Test development of any nature requires a stage for the investigation of the items designed, which must be carried out considering construct validity as the main criterion. In this article we bring an analysis of test itens designed to assess the reading ability in English as a foreign language. We divide the items into defensible and non-defensible: defensible items allow for the demonstration of the reading ability as described in the construct, whereas non-defensible items do not allow for collecting evidence for a more valid interpretation of the relevant ability. We suggest a division of the non-defensible itens into 3 categories: text-independent items, items with unintended clues for the answer, and vocabulary items, each one presenting some problems for construct validation. It is possible to conclude that items used in tests may not be adequate for the purposes they were designed. The analyses and conclusions presented here aim at collaborating with the development of tests that allow for the collection of evidence for more valid interpretations of the ability of reading in English as a foreign language.
reading; tests; validity; foreign language