BARLOW: REPRODUCTIVE RATES OF SPOTTED DOLPHINS 



Samples sizes for the southern stock are highly 

 variable between years, and several years have too 

 few specimens to reliably estimate the fraction of 

 the population that would be pregnant. For 1973-83, 

 the overall percentage of pregnant females is, 

 however, significantly higher for the southern stock, 

 36.4%, than for the northern stock, 31.4% Oc? = 

 3.97, P = 0.05). Because of this difference, northern 

 and southern stocks were not pooled in subsequent 

 analyses. Due to the small sample size from the 

 southern stock, examination of trends and biases in 

 female reproductive rates was limited to the north- 

 ern stock. 



Annual estimates of the percentage of pregnant 

 females are illustrated in Figure 2 for the northern 

 offshore stock. The regression is significant; however, 

 the residuals do not appear randomly distributed. 

 The negative slopes of the regression lines are large- 

 ly due to high pregnancy rates in 1971-73. Expected 

 values for the percent pregnant in each year were 

 generated in two ways: from the overall percent preg- 

 nant and from the "de-trended" regression predic- 

 tions. Chi-square tests using these expected values 

 show the annual variability in percent pregnant is 

 greater than would be expected from random sam- 

 pling of a population with a constant (P < 0.001) or 

 linearly decreasing (P < 0.001) pregnancy rata 



Although natural year-to-year variability in 

 pregnancy rates cannot be ruled out, a changing bias 

 in sampling could also cause larger than expected 

 variability in percent pregnant. Td look for such a 

 bias, the sample from the northern stock was 

 stratified by the six sampling factors described 



abova The percent pregnant in each of these strata 

 is given in Tkble 2, with the chi-square probabilities 

 that the samples could have been drawn randomly 

 from the pooled sample Of the factors examined, 

 pregnancy rate was significantly related to sampling 

 season, dolphin kill-per-set, and tuna catch-per-set 

 (henceforth the latter two are referred to as kill and 

 catch). 



Because sampling seasons, mean dolphin kills, and 

 mean tuna catches vary significantly between years, 

 these factors cannot be considered independent of 

 year. For instance, the interaction between preg- 

 nancy rates and dolphin kill might appear significant 

 due to high kill rates or high catch rates in a year 

 (or years) that coincidentally had high pregnancy 

 rates. Conversely, high pregnancy rates in one year 

 may be due to a sampling bias related to dolphin 

 kill. 



Multiway tests were used to identify possible inter- 

 actions between the three significant factors and 

 year effects. In all cases, 3-way tests indicate that 

 year effects are significant (Tkble 3). First order ef- 

 fects of kill, catch, and season were not significant; 

 however, higher order effects involving the latter two 

 were important (Tkble 3). A 4-way test using catch, 

 season, and year also shows significant higher order 

 interactions involving both catch and season (Ikble 

 3). 



Higher order interactions involving year and 

 another factor indicate that effect of that factor 

 changes with year. Since pregnancy rates appear to 

 have changed markedly from 1971-73 to 1974-83 

 (Fig. 2), the effect of the significant factors was 



< 

 z 

 o 



UJ 



c 



Q. 



»- 

 Z 

 UJ 



o 



oc 



UJ 



& 



1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 



YEAR 



Figure 2.— Percentages of mature females that were pregnant for the northern offshore stock of spotted dolphins 

 from 1971 to 1983. Solid line represents weighted regression. Sample sizes are in parentheses. 



661 



