Gas Exchange of Captive Freely Diving Grey Seals (Halichoerus Grypus)

Recent studies of some pinniped species (elephant seals, grey seals) have found that individuals may spend up to 90% of the time at sea submerged, with some dives being longer than the predicted aerobic dive limit, but without extended recovery periods at the surface (Le Boeuf et al. 1988, 1989; Hindell et al. 1991, 1992; McConnell et al. 1992a; Thompson and Fedak, 1993). Thompson and Fedak (1993) recorded extremely low heart rates from freely diving grey seals during these extended dives (down to 4 beats min⁻¹), which are similar to those recorded from forcibly dived seals (Scholander, 1940). Similarly, low heart rates have been recorded occasionally during diving in northern elephant seals (Andrews et al. 1991).

It is not known how these very long dives are achieved. Our understanding of how marine mammals are able to remain submerged has changed over the last 50 years. Early laboratory studies suggested that seals performed long anaerobic dives with long recovery periods (based on data from forcibly submerged animals; Scholander, 1940; Scholander et al. 1942). Then, fieldwork studies of freely diving animals found that aquatic mammals and birds generally carry out relatively short aerobic dives with short recovery periods (e.g. Kooyman and Campbell, 1972; Butler and Woakes, 1979; Kooyman et al. 1980; Kanwisher et al. 1981; Fedak et al. 1988). This led to the concept of the aerobic dive limit (ADL; Kooyman et al. 1980, 1983), used to define the longest dive that an animal could perform without accumulating lactate in the blood.

Recent data indicate that some species can carry out very long dives with short recovery periods, during long bouts of continuous diving (Hindell et al. 1991, 1992; Thompson and Fedak, 1993). These different diving patterns suggest that seals use a number of physiological strategies during diving. There appear to be differences between species in the strategy used for diving. For example, both grey and Weddell seals exhibit very long dives but, whereas in Weddell seals surface periods increase in duration rapidly for dives over 20 min (which correlates to blood lactic acid levels; Kooyman et al. 1980), in grey seals surface periods are reasonably constant for dives over 7 min in duration (Thompson and Fedak, 1993). This may be due to different foraging patterns (i.e. sit-and-wait predators versus actively swimming predators; Fedak and Thompson, 1993; Thompson et al. 1993).

This study aimed to examine oxygen loading, storage and utilisation in order to understand how grey seals carry out extended dives without the need for long surface periods. Gas exchange variables were measured at the surface (i.e. respiratory frequency, tidal volume, respiratory flow rates). Rapid-response breath-by-breath end-tidal O₂ and CO₂ values were measured, and it was thus possible to compute respiratory minute volumes and gas exchange.

Breath-by-breath oxygen and carbon dioxide concentrations were measured in the expired air of freely diving seals, using a custom-made mask. This enabled the rapid changes in breath-by-breath oxygen and carbon dioxide levels and respiratory airflow to be followed over a single breath and over a single surface breathing period.

The work was carried out between June 1989 and July 1992 at the Institute for Forestry and Nature Management (IBN) in Texel, Netherlands. Four adult grey seals, two male and two female (mass ranges: Luke 235–250 kg; Snot 160–185 kg; Splodge 150–165 kg; Heinz 140–160 kg) were used. Seals were kept in unheated outdoor seawater tanks, 35 m by 6 m and 1.75 m deep. One tank was partitioned into four interconnecting sections, and one section served as the respirometry chamber. This consisted of an area 8 m by 6 m, covered with netting, with one area roofed with plywood in which there was a hole. This breathing hole was covered by a mask designed to follow the shape of a seal’s head and was freely available to the seal. Initially, the dead space in the mask when the seal was breathing was approximately 1–1.5 l; this variability was due to the size and height of the seal’s head in the box. The dead space was reduced as much as possible by inner contouring of the corners of the mask. The minimum dead space was approximately 0.5 l (without accounting for the seal’s nose); further reduction in dead space restricted the seals’ access to the mask. The effect of dead space on the concentrations of end-expired O₂ and CO₂ was tested using the flushing mechanism of the mask (see system operation below). Different gas mixtures were fed into the mask to set the initial dead space gas composition (i.e. the mixture of gas which is present in the mask when the seal first begins the breathing bout). Two different flush gases were used: pure N₂ and a calibrated mixture of 12% O₂ and 5% CO₂.

