A Laser-beam Detection System for Measuring Burst Swimming
Performance of Fish.

 

Abstract

-Locomotor performance of animals is of considerable interest from management, physiological, ecological and evolutionary perspectives. Yet, despite the extensive commercial exploitation of fishes, the relationships between performance capacity, natural selection,
ecology and physiology are far better known for terrestrial vertebrates than for fishes. One reason for this may be the additional technical challenges faced when measuring locomotor capacities in aquatic species. Here we report on the successful adaptation of a computer driven photocell beam timing technique, which has been used extensively to measure acceleration and sprint velocity in terrestrial animals, to the aquatic medium. By employing light-emitting laser diodes and photodarlington detectors from commercial applications, we were able to construct a computer-controlled aquatic "drag strip" that successfully tracked and recorded animal position with one-hundredth millisecond accuracy. Our chamber was designed for approximately 1Kg Atlantic cod (Gadus morhua), and cost about $5,000 in materials. Data are presented showing that, for Atlantic cod, fast starts measured with our apparatus are significantly repeatable over three months and that inter-fish variance in performance is greater than the intra-fish performance of repetitive trials. The relative advantages and disadvantages of the method are discussed with the conclusion that our method should become the method of choice for quantitative genetic and evolutionary studies of locomotor performance in aquatic animals.

Introduction

Locomotor performance of animals is of considerable interest from management, physiological, ecological and evolutionary perspectives. For many animals, success in predator-prey interactions and dominance hierarchy encounters depends upon locomotor ability (.i.Garland et al. 1990;;.i.Webb 1986;). Similarly, the first response of motile animals to environmental perturbation is usually behavioral; successful movement to more suitable environments and therefore survival may depend upon locomotor abilities (e.g. .iBreitburg 1992;) Therefore, scientists have expended considerable energy over the past twenty years trying to measure an animal's relative ability to move in several temporal contexts (see .i.Bennett and Huey 1990;;.i. Garland and Losos 1994;; .i.Garland and Carter 1994; for reviews). There is also considerable interest in determining inter and intra-specific variation in the physiological variables which may constrain the various locomotor capacities (acceleration, ability of an animal to sprint short distances, and prolonged locomotor (endurance) capacities are the most common measurements; e.g. .i.Huey andDunham 1987;; .i.Gleeson and Harrison 1988;; .i.Bennett et al. 1989;; Bennett and Huey 1990; Garland and Losos 1994). The primary impetus of this work has been the hypothesized link between performance capacity and Darwinian fitness (.i.Arnold 1983;) coupled with the desire of scientists to understand natural selection in feral populations.

While the theoretical linkage between natural selection, performance capacity and physiology has been largely investigated using terrestrial vertebrates ( reviewed by Bennett andHuey 1990; Garland and Carter 1994; Garland and Losos 1994), the extensive commercial exploitation of fishes suggests that interest in understanding the constraints and implications of locomotor abilities and the relevance of these measures to Darwinian fitness should be even greater for these animals. Since so little work has actually proceeded in this arena ( e.g.. Plaut and Gordon 1994;; .i.Kolok 1992a;; .i.Kolok and Farrell 1994;; .i.Nelson 1989;;.i. Nelson 1990;) one can infer that measurement of various locomotor capacities in fishes is not technically trivial.

The study oflocomotor capacity in fishes actually has a long and stellar history (see reviews by .i.Beamish 1978; ; .i.Randall & Brauner 1991;); most of thiswork has focused on the mechanism of propulsion by fish and the use of exercise performance as a gauge of fishhealth or stress level. Very little attention has been given to the raw material of natural selection: variation of performance among individual fish and the sources of that variation. Fast-start performance is thought to be the measure of performance with the most predictive value for predator/prey interactions (.i.Webb 1986;), yet has been studied little. The majority of fish locomotion studies have employed a graded water velocity increment test first developed by J.R. Brett (1964; critical swimming speed; Beamish 1978). The few studies on fast-start acceleration have employed hydrodynamic kinematics (.i.Gero 1952;; .i.Gray 1953), high-speed filming, or high-speed filming coupled with digital image analysis as the method (.i.Wardle 1975;; .i. Webb 1975;,1978,1983; .i.Taylor and McPhail 1985;; .i.Harper and Blake 1990;; .i.Gamperl etal. 1991;). Recent technological advances have also allowed the use of piezoelectric accelerometers (.i.Harper and Blake 1989;,1990) to accurately measure fish acceleration. Unfortunately, none of these techniques are optimal if the goal is to measure fast-start performance in a large number of animals under field-relevant conditions; prerequisite bounds for evolutionary and ecological studies. The piezoelectric techniques require extensive animal handling and, therefore, to measure large numbers of animals with proper recovery times is difficult. Likewise, although large numbers of fish can be filmed relatively quickly, analysis of films and videotapes to extract acceleration and velocity data can take an inordinate amount of time. High-speed filming also requires that the fish perform under bright lights; this condition is ecologically irrelevant for the vast majority of fishes.

