Difference between revisions of "CNP RL"

From Pheno Wiki
Jump to: navigation, search
(Means and SDs (N=699))
(Task Structure Detail)
 
(7 intermediate revisions by the same user not shown)
Line 79: Line 79:
 
* Task Structure
 
* Task Structure
 
**    Participants performed 4 blocks, each corresponding to a different stage in the PST/PRLT tasks.  
 
**    Participants performed 4 blocks, each corresponding to a different stage in the PST/PRLT tasks.  
***  Training 1: Participants performed as many trials as required for reaching learning criteria or until 160 trials were completed (''NEED TO VERIFY IF BLOCK TERMINATES AT 160 trials'').
+
***  Training 1: Participants performed as many trials as required for reaching learning criteria or until 160 trials were completed.
 
**** Performance criteria calculation: cumulative performance accuracy was calculated once 60 trials were completed and was continuously calculated after each trial thereafter until criterion had been reached. The table below indicates how many trials were completed for each pair for each group list in the first 60 trials. <B>Learning criteria were the following: 70% accuracy on 100/0 pair, 65% accuracy on 80/20 pair, 60% accuracy on 70/30 pair, and 55% accuracy on 60/40 pair. </B>. Once criterion had been reached for a given pair, accuracy calculations were discontinued for that pair, but trial presentations continued until criterion on other pairs had been reached. Eighty trials specified in the group list were sequentially presented. If criteria were not reached at 80 trials, trial presentations began again from the beginning of the list.  
 
**** Performance criteria calculation: cumulative performance accuracy was calculated once 60 trials were completed and was continuously calculated after each trial thereafter until criterion had been reached. The table below indicates how many trials were completed for each pair for each group list in the first 60 trials. <B>Learning criteria were the following: 70% accuracy on 100/0 pair, 65% accuracy on 80/20 pair, 60% accuracy on 70/30 pair, and 55% accuracy on 60/40 pair. </B>. Once criterion had been reached for a given pair, accuracy calculations were discontinued for that pair, but trial presentations continued until criterion on other pairs had been reached. Eighty trials specified in the group list were sequentially presented. If criteria were not reached at 80 trials, trial presentations began again from the beginning of the list.  
  
Number of trials completed for each stimulus pair before criterion accuracy is first checked (criterion checking begins at the 60 trial mark)
+
Number of trials completed for each stimulus pair before criterion accuracy is first checked (criterion checking begins at the 60 trial mark). See [[RL_subject_group_list_assignment]] to determine to see the group assignment for each subject.
 
{| class="wikitable"
 
{| class="wikitable"
 
|+  
 
|+  
Line 1,587: Line 1,587:
  
 
=== Data Distributions ===
 
=== Data Distributions ===
 +
 +
<B>Training 1</B>
 +
 +
Based on sample of 1135 subjects.
 +
 +
Histogram of mean accuracy per stimulus pair. "Accuracy" refers to performance based on experienced positive/negative feedback.
 +
 +
[[File:AccuracyHist.png|500 px]]
 +
 +
Histogram of mean response times per stimulus pair.
 +
 +
[[File:RL RT Hist.png|500 px]]
 +
 +
Histogram of trials completed per stimulus pair
 +
 +
[[File:RL TrialsCompleted.png|500 px]]
 +
 +
Distribution of subjects who met training criteria (>=70% accuracy on 100/0 pair, >=65% accuracy on 80/20 pair, >=60% accuracy on 70/30 pair, and >=55% accuracy on 60/40 pair).
 +
 +
[[File:CritMet.png|500 px]]
  
 
=== Means and SDs (N=699) ===
 
=== Means and SDs (N=699) ===
Line 1,596: Line 1,616:
 
''Training 1''
 
''Training 1''
  
[[File:CNP RL Training1.png]]
+
[[File:CNP RL Training1.png | 600 px]]
  
 
''PST Probe''
 