For an experiment, an individual animal was introduced to the respirometry chamber and allowed to acclimate to the chamber for at least 1 h before recording began. The animals were fasted for at least 6 h before data recording, to eliminate any specific metabolic effects of feeding. Prior to each day’s set of experiments, the flow and gas analysers were calibrated and checked (see section below). Recordings began once the equipment had been set up and the animal had been acclimated to the chamber. Behavioural observations were noted by hand on a chart recorder, although it was generally not possible to observe the animal in the chamber unless the water was very clear. Recordings were made at various times during the day (between 09:00 and 24:00 h) for periods of 1–8 h. At the end of an experimental run, the gates of the respiratory chamber were opened and the seal was released into the uncovered area of the tank.

A schematic diagram of the breath-by-breath system used to sample the expired gases is shown in Fig. 1. The inspiratory and expiratory flows were measured using an ultrasonic flowmeter (BRDL Ltd, Birmingham University; 60 l s-1 version; see Woakes et al. 1987) mounted on the top of the mask so that the seal could breathe freely through the flowmeter. The gas analysers were positioned on a platform mounted on the mask to minimise the delay between sampling and analysis.

A continuous subsample of the seal’s inspiratory and expiratory flows was drawn at a constant flow rate (600 ml min⁻¹) through a small-bore capillary tube (approximately 1.5 mm diameter and 20 cm length) from the centre of the seal’s respiratory gas flow, passing through a Minisart drying filter (Sartorius; dead space approximately 0.1 ml), to two Servomex gas analysers (model 1505 Miniature Infrared CO₂ and 728 Zirconia O₂; Servomex UK Ltd, Crowborough, Sussex). The oxygen sensor required a reference gas, and the air in the vicinity of the respirometry mask was used after it had been passed through a separate drying filter. The maximum possible sample gas rates were used to minimise the delay of the system (400 ml min⁻¹ for the O₂ sample, 250 ml min⁻¹ for the O₂ reference gas, 200 ml min⁻¹ for the CO₂ sample).

The data streams from the three analysers (flow, oxygen and carbon dioxide) were sampled at a rate of 100.37 Hz, using a PCL-812 12-bit analogue-to-digital converter (PC LabCard); the data sampling and digital storage were controlled by a 16 MHz 286 Dell PC with specifically designed and written software. The output from the analysers was also recorded onto chart paper using a two-channel Watanabe Linearcorder Mk VII chart recorder (Watanabe Instruments, Japan). Respirometry data were recorded simultaneously with any heart rate data (see below) onto magnetic audio-type tapes using a four-channel FM tape recorder (Data Acquisition Ltd), so that analysis of the heart rate recording (which was carried out after the experiment) could be matched up to surface and dive periods.

The ultrasonic flowmeter calibration was checked against a Greer Electromanometer M12 with pneumotachograph flowhead (Mercury), which was only used for these calibration measurements and which had also been calibrated using commercial flowmeters (KDG flowmeters, Burgess Hill, UK). A custom-built gas syringe of 7 l was also used for volume calibration of the ultrasonic flowmeter. The overall respirometry system was calibrated in three stages. (1) The analogue-to-digital converter (ADC) input voltage gain was calibrated using constant voltage inputs and monitoring the digital output with a precision 4.5 digit electronic multimeter (Siemens B1021). (2) The gas sensors were calibrated using gas mixes (pure N₂; 12% O₂ and 4% CO₂ in N₂; 10% CO₂ in N₂; supplied and certified to 0.01% by Hoekloos Gases, Amsterdam, Holland), and the output voltages were monitored using the precision multimeter. (3) Overall system integrity was verified by monitoring the ADC output when calibration gases were passed through the gas analysers.

