1. The cost of reproduction due to limiting of the reproductive female’s locomotion capability has been suggested many times, but has rarely been directly examined, especially in fishes. Here, the effect of pregnancy on swimming performance in the viviparous Mosquitofish, Gambusia affinis, was studied.
2. Eight females of G. affinis were isolated, each in a separate aquarium, and critical swimming speed (Ucrit), body mass (BM) and cross-section area were measured every 5 days from the beginning of the pregnancy until 2–4 days after parturition.
3. Swimming kinematics (tail beat frequency and amplitude) was measured in non-pregnant and pregnant females at different swimming speeds.
4. BM increased during pregnancy from 0·47 ± 0·13 g to 0·72 ± 0·19 g, and the cross-section area also increased during pregnancy from 0·21 ± 0·06 cm2 to 0·32 ± 0·07 cm2. Ucrit decreased from 25·0 ± 1·3 cm s−1 before pregnancy to 20·1 ± 1·5 cm s−1 just before parturition, and returned to 24·7 ± 1·4 cm s−1 2–4 days after parturition. Interindividual variation was repeatable and reflects real differences among individuals.
5. Swimming kinematics was not affected by pregnancy.
6. The results suggest that reductions in Ucrit are probably because of aerobic constraints and not necessarily due to hydrodynamic changes resulting from changing in body form or plasticity. Moreover, the reduction in Ucrit is, potentially, a ‘cost of reproduction’ owing to decrease in the ability to gain food during pregnancy in G. affinis females.
The cost of reproduction is defined as a trade-off between present and future reproduction (Magnhagen 1991). This cost may be due to changes in energy allocation and decrease in growth rate during present reproduction, which might lead to smaller clutch size in future reproduction (Berglund & Rosenqvist 1986), or due to decreased survival rate which, in turn, decreases the probability of future reproduction (Madsen 1987; Magnhagen 1991; Shine et al. 1998). During pregnancy or gravidity, changes in female body mass, shape, coloration, hormonal situation and/or physical conditions occur (James & Johnston 1998). It has been suggested that these changes increase predation risk and thus increase cost of reproduction (Magnhagen 1991). However, the effect of pregnancy or gravidity on female locomotion ability has rarely been studied. Gravid or pregnant females are expected to be less agile than other individuals, which increases predation risk and adds to the cost of reproduction (Magnhagen 1991).
James & Johnston (1998) listed a number of possible mechanisms that could be responsible for the decrease in swimming performance in gravid or pregnant fish. These include changes in the contractile properties of the muscle, decreased ratio of skeletal muscle mass to body mass and/or alterations in the patterns of drag or bending during locomotion due to increased girth, and resultant increased cross-section area of the fish, which may increase the power necessary to create thrust (Videler 1993).
In the present study, the effects of pregnancy on critical swimming speed (Ucrit) of the Mosquitofish, Gambusia affinis (Baird & Girard 1853), has been studied. Gambusia affinis, a viviparous teleost belonging to the family Poeciliidae, is native to North America, its habitats ranging from New Jersey to central Mexico (Koya, Itazu & Inoue 1998). It was introduced into many countries in the twentieth century to control anopheline mosquitoes, since the fish thrive on mosquito larvae (Kurmholtz 1948) and can tolerate a wide range of environmental conditions (Cech et al. 1985). Gambusia affinis was introduced into Israel in 1924 (Goren 1973; Golani & Darom 1997) and is presently a dominant species in most shallow freshwater habitats in lakes and ponds (Goren & Ortal 1999) and rivers (Gafny, Goren & Gasith 2000) in Israel.
Gambusia affinis pregnancy duration varies from 22 days at water temperature of 25 °C and 16 : 8 L : D photoperiod (Koya et al. 1998) up to 39 days at water temperature of 25 °C under natural photoperiod (Carlson 1969). The female can become pregnant immediately after parturition either by copulating with a male or by fertilizing new oocytes with sperm from spermatozoa previously stored internally in the ‘delle’, an intraovarian structure for sperm storage (Koya et al. 2000).
During pregnancy, the female’s body changes in mass and shape. In G. affinis, as an open-water dweller, gaining food and avoiding predators is largely dependent on the fish’s swimming capability (Swanson, Young & Cech 1998).
