Testing Hypotheses Concerning The Phenotypic Plasticity Of Escape Performance In Fish Of The Family Cottidae

Ectothermic organisms in the temperate zone exhibit phenotypic plasticity in response to changes in environmental factors such as temperature, producing what has been termed a norm of reaction over an environmental gradient (Scheiner, 1993). Seasonal temperature fluctuations have long been known to produce physiological compensatory changes, known as acclimation responses, in a diversity of rate functions and in a wide variety of species (see Bullock, 1955, for an early review). Fry and Hart (1948) carried out a classic study examining the effect of thermal acclimation on fish swimming speed. Following acclimation to different temperatures for several weeks, maximum sustained cruising speed of goldfish (Carassius auratus L.) was measured at a range of temperatures. Acclimation extended the thermal range of activity and there was a shift in the optimum temperature for performance, between high- and low-temperature acclimation. Increases in speed at low temperature following cold acclimation were at the expense of swimming performance at high temperature, compared with warm-acclimated fish. Following cold acclimation, the mechanisms underlying altered swimming performance include, inter alia, hypertrophy of red muscle fibres (Johnston and Lucking, 1978; Sidell, 1980), increases in mitochondrial volume density (Johnston and Maitland, 1980; Egginton and Sidell, 1989), altered expression of myosin light chain (MLC) isoforms (Crockford and Johnston, 1990; Langfeld et al. 1991; Hirayama et al. 1997) and of myosin heavy chain (MHC) isoforms (Gerlach et al. 1990; Hwang et al. 1991; Johnson and Bennett, 1995; Imai et al. 1997) and associated increases in fast muscle myofibrillar ATPase activity (Johnston et al. 1975; Heap et al. 1986; Crockford and Johnston, 1990).

Not all temperate species of fish exhibit thermal acclimation. In fact, fish from habitats with large daily temperature fluctuations may exhibit reduced, or no, thermal acclimation. In the killifish (Fundulus heteroclitus), rapid daily fluctuations in water temperature of salt marsh habitats have been associated with a lack of acclimation in myofibrillar ATPase activity over the temperature range 5–25 °C (Sidell et al. 1983). Over a broader test range, 10–35 °C, ATPase activity showed some acclimation, although reduced in comparison with that of goldfish. This was reflected in a similarly reduced acclimation response in escape performance (Johnson and Bennett, 1995). It has been suggested that acclimation modifications are avoided if the contractile complex has a low thermal sensitivity to acute temperature fluctuations experienced by the organism (Sidell et al. 1983). Furthermore, the high degree of short-term temperature variation may mask any stable cues for acclimation.

During ontogeny, seasonal shifts in temperature and/or changes in habitat can profoundly alter the thermal optimum for physiological processes in ectotherms. For example, the thermal optimum for growth increases during ontogeny in larval plaice (Pleuronectes platessa) (Hovenkamp and Witte, 1991) and winter flounder (Pseudopleuronectes americanus) (Buckley, 1982). In the dragonfly (Libellula pulchella), maturation is associated with a more defined thermal sensitivity and an increase in optimum thoracic temperature and upper lethal temperature (Marden, 1995), paralleling the increase in thoracic temperature associated with the more active lifestyle of adult stages (Marden et al. 1996).

The evolutionary significance of acclimation has been less rigorously examined than have the mechanisms involved (Huey and Berrigan, 1996). Precht (1958) classified acclimation responses according to the difference in physiological rate between the previous and current thermal regime. If the rate was the same, then compensation was perfect; if the rate changed initially on introduction to the new temperature, but did not change thereafter, there was no compensation. Compensation could also be partial, excess or inverse. Despite a cautionary note by Fisher (1958), that ‘when acclimations occur it does not follow that these necessarily have obvious survival value’, it has often been assumed that acclimation responses are adaptive (Hazel and Prosser, 1974) and that fitness is improved following a period of acclimation.

This has become known as the ‘beneficial acclimation hypothesis’ (Leroi et al. 1994; Huey and Berrigan, 1996) and, until recently, has been proffered as the primary evolutionary explanation of acclimation. However, it has received recent criticism by Huey and Berrigan (1996) since, in reality, post-hoc adaptive stories can be created for any of Precht’s above-mentioned acclimation responses. According to Leroi et al. (1994), an acclimation response is only to be regarded as beneficial if it enhances differential reproduction, also known as Darwinian fitness. Furthermore, when specifically tested, the assumption has often been rejected (Leroi et al. 1994; Zamudio et al. 1995; Bennett and Lenski, 1997). Huey and Berrigan (1996) suggest that an examination of the evolutionary significance of patterns of acclimation should use more rigorous a priori hypotheses based on the natural history of the species. Using this rationale, we wished to test a set of hypotheses concerning thermal acclimation in two fish species of the family Cottidae.

