Populations of Gambusia affinis differ substantially in predation on newborns, birth weight, and interbrood interval. Populations of G. geiseri also differ significantly, but to a lesser degree, in the same three factors. Populations of G. nobilis differ in predation and birth weight. These factors are not concordant among species and localities; thus these variations are genetic, not environmental. Overall, G. affinis is less predaceous, has less heavy young, and has shorter interbrood intervals than G. nobilis, and G. geiseri is intermediate. However, the factors can be reversed by choosing extreme populations. Data are available on interbrood intervals for six species; they range from long to short by G. nobilis, G. heterochir, G. gaigei, G. geiseri, G. speciosa, and G. affinis. In the same sequence, birth weights vary from heavy to light and predation rates vary from major to less. However, intraspecific variation is not correlated among these factors. Predation susceptibility varies similarly with the heavier newborn being less likely to be eaten than the lighter newborns. However, predation on the relatively heavy but less active Poecilia young is the greatest of all.

Las poblaciones de Gambusia affinis difieren substancialmente en la depredación de reciennacidos, peso al momento de nacer e intervalo de las gestaciones. Las poblaciones de G. geiseri tambien difieren significativamente, pero a un grado meno, en los mismos tres factores. Las poblaciones de G. nobilis difieren en depredación y peso al momento de nacimiento. Estos factores no son concordantes entre especies y localidades; por lo tanto estas variaciones son basadas en genética, no en medio ambiente. En general G. affinis es menos depredador, tiene crías con menos peso y tiene intervalos de gestación más cortos que G. nobilis, mientras tanto G. geiseri se encuentra entre ambos. Sin embargo, los factores pueden ser reversidos al escoger poblaciones extremas. Data de los intervalos de gestación de seis especies es disponible; estos se extienden de largos a cortos de las especies G. nobilis, G. heterochir, G. gaigei, G. geiseri, G. speciosa y G. affinis. En las misma sequencias, el peso al momento de nacimiento varía de pesado a liviano y las proporciones de depredación varían de mayor a menor. Sin embargo, variación intraespecífica no está correlacionada entre estos factores. Susceptibilidad en la depredación varía similarmente con las crías que son más pesadas al haber menos posibilidad de ser comidos que los son más ligéros. Sin embargo, depredación en las relativamente pesadas pero menos activas crías de Poecilia es la más grande de todas.

Traditionally, reports on biological problems emphasize the species studied. Some reports, mostly recent, have demonstrated intraspecific variation of life history traits. The classic lizard studies by Tinkle (1969) have shown that variation correlated with environmental factors such as the latitude at which the stocks lived. A few fish studies (Haskins et al., 1961; Houde, 1988; Foster, 1994; Endler and Houde, 1995; Reznick et al., 1996; Reznick and Bryga, 1996) have shown variation in reproductive or predatory activities. Previously, I have shown that populations of Gambusia affinis vary in predation on congeneric newborns (Hubbs, 1992) that can be called cannibalism (Hubbs, 1991). In this report I expand on those data by using two other life history traits: birth weight of young and time between broods from females isolated from males. I also expand the number of species used for interpopulation comparisons to include Gambusia geiseri and G. nobilis. This report also includes comparable data on one population of G. speciosa, G. senilis, G. longispinis, G. sp., G. heterochir, and G. gaigei. The last three species have (or had) limited geographic ranges and consequently only one population could be used. Six of the species (G. nobilis, G. speciosa, G. longispinis, G. senilis, G. gaigei, and G. sp.) are native to the Chihuahuan Desert and five are listed as endangered by the appropriate federal government (all but the extinct G. sp.).

Gambusia affinis is widely distributed in the south-central North America. It is used extensively by public health agencies as a biological control for mosquitos. It is possible that some of my populations may have been introduced (one certainly has), but the intraspecific variation in three life history traits suggests limited replacement of native populations by introduced stocks.

