FISHERY BULLETIN: VOL. 78, NO. 3 



titative measures allows qualitative separation of 

 treatment group behavior, illustrates the high 

 variability in behavior within the 5.0 mm refuge 

 fence group, and shows that measures of numeri- 

 cal dynamics generally conform to anticipated 

 differences. 



The larger variability in population behavior 

 among the 5.0 mm treatment group was witnessed 

 earlier among the original 10 replicate popula- 

 tions equipped with 5.0 mm refuge fence. Laakso 

 (1959) speculated that the tendency toward canni- 

 balism increased with age of guppies and it may be 

 that at the J45 stage such tendencies are first 

 beginning to be expressed. The exact age or size at 

 which they become fully expressed may be highly 

 variable. An alternative possibility here would 

 certainly include possible imperfections in refuge 

 fence construction which might have allowed J50 

 fish, which clearly exhibited antagonistic behav- 

 iors toward fry, to remain within 5.0 mm refuge 

 areas beyond the size at which they should have 

 been excluded. 



(LENGTH) X 10 



Figure 11 — Number of embryos plotted against the cube of 

 female guppy length (stEtndard length in millimeters). Regres- 

 sion line is forced through the origin. 



SURVIVAL RATES. — Since the juvenile-fry 

 interactions often created pulses of numerical 

 increase in the guppy populations, there were 

 frequent intervals in which there were no new 

 individuals entering populations. Survival rates 

 for most populations during Phase I and Phase II 

 were estimated by comparing the numbers of fish 

 present at the beginning of such an interval to 

 numbers present at the beginning of the next 

 biweekly interval during which new numerical 

 growth was recorded. Survival rates for the 4.5 

 mm group and for certain 5.0 mm populations 

 during Phase II could not be estimated due to the 

 continuous nature of numerical increase. Esti- 

 mates of survival rates for intervals > 2 wk were 

 converted to biweekly estimates assuming con- 

 stant biweekly survival over the longer period. 

 Mean biweekly survival rate estimates ranged 

 from 0.972 to 0.995 during Phase I and from 0.953 

 to 0.984 during Phase II, averaging 0.984 and 

 0.970. These biweekly rates indicate roughly 50% 

 annual survival, a reasonable figure for laboratory 

 populations of guppies. 



FECUNDITY. — Dissection of a wide size range 

 of gravid guppy females at termination showed 

 that numbers of embryos were linearly related 

 to the cube of female length. Variability in 

 embryo counts appeared to increase with female 

 size (Figure 11). A linear regression of num- 



bers of embryos against the cube of female length 

 was forced through the origin and could be ex- 

 pressed as: 



E = 5.328 (L^ X 10""*) 



where E = number of embryos 



L = standard length in millimeters. 



The regression was forced through the origin to 

 ease manipulation of the fecundity relation in 

 further analyses and, over the majority of the size 

 range, did not differ appreciably from a regression 

 with intercept. The empirical relation obtained 

 was similar to earlier results of Felin (1935) and 

 Laakso (1959). 



A mean condition factor {K in the formula 

 weight = /C • L^) for gravid and nongravid fe- 

 males in A5 5 and larger size categories at weeks 

 36 and 58 was applied to the above relation to 

 obtain an expression relating embryo counts to 

 female weight in grams: E = 22.347  female 

 weight. In the following analyses the fecundity 

 relation was assumed to have held constant 

 throughout the 58- wk experimental period. 



ANALYSIS: NUMERICAL DYNAMICS 



Behavior of the treated guppy populations dur- 

 ing Phase II generally supported the hypothesis of 



568 