''PST Probe''
Line 1,602: Line 1,622:
 
'''Note that the Avoid calculations are incorrect below; on list of things to correct with Stone'''
 
'''Note that the Avoid calculations are incorrect below; on list of things to correct with Stone'''
  
[[File:CNP_RL_Probe.png]]
+
[[File:CNP_RL_Probe.png | 600 px]]
  
 
''Training 2''
 
''Training 2''
  
[[File:CNP_RL_Training2.png]]
+
[[File:CNP_RL_Training2.png | 600 px]]
  
 
''PRLT''
 
''PRLT''
  
[[File:CNP_RL_Reversal.png]]
+
[[File:CNP_RL_Reversal.png | 600 px]]
  
 
PRLT: Trials till first correct  
 
PRLT: Trials till first correct  
  
[[File:CNP_RL_Rev_trials_till_corr.png]]
+
[[File:CNP_RL_Rev_trials_till_corr.png | 400 px]]
  
 
=== References ===
 
=== References ===

Latest revision as of 11:38, 30 November 2012

go back to HTAC


Basic Task Description

The CNP "RL" task contains two tasks embedded in one - a probabilistic selection task (PST) and a probabilistic reversal learning task (PRLT). These tasks both involve reinforcement learning and were designed to assess feedback sensitivity and behavioral flexibility, respectively. Both have been extensively used to determine reinforcement learning biases and behavioral flexibility in both healthy and patient populations. Initially designed by Michael Frank, the probabilistic selection task is specifically used to determine participants' tendencies to learn either from positive or negative feedback (e.g., Frank et al., 2004). The probabilistic reversal learning task, originally developed by Trevor Robbins and Robert Rogers (Lawrence et al, 1999; Swainson et al., 2000), examines participants' ability to adapt to changes in learned contingencies. Both tasks involve initial training periods in which participants must learn appropriate responses given probabilistic feedback ("noisy" feedback).

Probabilistic Selection Task

In the typical implementation of the PST (Frank et al., 2004), three pairs of cards are presented and participants must learn the "correct" card in each pair. Each pair is associated with different probabilities. For pair AB, choosing A is associated with positive feedback 80% of the time (B 20% of the time). For CD, choosing C leads to positive feedback 70% of the time, and E in EF 60% of the time. Over time, participants learn to choose the higher probability cards - choosing A, C, and E most of the time. Learning may be achieved either by choosing the card associated with positive feedback or by avoiding the card associated with negative feedback. After training, to assess whether participants learn more from positive or negative feedback, the cards are recombined in a "probe" phase, such that each card is paired with every other card. Participants are required to make a choice given these novel pairs without receiving feedback. A bias towards learning from positive feedback is determined by the number of times participants choose the highest probability card (the card receiving the most positive feedback) relative to the others (A vs. B, A vs. C, A vs. D, A vs. E), and a bias towards learning from negative feedback is derived by the number times that the lowest probability card is avoided (B vs. C, B vs. E, B vs. F). The tendency to choose A versus avoiding B is associated with several neuropsychiatric phenotypes, most notably observed in Parkinson's patients (Frank et al., 2004). When off dopamine agonist medications, PD patients are more likely to learn by avoiding negative feedback, but rely on positive feedback more when on medications. Evidence for genetic associations with this feedback sensitivity bias has also been found (Frank et al., 2007).

Probabilistic Reversal Learning Task

The PRLT, like other reversal learning tasks, includes initial learning stages (acquisition) followed by a reversal stage in which stimulus-response contingencies change and participants must re-learn new associations. For example, in a concurrent discrimination task in which participants must choose between A or B and A is initially reinforced during acquisition, participants must learn that A is no longer reinforced during the reversal stage and choose B. Although there are several variants of the PRLT, the most commonly used involves choosing the correct image in a pair of simultaneously presented images given probabilistic feedback (Swainson et al., 2000). Typically, one of the images is correct 80% of the time. After learning criterion has been reached (e.g., 8 consecutively correct responses), the probabilities of receiving positive feedback reverse such that the image that was previously correct 80% of the time is now only correct 20% of the time. The errors made after the reversal are often termed "perseverative errors". These errors are considered a measure of participants' ability to adapt to contingency changes, and they are the most commonly used indices of reversal learning performance across species. Another common measure is failure or success in reaching learning criteria during acquisition and reversal stages.

