EHRLICH ET AL: THERMAL BEHAVIORAL RESPONSES OF FISHES 



Reynolds and Thomson (1974) or Reynolds and 

 Casterlin (1976). The differential, however, be- 

 tween the acclimation and the final preferendum 

 must be considered. Reynolds and Thomson ( 1974 ) 

 tested fish acclimated 17°C below their final pref- 

 erendum. Crawshaw (1975) used a range of ac- 

 climation temperatures from 22°C below to 3°C 

 above the final preferendum and found that as the 

 temperature differential diminished so did the 

 time required to reach the final preferendum. Dif- 

 ferentials of 5°C required as little as 1 h and 3°C 

 only 0.5 h (Crawshaw 1975). Based on the temper- 

 ature differences between acclimation and the 

 final preferendum (Table 1), Reynolds' and our 

 results generally fit the pattern described by 

 Crawshaw. 



Behavioral Criteria 



Most studies pertaining to behavioral responses 

 of fishes to thermal gradients have been concerned 

 with only one factor: the final preferendum. Addi- 

 tional information, however, can be obtained from 

 examination of parameters associated with the 

 frequency distribution of the selected tempera- 

 tures, particularly: skewness (degree of distortion 

 from symmetry) and kurtosis (peakedness). Ivlev 

 and Leizerovich (1960) compared the percent of 

 the area under the curve of number of fish per 

 temperature against the mode of the distribution 

 as well as the percentage of the curve on either 

 side of the mode. Reynolds and Casterlin (1976) 

 and Reynolds (1977) discussed the relationship 

 between various measures of central tendency 

 (mean, mode, and median) and skewness. They 

 also improved descriptions of thermal behavioral 

 responses by quantifying skewness but did not 

 state the statistical significance of the skewness. 

 Sokal and Rohlf (1969) stated that the absolute 

 value of coefficients of skewness and kurtosis have 

 little meaning and that they must be tested for 

 statistical significance. We identified distinct be- 

 havioral types with respect to the frequency dis- 

 tribution of selected temperatures by examining 

 skewness and kurtosis. The responses were, in 

 part, species-specific but also varied with on- 

 togenetic stage and nutritive condition. Reynolds 

 and Casterlin (1976) showed that skewness also 

 varied diurnally. Kurtosis can be used to assess 

 whether the test organisms display eury- or steno- 

 thermal behavioral responses (Ivlev and Leizero- 

 vich 1960). A narrow preferred temperature range 



will be overly peaked about the mean (leptokur- 

 tic), and a broad range of preferred temperatures 

 will show no obvious mode or only a very slight one 

 (platykurtic). The coefficient of kurtosis is particu- 

 larly useful for quantifying the strength of the 

 temperature selection response in populations 

 that are not normally distributed where normal 

 parameters such as mean and standard deviations 

 are inappropriate. 



A normal bell-shaped frequency distribution is 

 representative of species with a wide preferred 

 temperature range that is not close to lethal or 

 avoided temperatures. Speckled sanddabs dis- 

 played this type of behavior (Table 1, Figure 7). 

 Newly hatched larvae, however, of species such as 

 California grunion showed little temperature 

 selection acuity and preferred an even wider range 

 of temperatures (^i = 0.003, 0.5<P), (^2 = l-^S, 

 0.01<P<0.025). This behavior resulted in a 

 platykurtic distribution (Figure 7). Reynolds and 

 Thomson ( 1974) reported that newly hatched Gulf 

 grunion, Leuresthes sardi?2a , also showed no acute 

 temperature preference. The precision with which 

 larval California grunion selected temperatures 

 increased by 2 days posthatching, producing a 



20 



10- 



o 



1^ 



Cithartchthys 

 stigmaeus 



m 



Jrf^n 



expei imentol 

 -temperoture 

 range 



T — I — I T 1 — n — n — I — n — r-r~ 



10 15 20 25 



TEMPERATURE (°C) 



Figure 7. — Temperature-specific occurrences. Examples of 

 three types of frequency distributions: normal (based on 220 fish 

 observations), adult speckled sanddabs; platykurtic (based on 

 281 fish observations), newly hatched California grunion; and, 

 skewed to the left (based on 241 fish observations), juvenile 

 topsmelt. 



845 