The computer continuously monitored the analyser outputs and data were buffered until a breath was detected. This triggered data storage until the PC detected the end of the breathing period. Each surface period was saved as a separate file, with time, seal identification, temperature, calibration data etc. in the file header. As soon as the seal submerged, the respirometry mask was automatically flushed with fresh air using a high-flow pump, by energising a solenoid valve attached to the base of the breathing box. This valve also enabled different gas mixtures to be introduced to the box, for the first inspiration or expiration of a seal upon surfacing, for various control experiments.

During the course of a data collection period (1–8 h), the zero point of the flow and gas sensors tended to drift. The flow offset was corrected by assuming that the true average flow over a surface period (some 6000 samples) was zero and making the corresponding offset adjustment in software for processing. CO₂ and O₂ offsets were corrected by using the inspired air as a baseline (20.98% O₂ and 0.03% CO₂) and assuming that the sensors had returned, during the course of the 1.3 s (or so) of inspiration, to this value. The sensors and the capillary tubing introduced lags and delays into the system. The response time of the ultrasonic flowmeter was 12 ms (for 100% response). The response time of the oxygen analyser (installed in the system) was approximately 500–600 ms, while that for CO₂ was approximately 300–400 ms. Both these values include the lag times for gas transport etc. The delays introduced by the sensors and the capillary pathways were estimated by calculating the delay between the flow transition from expiration to inspiration (where there is a very rapid change in the O₂ and CO₂ levels) and the peak end-tidal CO₂ and minimum end-tidal O₂.

The data files saved by the sample program were analysed using a separate analysis program. Each breath cycle (expiration, inspiration, pause) was identified by the corresponding transitions of the flowmeter output. The diving pattern of the seals was analysed in cycles: a dive cycle was defined as a dive plus the following breathing bout. Since the animals’ activity levels could only occasionally be observed, it was not possible to categorise the submersions into types of dives (e.g. active or asleep). Therefore, the term ‘dive’ is used throughout this paper for these voluntary submersions and does not indicate any level of activity.

Dive durations were measured from the chart paper recordings. The analysis program automatically calculated surface durations, durations of inspiration and expiration (in seconds), respiratory frequency at the surface (fR; breaths min-1), peak expiratory and inspiratory flow rates (PEFR. and PIFR, in l min⁻¹ at BTPS) and tidal volumes in litres (VT, at BTPS). Minute volume (⁠⁠, l min⁻¹) was calculated as VT×fR. Oxygen and carbon dioxide data were output as percentage end-tidal (minimum) O₂ and percentage end-tidal (maximum) CO₂ and converted to kPa (assuming relative humidity of air at 37°C is 6.26 kPa). The breath-by-breath volumes of O₂ uptake and CO₂ production were calculated by integrating the product of the instantaneous fractional percentage of oxygen (or CO₂) and the expiratory flow rate at that instant. Owing to the response times of the gas analysers, the output signal was ‘smeared’ over a longer interval than that of the input (known as convolution). Thus, the calculations for gas volumes are an underestimate (by less than 5% for CO₂ and less than 8% for O₂). The mass-specific metabolic rates for each animal (i.e. rate of oxygen consumption, ⁠, in ml O₂ min⁻¹ kg⁻¹) were calculated by summing the volume of oxygen utilised in each of the breaths over one surface period and assuming this was utilised over the whole dive+surface period. Gas exchange volumes are tabulated at STPD.

