FISHERY BULLETIN: VOL. 80, NO. 3 



geneous pairs disclosed that stage-I fish moved 

 vertically to compensate for the imposed pres- 

 sure changes less often than those possessing a 

 more developed bulla system (Table 2). No advan- 

 tage for fish possessing gas-filled bullas rather 

 than liquid-filled bullas was detected. Although 

 there was a consistent tendency for herring pre- 

 viously exposed to 5 atm to move vertically to 

 compensate less frequently than those not so 

 exposed, this tendency was not statistically sig- 

 nificant. The overall intraclass correlation co- 

 efficient was 0.67, implying that the average 

 variation among pairs within a treatment group 

 was twice that between members of a pair. 



The hypothesis testing reported above ignores 

 the fact that a larval herring could respond in 

 three ways to the 47 changes in pressure; it could 

 move vertically to compensate, it could move ver- 

 tically in the opposite direction, or it could 

 simply maintain its current position within the 

 cylinder. A plot of the data for pairs of larvae 

 shows that nonresponse to changing pressure 

 was more frequent for stage-I larvae than for the 

 more developed fish (Fig. 2). Two pairs of larvae 

 in particular, one for the stage-I control group 

 and one from the stage-I treated group, are 

 clearly outliers, showing no response during 49% 

 and 86% of the trials, respectively. If these two 

 pairs are dropped from the analysis, then the dif- 

 ferences reported above are no longer significant 

 and the means are 27.7 and 25.2, respectively (in- 

 stead of 23.8 and 19.8) and are close to those for 

 the corresponding stage-II groups (Table 2). 



A plot of moving averages of the percentage of 

 compensatory responses against the sequential 

 series of pressure changes reveals that the stage- 

 I herring exhibited a rather consistent pattern of 



Table 2. — Analysis of variance of number of com- 

 pensatory moves of fish during pressure sensitivity 

 tests, and table of means: I = bulla absent, II = bulla 

 liquid filled, III = bulla gas filled. 



CO 

 LU 

 > 



O 



> 

 rx 

 o 



< 



co 



z 



LU 

 Q- 



o 

 o 



u. 

 O 



rx 



LU 

 CO 



Z 



■P<0.05; "P<0.001. 



10 20 30 40 50 60 70 80 90 



NUMBER OF ANTICOMPENSATORY MOVES 



Figure 2.— Total numbers of compensatory vertical move- 

 ments of a pair of herring plotted against the corresponding 

 number of anticompensatory movements for that pair. The 

 dashed line indicates the evenly divided response expected 

 under the null hypothesis. Nonresponse is indicated by the ver- 

 tical (or horizontal) distance from the hypotenuse. Control and 

 preexposed pairs are not plotted separately. Arrows indicate 

 two outliers. 



response over the 20 min test period for a given 

 AP, especially for the larger AP's (Fig. 3). Per- 

 formance improved at the onset of a new incre- 

 ment or decrement of pressure and then fell off 

 as the test continued, only to improve when the 

 next increment or decrement was used. The 

 other two developmental stages showed a rela- 

 tively high initial frequency of compensation 

 that rapidly decreased, then subsequently in- 

 creased until the test of the final pressure 

 change. 



Discussion 



The results in Figures 2 and 3 clearly show 

 that some larvae are responsive to pressure. 

 However, the relatively small sample size, the 

 correlation in the behavior between members of 

 a pair simultaneously tested, and the relatively 

 high variation in response among experimental 

 units within a given treatment group reduced 

 our ability to distinguish differences in the re- 

 sponse to changes in pressure of herring of dif- 

 ferent developmental stages. Additional prob- 

 lems in interpreting the first experiment arose 

 from the evidence that tests of a given pressure 



570 



