1. We recorded phrenic nerve activities and single unit firing of mesencephalic neurones in unanaesthetized supracollicularly decerebrated, paralysed and ventilated cats, in which vagi and carotid sinus nerves had been ablated. We made these measurements first at low levels of respiratory drive associated with normal PCO2 levels, then with increased respiratory drive and levels of phrenic activity produced by hypercapnia or by carotid sinus nerve stimulation. 2. We found that at least a quarter of the neurones in the central tegmental field of the mesencephalon, which were irregularly tonic or silent at low respiratory drives, developed a rhythmic increase of firing associated with each respiration. There appeared to be a threshold at about 50% of maximum respiratory activity, below which the respiratory-associated rhythm did not occur. Above this level, neuronal firing increased in graded fashion with increasing magnitude of respiratory activity. The latency from onset of phrenic activity to onset of increased neuronal firing was quite long (1.0 s) at drives just above the threshold but shortened to as little as 0.3 s as drive increased towards its maximum. 3. Cutting the spinal cord at C1-C2 had no effect on the ability of increased respiratory activity to generate a respiratory-associated rhythm in mesencephalic neurones. 4. Short-lasting anaesthesia with the agent Saffan caused mesencephalic neurones to lose the respiratory-associated rhythm with little change in phrenic activity and no change in respiratory cycle timing. 5. We also found a mesencephalic response to ventilator-induced chest expansion. The latency of the response from onset of expansion, indexed by fall of airway PCO2, to onset of neurone firing was shorter (0.2 s) than that found with the respiratory-associated rhythm. In seventeen neurones we found both the respiratory-associated rhythm and the independent ventilator-associated rhythm. 6. We interpret our findings to show that the respiratory-associated rhythmic firing of midbrain neurones is not primarily involved in generation or modulation of the motor function of the respiratory oscillator. We believe, instead, that these neurones are part of a sensory pathway conveying information about the magnitude of central neural respiratory drive, as well as spinally transmitted information from receptors in the chest wall, to thalamus and cortex. We suggest that the sensation ultimately generated may be that of 'air hunger' or dyspnoea.