The CNP RL task was designed and created by Russ Poldrack, Robert Bilder, Marissa Geoghegan, and Naomi Kenner. It was programmed in E-Prime 2.0 (Release: 2.0.1.127) by Marissa Geoghegan.

Task Procedure

For general testing procedure, please refer to LA2K General Testing Procedure [here?].

In the LA2K "RL" task, participants perform two training sessions with the PST and PRLT following each (see Figure 1).

Training 1 -> PST Probe -> Training 2 -> PRLT
Figure 1. Stages of the CNP "RL" Experiment

Training 1: In the first training session, participants perform a probabilistic object discrimination task in which they must select one of two simultaneously presented images (abstract visual patterns presented to the left and right of each other). One image is more likely to be "correct" than the other. Four pairs of images are presented with the following respective feedback probabilities: 100/0, 80/20, 70/30, and 60/40. For example, in the 80/20 pair, one card is correct 80% of the time. Subjects are trained to criteria (70%, 65%, 60%, and 55% correct, respectively for each card pair). Accuracy is calculated based on cumulative performance. At the start of the training session, the following instructions appear on the screen and are read to the participant by the experimenter:

Training 1 Instructions: "Learning Session: In this test you will be shown sets of two images at a time. Try to choose the one that is correct. You will learn if it was correct or not after you make your choice. At first you will have to guess until you learn which image is more likely to be correct. Press the LEFT key for the LEFT image. Press the RIGHT key for the RIGHT image. The feedback will tell if your choice was correct, but the feedback is not perfectly reliable. Sometimes, your choice may be wrong even though it was correct many times in the past. Place your hand on the table with your fingers extended and resting comfortably on the LEFT and RIGHT keys."

PST probe: In the PST probe phase, the pairs are recombined such that each image is presented along with all of the other images in the training (28 in all, including the original pair, see below). Each pair is presented once without feedback. Whether subjects learn more using positive or negative feedback is determined by how often the the higher probability item in the pair is chosen (learning from pos. feedback) versus how often the lower probability item in the pair is avoided (learning from neg. feedback).

Recombined cards presented during PST probe:
100	0
100	20
100	30
100	40
100	60
100	70
100	80
80	0
80	20
80	30
80	40
80	60
80	70
70	0
70	20
70	30
70	40
70	60
60	0
60	20
60	30
60	40
40	0
40	20
40	30
30	0
30	20
20	0

The following instructions appear on the screen and are read to the participant by the experimenter:

PST Probe Instructions: "Testing Session: Again you will be shown sets of two images. As before, try to choose the image that is most likely to be correct. Press the LEFT button for the left image. Press the RIGHT button for the right image. You will not be receiving feedback during this session. Just try to pick the correct one based on what you have learned so far."

Training 2: Training 2 is identical to Training 1 except that a fixed number of trials are presented (40 trials) without requiring participants to train to criteria. Instructions for Training 2 are:

Training 2 Instructions: "Learning Session 2: Again you will be shown sets of two images. Try to to choose the image that is most likely to be correct. Press the LEFT button for the left image. Press the RIGHT button for the right image. The feedback will tell if your choice was correct, but the feedback is not perfectly reliable. Just try to pick the one that is more likely to be correct."

PRLT: The reversal phase occurs after an additional training phase (Training 2) that is identical to the first training session, using the same stimuli. During reversal, the correct image in half of the original 4 pairs is reversed. These are the 100/0 and 70/30 pairs. For these pairs, the probabilities are reversed, such that the card that was previously correct 100% of the time is never correct, and the card correct 70% of the time is now only correct 30% of the time. Each of the four pairs is presented 10 times (40 trials total). Instructions for the PRLT are similar to those for Training 2.

