Do placental species abort offspring? Testing an assumption of the Trexler–DeAngelis model


*Correspondence author. E-mail:


  • 1We investigate how resource level affects reproduction in matrotrophic (Poeciliopsis prolifica) and lecithotrophic (P. monacha) fishes.
  • 2One of our goals was to test an assumption of the Trexler–DeAngelis model for the evolution of matrotrophy, which was that matrotrophic species can adjust litter size by aborting offspring in low food conditions. Our more general goal was to elucidate other differences between the reproductive modes.
  • 3Both species have superfetation and c. 30-day development time. Females of each species were assigned to high or low food availability for 30 days, or one gestation period. Any young born during that time interval would have initiated development before the initiation of the experiment. If embryos are aborted, then this would be seen as a reduction in brood size in the low food treatment relative to the high food treatment within this period.
  • 4Our results suggest P. monacha responds to low food by sacrificing reproduction to maintain lipids, while P. prolifica maintains reproduction at the expense of lipids. Neither species showed a significant reduction in offspring number over the course of the experiment, suggesting that these species do not abort offspring in low food conditions.


Understanding variation in life-history traits is a central focus of research in evolutionary biology because these traits play such a direct role in determining fitness. A central feature of life-history traits is that they are often functions of one another; these functional relationships define the trade-offs that we think play a fundamental role in shaping the way life-histories evolve (Roff 1992; Stearns 1992). For example, fecundity is often a function of some other aspect of the organism, such as body size. As a consequence, there can be a trade-off between current and future fecundity that is mediated by the way the current investment in reproduction affects growth, future body size, and hence future fecundity (Roff 1992; Stearns 1992). Fecundity is also strongly influenced by the mode of reproduction, such as egg-laying vs. livebearing, and hence is likely to tilt the balance in the favour of the evolution of one or the other of these modes of reproduction (Tinkle 1969). Tinkle observed, for example, that livebearing lizards only produce one clutch of eggs per year, while egg-layers can produce four or more clutches per season, so livebearing implies a large potential loss of fecundity. He suggested that livebearing may most often evolve in those populations of lizards that normally only produce a single clutch of eggs per year because they would suffer little or no loss of fecundity. More generally, the reproductive mode we see is presumably an adaptation that maximizes fitness in the face of these trade-offs.

Many hypotheses have been proposed to explain the evolution of livebearing and many empirical research programmes have been devoted to the testing and development of these hypotheses (Shine 1980, 1983, 2004; Qualls 1997; Hodges 2004). However, a less visually apparent, but perhaps equally important difference in reproductive mode has to do with timing of provisioning. In viviparous species, timing of maternal provisioning ranges from strict lecithotrophy (yolk-feeding) to extreme matrotrophy (mother-feeding). In lecithotrophic species, females produce a fully provisioned egg that is fertilized, and then retained inside of the female throughout development. The only post-fertilization investment involves gas exchange and waste disposal. In matrotrophic species, the egg size at fertilization is greatly reduced and does not contain sufficient resources to sustain growth and development. Most nutrients are provisioned after fertilization, by means of oophagy, histophagy or placentation (Turner 1947; Wourms, Grove & Lombardi 1988; Trexler & DeAngelis 2003). Lecithotrophy is generally considered to be the ancestral state (Reznick, Mateos & Springer 2002), however, matrotrophy has evolved independently in at least 24 clades of viviparous vertebrates (Blackburn 1992). The shift between matrotrophy and lecithotrophy represents a distinct change in the timing of provisioning, making when and where resources for reproduction come from important in determining whether a reproductive bout is successful. The multiple independent origins of matrotrophy are suggestive of strong selection (Losos et al. 1998). While mammalian reproduction is the most well-known example of matrotrophy, it exists in a wide range of taxa, including reptiles, fishes, insects and plants (Turner 1947; Lloyd 1980; Wourms et al. 1988; Farley 1998; Jerez & Ramirez-Pinilla 2001; Chen & Caswell-Chen 2004). Despite this, very little theoretical or empirical work has addressed the factors that have shaped the evolution of matrotrophy (Trexler & DeAngelis 2003). Currently, only one model exists in the literature, and neither its assumptions nor its predictions have been empirically tested.