Heart rates (fH) were also recorded during some of the experiments, by real-time short-range telemetry using acoustic transmitters (Bard Holand, SINTEF, Trondheim, Norway). The transmitters were mounted laterally, just dorsal and posterior to the base of a foreflipper. Bioelectric potentials created by each heartbeat were detected via two electrodes mounted dorsally and ventrally, along the midline of the animal. The electrodes consisted of 15 mm copper discs and were pressed closely to similar-sized areas of shaven skin and held in place and insulated from the water by epoxy resin, which glued the electrode mounting to the surrounding fur (Fedak et al. 1988). The transmitters produced a signal (70–90 kHz) which was modulated by the heartbeat potential. Signals were received using an omnidirectional hydrophone and receiver (Vemco), and the output was recorded on magnetic tape. The electrocardiogram (ECG) signals were later manually decoded and logged during playback of the tape, using the Dell PC as an event recorder. The mean of 10 inter-beat intervals for each surface period was calculated and converted to surface breathing heart rates; the mean dive heart rate was calculated similarly (see Chabot et al. 1991 for a discussion on methodologies of heart rate measurements).

Blood was sampled on an opportunistic basis, whenever the seals were being restrained (on land) for some other purpose. Samples were taken either from the hind flipper plexus or from the epidural vein and were analysed for haematocrit (using a Whatman mini-centrifuge) and for haemoglobin content (using a standard Sigma kit for the spectrophotometric assay of haemoglobin concentration, [Hb]); each sample of blood was evaluated in triplicate.

The output from the analysis program was read into a Minitab spreadsheet for statistical analysis (Minitab: Pennsylvania State University 1981). Heart rates were also analysed using Minitab. The data are presented as the overall means (± standard error of the mean), calculated from the means of the four individual animals. For comparisons between animals, Student’s t-tests were carried out; the level of statistical significance was taken at the 95% (P<0.05) confidence level.

Overall, the mean dive duration was 3.8±0.65 min, the mean surface duration was 0.8±0.14 min and the mean percentage of time spent submerged was 80.5±0.43%. The data are plotted as frequency histograms in Fig. 2. There was considerable individual variation in the dive patterns recorded. Mean dive times for the various individuals ranged between 2.9±0.3 and 5.7±0.2 min; maximum recorded dive times for each individual varied and ranged from 13.1 min (Heinz ♀) to 9.2 min (Snot, ♂). The longer dives tended to occur in the late afternoon or evening when there were few outside events to disturb the animals.

The overall mean diving heart rate was 28±2 beats min⁻¹ and the mean surface heart rate was 99±2 beats min⁻¹. The heart rates for individual seals ranged from 93±0.4 to 102±0.4 beats min⁻¹ while they were breathing at the surface and from 20±0.9 to 28±0.5 beats min⁻¹ during diving. Mean haematocrit for the four seals was 50.8±2.17%, and mean haemoglobin concentration was 18.3±0.85 g 100 ml⁻¹ blood.

Table 1 presents mean values of the respiratory variables for the four seals. Fig. 3 presents the gas and flow traces recorded over a typical single breathing period; the pattern of breathing (expiration, inspiration, pause) can be clearly seen. The respiratory frequencies for the individual seals ranged from 18.0±0.3 to 20.6±0.3 breaths min⁻¹, and the tidal volumes ranged from 9.75±0.04 l (for the 250 kg male seal) to 4.86±0.02 l (for the 160 kg female seal). The total lung capacity (TLC) for each seal was estimated from the scaling relationship TLC=0.10M_b^0.96, where M_b is body mass (Kooyman, 1989), and the tidal volumes relative to TLC for each seal ranged from 49% (for the 250 kg male seal) to 35% (for the 185 kg male seal). For three of the four seals, the mean peak flows were significantly higher for expiration than for inspiration (for the 250 kg male seal: PEFR was 12.3±0.09 l s⁻¹, PIFR was 9.3±0.05 l s⁻¹). Peak flow rates increased with increasing tidal volume. There were no significant correlations between tidal volumes, peak flows or times for inspiration or expiration as a function of the time during the surface period (see Figs 4, 5).

The overall mean end-tidal partial pressure of oxygen was 11.5±0.23 kPa, and that of carbon dioxide was 4.6±0.38 kPa. The variation in end-tidal during a surface breathing period is presented in Table 2. End-tidal gas values were not significantly correlated with tidal volumes or expiratory flow rates. The end-expiratory varied throughout the surface interval, producing a wide approximately U-shaped curve (see Table 2; Figs 4, 6A), with the breaths with the lowest end-tidal occurring towards the middle of the breathing bout (about 40% of the way across the surface period; i.e. during the third to the fifth breath after surfacing; see Table 2). This phenomenon was also reflected by the (inverse) shape of the end-expiratory curve (see Figs 4, 6B).

