Structure of cognitive abilities

Evidence based on behavioral paradigms and how to measure them effectively

Zhang, Liang

2023-10-19

Introduction

“Old-fashioned” functional Magnetic Resonance Imaging

Call for Cognitive Ontology

  • Issues of identification of neural correlates of cognitive processes

  • Cognitive ontology underlies structure-function mapping

    • The structure of the mind that specifies the component operations that comprise mental function (Poldrack 2010)

    • A formal ontology requires lots of work because of its requirements of controlled vocabulary and framework (Poldrack and Yarkoni 2016)

    • Data-driven ontology seems promising (Eisenberg et al. 2019)

Method

We collected cognitive tests commonly used in cognitive research, with which behavioral data on thousands of people were collected, to directly examine their relationships.

Sample Description based on CAMP

  • Tasks (creative ability not included)

    • 80 distinct tasks
    • 81 indices

  • Participants

    • 1204 participants (Mean age = 20.6, SD = 1.7, range: [17.5, 31.38]; Sex: 52.0% females, 48.0% males, 0.0% other)
    • 北京大学: 97; 北京联合大学: 98; 四川师范大学: 478; 天津师范大学: 537

Exploratory Factor Analysis

How many factors?

  • The traditional way to determine number of factors is indecisive for us
    • Some tasks load on many factors
    • Some tasks have low reliability

Factors Resampling

  • Here we use a new method based on resampling
    1. Resample with replace a sample of the same size as the original data
    2. Do exploratory factor analysis with 4-10 factors (reasonable based on previous slide)
    3. Store the factor attribution
    4. Repeat the process for 100 times
    5. Across the 100 samples, calculate the probability of each pair of tasks belonging to the same factor
    6. Average the results across all 4-10 factor solutions

Factor Convergence

Average across 4-10 factors

Average across 8-10 factors

Structure based on the convergence

Confirmatory Factor Analysis

Loading

Latent Correlation

Goodness of Fit

The desired CFI and TLI is above 0.90, and RMSEA below 0.05.

hierarchical chisq df pvalue CFI TLI RMESA
none 3865.42 1196 0 0.84 0.83 0.04
bifactor 3820.95 1173 0 0.85 0.83 0.04
highorder 4372.66 1216 0 0.82 0.81 0.05

Improve Fitting

  • Three means were tried
    • Trim low loading tasks: poor (> 0.32), fair (> 0.45), good (> 0.55)
    • Keep equal number of tasks for each dimension: 3 or 4
    • Drop certain dimension

Loading Cutoff

Loading Cutoff - Best Model

Task Number

Task Number - Best Model

Trim Dimension

Trim Dimension (Continue)

Based on the best model from task number

Trim Dimension - Best Model

Trim Dimension - Best Model

Measure Efficiency: Less Time-consuming

Example: \(R^2\) with G

Example: \(R^2\) with G

Task Selection Results

Exploratory Factor Analysis on Short Version

Factor Convergence

References

Eisenberg, Ian W., Patrick G. Bissett, A. Zeynep Enkavi, Jamie Li, David P. MacKinnon, Lisa A. Marsch, and Russell A. Poldrack. 2019. “Uncovering the Structure of Self-Regulation Through Data-Driven Ontology Discovery.” Nature Communications 10 (1): 2319. https://doi.org/10.1038/s41467-019-10301-1.
Poldrack, Russell A. 2010. “Mapping Mental Function to Brain Structure: How Can Cognitive Neuroimaging Succeed?” Perspectives on Psychological Science 5 (6): 753–61. https://doi.org/10.1177/1745691610388777.
Poldrack, Russell A., and Tal Yarkoni. 2016. “From Brain Maps to Cognitive Ontologies: Informatics and the Search for Mental Structure.” Annual Review of Psychology 67 (1): 587–612. https://doi.org/10.1146/annurev-psych-122414-033729.
Price, Cathy J., and Karl J. Friston. 2005. “Functional Ontologies for Cognition: The Systematic Definition of Structure and Function.” Cognitive Neuropsychology 22 (3-4): 262–75. https://doi.org/10.1080/02643290442000095.