For much of their pelagic larval dispersal (PLD) stage, larval perciform fishes are able to directly influence their dispersal by horizontal swimming, but it is unclear which means of measuring swimming ability is most appropriate for modelling dispersal and studying demographic and genetic connectivity. Most studies use critical speed (Ucrit), a laboratory flume measure derived by increasing flow until larvae can no longer maintain their position. Most swimming ability data on fish larvae are Ucrit, usually for larvae nearing the end of PLD. Recognizing that a forced laboratory measure is inappropriate for dispersal, researchers have used decreased Ucrit values, usually by 50%, and have argued that Ucrit is strongly correlated with more relevant swimming measures. Here I examined the suitability of Ucrit versus in situ speed (ISS), wherein speed of larvae is measured by divers following them in the ocean with a flow meter. Considerations of dispersal require inclusion of swimming ontogeny. Swimming speed regressions of speed on size of 10 species in 8 families showed that Ucrit and ISS are not well correlated. The Ucrit:standard length (SL) slope was greater than the ISS:SL slope in 6 species, and did not differ in the other 4 species. No overall metric, e.g. X% of Ucrit = ISS, was appropriate for conversion of Ucrit to ISS. Conversion of Ucrit to ISS is not straightforward. Ucrit measures swimming potential, not what larvae do in the ocean, whereas ISS directly measures larvae swimming in the ocean. Ucrit ontogeny is less variable, but ISS ontogeny is more relevant to dispersal. Ucrit may be useful for other purposes.