Gambusia nobilis occurred in several spring-fed waters in the Pecos valley of West Texas and eastern New Mexico. Presently it occupies two New Mexico areas: Blue Spring and the Bitter Lakes National Wildlife Refuge. It also occurs in two areas in Texas: Diamond-Y Spring and the spring complexes around Balmorhea (Hubbs and Springer, 1957). I have samples from all of these regions except Blue Spring. Two, Bitter Lakes and the Balmorhea complex, have two or more separate populations studied.

G. geiseri occurred in two spring areas in Central Texas: Comal and San Marcos springs. Some time about 1930, stocks were widely released elsewhere in Texas, presumably by public health agencies (Hubbs and Springer, 1957).

Consequently, I have data on native populations of two species (G. affinis and G. nobilis) and data on two native populations and six introduced sixty years ago of a third species (G. geiseri).

Materials and Methods

Stocks were obtained from 48 populations of Gambusia affinis, 7 of G. nobilis, 10 of G. geiseri, and one each of the other six species (Table 1). Five populations had only adults used. Additionally, predation studies were also performed using Poecilia young. Most of the samples were from widely-distributed localities in Arizona, New Mexico, Texas, and Arkansas (Map 1). Several localities were relatively close together (Map 1, Insets). The fish were brought into the Austin laboratory and fed heavily with Tetra Min and Drosophila larvae and adults.

Predation studies: Three to twenty newborn young were randomly isolated in aquaria with one adult that had been in the laboratory at least one week. Each aquarium (49 by 15 by 17 cm (deep)) had an airstone and a series of snails (Physa) that could consume excess food and provide supplemental food for the fish (Hubbs, 1990). Each experiment was fed ca. 20% of total fish biomass daily with flake food and with Drosophila larvae. The surviving young were counted 31 days later. About 7% of the experiments had no adult at the end of the 31-day interval; those experiments were excluded from the predation data presentations (Table 2). Gambusia seldom live more than two years, and it is presumed that most of the adult mortality was from natural causes. It is likely that some natural mortality occurred with young as well. I presume that 90% survival is close to the maximum survival rate possible in these types of experiments. At the end of an experiment, the adult was returned to the stock tank and may have been used again. As ten or fewer young were used in most experiments, the number of experiments is about 10% of the number of young reported. The data reported here are survival percentages. Predation would be the inverse.

Young used in predation experiments were tested with available adults. For example, Woman Hollering young might be tested with Egg Nog adults on one date and Cow Creek adults the next. Similarly, Cost adults could prey on Too Much Pond young and then on Fairy young at a later date, or Cut 'n' Shoot combined with Hog Eye and then Uncertain.

Birth weight: Females were isolated in floating breeding chambers that were checked daily. Most of the young were used in predation studies, but some young (at random) were placed in a 40deg.C drying oven for at least two days and weighed. The birth weight of some individuals in a brood might vary by as much as 800%, but variation among individuals was usually about 10%. The average weight of individual young was recorded. Sample sizes are listed in Table 1.

Interbrood interval: Isolated females were continuously maintained in the breeding chambers. The date of birth of the first individuals in a brood was recorded. The occasional young that were found on the next day were considered part of that brood. A female would then have another brood after an interval of more than three weeks. That was considered a second (third, etc.) brood, and the difference between the dates is reported as the interbrood interval. Occasionally, a female captured during a nonbreeding interval in nature would have a second brood shortly (5 - 10 days) after the first. That interbrood interval was excluded from the data. If a population had more than 10 interbrood intervals, the longest and shortest were deleted, if more than 20, the two longest and shortest were deleted, etc. Sample sizes are listed in Table 1.

Results

Predation: Although many authors (Seale, 1917; Krumholz, 1948; Koster, 1957; Myers, 1965; Axelrod and Schultz, 1971 and 1983; Minckley, 1973; Walters and Legner, 1980; Schoenherr, 1981; Harrington and Harrington, 1982; Meffe and Snelson, 1989) have reported that Gambusia adults avidly prey on their own young, Hubbs (1992) reported that half of the experimental young survived a week. Additional data provide a similar result (Table 3). Similarly, I have shown that young with male predators had higher survival than those with females (70% versus 30%) (Table 4). I also contrasted the individual predator-predation comparisons and got the same results (Table 5). This excluded the possibility that tests with one sex as predator came from a more predaceous population than those of the other. Many more individual tests had chi-squared values favoring males having a lower predation rate. It should not be surprising that two of more than 800 contrasts were statistically (p > 0.01) the converse of the majority. The same statistical level favoring the conclusion that females were more predaceous was obtained by 394 contrasts. Consequently, predation by adults on young is primarily by Gambusia females.

