This paper investigates the usefulness of a verbal protocol approach in examining the underlying construct of a doze test, i. e. the reasons that a test writer had for deleting some lexical items from a passage to construct a doze test. The informants were asked to
'think aloud' while they were doing the doze test. The observation of
the informants verbalising their thoughts revealed inadequacies in using this approach. To compensate for this inadequacy, retrospection interviews in which the informants were asked about their choices after their verbal protocols, were conducted. The analyses of the informants' think aloud and their retrospection showed that they could not verbalize all the mental processes they utilized in taking the test in their think aloud. The results, however, suggest verbal protocols as useful instruments for collecting a particular type of data which cannot be accessible in using other approaches.
)