Two distinct groups of mechanisms have been proposed to account for the ventilatory response (V˙_E) to ERCs, and are therefore believed to be pertinent to the control of exercise hyperpnoea. The first group of mechanisms suggests that no specific exercise stimuli are required to explain the V˙_E–V˙_CO2 coupling during ERC, but rather, that the V˙_E response to ERC relies mainly on the chemical consequences of the contractions acting on the arterial chemoreceptors (Cross et al. 1982a). The rise in CO₂ flow, encoded through the differential receptive properties of the carotid bodies, as suggested by the study of Yamamoto & Edwards (1960), was one of the prominent hypotheses in this group of proposed mechanisms (Phillipson et al. 1981; Cross et al. 1982a; Kumar et al. 1991). Such a system, optimized with information on the mechanical properties of the respiratory system, has even been suggested to account for the entire V˙_E response, not only to ERC but also to a voluntary exercise (Poon, 1987) (the limits of this hypothesis are presented in the Discussion).

The second group of theories challenges the above mechanism, and was originally championed by Kao (Kao et al. 1963). It was proposed that non-chemoreceptor mechanisms, possibly related to the mechanical, metabolic or circulatory changes induced by the contractions (McCloskey & Mitchell, 1972), are essential to maintaining CO₂/H⁺ homeostasis during ERC and therefore during exercise. The main argument formulated to support such a contention was that it is impossible to reproduce the characteristics of the V˙_E response to ERC by changing the venous CO₂ content. Indeed, according to much of the literature (Lewis, 1975; Greco et al. 1978; Ponte & Purves, 1978; Reischl et al. 1979; Fordyce & Grodins, 1980; Huszczuk et al. 1983; Shors et al. 1983; Bennett et al. 1984; Tallman et al. 1986), the V˙_E response to venous CO₂ loading appears to be hypercapnic and hypocapnic during CO₂ unloading both at rest and during exercise. Secondly, many of these authors found that spinal cord section procedures or the destruction of the antero-lateral pathway resulted in a blunted V˙_E response or the abolition of normal response to ERC, which was reduced to its chemical component (Grodins & Morgan, 1950; Kao, 1963; Cross et al. 1982b; Shors et al. 1983). The nature of the signal capable of accounting for a coupling between V˙_E and factors proportional to the rate at which CO₂ is ultimately produced and exchanged (Dejours, 1959; Whipp & Ward, 1991) remains unclear. More recent observations offer a means of reconciling the control of P_a,CO2 during ERC with a non-chemically mediated regulation of V˙_E. The hypothesis is that the gas exchange rate in the muscles can be sensed during ERC, via a local (intramuscular) circulatory signal and could reflexly stimulate breathing through group III and IV muscle afferent fibres (Haouzi et al. 1999). In other words, it is proposed that the significant stimulus to breathe during physiological exercise and ERC may have its origin in the muscles with an afferent signal linked not only to the mechanical consequences of the contractions (Mense & Meyer, 1985; Pickar et al. 1994; Mense, 1996; Adreani et al. 1997) but also to the magnitude of the local vascular response (Huszczuk et al. 1993; Haouzi et al. 2004b), proportional to the metabolic load. Such a ventilatory regulatory system need not involve disturbance of P_a,CO2.

Using a model of functional isolation of the cephalic circulation in sheep (Haouzi et al. 2003), the present study addresses the following questions. (1) How is P_a,CO2 regulated during ERC when the chemical control of respiration is functional but unaware of the possible consequences of an increased metabolism? (2) Is the ventilation response to ERC comparable with the effects of increasing CO₂ flow in the central circulation with no CO₂/H⁺ changes in the cephalic circulation? (3) What are the effects of an acute change in the muscle circulation during ERC with a constant haemodynamic in the cephalic circulation?

Eight 2-year-old female sheep (45.9 ± 10.5 kg) were studied. All the experiments were performed in compliance with the recommendations of the Council of European Communities, and under licence from the Ministère Français de l'Agriculture et de la Pêche (registration number 54-42).

