Difference between revisions of "Penn Electrophysiology of Encoding and Retrieval Study"

From Computational Memory Lab
Jump to: navigation, search
(Full Methods)
Line 10: Line 10:
  
 
The following publications draw on the PEERS dataset:
 
The following publications draw on the PEERS dataset:
 +
* Aka, A., Phan, T., and Kahana, M. J. (2021). Predicting recall of words and lists. Journal of Experimental Psychology: Learning, Memory, and Cognition [[Publications#AkaEtal21|(more)]]
 +
* Katerman, B. S., Li, Y., Pazdera, J. K., Keane, C., and Kahana, M. J. (2021). EEG biomarkers of free recall. NeuroImage [[Publications#KateEtal21|(more)]]
 +
* Weidemann, C. T. and Kahana, M. J. (2021). Neural measures of subsequent memory reflect endogenous variability in cognitive function. Journal of Experimental Psychology: Learning, Memory, and Cognition [[Publications#WeidKaha21|(more)]]
 +
* Kahana, M. J., Aggarwal, E. V., and Phan, T. D. (2018). The variability puzzle in human memory. Journal of Experimental Psychology: Learning, Memory, and Cognition [[Publications#KahaEtal18|(more)]]
 +
* Kuhn, J. R., Lohnas, L. J., and Kahana, M. J. (2018). A spacing account of negative recency in final free recall. Journal of Experimental Psychology: Learning, Memory, and Cognition [[Publications#KuhnEtal18|(more)]]
 +
* Long, N. M. and Kahana, M. J. (2017). Modulation of task demands suggests that semantic processing interferes with the formation of episodic associations. Journal of Experimental Psychology: Learning, Memory, and Cognition [[Publications#LongKaha17|(more)]]
 
* Healey, M. K. and Kahana, M. J. (2015) A four–component model of age–related memory change. Psychological Review. [[Publications#HealKaha15|(more)]]
 
* Healey, M. K. and Kahana, M. J. (2015) A four–component model of age–related memory change. Psychological Review. [[Publications#HealKaha15|(more)]]
 
* Healey, M. K., Crutchley, P., and Kahana, M. J. (2014). Individual differences in memory search and their relation to intelligence. Journal of Experimental Psychology: General, 143(4), 1553–1569  [[Publications#HealEtal14|(more)]]
 
* Healey, M. K., Crutchley, P., and Kahana, M. J. (2014). Individual differences in memory search and their relation to intelligence. Journal of Experimental Psychology: General, 143(4), 1553–1569  [[Publications#HealEtal14|(more)]]

Revision as of 18:44, 19 January 2022

The Penn Electrophysiology of Encoding and Retrieval Study (PEERS) is a series of studies carried out between 2010 and 2020 to elucidate the EEG correlates of memory processes with a particular focus on retrieval dynamics in free recall. In conducting PEERS we set out to do an experiment in big science where many subjects (over 150 in PEERS1-3 and 98 in PEERS4) each contributed a large quantity of data (e.g., 6-8 sessions each in PEERS1-3, and 24 sessions in PEERS4) allowing for detailed analysis of individual subject behavior and physiology. The entirety of the PEERS dataset is available here. Please see Data Archive for datasets used in individual publications.


The following publications draw on the PEERS dataset:

  • Aka, A., Phan, T., and Kahana, M. J. (2021). Predicting recall of words and lists. Journal of Experimental Psychology: Learning, Memory, and Cognition (more)
  • Katerman, B. S., Li, Y., Pazdera, J. K., Keane, C., and Kahana, M. J. (2021). EEG biomarkers of free recall. NeuroImage (more)
  • Weidemann, C. T. and Kahana, M. J. (2021). Neural measures of subsequent memory reflect endogenous variability in cognitive function. Journal of Experimental Psychology: Learning, Memory, and Cognition (more)
  • Kahana, M. J., Aggarwal, E. V., and Phan, T. D. (2018). The variability puzzle in human memory. Journal of Experimental Psychology: Learning, Memory, and Cognition (more)
  • Kuhn, J. R., Lohnas, L. J., and Kahana, M. J. (2018). A spacing account of negative recency in final free recall. Journal of Experimental Psychology: Learning, Memory, and Cognition (more)
  • Long, N. M. and Kahana, M. J. (2017). Modulation of task demands suggests that semantic processing interferes with the formation of episodic associations. Journal of Experimental Psychology: Learning, Memory, and Cognition (more)
  • Healey, M. K. and Kahana, M. J. (2015) A four–component model of age–related memory change. Psychological Review. (more)
  • Healey, M. K., Crutchley, P., and Kahana, M. J. (2014). Individual differences in memory search and their relation to intelligence. Journal of Experimental Psychology: General, 143(4), 1553–1569 (more)
  • Healey, M. K. and Kahana, M. J. (2014). Is memory search governed by universal principles or idiosyncratic strategies? Journal of Experimental Psychology: General, 143, 575–596 (more)
  • Lohnas, L. J. and Kahana, M. J. (2014a). Compound cuing in free recall. Journal of Experimental Psychology: Learning, Memory and Cogntion, 40(1), 12-24 (more)
  • Lohnas, L. J. and Kahana, M. J. (2013). Parametric effects of word frequency effect in memory for mixed frequency lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39(6), 1943–1946. (more)
  • Miller, J. F., Kahana, M. J., and Weidemann, C. T. (2012). Recall termination in free recall. Memory & Cognition, 40(4), 540–550. (more)

