The calcium-binding protein calretinin (CR) is present in a subpopulation of local-circuit neurons in the mammalian cerebral cortex containing gamma-aminobutyric acid. This light microscopic investigation provides a detailed qualitative and quantitative morphological analysis of CR-immunoreactive (CR+) neurons in the medial prefrontal cortex (mPFC; areas 24a,b,c, 32', and 25) of the normal adult human. The morphology of CR+ neurons and their areal and laminar distributions were consistent across human mPFC. The principal organisational features of CR+ labelling were the marked laminar distribution of immunoreactive somata and the predominantly vertical orientation of labelled axon-like and dendritic processes. Several types of CR- neurons were present in layer 1, including horizontally aligned Cajal-Retzius cells. In layers 2-6, CR+ neurons displayed a variety of morphologies: bipolar cells (49% of CR+ population), vertically bitufted cells (35%), and horizontally bitufted cells (3.5%). These neuron types were mainly located in layer 2/upper layer 3, and their dendritic processes were commonly aspiny and sometimes highly beaded. Aspiny (8%) and sparsely spiny multipolar (5%) CR+ neurons were also found. The mean somatic profile diameter of CR+ cells was 11.6 +/- 0.3 microm (mean +/- S.D). CA+ puncta formed pericellular baskets around unlabelled circular somatic profiles in layers 2/3 and around unlabelled pyramidal-shaped somata in layers 5/6. The somatic sizes of these unlabelled cell populations were significantly different. Immunolabelled puncta were also found in close contact with CR+ somata. Cortical depth distribution histograms and laminar thickness measurements defined the proportions of the overall CR- cell population in each layer: 7% in layer 1, 78% in layers 2/3, 14% in layers 5/6, and 1% in the white matter. In the tangential plane, CR+ neurons were distributed uniformly at all levels of the cortex. By using stereological counting procedures on immunoreacted Nissl-stained sections, CR+ neurons were estimated to constitute a mean 8.0% (7.2-8.7%) of the total neuron population in each cortical area. These data are compared with similar information obtained for the mPFC in monkey and rat (Gabbott and Bacon [1996b] J. Comp. Neurol. 364:657-608; Gabbott et al., [1997] J. Comp. Neurol. 377:465-499). This study provides important morphological insights into a neurochemically distinct subclass of local-circuit inhibitory neurons in the human mPFC.