Predation by males on poeciliid young had results similar to but more extensive than those reported previously (Table 6). Survivorship of Poecilia young exposed to G. affinis predation was the lowest of any with a sample size of over 100 young. Similarly, survivorship of Poecilia young was lowest in one of the other three comparisons. Predation rate may be affected by feeding by the predator or escape by the prey (Fuiman and Magurran, 1994). Although Poecilia young are relatively large, they are less active in aquaria than are Gambusia young. The second-lowest survivorship with male G. affinis as predator was conspecific. Predation by G. nobilis males was relatively high as young survival with G. nobilis as predator was lowest in all but three (G. nobilis with G. nobilis, G. heterochir and Poecilia young) with sample sizes of more than 50. In general, G. nobilis, G. heterochir and G. gaigei males ate more young that did males of G. affinis, G. speciosa, and G. geiseri. Conversely, G. nobilis, G. heterochir and G. gaigei young had a higher survival rate than those of the other three species.

Similarly, predation by females followed the same pattern as that reported previously (Table 7). Survival of Poecilia young exposed to G. affinis females was lower than that of any Gambusia species. Predation by other species was similar, with most survival of G. affinis young among the lowest. Predation on G. affinis young by congeneric females was substantially higher than that of males. Congeneric predation on G. affinis young was higher with G. geiseri, G. nobilis or G. gaigei females than using G. affinis, G. speciosa, or G. heterochir females. In this test, G. geiseri went from low to high and G. heterochir went from high to low in comparison with male predation. Again, survivorship with G. nobilis females was lowest in four of the six tests with sample sizes over 100. The exceptions were G. nobilis and G. gaigei young, where G. nobilis predation was second highest, and G. nobilis with G. heterochir young. Again, G. nobilis, G. heterochir and G. gaigei young had a higher survival rate than those of the other three species. Conspecific predation was highest by G. affinis females (= cannibalism); G. heterochir and G. gaigei conspecific predation was also quite high.

Extensive variation of survivorship occurred when G. affinis young were exposed to female G. affinis predation (Table 8). In two tests (Lost River and Big Bend), more than two-thirds of the young survived. In contrast, in seven tests (Bitter #3, Falcon, El Tigre, Fairy, Junction, Middle Creek, and Uncertain), survivorship was below 10%. Survivorship with males as predators tended to be similar with the two high survivors having 79 and 88% survivorship and the six with low figures having 85, 76, 66, 73, 64, 77, and 68% survivorship. Six of the seven survival percentages with male predators were lower than either of the two with high survivorship with female predators. In two other populations (Pecos and Clear Creek) where the survivorship with female predators had percentages above 50, the comparable male tests had survivorships of 81 and 85%. Similarly, Big Bend and Junction females were tested with G. speciosa and G. geiseri young. In each instance, survivorship with Big Bend females was higher than those with Junction females (61 vs. 18% and 73 vs. 8%). These results remained consistent whether either of two sets of field-caught females or laboratory-raised females (from Big Bend) were used. There is a distinct difference in predation on newborns depending upon the population of adults used. Clearly, the choice of G. affinis stocks used for mosquito control would have great influence on average predation on young fishes. It is possible that G. affinis predation on other prey such as mosquitos may vary among populations equally. Such tests by mosquito control agencies are now mandated.

The variation of predation by G. affinis adults does not have a geographic or ecologic pattern. The two New Mexico populations are from the Bitter Lakes National Wildlife Refuge: one is in the low and the other in the high group. Lazy Pond is one kilometer from the Pecos River site. Survivorship with female predators differs by 43% and is significant at beyond the 0.00001 level ( ² = 81). Survivorship based on individual tests is also significant beyond the 0.001 level. Too Much Pond is intermediate geographically and in predation rate.

