The Impact of Predation on Life History Evolution in Trinidadian Guppies: Genetic Basis of Observed Life History Patterns Author(s): David Reznick Source: Evolution, Vol. 36, No. 6 (Nov., 1982), pp. 1236-1250 Published by: Society for the Study of Evolution Stable URL: http://www.jstor.org/stable/2408156 Accessed: 03-03-2016 15:29 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Wiley and Society for the Study of Evolution are collaborating with JSTOR to digitize, preserve and extend access to Evolution.

http://www.jstor.org

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

Evolution, 36(6), 1982, pp. 1236-1250

THE IMPACT OF PREDATION ON LIFE HISTORY EVOLUTION IN

TRINIDADIAN GUPPIES: GENETIC BASIS OF

OBSERVED LIFE HISTORY PATTERNS

DAVID REZNICK'

Department of Biology, University of Pennsylvania, Philadelphia, Pennsylvania 19104

Received September 25, 1981. Revised March 6, 1982

The evolution of life history patterns

studies claiming to have established a ge-

often has been inferred from correlations

netic basis for life history differences con-

between the life histories and either pop-

found genetic differences with maternal

ulation survivorship curves (e.g., Tinkle,

effects (e.g., Ramakrishnan, 1960; Law et

1972; Tinkle and Ballinger, 1972; Vine-

al., 1977; Naevdal et al., 1978; Berven et

gar, 1975) or environmental factors sus-

al., 1979; exceptions include Alm, 1949;

pected of affecting survivorship (Con-

Birch et al., 1963; Reznick, 1981; Berven,

1982).

stantz, 1976, 1979; Maiorana, 1976;

Reznick and Endler (1982) found inter-

Stearns, 1976; Law et al., 1977). One dif-

ficulty with basing interpretations on sur-

population differences in guppy life his-

vivorship curves is that one cannot distin-

tories that were associated with differ-

guish between interpopulation differences

ences in predation. The purpose of the

in mortality due to external factors (pos-

present study was to determine if these life

sible sources of selection) and genetic fac-

history patterns have a genetic basis.

I will compare guppies from two types

tors that influence mortality (possible re-

sponses to selection). A general difficulty

of localities. In the first type of locality,

with the second approach is that the re-

the killifish Rivulus hartii is the only po-

lationship between the suspected environ-

tential guppy predator. Rivulus predom-

mental variable and the pattern of mor-

inantly eats small, immature size classes

tality that it causes in the study organism

of guppies (Seghers, 1973; Liley and Segh-

is either unknown or poorly characterized;

ers, 1975; Endler, unpubl.). At the second

e.g., is the survivorship of adults only, ju-

type of locality, guppies co-occur with the

veniles only, or all age classes affected?

pike cichlid Crenicichla alta and a variety

The expected response of the affected pop-

of other potential predators; these guppies

ulations depends strongly on this pattern

experience increased predation across all

age classes, plus selective predation on

of mortality (e.g., Gadgil and Bossert,

large, mature size classes (Seghers, 1973;

1970; Schaffer, 1974a, 1974b; Law, 1979;

Michod, 1979; Charlesworth, 1980). In few

Liley and Seghers, 1975; Endler, un-

systems is the pattern of mortality known

publ.).

Reznick and Endler (1982) found that

with confidence.

In addition to these difficulties, it is

field-collected guppies from "Crenicichla

rarely known whether there is a genetic

localities" mature at a smaller size, devote

basis to observed interpopulation differ-

a larger percentage of their total body

ences in life histories. Establishing this ge-

weight to each litter, reproduce more fre-

netic basis is fundamental to concluding

quently, and produce more and smaller

that these differences are an evolved re-

offspring than their counterparts from

sponse to demographic selection. Most

"Rivulus localities."

Environmental differences are associ-

ated with the different predator "treat-

ments." Rivulus streams tend to be small-

er, have slower currents, more canopy I Current Address: Department of Zoology, Uni-

cover, and higher densities of guppies than

versity of Maryland, College Park, Maryland 20742.

1236

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1237

Crenicichla streams (Reznick and Endler,

total body weight that consists of devel-

1982). These environmental factors are a

oping embryos), where they no longer dif-

potential, independent source of selection

fer significantly. The four localities thus

that can induce phenotypic changes in life

represent the major trends exhibited by

history patterns. For example, the higher

the entire sample (see Reznick, 1980 for

population densities of guppies and lower

details). If the observed patterns have a

light levels at Rivulus localities can reduce

genetic basis, then the F2's should display

the quantity of resources available per in-

similar differences for all variables, except

dividual. Lower food availability and the

possibly RA.

associated increase in intraspecific com-

Husbandry of Laboratory Stocks petiton is a potential source of selection

(Falconer and Latyszewski, 1952; Mac-

Twelve to 14 adult females were col-

Arthur and Wilson, 1967). In addition, fish

lected from the two Rivulus and two

have highly plastic life histories; differ-

Crenicichla localities. Female guppies re-

ences in food availability can cause phe-

tain viable sperm (e.g., Rosenthal, 1952),

notypic changes in growth rate, fecundity,

so all wild-caught individuals could be ex-

and age and size at maturation (e.g., Alm,

pected to produce one or more broods of

1959; Scott, 1962; Bagenal, 1969; Hislop

young. The fish were housed individually,

et al., 1978). The observed life history pat-

assigned to aquaria at random so that dif-

terns could be the product of such phe-

ferences between localities would not be

notypic plasticity. Alternatively, these en-

confounded with micro-environmental

vironmental differences could be an

differences across the laboratory. The de-

independent source of selection correlated

scendants of each female were considered

with the differences in predation.

a separate pedigree. Each locality was

This study assessed the life history pat-

therefore represented by a minimum of 24-

terns in the second generation of labora-

28 diploid genotypes. Guppies are likely

tory-born guppies so that any differences

to be multiply inseminated, as are other

between localities can be assumed to be

poeciliids (Borowsky and Kallman, 1976;

genetic. The fish received multiple, con-

Borowsky and Khouri, 1976), so the num-

trolled levels of food availability to allow

ber of genomes represented was probably

an estimate of the proportion of consumed

greater than this minimum.

calories devoted to reproduction and to

The F1's were sexed and the sexes were

mimic a potential difference in the natural

isolated prior to maturation. The criteria

environment of Crenicichla and Rivulus

used for sexing and the time of gonadal

guppies.

maturation included the morphogenetic

changes of the anal fin (Turner, 1941;

MATERIALS AND METHODS

Kallman and Schreibman, 1973) and the

Source of Laboratory Stocks accumulation of dark pigment in the anal

The experimental subjects were the de-

scendants from two Crenicichla (Aripo 6

and Oropuche 2) and two Rivulus (Aripo

1 and Quare 6) localities. These localities

will hereafter be referred to as Cren 1,

Cren 2, Riv 1, and Riv 2, respectively.

region of females.

