A Rotating Respirometer to Monitor Voluntary Activity and Associated Exchange of Respiratory Gases in the Land Hermit Crab (Coenobita Compressus)

In recent years much interest has focused on the physiological changes which accompany exercise in decapod crustaceans (see review articles by Herreid, 1981; McMahon, 1981; McMahon & Wilkens, 1983). In many of the original studies (e.g. McDonald, McMahon & Wood, 1979; McMahon, McDonald & Wood, 1979; Rutledge, 1980) distance, speed and force of locomotion were not quantified; animals were simply exercised to exhaustion which generally occurs within 10 min. This has made a comparison with the literature on ectothermic vertebrates difficult (reviewed by Brett, 1972; Bennett, 1978). Some of the more recent investigations on the land crabs Cardisoma (Wood & Randall, 1981a,b; Herreid, O’Mahoney & Full, 1983), Birgus (Smatresk & Cameron, 1981), Ocypode (Full & Herreid, 1983) and Uca (Full & Herreid, 1984) have quantified speed of enforced locomotion using classical treadmill techniques. The focus however has remained on 10- to 20-min bursts of exhausting activity (speeds of around 10 cm s⁻¹). While bursts of sprinting activity are important in escaping predators, submaximal but sustained locomotion may be far more typical of the natural behavioural repertoire. The only existing study on maintained activity is on the blue crab, Callinectes sapidus, (Booth, McMahon & Pinder, 1982) which can be forced to swim for up to 1 h.

Treadmill studies in all animals are somewhat questionable since speed is generally preset by the investigator, frequently unsupported by measurements of field activity. Furthermore, recent studies in humans have demonstrated that even trained treadmill runners experience balance problems which are not associated with ground running (Woolley & Winter, 1980). Thus treadmill exercise is undoubtedly more stressful than voluntary activity.

At night, on beaches in Panama, we were impressed by the sustained pedestrian locomotion of a local terrestrial hermit crab (Coenobita compressus H. Milne Edwards). Previous studies of short bursts of forced activity of this species have been made using a treadmill (C. F. Herreid & R. J. Full, unpublished). Hermit crabs are interesting from an energetic viewpoint due to the additive locomotor costs associated with carriage of the adopted molluscan shell which is frequently equivalent to body mass (Wheatly, 1984). As such, they provide a unique opportunity to assess the natural effects of loading in an invertebrate species. Furthermore, hermit crabs provide an interesting comparison with other decapods since they do not use sideways octapedal locomotion but move in a forwards direction using the chelipeds, second and third pairs of legs for walking. The fourth and fifth pairs of legs are reduced and modified to assist in shell retention.

The present paper describes observations we made in the field on natural locomotory patterns in Coenobita. We subsequently developed a rotating respirometer to study the energetics of voluntary activity in this species. This apparatus enabled us to correlate exercise performance with exchange of respiratory gases.

The study was conducted on land hermit crabs Coenobita compressus (H. Milne Edwards-identified by reference to Ball & Haig, 1973) which were collected at Boy Scouts’ Beach on Naos Island, off the Pacific coast of Panama in April and May of 1983. In this population, crabs had a mean body mass of 10·5 ± 1·1 g (N = 29) and most frequently inhabited Nerita scabricosta shells (identified from Keen, 1971) with a mean weight association index (WI = mollusc shell mass/crab body mass; Wheatly, 1984) of 0·96 ± 0.06 (N = 22). Less frequently Coenobita were found inside shells of Muricanthus, Polinices, Thais, Nattica, Northia and Opeatostoma providing WI ratios ranging up to 1·45. Activity patterns were observed in the natural environment. For physiological investigation, animals were transported to the nearby Naos Laboratory of the Smithsonian Tropical Research Institute where they were maintained under ambient conditions (i.e. air temperature ≃ 30 ± 4°C; humidity ≃ 70 ± 10%) in glass terraria (60 × 30 ×45 cm) for 1 week prior to experimentation. During this period they had free access to 10% sea water and fruit. A scaffolding of ceramic tiles was erected inside the terrarium to provide shelter and internal geography.

On the evening of 27th April 1983, a day which had experienced tropical weather typical of the season (i.e. brief but heavy rainfall around 10.00 h), a transect survey was run on the central portion of Boy Scouts’ Beach (8° 54·8’ N –79°31·9’ W). We had previously observed that crabs spent daylight hours sheltering in a patch of rocks high up on the beach. A 6-m wide transect was marked off in 2-m grids from these rocks down to the water line - a total distance of 30 m. Numbers of crabs in each grid were recorded at hourly intervals from 19.00 h until midnight.

