Fundamental problems in language testing concern reliability and validity of test formats. A first focus of discussion is the question whether language proficiency is a unitary competence or rather a divisible competence consisting of the 'four skills' (reading, listening, speaking, writing) and of 'abilities' (syntactic competence, lexical competence, morphological competence, phonological competence, etc.). Depending on the choice, language test formats choose for 'integrative tests' (cloze test e.g.) or skills' tests (reading, listening, speaking, writing). A second issue in the discussion concerns 'direct testing' vs. 'indirect testing'. Direct tests present a task that resembles the actual task in practical language use: 'speaking' is tested by a real speaking test format. Indirect tests present a task that differs from the real language use tas, e.g. testing 'speaking' by a written fill-in format. The discussion focuses on reliability and validity of the test (predictive validity, concurrent validity, face validity, construct validity and content validity). A critical evaluation of the discussion is presented in the article. Especially the problems of so-called communicative tests are discussed in the light of 'norm-referenced testing' vs. 'criterion-referenced testing'