Figure 1 Schematic representation of the circulation in the model

• A cephalic compartment which is supplied by blood at any constant desired level of P_a,CO2, P_a,O2 and perfusion pressure. This compartment contains the carotid and the central chemoreceptors as well as the carotid baroreceptors.
• A systemic compartment that is supplied by blood coming directly from the pulmonary veins mixed with a constant flow of blood coming through the occipital arteries back to the aorta.

The model used in the present study was modified from the original description as follows. Firstly, the level of P_a,O2 in the cephalic and systemic circulation was maintained at 150 Torr (instead of 400 Torr in the previous series of experiments). This was intended to prevent any stimulatory effect of hyperoxia in the central nervous system. Such an effect could have been responsible for a drift in V˙_E during some of our previous tests. Second, a longer loop was used in the carotid circuit producing a delay of 20 s between the changes occurring in the systemic blood and at the site of re-injection in the carotid circuit. As a result, changes occurring in the composition of the systemic blood could be more accurately corrected in the blood supplying the cephalic circulation. Third, the vagus nerves were not cut.

The animals were tracheostomized and an inflatable-cuff tracheal cannula (Shilley no. 7, Mallinckrodt Medical, Irvine, CA, USA) was inserted through the tracheostomy. A catheter (Plastimed 3F, Plastimed, Saint-Len-La-Foret, France) was placed in one jugular vein for additional injection of anaesthetic agent.

After incision of the neck, the common carotid arteries were tied about 10–15 cm from the sinus region and incised on both sides of the ligature. Two large catheters (Bardic 18 Fr, Bard Cardiopulmonary Products Inc., Haverhill, MA, USA) were advanced rostrally and caudally into these incisions and tied to the carotid vessels. The cannulae were equipped with thin catheters that were introduced into each of the cephalic cannulae. After several pilot experiments, it was possible to select the length of the thin cephalic catheter according to the position of the cannulae to allow their tips to be close to the origin of the occipital and external carotid artery. A thin catheter was also introduced through one of the caudal cannulae and advanced into the abdominal aorta to measure systemic blood pressure.

The cephalic and caudal cannulae were connected through a double peristaltic rotor pump (Rhone-Poulenc 03, Hospal, Lyon, France). The flow delivered by the pump was determined by an electromagnetic flowmeter (Delaland Electronique, Courbevoie, France). The circuit consisted of a low-resistance hollow-fibre blood oxygenator (Terumo, Capiox SX 18, Terumo Cardiovascular Systems Corporation, USA). Blood temperature was maintained at 38.4°C by the gas exchanger and monitored (see below). The flow of O₂ and CO₂ into the inlet port of the gas compartment could be modified using a calibrated flowmeter.

To ensure permanent monitoring of the change in P_a,CO2 and P_a,O2 of the blood entering (‘systemic’ blood) and leaving (‘cephalic’ blood) the circuit, both these variables were measured continuously with a 3M CDI System 500 (Terumo, Cardiovascular Systems Corporation). This AC-powered microprocessor-based monitor, which is used during cardiopulmonary bypass in humans, allows a reliable estimate of blood P_CO2, P_O2, pH and temperature with a 20 s time-constant response (Mahutte et al. 1994; Southworth et al. 1998). Two disposable CDI^TH products, Tervhocardio-vascular system (CDI) shunt sensors made up of light-emitting diodes and fluorescent chemical microsensors were placed in a shunt bypass line, one in the circuit supplied with the systemic blood, the other downstream of the blood gas exchanger. Calibration was performed before each experiment using tonometer gases. The values obtained online by the CDI system were used to decide the time at which blood could be sampled and analysed (ABL 500 Radiometer, Copenhagen, The Netherlands), and to adjust the level of CO₂ flow in the gas port during the transient phase when systemic P_a,CO2 changes. The relationship between the blood samples analysed with the ABL 500 and the data obtained from the 3M CDI System 500 is presented in Fig. 2.

