EHRLICH ET AL THERMAL BEHAVIORAL RESPONSES OF FISHES 



term can be attributed to different species exhibit- 

 ing various behavioral patterns in a gradient 

 tank, just as they do in their natural habitats. This 

 makes it impossible to use only one procedure to 

 determine the preferred range for all species 

 under all conditions. We determined the preferred 

 range and final perferendum by evaluating, on a 

 case-by-case basis, the behavioral responses of a 

 species subjected to known conditions such as ac- 

 climation temperature, feeding patterns, and cap- 

 tivity environment. 



Experiments with larvae lasted 5-6 h, but 

 juveniles and adults were tested for approxi- 

 mately 7-8 h. An "experiment," in this study, con- 

 sisted of a set of individual runs, with each "run" 

 being the observation of the position and water 

 temperature selected by each fish in the gradient 

 at a fixed point in time. We employed constant 

 time intervals between runs for any experiment: 5 

 min for larvae and 15 min for juveniles and adults. 

 Run selected temperatures were the primary data 

 source, and we calculated their mean, mode, and 

 variance prior to combining them with data from 

 other runs to determine the preferred tempera- 

 ture. 



DATA ANALYSES 



The frequency of occurrence of all experimental 

 temperatures was not uniform due to the shifting 

 of the gradient as well as having a variable 

 number of degree intervals per run and generally 

 fewer than the 21 compartments. This caused a 

 bias in the number offish observed at a particular 

 temperature when summed over an entire exper- 

 iment. To compensate for this, prior to calculation 

 of the mean and modal selected temperatures, we 

 adjusted the data by using the number offish per 

 total occurrence of a particular temperature in all 

 experimental compartments rather than the ac- 

 tual number of fish at each temperature. 



We defined the "initial selected temperature" as 

 that chosen by the fish immediately following es- 

 tablishment of a gradient of lO'^C. The modal 

 selected temperature was determined from the 

 percent occurrence frequency distribution derived 

 from adjusted data. After an initial time of appar- 

 ent searching and testing of water conditions the 

 experimental animals selected a temperature or 

 range of temperatures at which they remained for 

 the duration of the experiment. We called this the 

 final selected temperature (or temperature range) 

 and determined it from plots of selected tempera- 

 ture against time. The mean selected temperature 

 was derived by methods presented in Appendix 

 Table 1. 



Reynolds (1977) reported that skewness of 

 temperature preference frequencj' distributions 

 required a complete description of the distribu- 

 tion. We examined the following parameters to 

 delineate thermal behavioral responses, the ini- 

 tial, mean, modal, and final selected tempera- 

 tures, standard deviation about the mean, 

 coefficients of skew^ness and kurtosis, and 

 coefficient of dispersion. The first four parameters 

 defined the preferred temperature range. The 

 standard deviation about the mean selected temp- 

 erature quantified movement through a range of 

 temperatures and gave a measure of the degree of 

 eury- or stenothermal preference. We used 

 coefficients of skewness and kurtosis (Sokal and 

 Rohlf 1969) in testing for normality and then to 

 help define the shape of the temperature-specific 

 fish frequency of occurrence distribution and to 

 refine interpretation of behavioral types. The 

 coefficient of dispersion quantified the tendency of 

 a species to aggi-egate or school and gave the per- 

 centage of use of the experimental chamber by all 

 fish within one standard deviation of the run 

 selected temperature. 



EXPERIMENTAL TECHNIQUES AND 

 BEHAVIORAL RESPONSES 



Our experimental techniques and data in- 

 terpretation methods are useful for a wide variety 

 of behavioral tj-pes. There are three salient fea- 

 tures of this methodology- 1) the shifting and re- 

 versal of the temperature gradient to partition 

 position preference from thermal preference, 2) 

 the extended duration of the experimental period 

 and its relationship to the thermal histon,- of the 

 test organisms, and 3) the criteria for behavior 

 evaluation. 



Shifting and Reversal of 

 Temperature Gradient 



Hasler ( 1956) pointed out that fishes in experi- 

 mental gradients can position themselves accord- 

 ing to small deformities in the tank structure. We 

 employed two methods to eliminate this factor: 

 shifting the position of a given isotherm in the 

 gradient during an experiment, and reversing the 

 hot and cold ends between replicate experiments. 



841 



