This study set out to investigate intellectual domains as well as the use of measurement and validation methods in language assessment research and second language acquisition published in English in peer-reviewed journals. Using Scopus, we created two datasets: a dataset of core journals consisting of 1,561 articles published in four language assessment journals, and a dataset of general journals consisting of 3,175 articles on language assessment published in the top journals of SLA and applied linguistics. We applied document co-citation analysis (...) to detect thematically distinct research clusters. Next, we coded citing papers in each cluster based on an analytical framework for measurement and validation. We found that the focus of the core journals was more exclusively on reading and listening comprehension assessment, facets of speaking and writing performance such as raters and validation, as well as feedback, corpus linguistics, and washback. By contrast, the primary focus of assessment research in the general journals was on vocabulary, oral proficiency, essay writing, grammar, and reading. The secondary focus was on affective schemata, awareness, memory, language proficiency, explicit vs. implicit language knowledge, language or semantic awareness, and semantic complexity. With the exception of language proficiency, this second area of focus was absent in the core journals. It was further found that the majority of citing publications in the two datasets did not carry out inference-based validation on their instruments before using them. More research is needed to determine what motivates authors to select and investigate a topic, how thoroughly they cite past research, and what internal and external factors lead to the sustainability of a Research Topic in language assessment. (shrink)

Standard psychological scaling methods have been widely used as knowledge elicitation tools to uncover structural characteristics of a given domain. However, these methods traditionally rely on relatedness ratings from human experts, which is often time-consuming and tedious. We describe an integrated approach to knowledge elicitation and representation using Latent Semantic Analysis and Pathfinder Network Scaling techniques. The semantic structure of a subject domain can be automatically characterised from a collection of published documents in the domain. The method is illustrated with (...) an example of organising a digital library in accordance to latent semantic structures. (shrink)