Figure 2 A, relationship between the Pa,CO2 values obtained from the 3M CDI System 500 (online Pa,CO2) and the Pa,CO2 values from arterial blood samples (Pa,CO2) (91 samples). B, frequency distribution of the difference between the online and direct (blood) measurements (more ...)

The left and right carotid blood pressure and aortic blood pressure were continuously recorded (Statham pressure transducer, Gould Goddard, Oxnard, CA, USA). The fraction of CO₂ delivered to the inlet port of the oxygenator and the pump flow rate signals were fed, along with the respiratory signals, to an analog-to-digital converter (Mac Laboratory System Adinstrument Ltd., Castle Hill, Australia) and displayed online on a monitor. The signals were digitized at 200 Hz. In addition, the data obtained from the CDI blood parameter monitoring system were fed to a microcomputer using a digital-to-analog converter.

Carotid P_CO2 and P_O2 (downstream of the gas exchanger) will be referred to as cP_CO2 and cP_O2, whereas systemic P_CO2 and P_O2 (upstream of the gas exchanger) will be referred to as sP_CO2 and sP_O2. Blood pressure from the raw data (MacLab, Maclaboratory System) was temporally aligned to the ventilatory data. The computed variables were stored on disk for further analysis.

CO₂ inhalation The animals breathed a gas mixture containing oxygen to ensure that P_a,O2 was maintained at around 150 Torr. A control arterial blood sample was taken from the carotid and from the systemic circulation, the animal was then given CO₂ to breathe for 5–6 min (mixture containing 7% CO₂). Two types of tests were performed. In the first test, cP_a,CO2 was maintained constant by adjusting the fraction of CO₂ in the gas flowing in the extra-corporeal circuit. cP_a,CO2 could be maintained constant even during the first minute of CO₂ exposure by anticipating the change in cP_a,CO2 (see Results). Online cP_a,O2 was maintained at around 150 Torr. In the second type of test, cP_a,CO2 was allowed to increase.

CO₂ tests were performed prior to ERC and were repeated on three occasions during the experiments. The three tests were averaged for each animal.

ERC with the cephalic circulation isolated We modified our technique for contracting the hindlimb muscles (Haouzi et al. 1997) to be as close as possible to the approach used by Cross et al. (1982b) in anaesthetized dogs. The protocol of stimulation, the equipment and the electrodes were identical to those used for muscle training in patients with spinal cord lesion (Stimulator Parastep, Sigmedis Inc, Fairborn, OH, USA). The hindlimbs were shaved, and two pairs of large electrodes with conductive gel were placed on both thighs (Saint Cloud International Chantonnay, France). Muscle contractions were generated using a modulated stimulation frequency of 40 Hz, applied for 2 s every 4 s. Current intensity was set to between 10 and 15 mA with a rise and fall time of 0.5 s. Muscle contractions during this stimulation produced a rhythmic extension and adduction of the hindlimbs resembling dynamic exercise. The contractions were performed against a resistance consisting of rubber bands attached to the distal part of the limbs to double V˙_E and metabolism, as previously reported (Haouzi et al. 1997). Contractions were performed with functional cephalic isolation for a period of 4–5 min. Each animal breathed either the mixture containing air and O₂ or a mixture containing 7% CO₂, which was administered 5 min prior to the onset of ERCs. The latter condition was intended, as at rest, to add an external source of CO₂ flow to the central circulation, but through both an increase in CO₂ content and flow (see Fig. 1 for additional explanations).

Vascular occlusions In three animals, a balloon-tipped catheter (Meditech, occlusion balloon catheter, 8F, Boston Scientific, Matick, MA, USA) was advanced into the inferior vena cava (IVC) via a saphenous vein and placed rostral to the iliac bifurcation. This balloon catheter allowed, on full inflation, the complete occlusion of the IVC (Haouzi et al. 1997). In two animals, a similar balloon was placed at abdominal aorta level.

