Underwater census and single‐pass electrofishing were compared for estimating relative abundance of juvenile brown trout in the Kakanui River, South Island, NZ. Mean sampling efficiency was lower, and the variability of sampling efficiency was much greater, for underwater census (0+ trout: x = 0.38,s =0.368; 1+ trout: x = 0.62,s= 0.822) than for single‐pass electrofishing (0+ trout: x = 0.61,s= 0.143; 1+ trout: x = 0.74, s = 0.171). Sampling efficiency of both methods was dependent on temperature. Electrofishing became less efficient at higher temperatures whereas underwater census became less efficient at colder temperatures. The low, and highly variable, sampling efficiency for underwater census of 0+ brown trout was related to substrate hiding behaviour which is dependent on temperature. A ratio method for comparing relative abundance estimates is presented. Minimum significance values for the ratio (R) were derived for 0+ trout using temperature adjusted sampling efficiencies. To be statistically significant, relative abundance estimates made by underwater census had to differ by a factor of 6–7 times, whereas those made by single‐pass electrofishing had to differ only by about 2 times, depending on the number of fish counted. By confining comparisons of relative abundance estimates made by underwater census to the summer period, differences of about 3.5–4 times could be detected statistically. It was concluded that single‐pass electrofishing is superior to underwater census for estimating the relative abundance of juvenile brown trout in shallow (< 1 m) river habitat, especially when temperature varies widely as with season and time of day.