The dynamic nature of synaptic connections has presented morphologists with considerable problems which, from a structural perspective, have frustrated the development of ideas on synaptic plasticity. Gradually, however, progress has been made on concepts such as the structural remodelling and turnover of synapses. This has been considerably helped by the recent elaboration of unbiased stereological procedures. The major emphasis of this review is on naturally occurring synaptic plasticity, which is regarded as an ongoing process in the postdevelopmental CNS. The focus of attention are PSs, with their characteristically discontinuous synaptic active zone, since there is mounting evidence that this synaptic type is indicative of synaptic remodelling and turnover in the mature CNS. Since the majority of CNS synapses can only be considered in terms of their relationship to dendritic spines, the contribution of these spines to synaptic plasticity is discussed initially. Changes in the configuration of these spines appears to be crucial for the plasticity, and these can be viewed in terms of the significance of the cytoskeleton, of various dendritic organelles, and also of the biophysical properties of spines. Of the synaptic characteristics that may play a role in synaptic plasticity, the PSD, synaptic curvature, the spinule, coated vesicles, polyribosomes, and the spine apparatus have all been implicated. Each of these is assessed. Special emphasis is placed on PSs because of their ever-increasing significance in discussions of synaptic plasticity. The possibility of their being artefacts is dismissed on a number of grounds, including consideration of the results of serial section studies. Various roles, other than one in synaptic plasticity have been put forward in discussing PSs. Although relevant to synaptic plasticity, these include a role in increasing synaptic efficacy, as a more permanent type of synaptic connection, or as a route for the intercellular exchange of metabolites or membrane components. The consideration of many estimates of synaptic density, and of PS frequency, have proved misleading, since studies have reported diverse and sometimes low figures. A recent reassessment of PS frequency, using unbiased stereological procedures, has provided evidence that in some brain regions PSs may account for up to 40% of all synapses. All ideas that have been put forward to date regarding the role of PSs are examined, with particular attention being devoted to the major models of Nieto-Sampedro and co-workers.(ABSTRACT TRUNCATED AT 400 WORDS)