It has been hypothesised that tau protein, when hyper-phosphorylated as in Alzheimer's disease (AD), does not bind effectively to microtubules and is no longer able to stabilise them; thus microtubules break down, and axonal transport can no longer proceed efficiently in affected brain regions in AD and related tauopathies (tau-microtubule hypothesis). We have used Drosophila models of tauopathy to test all components of this hypothesis in vivo. We have previously shown that upon expression of human 0N3R tau in Drosophila motor neurons it becomes highly phosphorylated, resulting in disruptions to both axonal transport and synaptic function which culminate in behavioural phenotypes. We now show that the mechanism by which the human tau mediates these effects is twofold: first, as predicted by the tau-microtubule hypothesis, the highly phosphorylated tau exhibits significantly reduced binding to microtubules; and second, it participates in a pathogenic interaction with the endogenous normal Drosophila tau and sequesters it away from microtubules. This causes disruption of the microtubular cytoskeleton as evidenced by a reduction in the numbers of intact correctly-aligned microtubules and the appearance of microtubules that are not correctly oriented within the axon. These deleterious effects of human tau are phosphorylation dependent because treatment with LiCl to suppress tau phosphorylation increases microtubule binding of both human and Drosophila tau and restores cytoskeletal integrity. Notably, all these phospho-tau-mediated phenotypes occur in the absence of tau filament/neurofibrillary tangle formation or neuronal death, and may thus constitute the mechanism by which hyper-phosphorylated tau disrupts neuronal function and contributes to cognitive impairment prior to neuronal death in the early stages of tauopathies.