of the school was filmed, and the distance from the center of the 

 group of responding fish to the model was calculated from model speed 

 and reaction time. 



Model trajectory was controlled by guide wires, and model speed was 

 controlled by a weight and pulley system. The speed of the model was 

 variable between 0.3 m/sec and 2 m/sec, and was constant for any 

 given run. Three different sizes of models were constructed for each 

 of three shapes: circular, square, and elliptical (with major axis 

 equal to twice the minor axis) . „ These were designated "small" (area 

 100 cm^), "medium" (area 200 cm ), and "large" (area 400 cm ). These 

 models were all opaque and white and are referred to as the size- 

 shape series. A set of medium ellipses was constructed with opaque 

 models of colors black, red, silver, yellow, and white. In this 

 "color series" was a model of clear plexiglas ("clear") and a white 

 model which was wrapped with monofilament nylon at 2 cm intervals 

 ("strings"). These last two models were controls designed to evalu- 

 ate the role of vibratory and visual stimuli in evoking the escape 

 responses. 



A super- 8 movie camera in an underwater housing with remote cable 

 control was mounted on a tripod and used to record both the speed of 

 the model and the escape response of the fish. High speed Ektachrome 

 film was used. At the depth of the habitat, the camera's automatic 

 meter system maintained a lens opening between f 2 and f 4.5. The 

 depth of focus under these conditions is not ideal, but was adequate 

 to show escape responses. Suspended matter in the water was a more 

 severe problem. About 25% of our data films were difficult to 

 analyze, and an additional 25% of them were useless. 



To analyze the films, we modified a projector to permit viewing at 2 

 to 3 frames per second. Frames were counted electronically so that 

 the time required for the model to pass from one edge of the screen 

 to the other could be determined, as could the time from the instant 

 of the fishes' response to the time when the model passed the location 

 where the response started ("reaction time"). From the model's screen 

 transit time and a distance calibration sequence, the model's speed 

 could be accurately determined. Reaction time and model speed were 

 used to calculate the distance of the model from the fish at the 

 instant of response (reaction distance). 



In the analyses where lines were fit to data points, a computer 

 program (borrowed from Dr. E. W. Fager) employing the Bartlett linear 

 regression procedure was used. This procedure assumes normal distri- 

 bution of error in both x and y measurements. Because x values 

 (speed and reciprocal speed) were essentially accurate, and y 

 deviations from the Bartlett lines were approximately normal, the 

 confidence limits set on the slope by the Bartlett method are likely 

 to be reasonable. 



VI-202 