Trexler & DeAngelis (2003) used a combination of analytical and simulation techniques to produce the first non-verbal ecological model for the evolution of matrotrophy, focusing on what environmental conditions would favour matrotrophy over lecithotrophy, given a set of assumptions. They assumed that matrotrophy increases the number of offspring a female could have per reproductive bout. This is because a lecithotroph allocates all or most energy to an egg in a short interval of time before fertilization, while a matrotroph starts with smaller, less expensive eggs and allocates throughout gestation. From the same starting costs, a matrotroph can make more eggs initially and, given enough resources during the gestation period, will carry more embryos to term (Trexler & DeAngelis 2003). Additionally, they find that the conditions that favour the evolution of matrotrophy are more easily met if a female can diminish brood size via abortion, should resource levels drop below that required to sustain the brood. They consider a range of values for ability to resorb energy from these aborted embryos, and model short- and long-term lipid stores for each strategy. Finally, the model assumes that terminal offspring size is the same for both strategies, irrespective of resource availability.

The assumption that matrotrophs can abort embryos in the Trexler–DeAngelis model is an important one. Based on the model, without this ability, matrotrophy will be favoured only in an extremely narrow range of conditions, where food availability is consistently equal to, or higher than that needed to meet maintenance and reproduction needs. Any deviation from consistently high resource availability results in a catastrophic loss of fecundity because resources will be spread too thin and all offspring in a given brood will be lost. More generally, abortion of embryos has been suggested to confer a fitness advantage when optimal brood size is unknown or when the mother can identify the fitness potential of offspring during development and abort those with low potential (Lloyd 1980; Diamond 1987; Stearns 1987; Kozlowski & Stearns 1989).

However, empirical data on the advantage, or even existence, of embryo abortion in this context is lacking. For example, Borowsky & Kallman (1976) suggest that stress-induced abortion occurs in the platyfish, Xiphophorus maculatus because field-caught females had approximately twice the interbrood interval as laboratory-reared females. They interpreted this difference as indicative of abortion in the field-caught females. However, without common garden conditions, it is impossible to conclusively determine the cause of this difference. Trexler (1997) found that sailfin mollies, Poecilia latipinna, which are either lecithotrophic or facultatively matrotrophic, had fewer offspring at parturition than fertilization, regardless of whether they were reared on high or low food levels, which again suggests that some offspring were aborted between fertilization and birth. Meffe & Vrijenhoek (1981) concluded that abortion and resorption of embryos did not occur in several species of starved poeciliid fishes with a range of reproductive modes.

Seed plants have the equivalent of a placenta since extensive maternal provisioning occurs after the seed is fertilized. Lalonde & Roitberg (1989) looked for seed abortion due to decreased resource level in the dioecious plant Cirsium arvense. They defined abortion as an ovary that displayed pericarp development (ripened fruit wall), but that did not contain a healthy embryo. They found that abortion did occur, but the level of abortion was consistent between seeds in high and low resource conditions. Instead, significant differences in the level of abortion were found between clones, suggesting that abortion was due to non-viable gene combinations, rather than resource availability. Manzur et al. (1995) found that in humans, spontaneous abortion of 1–3 embryos occurred in 52·6% of women carrying triplets.

Our goal is to test the assumption of the Trexler–DeAngelis model that matrotrophic females have the ability to abort offspring under conditions of low food. Here we do so by determining how matrotrophs and lecithotrophs differ in their response to food levels. We predict that lecithotrophs will not abort offspring in response to reduced food availability because nutrients are pre-packaged before fertilization. In contrast, we predict that matrotrophs will abort offspring in response to reduced rations because they require a continuous supply of nutrients to sustain developing young. We examined closely related matrotrophic and lecithotrophic species of Poeciliopsis fishes to determine whether embryo abortion occurs in these respective reproductive modes.


