Global ischaemia induced by interruption of cerebral blood flow results in damage to vulnerable cells, notably in the CA1 and hilar hippocampal fields, and is frequently associated with memory deficits. This review examines cognitive deficits that occur in animal models of global ischaemia in rats and monkeys, the extent to which these deficits are associated with CA1 cell loss, and the evidence for functional recovery following transplants of foetal CA1 cells and grafts of conditionally immortalised precursor cells. In rats, impairments are seen most consistently in tasks of spatial learning and spatial working memory dependent on use of allocentric environmental cues. In monkeys, ischaemic deficits have been shown to a moderate extent in delayed object recognition tasks, but animals with a selective excitotoxic CA1 lesion show a profound impairment in conditional discrimination tasks, suggesting that these may be a more sensitive measure of ischaemic impairments. Several studies have reported correlational links between the extent of CA1 cell loss following two or four vessel occlusion (2 VO, 4 VO) in rats and behavioural impairments, but recent findings indicate that at intermediate levels of damage these relationships are weak and variable, and emerge clearly only when animals with maximal CA1 cell loss are included, suggesting that the deficits involve more than damage to the CA1 field. Nevertheless, ischaemic rats and CA1-lesioned marmosets with grafts of foetal CA1 cells show substantial improvements; in rats these are not found with grafts from other hippocampal fields. Conditionally immortalised cell lines and trophic grafts are currently being assessed for their functional potential in animal models, because clinical use of foetal cells will not be practicable. Recent findings suggest that an expanded population of neuroepithelial cells derived from the conditionally immortalised H-2Kb-tsA58 transgenic mouse improve spatial learning as effectively as CA1 foetal grafts in rats subjected to 4 VO, and clonal lines from the same source show similar promise. Lines derived from precursor cells have the potential to develop into different types of cell (neuronal or glial) depending on signals from the host brain. These cell lines may therefore have the capacity to repair damaged host circuits more precisely than is possible with foetal grafts, and offer a promising, approach both to functional recovery and to elucidating graft-host interactions.