Respiratory-related activation of human abdominal muscles during exercise

Abstract

We tested the hypothesis that abdominal muscles are active during the expiratory phase of the respiratory cycle during exercise. Electromyographic (EMG) activities of external oblique and rectus abdominis muscles were recorded during incremental exercise to exhaustion and during 30 min of constant work rate exercise at an intensity of 85 % of the peak oxygen consumption rate (V̇_O2). High amplitude intramuscular EMG activities of both abdominal muscles could be evoked with postural manoeuvres in all subjects. During cycling, respiratory-related activity of the external obliques was evoked in four of seven subjects, whereas rectus abdominis activity was observed in six of the seven subjects. We measured only the activity that was confined exclusively to the expiratory phase of the respiratory cycle. Expiratory activity of both muscles increased with exercise intensity, although peak values averaged only 10-20 or 20-40 % of the peak activity (obtained during maximal, voluntary expiratory efforts) for the external oblique and rectus abdominis muscles, respectively. To estimate how much of the recorded abdominal muscle activity was supporting leg movements during exercise, we compared the activity at the very end of incremental exercise to that recorded during the first five respiratory cycles after the abrupt cessation of exercise, when ventilation was still very high. Although external oblique activity was reduced after exercise stopped, clear expiratory activity remained. Rectus abdominis activity remained high after exercise cessation, showing a gradual decline that approximated the decline in ventilation. During constant work rate exercise, EMG activities increased to 40-50 and 5-10 % of peak in rectus and external oblique muscles, respectively, and then plateaued for the remainder of the bout in spite of a continual upward drift in V̇_O2 and pulmonary ventilation. Linear regression analysis showed that the rise in respiratory-related expiratory muscle activity during progressive intensity exercise was significantly correlated with ventilation, although weakly. In constant work rate exercise, expiratory EMG activities increased, but the changes were highly variable and did not change as a function of exercise time, even though ventilation drifted significantly with time. These experiments suggest that abdominal muscles play a role in regulating the ventilatory response to progressive intensity bicycle exercise, although some of the observed activity may support postural adjustments or limb movements. The contribution of abdominal muscles to ventilation during constant work rate exercise is variable, and expiratory activity does not ‘drift’ significantly with time.

At rest, expiration depends primarily on the elastic recoil of the lungs and chest wall, with minimal recruitment of abdominal muscles. However, during heavy exercise the expiratory period is shortened markedly, and recruitment of expiratory muscles may be needed for adequate lung emptying and maintenance of the end-expiratory lung volume (EELV). Indeed, increases in end-expiratory oesophageal and gastric pressures have been reported during heavy exercise in humans, suggesting that expiratory muscle activity increases under these conditions (Henke et al. 1988). The increase in end-expiratory gastric and/or pleural pressure (and decline in EELV) that has been reported in humans during exercise is probably the result of either increased tonic or phasic respiratory-related expiratory muscle activity.

There is some evidence that expiratory muscles of the abdomen and chest wall show phasic, respiratory-related activity during exercise. Ainsworth et al. (1989) reported that mild and moderate exercise caused phasic expiratory electromyographic (EMG) activity of the triangularis sterni and transversus abdominis muscles in dogs. Gutting et al. (1988) showed that transversus abdominis EMG activity increased as a function of exercise intensity in ponies. However, the level of activation of abdominal muscles in quadrupeds during exercise may not be the same as that observed in exercising human subjects. Indeed, the gravitational forces that act on the abdominal wall differ markedly between upright and supine positions. Accordingly, motor control strategies for regulating ventilation during exercise may be very different in humans and other mammals. Choukroun et al. (1993) report EMG recordings of the transversus abdominis muscle with surface electrodes during cycling exercise in human subjects, and report a 2- to 3-fold increase in activity between exercise at 40 and 80 % of the maximal capacity. However, of the four major abdominal muscles the transversus abdominis is furthest from the skin surface, and therefore furthest from the surface electrodes. In our hands, surface recordings of even the superficial abdominal muscles (i.e. the rectus and external oblique muscles) are very difficult to obtain during exercise due to current spread from the diaphragm and intercostal muscles (see Fig. 5). Moreover, Choukroun et al. (1993) did not have their subjects perform pre-exercise volitional manoeuvres designed to activate the abdominal muscles and help with electrode placement, nor did they publish actual recordings of the electrical activity that they obtained during exercise. As a result, it is impossible to assess the quality and phasic bursting pattern of the EMG in the work of Choukroun et al. (1993), rendering the data very difficult to interpret. In 1990, Naus et al. published an abstract stating that expiratory EMG activities of internal oblique, external oblique, rectus abdominis and internal intercostal muscles recorded with surface electrodes in healthy human subjects increased during incremental exercise, and Dempsey et al. (1990) published recordings of rectus and external oblique abdominal muscles in one subject performing ‘moderate exercise’. Although the recordings in the latter study were made with intramuscular electrodes and were of very high quality, they were made in just the one subject. Moreover, the influence of both incremental and constant work rate exercise on abdominal muscle activities has not been reported in any species, including humans.