.i.Huey et al. (1981); developed a computer driven photocell beam timing technique which allowed acceleration and sprint velocity to be repeatably and accurately measured in large numbers of terrestrial animals relatively quickly ( e.g. .i.Hertz et al. 1983;; .i.Huey and Dunham 1987;; .i.Bennett and Huey 1990;). The development of this computerized "drag strip" undoubtedly contributed to the flourishing of knowledge concerning various aspects of locomotor capacity and its evolutionary and ecological relevance that occurred for terrestrial vertebrates throughout the 1980's and continues today. This article describes the development of a system, similar to that described by Huey et al. (1981), but operative in an aquatic medium. Here we also show that we can use our system consisting of relatively inexpensive commercial lasers and an inexpensive detection system to measure significantly repeatable fast-start swimming speeds in individual Atlantic cod (Gadus morhua). We hope that the information contained herein will help lead to a level of understanding concerning the ecology of fish fast-starts on par with what is already known for terrestrial vertebrates.

Methods & Materials

Figure 1a- the swim chamber side view

Swim chamber designed by myself and Dale Webber (Dalhousie University) and built by us along with help from Shannon Reidy, now at University of Ottawa, to test hypotheses concerning the fast start capacity of Atlantic cod.

Figure 1b- the swim chamber top view

Chamber construction:
The fast start chamber was constructed from 1/4" and 3/8" opaque polyvinyl chloride "flat stock" (Fig.1) The dimensions of the actual raceway were (2.2m length x 0.3m width x 0.3m height) which separated an acclimating chamber and a receiving chamber each of equal dimension (1.0m l x0.3m w x 0.41m h; Fig.1). We designed our chamber for use on adult Atlantic cod; these dimensions should be scaled appropriately for fishes of different size. To allow passage of laser light, transparent windows were cut from Lexan® Plexiglas and secured to the raceway section of the chamber with aquarium-grade silicon cement (Fig.1). The gate separating the acclimating chamber from the raceway was fabricated from stainless steel hardware cloth. All of the construction materials are readily available from hardware stores and plastics wholesalers in North America. Light-emitting laser diodes ( Applied Laser Systems, 2160 NW Vine Street, Unit A, Grants Pass, OR, 97526 USA) of 3 milliwatt power output,600-720 nanometer wavelength and 3 millimeter beam width wereplaced at 0, 0.3, 0.9, 1.5, and 2.1m positions along the runway (Fig.1). These are the same lasers used in teaching and lecturing applications, therefore they can be purchased at a reasonable price. The lasers were mounted infront of the clear Plexiglass windows on one side of the raceway. A 3mm glass rod was attached to the front of the laser lens. This rod refracted the beam to project a vertical plane or "curtain of light" through the window and across the raceway. The beam width, height and intensity could be modified by changing the diameter of the glass rod and altering the distance of the laser to the plexiglass window.

A group of 6 photodarlington detectors of detection wavelength 580-720nm were obtained from a local electronics retailer and positioned vertically 2.5cm apart directly across from each plexiglass window (total of 30 detectors). This separation distance assured that a beam would be broken with the first two centimeters of a fish which crossed it (for the size and shape of fish we were using to test the chamber). Smaller fish would require a greater density of detectors. The detectors we used are also used commercially in remote-control applications and are therefore readily available and reasonably priced. To minimize the diminution of the light signal before it reached a detector, each detector was set into a hole drilled into the polyvinyl chloride on the opposite side of the raceway from the lasers and covered with a thin layer of glass.



