According to some prominent accounts of scientific progress, e.g. Bird's epistemic account, accepting new theories is progressive only if the theories are justified in the sense required for knowledge. This paper argues that epistemic justification requirements of this sort should be rejected because they misclassify many paradigmatic instances of scientific progress as non-progressive. In particular, scientific progress would be implausibly rare in cases where (a) scientists are aware that most or all previous theories in some domain have turned out to be false, or (b) the new theory was a result of subsuming and/or logically strengthening previous theories, or (c) scientists are aware of significant peer disagreement about which theory is correct.A second clarification concerns the term 'constitutes'. It is a truism that there are many ways for science to progress -even cognitively, i.e. with respect to its theories (representations). For example, there is surely a sense in which scientists make progress as they collect more evidence or develop new formalisms. These are all forms of progress… in a sense. But I say that we should distinguish these from the type of progress that occurred, for example, when J.J. Thomson's plum pudding model of the atom was replaced with Ernest Rutherford's nuclear model. The difference is that the former type of progress (e.g. collecting evidence) is progress because and in so far as it helps to achieve the latter type (e.g. improving atomic models).By contrast, the latter type of progress counts as progress regardless of whether it leads to some other instances of progress. To mark this distinction, I say (following Bird 2008, 280) that some improvements promote progress, whereas others constitute progress. It is the latter that accounts of scientific progress are accounts of.A final clarification. In the question above, I use the term 'improvement' instead of 'progress' to emphasize two related points about our topic. The first is that accounts of scientific progress are not attempts to analyze the term 'scientific progress' as it is used either in common parlance or scientific practice. Even if there was no such