Marine Cottidae are partial residents of the littoral zone (Gibson, 1969). Both short-horn sculpin (Myoxocephalus scorpius L.) and long-spined sea scorpion (Taurulus bubalis Euphr.) are benthic, marine teleosts of the temperate zone. In the north-east Atlantic, short-horn sculpin have a distribution between 45 and 78 °N, whilst long-spined sea scorpion occupy a somewhat lower clime, between 40 and 68 °N (Unesco, 1986) (Fig. 1). Around the British Isles, adult short-horn sculpin are caught offshore (Foster, 1969; King et al. 1983) between approximately 30 and 50 m (Unesco, 1986). In contrast, juvenile sculpin and all stages of long-spined sea scorpion are found in rock pools and in the shallow sublittoral zone (Foster, 1969; King and Fives, 1983; Unesco, 1986). Surface water temperatures in St Andrews Bay, Scotland, range from 3–5 °C in the winter to 15–18 °C in the summer (J. Murdoch, unpublished observations). Both sea scorpion and juvenile sculpin also experience acute changes in temperature, particularly in the intertidal zone. In the same bay, rock pools where these species are found have mean water temperatures in spring ranging from 5.7 to 12.5 °C over a 24 h period (G. Temple, unpublished observations). Pool temperatures have been found to be even more variable during the summer months (Morris and Taylor, 1983).

The short-horn sculpin and long-spined sea scorpion are ‘ambush predators’ and engage in very little steady swimming. Despite other forms of predator avoidance, such as camouflage and spines, both species commonly employ fast-starts for escape. Measurement of performance has been proposed as a practical method of yielding information on fitness and physiological compensation (Huey and Stevenson, 1979; Arnold, 1983). Several studies have shown or suggested a positive correlation between burst speed and survival, which can have a perceptible influence on fitness. Watkins (1996) found burst swimming speed of anuran tadpoles to be an important determinant of surviving attacks by garter snake predators. Thus, he thought it likely that selection acted directly on burst speed. Andraso (1997) suggested that the apparent equal success of an unarmoured morph of the brook stickleback (Culaea inconstans) in a polymorphic population of two morphs, armoured and unarmoured, was due to the absence of the protective pelvic girdle, therefore enabling a superior escape performance. Swain (1992) actually measured a correlation between escape performance and survival in stickleback (Gasterosteus aculeatus) larvae. Superior burst swimming performance by certain vertebral phenotypes was matched by increases in the frequencies of those phenotypes in the field.

In our study, the following predictions were made, using the escape response as a correlate of fitness: (1) in the short-horn sculpin, warm acclimation improves escape performance at high temperature, at the expense of performance at low temperature, compared with cold-acclimated fish, (2) escape performance in the intertidal long-spined sea scorpion exhibits reduced thermal sensitivity and therefore does not vary with acclimation temperature over the average range of seasonal temperatures, and (3) in short-horn sculpin, the ability to acclimate maximum speed thermally is acquired during ontogeny in parallel with the change in thermal habitat due to offshore migration of late juveniles.

All fish were collected between March 1995 and November 1996. Adult short-horn sculpin (Myoxocephalus scorpius L.) were caught using trawls and creels off the Fife coast and Isle of Cumbrae, Scotland. Long-spined sea scorpion (Taurulus bubalis Euphr.) and juvenile short-horn sculpin were collected on rocky shores around St Andrews, Scotland, using hand nets. In one series of experiments, adult fish of a discrete size class (sculpin, 157.2±2.7 mm total length, L, 59.2±3.4 g wet mass, N=38; sea scorpion, L=116.0±1.2 mm, 27.0±1.5 g wet mass, N=35: values are means ± S.E.M.) were acclimated to one of three temperatures. During the winter months (November to March), fish were caught and acclimated to 5±0.5 °C for a minimum of 6 weeks in a recirculating seawater aquarium. Fish were kept under a 10 h:14 h light:dark photoperiod regime. Similarly, during the summer months (June to September) fish were caught and acclimated to 15 or 20±0.5 °C and kept under a 13 h:11 h light:dark photoperiod regime. In both cases, fish were fed twice a week on prawns and squid.

To examine the scaling of temperature acclimation responses, fish of a range of lengths (sculpin, L=43–270 mm; sea scorpion, L=45–160 mm) were either caught in winter and acclimated to 5±0.5 °C or caught in summer and acclimated to 15±0.5 °C. Acclimation was again carried out for a minimum of 6 weeks, under the photoperiod and feeding regimes described above. This investigation included the discrete size class of fish decribed above that were acclimated to 5 or 15±0.5 °C.

Following thermal acclimation to 5, 15 or 20±0.5 °C, fish were transferred to a 2.0 m×0.6 m×0.25 m static filming arena at their acclimation temperature. Escape responses were elicited by tactile stimulation. A 4 mm diameter rod was presented randomly from either side of the caudal fin of a stationary fish. Filming was carried out using a high-speed ciné camera (NAC, Japan) operating at 500 frames s_⁻¹, via a mirror set at 45 ° above the tank. Lighting from below the tank enabled sharp silhouettes of the fish to be recorded on film. Adult short-horn sculpin (L=157.2±2.7 mm) and long-spined sea scorpion (L=116.0±1.2 mm) acclimated to 5 and 15±0.5 °C, were filmed at 0.8, 5.0, 15.0 and 20.0 °C. Fish acclimated to 20±0.5 °C were only filmed at their acclimation temperature.