The genus Poeciliopsis consists of small, livebearing fishes with internal fertilization. This genus contains reproductive strategies that range from lecithotrophy to extensive matrotrophy, with well-developed placentas (Turner 1940; Reznick et al. 2002). Because of this diversity of modes of reproduction, they are an ideal group to study matrotrophy and the evolution of post-fertilization provisioning. We used Poeciliopsis prolifica (matrotrophic) and P. monacha (lecithotrophic) to investigate differences between reproductive modes. These species are more closely related to one another than each would be to another species with the same reproductive mode in the genus. A phylogenetic reconstruction of the evolution of maternal provisioning in this genus shows that P. prolifica represents one of the three independent origins of extensive matrotrophy in Poeciliopsis (Reznick et al. 2002). The offspring of the P. prolifica in our experiment display an eightfold increase in dry mass between fertilization and birth (Pires, McBride & Reznick 2007). This increase is associated with an elaboration of maternal and embryonic tissues, which function together as the equivalent of a placenta (Turner 1940). In contrast, the embryos of P. monacha lose c. 40% of their dry mass during development, presumably reflecting the costs of metabolism and anabolism (Thibault & Schultz 1978; Reznick et al. 2002). This degree of weight loss is comparable to what is seen in the mass of a freshly laid egg vs. newly born embryo in egg-laying species (Wourms 1981) and hence is interpreted as lecithotrophy. Both species have superfetation, or the ability to carry multiple broods in different stages of development, making them virtual conveyor belts of offspring. This trait allowed us to look at the effect of a given experimental treatment on embryos that are in different stages of development. Both species have the ability to store sperm for an indeterminate amount of time (Winge 1937; Turner & Snelson 1984; Constanz 1989) allowing us to keep females in isolation during the course of a short-term experiment without remating.

To determine the effect of food on developing embryos in matrotrophic and lecithotrophic species, we used a factorial design with factors being food level (high/low) and reproductive mode (matrotrophic/lecithotrophic) as fixed factors. The gestation period for any given litter is c. 30 days (M.N. Pires and D.N. Reznick, unpublished data), so the effect of food reduction on developing offspring will be reflected in offspring born within 30 days of the initiation of the treatment. For each female, we recorded 15 days of reproduction on high food levels to establish a baseline, or ‘before’ rate of reproduction. Females were then randomly assigned either high or low rations. This treatment lasted 30 days, or one gestation period. A key to our design is that when a female begins her treatment, she will contain multiple litters of young that are already developing. All of the young born during the course of the experiment will be from eggs that were fertilized before the experiment began. Young that were in an advanced stage of development will be born early in the experimental period while those that were in earlier stages will be born progressively later. Because each litter was exposed to reduced maternal food availability for different durations of development, we will also be able to ascertain not only if abortion occurs, but also whether or not the stage of development of the embryo at the time of food deprivation affects whether or not it will be aborted. Thus, a reduction in brood size in the low food group as compared to the high food group is interpreted as the abortion of developing embryos due to reduced resource level. A subset of females was kept in the experiment for 30 additional days. Offspring born during this second 30-day interval were from litters that were initiated after experimental food treatments began and reflect the added effect of food level on the number of eggs that are fertilized. Poeciliopsis monacha sample sizes were 16 and 19 females for high and low food groups, respectively. Poeciliopsis prolifica had 22 females per treatment group. See Fig. 1 for a schematic of the design.

Figure 1.

Schematic of experimental design. For each species, 15 days of data were collected before the experimental treatments began. Treatments began when approximately half of each species was switched to low food, and lasted 30 days, or approximately one gestation period. Thus, offspring born within the treatment period were initiated before the treatment began, allowing us to see the effect of food level on already developing offspring. Analyses divide data into ‘pre’, ‘post 1’, and ‘post 2’ time periods. A subset of females was allowed to remain in the experiment for an additional 30 days (not shown in figure).