Figure 5. Influence of abrupt exercise cessation on abdominal muscle EMG activities.

Representative recordings of intramuscular (IM) RA and EO EMG activities in subject C. W. (top) and J. J. (bottom) at the end of exercise, and immediately after the abrupt cessation of exercise (‘End test’, arrow). In C. W. the surface EO activity (S EMG) was recorded, and in J. J. the surface RA activity was recorded. Inspiration is indicated by upward deflections in mouth pressure in C. W., and downward deflections in J. J. See text for more detailed explanation of the figure.

Accordingly, our goal was to test the following hypotheses: (1) human external oblique and rectus abdominis muscles show phasic, expiratory-related activity during cycling exercise, and the magnitude of the activity is correlated with the pulmonary ventilation rate; (2) the expiratory-related activity of abdominal muscles will be sustained throughout heavy constant work rate exercise, but will increase in accordance with the time-dependent increase in pulmonary ventilation (e.g. ‘ventilatory drift’, Hanson et al. 1982).

METHODS

We studied seven healthy male volunteers after they had given their informed consent in accordance with the rules established by The Human Subjects Committee at the University of Arizona, and the Declaration of Helsinki. The subjects ranged in age from 22 to 44 years, had normal lung capacity and function, and were very fit recreationally active cyclists with high peak oxygen consumption rates (V̇_O2,peak) on the cycle ergometer (Table 1).

Table 1.

Subjects' anthropometric and performance data

Subject	Age (years)	Forced vital capacity (ml)	Wpeak (W)	V̇o2,peak (ml kg−1 min−1)
C. K.	37	5092	325	57
C. R.	23	5325	350	54
T. K.	24	5800	350	59.5
S. C.	24	5600	350	63.4
J. C.	26	6250	325	49.7
J. J.	22	6400	350	63.1
A. M.	44	4650	300	49.8

Experimental protocols

The experiments involved three laboratory sessions, each taking place on separate days. The first session was used to determine the subjects’ peak work rate (W_peak) and V̇_O2 on an electrically braked bicycle ergometer (Excalibur model, Lode, Groningen, The Netherlands). Subjects began to exercise at an intensity of 100 W and the power requirement was increased 25 W each minute until the subjects were unable to complete the workload. The peak work rate was defined as the maximum power requirement maintained for 1 min. Expired minute ventilation (V̇_E) and V̇_O2 were measured and monitored online using conventional open circuit spirometry and computer technology.

The second session consisted of incremental exercise performed to exhaustion on the bicycle ergometer. The initial work rate was 10 % of each subject's peak, and was increased in 10 % increments every 2 min. Subjects were allowed to determine their pedal rates and were encouraged to maintain a constant rate throughout the test. To minimize upper body movement, subjects were instructed to grasp the handlebars and lock their elbows, and to maintain this position throughout the test. Prior to initiation of the exercise bout, the subjects were asked to perform various manoeuvres designed to elicit peak EMG activity of the abdominal muscles (see Measurement of EMG activities, below).

The third session consisted of a constant work rate exercise test on the bicycle ergometer. Control recordings were made for 6 min, followed by 3 min of warm-up exercise at 75 W, and 27 min of exercise at a work rate designed to require 85 % of the peak V̇_O2.

