The Development and Initial Validation of O-WSVLT, a Meaning-Recall Online L2 Spanish Vocabulary Levels Test

ABSTRACT

Recent literature in the field of L2 vocabulary assessment has advocated for the development of written receptive vocabulary tests such as Vocabulary Levels Tests (VLTs) that use: (a) meaning-recall item formats, (b) a minimum of 40 item counts per 1,000-frequency band to improve level estimates, and (c) lemmas (not word-families) as the lexical unit . With such recommendations in mind, this study presents the development and initial validation of Online written meaning-recall Spanish Vocabulary Levels Test (O-WSVLT), the first 120-item meaning-recall vocabulary levels test that measures knowledge of the 3,000 most frequent words in Spanish. A total of 209 L1-English learners of L2-Spanish participated in the study. Focusing on internal technical qualities of the test, Rasch measurement analysis was employed to provide evidence regarding four aspects of construct validity: content (i.e. representativeness and technical quality), substantive, structural, and generalizability . Results showed that (1) the items presented adequate spread of difficulty, (2) items demonstrated high levels of unidimensionality, and (3) O-WSVLT displayed an excellent fit to the Rasch model, with Rasch person and item reliability coefficients of 0.97 and 0.99 respectively. O-WVLT fills a gap by providing L2 Spanish teachers and researchers with a reliable tool to measure students’ written receptive meaning-recall vocabulary knowledge.

RESUMEN

La literatura reciente en el campo de la evaluación del léxico en una L2 ha abogado por el desarrollo de pruebas de vocabulario que utilicen: (a) un formato basado en el recuerdo del significado, (b) un mínimo de 40 ítems por cada banda de 1,000 palabras y (c) lemas como unidad léxica (Stoeckel et al., 2021). Con estas recomendaciones en mente, este estudio presenta el desarrollo y validación de O-WSVLT, la primera prueba de vocabulario de recuerdo del significado que mide el conocimiento de las 3,000 palabras más frecuentes en español. Un total de 209 estudiantes de español como L2 ingleses participaron en el estudio. Se empleó el modelo Rasch para proporcionar evidencia sobre cuatro aspectos de la validez de constructo: contenido (representatividad y calidad técnica), sustantiva, estructural y carácter generalizable (Messick, 1995). Los resultados mostraron que los ítems presentaban una distribución adecuada de la dificultad y altos niveles de unidimensionalidad. Además, la prueba mostró un ajuste excelente al modelo Rasch, con coeficientes de fiabilidad de sujeto e ítem de 0.97 y 0.99 respectivamente. O-WVLT proporciona a los profesores e investigadores de español una herramienta fiable para medir el conocimiento léxico receptivo del recuerdo del significado.

Disclosure statement

We have no known conflict of interest to disclose.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15434303.2024.2311724.

Notes

¹ This project has undergone all the necessary ethics approvals by the Office of Research Ethics at the University of Toronto Mississauga (approval number 24,020).

² Despite integrating 500-word bands to explore the hypothesis that word difficulty increases as frequency decreases, this manuscript underscores the need for test subsections to include sufficient items for precise vocabulary knowledge estimation per frequency band. Consequently, full item statistics for 500-word bands were not included to avoid any potential misinterpretation of endorsing their individual use.