528 



EHRLICH, McGOWEN, AND MUSZYNSKI 



20 25 



TEMPERATURE, °C 



30 



Fig. 3 Temperature-specific occurrences. The skewed frequency 

 distribution of 192-day-old fish is shown. 



juveniles to aggregate and school is shown by the smaller variation 

 about the temperature selected at each observation than was 

 demonstrated by ll-day-old and younger larvae. The older fish also 

 showed greater precision of temperature selection despite their 

 increasing activity with age. 



The habit of topsmelt of remaining in a given position 

 independent of temperature until conditions became extreme pre- 

 sents difficulties with respect to data interpretation. This response, in 

 conjunction with the experimental technique of moving isotherms, 

 gives little value to mean and modal preferred temperatures since 

 these are a function of the rate of shifting of the gradient. 

 Furthermore, once the fish started following an isotherm of 26 to 

 27° C, these temperatures became heavily weighted with respect to 

 fish occurrence; this resulted in a skewed frequency distribution 

 (gj =—0.72, P< 0.005) (Fig. 3). Apparently, selection of tempera- 

 tures near lethal limits produces skewed frequency distributions, 

 which vary with the duration of the experiment. Because of this, the 

 common practice of reporting mean and modal preferred tempera- 

 tures may not be adequate. Various investigators have reported that 

 the final preferendum is often near the upper lethal limit (Lowe and 

 Heath, 1969; Reynolds and Thomson, 1974; Reynolds and Casterlin, 

 1976). Beitinger (1977) stated that bluegUl (Lepomis macrochirus) 