The four high-survival populations (Lost River - saline, stenothermal; Big Bend - low salinity, stenothermal, elevated temperature; Pecos - moderate salinity, eurythermal; Clear Creek - low salinity stenothermal) have little in common. The seven low survival populations (Bitter #3 - saline, eurythermal; Falcon - moderate salinity, eurythermal; El Tigre - saline, eurythermal; Fairy - low salinity, stenothermal; Middle Creek - low salinity, stenothermal; Uncertain - very low salinity, eurythermal) are equally variable. None of these environmental factors are associated with predation rates.

Predation by female G. geiseri on G. affinis young also varies substantially. All but one of the populations had 7 to 28% survival (Table 9). The exceptional population, East Sandia Spring, has a survival rate twice that of the next highest (Toe Nail). It also has the highest survival when males are used as predators. East Sandia Spring has the smallest water volume and presumably the lowest population numbers. On my visit, G. geiseri was relatively rare. Conspecific predation by G. geiseri also varies widely but has little correlation with predation on G. affinis. Variation in predation by G. geiseri on G. affinis young is substantially less than predation by G. affinis when more than 100 young are used: 7% to 28% versus 0% to 69% survival. Presumably the consistency of predation by G. geiseri reflects the recent transfer of the fish from San Marcos. Predation rates of females from East Sandia Spring may reflect evolutionary changes in the 50 years since they were released.

Predation by female G. nobilis on G. affinis young again varies significantly. The five samples with sample sizes more than 100 vary between 0 and 14% survival. The small sample sizes have survival rates between those for large samples. Although the range of survival of G. affinis young preyed on by G. nobilis is similar to that of young preyed on by G. geiser, 14% versus 21%, the relative variation is infinity versus three-fold. If few young survive, a high upper figure cannot occur.

Birth weight: Average birth weights of Gambusia affinis also vary extensively from less than 10 milligrams to more than 20 milligrams (Table 10). The two New Mexico populations (Lost River and Bitter Lakes #3) have tiny babies (6.0 and 8.0 milligrams). The other small young are Contrabando Canyon, Fairy, Clymer Meadow, Hanks Bull, Patty's Ranch, Big Brown, Hi Island, and Uncertain. Many of the small young are from east Texas, yet other east Texas fish (Village Creek and Egg Nog) have relatively large young. Only Fairy is stenothermal, but commonly G. affinis is absent at stenothermal locations (Hubbs, 1995). All sites have low salinities.

The heaviest young are from Heart of the Hills and San Marcos. Fish from Heart of the Hills (eurythermal) have larger young than those from the nearby Fessenden Spring (stenothermal). The sample location at San Marcos is under the I-35 bridge at a site that is relatively eurythermal for the San Marcos River (Hubbs and Peden, 1969). Again, the variation has little apparent association with geographic or environmental circumstances.

Average birth weights of Gambusia geiseri range between 17.1 and 23.3 milligrams. These weights are of the heavier end of those for G. affinis but clearly vary far less. No birth weight data are available for East Sandia Spring.

Average birth weights for G. nobilis vary but only three populations have data, so comparisons are not readily available.

In general, G. affinis and G. speciosa birth weights are about 16 milligrams, G. geiseri about 20 milligrams, Poecilia about 30 milligrams, G. longispinis about 25 milligrams, G. gaigei about 30 milligrams, G. heterochir about 30 milligrams, and G. nobilis about 40 milligrams.

Poecilia formosa from San Marcos has heavier young than P. latipinna from San Marcos; sailfin molly young from Comal Springs are heaviest of all. Heart of the Hills sailfin Molly young are about the same weight as Comal amazon Molly. Aransas County P. latipinna young are much lighter than those from the San Marcos River system.

Interbrood intervals: Average interbrood intervals for G. affinis females varies between 27 and 41 days (Table 11). The geographically proximate Phantom and Carpenter Hill populations have no overlap (Hubbs, 1996). Again there is no association of the results with environmental or geographic factors.