When the F1 females were sufficiently

large to have attained maturity, they were

mated to mature F1 males derived from

the same locality but a different pedigree.

Some offspring from all P1 females were

Because these localities are only a subset

represented in these crosses to maximize

of the seven Crenicichla and five Rivulus

the genetic diversity of the fish in the ge-

localities considered by Reznick and En-

dler (1982), the results for this subset were

reanalyzed to determine if the patterns ob-

netics experiment. There was one cross per

aquarium. Crosses were assigned to

aquaria at random. Twenty-four crosses

served for the whole data set persist. The

were made per locality; only five failed to

trends for this subset parallel those of the

produce any offspring.

full data set for all variables except Re-

productive Allotment (RA, the percent of

When 25 days old, all F,'s were mea-

sured for standard length (mm) and weight

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1238 DAVID REZNICK

(mg), and sexed. Two females and two

tissues, reproductive tissues, and pre-

males were chosen from the middle of the

served newborn offspring were oven dried

size distribution of each litter for the ge-

overnight at 60 C. Each of these items was

netics experiment. The remaining off-

weighed to the nearest 0.1 mg (weight 1).

spring were preserved.

The tissues were extracted with anhy-

The four chosen offspring were placed

in two divided aquaria, with one male and

drous ether to remove triglycerides until

they reached a constant weight (weight 2),

one female in each tank. The two subdi-

ashed at 550 C in a muffle furnace, and

vided aquaria were assigned at random to

again reweighed (weight 3). The differ-

"high" and "low" levels of food availabil-

ence between weight 1 and weight 2 equals

ity. The control of food availability is ex-

the fat content of this tissue, and is mul-

plained below, under "Controlled Food

tiplied by 9.5 for conversion to the number

Availability." Seven to ten such litters

of calories (Kleiber, 1975). Weight 2 minus

of offspring were initiated for the four lo-

weight 3 equals the protein content, and

calities. The placement of these litters in

is multiplied by 5.7 to yield the number of

the laboratory was again random.

Females were measured every other

calories (Kleiber, 1975). The carbohydrate

content of the lean tissues is assumed to

week and after the birth of each litter.

be negligible. Similar measurements were

They were mated once a week with a male

made on the food given to the fish (Rez-

from a different pedigree. Females were

nick, 1980, Appendix 2).

maintained until they produced three

broods of young, then preserved with the

The estimated caloric content of new-

born young, and hence their contribution

third brood. The entirety of most F3 litters

to RE, had to be "scaled up" to account

were preserved.

for dry weight loss over the course of de-

The variables measured in this study in-

velopment. In wild-caught fish the weight

cluded: 1) size of the F2 fish at 25 days,

of developing offspring was at a maximum

the beginning of controlled food avail-

at the earliest stage of development, and

ability; 2) dry weights of newborn F3 fish;

decreased by an average of 38.3% from

3) number of young in each of the three

mature ova through very late-eyed em-

litters; 4) interbrood intervals; 5) age and

bryos. Because the percent fat or ash con-

weight at maturation in F2 males; 6) age

tent did not decrease or increase signifi-

and weight at first parturition in F2 fe-

cantly over the course of development, the

males; 7) reproductive allotment (RA) (the

caloric content of newborn offspring was

percent of total dry weight devoted to off-

converted into the estimated number of

spring) for F2 females; and 8) Reproduc-

calories invested in the litter by the mother

tive Effort (RE). Weight, rather than

by simply multiplying the former figure by

length, was always used as a measure of

size in this study because the measurement

1/(1 - 0.383) or 1.62.

RE was estimated at 6, 8, 10, and 12

error for weight was much smaller. RA

weeks after the initiation of controlled food

was analyzed for only the third litter be-

availability, or when the fish were 67, 81,

cause this is the the only litter for which

95, and 109 days old. These ages were

the dry weights of the mother and the off-

chosen because they correspond to routine

spring were available. RE was estimated

as the percent of consumed calories that

measurements of female size and changes

in food availability. At 4 weeks most fe-

were devoted to reproduction (Hirshfield

males had not yet matured; by 14 weeks

and Tinkle, 1975).

many females had already produced three

litters and had been preserved.

Estimation of Reproductive Effort The number of calories devoted to re-

To estimate RE, it was necessary to de-

production at a given age equaled the ca-

termine the caloric content of the food, the

loric content of any young born prior to

three litters of F3 offspring, and the re-

that age plus the caloric content of the de-

productive tissues present in the F2 fe-

veloping litter. Estimating the caloric con-

males after their third litter. The somatic

tent of the developing litter required sim-

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1239

plifying assumptions concerning the

A record was kept of each day that an

pattern of the mother's energy investment

individual did not consume its full ration.

in the young. Evidence indicated that

Consistently poor feeders were deleted

Trinidadian guppies did not begin yolking

from some analyses (see below).

a new set of ova until the current litter

Food availability was rescaled to the

was very near parturition (pers. observ.),

average size of all fish in the experiment

and that the new clutch was fully yolked

after each biweekly measurement. Be-

cause the objective of the experiment was

and initiated development within six days

after the birth of the previous litter (Ro-

to determine if there are systematic differ-

senthal, 1952). The simplest model for the

ences between localities in how a constant

pattern of energy investment by the moth-

resource base is utilized, all fish received

er, which was used for the results reported

the same quantity of food for each 2-week

below, was to credit a female for the com-

period, regardless of an individual's size

plete caloric content of the i + 1th litter on

at the beginning of the period. This scal-

the day that the ilh litter was born. A more

ing was based on a series of short-term

realistic model involved averaging the in-

feeding studies where I determined the

vestment in litter i + 1 over a time period

quantity of food required to sustain a giv-

both before and after the birth of litter i;

en percentage of the maximum growth rate

analyses of these two models gave quali-

(Reznick, 1980). These feeding studies also

tatively similar results.

demonstrated that the available "uncon-

If a female had not yet produced her

first litter, she was credited for the full

trolled" food sources, such as algae, were

likely to be minimal. The "high" and "low"

caloric content of that litter if the age in

food treatments initially received suffi-

question fell within one interbrood inter-

cient rations to sustain 85-90% and 60-

val of the first litter. The value of the in-

65 % of the maximum growth rate, re-

terval used was the average of each indi-

spectively.

vidual's two intervals. For example, if a

Other details of aquarium maintenance

female produced her first litter when 80

and environmental condition in the labo-

days old and produced the next two at 21

ratory can be found in Reznick (1980) and

day intervals, she was credited with the

Reznick and Endler (1982).

number of calories in her first litter at 67

Statistical Design and Analysis days, the first two litters at 81 and 95 days,

The life history variables were analyzed

and all three litters at 109 days.