The objectives of the present study were to determine if, and to what extent, swimming capability is reduced during the pregnancy period, and to monitor changes in body shape and swimming kinematics that may or may not support James & Johnston’s (1998) suggestions for the reasons for decreased swimming performance capability during pregnancy. Swimming performance was measured during the pregnancy period as critical swimming speed (Ucrit, Brett 1964), which measures aerobic swimming ability (Plaut 2001). Changes in the body mass, cross-section area, tail beat frequency and amplitude were also recorded during the pregnancy period to detect changes that may affect swimming performance. The working hypothesis was that, if the swimming reduction occurs because of alterations in the patterns of drag or bending during locomotion due to the increased girth, or changes in the contractile properties of the muscle, swimming kinematics (tail beat amplitude and frequency) at a given swimming speed should differ between pregnant and non-pregnant females.
Materials and methods
For the study of the critical swimming speed and changes of body mass and cross-section area, females of G. affinis (standard length (SL) 3·1 ± 0·3 cm, range 2·8–3·5 cm) were collected from a freshwater pond in the Botanical Gardens of Oranim College, Israel, during October 2000. The fish were introduced to the ponds more than 10 years ago and since then have survived there under natural conditions. Water temperature in the pond was 22·9 °C during collection. Eight females were collected and each of them was placed in a different 10-l aquarium filled with dechlorinated water at a temperature of 24 ± 1 °C, with continuous biological filtration. Since all the collected females were at different stages of pregnancy, the aquaria were monitored daily to detect parturition. Fish were fed twice a day with commercial tropical fish flake food (produced by Sera Vipan, D52518 Heinsberg, Germany), live Daphnia spp. and live mosquito larvae.
When parturition was detected, the fries were removed and two male G. affinis were added into the aquarium. The day after parturition (day 0), the female was placed in a water tunnel and her critical swimming speed (Ucrit) was measured (described in detail below). The female was then slightly anesthetized with 0·11 mg l−1 2-phenoxyethanol (Summerfelt & Smith 1990), measured for total length (TL) and standard length (SL) with a mechanical calliper (±0·1 mm) and weighed using an electronic balance (±0·01 g) for body mass (BM). The female was then returned to its aquarium. At this stage sexual behaviour was observed in the aquaria. After 2–3 days a small pregnancy spot appeared on the females’ abdomen and the males were removed from the aquaria. From this stage on, Ucrit was measured every 5 days until parturition (total of seven measurements during 35 days for each female). In the measurements following the first one, SL, body height (BH) and body width (BW) were measured by video recording of the swimming female to prevent excessive handling. BM was measured by gently drying the fish in the experimenter’s hand and placing it into a preweighed beaker of water, placed on the electronic balance. The female was then returned to its aquarium. Parturition occurred on the 31st−33rd day of pregnancy. Two to four days after this parturition, the females’Ucrit, TL, SL and BM were measured once again.
Critical swimming speed( u crit)
Ucrit was measured in a water tunnel previously described by Plaut (2000). Working section dimensions were 5 × 5 × 70 cm3 (height, width and length, respectively). A G. affinis female was placed in the working section, which was then sealed, and water velocity was set to 4 cm s−1 for 90–120 min. This duration was found to be sufficient for the fish to recover from handling and to perform a Ucrit test with no effects of handling stress (Kolok 1991; Peake et al. 1997). Water temperature was maintained at 24 ± 0·6 °C. After the recovery period, water velocity was increased to 8 cm s−1 and after 10 min increased again to 12 cm s−1 for the next 10 min, and so on at increments of 4 cm s−1 and intervals of 10 min until the fish fatigued, could not hold position in the centre of the working section and was swept to the downstream screen. At this point the water current was stopped, and the fish was weighed and returned to its aquarium. Ucrit was then calculated by the equation (Brett 1964):
Ucrit = Ui + [Uii(Ti/Tii)],(eqn 1)
where Ui is the highest velocity maintained for the whole 10-min interval (cm s−1), Uii is the velocity increment (4 cm s−1), Ti is the time elapsed at fatigue velocity and Tii is the intervals between velocity incremental change (10 min).
