Effects of questions' repetition and variation on the efficiency of the guilty knowledge test: A reexamination

The effect of question repetition and variation on the efficiency of the Guilty Knowledge Test (GKT), based on electrodermal and respiration measures, was examined in a between-subjects experiment with 3 conditions. Each participant was presented with a sequence of 12 biographical questions. In Condition 1, a single question was repeated 12 times; in Condition 4, each of 4 different questions was repeated 3 times; and in Condition 12, 12 different questions were used. A monotonic relationship between the number of different questions used and detection efficiency was observed only with the electrodermal measure (the areas under the receiver operating characteristic curves, obtained with this measure in Conditions 1, 4, and 12 were .68, .81, and .99, respectively). These results demonstrate that a GKT based on multiple questions is superior to the use of many repetitions of a single or a few questions, and it can reach an almost perfect detection efficiency.

Between-subjects experimental design with three conditions varying question repetition (N participants unspecified in available sources), measuring skin conductance and respiration responses to biographical questions, analyzed using ROC curve methodology and signal detection theory.

Background & Context

The Concealed Information Test (CIT), also known as the Guilty Knowledge Test (GKT), has long been recognized as a scientifically sound alternative to traditional deception detection methods. Based on sound theoretical principles and proper controls, the GKT satisfies the necessary requirements of an objective test, yet a crucial practical question remained unresolved: How should test questions be structured to maximize detection efficiency?

This 2002 study emerged as a critical reexamination of earlier findings. Ben-Shakhar and Elaad conducted a constructive replication of their 1997 study and obtained different results, indicating that a GKT based on multiple questions is superior to the use of many repetitions of a single, or a few questions. The research addresses a fundamental design question facing examiners: whether to increase test length by repeating the same question multiple times or by introducing different questions about the same concealed information.

The study's significance extends beyond technical optimization. Understanding how question repetition and variation affect detection efficiency has direct implications for test construction in field settings, where the availability of crime-relevant details may be limited and examiners must make strategic decisions about resource allocation.

Research Design & Methodology

The effect of question repetition and variation on the efficiency of the Guilty Knowledge Test (GKT), based on electrodermal and respiration measures, was examined in a between-subjects experiment with 3 conditions. Rather than using a traditional mock-crime paradigm, the researchers innovatively employed biographical information as stimulus material.

Each participant was presented with a sequence of 12 biographical questions. The experimental design manipulated question variety across three between-subjects conditions:

• Condition 1: A single biographical question repeated 12 times
• Condition 4: Four different biographical questions, each repeated 3 times
• Condition 12: Twelve different biographical questions, each presented once

This design elegantly controlled for total test length (all conditions involved 12 question presentations) while systematically varying the degree of question diversity. Physiological measurements included electrodermal activity (skin conductance response) and respiration line length (RLL), standard measures in GKT research. Detection efficiency was evaluated using signal detection theory metrics, specifically the area under the Receiver Operating Characteristic (ROC) curve.

Results & Key Findings

A monotonic relationship between the number of different questions used and detection efficiency was observed only with the electrodermal measure (the areas under the receiver operating characteristic curves, obtained with this measure in Conditions 1, 4, and 12 were .68, .81, and .99, respectively). This represents a striking progression from moderate to nearly perfect detection accuracy.

Key statistical findings include:

• Single question repeated 12 times (Condition 1): ROC area = 0.68 (electrodermal)
• Four questions × 3 repetitions (Condition 4): ROC area = 0.81 (electrodermal)
• Twelve different questions (Condition 12): ROC area = 0.99 (electrodermal) — nearly perfect detection
• The respiration measure was unaffected by variations in the questions, producing a ROC area of 0.85 with RLL measure

Inspection of Tables 3 and 4 reveals that a combination of the two physiological measures (SCR and RLL) increases detection efficiency beyond that obtained with the electrodermal measure alone. The electrodermal measure showed clear sensitivity to question diversity, while respiratory responses remained robust regardless of repetition pattern.

Discussion & Significance

These results demonstrate that a GKT based on multiple questions is superior to the use of many repetitions of a single or a few questions, and it can reach an almost perfect detection efficiency. The finding of a 0.99 ROC area with 12 different questions represents detection accuracy approaching the theoretical ceiling, a remarkable achievement in psychophysiological deception detection.

The divergent pattern between electrodermal and respiratory measures is theoretically informative. A number of experimental manipulations were found to divergently affect the SCR and cardiorespiratory (RLL and HR) measures (e.g., overt deception, interfering task, question repetition), suggesting these measures tap into different underlying mechanisms. The electrodermal system appears more sensitive to novelty and stimulus variation, consistent with orienting response theory.

These findings have been influential in subsequent CIT research and practice. Some authors have argued that examiners should integrate results across multiple CIT questions in order to obtain more statistically reliable and robust results. The study provides empirical support for test protocols that maximize question diversity rather than relying heavily on repetition of limited items.

Limitations & Considerations

The study employed biographical information rather than crime-related details from a mock crime scenario. While this approach offers excellent experimental control and ensures participants possess genuine knowledge of critical items, it may not fully capture the motivational and emotional states present in actual criminal investigations. The ecological validity of biographical questions as proxies for crime-relevant knowledge warrants consideration.

The study used a between-subjects design, meaning each participant experienced only one condition. While this avoids carryover effects, it requires larger sample sizes and assumes equivalent groups across conditions. The differences in the electrodermal responses between probes and controls decreased with more item repetitions, but the overall classification accuracy increased, highlighting the complex relationship between mean response magnitude and diagnostic efficiency.

The findings specifically apply to the GKT/CIT format with its multiple-choice question structure. Generalization to other polygraph formats or deception detection paradigms should be made cautiously. Additionally, the study examined only two physiological channels; modern CIT protocols often incorporate additional measures including heart rate and finger pulse volume.

Practical Applications

For polygraph examiners and forensic practitioners, the message is clear: when constructing a GKT, prioritize question diversity over repetition. Rather than asking the same question about a single crime detail multiple times, examiners should identify as many distinct, salient crime features as possible and construct separate questions for each. This approach yields substantially higher detection accuracy, with the potential to achieve near-perfect discrimination between knowledgeable and unknowledgeable examinees.

The findings also inform decisions about test construction in resource-limited scenarios. When only a few crime details are available, examiners face a strategic choice. The data suggest that even moderate question diversity (four different questions) substantially outperforms pure repetition, achieving a respectable ROC area of 0.81. This provides practical guidance for real-world investigations where crime scene details must be carefully protected from public disclosure. Combining multiple physiological measures further enhances accuracy, suggesting that multi-channel recording should be standard practice whenever feasible.