In spite of the guppy's reputation for

eating its young, isolated females almost

as a two-way analysis of variance, with

localities and food availability as fixed ef-

never did so; of 154 females who were pre-

fects. The three locality degrees of free-

served with their newborn young, only

dom were divided into three single degree

three were found to have eaten any. No

of freedom contrasts, one comparing the

precautions were taken to prevent canni-

predator "treatments," one comparing the

balism in this experiment.

two Rivulus localities, and one comparing

the two Crenicichla localities. All analyses

Controlled Food Availability of variance were executed with the SAS

When food availability was controlled

(after 25 days for the F, generation), each

fish received measured quantities of liver

paste (Gordon, 1950) in the morning and

live brine shrimp nauplii in the afternoon.

The food was measured volumetrically to

the nearest 0.5 micro-liter with a Hamil-

ton micropipette. The coefficient of vari-

ation of the dry weight of a typical ration

was approximately 3-5% (Reznick, 1980).

General Linear Models Procedure (Helwig

et al., 1979). Because of the unequal sam-

ple sizes (see below), all F-tests were based

on the Type IV sums of squares.

The distributions of the residuals of all

analyses of untransformed data complied

with the assumptions of the analysis of

variance, being normally distributed and

homoscedastic. All analyses presented be-

low were therefore performed on untrans-

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1240 DAVID REZNICK

TABLE 1. Offspring dry weight (mg) and number for the first through third litters in F2 females.

Offspring weight (mg) Offspring number

Source df 1st 2nd 3rd 1st 2nd 3rd

Y wt. (covariate) 1 25.91*** 10.25** 0.30 14.43*** 10.91** 0.54

Locale 3 6.47** 2.77t 2.83* 5.17** 1.69 1.76

Planned comparisons

Riv vs. Cren 1 15.46*** 6.54* 8.34** 8.14** 0.39 0.01

w/in Riv 1 2.09 0.96 0.48 3.60t 1.24 0.16

w/in Cren 1 0.05 0.01 0.00 3.57t 3.18t 4.94*

Food 1 0.12 1.69 0.47 0.77 3.48t 7.59**

Interaction 3 1.09 0.60 0.45 0.66 0.12 0.23

Residual sum of squares 45 2.59 5.32 9.73 216.19 446.65 788.84

t .05 < P < .10. * P < .05: **P < .01, *** p < .001.

formed data. The additional assumption

chronically poor feeders. This behavior

of homogeneous slopes was met in all

was most apparent after the age at first

analyses of covariance.

parturition. Irregular feeding artificially

The results reported in the text include

reduced energy allotment to growth and

the F-ratios for all analyses (Tables 1-3),

reproduction; however, there was no ap-

the predator means (Fig. 1), plus other rel-

parent impact of the abnormal feeding on

evant group means (in text). In addition,

the weight or age at first parturition or

the means for both levels of food avail-

interbrood interval. The latter three vari-

ability and all four localities are reported

ables were analyzed both with and with-

in the appendix. All reported means are

out these individuals; the results were

least square means, which are "estimates

qualitatively the same. The reported anal-

of the class or subclass arithmetic means

ysis for these three variables (Table 3) in-

that would be expected had equal subclass

cluded these six Riv 1, "high" food indi-

numbers been obtainable" (Helwig et al.,

viduals. The six individuals were deleted

1979 p. 249). Where appropriate, these

from the remaining analyses.

means are also adjusted for the effects of

Additional missing values were due to

deaths (three individuals), infertility (one

covariates.

individual), errors in sexing (one individ-

Missing Values and Outliers ual) or the unavailability of fish at the be-

ginning of the experiment (three individ-

Six of the nine females in the "high"

uals). Four fish were outliers in some

food treatment in the Riv 1 group were

TABLE 2. Age at maturity (weeks) and weight at maturity (mg) in F2 males.

Source df Age Wt

Wt at 25 days 1 17.26***

Locale 3 7.51*** 3,93*

Planned Comparisons

Riv vs. Cren 1 13.14*** 10.22**

w/in Riv 1 0.95 0.01

w/in Cren 1 1.81 1.75

Food 1 2.41 1.97

Interaction 3 0.55 0.32

Residual sum of squares

(degrees of freedom) 34.12 (56) 12,916.10 (57)

* P < .05; ** P < .01; *** P < .001.

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1241

TABLE 3. Life history characteristics of F2 females: Age (days) and weight (Wt - mg) at first parturition,

interbrood interval (Int - days), reproductive allotment (RA - I%), and reproductive effort (RE - I).

Source df Age Wt Int RA RE

Locale 3 5.77** 15.76*** 12.90*** 6.37** 7.22***

Planned comparison

Riv vs. Cren 1 14.49*** 19.96*** 32.54*** 1.15 6.19*

w/in Riv 1 3.86t 31.51*** 0.43 6.55* 10.30**

w/in Cren 1 0.40 0.67 6.24* 12.26** 1.64

Food 1 4.04* 3.85t 0.55 2.13 1.92

Interaction 3 0.41 0.49 0.67 0.08 0.27

Residual sum of squares

(degrees of freedom) 6,175.8 (54) 113,236.7 (54) 153.3 (54) 0.0410 (46) 0.2999 (46)

t 05 < P < 10; * P < .05; ** P < .01; *** p < .001.

trends in the correlation coefficients for the

analyses because of an exceptionally late

age at first parturition (one individual), an

remaining variables. The size at 25 days

exceptionally long interbrood interval, ap-

is therefore used as a covariate only in the

parently due to a skipped litter (one indi-

analysis of age at maturation in males.

vidual), or exceptionally low fecundity (two

individuals). Further details are given in

Offspring Weight

Reznick (1980). These individuals were Rivulus guppies produce significantly

only deleted from those analyses where the heavier offspring than Crenicichla guppies

value of their dependent variable was an (Fig. 1, Table 1). The female's post-par-

outlier. As a result, the total degrees of tum wet weight is a covariate in these

freedom are not the same for all analyses. analyses because larger females tend to

produce larger young. In addition, off-

RESULTS

spring weight tends to increase from the

Weight at 25 Days first to the third litter and as food avail-

The F2 offspring were fed ad lib until

ability decreases. Both trends were dupli-

25 days old and weighed prior to being

cated and the food trend was more prom-

kept singly with controlled food availabil-

inent in a second experiment employing a

ity. There are significant differences be-

wider range of food availability (Reznick,

tween localities for the 25-day weight.

