The Psychometrics of Interval Responses

Kloft, Matthias

Thesis

Open Access

The Psychometrics of Interval Responses

Files

Primary dmk.pdf (4.43 MB)

Date

2026-01-19

Authors

Kloft, Matthias

Supervisors

Person

Heck, Daniel W.

Abstract

Response formats, such as Likert-type scales or visual analog scales, which offer one response value to choose, have been widely used for psychological research and applied fields. However, in certain scenarios, a single response value may hide the underlying influences that are relevant to the response, such as uncertainty or ambivalence. In such cases, using an interval response format allows respondents to incorporate these influences by setting lower and upper bounds for a range of permissible values. In my thesis, I propose new psychometric models for continuous bounded interval responses, building upon previous research in item response theory and cultural consensus theory. These models address the statistical challenges posed by the interval response format. They also provide parameter estimates with psychologically meaningful interpretations. I further present studies examining the convergent and discriminant validity, as well as the test-retest reliability, of interval response measures. The study results suggest a high degree of convergence between measures using an interval response format, the dual-range slider, and measures using a single-valued response format, the single-range slider, when measuring the central tendency of personality or ambivalence of valence judgments. To examine discriminant validity, I investigated whether interval width measures could distinguish between different constructs when used across various applications. Overall, I found that the dimensional structure of interval widths aligns with the structure of task types. Additionally, I did not find a strong global response-style factor influencing all responses across various tasks. However, tasks of the same type share common influences on interval widths, which may indicate local response-style factors. Regarding test-retest reliability, my longitudinal study using two personality scales showed that interval location measures were reliable and comparable to single-valued measures. Interval width measures still showed moderate to good reliability. In conclusion, my thesis advances the psychometric modeling of interval responses and offers new insights into the psychometric properties of the resulting measures in applied research.