(List is current as of Oct 2015. See Publications for all lab publications.)


Detailed PEERS 1-3

Introduction

This page provides a comprehensive overlook of the experimental methods for the PEERS (AKA ltpFR) experiment. This page is intended for reference by lab members, for a formal writeup of the methods suitable for a journal article, checkout the SVN project: /home/svn/experiments/ltpFR.

ltpFR is a 20 session experiment using scalp EEG technology, with 2 additional behavioral sessions (WAIS and WMS-CVLT). Participants are screened for ltpFR using the pyFR paradigm in a one session preliminary study. There is a free recall section of the experiment and a recognition section of the experiment. View a diagram of the session

Python code for running the experiment can be found here.

For information about how to analyze the LTP data, please see the Scalp analysis page.

Each of the 20 sessions includes several different tasks: track ball, free recall, final free recall, recognition.

Trackball

Each session begins with the trackball task, which is intended to provide information on the EEG signals characteristic of different types of eye movements. Track ball proceeds as follows:

1. Asterisks are drawn at the center of the screen and at 12, 3, 6 and 9 o'clock positions.

2. The pre-recorded words "close" and "open" are played in sequence 5 times. The duration of the "close" and "open" clips are approximately 1480 and 1400 ms, respectively. Each word is followed by a delay interval of 5000 ms + uniformly distributed jitter drawn from the interval [0,300] ms.

3. The pre-recorded words "left" and "right" are played in sequence 5 times. The duration of the "left" and "right" clips are approximately 990 and 920 ms, respectively. Each word is followed by a delay interval of 2000 ms + uniformly distributed jitter drawn from the interval [0,300] ms.

4. The asterisks at the edge of the screen are hidden, and the asterisk at the center of the screen is caused to trace a curve defined by a B-spline fit to 10 randomly-selected screen coordinates. This is followed by a delay of 5000 ms + uniformly distributed jitter drawn from the interval [0,300] ms.

5. The asterisks described in 1. are displayed again.

6. The pre-recorded words "up" and "down" are played in sequence 5 times. The duration of the "up" and "down" clips are approximately 1060 and 1050 ms, respectively. This is followed by a delay of 5000 ms + uniformly distributed jitter drawn from the interval [0,300] ms.

7. Another curve-tracing stimulus is shown as in 4, followed by a delay with the same parameters.

8. The pre-recorded word "blink" is played 5 times. The duration of the clip is 695 ms, approximately. This is followed by a delay of 5000 ms + uniformly distributed jitter drawn from the interval [0,300] ms.

9. The close/open task described in 2. is repeated, with the same delay parameters.

Free Recall

Across all sessions, participants study lists of words and recall them. On a subset of lists, participants must perform a size or animacy judgment task. The task type is denoted by color, purple or green, and font (assigned randomly for each participant). Participants must decide if the word presented is an object that is SMALL/BIG (can/cannot fit in a shoebox) or LIVING/NONLIVING. The participant is told in advance which color represents which judgment, and this assignment is consistent throughout all 20 of their sessions. For each word, they press one of four labeled buttons on the keyboard to make the appropriate judgment. There are no-task/control lists during which there is no judgment task, task-lists, during which participants perform either a size or animacy judgment on all list items, and switch-lists, during which participants alternate between size and animacy judgments within list.

The experiment has three phases, where different sessions had different manipulations, see below. The transition between phases changed after subject LTP209.

