Abstract

The purpose of this dissertation research was to examine variability in coefficient alpha estimates across the 16 Learning Transfer System Inventory (LTSI) constructs and evaluate reliability-reporting practices within the literature using a reliability generalization meta-analysis. Studies published between 1997 and 2025 were identified through database searches and a review of Academy of Human Resource Development conference proceedings. Following screening and listwise deletion, 118 studies were included in the final analyses. Unweighted mean coefficient alpha estimates were calculated, and ordinary least squares regression and commonality analyses examined the influence of instrument version, translation status, and full or partial use. Reliability estimates varied across constructs. Among the seven constructs retained for interpretation, motivation to transfer and performance coaching and feedback met or exceeded the .80 threshold, while personal capacity for transfer demonstrated the lowest mean reliability. Version was the most consistent predictor, with Version 4 generally producing higher reliability estimates than earlier versions. Translation had a significant, construct-specific effect only for personal outcomes positive, while full versus partial use had minimal influence. Reporting practices were inconsistent, and 61 studies using the LTSI did not report usable construct-level, sample-specific reliability estimates. These findings reinforce that reliability is a property of scores produced within a specific context rather than an inherent characteristic of the instrument and underscore the need for sample-specific reliability estimates, greater methodological transparency, and more consistent reporting practices in learning transfer research.

Date of publication

8-2026

Document Type

Dissertation

Language

english

Persistent identifier

http://hdl.handle.net/10950/5130

Committee members

Kim Nimon, Greg Wang, Forrest Lane

Degree

PhD in Human Resource Development

Share

COinS