Measurement of ventilation and gas exchange

Inspiratory and expiratory airflow rates were measured breath-by-breath with a turbine flowmeter (Vacumed) attached to a face mask that allowed the subjects to breathe orally or nasally. This configuration is designed for very low dynamic flow resistance, which was calculated as 0.25 cmH₂O l⁻¹ s⁻¹ at a flow of 6 l s⁻¹. A small bore tube connected to a vacuum pump continuously sampled gases in the mask at a flow rate of 200 ml min⁻¹, and the fractional concentrations of CO₂ (F_ET,CO2) and O₂ (F_ET,O2) were measured with Vacumed analysers. End-tidal CO₂ was monitored with a rapidly responding CO₂ analyser (model CD-3A, Ametek, Pittsburgh, PA, USA) that sampled gas via a second port in the face mask. Gas analysers were calibrated with precision gas and the flow meter was calibrated with a precision 3.0 l syringe (W. E. Collins Inc., Braintree, MA, USA). Mouth pressure was measured from a small port in the mask to delineate the inspiratory and expiratory phases of the breathing cycle. Small bore tubing connected the port to a Validyne MP 45 pressure transducer with a diaphragm that was sensitive to ±2.5 cmH₂O. For reference, the laboratory barometric pressure ranged from 699 to 706 mmHg for all experiments, and all ventilatory variable values were corrected to body temperature and pressure, saturated (BTPS) conditions; V̇_O2 was expressed under standard temperature and pressure, dry (STPD) conditions, and end-tidal CO₂ data as ambient temperature and pressure, saturated (ATPS).

Measurement of EMG activities

In all subjects, intramuscular EMG recordings of the external oblique and rectus abdominis muscles were made by inserting two fine-wire electrodes (California Fine Wire, 0.002 in gauge) into each muscle. The wires were insulated with Formvar® except for approximately 1 mm at each end of the wire. In five subjects, surface recordings of the rectus abdominis or external oblique muscles were made with silver-silver chloride electrodes (1.0 cm diameter) placed on the skin adjacent to the intramuscular wires so that we could compare the quality of surface and intramuscular EMG activities. The electrodes were placed within 2 cm superior or inferior to the umbilicus and approximately 2.5-4.0 and 10.0-12.0 cm lateral to the umbilicus for rectus abdominis and external oblique, respectively, and securely taped.

Electrode placement was confirmed by having the subjects perform trunk rotations, seated leg lifts, and maximal expiratory efforts from end-expiratory lung volume, with and without an occluded breathing valve. These manoeuvres were repeated after each exercise test to ensure that the intramuscular activity obtained before and after exercise was the same, confirming that electrode position did not change. If the activities were different, the data were not used and the subject was asked to return on another day to repeat the test.

Data analysis

All signals were recorded on a polygraph chart recorder (model TA-11, Gould Instruments, Valley View, OH, USA) and a computerized data acquisition system (Spike II, Cambridge Electronic Design, London, UK). For the incremental exercise tests, the last 6-10 breaths of each workload were sampled, analysed, and averaged across all subjects. The endurance time for the constant work rate exercise (CWE) test was divided into 10 % epochs, and 6-10 consecutive breaths from the latter part of each epoch were chosen for analysis. Tidal volume (V_T), breathing frequency (f), and F_ET,CO2 were derived with custom-programmed data acquisition software. The EMG activity during the expiratory phase was rectified and integrated, and the average EMG was computed by dividing the area of the burst by the burst duration. This method corrects for differences in amplitude that may be the result of changes in burst duration. To avoid contamination of the phasic burst due to changes in tonic EMG activity, the average baseline for each burst was located, and only changes in activity above that baseline were subjected to the above analysis. EMG activity was expressed as a percentage of the activity obtained during the maximal non-occluded expiratory efforts. In pilot experiments we found that the non-occluded expiratory efforts provided the most reproducible activity, although the values for non-occluded and occluded expiratory efforts were not significantly different. However, the EMG amplitude recorded during the non-occluded expiratory efforts was chosen as the index of peak activity because we reasoned that these manoeuvres were more meaningful than the other two in terms of respiratory-related abdominal muscle activity. It is important to point out, however, that the EMG amplitudes recorded during the expiratory manoeuvres were considerably lower than those recorded during leg lifts, especially for the rectus abdominis muscle, which was intensely activated by leg lifts. Thus rectus abdominis activity during forced expiration averaged 43 % of the leg lift value (range 3-68 %), and external oblique activity averaged 62 % of the leg lift level (range 45-100 %).