LTP063 to LTP209

Phase 1: Task Manipulation

Session 1 (practice)

Quantity per session Task Order in session
4 NT Lists 1-4
4 Size Either 5-8 or 9-12
4 Anim Either 5-8 or 9-12
4 Switch Lists 13-16

Sessions 2-7

Quantity per session Task Order in session
4 NT Random in lists 3-16
4 Size Random in lists 3-16
4 Anim Random in lists 3-16
4 Switch Lists 1-2. remainder random in 3-16

Phase 2: Distractor Manipulation

In Phase 2, arithmetic is introduced to the experiment in addition to everything the participant was performing in Phase 1. The arithmetic consists of adding three numbers 1 through 9, inclusive (A + B + C). It is ALWAYs three numbers and always addition. Participants are asked to answer as many math problems correctly as they can (i.e. they should balance speed and accuracy). There are 3 different math lists: inter-item distractor lists, in which the participant does a block of arithmetic before each word in the list; end-of-list distractors, in which the participant does a block of arithmetic after the list/right before the recall period; and no math lists. There is a counter on the top right of the screen, allowing the participant to see the number of correctly answered math problems throughout the session. The list number is reduced during this phase because the math adds additional time to the experiment. Take note: In this phase, there are no control (NT) lists, only 4 switching lists, and 8 single judgement lists (4 BIG/SMALL lists and 4 LIVING/NON-LIVING lists) for a total of 12 lists per session. Numbers in each cell of the table represent the number of seconds that each distractor type was presented.

Session 8 (practice) Task: 7 size lists, 7 animacy lists, randomly (not counterbalanced across distractor tasks)

Quantity per session Inter-item distractor time (s) End of list distractor time (s) Order in session
2 0 0 Never list 1, otherwise random
4 0 8 Random
2 0 16 Random
4 8 8 Random
2 16 16 Random

Sessions 9-14 Task: 6 switch, 3 size, 3 animacy, randomly (not counterbalanced)

Quantity per session Inter-item distractor time (s) End of list distractor time (s) Order in session
2 0 0 Never list 1, otherwise random
3 0 8 Random
2 0 16 Random
3 8 8 Random
2 16 16 Random

Phase 3: Externalized FR

In Phase 3, participants stop doing arithmetic and return to only doing the judgement task. In the original Phase 3, participants were asked to recall any words that are salient to them in addition to those that they remember seeing from the list. It is important to note that the participants are not doing a stream of consciousness recall in this task, but that they were instructed to recall only words that are particularly salient.

Sessions 15-20

Quantity per session Task Order in session
4 NT Random in lists 3-16
4 Size Random in lists 3-16
4 Anim Random in lists 3-16
4 Switch Lists 1-2. remainder random in 3-16

LTP210 to LTP293

Phase 1: Task Manipulation (taskFR)

Session 1 (practice)

Quantity per session Task Order in session
4 NT Lists 1-4
4 Size Either 5-8 or 9-12
4 Anim Either 5-8 or 9-12
4 Switch Lists 13-16

Sessions 2-7

Quantity per session Task Order in session
4 NT Random in lists 3-16
4 Size Random in lists 3-16
4 Anim Random in lists 3-16
4 Switch Lists 1-2. remainder random in 3-16

Phase 2: Distractor Manipulation (CDTFR)

Session 8 (practice) Task: 7 size lists, 7 animacy lists, randomly (not counterbalanced across distractor tasks)

Quantity per session Inter-item distractor time (s) End of list distractor time (s) Order in session
2 0 0 Never list 1, otherwise random
4 0 8 Random
2 0 16 Random
4 8 8 Random
2 16 16 Random

Sessions 9-16 Task: 6 switch, 3 size, 3 animacy, randomly (not counterbalanced)

Quantity per session Inter-item distractor time (s) End of list distractor time (s) Order in session
2 0 0 Never list 1, otherwise random
3 0 8 Random
2 0 16 Random
3 8 8 Random
2 16 16 Random

Phase 3: Control for EFR -- identical to Phase 1 (EFR/TaskFR2)

This Phase 3 differs from the original Phase 3 in that there is no externalized free recall. Participants are doing the exact same thing as in Phase 1 of the experiment. Sessions 17-20

Quantity per session Task Order in session
4 NT Random in lists 3-16
4 Size Random in lists 3-16
4 Anim Random in lists 3-16
4 Switch Lists 1-2. remainder random in 3-16

Recognition

This section tests the participant's recognition of words. During each trial, words are presented onscreen one at a time. For each word, participants are asked to report vocally whether or not they recognize the word from that day's session by saying "PESS" for YES and "PO" for No, and how confident they are in that response, using a scale of least confident (1) to most confident (5). Then, the participant receives immediate onscreen feedback about whether their response was correct or incorrect and how long it took them to respond. Some words in the list are targets, which are words that the participant saw in their word pool for that session. Some words are lures, which are words that were not in their session that day. During the preliminary session, each participant trains the voice recognition software to their voice saying "PESS" and "PO". This allows accurate immediate feedback; the words YES and NO are much more difficult for the voice recognition software.