Locomotory activity and associated changes in gas exchange were studied in a closed respirometer (Fig. 1) which the crab could rotate at will by walking along the inside inclined surface. The apparatus was fashioned after a system used by Jensen & Holm-Jensen (1980). The respirometer was a 316 ml glass storage jar (i.d. 85 mm), with a counterbalanced lid, which would float sideways and approximately 50% submerged in a surrounding water jacket. The dimensions of the outer container only marginally exceeded the depth and external diameter of the respirometer so that movements other than rotational were eliminated. A plastic meshwork was glued onto the inside wall of the jar to provide traction for the crab’s walking movements. The water jacket was heated with a small aquarium filament heater to maintain air temperature inside the respirometer at the experimental temperature which was 30 ± 2°C. A hole drilled through the lid of the jar was sealed with dental dam serving as a sampling port for respirometer gas.

A 3-cm strip of aluminium foil was glued along the outside of the jar in parallel with the axis of rotation so that it was immersed in the surrounding water during half of each complete revolution. Two large flattened copper electrodes were positioned at right angles to this strip and led via an impedance convertor (UFI Model 2991, Morrobay, CA) to a Gilson biscriptual polygraph (ICT-SH). Complete revolutions were detected as changes in the impedance of the water each time the foil strip was immersed and were recorded as positive square-based deflections in the polygraph trace. Distance covered, duration and velocity of voluntary locomotion could, in this way, be quantified.

Based on preliminary field observations (see Results), animals were placed in the respirometer in the early morning and allowed to acclimate to the experimental chamber during a period of the day when they were normally inactive. Pre-exercised (control) rates of O₂ consumption and CO₂ production were estimated 3 h later by analysing the gas content of samples sequentially removed from the inside of the respirometer. ‘Steady-state’ active or recovered levels of and (as defined by Stainsby & Barclay, 1970) were measured over the two time periods 0·5–1·0 and 1·0–1·5 h from the start or end of spontaneous activity. Trials were performed on loaded and unloaded crabs. In the case of loaded crabs, animals were removed from their shells for determination of body mass only upon completion of measurements.

Gas samples (50–100 μl) were removed from the respirometer using gas-tight Hamilton syringes. Contents of O₂ and CO₂ were measured using Applied Electrochemistry analysers (S-3A O₂ sensor N-37M and CD 3A CO₂ sensor P61B, Sunnyvale, CA) electrically interfaced with a two-channel Linear Instruments Model 595 flat bed chart recorder. Flow of gas through the analysers was generated by two suction pump/flow meter systems (R/2a Applied Electrochemistry) which were set to have identical flows of 12·5 ml min⁻¹. These two systems effectively operated in parallel. One drew dehumidified room air first through a CO₂ analyser and then through channel 2 of a differential O₂ analyser. The second meanwhile drew dehumidified room air directly through the reference channel (1) of the O₂ analyser. The O₂ analyser was set to display the differential between the two channels. A diagram of this arrangement is given by McMahon & Burggren (1985).

The two-analyser system was first calibrated using a standard gas which was 3% CO₂ in N₂. Different volumes (10, 20, 30, 40, 50 μl) were injected as a discrete bolus into the gas inflow at an injection port. Since all tubing was small bore, the gas passed through the analysers as a bolus and appeared on the chart as peaks of higher CO₂ and lower O₂ concentration (as compared with room air). The CO₂ analyser system could be calibrated relatively simply since the area beneath CO₂ peak was directly proportional to the CO₂ content of the injected gas.

The O₂ analyser was also calibrated using unitary injections of 3% CO₂ in N₂ since this was the only certified gas mixture available to us in the field. The negative peak on the O₂ differential trace effectively represented displacement of O₂ from room air (termed O₂ ‘deficit’). This was again calibrated by integrating the area beneath the curve. In calculating O₂ consumption inside the respirometer it was important to remember that O₂ ‘deficit’ was attributable to CO₂ addition as well as O₂ removal by the crab. In calculating that part of the deficit which was due to O₂ removal it was necessary to know proportional changes in CO₂ content of respirometer gas and the total amount of CO₂ which had been evolved.