The subjects were also asked to rate their perception of breathing effort during both exercise tests. Subjects were presented with a visual analog scale that ranged from 8 (very easy) to 20 (very, very hard); they provided scores at each workload of the incremental test and approximately every 3 min during CWE.

Statistical analysis

The average data obtained for each subject at each work rate (incremental tests) and each 10 % epoch of endurance time (CWE) were entered into a spreadsheet and subjected to one-way repeated measures ANOVA (SigmaStat). When the F statistic was significant, Dunnet's post-hoc test was used to determine which of the exercise conditions were significantly different from the corresponding control condition. To test the hypothesis that expiratory muscle EMG activity was correlated with V̇_E, we conducted linear regression analyses during progressive intensity exercise, with EMG activity as the dependent variable and V̇_E as the independent variable. To test the hypothesis that abdominal EMG activities drifted with time during constant work rate exercise, we conducted linear regression analyses with EMG activity as the dependent variable and the per cent endurance time as the independent variable (see Fig. 7). Values were considered significantly different if the P value was 0.05 or less. All values presented in the text, table and figures are means ± s.e.m., unless indicated otherwise.

Figure 7. Average changes in V̇_O2 and abdominal muscle EMGs during constant work rate exercise.

Average changes in V̇_O2 and RA and EO EMG activities during constant work rate exercise. * Different from rest (Power = 0).

RESULTS

Incremental exercise

Exercise performance

The incremental exercise tests were stopped when the required power output increased to the point where the subjects could no longer maintain the pedal rate within the selected range (i.e. the peak work rate, W_peak). All seven subjects reached 300 W, six of the seven reached 325 W, and four subjects reached 350 W (Table 1).

Changes in abdominal muscle EMG activities and gas exchange during incremental exercise

Phasic, respiratory-related abdominal muscle activity was not observed under resting conditions in any subject, and the work rate associated with the appearance of activity varied widely. On average, phasic external oblique EMG activity occurred at 57 ± 31 % of peak, and rectus abdominis activity was first noted at 48 ± 20 % of peak. In fact, only four subjects showed measurable phasic external oblique activity at any time in the test, whereas six of the seven subjects had clear rectus abdominis activity at some time during the test.

Figure 1 shows recordings of the external oblique and rectus abdominis EMGs, and mouth pressure obtained at rest and during incremental exercise in a single subject; the abdominal muscle activities obtained during a volitional, non-occluded maximal expiratory effort are also shown. In this subject, external oblique activity appeared at a power output of 100 W, rectus abdominis activity was noted at 200 W, and the activity of both muscles was brisk at W_peak, in this case 350 W.

Figure 1. EMG activities during incremental exercise in one subject.

Intramuscular (IM) rectus abdominis (RA) and external oblique (EO) EMG activities and mouth pressure (P_mouth, inspiration upwards) at rest and during incremental exercise at three work rates. The EMG responses to unoccluded maximal exhalation manoeuvres, which were used as our index of peak EMG activity, are also shown.

Figure 2 shows the average changes in V̇_O2 and in the activity of the external oblique and rectus abdominis muscles as a function of exercise power output. V̇_O2 rose linearly as expected, but did not plateau, suggesting that the subjects stopped at V̇_O2,peak, with the true V̇_O2,max unattainable, probably because of local quadriceps muscle fatigue, which is a well-known phenomenon in exercise physiology. Rectus abdominis activity increased as a function of power, but the intersubject variability was large, resulting in significant differences only at the higher work rates. The activity of the external oblique muscle was less variable, but much smaller, and significance was achieved at only the two highest work rates. Figure 3 shows the corresponding changes in V̇_E and the rating of perceived breathing effort. Both variables rose linearly as a function of power output, and were highly correlated with one another. Figure 4 shows changes in f, V_T and end-tidal CO₂ as a function of power output. Frequency and V_T rose monotonically, and at the highest work rates f rose even more steeply and V_T reached a plateau. End-tidal CO₂ rose at the lighter work rates, reached a plateau and then decreased at the higher work rates, although it never declined below the resting level.

Figure 2. Average changes in V̇_O2 and abdominal muscle EMGs during incremental exercise.

Average changes in V̇_O2 and RA and EO EMG activities during incremental exercise. Abbreviations are as given in the legend to Fig. 1. * Different from rest (Power = 0).