During the Ucrit measurement a RedLake S-1000 high-speed digital video camera with f1·8, 12·5–75 mm zoom lens, connected to a RedLake MotionScope monitor (Redlake MASD Inc., San Diego, CA, USA) and placed 1 m above the working section, recorded the fish at low swimming speed. A grid background on the working section floor, and a mirror located at an angle of 45° beside the working section enabled analysis of the records for TL, SL, BW and BH, using Scion 4·0 image analysis software (Scion, Frederick, Maryland 21701, USA). All eight females were tested for one full pregnancy cycle.
Eight G. affinis females of similar size (mean SL = 3·4 ± 0·1 cm, range 3·2–3·5 cm) were collected and each was held in a different aquarium as described above. Each female was tested twice for swimming kinematics: 1–2 days after parturition and 2–4 days before the next parturition.
Tail beat frequency and amplitude were measured by video recording (250 frames per seconds) the swimming fish at three prescribed swimming speeds: 8, 16 and 24 cm s−1 in the water tunnel described above. A female G. affinis was placed in the water tunnel and held there at water velocity of 4 cm s−1 for 90–120 min to recover from handling. Water velocity was then increased to 8 cm s−1 and a recording of a >2 s sequence of the fish swimming steadily was obtained. After the first recording was complete, water velocity was increased to 16 and 24 cm s−1 and the swimming fish was recorded as described. Records were transferred to a PC as a video file (AVI format) and tail beat amplitudes and frequencies were analysed using RedLake Imaging Player ver. 2·14.
The data of Ucrit were analysed using an anova test with repeated measurements. When significant differences between measurements at different pregnancy stages were found (P < 0·05), pairwise t-tests were performed to locate the stage of pregnancy at which the changes occurred. In order to detect interindividual variations, the speed of each fish was standardized for SL by regression residual analysis, to eliminate size effect, ranked (8 for the fastest, 1 for the slowest) for each pregnancy stage, and ranks were analysed by Kendall’s coefficient of concordance (Sokal & Rohlf 1995).
The differences in tail beat frequency and amplitude between non-pregnant and pregnant females were analysed using an anova test.
The body mass of G. affinis females a day after parturition was 0·47 ± 0·13 g (range 0·31–0·73) and SL was 3·1 ± 0·3 cm (range 2·8–3·5). At day 30, 1–3 days before parturition, body mass was 0·72 ± 0·19 (range 0·48–1·09). During the pregnancy body mass of the fish increased continuously (Table 1). The increase in body mass was 55·3 ± 9·4% and not significantly related to fish size (SL, R2 = 0·046, F1,6 = 0·291, P = 0·609, linear regression analysis). However, absolute increase in body mass during the pregnancy was significantly related to initial body mass [added mass = 0·033(±0·04) + 0·31(±0·05)BMinitial, R2 = 0·847, F1,6 = 33·2, P = 0·001] and to SL [added mass = 0·25(±0·13) + 0·17(±0·04)SL, R2 = 0·715, F1,6 = 15·01, P = 0·008, linear regression analysis]. Two to four days after parturition BM decreased to 0·48 ± 0·13, which is similar to BM at day zero.
Table 1. Changes in body mass and cross-section area during and after pregnancy of Gambusia affinis females (mean ± SE)
Day of pregnancy
Body mass (g)
Cross-section area (cm2)
0·47 ± 0·05
0·21 ± 0·02
0·51 ± 0·05
0·21 ± 0·02
0·56 ± 0·05
0·23 ± 0·02
0·61 ± 0·06
0·24 ± 0·02
0·65 ± 0·06
0·27 ± 0·02
0·69 ± 0·07
0·30 ± 0·02
0·72 ± 0·07
0·32 ± 0·02
0·48 ± 0·05
0·21 ± 0·02
Cross-section area at the widest part along the fish body (calculated as the ellipse area where r1 is the body height and r2 is body width) also continuously increased during pregnancy (Table 1). Cross-section area was 0·21 ± 0·06 cm2 (range 0·13–0·31 cm2) on day 0 and significantly increased to 0·32 ± 0·07 cm2 on the 30th day of pregnancy (range 0·22–0·42 cm2), an increase of 52%. Two to four days after parturition the cross-section area decreased to 0·21 ± 0·07 cm2, which was similar to cross-section area on day 0.