Operational details:

A flow diagram detailing the software protocol is illustrated in Figure 3.

In summary, when activated by light, the photo darlington detector signal is amplified and triggers a 2N2222 transistor which puts out a 5 volt TTL signal to one of 8 inputs into an 8-input
NAND Gate (7430). When all 6 detectors in a bank are saturated, the NAND gate output is low (<0.3 volt). However, if one of the beams is broken, the corresponding input to the NAND gate goes low
and forces the output of the NAND gate to go high (>0.3 volts). The computer and digital timer board (MCS6522 Peripheral Interface Adapter, Interactive Microware Inc. P.O. Box 771, State College, PA, 16801 USA) continuously scans the outputs from NAND gates associated with each bank of detectors. An interrupt driven timer software routine in assembler code, operating at 1.023 MHz,
times the rising edge(2/3 V+) of the signal. The software-timing cycle was capable of distinguishing events 10-5s apart and would initiate upon breaking of the first light beam by the fish. Resistor R1 sets the zero level and resistor R2 controls the amplification or sensitivity of light detection (Fig. 2).

The system described has the capability of measuring from 8 detectors in each bank and up to 8 banks of detectors. This limitation is based on the number of digital inputs on the computer analog digital board. However, the number of inputs can easily be increased for applications which require a greater density of detectors or a longer raceway.

The photo detection response time, which was considered to be the response of the electronic circuit to light changes, was determined by instantly switching on a light source (light emitting diode) connected to the trigger channel of a dual input oscilloscope (Philips). The output of the circuit could then be timed with microsecond accuracy. The response time of the detector circuitry was determined to be 10-6s based upon a trigger of 2/3 V+ on the rising edge of the signal to the analog digital board.
Figure 3. Flow diagram of software protocol used to detect hardware laser beam breakage and timing between banks. Software was written in BASIC and Assembly Code by DMW.

Seven adult Atlantic cod were selected from our laboratory stock of Scotian Shelf cod, trawled off of Eastern Passage Nova Scotia in September of 1993, individually tagged, and held in 6 m3 insulated circular holding tanks until the fast-start trials were conducted in May/June of 1994. The holding tanks were continuously supplied with temperature regulated (5°C), air-saturated, filtered sea water. The tanks contained submersible pumps which maintained a constant water current of approximately 15 cm s-1. The fish were exposed to their natural photo period and fed a diet of chopped mackerel(Scomber scombrus L.) and chopped squid (Illex illecebrosus L.) 3 day week-1. The fish did not receive food for 4 days prior to an experimental trial and all performance tests were conducted without investigator knowledge of that particular individual's performance in any previous test.

Experimental protocol:
Twenty four hours prior to the initiation of the experiment, a fish was lightly anaesthetized with MS-222 (50 mg/L Sigma®) and placed in the acclimation chamber of the fast-start chamber (Fig 1.). The fish was kept in this area by a lowered gate which was used to separate the acclimation chamber from the raceway. Temperature controlled (5°C), filtered sea water flowed slowly, yet continuously, through the fast-start chamber in order to maintain water temperature, O2, and NH3 levels acceptable. The following morning, the gate was raised and the fish startled by grasping its caudal peduncle, thereby initiating the
characteristic "Mauthner startle response" (.i.Eaton et al. 1977;). Electrical and optical stimuli were also tried as ways to initiate a fast-start by Atlantic cod, but tactile stimulation elicited both the most intense and reproducible response. The tactile stimulation caused the fish to burst down the raceway into the receiving chamber (Fig. 1) where another gate could be closed allowing the fish to rest in this chamber before being returned to its holding tank or the acclimation chamber for a subsequent trial. A trial lasted 1-2 sec. Sprint swimming velocity and acceleration profiles were calculated by the computer software from the time elapsed after the first laser beam was broken and the distance between the laser banks and subsequent breakage of laser beams. If additional trials were desired, the fish was returned to the start chamber and allowed to rest for an additional 3 h. The procedure could then be repeated, and 3 h later repeated again for a total maximum of 3 different trials on the same fish in a period of 1 day. Three hours was considered a minimum time to recover from a fast-start trial and subsequent handling (.i.Reidy et al. 1995);. We conducted our trials in low ambient red light so that the fish were in complete darkness. In a separate experiment, this procedure was repeated three months apart with a different group of 17 Atlantic cod (Reidy and Nelson unpublished). Some data from this latter group are also presented here to support our contention of methodological reproducibility over time.

