consolidation of the rewarded TDT are unique to the presence of reward, and that any reward processing on VPL that occurred required sleep for initiation.”
“The second possibility is that the reward system effective on VPL is inhibited during wakefulness.”
“This would suggest that sleep may not be the defining factor for reward processing of VPL, but that there is an inhibition process during wakefulness preventing the effects of reward on VPL.”
“These speculations are also supported by our wake groups, where no significant difference in performance improvement was observed between the rewarded and non-rewarded groups.”
“Together, the findings of this study warrant additional exploration of the interaction between reward and sleep on VPL. We hope to continue this line of research in order to uncover the mechanisms behind the presented behavioral evidence.”

17. Acknowledgments Laboratory for Cognitive and Perceptual Learning
NIH Support:
R01EY015980
R01MH091801
“I would like to take a moment to acknowledge my colleagues in the laboratory for cognitive and perceptual learning for all the help and support on this project.”
“I would also like to thank our NIH funding sources, as well as all of you for listening. I would be happy to take your questions at this time.”
“Thank you.”

18. Supplementary Slides
Mednick et al., 2003

19. Supplementary Slides
Abe et al., 2011

20. Supplementary Slides
Our findings suggest that sleep was necessary for VPL to occur, but not reward
Sleep  VPL
Reward  VPL
This leaves us with two possible models
Sleep  Reward  VPL
Reward  Sleep  VPL
Future directions include examining these models
Script:
“Our first two models were derived from our primary findings, where the results suggest that sleep was necessary for VPL to occur, but not reward.”
“Therefore, we are left with two possibilities.”
“First, sleep may be triggering the reward system on which VPL interacts – suggesting that the effects from reward on VPL will not exist until sleep has occurred.”
“Second, reward may be enhancing the processes during sleep that contribute to the consolidation of VPL – suggesting that there is a specialized form of sleep consolidation that takes place when the reward system is activated.”
Further studies are needed to dissociate these last two models. Future directions will include examining our findings to further understand the mechanisms at work.”
----- Meeting Notes (5/4/15 15:58) -----
get rid of this slide and instead
put it in supplemntory
keep for next experiment

21. Supplementary Slides Additional models to consider for our results
Recursive properties of sleep and reward
Sleep  Sleep/Reward  VPL
Reward  Sleep/Reward  VPL

22. Supplementary Slides VPL VPL Interaction Model Independent Model Sleep
MOVE TO LAST SLIDES
Reward
Reward