The scaling of fast-start performance was assessed in fish acclimated to 5 and 15±0.5 °C, and tested acutely at 5.0 and 15.0 °C, in a reciprocal experimental design.

Prior to analysis, it was necessary to calculate the position of the mid-point (centre of mass) of the fish relative to the snout. A long pin with a weighted string attached was placed transversely through the head of a straight, frozen fish. The pin was balanced on two parallel bars. The fish, inclined at an angle of approximately 45 ° to the horizontal, was photographed. The pin was then removed and placed through the body beneath the first dorsal fin, and the fish was photographed again. On superimposing one photograph on top of the other, the point at which the strings crossed was found and used as the mid-point. This procedure was carried out on three fish of varying sizes for each species.

After processing, the high-speed films were digitized using MOVIAS-NAC Corp. software. Escape sequences from adult fish of the discrete size classes were analysed by digitizing ten points down the centre of the fish (the approximate position of the spine; J. Wakeling, unpublished results) from snout to tail tip. This was done for every fourth frame (8 ms) owing to the large number of points being digitized. Thirty frames were digitized before the start of the response and after the end of the second half-tailbeat, guaranteeing that the start and end of the response were included and avoiding the problems associated with smoothing the first few and last few frames (Harper and Blake, 1989). The end of the second half-tailbeat was taken as the maximum angular displacement of the snout to the left or right before myotomal contraction caused bending on the contralateral side of the fish at the start of the third half-tailbeat or glide (Figs 1C, 2). The x and y position data were smoothed using piecewise cubic regressions (Mathmatica, Wolfram Research Inc. USA). Length-specific maximum velocity (Û_max, s^₋₁) and length-specific maximum acceleration (Â_max, s₋₂) of the centre of mass were calculated from the first and second differentials of each cubic regression. The correct smooth width was established by plotting values of Û_max calculated using a number of different widths. As the smooth width was increased from three, values for Û_max showed an initial decrease, because of the digitizing errors being smoothed. Û_max then reached a plateau and finally decreased a second time owing to oversmoothing using smooth widths that were too wide. The correct smooth width was thus taken from the plateau region of graphs of Û_maxversus smooth width. This resulted in 13-point smoothing for short-horn sculpin and 9-point smoothing for long-spined sea scorpion. The fastest escape sequence from each fish was selected for further analysis. The cumulative turning angle (CTA, in degrees), defined as the sum of the absolute angles of turning of the snout relative to the mid-point throughout the first and second half-tailbeats (Fig. 2), and the maximum angular velocity (ω_max, rad s₋₁) were determined in a similar way to the position and velocity. The CTA was only calculated for sequences in which escape responses consisted of two complete half-tailbeats.

To examine the scaling of swimming velocity, a second method was used to calculate Û_max. This involved digitizing the mid-point of the fish frame-by-frame using an overhead projector overlay on which the distance of the midpoint from the snout had been marked. This method was used because only maximum velocities and the response durations (the time to the end of the second half-tailbeat) were calculated from these sequences; the first method used above enabled further calculations not discussed here. Once again, digitizing was started before the fish first moved and after the end of the second half-tailbeat. The x and y position data were smoothed using cubic piecewise regressions (LabView, National Instruments, USA) and Û_max was calculated. The width of each piece was set to ensure that the standard error of all points from their smoothed positions matched the standard error of digitizing. This resulted in smoothing ranging from 21 points for large fish to 15 points for small fish. The fastest escape response from each fish was again selected. Maximum absolute velocity was also derived from each response. Only sequences in which two half-tailbeats were completed were used to measure the response duration.

To check that the two methods of calculating Û_max were comparable, fast-starts from the discretely sized adult fish which had been acclimated to 5 and 15 °C and swum at 5.0 and 15.0 °C were analysed using both methods and compared using a one-way analysis of variance (ANOVA); there was found to be no significant difference between values from the two procedures (P>0.2).

A two-way general linear model (GLM) ANOVA was used to examine the effects of acclimation temperature (5 and 15 °C) and acute (test) temperature on Û_max, Â_max, ω_max and CTA. A one-way ANOVA was used to examine the effects of acclimation temperature at all four test temperatures. If fish were unable to swim (scored as zero), Mann–Whitney U-tests were used to test for differences between acclimation to 5 and 15 °C at that particular test temperature. All statistical tests were computed using Minitab software (Minitab Inc., USA).

