Code or (not code): separating formal and natural language in CS education

This paper argues that the "institutionalised understanding" of pseudo-code as a blend of formal and natural languages makes it an unsuitable choice for national assessment where the intention is to test program comprehension skills. It permits question-setters to inadvertently introduce a level of ambiguity and consequent confusion. This is not in keeping with either good assessment practice or an argument developed in the paper that CS education should be clearly fostering the skills needed for understanding formal, as distinct from natural, languages. The argument is backed up by an analysis of 49 questions drawn from the national school CS examinations of a single country, spanning a period of six years and two phases -- the first in which no formal pseudo-code was defined, the second in which a formal reference language, referred to as a "formally-defined pseudo-code", was provided for teachers and exam setters. The analysis demonstrates that in both phases, incorrect, confusing or ambiguous code was presented in questions. The paper concludes by recommending that the term reference language should be used in place of pseudo-code, and an appropriate formally-defined language specified, in national exam settings where a common language of assessment is required. This change of terms emphasises the characteristics required of a language to be used for assessment of program comprehension. The reference language used in the study is outlined. It was designed by the authors for human readability and also to make absolutely explicit the demarcation between formal and informal language, in such a way that automated checking can be carried out on programs written in the language. Formal specifications and a checker for the language are available.