Breathing in mammals is a fundamental behaviour produced by movements generated and controlled by the central nervous system. The formation of the diaphragm in mammals has led to a unique phenomenon, inspiratory led breathing. Indeed, in most mammals at rest, only inspiration is active – expiration is passive. As such, research into mechanisms of respiratory rhythm generation has focussed on the inspiratory oscillator, the preBötzinger Complex (preBötC). Although we now understand a large number of fundamental properties underlying the behaviour of the preBötC, much less is known about its expiratory counterpart that generates active expiration under conditions of increased respiratory drive, such as exercise. At birth, the expiratory oscillator, referred to as the parafacial respiratory group, is comprised of at least two neuronal subpopulations that discharge during the late expiratory phase (active expiration). One subpopulation is indistinguishable from chemosensitive neurons in the retrotrapezoid nucleus (RTN); they are CO2 sensitive, glutamatergic and have the same molecular profile; for example, they express NK1R, Phox2b and galanin but do not express TH (Onimaru et al. 2008, 2014; Bautista et al. 2018). The other subpopulation is CO2 insensitive, inhibitory and does not express NK1R, Phox2b or galanin (Onimaru et al. 2014), clearly distinct from chemosensitive RTN neurons. During development, the expiratory oscillator changes dramatically: its activity becomes suppressed at rest (Pagliardini et al. 2011); it loses its chemosensitivity; the Phox2b positive subpopulation of neurons appears to be lost; and it contains both glutamatergic and GABAergic neurons (Pagliardini et al. 2011; Biancardi et al. 2021). Rhythmic respiratory activity of the chemosensitive neurons also disappears. Based on these differences, the adult expiratory oscillator is referred to as the lateral parafacial region (pFL). The mechanisms underlying this transformation are unknown because the majority of studies into active expiratory behaviour have focussed on the motor output and defining the conditions that recruit the pFL, including REM sleep, elevated CO2 and reduced O2. In this issue of The Journal of Physiology, Magalhães et al. (2021) report performing the first electrophysiological characterisation of pFL neurons and the unmasking of the synaptic and intrinsic membrane properties that shape their expiratory discharge. A single cell polymerase chain reaction revealed not only a lack of markers of RTN neurons, but also the absence of TASK2 or GPR4 that have, somewhat controversially, been linked to the intrinsic chemosensitivity of RTN neurons. These data provide molecular evidence that the pFL and RTN are functionally distinct, and may explain the CO2/pH insensitivity in pFL neurons (Magalhães et al. 2021). Most importantly, the impressive 3 h long whole-cell recordings in an in situ preparation performed by (Magalhães et al. 2021) have revealed that the rhythmic bursting behaviour of pFL neurons during hypercapnia is determined by an interaction between intrinsic membrane properties and changes in inhibitory synaptic input. pFL neurons lack persistent sodium channels (Magalhães et al. 2021), which underlie the intrinsic oscillatory properties of many neurons, including respiratory neurons. They share many conductances with cardiac pacemakers, including TREK-1, HCN, NALCN and Cav3.1/3.2, where oscillatory bursting is induced by a TREK-1 hyperpolarising current (Unudurthi et al. 2016), which activates HCN (Wahl-Schott et al. 2014), working in conjunction with NALCN (Lu and Feng, 2012; Liang et al. 2021), to generate a slow membrane depolarisation that activates Cav3.1/3.2 and triggers the burst. Importantly, although these conductance may facilitate oscillatory firing of pFL neurons during active expiration, their relative densities do not support intrinsic oscillatory behaviour. When synaptically isolated, pFL neurons fire tonically, perhaps explaining why abdominal activity becomes tonic in the absence of network activity. Thus, the recruitment of active, rhythmic expiration by elevated CO2 does not result from the activation of intrinsic oscillatory properties but, instead, from changes in the pattern of synaptic inhibition. The expiratory oscillator is silent at normal levels of CO2 (normocapnia) because the membrane potential of pFL neurons remains subthreshold as a result of post-synaptic GABAergic and glycinergic inhibition. Although always subthreshold, their membrane potential varies throughout the respiratory cycle; they are most hyperpolarised during inspiration, more depolarised during post-inspiration as GABAergic inhibition is released and depolarised even further during phase two expiration when glycinergic inhibition is also removed (Magalhães et al. 2021). When CO2 levels increase, pFL neurons hyperpolarise even further during inspiration as a result of the increased frequency of GABA and glycinergic IPSCs. Release from this inhibition results in a greater than normal post-inspiratory depolarisation (and discharge in some cycles) as a result of a Cav3.2 T-type Ca2+ channel-mediated post-inhibitory rebound (PIR), which is enhanced as a result of greater inhibition during the inspiratory phase and reductions in post-inspiratory glycinergic IPSCs. As expiration proceeds, inhibitory signalling is further diminished (Magalhães et al. 2021) and pFL neurons continue to depolarise and discharge during phase two expiration. The data reported by Magalhães et al. (2021) significantly advance our understanding of the regions and mechanisms underlying active expiration. The proposed model also differs markedly from the contemporary hypothesis that active expiratory firing of pFL neurons is driven by a decrease in inhibitory input coupled to increased excitatory input from RTN neurons with a strong local projection to the pFL (Biancardi et al. 2021). The data also suggest that excitatory input into the pFL does not play a major role in the generation of active expiration. Glutamatergic sEPSCs were observed during all phases of the respiratory cycle and were not altered by hypercapnia. Moreover, release from fast synaptic excitation did not alter pFL neuronal firing (Magalhães et al. 2021). Thus, excitatory input may simply provide a tonic depolarising input that works in conjunction with release from inhibitory input and PIR to induce rhythmic pFL behaviour (Magalhães et al. 2021). An important caveat of the study by Magalhães et al. (2021) is that they only recorded from glutamatergic neurons, although we know local GABAergic interneurons are also present (Magalhães et al. 2021). It is therefore possible that excitatory RTN neurons project to GABAergic pFL neurons, which in turn provide inhibitory drive to pFL neurons, from which excitatory pFL neurons may rebound. This would provide a source of GABAergic input that could work in conjunction with glycinergic inputs originating from the BötC as proposed by Magalhães et al. (2021) or from inhibitory preBötC neurons (Yang et al. 2020). In summary, the exciting study by Magalhães et al. (2021) has inverted our thinking about the inner workings of the expiratory oscillator. Rather than a tonically supressed oscillator that escapes inhibition when respiratory drive is high, it now appears that inhibition is necessary to supress active expiration at rest, and also that increases in inspiratory phase inhibition and decreases in expiratory phase inhibition are necessary to drive active expiration under conditions of elevated CO2. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. The authors declare that they have no competing interests. R.T.R.H.: wrote the original article. G.D.F.: contributed through revision. Both authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed. None.