The P. prolifica were the second and third laboratory generation derived from adults, collected from el Palillo River in 2004. Poeciliopsis monacha were from a laboratory population that was derived from adults, collected from the Rio Fuerte drainage and that had been in laboratory culture since 2001. Females of similar size and age were reared with mature males in 38-liter community tanks and were fed ad libitum with liver paste in the morning and brine shrimp in the evening. Once actively reproducing, females were isolated in 8-liter aquaria and fed high food rations (just under average ad libitum levels for the laboratory population). Females were kept on quantified food for one gestation period, or c. 30 days, before the study began, to ensure that broods used in the analysis were initiated after quantified food rations began. We then collected 15 days of baseline data, followed by the 30-day treatment.

We used Hamilton micropipettes to ensure accurate allocation of food. For P. prolifica, high food levels were 40 µL per meal, and low food levels were 15 µL per meal. For P. monacha, high food levels were 50 µL per meal, and low food levels were 15 µL per meal. Initially, all high food treatments were 50 µL, but it was discovered that P. prolifica females were not eating all rations, so the level was reduced to prevent detrimental effects in water quality due to overfeeding. Poeciliopsis monacha are larger on average, which probably accounts for the differences in the rate of food consumption.

We measured length and mass of females on days 1 and 30 of the treatment. Females were preserved at the end of day 30. All offspring born after females were isolated in 8-liter aquaria were preserved immediately after birth. Fish were euthanized using an overdose of MS-222 and preserved in 5% formaldehyde. The main dependent variable of interest for determining whether abortion occurs was offspring number. However, matrotrophic females may respond to low food in a number of unpredicted ways, including extending the gestation period, producing smaller, leaner offspring, or by sacrificing lipid stores to provision young. Thus, we measured female dry mass, female fat content, offspring dry mass and offspring fat content following the methods in Reznick & Yang (1993) to provide a more complete picture of the reproductive biology of each species. A repeated measures ancova was used to analyse offspring number, with species and food treatment as independent variables, and female size as a covariate. The repeated measures were the 15 days immediately preceding the treatments, days 1–15 of treatment, and days 16–30 of treatment. A similar anova was used to analyse offspring dry mass and offspring fat content, where the covariate was not needed. A two-tailed t-test was used to analyse changes in interbrood interval, reproductive allotment, degree of superfetation, female dry mass, female fat content and number of developing young. Some analyses below omit females either because she did not give birth in a given time interval, or because she was part of the subset that was allowed to continue in the experiment for 60 days. Samples sizes are smaller for offspring dry mass and fat content analyses because several broods were inadvertently destroyed during processing. These deletions are noted in the degrees of freedom, reported below.


female size

Low food females weighed approximately three quarters of their high food counterparts at the end of the experiment, despite no differences initially (repeated measures anova: P. monacha (lecithotrophic): F(1,33) = 91·184, P < 0·001; P. prolifica (matrotrophic): F(1,42) = 96·724, P < 0·001), indicating that food levels were sufficient enough to invoke a response (Table 1).

Table 1.  Summary of means, standard errors (in parentheses) and sample sizes for dependent variables. Pre, post 1 and post 2 represent 15 days prior to treatment, days 1–15 of treatment and days 16–30 or treatment, respectively. For repeated measures tests, sample sizes are listed on the first line only
Dependent variablePoeciliopsis prolificaPoeciliopsis monacha
  • Represents dissection data at the end of the 30-day treatments. Thus, superfetation is an index of how food level affects the degree of superfetation on broods that were initiated after the food treatments began.