To examine the scaling of Ûmax, Umax (absolute swimming velocity, m s₋₁) and response duration with temperature acclimation, linear regressions (y=aL_b, where L is total body length) for the following groups, 5@5.0, 5@15.0, 15@5.0 and 15@5.0 (acclimation temperature at test temperature, °C), were calculated on log-transformed data using least-squares regression analysis. The significance of all regression lines was assessed by means of the ANOVA F-statistic. To test for differences in acclimation ability between short-horn sculpin from the two different habitats, a two-way GLM analysis of covariance (ANCOVA) (Minitab Inc., USA) was used, with habitat in which the fish were caught and temperature group as between-subject factors and length as a covariate. Furthermore, ANCOVA was used to test for differences in regression lines. Slopes (exponent b) were compared using the first stage of ANCOVA. Tukey multiple-comparison tests were employed to determine between which groups any differences lay. Where no differences were found, the second stage of ANCOVA was used to test for differences in regression elevations (intercepts, proportionality coefficient a). Tukey tests were again used to determine where differences lay. These tests were carried out following Zar (1996).

Escape responses fitted the description of Weihs (1973) and were typical ‘C-starts’. Initial contraction of the trunk muscles on one side of the fish (stage 1) bent the fish into a curved shape. Following this was a propulsive stroke or strokes (stage 2), consisting of one or more tailbeats, and then a final gliding or steady swimming stage (stage 3) (Figs 1C, 2).

The results for the eight two-way GLM ANOVAs are given in Table 1. Test temperature had a significant effect on Ûmax and Â_max in both short-horn sculpin (Û_maxF_(3,59)=8.38, P<0.0005; Â_maxF_(3,59)=6.76, P=0.001) and long-spined sea scorpion (Û_maxF_(3,58)=36.05, P<0.0005; Â_maxF_(3,58)=21.07, P<0.0005) (Fig. 3). In the short-horn sculpin, the adjusted mean Û_max and Â_max for both acclimation groups increased by 54.8 % and 53.1 %, respectively, between 0.8 and 15.0 °C. In long-spined sea scorpion, these values increased by 177.8 % and 161.6 %, respectively, over the same temperature range. Test temperature had a significant effect on ωmax (F(3,58)=5.55, P=0.002) and CTA (F_(3,54)=4.75, P=0.005) in long-spined sea scorpion, but not in short-horn sculpin (P=0.441 for ω_max, P=0.393 for CTA) (Table 1; Fig. 4). In sea scorpion, the mean value of ω_max increased by 136.6 % and the mean value of CTA by 81.3 % between test temperatures of 0.8 and 5.0 °C.

Acclimation temperatures of 5 and 15 °C had a significant effect on Û_max (F_(1,59)=13.52, P=0.001) and Â_max (F_(1,59)=7.68, P=0.007) in short-horn sculpin (two-way ANOVA) (Table 1; Fig. 3A,B). However, there was only a significant interaction between test and acclimation temperature for Û_max (F_(3,59)=6.02, P=0.001) (Fig. 3A). For 5 °C-acclimated short-horn sculpin, mean Û_max was 3.7 s₋₁ at 0.8 °C and 3.8 s₋₁ at 5.0 °C, giving a Q₁₀ of 1.07 (Table 2). In contrast, the Q₁₀ for the 15 °C-acclimated group was 3.02, with mean Û_max increasing from 2.9 s₋₁ at 0.8 °C to 4.6 s₋₁ at 5.0 °C. Similar differences in Q₁₀ were found for Â_max (Table 2). However, Tukey tests revealed no significant differences in Û_max or Â_max at 0.8 or 5.0 °C between the two acclimation groups (Fig. 3A,B). In contrast, at 20.0 °C, the 20 °C- and 15 °C-acclimated fish had a significantly higher Û_max (5.2 s₋₁) than the 5 °C-acclimated fish (2.5 s₋₁) (P<0.001 for the 20 °C- and 15 °C-acclimated fish, P<0.05 for the 5 °C-acclimated fish; Tukey multiple-comparison test) (Fig. 3A). At this test temperature, only 20 % of 5 °C-acclimated fish completed two half-tailbeats compared with 90.9 % of the 20 °C-acclimated fish and 80.0 % of 15 °C-acclimated short-horn sculpin. At a test temperature of 15.0 °C, the 15 °C-acclimated sculpin had a mean Û_max of 5.8 s₋₁ compared with 4.4 s₋₁ for the 5 °C-acclimated fish (P<0.1, Tukey multiple-comparison test). In this species, CTA and ω_max were not influenced by acclimation temperature (P=0.441, P=0.097, respectively, two-way ANOVA) (Table 1; Fig. 4A,B).