Average interbrood intervals for G. geiseri females is at the long end of the G. affinis range, but the variation is less than for G. affinis (8 days vs. 14). The population extreme in predation rate (East Sandia) has no data on interbrood interval.

The interbrood intervals for G. nobilis females are similar, but small sample sizes make any interpretations inconclusive.

In general, interbrood intervals are about 35 days for G. affinis, 40 days for G. geiseri, 44 days for G. gaigei, 47 days for G. heterochir, and 52 days for G. nobilis.

Numerous interbrood intervals were recorded for G. affinis and G. geiseri. The maximum number of broods for an isolated G. affinis female was 5 (4 interbrood intervals) and 4 (3 interbrood intervals) for isolated G. geiseri females. A larger fraction of the isolated G. affinis females attained 5 broods than isolated G. geiseri females attained 4 broods. Commonly, the last G. geiseri brood had one young while the last G. affinis brood had a normal number of young. Females of both species have been held in isolation for more than three additional months without producing more broods.

Discussion

Populations of Gambusia affinis vary widely in predation on newborn, birth weight, and interbrood intervals. Populations of G. geiseri also vary in each factor but at a reduced rate. Populations of G. nobilis vary in predation and birth weight. The degree of variation is substantially greater for G. affinis than for most introduced populations of G. geiseri. The amount of variation for G. nobilis is less certain, primarily due to relatively small sample sizes. Several locations have samples of two or three species. The variations of life history traits are not concordant, thus demonstrating that the variation is internal to the species (genetic) and not controlled by the habitat (environmental). For example, I have data on populations from Phantom Cave and Diamond-Y Refugium for all three species and all three traits: Birth weight: G. affinis heavy at Diamond-Y, G. nobilis and G. geiseri heavy at Phantom Cave. Interbrood interval: G. geiseri long at Phantom Cave, G. affinis and G. nobilis long at Diamond-Y. Cannibalism: G. affinis extensive at Phantom Cave, G. geiseri extensive at Diamond-Y, G. nobilis virtually the same.

This report includes comparisons of variation in life history traits for three factors for two species and two factors for one species. All eight vary sufficiently to provide unusual data if one population were to represent the species. For example, G. affinis female conspecific predation is about 30% survival, G. geiseri predation on G. affinis is about 20% survival, and G. nobilis predation on G. affinis has about 9% survival. In contrast, however, if G. affinis were to be represented exclusively by Falcon females, G. geiseri and G. nobilis by Balmorhea females, the survival rates would be 0% (versus 30%), 12% (versus 20%), and 14% (versus 9%), a reversal of the relative predation rates for the three species as a whole. Similarly, birth weights of newborn Poecilia can vary by species depending upon the population used.

Predation rate is associated with birth weight and interbrood interval by species. The most predaceous species, G. nobilis, has the heaviest young and takes longer between broods. Conversely, most populations of G. affinis have low predation, lighter young and shorter interbrood intervals. The interspecies correlations do not extend to population studies. Contrasts look like a shot gun blast.

Similarly, the birth weight and interbrood intervals for the species as a whole can be reversed by the use of selected populations of G. affinis and G. geiseri.

Stockwell (1995) showed that introduced populations of G. affinis in Nevada varied in minimum size of female maturity and fat content. His populations were introduced about 55 years ago from central Texas. They were initially introduced into one location and then transferred to four others. Two populations were from thermally stable environments, one from a warm spring, and one from a thermally unstable environment. The two from thermally stable environments were similar to each other and differed substantially from those from the unstable environment; the warm spring population was intermediate. He found other differences in field samples that did not recur in laboratory experiments.

These results resemble those reported by Stearns (1983) who showed that introduced populations of Gambusia affinis in Hawaii varied in fecundity and in the dry weight of females and of embryos. All of the stocks he analyzed had been introduced about 70 years earlier (ca. 150 fish from somewhere in Texas). Our dry weight data for G. affinis offspring vary to a similar degree. He used embryos and I used newborn young, which should be slightly heavier than embryos (Hubbs, 1971). The fish released into Hawaii came from "Texas," presumably either from the same location or were mixed prior to release, and his results were not caused by differences among the source populations. Stearns tested fish living in stable environments or environments with fluctuating water levels. Clearly, the influence of an environmental variable influenced the life history traits he observed. I found similar levels of variation among native populations of G. affinis that did not correlate with environmental factors.