Results

The results from our trials with Atlantic cod show that the chamber described here can effectively measure fast-start performance in an aquatic medium and should be of use to investigators interested in the ecological and evolutionary ramifications of fast-start performance of fishes. We support this contention with evidence that the method is repeatable over a period of several months and a finding that the intra-individual variances of both acceleration and top swimming speed in repetitive trials are smaller than the inter-individual variances in these measures (see below).

Methods Critique:

Method Advantages:
The major advantage of this technique is that it allows the investigator to obtain acceleration and swimming speed data on a large number of fish under ecologically relevant light levels very quickly. The rate at which animals can be processed can be increased from that described above (three per day) by reducing acclimation time through techniques we have described earlier (e.g. .i.Nelson et al. 1996;). In brief, animals can be held in large diameter polyvinyl chloride tubes under the experimental conditions of interest and then, at the appropriate time, gently slid into the acclimation chamber without ever seeing or touching a human. Based upon heart rate, ventilation, and blood pressure measurements, the stress incurred through such a transfer between tanks is minimal for Atlantic cod (J.A. Nelson, Y. Tang and R.G. Boutilier unpublished observations). By employing this type of methodology, it should be possible for a well staffed laboratory to obtain fast-start data from as many as ten animals per day.
In contrast, while filming the fast-starts of fish is no more time-intensive than our method, high-speed cinematography must occur at light levels which are appropriate only for neustonic fishes, and may produce unnatural responses in fish not accustomed to such bright light. In addition, extracting acceleration and swimming speed data from films can take hours per fish; our method produces swimming speed and acceleration data which can be stored in a computer file, saved to a spreadsheet, or printed instantly. Filming also has a number of technical difficulties resulting from image distortion, parallax and measurement error associated with mechanical and electrical imperfections in cameras and image digitizing systems as well as inadequate film speeds. These limitations have been described thoroughly by .i.Harper and Blake (1989) ;and will not be reiterated here.

Accelerometers, when properly deployed, are the optimum way to obtain an accurate measurements of a fish's ability to fast-start .i.(Harper and Blake 1990) ;. However, their use is precluded for small fishes and the extensive animal handling and surgery required for accelerometer implantation render their use impractical for evolutionary or ecological studies requiring large sample sizes. This method also requires time and labor-intensive calibrations. Furthermore, to obtain the ultimate degree of accuracy these instruments are capable of, one must also film the fish and analyze the films to correct for tangential accelerations (Harper and Blake 1990). We believe that for studies requiring only relative measures of acceleration and fast-start velocity, the method we describe here is optimal.

Method disadvantages:
The major disadvantage of the fast-start chamber we describe here is that it can be used only for relative performances. Errors induced by wall effects .i.(Webb 1993); and non-linear swimming paths by the fish, as well as physical and monetary factors which limit how densely the lasers and photo detectors can be positioned, compromise the ability of this technique to measure absolute values of acceleration or fast-start velocity. Furthermore, fish with more pointed snouts will break the
beam with greater variance than those with blunter snouts. Thus, deviations of measured speeds and accelerations from reality will be specific to each species and size class of animal measured. This source of error can be limited by reducing the vertical and horizontal distance between the photo darlington detectors (Fig. 1), but each increment of error minimization incurs further costs. For example, the 0.3m horizontal distance between the first two detector banks in our prototype was insufficient to resolve the maximum acceleration capability of Atlantic cod with confidence (see below). It is also
possible with more complex circuitry, to monitor each photo transistor in a bank and
thereby quantify and correct for any error due to a vertical swimming component, if present.