Figure 3. Ventilation and perception of breathing effort during incremental exercise.

Average changes in V̇_E and rating of perceived breathing effort (RPE) during incremental exercise. See text for explanation.

Figure 4. Average changes in breathing frequency and volume, and tidal CO₂ during incremental exercise.

Average changes in breathing frequency (f), tidal volume (V_T) and end-tidal CO₂ during incremental exercise. See text for explanation.

Abdominal muscle activities immediately after the cessation of peak exercise

Five of the seven subjects were able to stop abruptly at the end of exercise, giving us an opportunity to determine if the drive to the abdominal muscles was related to posture or active expiration. Figure 5 shows abdominal muscle activities and mouth pressure recordings at the end of the incremental exercise test in two of the seven subjects. In subject C. W., mouth pressure and ventilation fell modestly upon exercise cessation, and the abdominal muscle activity also fell. Nevertheless, the activity still remained elevated, even though the legs were not moving. Mouth pressure and ventilation did not fall as abruptly in subject J. J., and the abdominal activity also remained brisk, particularly in the rectus abdominis. The magnitude of change in abdominal muscle activity between peak exercise and the abrupt termination of exercise in all five subjects is shown in Fig. 6. The data represent the average of the last five breaths of the exercise test, and the first five breaths after abrupt termination of exercise. Four of the five subjects maintained rectus activity, while the fifth showed a modest reduction. In contrast, external oblique activity fell markedly upon exercise cessation in all four subjects (the fifth subject did not have a viable external oblique recording after exercise).

Figure 6. Average EMG activities during and immediately after peak incremental exercise.

The average of the five abdominal muscle EMG bursts that preceded exercise cessation (Peak exercise) are compared with the average of the first five EMG bursts following the abrupt cessation of exercise, and are connected by continuous lines. Each set of points represents data from one subject. We obtained data from five subjects for RA activity, and four subjects for EO activity. See text for further details of this analysis.

Constant work rate exercise

Exercise performance

Subjects exercised at a work rate designed to require a V̇_O2 of 85 % of peak. The work rates were well chosen, as direct measurements of V̇_O2 in the steady state showed that, on average, the subjects exercised at an intensity that required 85.3 ± 2.5 % of V̇_O2,peak (mean ±s.d.). Since all subjects were trained cyclists, all were able to complete the 27-min bout, although three of the seven were very fatigued at the end of the test.

Changes in abdominal muscle EMG activities, perceived exertion and pulmonary gas exchange during constant work rate exercise

V̇_O2 rose abruptly and reached the target level within 10-20 % of the endurance time (Fig. 7). Although V̇_O2 appeared to stabilize at the target level, there was in fact a small but significant upward drift as the exercise proceeded. Five of the seven subjects showed rectus abdominis activity, and four of the seven showed external oblique activity during the test; three of the subjects showed both rectus and external oblique activities. On average, the activities of the external oblique and rectus abdominis muscles increased with exercise intensity, but the magnitude and time course of the change appeared to be different in the two muscles (Fig. 7). The rectus activity was relatively low but consistently observed, whereas the external oblique activity was on average greater, but much more variable. The rating of perceived breathing effort rose throughout the test, but abdominal activities were at a steady plateau, suggesting that the two variables were not causally related. On the other hand, minute ventilation showed a persistent upward drift that paralleled the steady rise in perceived breathing effort (Fig. 8). The upward drift in ventilation was mediated by a rise in f with no change in V_T (Fig. 9). End-tidal CO₂ also drifted downward, although at exercise onset it actually increased above the control level and remained significantly greater than control throughout the test (Fig. 9).

Figure 8. Ventilation and perception of breathing effort during constant work rate exercise.

Average changes in V̇_E and rating of perceived breathing effort (RPE) during constant work rate exercise. See text for explanation.

Figure 9. Average changes in breathing frequency and volume, and tidal CO₂ during constant work rate exercise.

Average changes in breathing frequency (f), tidal volume (V_T) and end-tidal CO₂ during constant work rate exercise. See text for explanation.