In long-spined sea scorpion, two-way ANOVA revealed that acclimation temperature had no significant effect on Û_max, Â_max or CTA (P=0.751, P=0.136, P=0.068, respectively) (Table 1; Figs 3C,D, 4C), but did significantly affect ω_max (F_(1,58)=8.37, P=0.005), with higher values for 5 °C-acclimated fish (Fig. 4D). There was a significant interaction between acclimation and test temperature in Û_max (F_(3,58)=5.87, P=0.001) (Fig. 3C). Only two out of five fish acclimated to 15 °C were able to swim at 0.8 °C, although all individuals recovered fully following a rise in temperature. When zero was scored for fish that did not swim, Mann–Whitney U-tests revealed significantly higher values for all variables for 5 °C-than for 15 °C-acclimated fish at 0.8 °C (P<0.05) (Figs 3C,D, 4C,D). Using mean values for fish that swam, Û_max was 3.7 s₋₁ at 0.8 °C and 5.7 s₋₁ at 5.0 °C in 5 °C-acclimated fish (Q₁₀=2.83) (Table 2). For 15 °C-acclimated sea scorpion, Û_max was 1.7 s₋₁ at 0.8 °C and 5.1 s₋₁ at 5.0 °C (Q₁₀=12.88) (Table 2). At the other end of the temperature range, no variable showed any significant difference between acclimation temperatures (P>0.05).

In both species of fish, maximum length-specific velocity of the midpoint decreased and response duration (in ms) increased with increasing total body length at all temperature regimes (5@5.0, 5@15.0, 15@15.0, 15@5.0 °C; acclimation temperature at test temperature) (P<0.05, F-statistic) (Fig. 5; Table 3). Absolute velocity (U_max, m s₋₁) increased with increasing total fish length for all groups of short-horn sculpin (P<0.01, F-statistic) apart from those acclimated to 5 °C and swum at 15.0 °C (P=0.865, F-statistic; slope exponent b of 0.02) (Table 3). In the long-spined sea scorpion, U_max increased significantly with total fish length for all groups (P<0.05, F-statistic) except those acclimated to 5 °C and swum at 5 °C (P=0.34, F-statistic; slope exponent b=0.11) (Table 3). Using Û_max of short-horn sculpin as the response variable, the two-way GLM ANCOVA revealed that there was a significant interaction between habitat and temperature group (F_(3,104)=4.41, P=0.006), thus indicating differences in the acclimation ability of sculpin from the two localities (Table 4). Û_max scaled to a common slope exponent of −0.55 for 5 °C-acclimated fish at a test temperature of 5.0 °C and for 15 °C-acclimated fish at test temperatures of 15.0 °C and 5.0 °C. In the latter acclimation group, values of the proportionality coefficient a were significantly different, with log₁₀Û_max being 26.1 % higher in fish swimming at their acclimation temperature (test temperature) of 15.0 °C than at an acute temperature of 5.0 °C (P<0.05, Tukey multiple-comparison test) (Fig. 5C; Table 5). However, escape performance was not significantly different between acclimation groups at a test temperature of 5.0 °C (Fig. 6C; Table 6). The scaling relationship for Û_max changed to aL_−0.98, when 5 °C-acclimated short-horn sculpin were swum at an acute temperature of 15.0 °C (Figs 5A, 6A). This caused the scaling relationships (exponent b) to differ significantly (P<0.01, Tukey multiple-comparison test) between fish swum at an acute temperature of 15.0 °C and those swimming at their acclimation temperature of 15.0 °C (Fig. 6A; Table 6). The first part of this study revealed that Û_max at 15.0 °C was higher in adult fish acclimated to 15 °C than in those acclimated to 5 °C. However, the scaling of Û_max changed with temperature such that the regression lines for these two groups converged for the small size range of fish, indicating that this was not true for the smaller fish; Û_max at 15.0 °C was not higher in sculpin acclimated to 15 °C compared with those acclimated to 5 °C.

Using response duration as the response variable, the two-way GLM ANCOVA revealed there to be no significant interaction between habitat and temperature group. Thus, there was no significant change in the scaling of response duration in short-horn sculpin with acclimation temperature, the common slope exponent being 0.41. However, there was significant regression elevation (Table 5; P<0.01, Tukey multiple-comparison test), with fish swimming at 15.0 °C having an approximately 9 % shorter response duration than fish swimming at 5.0 °C irrespective of previous acclimation temperature. This was also the pattern found in long-spined sea scorpion. Response duration scaled to a common slope expononent of 0.28, with fish swimming at 15.0 °C having significantly shorter response durations (by approximately 11 %) than fish swimming at 5.0 °C, irrespective of thermal acclimation temperature (P<0.01, Tukey multiple-comparison test). Furthermore, Û_max of sea scorpion scaled to a common slope exponent of −0.76. Regardless of acclimation temperature, fish swimming at 15.0 °C had a significantly higher Û_max than fish swimming at 5.0 °C (by approximately 18 %) (Tables 5, 6; Figs 5B,D, 6B,D).

At a test temperature of 15.0 °C, short-horn sculpin achieved maximum length-specific velocities and accelerations during the escape response that were 34 % and 22 % higher, respectively, in 15 °C-acclimated than in 5 °C-acclimated fish. At a test temperature of 20.0 °C, these values increased to 110 % and 55 %, respectively. However, at cold temperatures of 5.0 and 0.8 °C, 5 °C-acclimated fish did not have significantly improved escape performance over 15 °C-acclimated fish.