All of us demonstrated variation in life history traits for introduced populations (G. affinis for Stearns and Stockwell and G. geiseri for me). The degree of variation of life history traits for G. geiseri was less than Stearns and Stockwell showed for G. affinis. This may have resulted from a) a shorter time since release (50 versus 55 or 70 years), b) the species used, or c) the differences among the environments (quite different for Stearns and Stockwell, virtually identical for me).

It is therefore essential that reports of life history (and perhaps all) traits include a consideration of the population used as well as the species. Furthermore, it is essential that the use of fish for practical applications consider the population as well as the species. This applies to the use of hatchery fish for recreational activities as well as for public health concerns.

It is possible, but unlikely, that the variation of G. affinis life history traits reflects species level differences, (i.e., G. affinis sensu stricto is a complex of numerous species). If that hypothesis is valid, there should be a correlation among the life history traits and with geographic location; these do not occur. Certainly that cannot apply to the East Sandia Spring population of G. geiseri. Even if G. affinis is a species complex, these results are merely raised to another evolutionary

level, and this variation is among populations of a morphologically-recognized species.

Acknowledgments

The field work to capture stocks benefitted from the assistance of numerous individuals including: Francisco Abarca; Daniel R. Brooks; Tony Castillo; Pat Connor; Laurie Dries; David Edds; Debby and Robert J. Edwards; Alice F. Echelle; James Fries; Linda Fuselier; Noeleonel Garcia; Cole and Gary Garrett; Lawrence Gilbert; Baile Griffith; Sam Hamilton; Tom Hayes; Dean, Garrett and Jacob Hendrickson; Ereckson and David Hillis; Anson and A. Ryland Howard; Catherine S. Hubbs; James E. Johnson; John Kargis; James King; Andy and Edie Marsh; William J. Matthews; Hanna Robin Morgan; James Peoples; Amy and A. Lee Pfluger; Andy Price; Hoven Riley; Michael J. Ryan; Michael J. Ryan; Glen A. Sachtleben; David Schlesser; John Tilton; Arcadio Valdes-Gonzales; David Van Meter; Robert Wienecke; and David C. Wilson. The laboratory work involved the assistance of Laurie Dries, Catherine A. Marler, Deborah A. McLennon, Molly R. Morris, Tanya Peterson, Gil Rosenthal, and Paige Warren.

I am also indebted to landowners, principally Ford and Edie Boulware and A. Lee Pfluger for permission to collect fishes from their ranches. Deborah J. Miller patiently word processed numerous versions of this report, including the tables, and Janet Young prepared the map.

Axelrod, H. R., and L. P. Schultz. 1971. Handbook of tropical aquarium fishes. Second ed. TFH Publications, Jersey City, New Jersey, 718 pp.

Axelrod, H. R., and L. P. Schultz. 1983. Handbook of tropical aquarium fishes. Third ed. TFH Publications, Jersey City, New Jersey, 718 pp.

Endler, J. A., and A. E. Houde. 1995. Geographic variation in female preferences for male traits in Poecilia reticulata. Evolution 49:456-468.

Foster, S. A. 1994. Inference of evolutionary pattern: Diversionary displays of three-spined sticklebacks. Behavioral Ecology 5:114-121.

Fuiman, L. A., and A. E. Magurran. 1994. Development of predator defences in fishes. Reviews of Fish Biology and Fisheries 4:145-183.

Harrington, R. W., Jr., and E. S. Harrington. 1982. Effects on fishes and their forage organisms of impounding a Florida salt marsh to prevent breeding by salt marsh mosquitos. Bulletin of Marine Science 32:523-531.