Intramuscular vs. surface electrodes for recording the abdominal muscle EMG during exercise

Five subjects had simultaneous recordings of intramuscular and surface electrodes in one of the two muscles. Figure 5 shows recordings at the end of an incremental exercise test in two subjects, one with surface external oblique activities (subject C. W.) and the other with surface rectus abdominis activities (J. J.). In both cases, it is clear that the intramuscular recordings are composed of a wider range of motor unit sizes, higher signal-to-noise ratios and fewer artifacts during exercise; this general pattern was also observed in the other three subjects.

Linear regression analysis of expiratory muscle EMG activities in progressive intensity exercise

The results of the linear regression analysis for progressive intensity exercise are summarized in Table 2. Although the correlation between EMG activity and ventilation rate was not particularly strong for either muscle, the slope of the relation for both muscles was statistically significantly different from zero. In this population, changes in the exercise ventilation rate can explain 21 % of the change in external oblique activity, and 17 % of the change in rectus abdominis activity.

Table 2.

Results of linear regression analysis for external oblique (EO) and rectus abdominis (RA) EMG activities in progressive intensity exercise

Dependent variable	External oblique EMG
Independent variable	V̇e
Number of observations	87
Total degrees of freedom	86
Regression equation	EO EMG =−9.7 + (0.30 V̇e)
R	0.46
F statistic	22.6
t statistic for slope	4.75
P value for slope	< 0.001

Dependent variable	Rectus abdominis EMG
Independent variable	V̇e
Number of observations	87
Total degrees of freedom	86
Regression equation	RA EMG =−13.2 + (0.66 V̇e)
R	0.41
F statistic	16.8
t statistic for slope	4.1
P value for slope	< 0.001

Linear regression analysis of expiratory muscle EMG activities in constant work rate exercise

To determine if EMG activity drifted as a function of time in constant work rate exercise, we conducted linear regression analyses with EMG activity as the dependent variable and per cent endurance time as the independent variable (e.g. Fig. 7). The results of these analyses are given in Table 3. Neither external oblique nor rectus abdominis activity changes significantly with time, although a trend was noted for the external oblique (P = 0.065).

Table 3.

Results of linear regression analysis for external oblique (EO) and rectus abdominis (RA) EMG activities in constant work rate exercise

Dependent variable	External oblique EMG
Independent variable	V̇e
Number of observations	77
Total degrees of freedom	76
Regression equation	EO EMG =−20.8 + (0.3 % endurance time)
R	0.21
F statistic	3.5
t statistic for slope	1.87
P value for slope	0.065

Dependent variable	Rectus abdominis EMG
Independent variable	V̇e
Number of observations	77
Total degrees of freedom	76
Regression equation	RA EMG = 4.8 + (0.016 % endurance time)
R	0.06
F statistic	0.32
t statistic for slope	0.57
P value for slope	0.57

DISCUSSION

Our data show that, while seldom present in light exercise, moderate to heavy bicycle exercise elicits phasic activation of the external oblique and/or rectus abdominis muscles in the majority of subjects. Persistent EMG activity immediately after the abrupt cessation of exercise suggests that the muscles were activated primarily for breathing as opposed to postural control or locomotion. Although activity could be obtained with either surface or intramuscular electrodes, the signal-to-noise ratio and the detection threshold for activity was better in the latter.

Critique of methods

We detected activity in one or both of the two muscles studied in six of seven subjects during incremental exercise, and in four of seven in submaximal constant work rate exercise, which raises the possibility that our methods may be insensitive. However, with intramuscular electrodes, voluntary leg lifts and trunk twists evoked brisk, highly reproducible activity in both muscles in all subjects, with the rectus abdominis being more prominent with leg lifts and the oblique muscles with trunk twists. Similarly, voluntary forced expiratory efforts with or without an occluded airway reproducibly evoked brisk activity from both muscles in every subject studied. Although these manoeuvres also resulted in easily detectable activity from the surface electrode recordings, the magnitude of the activity was not as large and the signal-to-noise ratio was smaller compared with the intramuscular electrode recordings.

