The behaviour of inspiratory motoneurones is poorly understood in humans and even for limb muscles there are few studies of motoneurone behaviour under concentric conditions. The current study assessed the discharge properties of the human phrenic motoneurones during a range of non-isometric voluntary contractions. We recorded activity from 60 motor units in the costal diaphragm of four subjects using an intramuscular electrode while subjects performed a set of voluntary inspiratory contractions. These included a range of inspiratory efforts above and below the usual tidal range: breaths of different sizes (5-40% vital capacity, VC) at a constant inspiratory flow (5% VC s−1) and breaths of a constant size (20% VC) at different inspiratory flows (2.5-20% VC s−1). For all the voluntary tasks, motor units were recruited throughout inspiration. For the various tasks, half-way through inspiration, 61-87% of the sampled motor units had been recruited. When the inspiratory task was deliberately altered, most single motor units began their discharge at a particular volume even when the rate of contraction had altered. The initial firing frequency (median, 6.5 Hz) was consistent for tasks with a constant flow regardless of the size of the breath. However, for breaths of a constant size the initial firing frequencies increased as the inspiratory flow increased (range across tasks, 4.8-9.3 Hz). The ‘final’ firing frequency at the end of inspiration increased significantly above the initial frequency for each task (by 0.8-5.2 Hz) and was higher for those tasks with higher final lung volumes and higher inspiratory flows (range across tasks, 7.8-11.0 Hz). There was no correlation within a task between the time of recruitment and the initial or final firing frequency for each motor unit. However, for each inspiratory task, initial and final firing frequencies were positively correlated. Because the discharge of three to four units could be recorded simultaneously in a range of tasks, a quantitative ‘shuffle’ index was developed to describe changes in their recruitment order. Recruitment order was invariant in the task with the slowest inspiratory flow, but varied slightly, but significantly, in tasks with higher inspiratory flows. The discharge rates of single motor units were compared for targeted voluntary breaths and non-targeted involuntary breaths which were matched for size. There were no significant differences in the initial or final firing frequencies, but recruitment order was not always the same in the two types of breath. In the anaesthetized cat, the resting membrane potential of inspiratory motoneurones fluctuates with the respiratory cycle as they are depolarized during inspiration and hyperpolarized during expiration (Sears, 1964; Smith et al. 1988; for review see Monteau & Hilaire, 1991). In the cat, two separate pools of phrenic motoneurones have been described, those recruited ‘early’ and those recruited ‘late’ in inspiration (Hilaire et al. 1972; Nail et al. 1972; Berger, 1979; Dick et al. 1987). Activation of the diaphragm is achieved both by increases in the discharge frequency of phrenic motoneurones and by recruitment of additional motoneurones (Hilaire et al. 1972, 1983; Nail et al. 1972; Iscoe et al. 1976; Road & Cairns, 1997). In spinal and anaesthetized cats, recruitment of the phrenic motoneurones is believed to follow the ‘size principle’ (Iscoe et al. 1976; Dick et al. 1987; see below). Much of the knowledge about the properties of single motor units in human limb muscles comes from experiments involving contractions under isometric conditions. Under these conditions, motoneurones increase their discharge rate as drive increases (see Kernell, 1965) and are usually recruited in a stable order (Milner-Brown et al. 1973; Desmedt & Godaux, 1977; cf. Grimby & Hannerz, 1977; Stephens et al. 1978; Nardone et al. 1989; Howell et al. 1995). The stable recruitment order is thought to reflect the ‘size principle’ by which, if all motoneurones in a pool receive the same excitatory and inhibitory drives, then recruitment is determined by factors related to the size of the motoneurones (Henneman, 1957; for review see Henneman & Mendell, 1981; Binder et al. 1996). Recent studies of motor unit recruitment in limb muscles have established the existence of task-dependent heterogeneous activation of the motoneurone pool and subsequent changes in recruitment patterns of motor units within a muscle (ter Haar Romeny et al. 1982, 1984; Chanaud et al. 1991; Riek & Bawa, 1992; Puckree et al. 1998), while in other muscles, such as first dorsal interosseous, recruitment order is stable regardless of the task (e.g. Enoka et al. 1989; Jones et al. 1994). However, for limb muscles there have been few studies of motor units during natural cyclic behaviour or when muscle length changes (Grimby, 1984). Biceps brachii and first dorsal interosseous have been studied during both isometric and isotonic contractions when motor unit recruitment order was not different (Thomas et al. 1987; Tax et al. 1989). The present study examined the behaviour of human diaphragmatic motor units during a range of carefully controlled voluntary breathing tasks, while the muscle shortened. Data on the firing rates of human inspiratory motoneurones come from recordings of single motor unit activity in the parasternal intercostal muscles (Whitelaw & Watson, 1992; Gandevia et al. 1996), the scalenes (Gandevia et al. 1996) and the diaphragm (De Troyer et al. 1997). During quiet breathing, the discharge frequency of motor units for the parasternal intercostal muscles is ∼11 Hz (Whitelaw & Watson, 1992; Gandevia et al. 1996). For the diaphragm, the discharge frequency is ∼10 Hz in control subjects but increases to ∼18 Hz during quiet breathing in patients with chronic obstructive pulmonary disease (De Troyer et al. 1997) and to ∼18 Hz in control subjects with increased chemical drive to breathe (Gorman et al. 1999). In human subjects, the recruitment order of parasternal intercostal motor units appears to be stable during normal breathing (Watson & Whitelaw, 1987). The present study was designed (i) to determine whether there are different groups of diaphragmatic motor units based on early and late recruitment; (ii) to measure the increases in discharge frequency of diaphragmatic motor units during voluntary increases in inspired volume and inspiratory flow; and (iii) to devise a method to assess recruitment order of multiple units and to apply it to determine the extent of changes in diaphragmatic motor unit recruitment. Some of the data have been presented as an abstract (Butler et al. 1997).