The U-shaped response of end-tidal occurred after most of the dives, although to varying extents [for example, during bouts of very short dives (1–2 min) the surface period may only be for 2–3 breaths]. The U-shaped curve was also evident over a range of mask volumes (1.5–0.5 l of mask dead space). Fig. 6C shows the data from dives by Heinz from 1992 (mask volume 0.5 l; compare with Fig. 6A, when mask volume was approximately 1.0 l) and demonstrates that the position of the minimum end-tidal value occurs approximately 35% through the surface period regardless of dead space volume. It can be seen, with the smaller mask, that the minimum value of the end-tidal oxygen partial pressure is lower, although the position of the minimum occurs at the same time in both cases. Analysis of the end-tidal during surface periods when either 12% O₂ or pure N₂ was used as the initial gas mixture in the mask demonstrated that the very first exhalation was significantly affected by the low oxygen content in only 9 out of the 25 trials. However, in all cases, subsequent breaths follow the previously observed U-shaped response, with both the position and value of the minimum being consistent.

The values for oxygen consumption rates are presented in Table 3, and it can be seen from this and from Fig. 7 that the rate of oxygen consumption decreases with increasing dive length, from 7.9±1.12 ml O₂ min⁻¹ kg⁻¹ for dives of less than 3 min to 3.5±0.26 ml O₂ min⁻¹ kg⁻¹ for dives of greater than 9 min.

The dive durations recorded in this study on captive seals are similar to those recorded in the field from freely diving grey seals (Thompson et al. 1991; McConnell et al. 1992b; Thompson and Fedak, 1993). The very long dives occasionally performed by grey seals in the wild did not occur in this study, but the distribution of most dives around the 3–7 min range corresponds to the diving behaviour in the wild. Thus, it is possible to relate the respirometry data from this study to that in freely diving grey seals at sea.

The heart rates recorded from the four seals followed the bimodal pattern seen both in freely diving wild seals (Fedak et al. 1988) and in other laboratory studies (Ponganis et al. 1990; Williams et al. 1991): high while breathing at the surface and low while submerged. The heart rate values recorded in the present study are lower than the typical rates recorded for grey seals both in the wild and in the laboratory [i.e. heart rate while breathing: 119±1 beats min⁻¹ (mean ± S.E.M.); heart rate while diving: 38±1 beats min⁻¹; Thompson and Fedak, 1993]. The low diving heart rate may be due to the restricted situation. If the seals were slightly stressed, this may have produced a decreased heart rate while diving (i.e. a more extreme dive response than would ordinarily be the case). The lower heart rates while at the surface may be because these animals are not foraging and therefore do not need to maximise their heart rate at the surface; lower surface heart rates have been recorded from freely diving grey seals when they are around haul-out sites (Sea Mammal Research Unit, unpublished data).

The ventilatory data indicate the capacity for these seals to exchange large volumes of gases during each breathing bout. The tidal volumes are high since VT is a large proportion of the total lung capacity (TLC). For the 250 kg male seal in this study, the estimated TLC is 20.8 l (from the scaling relationship TLC=0.10M_b^0.96; Kooyman, 1989), and the average tidal volume for this seal is 9.75 l (i.e. 47% of TLC). In comparison, for a 450 kg horse, an athletic animal, the estimated TLC would be 32 l (from the relationship TLC=0.063M_b^1.02; Tenney and Remmers, 1963); the resting tidal volume is approximately 5 l, and tidal volume at a fast canter (approximately 12 m s⁻¹) is 14 l (Woakes et al. 1987; Butler et al. 1993); these values are 16 and 44% respectively of TLC.

