# Validating a multinomial processing tree model for measuring confidence in lineups using a post-response feedback manipulation

## TL;DR

The paper introduces and validates a new model to measure confidence in eyewitness lineup responses, showing that confidence can be reliably assessed without affecting other cognitive processes.

## Contribution

The lineup confidence model extends an existing framework by incorporating confidence measurement while preserving the accuracy of other cognitive process estimates.

## Key findings

- Confidence levels were higher for responses based on detection compared to guessing or biased selection.
- Post-response feedback affected confidence without altering the underlying cognitive process parameters.
- The model successfully validated confidence measurement without compromising other process measurements.

## Abstract

Confidence in lineup responses is important in research and practice. Here we introduce the lineup confidence model, an extension of the well-validated two-high threshold eyewitness identification model. The two-high threshold eyewitness identification model serves to measure four cognitive processes underlying lineup responses: culprit-presence detection, culprit-absence detection, biased suspect selection and guessing-based selection. The lineup confidence model additionally incorporates the measurement of confidence. To validate the lineup confidence model, we conducted an experiment with a large sample size (N = 1565) using post-response feedback as a manipulation of confidence. Confidence followed a predictable and psychologically plausible pattern: responses based on detection were more likely to result in high confidence than responses based on guessing, and responses based on biased suspect selection were also more likely to result in high confidence than responses based on guessing. Importantly, post-response feedback selectively influenced confidence while leaving the parameters for culprit-presence detection, culprit-absence detection, biased suspect selection and guessing-based selection unaffected. Confidence can thus be measured with the model without compromising the measurement of the other processes specified by the model. This successful validation indicates that the lineup confidence model may be useful for examining how lineup characteristics and external factors influence confidence as a function of the processes underlying lineup responses.

Confidence plays a central role in how legal decision-makers interpret witnesses’ responses to lineups. In addition to making a lineup response (identifying someone or rejecting the lineup), witnesses are often asked to indicate how sure they are, using verbal phrases like “very sure” or numerical ratings like “90 %”. It is recommended in best-practice guidelines to collect confidence judgments immediately after all lineup outcomes, including suspect identifications, filler identifications and rejections. Expressed confidence can strongly influence the weight given to witness testimony. However, confidence can be affected by external factors such as feedback from police officers, highlighting the need to understand how such factors influence confidence. To support this research effort, we developed and validated the lineup confidence model, a formal cognitive model that builds on an existing framework for measuring the processes underlying lineup responses. This model allows examining how confidence varies depending on whether a lineup response is based on culprit-presence and culprit-absence detection, biased suspect selection or guessing. We show that confidence can be validly measured using this model without compromising the measurement of the processes underlying lineup responses. The model thus offers a new measurement tool for examining how lineup procedures and post-response influences affect confidence. This helps researchers and practitioners to better understand the factors that determine when confidence is informative and when it may be misleading.