This work was supported by the Rowland Institute for Science. I gratefully acknowledge the help of Winfield Hill for instrumentation, programming, and encouragement. I thank Dan Coutu for the MIDI music and Matt Maltzman for ongoing discussions and help with the manuscript. Thank you, Angel Peterchev, for suggesting Paganini's theme for the melody experiment. Thanks are due Chris Stokes and the Rowland Staff. Special thanks to Allen Neuringer for his inspiration and invaluable comments. Correspondence concerning this article should be addressed to A

As part of ongoing research into the ability of koi to categorize complex auditory stimuli, we have had to develop novel apparatus, The stimulus generator presents sound from two CD drives under computer control through a new underwater speaker. The operant manipulandum is a horizontal button that eliminates spurious triggering by water turbulence and problematic response topographies. The button's design has also been adapted for use by tilapia in an aquacultural food-preference study. The feeder uses a nipple to dispense food pellets reliably under water. In this paper, the apparatus is described in detail. Also discussed are methodological issues related to its design, as weIl as its usage in a pilot study in which koi learned to discriminate music from silence by using a single manipulandum with food reinforcement.We have been studying the ability of fish-specifically koi (Cyprinus carpio)-to discriminate and categorize complex auditory stimuli such as music. In this paper, we describe both the apparatus that we have developed and a preliminary experiment that demonstrates the use of the apparatus and illustrates some methodological issues. In designing the apparatus, we were guided by ethological as weil as practical considerations.In designing our aquarium, we were influenced by the fact that koi are social animals. Living in isolation can be stressful in itself, but transportation between a horne tank and a separate experimental tank can be stressful and experimentally impractical, especially once fish have grown to a length of more than a foot and a weight of several pounds. Therefore, our aquarium consists ofa group horne tank that can be partitioned into experimental and holding areas.We also developed a response button and feeder because we needed reliable equipment.