The ventilatory responses to ERC obtained during the occlusion were not used to compute the response to ERC. Occlusions were performed for 45–60 s to avoid the confounding effects of ischaemic contractions on muscle afferent fibres (Kaufman et al. 1984; Mense & Meyer, 1985).

Comparisons were made only on blood samples taken when online P_a,CO2 was stable. Statistical significance was based on P values less than 0.05.

Figure 3 Example of the breath-by-breath ventilatory response to the inhalation of CO2 in one sheep when cephalicPa,CO2was maintained constant

Air breathing Cephalic P_a,CO2 was 41.9 ± 2.1 Torr and sP_a,CO2 was 43.1 ± 6.4 Torr before starting ERCs. In approximately half of the tests, we observed a transient irregular breathing pattern at the onset of ERC. This transient change in V˙_E, which consisted of a moderate depression in breathing, disappeared within approximately 15 s, and was replaced by a progressive exponential-like increase in V˙_E. Ventilation reached a steady state within 2 min (Figs 4 and 5). V˙_CO2 increased from 192 ± 23 to 317 ± 84 ml min⁻¹ at the end of the fourth minute (Δ= 125 ± 83 ml min⁻¹, P < 0.01) and V˙_E increased significantly from 5.24 ± 1.81 to 9.27 ± 3.6 l min⁻¹ (Δ= 4.03 ± 2.9 l min⁻¹, P < 0.01). The relationship between changes in V˙_E and V˙_CO2 is given in Fig. 6. Systemic P_a,CO2 decreased systematically by −2.47 ± 1.9 Torr (ΔsP_a,CO2 ranged from −6.3 to −0.2 Torr, P < 0.01). No significant changes in P_a,CO2 occurred in the cephalic circulation (ΔcP_a,CO2=−0.75 ± 1 Torr, n.s.).

Figure 4 Example of the breath-by-breath ventilatory response to ERC in one sheep when cephalicPa,CO2was maintained constant

Figure 5 Flow signal in one sheep during transition from rest to ERC

Figure 6 ΔV˙E versus ΔV˙CO2 relationship for all the tests (•) when cPa,CO2was maintained constant

The circulatory response to hindlimb contractions consisted of a slight but significant rise in systemic mean arterial pressure (+7.0 ± 9.3 mmHg, P < 0.05), which typically occurred after an initial decrease in mean blood pressure. Heart rate increased significantly by 18.6 ± 5.5 beats min⁻¹ (from 97.5 ± 14.6 to 116.1 ± 13.7 beats min⁻¹, P < 0.01).

CO₂ breathing An example of the response to ERC against a background of high systemic P_a,CO2 but unchanged cP_a,CO2 is shown in Fig. 7. At rest, V˙_E was 6.76 ± 2.21 ml min⁻¹ during 7% CO₂ breathing (versus 6.24 ± 2.52 l min⁻¹ during air breathing), and sP_a,CO2 rose from 37.6 ± 10.3 to 59.2 ± 4.6 Torr with no change in cP_a,CO2 (44.1 ± 1.3 versus 43.1 ± 2.9 Torr). At the end of the ERC tests, V˙_E reached the level of 11.4 ± 3.4 l min⁻¹ versus 10.32 ± 3.85 l min⁻¹ during air breathing (the difference was not significant). However, in contrast to ERC during air breathing, sP_a,CO2 increased systematically by an average of 4.73 ± 1.38 Torr (from +1.9 to +6.1 Torr, P < 0.01). In other words, increasing the flow of CO₂ to the arterial side during ERC produced a hypercapnic response when cP_CO2 was maintained constant.

Figure 7 Example of the breath-by-breath ventilatory response to ERC in one sheep when c Pa,CO2was maintained constant against a background of high sPa,CO2

Figure 8 summarizes, in one graph, the V˙_E versus sP_a,CO2 and cP_aCO2 changes for all the above conditions.