The respiratory frequencies during the surface breathing periods were approximately 30–80% higher than would be estimated from scaling relationships for resting mammals (fR=0.84M_b^−0.26 where fR is in Hz and M_b is in kg; Worthington et al. 1991). For example, for a 250 kg mammal, this relationship would predict a respiratory frequency of 12 breaths min-1, which is 67% of that of the larger male (250 kg) in this study. Thus, the minute volumes (⁠ = VT×fR, in l min⁻¹) are much higher than would be estimated (from the allometric equation: lung ventilation rate=20M_b^0.75, in l h⁻¹; Schmidt-Nielsen, 1983). For a 250 kg mammal, the estimated minute volume would be 21 l min⁻¹, while the measured minute volume at the surface of the 250 kg male seal in this study was 159 l min⁻¹. However, if these data are recalculated as weighted values to include the time spent submerged (around 80% of the dive+surface cycle), then the overall minute volumes approach those estimated from the allometric equations. Thus, both large tidal volumes and high respiratory frequencies enable these seals to carry out the gas exchange needed for the whole dive+surface cycle during the relatively short surface interval.

The oxygen consumption rates recorded in this study fall within the range of metabolic rates measured for seals diving in the laboratory and for freely diving Weddell seals. Fedak (1986) and Fedak et al. (1988) present data for grey seals swimming in a water channel; their oxygen consumption rates range between 3.5 and 15 ml O₂ min⁻¹ kg⁻¹, increasing with increasing exercise load. Castellini et al. (1992) present data from freely diving Weddell seals, which show a decrease in rate of oxygen consumption with increasing dive duration (for short dives less than 14 min, is 5.0±0.9 ml O₂ min⁻¹ kg⁻¹; for long dives more than 14 min, is 3.4±0.6 ml O₂ min⁻¹ kg⁻¹). These are similar to the data from the present study (Fig. 7) and suggest that these seals can lower their metabolic rate during longer dives and thus conserve oxygen stores and reduce the need for anaerobic metabolism in longer dives.

The data on metabolic rates can be used to calculate the aerobic dive limits for these animals. The term ‘aerobic dive limit’ can be defined in three ways, and it should be made clear which usage is appropriate for different circumstances. The measured ADL is the longest dive duration possible, before the measured lactate levels in the blood increase appreciably after the end of the dive (Kooyman et al. 1980, 1983). The estimated ADL uses measured values of ⁠, and estimates of oxygen stores, to calculate the maximum dive time before whole-body anaerobic metabolism is required; and the predicted ADL uses allometric scaling relationships to obtain a value for (and estimates of oxygen stores), and thus to calculate the ADL. In this paper, since the metabolic rate (i.e. ⁠) has been measured, the aerobic dive limit calculated (as total available oxygen stores/rate of oxygen consumption) is called the estimated ADL.

The muscle oxygen stores for a 150 kg grey seal can be estimated as 3.26 l of O₂, assuming that the muscle mass is 30% of total body mass (Lenfant et al. 1970; Lydersen et al. 1992), the myoglobin concentration is 5.4 g 100 g⁻¹ wet muscle mass (J. Z. Reed, P. J. Butler and M. A. Fedak, in preparation), and the myoglobin combining capacity is 1.34 ml O₂ g⁻¹ myoglobin). The blood oxygen stores can be estimated as 3.31 l of O₂, assuming a blood volume of 120 ml blood kg-1 (Castellini et al. 1985), a haemoglobin concentration of 18.3 g 100 ml⁻¹ blood (this study), a haemoglobin combining capacity of 1.34 ml O₂ g⁻¹ Hb, and assuming that, since these seals hyperventilate at the surface, the venous and arterial blood will have similar levels of oxygen at the end of the surface breathing period and that the oxygen that could be available during the dive is from 95% to 20% saturation (Lenfant et al. 1970; Kooyman et al. 1983). The lung oxygen stores can be estimated as 0.92 l of O₂, assuming a TLC of 12.3 l for a 150 kg grey seal (Kooyman, 1989), a diving lung volume of approximately 50% of TLC (Kooyman et al. 1971; Kooyman and Sinnett, 1982) and that the utilisable oxygen in the lungs is 15% (Kooyman, 1973). This gives a total available oxygen store of 7.49 l of O₂.