There were no obvious differences in the pattern or magnitude of the ventilatory response to exercise in those subjects without demonstrable abdominal activity compared with those that had brisk activity. Thus the failure of some subjects to use their abdominal muscles during exercise indicates that a normal ventilatory response to cycle exercise can be obtained without the activation of abdominal muscles. It should be noted, however, that we have studied only the superficial abdominal muscles and have not addressed the activities of the transversus abdominis or internal oblique muscles. The four muscle groups of the abdominal wall have different anatomical arrangements and thus probably do not contribute equally to the generation of abdominal pressure during expiration. It is possible that the three subjects that showed no activity during exercise recruited the deeper abdominal muscles, whose activity we did not record. Support for this possibility comes from a study by Abe et al. (1996), showing that the transversus abdominis was active under supine resting conditions, and with CO₂ rebreathing the order of activation of the other abdominal muscles was internal oblique, external oblique and rectus abdominis; thus we recorded the activity of the two most superficial muscles which are easier to access, but may have higher activation thresholds.

Response to incremental exercise

None of the subjects showed respiratory-related abdominal muscle activity under resting conditions (see above). During incremental exercise, six of seven subjects recruited the rectus abdominis muscles, and the threshold for recruitment ranged from 25 to 86 % of the peak work rate. Only four of seven subjects recruited the external oblique abdominal muscles, and the threshold for recruitment ranged from 23 to 100 % of the peak work rate. These values are higher than those reported by Naus et al. (1990), who reported 8 % activation of expiratory muscles. However, Naus et al. (1990) normalized expiratory muscle EMG activities to the activity obtained during a maximal voluntary contraction of the abdominal muscles (leg lifts). In contrast, our data are expressed as a percentage of the peak EMG activity elicited during a maximal expiratory manoeuvre, which produced EMG amplitudes 40-60 % lower than those obtained during leg lifts. Thus if we chose to use the EMG activity recorded during leg lifts as our index of peak activity, our values would be remarkably similar to those reported by Naus et al. (1990).

Choukroun et al. (1993) used surface electrodes to record the activity of the transversus abdominis muscle in human subjects during cycling exercise at 40, 60 and 80 % of maximal power. They reported a 2- to 3-fold increase in activity at 80 % of maximal power, which may reflect the greater activity in the deep abdominal muscles (see above). However, our own surface recordings of even the superficial muscles had relatively low signal-to-noise ratios (Fig. 5), and it is difficult to determine how Choukroun et al. (1993) were able to record transversus muscle activity with surface electrodes in exercising subjects. They report no voluntary manoeuvres, and no records that would allow assessment of the quality and precision of the recordings. Accordingly, we are doubtful that their surface recordings reflect non-contaminated activity of the transversus abdominis muscle. Dempsey et al. (1990) reported original recordings of the rectus abdominis and ‘lateral quadrant’ muscles (presumably the external oblique) during moderate cycling exercise in a single subject. Similar to our results, they showed brisk activity that remained elevated upon cessation of exercise, suggesting a ventilatory rather than a postural role for the abdominal muscles during exercise. Hecker et al. (1989) performed bilateral anaesthetic blockade of human intercostal nerves from the T-6 to the T-12 level, and reported a normal ventilatory response to exercise. From the ventilatory data alone, they suggested that abdominal and expiratory intercostal muscles play a minor role in the ventilatory response to exercise. However, the abdominal muscles receive most of their innervation from spinal nerves L1 and L2. Thus although the expiratory intercostal muscles (and perhaps the rostral levels of the rectus abdominis muscles) play a minor role in the ventilatory response to exercise, these data do not exclude an important role for abdominal muscles. Finally, studies in animal preparations show increased abdominal muscle EMG activities with increasing ventilation during exercise or CO₂ rebreathing (Gutting et al. 1988; Ainsworth et al. 1989).

The above discussion leads to the conclusion that the ventilatory response to heavy exercise in the majority of subjects involves a contribution from the abdominal muscles. The observation that end-expiratory pleural and gastric pressures become more positive in heavy exercise (Henke et al. 1988; Dempsey et al. 1990) supports this idea. However, the observation that two of our subjects had a normal ventilatory response to exercise without recruitment of the superficial abdominal muscles, coupled with the observations of Abe et al. (1996; see above), suggests that the majority of the contribution may come from the deep abdominal muscles (i.e. the internal obliques and transversus abdominis muscles). Further studies combining detailed electromyography of all four abdominal muscles with simultaneous measurement of pulmonary mechanics are needed to clarify these issues.