Figure 8 Mean (±s.d.) V˙E versus sPa,CO2and cPa,CO2relationships in all the animals during ERC and during CO2 breathing

Venous and arterial occlusions An example of consecutive occlusions is shown in Fig. 9 and average data are shown in Fig. 10. Occluding the inferior vena cava for 1 min during ERC provoked a dramatic reduction in sP_a,CO2 that was difficult to anticipate and thus cephalic P_a,CO2 dropped by −4.3 ± 4 Torr (from −11 to −1 Torr, P < 0.01). V˙_CO2 dropped by 102 ± 45 ml min⁻¹ (P < 0.01). V˙_E increased significantly from 11.99 ± 4.11 to 13.01 ± 4.63 l min⁻¹ (P < 0.05) during the occlusion. Carotid pressure remained constant during the manoeuvre (+1.4 ± 3 Torr, 119.7 ± 27.3 versus 121.1 ± 25.3 mmHg). In other words, the increase in V˙_E was not related to any change in carotid blood pressure and occurred despite a reduction in cP_a,CO2. During aortic occlusion, V˙_CO2 decreased by 59 ± 77 ml min⁻¹ and V˙_E dropped systematically from 13.92 ± 2.56 to 11.89 ± 2.69 l min⁻¹ (P < 0.05) with a small reduction in cP_a,CO2 (−1.5 ± 1.5 Torr). No change in carotid pressure was observed during the arterial occlusion (138.3 ± 27.9 versus 137.9 ± 29.4 mmHg).

Figure 9 Example of the breath-by-breath ventilatory response to ERC with arterial occlusion (A) and obstruction of the vena cava (B)

Figure 10 V˙E versus V˙CO2 relationship for all the tests during venous occlusion (left panel) and arterial occlusion (right panel) during ERC

During ERC, when the peripheral and central respiratory chemosensitive areas were isolated from changes in CO₂/H⁺ related to a rise in V˙_CO2 (increases in either oscillations or mean value of P_a,CO2), we observed that ventilation increased in proportion to the rate of gas exchange, analogous to the closed-loop condition (Morgan & Grodins, 1950; Cross et al. 1982b; Huszczuk et al. 1983, 1986). A small decrease in ‘body’P_a,CO2 was observed. When the flow of CO₂ into the systemic circulation via the pulmonary circulation was increased, at rest or during ERC, we found that there was no increase in respiratory drive to offset the subsequent rise in systemic P_a,CO2. It therefore appears that increasing CO₂ flow to the central circulation, unrelated to a metabolic challenge, cannot trigger any stimulatory effects on V˙_E provided that cephalic P_a,CO2 is not allowed to rise. Finally, the V˙_E response was exaggerated when the venous return from the contracting muscles was blocked by venous occlusion, leading to a dramatic decrease in P_a,CO2 and systemic P_a,CO2. This effect was observed in the absence of arterial baroreceptor involvement. Opposite effects were observed during arterial occlusion, as already reported in dogs during ERC (Huszczuk et al. 1993).

Firstly, we shall briefly discuss the main characteristics of the model used and its chemosensitivity to CO₂. The nature of the information that could control P_a,CO2 during contractions will then be examined and the possible implications for exercise hyperpnoea regulation discussed.

The V˙_E response to an increase in cephalic P_a,CO2 was quite similar to values we published previously and to those available in the literature (Campbell & Vagedes, 1990). The CO₂ sensitivity of the extra-cephalic compartment during CO₂ inhalation decreases from 0.69 ± 0.25 l min⁻¹ Torr⁻¹ to almost zero (0.03 ± 0.04 l min⁻¹ Torr⁻¹) when P_a,CO2 changes were prevented in the carotid circulation. A residual increase in V˙_E could still be observed in several tests, where P_a,CO2 increased in the body while cP_a,CO2 was maintained constant. This effect was smaller than in our previous report (Haouzi et al. 2003). With the present circuit, we could prevent small rises in cephalic CO₂ more effectively due to the longer delay imposed by the extra-corporeal circulation compared with the circuit used in the past. The lack of pronounced hyperoxia may have also prevented V˙_E from drifting, making our measurements more reliable. We hypothesized that this remaining, albeit small, respiratory effect was related to some of the circulatory effects of the rise in systemic CO₂ (see Haouzi et al. 2003, for discussion).