Haskins, C. P., E. F. Haskins, J. J. A. McLaughlin, and R. E. Hewitt. 1961. Polymorphism and population structure in Lebistes reticulatus, a population study. In: Vertebrate Speciation, W. F. Blair, Ed.: 320-395.

Houde, A. 1988. Genetic difference in female choice between two guppy populations. Animal Behavior 36:510-516.

Hubbs, C. 1971. Competition and isolation mechanisms in the Gambusia affinis x G. heterochir hybrid swarm. Bulletin Texas Memorial Museum 19:iii + 47pp.

Hubbs, C. 1990. Snails as a food source for Gambusia. Texas Journal of Science 42:245-256.

Hubbs, C. 1991. Intrageneric "cannibalism" in Gambusia. Southwestern Naturalist 36:153-157.

Hubbs, C. 1992. Geographic variation in cannibalism of congeneric young by Gambusia adults. Proceedings Desert Fishes Council 22:43-52.

Hubbs, C. 1995. Springs and spring runs as unique aquatic systems. Copeia 1995:989-991.

Hubbs, C. 1996. Geographic variation of interbrood intervals in Gambusia affinis. Special Publication Charles Univiversity Praha, Czech Republic, in press.

Hubbs, C., and A. E. Peden. 1969. Gambusia georgei sp. nov. from San Marcos, Texas. Copeia 1969: 357-364.

Hubbs, C., and V. G. Springer. 1957. A revision of the Gambusia nobilis species group, with descriptions of three new species, and notes on their variation, ecology, and evolution. Texas Journal of Science 9: 279-327.

Koster, W. J. 1957. Guide to the fishes of New Mexico. Univ. New Mexico Press, Albuquerque, 116 pp.

Krumholz, L. A. 1948. Reproduction in the western mosquitofish Gambusia affinis affinis (Baird & Girard) and its use in mosquito control. Ecological Monographs 18:1-43.

Meffe, G. K., and F. F. Snelson, Jr. 1989. Ecological overview of poeciliid fishes. Pp.13-31, In Ecology and evolution of livebearing fishes (Poeciliidae) (G. F. Meffe and F. F. Snelson, Jr., eds.). Prentice-Hall, Englewood Cliffs, New Jersey, 453 pp.

Minckley, W. L. 1973. Fishes of Arizona. Arizona Department of Fish and Game, Phoenix, 293 pp.

Myers, G. S. 1965. Gambusia, the fish destroyer. Tropical Fish Hobbyist 13:31-32 + 53-54.

Reznick, D. N., F. H. Rodd, and M. Cardenas. 1996. Life-history evolution in guppies (Poecilia reticulata: Poeciliidae). IV. Parallelism in life-history phenotypes. American Naturalist 147:319-338.

Reznick, D. N., and H. A. Bryga. 1996. Life-history evolution in guppies (Poecilia reticulata: Poeciliidae). V. Genetic basis of parallelism in life histories. American Naturalist 147:339-359.

Schoenherr, A. A. 1981. The role of competition in the displacement of native fishes by introduced species. Pp. 173-203, In Fishes of the North American deserts (R. J. Naimen and D. L. Soltz, eds.). Wiley Interscience, New York, 552 pp.

Seale, A. 1917. The mosquito fish, Gambusia affinis (Baird and Girard) in the Philippine Islands. Philippine Journal of Science, 12:177-187.

Stearns, S. C. 1983. A natural experiment in life-history evolution: Field data on the introduction of mosquitofish (Gambusia affinis) to Hawaii. Evolution 37:601-617.

Stockwell, C. A. 1995. Evolutionary trajectories in recently established populations of western mosquito fish (Gambusia affinis). Ph.D. Dissertation, University of Nevada, Reno, 94 pp.

Tinkle, D. 1969. Evolutionary implications of comparative population studies in the lizard, Uta stansburiana. Systematic Biology, National Academy of Sciences 1692:133-160.

Walters, V. L., and E. F. Legner. 1980. Impact of the desert pupfish, Cyprinodon macularius and Gambusia affinis affinis on the fauna in pond ecosystems. Hilgardia 48:1-8.