Summary Ordinary language philosophers frequently draw on the fact that an appropriately selected sentential combination of the form p but not q can, or cannot, be uttered without absurdity; however, they do so without sufficient reflection on the methodology of such combination tests, which results in considerable shortcomings even in practical application. To improve things, I shall discuss two criteria for distinguishing ‘pragmatic’ from ‘non-pragmatic’ implications and for separating the latter into ‘linguistic’ (‘semantic’ and ‘syntactical’) and ‘non-linguistic’ ones (2–3); consider (...) the bearing of the principle of ‘sense-constancy‘ on the applicability of combination tests (4); and call attention to the important, though merely heuristic, function of investigating pragmatic, as opposed to semantic, implications (5.1–5.2). Finally, I shall hint at a striking analogy between the linguistic (yet non-empirical) method of combination tests, the phenomenological method of ‘fictional variation’, and the empirical method of ‘conjectures and refutations’ (5.3). (shrink)