The animals were not vagotomized and, if anything, the V˙_E effects of increasing body P_a,CO2 were smaller compared with those observed in our previous study with vagotomized animals. This confirms that the aortic chemoreceptors of sheep contribute little to breathing control. Finally, when P_CO2 was increased in the airways the activity of the stretch receptors may have been depressed (Bartoli et al. 1974), an effect we would not have observed in our previous study. No changes in ventilation were observed that were indicative of a reduction in vagally mediated reflexes that mimic the effects of a vagotomy. This, however, may be due to the fact that the interaction of alveolar P_CO2 with lung mechano-receptors operates mostly in the hypocapnic range.

A low-intensity current was used to minimize the direct stimulation of afferent fibres and to prevent nociceptive stimuli (Cross et al. 1982b). The main limitation of this approach is that the metabolic rate, and thereby V˙_E, can increase at best by a factor of two. We observed a small reduction in sP_a,CO2 in this preparation. This suggests, as in the model of Bennett and Fordyce (1988), that the gain of specific exercise stimuli could be slightly higher than the total gain of the exercise hyperpnoea when all the control system mechanisms are allowed to operate. The apparent V˙_E–V˙_CO2 coupling is still preserved and does not require the transduction of any CO₂-related signal resulting from the rise in V˙_CO2.

In the present preparation, since all the chemosensitive areas were kept intact but ‘unaware’ of any CO₂- (and O₂-) related changes occurring in the body, ventilatory control systems other than chemical increase V˙_E in proportion to the metabolic rate during ERC. However, this conclusion does not agree with all the relevant literature. Indeed, as mentioned in the Introduction, there is a lack of agreement on the contribution of the respiratory chemoreceptors towards the V˙_E response to ERC. This debate originates from an ongoing controversy.

The controversy pertains to the effects of spinal cord section on the V˙_E response to ERC. Without reviewing all the elements of this debate, it is clear that the experimental evidence for spinal cord transmission of information during ERC appears to fall into conflicting categories ranging from a nearly complete absence of effect, to a reduction of the exercise hyperpnoea to its ‘chemical’ component (i.e. resulting from the chemical consequences of an elevated V˙_CO2) (Grodins & Morgan, 1950; Kao, 1963; Cross et al. 1982b; Shors et al. 1983). The latter observation is to be expected if one accepts the proposition that denervation eliminates all neural information from the exercising limbs (transection of the spinal cord has been as high as T6). In such a situation, the animal should respond to electrical exercise in the same manner as the humoral dog in the cross-circulation experiments of Kao (1963), i.e. following the ventilatory response to hypercapnia.

Attempts to explain the first type of results, i.e. the persistence of an isocapnic hyperpnoea after denervation, have been numerous. The main problem is that a very similar relationship between V˙_E and V˙_CO2 is to be expected whether specific somatic information is operating or whether the responses are mediated exclusively through the chemical control of respiration (such as after spinal cord section). The direction of the deviation in P_a,CO2, small for moderate changes in V˙_CO2, requires careful measurement of arterial blood gas and appropriate statistical analysis, as in the study by Cross et al. (1982b). Indeed, although these authors advocated the contribution of the chemical control of respiration during ERC, they found that the V˙_E responses were always slightly hypocapnic (ΔP_a,CO2=−2.1 Torr) when contraction-specific stimuli were allowed to operate, whereas P_a,CO2 increased (by +1.2 Torr) after hindlimb denervation. This difference in P_a,CO2 is precisely what would be expected if somatic information is regulating V˙_E with the spinal cord intact. These deviations in P_a,CO2 have not always been studied and identified in every experiment dealing with spinal cord section, but were clearly acknowledged in early studies (Grodins & Morgan, 1950). Finally, it is intriguing that in some studies (Levine, 1979) that reported an ‘isocapnic’ increase in ventilation during ERC after spinal cord section, carotid body denervation did not affect the persisting V˙_E response.