Contents of O₂ and CO₂ in gas samples removed from the respirometer were thus measured in μmol and were subsequently used to calculate and knowing the volume of the respirometer, elapsed time and animal mass.

All data are expressed as mean ± 1 S.E. (number of observations). Samples were tested for homogeneity of variance (F test) and means compared by Student’s twotailed t-test (paired or unpaired design) using P = 0·05 as the fiducial limit. Linear regression was performed by the least squares method. Slopes from different treatments were compared using an analysis of covariance (Zar, 1974). In cases where regression lines were parallel, coincidence was tested by comparing elevations using the Newman-Keuls multiple comparison procedure (Newman, 1939; Keuls, 1952).

Fig. 2 illustrates the distribution of hermit crabs on the experimental grid from 19.00 h to midnight on a typical evening in the tropics. At 19.00 h no crabs were present. Over the next hour 35 animals emerged and occupied transects anywhere up to 21 m from the rocks. To establish this position it was necessary for them to move at a speed of 0·58 cm s⁻¹. The mean weighted distance travelled however was 3·86 m which equates to an average velocity of 0·11 cm s⁻¹. Over the next 2h the weighted mean distance increased to 6·20 m and then decreased to 3·30 m by midnight as the nightly excursion came to a close. From the difference in weighted mean distances covered per unit time, minimum running speeds of 0·09, 0·05 and 0·04cm s⁻¹ were calculated for successive hourly intervals between 21.00 h and midnight. These are probably conservative estimates of running speed since the movement of animals during the first hour of activity clearly indicates that individuals were capable of greater velocities. An in-depth survey would demand that individual crabs be marked and followed. Visual inspection did reveal however that the population as a whole remained active for a number of hours and so sustained submaximal exercise is a prominent feature of the normal activity pattern.

In the laboratory hermit crabs exhibited forward pedestrian locomotion using the chelae as levers for propulsion in addition to the second and third pairs of legs. Animals were virtually 100% successful in rotating the respirometer and entire actograph traces were obtained from 32 individual trials. Activity spontaneously commenced in the late afternoon and continued uninterrupted for an average of 148 ± 13 min. During this time crabs covered 52·9 ± 7·3 m at an average velocity of 0·6 cm or 0·2 BL (body lengths) s⁻¹ where BL is taken as the tip of the chelae to the projection of the shell apex.

As in the field, individual crabs were variable in their exercise performance. The maximum values observed for speed, distance covered and duration of spontaneous activity were respectively 2·5 cm s⁻¹ (0·7 BL s⁻¹), 150 m and 300min. Crabs were divided arbitrarily into high and low performers based on their ability to cover distances 2 standard errors in excess of the mean value. An analysis of velocity with time during the exercise bout (Fig. 3) revealed that high performers maintained a velocity of 0·81 cm s⁻¹ throughout the entire 221 min of activity whereas low performers exhibited a progressive decrease in running speed which became significantly reduced prior to cessation of activity at 119 min.

Standard and in inactive C. compressus occupying Nerita shells at 30°C were respectively 91·4 ±8·9 (23) and 41·6 ±4·4 (22) μ mol kg⁻¹ min⁻¹ producing an r value of 0·5. When removed from their shells, these same crabs maintained at 92·8 ± 9·7 μ molkg⁻¹ min⁻¹ but declined to 27·2 ±2·9 μ mol kg⁻¹ min⁻¹. Quiescent increased logarithmically by a factor of 0·86 with WI (sheikbody mass ratio) (Fig. 4). Habitation of the mollusc shells Muricanthus and Opeatostoma significantly increased the load rating (WI) to 1·24 ± 0·06 (N = 8, P= 0·013) and 1·45 ±0·08 (N=3, P = 0·007) respectively when compared with Nerita associations. Quiescent increased correspondingly to 138 ± 13 (8) and 245 ± 24 (3) μ mol kg⁻¹ min⁻¹. Both increases were highly significant compared to crabs inside Nerita shells (P = 9·9 × 10 ^{− 3}, 3·6 × 10 ^{− 5} respectively).