A second disadvantage of this technique is that no information about the type of fast start is obtained. Fish execute fast starts with a variety of biomechanical strategies (.e.g. i.Harper and Blake 1990);; knowledge of which fast-start type was used by a fish allows the investigators to group performances and thereby reduce the overall variance in their data.

Figure 4a.

Figure 4. Swimming speed of Atlantic cod as they burst through the 2.2 meter runway after tactile stimulation of the animal in the holding chamber. The equation and correlation coefficient of the least squares linear regression describing each line are included a) three consecutive trials of the same animal, all performed within a 12 hour period b) the fastest of three trials, all performed in a single day, for each of six additional cod.

Figure 4.b

General Results:

Fast-start velocity:
Swimming speed generally increased in a linear fashion, unlike the situation for rainbow trout (Salmo gairdneri) and northern pike (Esox lucius; .i.Harper and Blake 1990;). Therefore, the relationship of time versus swimming speed was fit with least-squares linear regressions, the lines and equations of which are presented on Figure 4. Figure 4a shows a representative result when an animal is swum repeatedly, three times in the same day. This graph illustrates that much of the variability in repetitive runs occurs with the initiation of the fast start; the three runs were virtually indistinguishable after the fish passed the second detector array (first data point). This latter result can also be seen numerically by comparing the slopes (acceleration, also plotted on Figure 5a) from the least squares linear regression lines (Fig. 4a). The heterogeneity of each of the three starts is aptly demonstrated by the 20% difference between the slowest and fastest swimming speed recorded at the second detector array (first data point; Fig. 4a).
Comparison of the velocity profiles for the other six fish (Figure 4b) demonstrates that most of the fish accelerated linearly through the last three detector arrays and that much of the inter-individual variance in fast starts measured with our apparatus also occurs in the initial 0.3 meters of the chamber. This graph (Figure 4b) presents the fastest of three performance trials, run in a single day for each fish. Four of the six fish had remarkably similar "best swims" after the first detector array. In contrast, Fish #2 accelerated better than any other fish through the first two detector arrays but then basically decelerated through the remainder of the chamber while Fish #6 had the slowest start of any fish, but had the greatest rate of acceleration (about 2 m·s-2) throughout the remainder of the chamber (Fig. 4b). These results can also be seen numerically by examining the equations of the "best fit" linear regression; the slope of the line is the acceleration through the last three detector arrays and the y-intercept can be considered a rough measure of the animal's starting ability (velocity at 0 time). Notice that the heterogeneity in these measurements between individuals exceeds that of multiple trials of the same individual (Fig. 4a).

Acceleration:
Figure 5a plots the acceleration data corresponding to the three trials depicted in figure 4a. Again, this was the same animal swum three times in a single day. These data show that the maximal rate of acceleration for this cod undoubtedly occurred before our first detector array (0.3m) and, that although the fish continued to accelerate throughout the swim chamber, the magnitude declined to a steady level after 0.3m (Fig. 5a). A graph of elapsed time versus acceleration was best fit with a declining power function; the lines and equations of the "best fit" power functions are included on Figure 5. These data are in accord with accelerometer data collected on rainbow trout (Salmo gairdneri) and northern pike (Esox lucius) by .i.Harper and Blake (1990;). These authors found maximum acceleration for all types of fast-starts to occur within the first 0.15s of swim initiation. To get a realistic number for maximal acceleration by Atlantic cod of this size, one would need to have the second photo detector array positioned much closer than 0.3m from the starting line.

Figure 5a

Figure 5. Acceleration of Atlantic cod as they burst through the 2.2 meter runway after tactile stimulation of the animal in the holding chamber. The equation and correlation coefficient of the "best fit" power function for each curve are included a) acceleration curves for the same three consecutive trials depicted in Figure 4a. b) acceleration curves for the six trials depicted in Figure 4b.



Figure 5b presents the acceleration data corresponding to the swimming speeds presented in Figure 4b. Again, these were the fastest of three trials run in a single day for each of six fish. Although the inter-individual heterogeneity in acceleration is not as visually evident as it is for swimming speed (Fig. 4b), the five-fold variance in power function exponents (Fig. 5b) for only six fish shows that each individual accelerated in a unique manner. In contrast, the power function exponents from repetitive swims of the same fish (Fig. 5a) are very similar. Even if one chose only to work with maximal acceleration values, there is a greater than five-fold difference between the top acceleration reached by fish # 6 and that reached by fish # 2. The corresponding differences for the fastest and slowest trial of any individual fish (Fig. 5a) were generally less than two fold.


