Vertical distributions of adult rainbow trout (> 25 cm fork length, FL) were determined with a SIMRAD ES470 split‐beam echosounder in two 80–90 m deep lakes differing in water quality. Between November 1993 and February 1994, most trout (> 80%) were between 10 and 40 m, within or close to the thermocline. However, a small group of fish occupied colder waters, deeper than 50 m. In February, surface water temperatures > 21.0°C and hypolimnetic oxygen levels < 2.5 g m−3compressed the habitable depth range for trout in Lake Rotoiti to 12–35 m compared with 12–80 m in Lake Rotoma. Deeper‐dwelling trout inhabiting waters over 50 m in Lake Rotoiti would have been forced into shallower waters at this time. However, the vertical distribution of the remaining trout in Lake Rotoiti was not compressed. In March 1994, adult trout were still present in waters 10–40 m deep in both lakes, but many of the smaller fish had moved into shallower waters (< 10 m deep), probably because of declining water temperatures in the epilimnion and increased densities of their preferred prey. At this time, the lower depth range for trout in Lake Rotoiti was limited to 28 m by the 2.5 g m−3oxygen level and trout occupied warmer waters than they did in Lake Rotoma. When oxygen was not limiting, water temperature was the main variable determining the depth of the trout, and monthly changes in the mean depth of trout between both lakes and months were explained by a thermo‐regulatory model for trout movement.