To remove gas samples, it was necessary to stop the rotation of the respirometer momentarily. Although momentum was instantly re-established, sampling was restricted to 30-min intervals to minimize any possible disturbing effect on the locomotory patterns. The and values for the initial half hour of activity were discounted since lactic acid accumulated at the start of exercise (see M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation) and we were uncertain of the extent to which anaerobic ATP production fuelled locomotion at this time. However, the increase in lactate was not cumulative and so the values of and recorded in the two subsequent time periods are representative of a ‘steady state’ and have been used to assess the total energy required to run at a given velocity. The mean values reported therefore correspond to measurements taken 45 min into the exercise bout and are plotted as a function of velocity in Fig. 5 for loaded and unloaded crabs. Velocity was calculated from the polygraph trace as revolutions per minute multiplied by the internal circumference of the jar.

and both increased linearly with velocity which is diagnostic of a number of birds, mammals and reptiles (Taylor, 1973). Crabs running with shells exhibited running speeds anywhere from 0·2 – 2·0cms⁻¹. Unloaded crabs displayed velocities at the lower end of this range. The gradient of this line is termed M_run (minimum cost of transport) and is commonly used to assess the energetic cost of locomotion (Taylor, Schmidt-Nielsen & Raab, 1970; Taylor, Heglund & Maloiy, 1982) since it is independent of an animal’s resting and and range of running velocities. As such, this index allows comparison of widely differing species. M_run represents the total amount of energy required to locomote at a given velocity and assumes that steadystate has been reached (Seeherman, Taylor, Maloiy & Armstrong, 1981; Rome, 1982). Both requirements were fulfilled in the present study.

In the case of O₂ uptake, M_run increased significantly from 0·21 – 0·55 mmol g⁻¹ km ⁻¹ as a result of shell carriage or loading. When regression coefficients for animals with and without shells were tested for homogeneity via analysis of covariance, the slopes were found to be significantly different (F_{(1), 1,36} = 9·14; 0·005 <P< 0·0025). The null hypothesis could not be rejected for vs velocity however (F_{(1), 1,24} = 1·22; P> 0·25) and a common weighted regression coefficient of 134·1 was computed. Statistically however, these two lines do not coincide, since their elevations are different (F_{(1), 1,25} = 13·9; P = 0·001).

The Y-intercept values (corresponding to zero velocity) extrapolated from these regression equations (Table 1) show good agreement with values measured in inactive animals (see above). measured in unloaded crabs was somewhat higher than predicted indicating that removal from the shell had traumatized the animal. In this same group Y-intercept values were very low suggesting that it is not advisable to extrapolate outside the measured range.

Interpolation of the mean locomotory velocity of 0·6 cm s⁻¹ onto the regression equations for crabs carrying shells produced mean active and values of 310·6 and 106·8 μ molkg⁻¹ min⁻¹. These represent respectively 3·4 and 2·6 factorial increases above inactive values. Corresponding values for crabs removed from their shells were 142·6 and 42·8 μmol kg⁻¹ min⁻¹, both of which constitute a 1·6-fold increase over resting rates.

The gross energetic cost of transport was calculated (in mmol g⁻¹ km⁻¹) from these regression lines by dividing or by velocity (see Herreid, 1981). This index exhibited an exponential decay which could be fitted, with greater than 90% confidence, by a second order polynomial for all four treatments (Fig. 6.). The asymptote was in all cases close to the value for M_run. At high velocity the gradient of this line for all treatments was consistently 1·39 mmol g⁻¹ km ^{− 2} h⁻¹. As velocity decreased, the gradient increased to as much as 26·4. The first significant departure from the asymptote was arbitrarily defined as the velocity at which the gradient doubled. This so-called ‘critical velocity’ was 1·0 cms⁻¹ for animals carrying shells (1·1 cms ^{− 1} for ⁠, 0·9cms⁻¹ for ⁠) which is fairly close to the mean velocity observed during spontaneous locomotion (0·6cms⁻¹) suggesting that animals voluntarily minimize energetic expenditure. Interestingly the critical value was lower when animals were removed from their shells (0·9cms⁻¹ for ⁠; 0·3cms⁻¹ for ⁠) which may explain why this group voluntarily ran at a lower speed.

recorded between 30min and 1h after activity ceased was 65·3 ±9·3 (13) μ molkg⁻¹ min⁻¹ which was not significantly different from the pre-exercised rate (P = 0·06). In some experiments, 20ml of 10% sea water was introduced into the respirometer by injection through the lid at the start of the recovery period (see M. G. Wheatly, B. R. McMahon, W. W. Burggen & A. W. Pinder, in preparation). Under these circumstances declined to 165·9 ± 39·4 (7) μ molkg⁻¹ min⁻¹ which was half the active rate. However, this rate was significantly higher than the preexercised level (P = 8·7 × 10 ^{− 7}). was 18·8 ± 2·6 (13) μ mol kg⁻¹ min⁻¹ in crabs recovering without water which was a significant reduction below settled values (P = 4·95 × 10 ^{− 5}). When crabs had access to water at this time the measured in the gas phase decreased further to 12·9 ±1·0 (7) μ molkg⁻¹ min⁻¹. This reduction was significant compared to settled levels (P = 7·17 × 10 ^{− 4}) but not compared to during recovery without water (P = 0·12).