Long-term repeatability:
We were confident enough in the technique described here that we included it in an ongoing study on the locomotor performance of Atlantic cod (Reidy and Nelson in preparation). As one of many measurements in this study, 17 Atlantic cod had their fast-start performance tested twice, with the trials falling approximately 3 months apart. Here we present only the maximal swimming velocity reached by the 17 cod for each of the two trials because it illustrates an important point (Fig. 6). Figure 6 shows that our method is significantly repeatable over a period of three months in a population of wild fish held in the laboratory. By correcting for differences in body size, this relationship becomes even more robust (data not shown). Notice that since nine of the fish had a faster second trial, seven fish had a faster first trial, and one fish had identical swimming speeds between trials, we can safely conclude that there was no learning effect to account for nor did the fish's health deteriorate over the three month period.

Figure 6

Figure 6. Maximum swimming speed recorded in each of two separate fast-start trials, run approximately three months apart, for each of 17 Atlantic cod. The equation and correlation coefficient of the least squares linear regression (solid line) and a line of perfect identity (dashed line) are included. Points above the dashed line are fish which swam faster in the second trial while points below the line are those fish which swam faster in the first trial.

Discussion

This work opens a door for investigators interested in the ecological and evolutionary implications of variance in fish locomotor performance as well as its mechanistic basis. Most of the work in this arena has employed reptilian models (see .i.Bennett and Huey 1990;;.i. Garland and Losos 1994;; .i.Garland and Carter 1994 for reviews) . The few fish studies which have expressed interest in variance of performance have used critical swimming speed as the measure of locomotor capacity (e.g. .i.Kolok 1992a;1992b;; .i.Nelson et al. 1996;;.i. Plaut and Gordon 1994;) which is a test of limited applicability to any locomotor pattern used by fish in nature. Furthermore, the physiological support of critical swims can vary significantly among conspecifics and with environmental history (e.g. Nelson et al. 1996), complicating studies where variance in performance capacity is the primary concern.

In contrast, fast start performance of fish is biologically important and
worthy of investigation in order to understand relative predation
efficiency, predator avoidance abilities etc. (.i.Webb 1986;). Presumably, the dearth of studies on factors contributing to variance in performance and its ecological/evolutionary relevance in fishes comes, in part, from the lack of a convenient method for studying fast-starts in large numbers of fish. Here we present a modestly-priced, minimal effort method for measuring relative fast-start performance in mobile aquatic organisms. The method described, allows investigation of fast-start swimming performance under realistic light levels, without a huge
investment of investigator time. We used our prototype fast-start chamber to measure swimming velocities and accelerations of seven Atlantic cod. The inter-individual variance in performance was greater than intra-individual variance of repetitive trials, suggesting that this chamber can be used to explore mechanistic differences in fast-start performance. Furthermore, the maximum fast-start velocity of 17 Atlantic cod measured by our method was significantly repeatable over a period of three months.
With this degree of success with a prototype and a species not generally known for its swimming prowess, we are confident that our method will become the method of choice for quantitative genetic and evolutionary studies of aquatic animals. Our chamber should be capable of measuring fast-start performance in any macroscopic aquatic organism that employs an escape response. Thus investigators can now begin unraveling the complex relationship between organismal performance, physiology and ecological determinants of success like predation, competition, and social dominance in aquatic organisms.

Acknowledgments

We thank Dr. N. Balch, the Dalhousie Aquatron staff, Todd Bishop for his excellent husbandry and Dr. B. Paton for his laser suggestions. This study was supported by a Department of Fisheries and Oceans/NSERC subvention grant to Dr. J. A. Nelson and Dr. S.R. Kerr and by Ocean Production Enhancement Network (OPEN) funding to Dr. R.G. Boutilier and Dr. S.R. Kerr.