PRLT Instructions: "Testing Session 2: Once more you will be shown sets of two images. Try to choose the image that is most likely to be correct. Press the LEFT button for the left image. Press the RIGHT button for the right image. The feedback will tell if your choice was correct, but the feedback is not perfectly reliable. Just try to pick the one that is most likely to be correct."

Counterbalancing: In all stages of the experiment, the image pairs were presented according to a list that contained a pre-determined sequence of trials. Four groups of lists were used and counterbalanced across participants. In each group, eight images are drawn from a set of twelve. Each image was assigned to a separate probability across groups.

Task Structure Detail

  • Task Structure
    • Participants performed 4 blocks, each corresponding to a different stage in the PST/PRLT tasks.
      • Training 1: Participants performed as many trials as required for reaching learning criteria or until 160 trials were completed.
        • Performance criteria calculation: cumulative performance accuracy was calculated once 60 trials were completed and was continuously calculated after each trial thereafter until criterion had been reached. The table below indicates how many trials were completed for each pair for each group list in the first 60 trials. Learning criteria were the following: 70% accuracy on 100/0 pair, 65% accuracy on 80/20 pair, 60% accuracy on 70/30 pair, and 55% accuracy on 60/40 pair. . Once criterion had been reached for a given pair, accuracy calculations were discontinued for that pair, but trial presentations continued until criterion on other pairs had been reached. Eighty trials specified in the group list were sequentially presented. If criteria were not reached at 80 trials, trial presentations began again from the beginning of the list.

Number of trials completed for each stimulus pair before criterion accuracy is first checked (criterion checking begins at the 60 trial mark). See RL_subject_group_list_assignment to determine to see the group assignment for each subject.

Group List 100/0 80/20 70/30 60/40
1 14 15 14 14
2 12 16 16 16
3 14 16 13 17
4 12 14 16 18
      • PST Probe: 28 trials consisting of recombined pairs (including original pairs presented during Training 1); one presentation per pair
      • Training 2: 40 trials of original pairs (10x each)
      • PRLT: 40 trials of original pairs (10x each)
  • Timing:
    • All trials began with presentation of the stimuli. Participants were not under time-pressure to respond (i.e., self-paced).
    • Immediately following the participant's response, feedback was presented for 1 second.
    • An inter-stimulus interval (time between off-set of feedback and onset of the subsequent trial) of 500 ms was used during which only the grey background appeared.
  • Stimulus Characteristics
    • sensory modality: visual
    • functional modality : spatial/categorical
    • presentation modality: computer (eprime)
  • Performance Feedback Characteristics
    • sensory modality: visual
    • functional modality: verbal
    • reward (e.g., none, points, money, food): none
    • description: feedback appears as "Correct!" in green font or "Incorrect!" in red font above the image pairs.
  • Response Characteristics
    • response required: yes
    • effector modality: manual
    • functional modality: keypress
    • response options: forced choice (left/right)
    • response collection: keyboard
  • Assessment/Control Characteristics
    • Timing: self-paced,
    • Average Run Time: 10 mins

Task Schematic

Screenshot of LA2K RL: stimulus presentation. Participant responding is self-paced
Screenshot of LA2K RL: positive feedback presentation. Feedback is presented for 1 second.
Screenshot of LA2K RL: negative feedback presentation. Feedback is presented for 1 second.

After feedback presentation, a grey background was displayed for 500 milliseconds prior to the next trial.

Task Parameters Table

Task parameters table to be inserted.

Stimuli

Twelve abstract computer-generated images (ArtMatic Pro, U&I Software LLC, http://uisoftware.com; used in Sala et al., 2009; Ghahremani et al., 2010, Cohen et al., 2010) were used in the tasks (shown below). Each image was 128x128 pixels. They were centrally presented side-by-side (separated by 81 pixels) on a grey background.