Additional experimental paradigm details

Timing

Word presentation duration: 3000 ms

ISI: 800 ms with additional 400 ms jitter

Delay between end of study and recall: 1200 ms

Recall duration: 75 s

Constant rest before new list: 1500 ms

Delay before FFR begins: 5000 ms

FFR duration: 300 s

Recognition: ~25 minutes

Grey boxes in the figure above denote variable rest periods, the length of which is controlled by the participant.


LtpFR2 (PEERS4) detailed methods

This page delineates the structure of the ltpFR2 experiment. ltpFR2 is a 24 session experiment using scalp EEG technology. Initially, participants were screened for ltpFR2 using the ltpFR paradigm. Only those participants who had completed ltpFR could be recruited for ltpFR2. Completion of ltpFR served as a built-in screening process to determine participant reliability and desirability, and signified their prior completion of the behavioral studies (WAIS and WMS-CVLT). Presently, we screen participants for ltpFR2 using a preliminary session of pyFR. There is a free recall section of the ltpFR2 experiment, as well as distractor tasks between word lists.

See here of a list of 98 subjects who completed all 24 sessions.

Free Recall

Throughout all sessions, participants view lists of words and recall them. Additionally, there is a distractor period where participants are asked to perform arithmetic equations (A+B+C). The distractor can take place during two times of the experiment: 1) before list 2) end-of-list. Both distractor periods are 24 seconds in duration. The before list distractor is presented 50% of the time, while the end-of-list distractor is always presented.

  • For relevant documents regarding ltpFR2 testing and collection go here
  • Experiment code including word pools and WAS similarity matrix can be found in /home/svn/experiments/ltpFR2/trunk
  • Data is located in /data/eeg/scalp/ltp/ltpFR2 and /protocols/ltp/subjects/<subject_code>/experiments/ltpFR2

Methods

ltpFR2 Timeline

Summary

  1. 24 sessions
    • 24 lists per session
    • Sessions 1-23 use same 576 words for all participants
    • Session 24 uses 288 old words and 288 novel (i.e. not used in sessions 1-23) words, and has three list types, evenly split:
      • 8 lists composed of 8 old/16 new words
      • 8 lists composed of 12 old/12 new words
      • 8 lists composed of 16 old/8 new words
  2. List Construction
    • WAS semantic similarity scores were calculated for all possible word pairs in the full PEERS word pool of 1638 words
    • Word pairs were divided into 4 linearly spaced semantic similarity bins (low, medium-low, medium-high, high similarity)
    • 24 words per list
      • 16 semantically controlled words
        • 2 pairs of words drawn from each similarity bin
        • 1 pair is presented contiguously, 1 pair is presented with at least 2 intervening items
      • 8 random, unpaired words
  3. Distractors:
    • a) All A+B+C arithmetic
    • b) Pre-list distractor: 24 seconds, present on ½ of lists
    • c) Post-list distractor: 24 seconds, present on all lists
  4. Stimulus duration + ISI w/ jitter
    • 1600 ms stimulus duration
    • 800-1200 ISI (800 fixed plus 0-400 jitter)

Methods

Participants performed a delayed free recall experiment consisting of 24 experimental sessions. Each session consisted of 24 trials, with each trial containing a list of 24 words, presented one at a time on a computer screen. A random half of the lists (excluding the first list) were preceded by a 24-second, distractor-filled delay, and all lists were followed by a 24-second distractor period. A free recall test followed the post-list distractor on each list. Lists 8 and 16 were followed by short breaks, during which experimenters entered the testing room to check on participants and adjust EEG electrodes as necessary.

