The quadratic relationship between difficulty of intelligence test items and their correlations with working memory

Area of Science:

• Cognitive Psychology
• Neuroscience
• Psychometrics

Background:

• Fluid intelligence (Gf) is vital for abstract reasoning and problem-solving.
• Gf is strongly linked to working memory (WM) effectiveness.
• A prevailing theory suggests WM storage capacity underlies the WM-Gf correlation, predicting increased correlation with higher test load.

Purpose of the Study:

• To investigate whether the correlation between working memory capacity (WMC) and fluid intelligence (Gf) changes with increasing test load.
• To critically evaluate the storage-capacity account of the WM-Gf relationship.

Main Methods:

• Analytical inference and numerical simulations were employed to model the WM-Gf correlation.
• Empirical analysis was conducted using a large sample and two WMC tasks, specifically examining the Raven's Progressive Matrices test.
• Item-wise correlations were analyzed in relation to test difficulty and item distribution (ceiling/floor effects).

Main Results:

• The study demonstrated that changes in the WM-Gf correlation with load are primarily influenced by item difficulty and ceiling/floor effects.
• A quadratic relationship was observed for the Raven's test, consistent with its item difficulty spanning from ceiling to floor.
• No significant changes in correlation were found due to varying WM/Gf load, challenging the storage-capacity hypothesis.

Conclusions:

• The observed load-dependent changes (or lack thereof) in the WM-Gf correlation do not support or refute specific theories of WM/Gf.
• The storage-capacity account of the WM-Gf relationship is not supported by the findings.
• Ceiling and floor effects significantly impact item-wise correlations, and moderate correlations are relatively robust to these effects.