Image A
Image B
Image C
Image D
Image E
Image F
Image G
Image H
Image I
Image J
Image K
Image L

In each of the four counterbalancing groups, eight images were drawn from the above set of twelve. The following shows which images were used for each group:

Group 1: C,E,F,G,H,J,K,L
Group 2: A,B,C,D,G,I,K,L
Group 3: B,C,D,E,F,H,J,K
Group 4: A,B,D,F,G,H,I,J

Performance feedback during training and PRLT stages appeared as text above the image pairs (Font: Gill Sans MT, Size: 24 point). Positive feedback was indicated by "Correct!" in green font, and negative feedback by "Incorrect." in red font.

Dependent Variables

PST Probe

The primary dependent variables for the PST Probe are:

1. The number of times subjects chose the 80% probability card in the following pairs (80/70, 80/60, 80/40, 80/30, 80/20). This is a measure of learning from positive feedback since it indicates how often subjects chose the stimulus most associated with positive feedback. (variable name: Choose_80)

2. The number of times subjects avoided the 20% probability card in the following pairs (20/80, 20/70, 20/60, 20/40, 20/30). This is a measure of how often the stimulus most often associated with negative feedback was avoided. (variable name: Avoid_20)

Of course, finer scaled measures of variables number 1 and 2 may be computed by examining how often participants choose the 70% card over 60, 40, 30, and 20, etc. or how often they choose the 30% card over the 80, 70, 60, 40, and 30.

As a control measure to assess learning of the original pairs (100/0, 80/20, 70/30, 60/40) during the training period, accuracy for the original pairs presented during the PST probe can be examined.

PRLT

The primary dependent variables for the PRLT are accuracy (proportion correct out of 10 trials) calculated separately for the 100/0 (deterministic) and 70/30 (probabilistic) trials (variable names: RL_REV_100_0R_MN and RL_REV_70_30R_MN)

Further dependent variables of interest include the number of trials required to reach criterion during Training1 and Training2. Response times compared across the card pairs can also serve as a measure of response conflict. For example, greater response conflict may be inferred when response times to 60/40 cards are longer than 80/20.

Table of all summary variables:

