Signal detection and threshold models of source memory.

Area of Science:

• Cognitive Psychology
• Memory Research
• Decision Making

Background:

• Recent claims suggest curvilinear receiver operating characteristics (ROCs) in rating-based source memory refute threshold models.
• However, rating-based ROCs may not be sufficiently diagnostic to invalidate threshold models.
• Source memory analyses often overlook crucial factors like old-new detection and response tendencies, necessitating direct parameter estimation.

Purpose of the Study:

• To investigate the diagnostic power of rating-based ROCs in source memory research.
• To evaluate the efficacy of direct parameter estimation models in capturing response bias.
• To compare the Two High Threshold Multinomial Model and the Multivariate Signal Detection Model for source monitoring.

Main Methods:

• Conducted five source monitoring experiments with varied response bias manipulations and materials.
• Applied the Two High Threshold Multinomial Model (Bayen et al., 1996) for parameter estimation.
• Utilized the Multivariate Signal Detection Model (DeCarlo, 2003) for comparative analysis.
• Incorporated rating-based ROC analyses alongside model fitting.

Main Results:

• Findings indicate that rating-based ROCs are not definitive evidence against threshold models.
• Both the Two High Threshold Multinomial Model and the Multivariate Signal Detection Model demonstrated validity in capturing bias processes.
• These models offer a preferable approach to direct parameter estimation in source monitoring.

Conclusions:

• Threshold models remain viable in source memory research, despite claims based on ROC shapes.
• Direct estimation models effectively capture response bias, offering a more robust measurement tool.
• Theoretical integration of these models may provide a more comprehensive understanding of source monitoring.