Female mass (g)
 Before treatment0·499 (0·023) 220·504 (0·023) 220·929 (0·031) 160·940 (0·024) 19
 After treatment0·554 (0·025)0·420 (0·017)0·935 (0·025)0·731 (0·017)
Offspring number
 Pre12·00 (1·26) 2211·86 (0·68) 229·88 (0·93) 1610·42 (0·98) 19
 Post 18·18 (0·90)8·41 (0·56)7·63 (1·08)8·42 (1·28)
 Post 28·95 (0·96)8·00 (0·82)7·94 (1·98)8·79 (2·02)
Offspring mass (g)
 Pre8·11e−4 (3·0e−5) 218·93e−4 (3·9e−5) 221·834e−3 (6·4e−5) 131·694e−3 (6·2e−5) 19
 Post 18·64e−4 (3·6e−5)8·00e−4 (2·3e−5)1·803e−3 (6·4e−5)1·598e−3 (4·3e−5)
 Post 29·05e−4 (4·2e−5)7·70e−4 (4·4e−5)1·786e−3 (4·6e−5)1·577e−3 (6·6e−5)
Offspring fat content (%)
 Pre9·39 (1·09) 2010·92 (2·12) 1918·17 (1·64) 1217·53 (1·26) 15
 Post 18·47 (2·25)7·48 (0·72)14·83 (1·57)14·14 (1·74)
 Post 212·32 (2·28)7·98 (0·87)11·78 (1·47)11·84 (1·48)
Change in interbrood interval0·43 (0·62) 211·55 (0·76) 221·13 (1·53) 131·17 (0·94) 17
Female % fat17·35 (1·11) 1810·09 (1·05) 165·00 (1·50) 132·82 (0·65) 13
RA0·11 (0·01) 180·11 (0·01) 170·27 (0·01) 130·17 (0·02) 13
Superfetation3·89 (0·31) 183·88 (0·31) 172·69 (0·17) 131·84 (0·15) 13
Total no. of developing young18·83 (3·61) 1812·76 (1·96) 1716·15 (1·88) 137·92 (1·27) 13

offspring number

All treatment groups showed a decline in fecundity between the pre-treatment and treatment period. This may be due to a net decline in food availability after the fish were isolated in 8-liter tanks and kept on quantified rations. No difference in offspring number was detected between food treatments for either species for the 30-day treatment period (Table 1; Fig 2; P. monacha (lecithotrophic): F(2,31) = 0·003, P = 0·997; P. prolifica (matrotrophic): F(2,40) = 0·337, P = 0·716, using females size as a covariate). Poeciliopsis prolifica does show a slight, non-significant decline in offspring production in the low food group relative to the high food group towards the end of the 30-day treatment period (Fig. 2b, partial-η-squared 0·017). The absence of a difference in fecundity suggests an absence of the abortion of developing embryos in both species. We did not perform formal analyses of the reproductive data for the subset of females that were retained in the experiment for an additional 30 days because of the small sample sizes (3–5 per treatment group) but the trends show that fecundity declines in the low food treatments of both species, likely because fewer offspring are initiated per litter. The rate of decline in P. prolifica was more gradual than in P. monacha; P. monacha ceased reproduction by the end of the 60-day interval (See Supplementary Fig. S1).

Figure 2.

Offspring number over time in high and low food (a) P. monacha and (b) P. prolifica. Times marked ‘pre’ represent the 15 days prior to treatment start. ‘Post 1’ represents days 1–15 of treatment, and ‘post 2’ represents days 16–30 of treatment. No significant differences were found between food groups in either species.

offspring mass

Response in dry mass of offspring did not differ between treatment groups for lecithotrophic P. monacha (Table 1; Fig 3a; F(2,26) = 0·807, P = 0·457). In matrotrophic P. prolifica, dry mass of offspring from high food females increased over the course of the experiment. In low food P. prolifica females, dry mass of offspring decreased (Fig 3b; F(2,40) = 0·8304, P = 0·001). This decline in offspring size in the low food treatment implies that P. prolifica females are relying on food consumption to support the growth of developing young rather than relying solely on fat stores.

Figure 3.