VARIABLE NAME STAGE Description
1 RL_TR1_100_0_MN Training 1 100/0 pair Mean Accuracy
2 RL_TR1_100_0_MD Training 1 Median RT
3 RL_TR1_100_0_STD Training 1 Stddev RT
4 RL_TR1_60_40_MN Training 1 60/40 pair Mean Accuracy
5 RL_TR1_60_40_MD Training 1 Median RT
6 RL_TR1_60_40_STD Training 1 Stddev RT
7 RL_TR1_70_30_MN Training 1 70/30 pair Mean Accuracy
8 RL_TR1_70_30_MD Training 1 Median RT
9 RL_TR1_70_30_STD Training 1 Stddev RT
10 RL_TR1_80_20_MN Training 1 80/20 pair Mean Accuracy
11 RL_TR1_80_20_MD Training 1 Median RT
12 RL_TR1_80_20_STD Training 1 Stddev RT
13 RL_PRB_100_0_MN PST Probe Accuracy - selection of highest probability card
14 RL_PRB_100_0_MD PST Probe Median RT
16 RL_PRB_100_20_MN PST Probe
17 RL_PRB_100_20_MD PST Probe
19 RL_PRB_100_30_MN PST Probe
20 RL_PRB_100_30_MD PST Probe
22 RL_PRB_100_40_MN PST Probe
23 RL_PRB_100_40_MD PST Probe
25 RL_PRB_100_60_MN PST Probe
26 RL_PRB_100_60_MD PST Probe
28 RL_PRB_100_70_MN PST Probe
29 RL_PRB_100_70_MD PST Probe
31 RL_PRB_100_80_MN PST Probe
32 RL_PRB_100_80_MD PST Probe
34 RL_PRB_20_0_MN PST Probe
35 RL_PRB_20_0_MD PST Probe
37 RL_PRB_30_0_MN PST Probe
38 RL_PRB_30_0_MD PST Probe
40 RL_PRB_30_20_MN PST Probe
41 RL_PRB_30_20_MD PST Probe
43 RL_PRB_40_0_MN PST Probe
44 RL_PRB_40_0_MD PST Probe
46 RL_PRB_40_20_MN PST Probe
47 RL_PRB_40_20_MD PST Probe
49 RL_PRB_40_30_MN PST Probe
50 RL_PRB_40_30_MD PST Probe
52 RL_PRB_60_0_MN PST Probe
53 RL_PRB_60_0_MD PST Probe
55 RL_PRB_60_20_MN PST Probe
56 RL_PRB_60_20_MD PST Probe
58 RL_PRB_60_30_MN PST Probe
59 RL_PRB_60_30_MD PST Probe
61 RL_PRB_60_40_MN PST Probe
62 RL_PRB_60_40_MD PST Probe
64 RL_PRB_70_0_MN PST Probe
65 RL_PRB_70_0_MD PST Probe
67 RL_PRB_70_20_MN PST Probe
68 RL_PRB_70_20_MD PST Probe
70 RL_PRB_70_30_MN PST Probe
71 RL_PRB_70_30_MD PST Probe
73 RL_PRB_70_40_MN PST Probe
74 RL_PRB_70_40_MD PST Probe
76 RL_PRB_70_60_MN PST Probe
77 RL_PRB_70_60_MD PST Probe
79 RL_PRB_80_0_MN PST Probe
80 RL_PRB_80_0_MD PST Probe
82 RL_PRB_80_20_MN PST Probe
83 RL_PRB_80_20_MD PST Probe
85 RL_PRB_80_30_MN PST Probe
86 RL_PRB_80_30_MD PST Probe
88 RL_PRB_80_40_MN PST Probe
89 RL_PRB_80_40_MD PST Probe
91 RL_PRB_80_60_MN PST Probe
92 RL_PRB_80_60_MD PST Probe
94 RL_PRB_80_70_MN PST Probe
95 RL_PRB_80_70_MD PST Probe
97 RL_TR2_100_0_MN Training 2 100/0 pair Mean Accuracy
98 RL_TR2_100_0_MD Training 2 Median RT
99 RL_TR2_100_0_STD Training 2 Stddev RT
100 RL_TR2_60_40_MN Training 2 60/40 pair Mean Accuracy
101 RL_TR2_60_40_MD Training 2 Median RT
102 RL_TR2_60_40_STD Training 2 Stddev RT
103 RL_TR2_70_30_MN Training 2 70/30 pair Mean Accuracy
104 RL_TR2_70_30_MD Training 2 Median RT
105 RL_TR2_70_30_STD Training 2 Stddev RT
106 RL_TR2_80_20_MN Training 2 80/20 pair Mean Accuracy
107 RL_TR2_80_20_MD Training 2 Median RT
108 RL_TR2_80_20_STD Training 2 Stddev RT
112 RL_REV_60_40_MN Reversal 60/40 pair Mean Accuracy
113 RL_REV_60_40_MD Reversal Median RT
114 RL_REV_60_40_STD Reversal Stddev RT
118 RL_REV_80_20_MN Reversal 80/20 pair Mean Accuracy
119 RL_REV_80_20_MD Reversal Median RT
120 RL_REV_80_20_STD Reversal Stddev RT
121 RL_REV_100_0R_MN Reversal 100/0 Rev pair Mean Accuracy
122 RL_REV_100_0R_MD Reversal Median RT
123 RL_REV_100_0R_STD Reversal Stddev RT
124 RL_REV_70_30R_MN Reversal 70/30 Rev pair Mean Accuracy
125 RL_REV_70_30R_MD Reversal Median RT
126 RL_REV_70_30R_STD Reversal Stddev RT
143 Choose_100 PST Probe proportion times 100 selected in the following pairs: 100/80 100/70 100/60 100/40 100/30 100/20
144 Choose_80 PST Probe proportion times 80 selected in the following pairs: 80/70 80/60 80/40 80/30 80/0
145 Choose_70 PST Probe proportion times 70 selected in the following pairs: 70/60 70/40 70/20 70/0
146 Choose_60 PST Probe proportion times 60 selected in the following pairs: 60/30 60/20 60/0
147 Choose_40 PST Probe proportion times 40 selected in the following pairs: 40/30 40/20 40/0
148 Choose_30 PST Probe proportion times 30 selected in the following pairs: 30/20 30/0
149 Choose_20 PST Probe proportion times 20 selected in the following pairs: 20/0
150 Avoid_80 PST Probe proportion times 80 avoided in 80/100
151 Avoid_70 PST Probe proportion times 70 avoided in 70/100 70/80
152 Avoid_60 PST Probe proportion times 60 avoided in 60/100 60/80 60/70
153 Avoid_40 PST Probe proportion times 40 avoided in 40/100 40/80 40/70
154 Avoid_30 PST Probe proportion times 30 avoided in 30/100 30/80 30/60 30/40
155 Avoid_20 PST Probe proportion times 20 avoided in 20/100 20/70 20/60 20/40 20/30
156 Avoid_0 PST Probe proportion times 0 avoided in 0/80 0/70 0/60 0/40 0/30 0/20
157 Choose_HI_Prob_MN PST Probe mean of choose vars above
158 Avoid_HI_Prob_MN PST Probe mean of avoid vars above