Over the test range 5.0–15.0 °C, the long-spined sea scorpion only showed acclimation responses in maximum angular velocity, this being greater in 5 °C-than in 15 °C-acclimated fish. Over the extended temperature range 0.8–20.0 °C, long-spined sea scorpion exhibited acclimation responses in all variables at low temperature (0.8 °C). For example, maximum length-specific velocity and acceleration were 111 % and 120 % greater, respectively, in cold-than in warm-acclimated fish. At 20.0 °C, there was no significant difference in any variable between acclimation temperatures.

In short-horn sculpin, the scaling of maximum length-specific velocity at 15 °C was significantly different between the two acclimation groups. Juvenile short-horn sculpin found on the shore and all stages of long-spined sea scorpion showed increased velocities upon acute exposure to 15.0 °C, whereas adult short-horn sculpin showed enhanced velocity only following a period of thermal acclimation to 15 °C.

Our first hypothesis stated that acclimation in short-horn sculpin improves escape performance at high temperature, but at the expense of performance at low temperature, compared with cold-acclimated fish. Consistent with this prediction, we found that acclimation to 15 °C and 20 °C clearly improved the swimming speed and acceleration at 15.0 and 20.0 °C over those of 5 °C-acclimated fish acutely exposed to high temperature. Our results echo those of Beddow et al. (1995), who examined the effect of acclimation to a summer temperature of 15 °C on the prey capture response in the same species. In their study, maximum velocity at 15 °C was 33 % higher in summer-than in winter-acclimated sculpin. This phenomenon is reflected in the power output of fast muscle fibres. Johnson and Johnston (1991) used the work loop technique to examine the maximum power output of fast muscle fibres performing oscillatory work in summer- and winter-acclimated short-horn sculpin. At 15 °C, summer-acclimated fish produced three times the power output of winter-acclimated fish. Similarly, Johnston et al. (1995) used the work loop technique under conditions simulating the first tailbeat of a fast-start. At 15 °C, power output increased from 6.3W kg₋₁ in 5 °C-acclimated fish to 23.8 W kg₋₁ in 15 °C-acclimated fish. Examination of the power output from the force–velocity (P–V) relationship, which gives a good estimate of the maximum instantaneous power output (Josephson, 1993), revealed that power was approximately six times higher in 15 °C-than in 5 °C-acclimated short-horn sculpin at 15 °C (Beddow and Johnston, 1995).

Contrary to our hypothesis, however, we found that acclimation to high temperature did not significantly compromise escape performance at low temperature in short-horn sculpin (Fig. 3A,B). Work loop studies using imposed sinusoidal length changes about resting muscle fibre length found that maximum power output at 4 °C showed little variation between summer- and winter-caught short-horn sculpin (Johnson and Johnston, 1991). The relaxation rates of fast muscle fibres in the short-horn sculpin were also found to be independent of acclimation temperature (Beddow and Johnston, 1995). However, these authors found that the P–V relationship was less curved at 5 °C in fast muscle fibres from 5 °C-than from 15 °C-acclimated short-horn sculpin. After normalising the curves for maximum force (P₀) and velocity (V_max), this change in curvature produced a 40 % increase in relative power output in 5 °C-acclimated fish. It was concluded that the contractile properties of fibres are modified at low, as well as at high, temperatures. Q₁₀ values in the present study did indicate a less thermally sensitive performance by 5 °C-than by 15 °C-acclimated fish. Between 0.8 and 5.0 °C, Û_max and Â_max in the 5 °C-acclimated fish were virtually unchanged, whereas Q₁₀ values were 3.02 and 2.24, respectively, in the 15 °C-acclimated fish (Table 2).

On the basis of these results, we can only accept part of our first hypothesis. Seasonal changes in fast-start performance occurred predominantly at high temperature. Following warm acclimation, escape performance was improved. The emphasis of the hypothesis was on the trade-off in performance between high- and low-temperature acclimation. At low temperatures, acclimation was not detrimental but it did not improve performance and therefore there was no trade-off. Consequently, we must reject this part of the prediction. Likewise, in terms of the beneficial acclimation hypothesis, only warm acclimation was beneficial (as regards fast-start performance). Bennett and Lenski (1997) similarly found acclimation in Escherichia coli to be beneficial in some cases but not in others. In their experiments, fitness was measured directly; in ours, we can only speculate that improved fast-start performance with warm acclimation may enhance individual survival and ultimately fitness.

Our second hypothesis concerned the inability of long-spined sea scorpion to acclimate over the average range of seasonal seawater temperatures (test temperatures 5.0–15.0 °C). The rationale was that temperature sensitivity is lower in organisms subjected to a high degree of short-term temperature variation, and that this variation would not provide a stable cue for acclimation. An increase in temperature accelerates most processes (Schmidt-Nielsen, 1990) and this did occur with the escape response, but it was unaffected by previous thermal history. At test temperatures of 5.0 and 15.0 °C, the values for Û_max and Â_max were not significantly different between acclimation groups (Fig. 3C,D). Furthermore, the study on the scaling of escape performance revealed there to be no significant differences in Û_max and response duration between acclimation groups; Û_max was always lower and response duration always longer at 5.0 than at 15.0 °C (Figs 5B,D, 6B,D; Tables 5, 6). Therefore, no effect of acclimation at these two test temperatures was found in these particular kinematic variables.

