Inside Teacher Assessment Decision-Making: From Judgement Gestalts to Assessment Pathways

Abstract

Assessment decision-making is an integral part of teacher practice. Issues related to its trustworthiness have always been a major area of concern, particularly variability and consistency of teacher judgment. While there has been extensive research on factors affecting variability, little is understood about the cognitive processes that work to improve the trustworthiness of assessment. Even in an educational system like Australia, where teacher-based assessment in mainstream classrooms is widespread, it has only been relatively recently that there have been initiatives to enhance the trustworthiness of teacher assessment of English as a second or additional language (EAL). To date, how teachers make their decisions in assessing student oral language development has not been well studied. This paper reports on the nature and dynamics of teacher decision-making as part of a larger study aimed at exploring variability of teacher-based assessment when using the oral assessment tasks and protocols developed as part of the Victorian project, Tools to Enhance Assessment Literacy for Teachers of English as an Additional Language (TEAL). Employing a mixed-method research approach, this study investigated the assessment judgements of 12 experienced NSW primary and secondary EAL teachers through survey, assessment activity, think-aloud protocols and individual follow-up interviews. The paper highlights the key role of teachers’ first impressions, or judgement Gestalts, in forming holistic appraisals shaping subsequent assessment decision-making pathways. Based on the data, a model identifying three assessment decision-making pathways is proposed which provides a new lens for understanding differences in teachers’ final assessment judgements of student oral language performances and their relative trustworthiness. Implications of the model for assessment theory and practice, teacher education, and future research are discussed.