Each trial began with a 10-second countdown, which was displayed onscreen. Participants were permitted to pause and resume this countdown at any time by pressing a key. After the countdown was complete, a fixation cross appeared on the screen for 1500 ms. For trials without a pre-list distractor, the fixation cross was immediately followed by the presentation of the first word. For trials with a pre-list distractor, this fixation cross was instead followed by a 24-second distractor period. After the distractor period, the screen went blank for a jittered 800--1200 ms (uniformly distributed), after which the first word was presented. Each word was presented on the screen in white text on a black background for 1600 ms, and was followed by a jittered inter-stimulus interval of 800--1200 ms (uniformly distributed). Following the inter-stimulus interval after the final word in each list, participants performed a distractor task for 24 seconds. This post-list distractor task was followed by a 1200--1400 ms (uniformly distributed) delay, after which a tone sounded and a row of asterisks appeared onscreen for 500 ms, indicating the start of the free recall period. Participants were given 75 seconds to recall aloud as many of the words from the current list as possible, in any order. A fixation cross was displayed onscreen for the duration of the recall period. Once 75 seconds had elapsed, the recall period ended and a blank screen was displayed for 1000 ms, after which the 10-second countdown for the next list began.

Both the pre-list and post-list distractor tasks consisted of answering math problems of the form A+B+C=?, where A, B, and C were positive, single-digit integers. Math problems were displayed onscreen one at a time in white text on a black background, and participants were instructed to type the answer to each equation as quickly and accurately as possible. New problems would continue to appear until the full 24 seconds had elapsed, at which point the final problem was immediately removed from the screen. Participants were given a monetary bonus based on the speed and accuracy of their responses.

Each session required 24*24=576 words. The word pool for this experiment consisted of a 576-word subset of the 1638-word pool used in ltpFR. This same 576-word pool was used to generate the lists for sessions 1 through 23. That is, participants saw the same 576 words in each of their first 23 sessions, but the ordering of these words was randomized for each session. The 24th session introduced a set of novel words, which were drawn from the remaining words in the larger 1638-word pool. Specifically, the 24th session included eight lists composed of 8 old words and 16 new words, eight lists composed of 12 old words and 12 new words, and eight lists composed of 16 old words and 8 new words. For all 24 sessions, lists were constructed such that varying degrees of semantic relatedness occurred at both adjacent and distant serial positions. Semantic relatedness between all words in the full word pool was determined using the Word Association Space (WAS) model described by SteyEtal04. WAS similarity values were used to group words into four similarity bins (high similarity: WAS score > 0.7; medium--high similarity: 0.4 < WAS score < 0.7; medium-low similarity: 0.14 < WAS score < 0.4; low similarity: WAS score < 0.14). Two pairs of words from each of the four bins were arranged such that the words of one pair occurred at adjacent serial positions and the words of the other pair were separated by at least two other items. These eight word pairs comprised 16 of the 24 words in each list. The remaining eight words in each list were selected and positioned at random.

VFFR (PEERS5)

This page outlines the structure and methodology of VFFR, a 10-session externalized free recall experiment. VFFR is also known as ltpFR2.5 or ltpFR2b (see prior FR studies ltpFR and ltpFR2). A number of participants who had completed ltpFR2 were recruited for VFFR.


Summary

  1. 10 sessions
    • 1 practice block x 10 trials + 24 blocks x 24 trials per session
    • Last 5 sessions: Externalized initial free recall period (10 minutes) prior to regular trials (recall instructions and bonus information initially given right before the 6th session begins)
    • Each session uses the same 576 ltpFR2 wordpool in randomized order
  2. Stimulus duration + ISI w/jitter
    • 1200-1800 ms stimulus duration
    • 1000-1600 ms jittered ISI

Full Methods

Participants completed 10 experimental sessions. Each session consisted of 1 block of 10 practice trials, followed by 24 blocks of 24 trials each. Each block began with a 10-second countdown. After the countdown was complete, the first trial of the block began. On each trial, a black screen was shown for a jittered 1000-1600 ms (uniformly distributed), after which a single word appeared onscreen in white text for 1200-1800 ms (uniformly distributed). Following presentation, the screen went blank again and participants were instructed to pause briefly, and then vocalize the word they had just seen. If they began speaking within 1 second of the word leaving the screen, the message “Too fast.” appeared on the screen in red text. Participants were instructed to minimize the number of trials on which this occurred. After the participant finished speaking, a tone sounded, marking the end of the current trial. Speech was detected using a volume amplitude threshold. In addition to the 10-second countdown between blocks, two 2-minute mid-session breaks were administered after block 8 and block 16.

In the last five sessions, the 1 practice block and 24 experimental blocks were preceded by a 10-minute initial externalized free recall period. Participants were instructed to recall as many words as possible from the previous sessions in any order, while also vocalizing any additional words that come to mind in their attempt to recall these items.