Cleaning Rules

The main cleaning rules involve eliminating trials for which response times are unrealistically low (e.g., 100 ms). Since training continues until criterion has been reached, it is assumed that subjects have adequately learned to the PST and PRLT. Since only one trial per card pair is presented during PST, it is important to determine that all responses were made for the trials containing the probabilities of interest. For example, if choose 80 and avoid 20 is of interest, it's important to make sure that all responses were made to card pairs with these values.

Code/Algorithms


History of Checking Scoring:

Jan 26, 2012: Dara met with Stone, requesting the following changes and additions:

  • Added the following variables:
    Training1:
    RL_TR1_X_Y_CRITMET (where X and Y refer to the 4 card probabilities (100/0, 80/20, etc.) - whether training criteria has been met for each of the four card pairs. Accuracy is computed for all trials for each pair and compared with the respective criteria for each pair.

    RL_TR1_X_Y_TRIALS_TILL_CRIT - number of trials required until criterion has been met. Accuracy is iteratively calculated starting from a minimum number of trials specified in the Group List table above until criteria has been met. For example, the 80/20 pair in Group List 1, 15 accuracy is calculated for trials 1 through 15; if accuracy is >= 65%, RL_TR1_X_Y_TRIALS_TILL_CRIT = 15, otherwise accuracy is recalculated for trials 1 through 16 and compared again against 65% and so on. If criteria is never met, RL_TR1_X_Y_TRIALS_TILL_CRIT =0.

    NOTE1: The experiment continued to present the pairs after criteria had been met, and especially with the low-probability pairs, accuracy often dropped after the RL_TR1_X_Y_TRIALS_TILL_CRIT window. Therefore, the RL_TR1_X_Y_CRITMET variable (based on all trials) may be zero in cases where RL_TR1_X_Y_TRIALS_TILL_CRIT is non-zero.

    NOTE2: Both of these variables are easier to interpret for the 100_0 pairs; still, variables for the probabilistic pairs are provided.