The transient reduction in breathing frequency that we observed in some tests at the onset of contractions (Fig. 5) has been found previously (Cross et al. 1982b). Although we did not examine the mechanisms of such an inhibitory effect, it is possible that a sudden muscle stimulation may have activated spinal reflexes which have been shown to inhibit breathing (Eldridge et al. 1981). This reflex may transiently override the stimulatory effect of the descending reticulo-spinal pathway at spinal motoneurone level (Eldridge et al. 1981).

The central nervous system therefore appears to be capable of encoding and using information related to factors proportional to V˙_CO2 independently of the chemical component of this parameter.

The effect of venous and arterial occlusions on the ventilatory response to ERC has already been reported during ERC in anaesthetized dogs (Huszczuk et al. 1993). Our results are consistent with these findings, since venous blockade stimulates breathing whereas ventilation is reduced during arterial occlusion. The present findings confirm that these effects do not rely on the arterial baroreflex.

Since a reduction in venous return to the central circulation is to be obtained during both the arterial and venous occlusions, why is the ventilatory response reduced during the arterial blockade and stimulated during the venous blockade? During the first minute of an arterial occlusion, the metabolic consequences of muscle ischaemia are minimized or absent (Kaufman et al. 1984; Mense, 1996), so the ventilatory response to contractions could only be affected by the putative consequences of the reduction in venous return, the reduction in peripheral vascular pressure and to a possible change in the magnitude of the contractions. Venous occlusion reduces venous return to the central circulation as well, but distends the venous ends of the muscle peripheral vascular bed. Therefore, in order to reconcile these two observations, it has been proposed (Huszczuk et al. 1993; Haouzi et al. 2004a) that a signal of circulatory nature originating in the peripheral vascular bed, rather than in the central circulation, could link the ventilatory response to the change in muscle circulation.

This speculation was supported by the observation that group III and IV afferent fibres take into account not only the level of muscle tension but also the level of the distension of the vascular structures in the muscle (Haouzi et al. 1999). The muscle post capillary bed appears to be an important site of mediation due to the large number of group IV endings located in the adventia of the muscular venular system.

The mechanical consequences of the contraction together with its local vascular response may well constitute the signal that is transduced, in the absence of local ischaemia, by the muscle afferents during ERC.

Finally, the present results point towards a model of respiratory control that is different from that based solely on a peripheral neural drive to breathing during exercise, independent of the metabolic production. In such a model, any error signal resulting from a discrepancy between the ventilatory outcome of this neural drive and CO₂ output would be controlled by chemoreflex feedback. Instead, it is proposed that information travelling through muscle afferent fibres, proportional to the vascular and thus the metabolic response, provides a neural link coupling V˙_E to gas exchange in the working muscles. The degree of vasodilatation in the muscle could be used by the central nervous system as a marker of flow and thus of the circulatory component of the metabolism in the exercising tissues.

It is concluded that the arterial blood CO₂/H⁺ composition during hyper-metabolism induced by rhythmic muscle contractions does not rely on the chemical regulation of breathing. Afferent fibres from the skeletal muscles appear to be capable of modulating their responses as a function of the degree of local vascular recruitment or vasodilatation. This system could provide the respiratory centres with an image of the metabolic rate changes in the tissues by sensing the extent of the vascular bed being perfused.

The authors are grateful to Bernard Chalon, Yvonne Bedez, Bernanrd Tousseul and Elizabeth Gerardt for their precious technical assistance. The study was supported by a grant from “Le Ministere de L'Enseignement et de la Recherche, EA3450” and from L'Univerisité H. Poincaré, Nancy, France.