1980).

(Means-Rv 1 = 46.9 mg, Riv 2 = 53.6

Offspring Number mg, Cren 1 = 47.7 mg, Cren 2 = 41.0 mg;

F3,126 = 18.25, P < .0001.) Differences in

Crenicichla guppies produce signifi-

size at the beginning of the feeding exper-

cantly more offspring than Rivulus gup-

iment have a potential impact on the val-

pies in the first litter (x Cren = 5.2; x

ues of all subsequent variables. Because

Riv = 3.2); this difference disappears in

all fish receive the same rations for a given

the second and third litters (Table 1). The

time interval, a fish that is larger at the

"high" food females produce more off-

beginning of an interval receives relatively

spring than the "low" food females in all

less food than a smaller individual whose

three litters. The magnitude of the differ-

maintenance costs are lower. The possible

ence increases from the first to the third

importance of these differences in initial

litter, becoming statistically significant in

weight was evaluated by correlating initial

the third litter (Xhigh = 18. 1, xlow = 14.0).

weight with all subsequent parameters in

There is also some significant heteroge-

each of the treatment groups. Male age at

neity between localities, within a predator

maturation is consistently negatively cor-

treatment, with Cren 1 females producing

related with size at 25 days. There are no

more offspring than Cren 2 females in the

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1242 DAVID REZNICK

Offspring weight Male age Male weight Female age

. (mg) (weeks) 110 (mg) (days)

09 f 7 f 70l 70n :

C R C R CR CR

Female weight 26 Interval (days) 0.17 RA (%) 0.35 RE (%)

(mg)

300-

24- 0.15-02-

200[ft_ _ _ _ _ _ i ,+2~~~~22 ] 0.13 < 0.15 l

C R C R C R CR

FIG. 1. Summary of lab results. Offspring weight, mean dry weight of third litter offspting (mg). Male

age (weeks), age of males at maturation. Male weight (mg), wet weight of males at maturation. Female age

(days), female age at first parturition. Female weight (mg), female wet weight after first parturition. Interval

(days), interbrood interval. RA, reproductive allotment. RE, reproductive effort. Horizontal line = mean,

vertical line= ? one standard error. C, populations from Crenicichla localities. R, populations from

Rivulus localities.

third litter. Because fecundity tends to in-

they first give birth (high = 73.9 days,

crease with female size, the female's post-

xlow = 79.4 days). "Low" food females also

partum weight is included as a covariate.

tend to be smaller than their "high" food

counterparts (xhigh = 256 mg, Xlow = 233

Age and Weight at Maturatzon in Males mg). Both differences were significant in

Males from Rivulus localities are sig-

a separate experiment (Reznick, 1980).

nificantly older and larger at maturation

The two Rivulus means also tend to be

than males from Crenicichla localities (Fig.

different from one another. Riv 2 females

1, Table 2). In addition, "low" food males

are larger (316 mg vs. 224 mg) and tend

tend to mature at a later age and smaller

to be older (85.6 days vs. 78.1 days) at first

size than "high" food males. Both of these

parturition than Riv 1 females. The Riv 1

trends were significant in a subsequent ex-

females are nearly equal in size to the two

periment employing wider ranges of food

sets of Crenicichla females. The mean age

availability (Reznick, unpubl.).

at first parturition in Riv 1 females is in-

termediate to the Crenicichla and Riv 2

Age and Weight at First Parturition in means; however, the localities rank order

Females in a fashion that is consistent with the

Rivulus females are older and larger at

overall differences between predators. All

first parturition than Crenicichla females

four Rivulus samples (two localities by two

(Fig. 1, Table 3). Food availability also

levels of food availability) have greater ages

has a significant effect; "low" food females

are older than "high" food females when

at first parturition than all four Crenicich-

la samples (see Appendix).

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GTTPPV LIFE -HISTORV G(ENETICS 1243

Interbrood Interval

.092

The dependent variable in the analysis

of interbrood interval is the mean of the

.08- Crenl

two intervals recorded for each female.

LuGren2

The lengths of the two intervals tend to

.071

be highly correlated. Rivulus females have

significantly longer interbrood intervals

.06than Crenicichla females (Fig. 1, Table 3).

a)

The difference in length is approximately

4- .052.2 days, so that the duration of the av-

erage Crenicichla interval is only 90% that

L0

0L of the average Rivulus interval. The two

a) .04

Crenicichla means differ significantly;

however, as with the age at first parturi-

.03

tion, all four Rivulus samples have longer

interbrood intervals than all four Creni.02-

cichla samples. Also, in two independent

experiments, one on wild-caught females .01

(Reznick and Endler, 1982) and one on F2

6 8 10 12 females (Reznick, 1980), these same Cren-

Age (weeks)

icichla localities are not significantly dif-

ferent from one another. FIG. 2. The response profile of reproductive effort

In a second study of the effects of food

(RE) in 67, 81, 95, and 109 day-old fish.

availability, the "low" food treatment has

significantly longer interbrood intervals

than the higher two levels (F2,30 = 8.16,

P < .01; Xhigh = 23.0, Xinediurn = 23.1,

Xlow = 24.3). One plausible interpretation

of this result is that, as food availability

decreases, the time required to yolk a new

.1544, Riv 2 = .1248, Cren 1 = .1576,

Cren 2 = .1096) implying that RA is a

fairly stable characteristic of these fish in

the laboratory environment.