    Probe:
    Rename MEAN_AVOID_HI to MEAN_AVOID_LO (because we are calculating how often subjects avoid the low probability cards)

    Reversal:
    RL_REV_X_X_1STCORR - first correct trial during reversal phase RL_REV_X_X_SWITCHES_AFTER_1STCORR - number of switches between responses after the 1st correct response (a measure of reversal learning stability) RL_REV_X_X_MN_AFTER_1STCORR - mean accuracy after first correct trial (another measure of reversal stability)

  • Removed unnecessary summary variables with no data (RL_REV_100_0_MN, RL_REV_100_0_MD, RL_REV_100_0_STD, RL_REV_70_30_MN, RL_REV_70_30_MD, RL_REV_70_30_STD)
  • Change code to fixed error in calculation of "AvoidX" variables so that it bases them on proportion correct.


    Dara Ghahremani checked summary variables for PRLT with Stone for accuracy in fall of 2009. Worked with Stone on Probe summary variables in fall 2010.

    Version History

    In Version 2, fixed two bugs with run order 4 (one involved extra events, the other with the 'R' trials not being properly noted for 100/0 conditions)

    Data Distributions

    Training 1

    Based on sample of 1135 subjects.

    Histogram of mean accuracy per stimulus pair. "Accuracy" refers to performance based on experienced positive/negative feedback.

    AccuracyHist.png

    Histogram of mean response times per stimulus pair.

    RL RT Hist.png

    Histogram of trials completed per stimulus pair

    RL TrialsCompleted.png

    Distribution of subjects who met training criteria (>=70% accuracy on 100/0 pair, >=65% accuracy on 80/20 pair, >=60% accuracy on 70/30 pair, and >=55% accuracy on 60/40 pair).

    CritMet.png

    Means and SDs (N=699)

    NOTE TO DATA MTG GROUP: As of 5/24/11, 1002 subjects were run on the CNP RL. 303 of these subjects had no summary variables available. (Data request submitted on 6/1/11 for trial-by-trial data only included data from 403 subjects).


    Training 1

    CNP RL Training1.png

    PST Probe

    Note that the Avoid calculations are incorrect below; on list of things to correct with Stone

    CNP RL Probe.png

    Training 2

    CNP RL Training2.png

    PRLT

    CNP RL Reversal.png

    PRLT: Trials till first correct

    CNP RL Rev trials till corr.png

    References

    Cohen JR, Asarnow RF, Sabb FW, Bilder RM, Bookheimer SY, Knowlton BJ, Poldrack RA. A unique adolescent response to reward prediction errors. Nat Neurosci. 2010 Jun;13(6):669-71.

    Frank MJ, Seeberger LC, O'Reilly RC. By carrot or by stick: cognitive reinforcement learning in Parkinsonism. Science 2004;306:1940-1943

    Frank MJ, Moustafa AA, Haughey HM, Curran T, Hutchison KE. Genetic triple dissociation reveals multiple roles for dopamine in reinforcement learning. Proc Natl Acad Sci USA 2007;104:16311-16316

    Ghahremani DG, Monterosso J, Jentsch JD, Bilder RM, Poldrack RA. Neural components underlying behavioral flexibility in human reversal learning. Cereb Cortex. 2010 Aug;20(8):1843-52.

    Lawrence AD, Sahakian BJ, Rogers RD, Hodges JR, Robbins TW. Discrimination, reversal, and shift learning in Huntington's disease: mechanisms of impaired response selection. Neuropsychologia 1999;37:1359-1374.

    Sala JB, Courtney SM. Flexible working memory representation of the relationship between an object and its location as revealed by interactions with attention. Atten Percept Psychophys. 2009 Oct;71(7):1525-33.

    Swainson R, Rogers RD, Sahakian BJ, Summers BA, Polkey CE, Robbins TW. Probabilistic learning and reversal deficits in patients with Parkinson's disease or frontal or temporal lobe lesions: possible adverse effects of dopaminergic medication. Neuropsychologia 2000;38:596-612

    go back to HTAC