Describing the relationship between the cognitive and linguistic complexity of a mathematical literacy examination and types of student errors

Much prior research has shown that if students have a poor command of the language in which they are taught and assessed, they experience a complex and deep learning disadvantage (Barton & Neville-Barton, 2003). Abedi (2006) mentions, in particular, that unnecessary linguistic complexity can threaten the validity of examination items and thus compromises the fairness of the assessment for English language learners. In Clarkson’s (1991, p. 31) research it was found that for the English language learners in the study “comprehension errors [made] up a high proportion of the errors made when…students attempt[ed] to solve mathematical word problems”. In an attempt to explore whether this was the case for a group of National Certificate (Vocational) [NC(V)] students at an FET college, the research conducted in this study focused on describing the cognitive and linguistic complexity of Level 4 Mathematical Literacy examination items as well as the types of responses from a sample of students. A mixed-methods case study design was selected. Student errors were classified as either due to mathematical literacy-related sources, or language-related sources and the question was asked as to how the cognitive and linguistic complexity of items might be related to the types of errors made. Statistically significant correlations were found between the linguistic complexity of items and language-related errors, and between the cognitive complexity of items and all types of errors. It was also possible to identify which language features, in particular, were statistically significantly correlated with linguistic complexity, namely: prepositional phrases; words of 7 letters or more and complex/compound sentences. As was expected, the majority of errors were categorised as mathematical literacy-related. However, as many as 19.22% of all errors made were identified as language-related. While the scope of the study prevents any generalisations from being made, the results indicate a need for a larger-scale study of this nature to determine if the complex and deep learning disadvantage mentioned by Barton and Neville-Barton (2003) does exist with regard to the assessment of Mathematical Literacy for NC(V) students who are English language learners (Barton & Neville-Barton, 2003)