Reproductive effort (RE).-RE is esti-

clutch of eggs increases. The two highest

mated as the percent of consumed calories

levels of food availability in this study are

that are devoted to reproduction at a given

similar to those of the previous study, are

age (see Materials and Methods) and is thus

nearly identical in value, and are appar-

a composite of the age at first reproduc-

ently too high to have an impact on inter-

tion, interbrood interval, offspring num-

ber, and offspring size. Because each in-

brood interval.

Reproductive allotment. -RA, or the

dividual is represented by four consecutive

percent of total dry weight that consists of

measures of RE (Fig. 2), the results are

a single litter of offspring, is only directly

analyzed as a multivariate (repeated mea-

estimable for the third litter. The only sig-

sures) analysis of variance. A preliminary

nificant effect in the analysis of RA is be-

analysis found no locality by response in-

tween localities within a predator treat-

teraction by doing a one-way, multivari-

ment (Fig. 1, Table 3). Riv 1 and Cren 1

ate analysis of variance on the differences

have substantially higher RA's than Riv 2

between the adjacent values of RE (Pillai's

and Cren 2 (Means: Riv 1 = .1723, Riv

2 = .1398, Cren 1 = .1670, Cren 2 =

Trace-F9,138 = 1.01, P = .4318; Mor-

rison, 1976 p. 207). The absence of a lo-

cality by response interaction simplifies the

.1267).

RA was estimated in a separate exper-

test for locality effects to a univariate

iment on F2 and F3 females being fed ad

analysis of variance on the sum of the four

lib (Reznick, 1980 Ch. 5) and, while the

RE values for each individual (Morrison,

values were uniformly lower, produced

remarkably similar results (Means: Riv 1 =

1976 p. 207-208). Crenicichla guppies have

significantly greater RE's than Rivulus

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1244 DAVID REZNICK

TABLE 4. Summary of life history differences between Rivulus and Crenicichla guppies in F2 (lab) and wild-

caught fish.

Lab Field

1) d's

a) Age at maturation Cren < Riv* not available

b) Size at maturation Cren < Riv* Cren < Riv*a

2)Y's

a) Age at first parturition Cren < Riv* not available

b) Size at first parturition Cren < Riv* Cren < Riv*b

3) Interbrood interval Cren < Riv* Cren < Riv*

4) Embryo weight Cren < Riv* Cren < Riv*

5) Fecundity Cren > Riv*c Cren > Riv*d

6) Reproductive allotment Cren -Riv ns Cren > Riv*

7) Reproductive effort Cren > Riv* not available

*P < .05.

ns = not significant (P > .05).

a Average size of mature males.

b Minimum size class in which the majority of females were gravid (see Reznick and Endler, 1980)

c For first litter only. Fecundities do not differ significantly for second and third litters.

I Based on non-parametric Anova of the expected fecundity of a female with a somatic dry weight of 30 mg. See Reznick and Endler (1980)

for details on data.

I will comment on the results for repro-

guppies (Fig. 2, Table 3). The two Rivulus

ductive allotment, reproductive effort, and

means are also significantly different (Ta-

food availability.

ble 3). As with the weight at first partu-

rition, the Riv 1 values for RE tend to

Reproductive Allotment equal those for the two Crenicichla sam-

ples (Fig. 2) and it is the Riv 2 value that

While the two predator "treatments" do

not differ in RA in the present study, an

is responsible for the difference between

analysis of the field results for this subset

predator "treatments."

of localities studied by Reznick and En-

DISCUSSION dler (1982) indicates that no difference

Male and female guppies from Rivulus

localities attain maturity at a later age and

larger size than guppies from Crenicichla

localities. Rivulus females also have long-

er interbrood intervals, larger offspring,

could be expected. The similar values of

RA in two independent experiments in-

dicate that this character may have a ge-

netic basis. This result reflects the rela-

tively small magnitude of the difference

lower fecundity in the first litter, and tend

between predator "treatments" for RA rel-

to devote less energy to reproduction than

ative to the other life history parameters.

Crenicichla females (summarized in Fig.

1). All of these differences parellel the

findings for field-collected guppies (Table

4) and, because they persist after two gen-

erations in a common environment, they

are assumed to have a genetic basis. Ear-

lier discussion by Reznick and Endler

(1982) of how these life history patterns

evolved was based on field collected fish

With most of the remaining variables

measured by Reznick and Endler (1982),

the means for the Crenicichla populations

are non-overlapping with the Rivulus

means, so that almost any subset of these

localities would produce qualitatively sim-

ilar results to the whole data set. With

RA, while the predator "treatments" are

different overall, the population means

and indirect estimates of age at maturity

show some overlap, so a randomly chosen

and RE. This discussion is now justified

subset will not necessarily reflect the over-

with direct measurements. Before consid-

ering the general significance of these data,

all pattern. RA may thus be less respon-

sive to predator-mediated selection or more

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1245

sensitive to other factors than other life

terval over which the investigator mea-

history parameters. The absence of a dif-

sures them. For example, if the trends in

ference in RA may also reflect the disap-

this study persist, then a single measure

pearance later in life of an initial differ-

of RE at 150 days might indicate no sig-

ence in the rate of investment in

nificant differences between predator

reproduction, as discussed below.

"treatments," while a single measure at 80

days certainly would. It thus seems more

Reproductive Effort

appropriate to consider the time course of

If the Riv 2 and two Crenicichla local-

ities prove typical of the predator "treat-

ments," then the differences in RE be-

RE with respect to the survivorship curves

of the organism in question, than to con-

sider a single measure of RE.

tween Crenicichla and Rivulus guppies

The Effects of Food Availability may only be evident in the first few months

of life. Their fecundities differ only for the

Food availability was controlled both to

first litter (Table 1) and the magnitude of

allow a direct measure of RE and because

the differences in RE between the three

differences in food availability represent a

localties narrows from 67 to 109 days; the

potential environmental difference be-

Cren 2 and Riv 2 means are no longer

tween the two types of localities. The per-

significantly different at 109 days. If this

formance of Rivulus and Crenicichla gup-

pattern persists to later ages, it may in-

pies on a constant resource base

dicate that the impact of Crenicichla se-

demonstrates that the two are genetically

lection is heaviest on the early age classes,

distinct independently of any such envi-

placing an emphasis on producing some

ronmental differences. Furthermore, re-

young as soon as possible.