This investigation presents an innovative approach to the study of exercise physiology in decapod crustaceans. Its most unique feature is that, by virtue of the design of the apparatus, the activity studied in the hermit crab C. compressus was entirely voluntary and not dictated by the experimentor. Furthermore an attempt was made to relate the laboratory exercise performance to activity patterns observed in the field.

The shell-dwelling habit creates an important distinction between the land hermit crab and other terrestrial crabs since it affords a built-in protection mechanism against adverse environmental stimuli or predators. If disturbed, crabs will withdraw into the shell using the major chela in addition to other flattened ambulatories to obstruct the aperture (Harms, 1929). The success of this species in the terrestrial environment is largely attributed to retention of the mollusc shell for purposes of concealment (Reese, 1969). For this reason terrestrial hermit crabs can avoid bursts of high speed pedestrian activity which are observed in other species such as Uca (Full & Herreid, 1984) or Cardisoma (Wood & Randall, 1981a,b; Herreid et al. 1983) and are shortlived since they are fuelled predominantly anaerobically (see review article by McMahon, 1981). This field has focused in the past on short periods of exhaustive activity (McDonald et al. 1979; Rutledge, 1980) since it is often difficult accurately to quantify exercise performance without sophisticated instrumentation.

Sustained submaximal pedestrian activity has not previously been studied since most species exhibit erratic movements when running at low speeds on treadmills (Full & Herreid, 1984). Marathon swimming activity has however been studied in the blue crab (Callinectes sapidus) which naturally undertakes long-distance migrations (Judy & Dudley, 1970). Booth et al. (1982) examined 1 h of forced swimming in this species. Their study revealed some important differences in the physiological responses to submaximal versus exhaustive activity and these have been further documented in the present study for pedestrian locomotion.

Their rapid and effective method of withdrawal from hostile conditions has enabled hermit crabs to colonize extensively the terrestrial environment. While their activity lacks the urgency of other terrestrial species, it is no less impressive especially considering the imposed load which they support in the form of the adopted molluscan shell. The conspicuous regimented motion of this particular species across beaches in the tropics has earned them the local name of ‘soldachi’ - ‘soldiers’ in translation. The nocturnal activity pattern presently reported confirms original observations made by Ball (1972) who reported that C. compressus became active at night in arid areas.

The rotating respirometer used in this study afforded us the unique ability to quantify voluntary locomotion in terms of time active, distance covered and running speed. There was a strong similarity between activity profiles in the laboratory and field. In the field at night, crabs were found as far as 24 m from their daytime retreat (Fig. 2), suggesting an upper limit of 48 m for a nightly excursion. The mean distance covered in laboratory trials was 53 m. Conservative estimates from transect data suggest that crabs moved at speeds anywhere from 0·11·0·58 cm s⁻¹. Mean speeds observed in the laboratory (0·6 cm s⁻¹) were at the upper end if this range. Compared to other terrestrial species these are moderate speeds. The fastest documented land crab is the ghost crab Ocypodegaudichaudii (Full & Herreid, 1983) which has a body mass of 2·5 g and can attain speeds of 21·1 cm s⁻¹. Larger species such as Gecarcinus lateralis (50 g) and Cardisoma guanhumi (150 g) display running speeds around 4·7cms⁻¹ (Herreid, Lee & Shah, 1979; Herreid et al. 1983). Full & Herreid (1984) attribute decreases in speed and endurance time to increase in body mass. While C. compressus has a mean body mass of 10·5 g, the total mass transported is 20 g due to the additional mass of the shell. Nonetheless the average speed reported is still comparatively low. Body length is another size criterion used in comparing exercise performance of different species. Based on the classification of Wood & Randall (1981a) a velocity of 0·2 BL s⁻¹ in Coenobita would equate to ‘moderate’ exercise for Cardisoma camifex. However, again work performed is not a simple function of BL in hermit crabs due to the load of the shell.