Acclimation temperature did affect ω_max, which is calculated from angle and time components and therefore reflects the CTA and response duration. As response durations were not significantly different between the two main acclimation temperatures, ω_max consequently mirrors CTA (although this was not significantly affected by acclimation temperature, P=0.068), both being greater in 5 °C-than in 15 °C-acclimated sea scorpion over the test range of 5.0–15.0 °C (Fig. 4C,D). The functional significance of this is not known. However, greater body curvature during fast-starts has been found in fish from cold habitats compared with those from warm habitats (J. Wakeling, unpublished results). In the present study, the fact that one measure of fast-start performance changed with acclimation between 5.0 and 15.0 °C encourages a rejection of our second hypothesis. However, the existence of acclimation in this sole variable could still be an indication of a reduced acclimation response in long-spined sea scorpion over this temperature range.

With regards to test temperature, CTA was greater at a test temperature of 5.0 °C than at 15.0 °C in both acclimation groups (Fig. 4C). A greater degree of turning has been associated with submaximal responses (Domenici and Blake, 1991). Therefore, the high CTA may be a result of the lower values of Û_max and Â_max at test temperatures of 5.0 °C compared with 15.0 °C. In addition, long-spined sea scorpion at 5.0 °C may undertake more turning throughout the first two half-tailbeats as an avoidance mechanism, used to confuse predators, when higher values of velocity and acceleration are not attainable.

Over the entire thermal test range, acclimation responses were evident in Û_max, Â_max, CTA and ω_max of adult long-spined sea scorpion, with 5 °C-acclimated fish having significantly greater values at 0.8 °C than 15 °C-acclimated fish. In the strict sense, ‘beneficial acclimation’ has been defined as conferring an advantage at the temperature that produced the phenotypic response (Leroi et al. 1994). However, acclimation at 5 °C was clearly advantageous to the fish at 0.8 °C, because only 40 % of 15 °C-acclimated sea scorpion could swim at this temperature. We therefore believe cold acclimation to be beneficial in this species. There was little evidence for a trade-off in performance optimum temperature with thermal acclimation. This might have been manifest at higher test temperatures. However, higher temperatures would have been above those normally experienced by these fish and would therefore have been ecologically irrelevant. Therefore, in contrast to the short-horn sculpin, performance curves crossed at high temperature in the sea scorpion because seasonal changes in swimming performance were principally manifest at low temperature.

It is possible that acclimation ability is retained in this species, despite diurnal temperature fluctuations, because the seasonal differences in water temperatures are so large. The general seasonal change in temperatures may be sufficient to cue acclimation, although other factors are also known to have some control. For example, photoperiod is known to influence acclimation in some species (Kleckner and Sidell, 1985; Kolok, 1991). In the pupfish (Cyprinodon neuadensis amargosae), a eurythermal desert stream fish, acclimation has been achieved using a cycling thermal regime. The fish adapted their metabolic rate to both high and low daily temperatures, increasing their thermal tolerance by 2 °C (Feldmeth et al. 1974).

Our third hypothesis was accepted over the temperature range tested. In short-horn sculpin, the ability to acclimate maximum speed Û_max thermally was acquired during ontogeny. Sculpin exhibited differential responses in reaction to temperature change depending on acclimation temperature and stage of development. In contrast to adults found offshore, juveniles from the intertidal and shallow subtidal zones increased Û_max upon acute exposure to 15.0 °C (Fig. 5A) but did not increase Û_max any further following acclimation to 15 °C (Fig. 6A). At a test temperature of 5.0 °C, all sizes of sculpin failed to improve Û_max following acclimation to 5 °C (Fig. 6C). All stages of long-spined sea scorpion lacked compensation in Û_max following acclimation to 5 or 15 °C. We conclude that thermal sensitivity in the short-horn sculpin changes during ontogeny, such that the potential to exhibit acclimation responses to these temperatures is acquired concomitant with a migration to deeper, more thermally stable but seasonally fluctuating, water mass.