Critical swimming speed (Ucrit) on day 0 was linearly related to fish standard length [Ucrit = 13·2(±2·2) + 3·81(±0·7)SL, R2 = 0·829, F1,6 = 29·1, P = 0·002, linear regression analysis], but the relation was less significant at the last measurement on day 30 (R2 = 0·557, F1,6 = 7·55, P = 0·03). During pregnancy Ucrit decreased as pregnancy developed (anova with repeated measurements, F7,49 = 37·7, P < 0·001, Fig. 1). Onset of the decrease, compared with day 0 (pairwise t-test, t(7) = 3·945, P = 0·006) was on day 10 and the decrease continued, especially between days 25 and 30 (pairwise t-test, t(7) = 10·71, P < 0·001). The average Ucrit at day 0 was 25·0 ± 1·3 cm s−1 and decreased to 20·1 ± 1·5 cm s−1 on day 30 (1–3 days before parturition). Thus, during pregnancy Ucrit decreased to 80·5 ± 4·6% of that on day 0. At day 35, 2–4 days after parturition, average Ucrit returned to 24·7 ± 1·4 cm s−1, which is not significantly different from Ucrit at day 0 (pairwise t-test, t(7) = 1·426, P = 0·197).
After standardizing the Ucrit data to SL, Kendall’s coefficient of concordance test revealed that there was significant repeatable variation among individual’s Ucrit ( χ(7)2 = 35·45, τ(8) = 0·633, P < 0·001).
Both tail beat frequency and tail beat amplitude (Table 2) at all tested swimming speeds were similar in non-pregnant and pregnant females. Tail beat amplitude increased from 8 cm s−1 to 16 cm s−1 and was not significantly changed between 16 cm s−1 and 24 cm s−1. Tail beat frequency increased significantly as swimming speed increased.
Table 2. Swimming kinematics (tail beat amplitude and frequency) in non-pregnant and pregnant Gambusia affinis females (mean ± SE)
Swimming speed (cm s−1)
Tail beat amplitude (cm)
Tail beat frequency (Hz)
0·62 ± 0·05
6·45 ± 0·65
0·64 ± 0·07
6·55 ± 0·66
0·92 ± 0·08
7·84 ± 0·75
0·90 ± 0·06
7·76 ± 0·61
0·98 ± 0·07
10·63 ± 0·80
1·01 ± 0·07
10·74 ± 0·71
The main findings of this study are that sustained or prolonged swimming performance (measured as Ucrit) of G. affinis females decreases during pregnancy to about 80% of that of non-pregnant female. Although many authors (for review see Magnhagen 1991) predicted reduction in the context of cost of reproduction and predation risk during reproduction duration, this is the first time that this parameter has been measured in fish as Ucrit and the second time that the effect of pregnancy on locomotion has been measured in fish (James & Johnston 1998). In addition, this study supports previous findings (Kolok 1992; Kolok & Farrell 1994; Gregory & Wood 1998; Kolok, Plaisance & Abdelghani 1998) that variation in swimming performance of a fish, a whole-animal activity, is a characteristic of an individual fish and repeatable over time. The results of the present study indicate that this interindividual variation is also repeatable during pregnancy.
Body mass of G. affinis females increased during pregnancy. The fact that the addition to body mass during pregnancy was linearly related to body size suggests (assuming that a fry body mass has a more or less constant value) that clutch size is positively related to the body size of the female (Brown-Peterson & Peterson 1990).
Maximum cross-section area of the pregnant female increased by 53% compared with that of the non-pregnant female. The effect of cross-section area on the drag created between the water and the body of the swimming fish is not yet clear (Videler 1993), but increase in cross-section is expected to cause an increase in drag, and may decrease the maximum swimming speed of the fish. However, if increased body cross-section area were the reason for decreased Ucrit, one would expect the pregnant fish to increase its tail beat frequency and/or amplitude at any given swimming speed to overcome the extra drag compared with non-pregnant fish. The results of the present study show that neither tail beat frequency nor tail beat amplitude differed between pregnant and non-pregnant fish at any given swimming speed; thus, it is suggested that increase in body cross-section area does not contribute significantly to the reduction in Ucrit.