Mertz (1975) observed a similar re-

sponses to decreasing food availability do

not entirely parallel the differences in life

sponse of flour beetles to an artificial re-

histories between Rivulus and Crenicichla

duction of adult survivorship. Adults in

localities. Lower food results in a tendency

selected lines were killed either 10 or 20

towards larger offspring, lower fecundity,

days after eclosion, while the control group

longer interbrood intervals, and later ages

was allowed to reproduce for its natural

at first parturition. These trends all par-

life span. After 11-12 generations of se-

allel the direction of the change from

lection, the 10-day line had significantly

Crenicichla to Rivulus localities, where

greater early fecundity than the other two

food availability is potentially decreasing.

lines; all three lines had equal fecundities

However, decreasing food also results in

thereafter. In both 10 day beetles and

smaller sizes at maturity while Rivulus

Crenicichla guppies, it is possible that the

guppies tend to be larger at maturity.

selection for reproductive performance

GENERAL DISCUSSION later in life is weak because few or no in-

dividuals normally survive to that point.

Charlesworth's (1980) theoretical consid-

eration of evolution in fecundity genes with

Predator-mediated differences in age-

specific survivorship potentially explain the

evolution of these life history patterns. By

age-specific effects also bears on this ob-

preying selectively on small, immature

servation. He found that where selection

guppies, Rivulus hartii reduce juvenile

favors a change in fecundity, it will be

survivorship. Crenicichla alta and asso-

stronger on genes expressed earlier in life.

ciated predators tend to prey selectively on

A measureable change in this character is

large, mature size classes of guppies, and

thus more likely to be evident in earlier

hence reduce adult survivorship (Haskins

et al., 1961; Liley and Seghers, 1975; En-

age classes.

These patterns of reproductive invest-

dler, pers. commun.). Crenicichla-like

ment are of practical significance because

mortality patterns are consistently pre-

they suggest that the nature of observed

dicted to favor earlier maturity and in-

differences in RE depend on the time in-

creased reproductive effort, while Rivu-

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1246 DAVID REZNICK

lus-like mortality patterns favor increased

intra-specific life history comparisons (see

age at maturity and decreased reproduc-

Introduction), may help explain why these

tive effort (Gadgil and Bossert, 1970;

comparisons frequently fail to fit theoret-

Schaffer, 1974a; Law, 1979; Michod, 1979;

ical predictions. The major problems are

Charlesworth, 1980). The results pre-

determining the interpopulation differ-

sented here and by Reznick and Endler

ences in demographic selection and estab-

(1982) are in accord with these predic-

lishing a genetic basis for the observed

tions. These facts, plus the assessment of

patterns. Since these problems are rarely

reproductive effort, address the criteria

addressed, the exercise of classifying com-

outlined by Stearns (1977) for evaluating

parisons as being consistent or inconsis-

empirical studies of life history evolution.

tent with theoretical predictions seems

In so doing, this work also provides a case

premature.

where theories of life history evolution will

eventually become testable.

Environmental differences (e.g., food

availability) associated with differences in

The prevailing difficulty with compar-

isons made between species or at higher

taxonomic levels is, as implied by Stearns

(1980), that it becomes impossible to pre-

predation may still have selected for the

cisely characterize the factors responsible

observed genetic differences in life histo-

for selecting for the observed patterns.

ries. Reznick and Endler (1982) presented

Since the underlying selective forces are

three lines of evidence favoring the direct

lost to history, I question the usefulness of

role of the predators in selecting for these

such comparisons beyond the fashion in

patterns. Field transplant experiments

which they have already been employed,

currently in progress will definitively dis-

which was to suggest genera.l patterns and

tinguish between these possibilities.

Stearns (1980) recently suggested that

stimulate further research in the field. The

only unambiguous arguments for life his-

evidence for life history evolution may be

tory evolution in response to specific forms

better sought in comparisons made be-

of demographic selection are those based

tween species or at higher taxonomic

levels, rather than within species. This re-

on within species comparisons, where po-

tential causes of life history patterns can

versal of his earlier argument (1976) was

be considered and tested. An outstanding

made because of the possibility that "de-

difficulty with intraspecific comparisons is

sign constraints," or interactions between

the possible role of "design constraints" and

individual life history parameters, might

how they limit the evolutionary response

limit the available avenues for adaptation.

of the organism. Rather than use such

In addition, Stearns proposed that the the-

constraints as a reason for substituting

ory of "punctuated equilibria," if appli-

weaker methods, the goal should be to

cable to life history characteristics, sug-

characterize these constraints and deter-

gested that major adaptive changes would

be associated with speciation events, while

mine their role in molding life history evo-

lution.

characters within a species would remain

SUMMARY fairly stable. This argument corresponds

well with the observation that the pre-

Genetically based life history differ-

dicted correlations between life history pa-

ences are described in guppies exposed to

rameters are more strongly evident in

differences in predator-mediated age-spe-

comparisons made at higher taxonomic

cific survivorship in their natural environ-

levels than in comparisons made within

ment in Trinidad. At one type of locality,

species (Stearns, 1980). However, the ob-

servation of genetic life history differences

within species reported here and in other

the killifish Rivulus hartii is the only po-

tential guppy predator. Rivulus preys se-

lectively on small, immature size classes

studies (see references in Introduction) ar-

of guppies. At a second type of locality,

gues against Stearns's suggestion.

the pike cichlid Crenicichla alta and other

Difficulties inherent in interpreting most

predators prey selectively on large, mature

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1247

size classes of guppies. The second gen-

periments possible. Finally, I thank Jeanie

eration of laboratory reared guppies from

Arciprete for preparing and editing the

two Rivulus and two Crenicichla localities

manuscript.

are genetically different for most of the

variables discussed in a previous study on

LITERATURE CITED

field collected fish (Reznick and Endler, ALM, G. 1949. Influence of heredity and environ-

ment on various forms of trout. Rep. Inst. Freshw.

1982). Crenicichla guppies have earlier

Res. Drottningholm 2 9:2 9-34. ages and smaller sizes at maturation in

1959. Connection between maturity, size

males and at first parturition in females.

and age in fishes. Rep. Inst. Freshw. Res. Drott-

They also have smaller offspring, more ningholm 40:5-145.

offspring in the first litter, shorter inter-

BAGENAL, T. B. 1969. The relationship between

food supply and fecundity in brown trout Salmo brood intervals, and greater reproductive

trutta L. J. Fish. Biol. 1:176-182. efforts (estimated as the percent of con-

BERVEN, K. 1982. The genetic basis of altitudinal

sumed calories devoted to reproduction). variation in the wood frog Rana sylvatica. II. An

These trends in life history patterns are experimental analysis of larval development.