References

Arnold, S.J. 1983. Morphology, performance and fitness. American Zoologist. 23: 347-361.


Beamish, F.W.H. 1978. Swimming capacity. Pages 101-187 in W.S. Hoar and D.J. Randall eds. Fish physiology. Vol. 7. Academic Press, New York.

Bennett, A.F. and R.B. Huey. 1990. Studying the evolution of
physiological performance. Pages 251-283 in D.J. Futyma and J.
Antonovics, Editors. Oxford surveys of evolutionary biology. Vol. 7. Oxford Univ. Press, Oxford.

Bennett, A.F., T. Garland Jr., and P.L.Else. 1989. Individual
correlation of morphology, muscle mechanics, and locomotion in a
salamander. American Journal of Physiology 256 (Regulatory, Integrative and Comparative
Physiology) 25:R1200-R1208.

Breitburg, D.L. 1992. Episodic hypoxia in Chesapeake Bay:
Interacting effects of recruitment, behavior, and physical disturbance. Ecological Monographs 62: 525-546.

Brett, J.R. 1964. The respiratory metabolism and swimming performance of young sockeye salmon. Journal of the Fisheries Research Board of Canada 21: 1183-1226.

Eaton, R.C., R.A. Bombardiere and D.L. Meyer. 1977. The Mauthner-
initiated startle response in teleost fish. Journal of Experimental Biology 66: 65-81.

Gamperl, A.K., D.L. Schnurr, and E.D.Stevens.1991. Effect of a
sprint-training protocol on acceleration performance in rainbow
trout (Salmo gairdneri). Canadian Journal of Zoology 69: 578-582.

Garland, T.J. and P.A. Carter. 1994. Evolutionary physiology. Annual Review of Physiology 56: 579-621.

Garland, T.J. and J. B. Losos. 1994. Ecological morphology of locomotor performance in squamate reptiles. Pages 240-302 in P.C. Wainwright and S.M. Reilly editors, Ecological morphology: integrative organismal biology. University of Chicago Press, Chicago.

Garland, T. JR., E. Hankins, and R.B.Huey. 1990. Locomotor
capacity and social dominance in male lizards. Functional Ecology
4:243-250.

Gero, D.R. 1952. The hydrodynamic aspects of fish propulsion. American Museum Novit. 1601:1-32.

Gleeson, T.T. and J.M. Harrison 1988. Muscle composition and its
relation to sprint running in the lizard Dipsosaurus dorsalis.
American Journal of Physiology 255 (Regulatory, Integrative and Comparative
Physiology
24): R470-R477.

Gray, J. 1953. The locomotion of fishes. Pages 1-16 in S.M.Marshall and
P.Orr editors, Essays in Marine
Biology. Elmhirst Memorial Lectures. Oliver and Boyd. Edinburgh.

Harper, D.G. and R.W. Blake.1989. On the error involved in high-speed film when used to evaluate maximum accelerations of fish. Canadian Journal of Zoology 67:1929-1936.

Harper, D.G. and R.W. Blake. 1990. Fast-start performance of rainbow trout Salmo gairdneri and northern pike Esox lucius. Journal of Experimental Biology 150: 321-342.

Hertz, P.E., R.B. Huey and E. Nevo. 1983. Homage to Santa Anita: thermal sensitivity of sprint speed in agamid lizards. Evolution 37: 1075-1084.

Huey, R.B. and A.E. Dunham 1987. Repeatability of locomotor performance in natural populations of the lizard Sceloporus merriami. Evolution. 41:1116-1120.

Huey, R.B., W. Schneider, G.L. Erie, and R.D. Stevenson. 1981. A
field-portable racetrack for measuring acceleration and velocity of small cursorial animals. Experientia 37: 1356-1357.

Kolok, A.S. 1992a. Morphological and physiological correlates with swimming performance in juvenile largemouth bass. American Journal of Physiology 263 (Regulatory, Integrative and Comparative
Physiology
32): R1042-R1048.

Kolok, A.S. 1992b. The swimming performances of individual largemouth bass (Micropterus salmoides) are repeatable. Journal of Experimental Biology 170: 265-270.


Kolok, A.S. and A.P. Farrell. 1994