Maximum length-specific velocities decreased with increasing total fish length in both species of Cottidae, with absolute velocities increasing with length for the majority of temperature groups. Bainbridge (1958) measured swimming speed in dace (Leuciscus leuciscus), trout (Salmo irideus) and goldfish (Carassius auratus) and found that absolute speed increased with length at any particular tailbeat frequency. Size similarly affected maximum absolute speed during escape responses in rainbow trout (Salmo gairdneri) (Webb, 1976) and turbot (Scophthalmus maximus) larvae (Gibson and Johnston, 1995), but not in angelfish (Pterophyllum eimekei) (Domenici and Blake, 1993). On examination of the literature, Wardle and He (1988) found that maximum velocity scaled to an average aL_0.85. In turbot larvae, Gibson and Johnston (1995) found that the maximum length-specific velocity scaled to aL_−0.69. Maximum length-specific velocity scaled to an average aL_−0.55 in short-horn sculpin (all tests except 5@15.0 °C) and to aL_−0.76 in long-spined sea scorpion. Response durations were found to increase with size, as was also found in rainbow trout (Webb, 1976) and angelfish fast-starts (Domenici and Blake, 1993). Archer and Johnston (1989) found that length-specific tailbeat amplitude in the Antarctic fish Notothenia coriiceps was also dependent on size and decreased with increasing fish length.

Changes in swimming performance with increasing size can be related to increases in myotomal area, mass and length, and fibre length (Webb, 1978; Webb and Johnstrude, 1988; Archer et al. 1990). Isotonic contraction (Wardle, 1975) and isometric twitch contraction times of fast muscle fibres increase with size (Archer et al. 1990), in part reflecting tailbeat duration and, hence, maximum speed (Wardle, 1975). James et al. (1998) have recently examined scaling in the short-horn sculpin and found myofibrillar ATPase activity of fast muscle to decline with length, a situation found in a number of marine teleosts (Witthames and Greer-Walker, 1982). However, this was not related to changes in MHC or MLC ratios with increasing size, but accompanied changes in Troponin I isoforms (James et al. 1998).

Swimming performance also alters with size because of the change in the ratio of inertial to viscous forces (Reynold’s number, Re) during growth. However, Re is most relevant to larvae, for which viscous forces dominate swimming performance. Although the viscous hydrodynamic regime has recently been found to extend to a higher Re than previously thought (Fuiman and Batty, 1997), the fish used in our study were juveniles whose swimming performance would have been dominated by inertial forces.

Phenotypic plasticity has evolutionary significance only if it enhances fitness (Leroi et al. 1994). The most compelling studies of the beneficial acclimation hypothesis have used organisms with short generation times, such as bacteria (Leroi et al. 1994; Bennett and Lenski, 1997), for which fitness could be assessed directly. In order to study any organisms with longer generation times, correlates usually have to be used. In the present study, we did not measure fitness directly but used the escape response as a correlate. Warm acclimation was found to be beneficial and to improve fast-start performance in adult short-horn sculpin, whereas cold acclimation appeared to offer no benefit. In the long-spined sea scorpion, these results were reversed. If we had measured fitness per se and obtained similar results, we would be encouraged to ‘reject the generality of the beneficial acclimation assumption’ (Leroi et al. 1994), since acclimation did not always confer an advantage in either species.

The influence of fast-start performance on fitness, however, is not known in these species, and perhaps we should question the assumption that this performance parameter is a good indicator of fitness. The short-horn sculpin extends to a high latitude and is a cold-tolerant species which excretes a polypeptide antifreeze from the liver during the winter months (Fletcher et al. 1989). This perhaps indicates an obvious adaptation to its environment, but not one that is manifest in a significantly improved escape performance following cold acclimation. For cold temperatures to restrict locomotor capacity is not uncommon (see Bennett, 1990, for a review). A combination of lower predation pressure in the winter and a preferential investment in cold tolerance rather than fast-start capability may perhaps explain why locomotor and muscle contractile compensations (Johnson and Johnston, 1991) are lacking at low temperature in this species. Only further studies including field experiments could adequately address this question.

It should be noted that these experiments were conducted at different times of the year that corresponded to a particular acclimation group. This was to avoid disrupting the natural seasonal rhythms of the fish too greatly and thereby to enhance our ability to detect acclimation responses if present. Although gravid fish were not used, our results may have been influenced by some factor other than temperature, such as hormonal status. However, in a previous study, Beddow and Johnston (1995) found fast muscle contractile properties in short-horn sculpin to be similar between 15 °C- and 5 °C-laboratory-acclimated short-horn sculpin and summer- and winter-caught fish, respectively. We therefore believe that the thermal acclimation conditions used in the present study are relevant to the field situation.

We conclude that escape performance was significantly altered by acclimation temperature in both the short-horn sculpin and the long-spined sea scorpion. Walsh et al. (1997) examined the temperature tolerance of a number of different species and populations of freshwater sculpin in order to explain habitat distributions. Stream-dwelling species and populations were expected to be more stenothermal than those inhabiting thermally labile springs. Although they found this in a number of cases, they did not find support for the theory that temperature alone accounts for the distribution patterns observed. In the present study, we found acclimation responses in both species of marine Cottidae that differed from predicted patterns. However, there were interspecific and intraspecific differences in these patterns which we believe may reflect local niche distributions, although other factors are undoubtedly important.

This work was supported by the Natural Environment Research Council. The authors wish to thank Dr James Wakeling for the kinematic programmes and Nicholas Cole for the photographs.