Crabs which fell into our category of ‘high performers’ (Fig. 3) maintained velocity throughout 3 h of activity. Based on physiological parameters which we subsequently measured (M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation) it is our opinion that exercise terminated in the laboratory as a consequence of natural activity patterns and not in response to any physiological limitation.

Standard agreed with values reported in other decapods at comparable acclimation temperatures (see McMahon & Wilkens, 1983). The discovery that standard correlated with WI (Fig. 4) was not altogether unexpected except that it was not a simple function of load rating. On transferring from a Nerita to an Opeatostoma shell, Coenobita increased the total load supported by only 25% yet increased by 2·6-fold. Most animals which carry shells are physically attached to them. In hermit crabs, the shell is held in various ways. The left uropod is enlarged and modified for hooking on to the columella of the shell. Contraction of the longitudinal abdominal muscles presses the surface of the abdomen against the inner wall of the shell and the last two pairs of legs are pushed against the wall of the shell opening. All of these gripping mechanisms are energetically expensive which means that there is a postural cost of shell habitation even when the animal is stationary. The increased cost of supporting visually larger or heavier shells is offset by the advantage afforded in territorial conflict (Hazlett, 1970).

Inactive has been reported in two other species of land crab (Herreid et al. 1979; Wood & Randall, 1981a). In both cases the respiratory quotient (RQ) measured in the gas phase was significantly less than unity, agreeing with the present findings. One major deficiency acknowledged in respiratory studies in land crabs is the extent to which branchial water is retained and subsequently used for CO₂ elimination. Based on low settled blood CO₂ tensions measured in another study (see M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation), the aquatic route could be a major pathway for CO₂ excretion in Coenobita compressus, and this would explain the spuriously low RQ values obtained in the gas phase. The problem may be exacerbated in hermit crabs since they have a greater volume of water at their disposal on account of the reservoir contained inside the whorls of the molluscan shell (see McMahon & Burggren, 1979; Wheatly, Burggren & McMahon, 1984). This being true, it is surprising that gaseous did not increase when crabs were removed from their shells. However animals were considerably traumatized in the process and so unloaded crabs were not truly quiescent. Wood & Randall (1981a) attribute the low RQ to retention of respiratory CO₂ for carapace information.

Since our protocol involved intermittent sampling of a closed respirometer, it was not our original intention to determine the kinetics of gas exchange. However, the values we obtained at 30 min after the start of exercise did not increase significantly in the remaining two sampling periods and so ‘steady-state’ rates must have been attained within that time. Since we know that lactate did not accumulate appreciably (see M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation), the aerobic response to exercise was relatively fast. A rapid increase in has been reported in response to submaximal exercise regimes in mammals (Cerretelli, Pendergast, Paganelli & Rennie, 1979), cockroaches (Herreid, Prawel & Full, 1981) and blue crabs (Booth et al. 1982). By comparison, the terrestrial crabs Cardisoma and Gecarcinus (Herreid et al. 1979, 1983) display sluggish and reduced aerobic responses. Booth et al. (1982) were able to demonstrate the rapidity of the response in Callinectes by recording scaphognathite and heart rates and O2 extraction. Physiological recordings were made on 50-g hermit crabs by McMahon & Burggren (1979). However, reducing the tubing diameter in order to fit a 10-g crab with a respiratory mask would undoubtedly increase resistance as well as being an energetic encumbrance. In the present investigation we therefore sacrificed this information in favour of obtaining reliable settled gas exchange data (see below and arguments raised by Herreid, 1981).

In Coenobita and were both proportional to velocity (Fig. 5) agreeing with a variety of other invertebrates (Herreid, 1981) and higher vertebrates (Taylor et al. 1970; Taylor, 1973; Fedak, Pinshow & Schmidt-Nielsen, 1974). Changes in gait can frequently explain a switch to a curvilinear relationship (Margaria, Cerretelli, Aghemo & Sassi, 1963). Minimum cost of transport (M_run) was calculated from the gradient of this line (Table 1) for all treatments. Expressed in terms of O₂ uptake, Mnw has previously been determined for two other species of land crab – Uca (Full & Herreid, 1984) and Ocypode (Full & Herreid, 1983). Both species have a body mass around 2·5g and an M_run value of 0·08mmol O₂g⁻¹ km⁻¹ (= 1·9 mlO₂g⁻¹ km⁻¹). Since M_run characteristically decreases with body mass on a double log plot, it is surprising that the M_run values for Uca and Ocypode are less than half the values presently measured in unloaded Coenobita and must reflect some fundamental differences in muscular efficiency which makes the smaller crabs more agile.

