Cloze tests and C-Tests are both tests of reduced redundancy based on the theory of general language proficiency. This paper presents the theory and shows first why cloze tests are unsatisfactory operalizations of the theory and the ways in which C-Tests are technically superior. It then reports the various investigations which have been performed in the construct validation of C-Tests and discusses their relevance to the original theory. Four hypotheses are set up relating to linerarity, parallelism, prediction of difficulty and processing strategies. The results obtained with the C-Tests support these hypotheses.