Offspring dry weight over time. (a) In P. monacha groups differed for unknown reasons before food treatments began, but there were no significant differences in response to food level between treatment groups. (b) High and low food groups of P. prolifica differed significantly in offspring dry mass over time.

offspring fat content

The composition of offspring born during the course of the experiment was not affected by food level. Offspring from high and low food females showed a similar percentage of lipids within each species (Table 1; P. monacha (lecithotrophic): F(2,24) = 0·031, P = 0·966; P. prolifica (matrotrophic): F(2,36)= 1·306, P = 0·284).

interbrood interval

Interbrood interval was analysed by taking the difference in days between the last full interval before treatments began and the last full interval before the experiment ended, and then comparing this number between high and low food groups. Interbrood interval did not differ between the food groups for either species (Table 1; P. monacha (lecithotrophic): t(28) = −0·013, P = 0·990; P. prolifica (matrotrophic): t(41) =–1·132, P = 0·264), indicating that neither species responds to low food by lengthening development time. Differences between species in interbrood interval is due to P. prolifica having more developing broods of young and hence a shorter interbrood interval than P. monacha.

female lipids

At the end of the experiment, P. monacha (lecithotrophic) showed no significant difference in composition between food groups (Table 1; t(24) = 1·336, P = 0·194). In contrast, high food P. prolifica (matrotrophic) had a significantly higher proportion of lipid content than low food P. prolifica (Table 1; t(32) = 4·714, P ≤ 0·001).

reproductive allotment

Low food P. monacha (lecithotrophic) females showed a significant reduction in reproductive allotment (reproductive dry mass/total dry mass) at the end of the 30-day treatment compared to high food females (Table 1; t(24) = 4·901, P < 0·001). Reproductive allotment in P. prolifica (matrotrophic) was not statistically different between food groups (Table 1; t(33) = −0·630, P = 0·533).


At the end of the 30-day treatments, lecithotrophic P. monacha showed a significant decrease in the number of developing litters in the low food treatment compared to the high food treatment (Table 1; t(24) = 3·633, P = 0·001). Food level did not have an impact on the number of developing litters in matrotrophic P. prolifica (Table 1; t(33) = 0·015, P = 0·988).

number of developing young

Developing young present in females at the end of the 30-day experiment would be ones who initiated development after the experiment began. Dissection of lecithotrophic P. monacha females revealed that low food females had fewer developing young than high food females (Table 1; t(24) = 3·624, P = 0·001). There was no difference between high and low food groups in matrotrophic P. prolifica (Table 1; t(33) = 1·453, P = 0·156). Stage zero eggs (yolked eggs with no embryonic development) were not used in this analysis because it is difficult to determine litter size before eggs are fully yolked and fertilized.


We did not find evidence of abortion in matrotrophic species. All offspring born in the course of the 30-day treatments were initiated before the treatments began. If abortion occurs, then low food groups should have a reduction in fecundity as compared to the high food control groups. However, neither species showed significant differences in offspring number between the high and low food treatments. If this is a general property of the Poeciliidae, then the conditions that favour the evolution of matrotrophy under the Trexler–DeAngelis model are much narrower, and matrotrophy will be favoured only when food availability is consistently equal to or higher than that needed to meet maintenance and reproduction needs. Deviation from consistently high resource availability will result in resources being spread too thin and all offspring in a given brood will be lost.

There was a hint of a reduction in the fecundity of the low food treatment of matrotrophic P. prolifica towards the end of the treatment period, which would suggest that they might have some ability to abort embryos, but only ones that are early in development. If P. prolifica is indeed capable of such embryo abortion, their ability is limited and a much larger experiment and/or brood size would be required to perceive it as statistically significant. Alternatively, it may be that abortion is a strategy that only occurs in more extreme conditions. Even restricting abortion to more extreme conditions would reduce the scope of conditions that favour the evolution of matrotrophy.

Trexler and DeAngelis modelled an organism that lacked superfetation and that produced offspring of constant size. The reality of the mode of reproduction in our study organisms leaves them with an enlarged spectrum of responses to a reduction in food availability. Matrotrophic species responded to a switch to low food by producing smaller offspring in those broods that were initiated before food level dropped. As expected, developing broods in the lecithotroph were buffered from the reduction in food level, and showed no decrease in mass. The decrease in mass in P. prolifica is due to an overall reduction in offspring size, rather than a reduction in lipid content. This reduction is consistent with findings in another matrotroph, Heterandria formosa. Reznick, Callahan & Llauredo (1996) found that H. formosa responded to low food by producing smaller offspring of similar composition, and suggested that this may be due to an inability of matrotrophs to allocate fat reserves to developing offspring. The two species did not respond to reduced food by modifying the rate of development of their young since the interbrood intervals of the high and low food treatments were the same.