Constant work rate exercise

This study is the first to examine abdominal muscle activity during prolonged constant work rate exercise in human subjects, so we have no basis for comparison with other studies. Nevertheless, it appears that the abdominal muscles help to support the ventilatory response to heavy, prolonged cycling exercise. The changes in V̇_E, V_T and f, and the differences in these variables between incremental and constant work rate exercise are similar to those reported by Syabbalo et al. (1994), suggesting that our subjects had normal ventilatory responses in spite of the breathing mask and presence of the intramuscular electrodes. The V̇_O2 rose slightly but significantly with time, as did ventilation, f and the rating of perceived breathing effort. The frequency-dependent upward drift in V̇_E has been reported previously (Hanson et al. 1982).

During constant work rate exercise, five of the seven subjects showed rectus abdominis activity, four of the seven showed external oblique activity, three of the subjects showed both rectus and external oblique activities, and two subjects failed to show phasic activity in either muscle at any time during the test. The subjects that did not recruit their abdominal muscles in this test also failed to recruit them during incremental exercise (see above). The rectus abdominis activity rose slightly but significantly as soon as the subjects achieved a gas exchange steady state, and remained at the same level throughout the test. External oblique activity was more variable, and although the activity appeared to be higher at the end of the test than at the beginning, the activity did not drift as a function of time (Table 3). These data, although equivocal, raise the possibility that the progressive rise in external oblique activity during exercise contributed to the time-dependent rise in V̇_O2, together with contributions from other accessory muscles and the heart. Indirect support for recruitment of the abdominal muscles during constant work rate exercise comes from the study of Fuller et al. (1996). They examined the endurance performance of the abdominal muscles during volitional, maximal expiratory efforts to the point of fatigue. This test was performed twice on separate days, once with and once without an exhaustive bout of prolonged bicycle exercise. They found that fatigue of the abdominal muscles was significantly faster in the trial performed after the exhaustive exercise bout. This was shown not to be due to ‘generalized’ fatigue, because control studies on the inactive forearm muscles revealed no difference in performance between the two conditions. We conclude that the abdominal muscles play a supporting role in determining the ventilatory response to heavy, constant work rate cycling exercise in healthy human subjects.

Possible mechanisms

The mechanism responsible for the increase in abdominal muscle activity during exercise is unknown. Increases in blood CO₂ concentrations (Bishop & Bachofen, 1972), end-expiratory lung volume (Bishop, 1964), and increased expiratory flow resistance (Baker et al. 1979) are all potent activators of abdominal muscles. In the present experiments end-tidal CO₂ concentrations did not change significantly, or changed in a direction that would reduce abdominal muscle activities. Although we did not measure changes in end-expiratory lung volume, previous studies have shown that end-expiratory lung volume (Henke et al. 1988; Johnson et al. 1992) and pulmonary resistance (Hussain et al. 1985) decrease slightly or do not change during heavy exercise in young healthy subjects.

Paterson (1992) has shown that various ions, organic metabolites, and hormones that are released into the blood at the anaerobic threshold stimulate carotid body chemoreceptors, and increase V̇_E and f. Recent studies have shown that electrical or chemical stimulation of carotid body afferents increases abdominal muscle activities in the cat (Fregosi, 1994). Taken together, these observations raise the possibility that increased carotid body discharge during heavy exercise mediates the increase in abdominal motor activity and the subsequent rise in f and V̇_E.

Finally, as the exercising leg muscles begin to fatigue in heavy exercise, enhanced recruitment of motor units could increase central motor command (Eldridge et al. 1981), which may in turn evoke sharp increases in the activities of both inspiratory and expiratory muscles. Some support for the central command hypothesis is provided by the observation that the rating of perceived breathing effort rose monotonically with ventilation and abdominal muscle activity.

In conclusion, the majority of our subjects showed increased abdominal muscle activity during heavy progressive intensity exercise, suggesting that the abdominal muscles contribute to the increased ventilation under these conditions. The rise in ventilation during very intense exercise (i.e. when the ratings of perceived breathing effort were very high) was mediated primarily by sharp increases in breathing frequency, and corresponding reductions in the time available for expiration. Under such conditions, activation of the abdominal muscles is needed to accelerate expiratory flow, which prevents the end-expiratory lung volume from rising. This important homeostatic adjustment ensures that gas exchange will not be impaired, and that increases in the mechanical work of breathing will be minimized (Johnson et al. 1992).