A hypothesis on change in prey size selectivity in relation to illumination level was tested on the basis of data on weight and size composition of the content of the digestive tract of larvae of perch Perca fluviatilis and on zooplankton in the layer 0–6 m (Wallersee Lake, Austria). Larvae foraging in the twilight-night period had almost two times more food in the intestine than those foraging in the daytime. The size composition of perch larvae and concentration and size composition of zooplankton hardly differed in the daytime and twilight-night samples. For the first time, it is shown on field material that more intensive feeding of larvae at twilight is related to selection of significantly larger prey than in daytime feeding. In the day-time, the larvae consumed more prey but their maximum size did not exceed 0.6 of the diameter of mouth opening of the fish; at twilight it was over 0.8. In case of feeding on so large prey, not only the weight of the consumed feed increases but the time used for capture and swallowing also considerably increases. The larvae which in the period of investigation did not yet form schools, which perform the principal defensive function, were especially vulnerable for predators feeding on relatively large prey. The decrease of the part of small-sized prey at twilight is not related to their lesser availability due to a low illumination. It is assumed that feeding on energetically more valuable but less available prey is shifted to the period of low illumination when the larvae are less exposed to predation risk. The obtained results are discussed from positions of the triotroph concept (Manteifel’, 1961).