Lactate has long been considered to be a potentially damaging final metabolite of anaerobic glycolysis and has received little interest by neuroscientists. Interest in lactate began to increase with the demonstration of non-oxidative glucose consumption (i.e. glycolysis) in the activated brain by a landmark PET study (Fox et al. 1988). Numerous studies have since measured lactate with the consensus that sustained focal neural activity in the CNS is inevitably accompanied by an increase in extracellular lactate (Prichard et al. 1988; Mangia et al. 2007). In this context, Pellerin & Magistretti (1994) proposed that lactate may not be a metabolic dead-end product but rather the dominant oxidative substrate for neurons. The formulation of the astrocyte–neuron lactate shuttle hypothesis has generated tremendous interest in the cellular source and fate of lactate, to the point that ‘lactate’ has become a contentious term and is frequently associated with academic disputes. In this issue of The Journal of Physiology, Caesar et al. (2008) open a new perspective on the generation of extracellular lactate. As is often the case, this intriguing study advances the field but also raises more questions than it answers. Since the original report by Pellerin & Magistretti, it has been widely assumed that lactate production takes place in astrocytes and is triggered by astrocytic glutamate transporters. In stark contrast to this view, Caesar et al. provide evidence that lactate production is the consequence of AMPA receptor activation. They report in vivo microdialysis measurements of extracellular lactate concentration in the molecular layer of rat cerebellum. These measurements are interpreted in the context of changes in cerebral blood flow, electrical activity, tissue oxygen tension and regional glucose utilization. The study design is straightforward: sustained climbing fibre stimulation at 5 Hz induces neural activity in Purkinje cells with a robust rise in extracellular lactate and a correlated rise in oxygen consumption in the molecular layer of the cerebellum. The effect of AMPA receptor blockade is investigated. According to the prevailing view that lactate production is triggered by astrocytic glutamate uptake, CNQX should have no effect. Surprisingly, climbing stimulation under CNQX did not lead to an observable rise in lactate. The wide-reaching conclusion is that lactate production in the cerebellar cortex is mediated by AMPA receptors rather than glutamate transporters. Importantly, Duan et al. (1999) have previously shown that CNQX has no effect on astrocytic glutamate uptake. Consequently, the proposed glutamate transporter-mediated pathway is left intact. With the proposal of a glutamate transporter-independent pathway and the possibility of a postsynaptic site for lactate production, the paper by Caesar et al. provides a double challenge to the lactate shuttle hypothesis. Nevertheless, the authors overtly state that their study still respects the possibility of lactate production and shuttling from Bergmann glia to neurons. Indeed, AMPA receptors are present on both neuronal (Purkinje cells) and glial cells (Bergmann astrocytes) in the cerebellar cortex. Consequently, the current study which relies on CNQX blockade of AMPA receptors alone cannot answer the pressing question of whether lactate production is of neuronal or glial origin. Future studies using NASPM, an antagonist of Ca2+-permeable AMPA receptors expressed by Bergmann glia may clarify this issue. While CNQX clearly abolishes transient lactate increases, it should be considered that CNQX essentially silences all measured responses in the cerebellar cortex with loss of electrical activity, the blood-flow response, tissue oxygen use and glucose uptake (Offenhauser et al. 2005; Caesar et al. 2008). This of course does not prove that all these processes are mediated by AMPA receptors. To effectively rule out (or verify) glutamate transporters as mediators of lactate secretion, direct manipulations of glutamate transporter activity or expression are necessary. In contrast to previous studies (Hu & Wilson, 1997; Mangia et al. 2003) lactate dips during the early phase of neuronal activation were not observed. These discrepancies can always be attributed to divergent stimulus protocols. However, microdialysis measurements report only bulk changes in extracellular metabolites with limited temporal and spatial resolution, and rapid and localized metabolic events may be averaged out. Given the cellular and metabolic heterogeneity of nervous tissue on the micrometer scale (Fig. 1), measurements that directly resolve cellular responses (Kasischke et al. 2004) are ultimately required. A closer look at the cerebellar cortex may explain why: astrocytes and neurons are the principal structural elements in the molecular layer (Fig. 1A). Their processes are intimately associated in a convoluted, sometimes parallel spatial arrangement, implying that the distances for trans-cellular diffusion and transport between adjacent processes are appreciably shorter than the distances within a single cell (Fig. 1B). Astrocytic glutamate transporters are strongly expressed in the cerebellar cortex and they are perfectly positioned as a coupling site for neuron–glia interactions. In fact, astrocytic glutamate update sites ensheath Purkinje cells so tightly that their somas are delineated by a continuous EAAT2 band (Fig. 1C). The cellular distribution of cytochrome oxidase in the cerebellar cortex (Fig. 1D and E) implies that neuronal elements are profoundly oxidative, while the Bergmann glia exhibit considerably lower oxidative capacities, indirectly supporting the Pellerin & Magistretti model. Glutamate transporters and oxidative capacities in the rat cerebellar cortex A, immunohistochemistry of neurons (MAP2) and astrocytes (GFAP) in the rat cerebellar cortex. The radially orientated Bergmann glia define the molecular layer. B, neuronal and astrocytic processes are closely intertwined. Astrocyte processes directly adjoin Purkinje cell somas. C, the astrocytic glutamate transporter EAAT2 is ubiquitously expressed in the molecular layer and delineates Purkinje cells. D, Purkinje cells are profoundly oxidative cells, as demonstrated by full colocalization with cytochrome oxidase (CO). Their dendrites and interneurons in the molecular layer also exhibit a high degree of colocalization. E, Bergmann glia colocalize scarcely with CO, indicating low oxidative capacities. Scalebars: A, 100 μm; B–E, 20 μm. We credit Patricia J. Fisher and Karl A. Kasischke (Cornell-University, Ithaca, NY) for the figure. The study by Caesar et al. suggests an unexpected pathway for lactate production in the CNS and diverts our attention to AMPA-mediated events in Purkinje cells and Bergmann glia. It will be exciting to see whether future studies can directly identify the cellular origin of extracellular lactate and its ultimate fate.