Oecologia 52:360-369.

in close agreement with predictions made

BERVEN, K. A., D. E. GILL, AND S. J. SMITH-GILL. by life history theory. Decreased adult

1979. Countergradient selection in the green frog,

survivorship (i.e., Crenicichla predation)

Rana clamitans. Evolution 33:609-623.

is predicted to favor increased reproducBIRCH, L. D., TH. DOBZHANSKY, P. 0. ELLIOTT,

tive effort and decreased age at maturity.

AND R. C. LEWONTIN. 1963. Relative fitness

of geographic races of Drosophila serrata. Evo-

Decreased juvenile survivorship (i.e., Ri-

lution 17:72-83. vulus predation) should favor the opposite

BOROWSKY, R., AND K. D. KALLMAN. 1976. Pat-

response. These are precisely the trends terns of mating in natural populations of Xi-

observed in the study on field collected fish phophorus (Pisces: Poeciliidae). I. Xiphophorus

muculatus from Belize and Mexico. Evolution 30:

(Reznick and Endler, 1982) and in the ge-

693-706. netic study presented here.

BOROWSKY, R., AND J. KHouRI. 1976. Patterns of

mating in natural populations of Xiphophorus. ACKNOWLEDGMENTS

II. Xiphophorus variatus from Tamaulipas,

This research was supported by NIH

Mexico. Copeia 176:727-734.

CHARLESWORTH, B. 1980. Evolution in Age training grant GH07071, an NSF predoc-

Structured Populations. Cambridge Univ. Press, toral grant, and the University of Penn-

Cambridge.

sylvania. This work was submitted to the

CONSTANZ, G. D. 1976. Life history strategy of

University of Pennsylvania, Department the Gila topminnow, Poeciliopsis occidentalis: a

of Biology, in partial fulfillment of the re-

field evaluation of theory on the evolution of life

histories. Ph.D. thesis. Arizona State University. quirements for a Ph.D. I wish to thank

1979. Life history patterns of a live bear-

my dissertation committee, Robert Rick-

ing fish in contrasting environments. Oecologia

lefs, George Constantz, Arthur Dunham, 40: 189-201.

John Gillespie, Neville Kallenbach, and

Thomas Uzzell and my wife, Susan Snell,

ENDLER, J. A. 1978. A predator's view of animal

color patterns. Evol. Biol. 11:319-364.

1980. Natural selection on color patterns for their support and encouragement

in Poecilia reticulata. Evolution 34:76-91.

throughout this study and for their helpful

FALCONER, D., AND M. LATYSZEWSKI. 1952. The

comments on the manuscript. I am espeenvironment in relation to selection for size in

cially grateful to Michael Hirshfield, Jo-

mice. J. Genet. 51:67-80.

GADGIL, M., AND W. BOSSERT. 1970. Life historseph Travis, and Keith Berven for their

ical consequences of natural selection. Amer. helpful discussions and critical reviews.

Natur. 104:1-24.

Stephen Stearns and an anonymous reGORDON, M. 1950. Fishes as laboratory animals,

viewer also made valuable comments on p. 345-449. In E. J. Farris (ed.), The Care and

the manuscript. Many of the suggestions

Breeding of Laboratory Animals. John Wiley and

Sons, N.Y. made by these individuals were incorpo-

HASKINS, C. P., E. F. HASKINS, J. J. A. Mc-

rated in the final version. Tom Connell's

LAUGHLIN, AND R. E. HEWITT. 1961. Poly-

and Bryan Woods' tireless support in the morphism and population structure in Lebistes

laboratory made this and concurrent ex-

reticulatus, a population study, p. 320-395. In

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1248 DAVID REZNICK

W. F. Blair (ed.), Vertebrate Speciation. Univ.

Texas Press, Austin.

HELWIG, J. T., K. A. COUNCIL, AND P. S. REIN-

ecology of Euphorbia hirta Linn. J. Ind. Bot.

Soc. 39:455-473.

REZNICK, D. N. 1980. Life history evolution in the

HARDT. 1979. SAS User's Guide. 1979 ed. SAS

guppy (Poecilia reticulata). Ph.D. Dissertation.

Inst., Raleigh.

Univ. Pennsylvania.

1981. "Grandfather effects": the genetics

HIRSHFIELD, M. F., AND D. W. TINKLE. 1975.

Natural selection and the evolution of reproduc-

tive effort. Proc. Nat. Acad. Sci. USA. 72:222 7-

223 1.

HISLOP, J. R. G., A. P. ROBB, AND J. A. GAULD.

1978. Observations on effects of feeding level on

growth and reproduction in haddock, Melono-

gramnus aeglefinus (L.) in captivity. J. Fish. Biol.

13:85-98.

KALLMAN, K. D., AND M. P. SCHREIBMAN. 1973.

of interpopulation differences in offspring size in

the mosquito fish. Evolution 35:941-953.

REZNICK, D. N., AND J. A. ENDLER. 1982. The

impact of predation on life history evolution in

Trinidadian guppies. Evolution 36:160-177.

ROSENTHAL, H. L. 1952. Observations on repro-

duction of the poeciliid Lebistes reticulatus (Pe-

ters). Biol. Bull. 102:30-38.

SCHAFFER, W. M. 1974a. Selection for optimal life

A sex-linked gene controlling gonad differentia-

histories: the effects of age structure. Ecology 55:

tion and its significance in determining the age of

291-303.

1974b. Optimal reproductive effort in fluc-

sexual maturation and size of the platyfish Xi-

phophorus maculatus. Gen. Comp. Endocrin. 24:

289-304.

KLEIBER, M. 1975. The Fire of Life-An Intro-

duction to Animal Energetics. Robert F. Krieger

Publ. Co., Huntington.

LAW, R. 1979. Optimal life histories under age-

specific predation. Amer. Natur. 114:399-417.

LAW, R., A. D. BRADSHAW, AND P. D. PUTWAIN.

1977. Life history variation in Poa annua. Evo-

lution 31:233-246.

LILEY, N. R., AND B. H. SEGHERS. 1975. Factors

tuating environments. Amer. Natur. 108:783-790.