Increasing cross-section area is not the only possible reason for the reduction of swimming speed during pregnancy. James & Johnston (1998) proposed: (a) changes in the contractile properties of the muscle; (b) decreased ratio of skeletal muscle mass to body mass; and/or (c) alterations in the patterns of bending during locomotion due to the increased girth. Changes in contractile properties of the muscles would result in changes in tail beat amplitude: if the muscle contraction becomes shorter and/or slower, tail beat frequency and/or amplitude at a given speed should be changed, respectively. Similarly, decrease in tail beat amplitude and increase in tail beat frequency are expected in pregnant compared with non-pregnant females at a given swimming speed, if pregnancy causes alteration in the pattern of bending during locomotion due to the increased girth. In this study, no significant changes were detected in either kinematics parameter. Thus, it is suggested that the reduction in Ucrit of pregnant G. affinis is not significantly affected by changes in contractile properties of the muscles and/or alterations in the patterns of bending due to increased girth. However, these factors may play a significant rule in fast-start and burst swimming speed (James & Johnston 1998).
The body mass of pregnant G. affinis females increases gradually during pregnancy to 155·3 ± 9·4% compared to non-pregnant females. As an ovoviviparous fish, this addition of developmentally very active tissue (embryos) within the female body is expected to expend a substantial portion of oxygen consumed by the pregnant female and to significantly increase its standard metabolic rate. However, the ability of fish to supply oxygen to the tissues is limited by gill surface area and vascular system capabilities (Jobling 1994). Swimming activity requires a significant addition of oxygen supply (Brett 1964; Beamish 1978), and Ucrit gives an estimation of maximum aerobic capacity (Beamish 1978). Thus, it is hypothesized that the progressive reduction in Ucrit during pregnancy is due to decrease of the amount of oxygen that can be allocated for swimming (scope of activity), and not necessarily because of hydrodynamic or biomechanical constraints imposed by changing body shape, muscle activity or plasticity.
As mentioned above, Ucrit gives an estimation of the maximum aerobic swimming capacity at which fish fatigue. Fish rarely, if ever, experience fatigue in their natural habitat (Beamish 1978; Plaut 2001). However, Plaut (2000) found a correlation between Ucrit and routine activity level, and Watkins (1996) showed that slower tadpoles were more vulnerable to predation than fast tadpoles. Thus, it is reasonable that reduction in swimming capability would result in reduction of activity rate and thus reduction in food gained and may be predator avoidance as well.
Ucrit, as expected, was positively correlated to SL in non-pregnant females, but the level of significance of this relation was decreased in pregnant females. These relations suggest that the swimming capability of larger females that have a larger clutch size is reduced more than is that of small females that have a smaller clutch size. This result may indicate that there is a positive correlation between reproduction output (number of offspring) and cost of reproduction (decreasing swimming performance; Magnhagen 1991; Veasey et al. 2001).
Bennett (1987) emphasized the importance of interindividual analysis, as a significant factor affecting fitness and survival. Since then, more data on interindividual variation have been collected, also in relation to different environmental conditions, and summarized by Spicer & Gaston (1999). In fishes, swimming capability has been shown to vary largely and repeatedly between individuals in the Largemouth Bass, Micropterus salmonides (Kolok 1992), Northern Squawfish, Ptychocheilus oregonensis (Kolok & Farrell 1994), and Rainbow Trout, Oncorhynchus mykiss (Gregory & Wood 1998). In G. affinis, Ucrit was also found to have significant repeatable variation between individuals, which has a slight tendency to decrease during pregnancy stages. This is the first time interindividual variation in relation to reproductive state in fish has been measured. The evolutionary importance of this source of variability has been previously discussed (Bennett 1987; Kolok 1991, 1999; Spicer & Gaston 1999).
In conclusion, this paper has documented that Ucrit of G. affinis decreases by about 20% during pregnancy and remains repeatable in the face of pregnancy. It is hypothesized that the reduction in Ucrit is mainly because part of the oxygen absorbed in the gills is used by the embryos in pregnant females, rather than to power swimming activity. However, this should be a subject for a further study. Such a reduction in swimming ability may decrease the ability to gain food, especially in open water, planktivorous fish. However, it is not yet clear if this reduction in swimming capability increases predation risk, since when avoiding predator attack, the fish usually use fast-start and burst, anaerobic swimming speed.
The author thanks Prof. N. Soker for supporting the establishment of the laboratory apparatus, and the University of Haifa Research Authority and the Committee for Research and Evaluation for financial support.