Based on the original equation of Taylor et al. (1970) relating M_run (in ml O₂g⁻¹ km⁻¹) to body weight (W ing) (i.e. M_run = 8·64 W ^{− 0·4}), a hermit crab of average mass 10·5 g should have an M_run value of 3·31 ml O₂ g⁻¹ km⁻¹ (=0·131 mmol O₂g⁻¹ km⁻¹). The value measured was 5·19mlO₂ g ^{− 1} km ^{− 1} (= 0·206mmolg ^{− 1}km ⁻¹), which is slightly higher than the value predicted from the empirical equation. This may relate to the unorthodox gait they exhibit. Careful observation revealed that hermit crabs brace the chelipeds in the substrate and drag the shell along in a fashion analogous to the use of handpoles by a skier. This mode of transport will increase frictional resistance and thereby energy expenditure (Denny, 1980). Together with data from Jensen & Holm-Jensen (1980) and Full & Herreid (1984), the present findings suggest that Taylor’s original equation for vertebrates is applicable to a range of invertebrates. If M_run is a direct function of mass supported, then carriage of a 10-g shell should theoretically increase M_run to 6·62mlO₂g⁻¹ km⁻¹ (= 0·262mmol O₂g⁻¹ km⁻¹). The value measured in loaded hermit crabs was double this which suggests that it requires more energy to carry each gram of an imposed load than an equivalent unit of body mass (cf. Taylor, Heglund, McMahon & Looney, 1980).

The fact that extrapolated Y-intercept values for and showed good agreement with values measured in inactive crabs suggests that the voluntary exercise regime we employed did not unduly stress the animals. Classical treadmill techniques customarily report Y-intercept values 1·7-fold in excess of settled rates (Taylor et al. 1970; Herreid et al. 1979). This discrepancy may be attributed to problems with balance when speeds are involuntarily imposed in addition to elevation in basal metabolic rates by the use of respiratory masks and recording electrodes, etc.

When running at an average speed of 0·6 cm s⁻¹, loaded crabs increased by 3·4fold producing an aerobic metabolic scope (AMS) of 219 μ molO₂kg⁻¹ min⁻¹. The majority of decapods studied have an aerobic factorial scope of 2·5 – 5·0 (see Table 1; McMahon, 1981) although Full & Herreid (1983) reporta 12-fold increase in Ocypode which is more in line with exercising fish (Brett, 1972) and mammals (Taylor et al. 1980). Carbon dioxide flux in loaded crabs did not increase correspondingly (2·6 factorial) which had a profound effect on acid-base balance. This aspect of the study is further examined in another paper (M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation). The factorial scope would appear to be a function of loading since values of 1·6 × were obtained in unloaded crabs. Furthermore, and changed concomitantly in this group suggesting that the inequality in and in loaded crabs relates to the presence of the shell water.

Gross cost of transport decreased with velocity to an asymptote of M_run (Fig. 6) which is thus largely hypothetical since it is only ever approached at high speed. Interestingly it transpired that M_run was first attained at around 1·0 cms⁻¹ which is close to the average velocity observed in the laboratory and in the field. Therefore crabs voluntarily elect to run at speeds which are least energetically expensive. Unloaded crabs could achieve the same energy economy at lower speeds which was exactly what we observed in laboratory trials (cf. Fig. 5).

recovered rapidly after the exercise bout as one would predict in the absence of an appreciable O₂ deficit. When water was made available during the recovery period it was used to replenish the branchial/shell reservoir and correct acid-base balance (see M. G. Wheatly, B. R. McMahon, W. W. Burggren & A. W. Pinder, in preparation). The intense period of exchange which ensued raised but not to exercised levels. The presence of water during recovery facilitated CO₂ excretion so that recorded in the gas phase was reduced.

This work was undertaken at the Naos Laboratory of the Smithsonian Tropical Research Institute, Republic of Panama in April and May of 1983 at which time the senior author was an Alberta Heritage Foundation Postdoctoral Fellow. Agencies gratefully acknowledged for financial support are AHMFR (MGW), NSERC (no. A 5762 to BRM), NSF (no. PCM-80-03752 to WWB) and the University of Massachusetts (WWB and AWP). The artwork was prepared by Sharon Harrison and Bill Adams.