In ‘The Trouble with Tarski’, The Philosophical Quarterly, 48 (1998), pp. 1–22, Jonathan Harrison attacks ‘Tarski-style’ truth theories for both formalized and natural languages, on the grounds that (1) truth cannot be a property of sentences; (2) if it could be, T-sentences would have to be necessary truths, which they are not; and (3) T-sentences are not necessarily true and can even can be false. I reply that (1) cannot be an objection to Tarskian truth theories, since these can be formulated in terms of whatever truth bearers might be. Thesis (2) is unjustified: Harrison's argument for it depends on an equivocation. Thesis (3) is false, since the right-hand side of a T-sentence is a meta-language translation of the object-language sentence described on the left-hand side, and this guarantees its truth.