Kiparsky has proposed that tense is a “stronger” category (i. e. resists levelling more) than verb agreement within the verb paradigm. In German there has been levelling of endings, stem vowels and stem consonants throughout the paradigm-sometimes across tense or mood differences, sometimes across number or person differences. This paper establishes that tense or mood has resisted levelling more than number or person in the history of German. This result is easily reconcilable with a model of grammar in which tense and mood are base categories and verb agreement in number and person is introduced transformationally. The “paradigm” must also be a real factor of linguistic change and hence must be expressible in a grammar if that grammar is to have diachronic relevance.