The field component of this study compared feeding preferences of Umbra limi with available resources in the environment. The mudminnows preferred benthic coleopterans, anisopterans, and amphipods while avoiding planktonic and other benthic food types. The laboratory component tested for possible switching behaviour by offering different ratios of two food types (Enchytraeus worms and Culex larvae). Both components indicated a benthic feeding preference in mudminnows. Although preferences were variable across individuals, switching did not occur, a result consistent with recent predictions on predatory behaviour. Feeding patterns reflected individual differences among fish, runs of feeding on one food type, and a constancy in the total number of the two food types eaten.
The population studied was located in ponds in flooded forest near the entry road to the Queen's University Biological Station 40 km northeast of Kingston on the Rideau River system . In order to assess feeding in the field, we set unbaited minnow traps between 18 June and 8 July 1982 at three sites and checked daily between 0900-1000 and 1900-2000 h. Data were gathered immediately upon removal of the fish from the trap. The total length of the fish was recorded in mm and the stomach contents from the esophagus to the pyloric sphincter were removed. Under a dissecting microscope, food items were identified to order or suborder and counted, and percent composition of total food volume by prey was estimated. Fifty-one stomachs were examined, of which 20 held appreciable (> 0.5 cc) food volumes.
Benthic food resources were estimated by taking three 15 x 15 cm Ekman dredge samples on 21 June from each site and preserving them in 10% formalin. Later, organisms were picked from the debris, sorted by size by passage through screens of decreasing mesh size, and identified under a dissecting microscope following Needham & Needham (1962). Percent composition of total resource was estimated by volume using a glass cylinder. Zooplankton food resources were estimated by taking ten Schindler trap samples on 25 June and preserving them in 5% formalin . Food items in the first sample were counted completely and identified under a dissecting microscope following Needham & Needham (1962). The remaining nine samples were subsampled by bringing the volume up to 100 ml and drawing off 20 ml of the resulting homogeneous suspension. Food items were again identified to suborder and counted, and percent composition of total resource volume was estimated using a glass cylinder.
For the laboratory study, unsexed fish between 7.2 and 8 .3 mm total length were housed on Queen's campus individually in visual isolation in glass aquaria measuring 51 x 27 x 18 cm high equipped with stones, snails, Ceratophyllum demersum, and air bubblers at 26± 1 ° C under a 12 : 12 light : dark cycle . Pilot studies of several weeks duration were used to acclimatize the fish to the laboratory and to determine suitable food types and feeding protocols. After these studies, six fish were randomly selected and fed mixtures of Enchytraeus worms and fourth instar Culex tarsalis larvae. At a fixed time of day between 1530 and 1730 h, every second day, food was offered to three of the fish (numbered 1, 2 and 3) individually for 15 min at one of the ratios of 40 :0, 30 : 10, 20 :20, 10 :30, and 0 :40 worms : larvae sequentially for at least five daily feedings (meals) at each ratio while three (4, 5 and 6) were fed in the reverse sequence of ratios. An observer who sat quietly in front of the aquaria recorded the sequence in which the worms and larvae were eaten.