At the overall mean metabolic rate of 5.2 ml O₂ min⁻¹ kg⁻¹ (Table 3), this would give an ADL of 9.6 min. This value relates well to the diving behaviour recorded from freely diving grey seals (Thompson and Fedak, 1993), where only 6% of the dives exceeded 10 min duration. However, since the rate of oxygen consumption, calculated for the whole dive+surface period, decreases with increased dive time, then using the mean metabolic rate possibly gives an underestimate of the ADL. If the lowest mean value for is used (3.5±0.26 ml O₂ min⁻¹ kg⁻¹), this gives an estimated ADL of 14.3 min for a 150 kg seal. The basal metabolic rate (BMR) can be estimated, according to the allometric relationship BMR=0.0101M_b^0.75 (in l O₂ min-1; Kleiber, 1975); for a 150 kg grey seal this would give a value of 430 ml O₂ min⁻¹, corresponding to an ADL of 17.3 min. These calculations assume that the all blood, muscle and lung oxygen stores are available to the seal. However, the breath-by-breath end-tidal data from this study indicate that not all these stores may be universally available within the seal.

Breath-by-breath measurements of end-tidal oxygen and carbon dioxide concentrations can be used to indicate the arterial O₂ and CO₂ levels in steady-state conditions, and to date there have been no data published on the variation of breath-by-breath end-tidal variables over a breathing bout from any pinniped. Kooyman et al. (1973) obtained end-expiratory O₂ and CO₂ measurements, using a syringe to sample the expiration of Weddell seals, and they found end-tidal values ranging between 16.5 kPa (pre-dive) and 5.3 kPa (post-dives for dives lasting longer than 20 min). The end-tidal data from this study fall within this range. However, the data from the present study suggest that the partial pressures of arterial gases vary over the breathing bout, with the minimum end-tidal occurring approximately 40% into the surface period (see Fig. 6, Table 4). This U-shaped curve was unexpected. Intuitively, it would be expected that, after a breath-hold dive, the first breath would be lowest in oxygen, and then end-tidal would increase monotonically throughout the surface period.

Experiments were carried out to assess whether this U shape could be due to dead space in the mask. The main objective was to determine whether the occurrence of relatively high initial values, preceding the minimum oxygen value, was a result of dead space. Hence, qualitatively, it is only required to investigate the presence of high initial values and lower subsequent values. These experiments demonstrated that, although there was some dead space mixing, the U shape is a ‘real’ response. The evidence that the U-shaped response is a physiological response, and is not an artefact of the sampling system, can be summarised as follows. (i) During the course of this study (1989–1992), dead space in the breathing box was reduced approximately fourfold, yet the same shape of curve persisted. (ii) The same shape of curve is seen over the range of tidal volumes in the four seals. (iii) The introduction of 12% oxygen (or nitrogen or air) into the box for the first breath had no effect on the following breaths, which still followed the same U shape as seen normally.

These experiments demonstrate that, although there was some dead space mixing in the respirometry mask (and therefore the end-tidal O₂ values quoted must be considered maximum values), the U-shaped response is a genuine animal response, as demonstrated by its invariant response over different mask volumes and different initial gas concentrations. These data can thus be interpreted as an indication of low levels of gas exchange between the lung air and the active tissues during submersion. This could be due either to a lack of gas exchange at the lung–capillary boundary or to low levels of perfusion of the active tissues.

Gas exchange at the lung–capillary interface could be reduced by shunting of the blood away from the diffusion surfaces. No anatomical lung shunts have been demonstrated, although there have been several detailed anatomical dissections of both phocids and otariids (e.g. Howell, 1930). Evidence of physiological lung shunts (caused by alveolar collapse) have only previously been related to deep diving (>50 m; Kooyman and Sinnett, 1982), and it is unlikely that this could be the cause in these shallow-diving laboratory seals. Maybe these seals are able to cause partial alveolar collapse by contracting the abdominal muscles and/or diaphragm and expelling air voluntarily from the alveoli into the lower airways (which are non-exchange surfaces; Denison and Kooyman, 1973).