SCOTT, D. P. 1962. The effect of food quantity of

fecundity of Rainbow trout. J. Fish. Res. Bd.

Can. 19:715-731.

SEGHERS, B. H. 1973. An analysis of geographic

variation in the antipredator adaptations of the

guppy, Poecilia reticulata. Ph.D. thesis. Univ.

British Columbia.

STEARNS, S. C. 1975. The evolution of life history

patterns: a critique of the theory, a review of the

data, and a case study-Gambusia afnis in Ha-

affecting the morphology and behavior of guppies

waii. Ph.D. dissertation. Univ. British Colom-

in Trinidad, p. 92-118. In G. P. Baerends, C.

bia.

1977. The evolution of life history traits:

Beer, and A. Manning (eds.), Function and Evo-

lution in Behavior. Oxford Univ. Press, Oxford.

MACARTHUR, R. H., AND E. 0. WILSON. 1967.

a critique of the theory and a review of the data.

Ann. Rev. Ecol. Syst. 8:145-171.

1980. A new view of life history evolution.

The Theory of Island Biogeography. Princeton

Univ. Press, Princeton.

MAIORANA, V. C. 1976. Size and environmental

Oikos 35:266-281.

TINKLE, D. W. 1972. The dynamics of a Utah

predictability for salamanders. Evolution 30:599-

population of Sceloporus undulatus. Herpetolog-

613.

ica 28:351-359.

MERTZ, D. B. 1975. Senescent decline in flour bee-

tle strains selected for early adult fitness. Physiol.

Zool. 48:1-23.

MICHOD, R. E. 1979. Evolution of life histories in

response to age-specific mortality factors. Amer.

Natur. 113:53 1-550.

MORRISON, D. F. 1976. Multivariate Statistical

Methods. McGraw-Hill, N.Y.

NAEVDAL, F., M. HOHN, 0. INGEBRIGTSEN, AND

P. D. MOLLER. 1978. Variation in age at first

spawning in Atlantic salmon (Salmo salar). J.

TINKLE, D. W., AND R. E. BALLINGER. 1972.

Sceloporus undulatus: a study of the intra-specific

comparative demography of a lizard. Ecology 53:

570-585.

TURNER, C. L. 1941. Morphogenesis of the gon-

opodium in Gambusia affinis affinis. J. Morph.

69:161-185.

VINEGAR, M. B. 1975. Life history phenomena in

two populations of the lizard Sceloporus undu-

latus in southwestern New Mexico. Amer. Midl.

Natur. 93:388-402.

Fish. Res. Bd. Can. 35:145-147.

RAMAKRISHNAN, P. S. 1960. Studies in the aut-

Corresponding Editor: S. J. Gould

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

GUPPY LIFE HISTORY GENETICS 1249

APPENDIX. Means for the life history variables measured in this study. All means are adjusted for unequal

sample sizes (Helwig et al., 1979) and for covariates, where applicable.

Offspring weight (mg)a:

Locale

1st litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 1.02 0.93 0.81 0.88 0.91

Low 1.06 0.97 0.88 0.79 0.92

Locale means 1.04 0.95 0.84 0.83

Pred. means 0.99 0.84

2nd litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 1.04 1.03 0.90 0.93 0.98

Low 1.11 1.01 1.01 0.97 1.02

Locale means 1.08 1.02 0.96 0.95

Pred. means 1.05 0.95

3rd litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 1.13 1.20 1.04 0.98 1.08

Low 1.16 1.21 1.02 1.09 1.12

Locale means 1.15 1.20 1.03 1.03

Pred. means 1.17 1.03

Fecunditya (mg; dry weight):

Locale

1st litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 3.8 3.1 6.5 4.7 4.5

Low 4.6 1.3 5.6 4.2 3.9

Locale means 4.2 2.2 6.0 4.4

Pred. means 3.2 5.2

2nd litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 12.3 9.9 13.1 10.6 11.5

Low 9.9 8.8 10.7 8.9 9.6

Locale means 11.1 9.3 11.9 9.8

Pred. means 10.2 10.9

3rd litter: Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 18.3 17.8 19.3 16.9 18.1

Low 14.6 13.2 16.6 11.7 14.0

Locale means 16.4 15.5 17.9 14.3

Pred. means 16.0 16.1

Male age at maturation (weeks)b:

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 7.9 8.4 7.5 7.1 7.7

Low 8.6 8.7 7.8 7.1 8.0

Locale means 8.2 8.6 7.6 7.1

Pred. means 8.4 7.4

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions

1250 DAVID REZNICK

APPENDIX. Continued.

Male weight at maturation (mg; wet weight):

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 102.1 101.3 97.3 84.7 96.4

Low 97.9 97.6 85.6 83.2 91.1

Locale means 100.0 99.4 91.4 84.0

Pred. means 99.7 87.7

Female age at first parturition (days):

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 77.2 82.1 70.9 65.4 73.9

Low 79.0 89.2 74.5 75.0 79.4

Locale means 78.1 85.6 72.7 70.2

Pred. means 81.9 71.5

Female weight at first parturition (mg; wet weight):

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 230.8 335.5 242.5 213.3 255.5

Low 217.1 296.7 207.3 209.0 232.5

Locale means 223.9 316.1 224.9 211.1

Pred. means 270.0 218.0

Interbrood interval (days):

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High 25.7 24.9 23.1 22.0 23.9

Low 24.6 24.7 23.6 21.6 23.6

Locale means 25.2 24.8 23.3 21.8

Pred. means 25.0 22.8

Reproductive allotment (%):

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High .1784 .1452 .1712 .1362 .1577

Low .1663 .1345 .1629 .1172 .1452

Locale means .1723 .1398 .1670 .1267

Pred. means .1561 .1451

Reproductive effort (% consumed calories)c:

Locale

Riv 1 Riv 2 Cren 1 Cren 2 Food means

Food High .2650 .1606 .2688 .25 15 .2365

Low .2284 .1125 .2695 .2071 .2044

Locale means .2467 .1366 .2691 .2293

Pred. means .1917 .2505

Adjusted for female post-partum wet weight as a covariate.

Adjusted for wet weight when 25 days old as a covariate.

c These means are for the sums of the four separate estimates of RE (see text)

This content downloaded from 132.174.254.72 on Thu, 03 Mar 2016 15:29:41 UTC All use subject to JSTOR Terms and Conditions