hard plaster (Castone; Dentsply, York, PA) was poured into the bottom of both bottom

trays. The plaster and sponges were moistened before each test run. Small 20 cm desk

fans were directed toward the vents of the test boxes so that high humidity did not cause

condensation on the glass lid or the sides of the trays.

 Flies were introduced into the trays as described previously. A small opaque

inverted cup (4 cm diameter) with a large wire loop glued to the top was placed on the

plaster in each of the two bottom trays. These cups were moved back and forth from one

end of a tray to the other with a plastic rod each time most of the ants had crawled under

a cup to hide. This procedure kept the ants trailing continuously from one end of a bottom

tray to the other so that the flies always had an opportunity to attack the ants. Smaller

workers were obtained by sieving as described above. We used 0.3 g of workers and 0.6

g of brood for each test group.

 Each test used ants from a different colony and received 10-15 female flies and an

equivalent number of males. Trials lasted about 3 h during which time we recorded the

number of females hovering in attack-mode over each species every 10 min during 1 100-

1500 EST to produce an average number of females in attack-mode. When possible, I

recorded the number of oviposition attempts in 20 sec intervals for individual flies

hovering in attack-mode over each group of ants to produce an average rate of attack per

minute per attacking female. When the test boxes were reused, workers from one species

of fire ant were not placed in a side that had been used by another species. At the end of

each trial, worker ants were transferred to small boxes and checked for pupating flies as

described in no-choice tests. The head capsules of dead workers were inspected for fly

larvae or pupae every 1-2 d for a period of 25 d so that most larvae had time to complete