The other, more likely, hypothesis is that the actively respiring tissues (such as the muscles) are being under-perfused during the dive, and that the blood oxygen stores are not being utilised by the tissues. Thus, there is little transfer of oxygen between blood and muscle during the dive; the muscles use oxygen stored in myoglobin, and these stores are depleted during the dive. On surfacing, therefore, the lung air oxygen levels are high. The normal muscle blood flow is restored, and the muscle oxygen stores are replenished, depleting the blood oxygen stores. Oxygen is taken up from lungs, and the end-tidal decreases. Once tissues are reoxygenated, the end-tidal oxygen increases. The relatively low diving heart rates recorded in these seals suggest that there is a reduction in perfusion during diving. Heart rate recordings from freely diving wild seals show that there is usually a pre-surface tachycardia, as the seal is swimming to the surface, and this may allow mixing of the depleted muscle oxygen stores with the blood oxygen stores, to minimise the surface time; in this study, no pre-surface tachycardia was recorded.

Resting muscle metabolic rates have been estimated at approximately 1.5–2.0 ml O₂ min⁻¹ kg⁻¹ (McGilvery, 1975); if the muscle oxygen store is approximately 70 ml O₂ kg⁻¹ (from 5.4 g myoglobin 100 g⁻¹ muscle and 1.34 ml O₂ g⁻¹ myoglobin), then this would last for 35 min, at resting muscle metabolic rate. Thus, the myoglobin stores would be adequate for muscle metabolism during long low-activity dives. However, if the lowest mean of 3.5 ml O₂ min⁻¹ kg⁻¹ recorded in this study is used, the ADL of a 150 kg seal (using only the muscle stores) would be 6.2 min, but this includes the metabolism of of tissues other than the muscles, which would be reliant on the blood oxygen stores. Studies of wild seals have recorded periodic tachycardia during diving (Thompson and Fedak, 1993), and it is possible that these bursts of higher heart rates are to pulse blood through the muscles, to ‘top up’ the muscle oxygen stores during a dive.

To test these hypotheses, it would be necessary to take blood samples from captive grey seals diving freely in similar situations to those used in the present study, to measure the blood gas levels and to compare the level of blood oxygen with the lung oxygen levels seen on surfacing. Whatever the mechanism may be, that such extreme physiological responses occur in a grey seal during shallow voluntary dives poses questions about the ‘normal’ dive response in these animals and suggests that they may utilise profound oxygen-conserving adaptations as a routine response to diving in some circumstances. That this is the case is also demonstrated by the extremely low heart rates recorded from freely diving grey seals, during normal foraging (Thompson and Fedak, 1993). Thus, these seals have a suite of physiological responses and can adapt flexibly to various diving situations.

In conclusion, this study has demonstrated that grey seals have high rates of gas exchange (i.e. rapid respiratory frequency and large tidal volumes) at the surface. Also, these seals have low rates of oxygen consumption during a dive bout, which decreased with increasing dive duration. These combined adaptations would enable rapid oxygen loading at the surface and low rates of oxygen utilisation while diving, thus allowing a high proportion of the time at sea to be spent submerged. Most importantly, when these grey seals are breathing at the surface, the end-tidal values follow a U-shaped response, which indicates that they are utilising relatively extreme cardiorespiratory responses, even during voluntary dives in shallow water.

We thank Peter Reijnders, Ditta Reis, Aad Sleutel, Piet-Wim van Leeuwen and Koos Zegers, of the Institute for Forestry and Nature Management (IBN) in Texel, Netherlands, for help with the seals and Colin Hunter, Bernie McConnell, Gilly Banks, Kevin Nicholas and the staff of the Sea Mammal Research Unit for technical assistance and general support. J.Z.R. was supported by a Natural Environment Research Council CASE research studentship.