Producing smaller offspring is likely to be maladaptive in low resource environments. The production of larger offspring has been predicted to be adaptive in resource poor environments (Smith & Fretwell 1974; Sibly & Calow 1983; McGinley, Temmer & Geber 1987; McGinley & Charnov 1988; Laurie & Brown 1990; Lalonde 1991; Charnov, Downhower & Brown 1995; Einum & Fleming 2004), and many studies have provided empirical evidence in a variety of species to support the predictions (Ferguson & Fox 1984; Gliwicz & Guisande 1992; Parichy & Kaplan 1992). Hassall et al. (2006) found that under favourable conditions, maternal fitness in the grasshopper Chorthippus brunneus was highest when females produced a large clutch with small eggs. However, under poor conditions, maternal fitness was highest when each clutch consisted of few, large offspring. Hutchings (1991) found that in brook trout, larger offspring had higher survival in low food conditions, but this advantage was reduced in high resource conditions. Given this evidence, a matrotroph that produces smaller offspring in low food conditions will likely have a selective disadvantage compared to a lecithotroph that does not reduce offspring size in low food conditions.

Reznick et al. (1996) found that lecithotrophic species (Po. reticulata, Priapicthys festae) responded to a reduction in food availability by producing larger offspring. Bashey (2002) showed that this size increase is likely to represent adaptive phenotypic plasticity because of the fitness advantage that these larger babies have in low food environments. The fact that matrotrophic species seem to respond to a sudden reduction in food availability in the opposite fashion suggests that this response represents a non-adaptive constraint associated with the evolution of matrotrophy, which would again restrict the conditions that favour the evolution of matrotrophy. Lecithotrophic P. monacha is expected to show a similar adaptive plasticity in low food conditions as Po. reticulata and Pr. festae. A formal test of this plasticity is in progress.

Dissection data of females at the end of the experiment give insight into how the food level would affect broods that were initiated during the treatment regime. Low food P. monacha (lecithotrophic) had reduced reproductive allocation, degree of superfetation and litter size, but not a reduction in lipid stores. Low food P. prolifica (matrotrophic) showed a reduction in the percent of lipids, but not in the number of developing broods of young, litter size or in reproductive allocation. This suggests that P. monacha responds to low resource conditions by defending maintenance at the expense of reproduction. Poeciliopsis prolifica instead reduces its reserves in favour of maintaining reproduction.

If our results represent a general trend for all matrotrophs, then they indicate a reduction in the range of conditions where a matrotroph could out-compete a lecithotroph compared to that suggested by the Trexler–DeAngelis model. A maladaptive response in terms of offspring size makes some sense in the light of Trexler and DeAngelis’ prediction that matrotrophy evolved in a consistent, high resource environment. If this is the case, then selection for such an adaptive response to low food should be relaxed. Alternatively, if matrotrophy requires a consistent level of food availability to produce appropriate sized offspring, then it may represent a constraint that limits the ability of matrotrophic species to persist in variable environments. This conclusion is consistent with Thibault & Schultz's (1978) narrative description of the habitats of P. monacha, P. prolifica and P. lucida. Lecithotrophic P. monacha is found in the harshest environments, where there are ‘dramatic seasonal fluctuations’ in temperature, water levels, light intensity and resource availability. Matrotrophic P. prolifica is found in more specialized habitats, often deep, permanent pools, where food resources are at consistently high levels. Poeciliopsis lucida, which has an intermediate level of matrotrophy, is found in environments of intermediate resource availability and predictability (Thibault & Schultz 1978).

