How do circuits generate behaviour? The first approach to find out how something works is to lesion it. This strategy has been successful in neuroscience, whether for looking at genes, cells or brain regions. Sometimes, the results of lesions are highly revealing; but lesions do not always produce clear-cut effects. Permanent knockouts or lesions, even if cell-type- and age-specific, can allow compensations to develop, so fogging the interpretation(s). To overcome compensations, rapid and reversible interference is valuable. “Reversible lesionings” of brain regions, whether by anaesthetics, cooling, or GABA_A receptor agonists have been used in thousands of studies. Reversible studies allow us to find out not only “where”, but also “when” and for “how long”. To take such analysis further, it would be ideal to reversibly inhibit or activate individual cell types within a chosen region on a fast time scale in vivo (Wulff and Wisden, 2005). We could then examine the animal’s behaviour and model how the activity of this particular type of cell contributes. We would need a receptor, or an activatable switch, that could be placed uniquely in the neurons whose activity we wish to influence. If we could have a drug specific enough, a pharmacological approach will be better than a genetic one because using drug antagonists is fast and reversible. We would give a chemical or drug which uniquely binds to this receptor, and only the neurons with the sensors will be activated or inhibited. Neurons without the receptor could not respond. This approach is called “pharmacogenetics”: the prefix “pharma” because we use drugs; and “genetics” because we determine the drug’s specificity by using genes to make the protein sensor in a restricted way, so that the sensor protein appears only in particular types of cell. Although conceptually simple, there have been technical problems to achieve this pharamacogenetic approach. Nevertheless, very good progress has been made. A peptide-receptor system from Drosophila (allatostatin and its receptor, which couples to potassium channels) can reversibly silence mammalian neurons. The method will have many applications, and can be used across species; on the other hand, the low tissue penetration of the allatostatin peptide will make it harder to hit distributed cell populations in the living animal. The allatostatin method has been used elegantly to investigate the function of spinal cord interneurons (Gosgnach et al., 2006). We and our collaborators invented a pharmacogenetic method which will provide researchers with a valuable new approach to investigating brain circuits. By making a point mutation (F77I) in the GABA_A receptor γ2 subunit gene, we engineered mice that were insensitive to zolpidem (Cope et al., 2004). Usually zolpidem works at many types of neurons in the brain, allosterically enhancing the actions of γ-aminobutyric acid (GABA) at a ligand-gated chloride channel – the GABA_A receptor. Zolpidem increases peak current amplitude prolongs the decay of the inhibitory postsynaptic current. In our “zolpidem insensitive mice”, GABA still works normally at its receptor, but the mice are resistant to the effects of zolpidem. We then used genetics to reintroduce zolpidem sensitivity back to just one type of neuron in the mice. This means that when the mice are given zolpidem (blood injection), instead of many neurons being affected all over the brain, now just one type of cell is affected. We can investigate how changing the activity of this one type of cell influences behaviour. We have demonstrated the zolpidem method’s feasibility by engineering mice so that only cerebellar Purkinje cells were selectively sensitive to the drug (Wulff et al., 2007). In these mice a cell-specific swap of GABA_A receptor γ2 subunits occurred, resulting in zolpidem sensitivity restricted to Purkinje cells. In these mice zolpidem produced ataxia, whereas no ataxia occurred in drug-insensitive littermates. We concluded that by making Purkinje cells uniquely zolpidem-sensitive, so that we can up-regulate the GABAergic drive selectively onto these cells, acute modulation produces ataxia – and thus basket/stellate cell input is continually modulating the Purkinje cell output. We also made Purkinje cell-specific γ2 knockouts – these mice are not ataxic; thus from the “static knockout”, we would conclude that the molecular layer interneuron input onto Purkinje cells is dispensable for motor co-ordination, an opposite conclusion from that reached with the “zolpidem method”. These two approaches on the same cell type, the static knockout of GABAergic input, and the reversible modulation of GABAergic input, illustrate the desirability of pursuing both strategies to get a full picture (Wulff et al., 2007). The “zolpidem method” can be applied to a wide range of synapses that use GABA_A receptor inputs.