A hierarchical process-dissociation model.

Area of Science:

• Cognitive Psychology
• Quantitative Psychology
• Psychometric Modeling

Background:

• Traditional process-dissociation models often aggregate data, potentially introducing biases as demonstrated by Curran and Hintzman (1995).
• Aggregation artifacts can lead to spurious support for psychological models, affecting the validity of research findings.
• This issue extends beyond process dissociation to other nonlinear models like signal detection and multinomial processing trees.

Purpose of the Study:

• To develop a hierarchical process-dissociation model that bypasses the need for data aggregation.
• To demonstrate that the critique regarding aggregation biases does not apply to the proposed hierarchical model.
• To establish hierarchical modeling as a general solution for accurately fitting various psychological processing models.

Main Methods:

• Development of a hierarchical process-dissociation model.
• Analysis of model performance without aggregating participant or item data.
• Examination of correlation patterns across participants and items to assess selective influence and dissociation.

Main Results:

• The hierarchical process-dissociation model successfully avoids aggregation biases.
• Selective influence is supported, showing a dissociation in correlation patterns: items better recollected also show higher automatic activation.
• Across participants, there is no correlation between recollection and automatic activation, highlighting the model's ability to capture nuanced relationships.

Conclusions:

• Hierarchical modeling provides an accurate and unbiased method for fitting psychological processing models.
• The developed hierarchical process-dissociation model overcomes limitations of traditional aggregation-based approaches.
• This approach offers a general solution for improving the analytical rigor of nonlinear psychological models.