Marsh-Matthews & Deaton (2006) looked at the effects of food level on the reproduction of Gambusia geiseri also with the goal of evaluating aspects of the Trexler–DeAngelis model. Gambusia geiseri has been termed an incipient matrotroph, meaning that while it is primarily lecithotrophic, it may provision very low levels of nutrients after fertilization if resource levels are high enough. They reared fish on high and low food treatments beginning c. 44 days before exposure to males. They found that high food females had larger brood sizes, larger embryos and a higher rate of nutrient transfer (measured by the injection or radiolabelled nutrients). They found no difference in the probability of embryo abortion, observed directly via dissection, between treatments. The increase in embryo size was attributed to incipient matrotrophy. This result is interpreted as support for the Trexler–DeAngelis model, since they predicted that matrotrophy is more likely to evolve in the presence of high resource availability. It differs from the model because they are reporting on phenotypic plasticity, while Trexler and DeAngelis were predicting the circumstances that favour the evolution of the trait.

Both the Trexler (1997) study on sailfin mollies, and the Marsh-Matthews & Deaton (2006) study looked for evidence of abortion in facultative matrotrophs that had been reared on high and low levels of food availability prior to the initiation of the litters that were the dependent variables in their experiment; both studies found evidence for abortion, but it was independent of food level. Our study instead quantifies the potential abortion caused by variation in food availability while the young are developing. The differences between the experimental designs can be thought of as possible scenarios that a pregnant female may encounter, and likely represents a small portion of the variation that might be encountered in nature.

The overall contrast in the life histories of P. monacha and P. prolifica are reminiscent of a more general trend that we see in life-history evolution. Lecithotrophic P. monacha responds to resource restriction by maintaining itself at the expense of reproduction while matrotrophic P. prolifica responds by maintaining reproduction at the expense of maintenance. Elsewhere it has been shown that P. prolifica is younger at first reproduction and produces offspring at a higher rate (Thibault & Schultz 1978). Poeciliopsis prolifica thus has the kind of life history that is predicted to evolve in environments with high extrinsic mortality (Charlesworth 1994). If this were a general property of species with matrotrophy, then it may be that placentation evolves because it facilitates earlier maturity; the quantity of resources that are required to initiate a litter of young and their volume at initiation is smaller, allowing development to begin in a smaller, younger individual. Vitt & Blackburn (2002) have suggested that placentation evolved in the lizard genus Mabuya for this reason, but these sorts of options have yet to be considered in a theoretical model. A positive outcome of this study is that it has highlighted the additional degrees of freedom that are available to organisms that were not considered by Trexler and DeAngelis, and hence has defined ways in which the model can be expanded in the future. Specifically, for matrotrophs, the assumption that terminal offspring size is the same should be changed to reflect the initial reduction in offspring mass in response to low food conditions, as shown in P. prolifica. Additionally, Trexler and DeAngelis assume that lecithotrophs draw lipid stores down to nearly zero before each reproductive bout. However, data from P. monacha indicate that lipid stores were maintained at an expense to reproduction. Finally, Trexler and DeAngelis did not include superfetation in their model. If all of these factors were incorporated in their simulations, they may change the relative advantage of the lecithotrophic strategy over the matrotrophic strategy in low food conditions.

While our goal is to address the general phenomena of matrotrophy and lecithotrophy, our study instead considers a single pair of species that differ in this fashion. It thus does not represent an absolute explanation of the differences between the two modes of maternal provisioning. The virtue of doing this kind of research in the genus Poeciliopsis is that it harbours three independent origins of extensive matrotrophy (Reznick et al. 2002). There is a second cluster of related species that also contains sister species that either do or do not have matrotrophy. These same experiments can thus be repeated on new species and the generality of the results can be evaluated.


We thank Kevin McBride, Tara Mastro, Samantha Natividad, Yuridia Reynoso, Trinity Ryan and many undergraduate volunteers for help in the laboratory. Reznick laboratory members and two anonymous reviewers provided valuable feedback on earlier versions of the manuscript. We thank Christopher Oufiero and Andrew Stoehr for constructive discussions. Financial support was provided by NSF, Division of Environmental Biology, Population, and Systematics Programs (grant no. DEB0416085).