Neurocognitive Psychometrics of Individual Differences in
Working Memory Processes
Kolloqium WS 23/24
Jan Göttmann
Project
Neurocognitive Psychometrics of interindividual differences in working memory
- Development of tasks which enable the estimation of the Memory Measurement Model (M3; Oberauer & Lewandowsky, 2019) and are suitable for EEG measurements
- Mutual validation of model parameters of the M3 model and neurocognitive correlates of working memory processes
Experiments:
Memory Measurement Model
Implementation Ressource
(MEMOIR)
MEMOIR
Memoir Toolbox
- Flexible toolbox of experiments tailored for the Memory Measurement Model (M3)
- Features Verbal, Visual and Numerical domain
- Customizable for different research enviroments (behavioral and electrophysiological)
- Customizable for different samples (e.g. general population vs. student sample)
- Extensive Tuning on several student samples
MEMOIR
Memoir Toolbox
- Random vs. Sequential recall possible
- 25 % filler trials with varying Set Size to discourage participants to predict the number of remaining Items
- Different Cue Types possible (pre- & post-cue)
- Different Set Sizes possible (3 to 6)
- Planned: Different Recall Types (Continoues Reproduction of colors)
Experiment: Verbal Cued Complex Span Task
?
Target Item
Distractor Item
Secondary Task: Jugde color hue (more blueish or redish?)
200 ms
1200 ms
650 ms
200 ms
Fixation + for 1000 ms before each Cue onset
Experiment: Verbal Cued Complex Span Tasks
Haus
Socke
Schlüssel
Nuss
Maus
Brettspiel
Auto
Flasche
Stuhl
Elefant
Nudel
Bonbon
Münze
Tasse
2.
The Memory Measurement Model (M3)
Memory Measurement Model (M3)
Measurement models for simple, complex, and updating working memory tasks.
Main Idea:
The probability of choosing the correct item from a set of retrieval candidates mainly depends on the strength of binding between the item and its context, such as its serial position.
Colour
Position
1
2
3
4
A(correct)=b+a+cA(otheritem)=b+aA(distractor.in.position)=b+f⋅(a+c)A(distractor.in.other.position)=b+f⋅aA(not.presented.lure)=b.
A(correct) = b + a + c \\
A(otheritem) = b + a \\
A(distractor.in.position) = b + f \cdot (a + c) \\
A(distractor.in.other.position) = b + f \cdot a \\
A(not.presented.lure) = b.\\
Memory Measurement Model
Different M3 parameters contribute to this activation, which drives retrieval
- The ability to form bindings of memory items to context information (c)
- The ability to filter task irrelevant information like distractors (f)
- The general feature activation of a memory item (a)
Experiment: Verbal Cued Complex Span Tasks
Current Study
- Synchronized alpha-activity after cue-onset is related to the filter parameter f of the M3 model
- Binding parameter c (or filter f) could be related to CRN (preparation of cognitive Ressources) or P300, or other ERP-components related to attention allocation and preparatory processes
Experiment: Verbal Cued Complex Span Tasks
Current Study
-
n = 36 (planned 60 subjects for visual and verbal domain)
- 80 Trials (360 retrievals) for each Task (25 % Filler Trials)
- Set Size 6 for Verbal Task, Set Size 4 for Visual Task
- 2 x Sessions for Verbal and Visual Cued Complex Span Task with EEG - Recording
- HMT (longform), Operation Span, and Posner Task
as secondary measures
Experiment: Verbal Cued Complex Span Tasks
EEG Recording
-
32 Electrodes 10 / 20 System
- 1000 Hz Sampling Rate
EEG PreProcessing
- automatic cleaning procedure
- high-pass filter at 1 Hz; 30 Hz low-pass filter
- noisy channel removal
- (line noise removal)
- re-reference to average reference
- artefactual IC removal
- interpolation of channels
- epoching around cue onset (-200 – 2300 ms)
- excluding trials (incorrect secondary repsonses, filler trials, EEG artifacts
Experiment: Verbal Cued Complex Span Tasks
Behavioral Results
- μACC−Verbal=.73
- μACC−Visual=.59
- Strong relationship between HMT Score and percentage correct of r = .66
- Strong Correlation of binding parameter c with HMT Score of r = .64
- Strong Correlation of filter parameter f with HMT Score of r = -.56
Experiment: Verbal Cued Complex Span Tasks
ERP Results
- Exploartive first analysis of ERP with different baselines:
- Trial Baseline: 500 - 800 ms for first after initial fixation cross
- Cue baseline: 200 ms before cue onset
- 30 Hz low pass filter
- grand avarage ERPs
- (1 Hz high pass)
Experiments: Cued Complex Span Tasks
Cue Baseline
Trial Baseline
Grand Avarage über P3, Pz & P4
Experiments: Cued Complex Span Tasks
Grand Avarage over P3, Pz, P4
315 - 410 ms
1360 - 1660 ms
1670 - 2100 ms
d = - 1.24,
p < .001
d = - .33
p < .001
d = - .56
p < .001
Experiments: Cued Complex Span Tasks
Grand Avarage over Fz
d = - .51
p < .001
508 - 765 ms
Experiments: Cued Complex Span Tasks
Parameter Correlations
CRN
- Filter parameter f is related to frontal CRN of Memory Cue Condition Activation,
r = .38 p < .05 and to Distractor Cue Condition Activation r = .42 p < .05 !
- No significant correlations of f with the effect (between conditions)
-
Binding strenght c is is related to frontal
CRN of Memory Cue Condition Activation,
r = - .30 p = .08 and to Distractor Cue Condition Activation r = - .30 p = .07 !
Experiments: Cued Complex Span Tasks
Parameter Correlations
CRN
- Filter parameter f is related to frontal CRN of Memory Cue Condition Activation,
r = .38 p < .05 and to Distractor Cue Condition Activation r = .42 p < .05 !
- No significant correlations of f with the effect (between conditions)
-
Binding strenght c is is related to frontal CRN of Memory Cue Condition Activation,
r = - .30 p = .08 and to Distractor Cue Condition Activation r = - .30 p = .07 !
Experiments: Cued Complex Span Tasks
Parameter Correlations
P300 (early)
- General activiation parameter a is related to P300 time window effect,
r = .31, p = .06
- No significant correlations of c and f for this time window
Experiments: Cued Complex Span Tasks
Parameter Correlations
P300 (Late)
- Binding strength parameter is realted to Memory Cue Condition Activation, r = - .28, p = .10
and to Distractor Cue Condition Activation r = - .33 p < .05 !
- No significant correlations for other parameters
Experiments: Cued Complex Span Tasks
Parameter Correlations
Late negativity
- Filter parameter f is realted to Memory Cue Condition Activation,
r = - .48, p < .05
and to Distractor Cue Condition Activation r = - .41 p < .05 !
- No significant correlations for other parameters
Fragen
Discussion
- Welche Baseline wäre bei relativ langen Trials (12 Stimuli) für Zeitfrequenz-Analysen sinnvoll
- Welche EKP-Komponenten wären interessant im Hinblick auf Cueing-Effekte
- Unterschiede im EKP / Zeitfrequenzverlauf für veerschiedene Fehlerkategorien (Distraktorfehler, Transpositionerrors)
Thank you for Your Attention!
github.com/jgman86
jan.goettmann@uni-mainz.de
Neurocognitive Psychometrics of Individual Differences in Working Memory Processes Kolloqium WS 23/24 Jan Göttmann