The trilaminar kinetochore directs the segregation of chromosomes in mitosis and meiosis. Despite its importance, the molecular architecture of this structure remains poorly understood [1]. The best known component of the kinetochore plates is CENP-C, a protein that is required for kinetochore assembly [2], but whose molecular role in kinetochore structure and function is unknown. Here we have raised for the first time monospecific antisera to CENP-A [3], a 17 kD centromere-specific histone variant that is 62% identical to the carboxy-terminal domain of histone H3 [4,5] and that resembles the yeast centromeric component CSE4 [6]. We have found by simultaneous immunofluorescence with centromere antigens of known ultrastructural location that CENP-A is concentrated in the region of the inner kinetochore plate at active centromeres. Because CENP-A was previously shown to co-purify with nucleosomes [7], our data suggest a specific nucleosomal substructure for the kinetochore. In human cells, these kinetochore-specific nucleosomes are enriched in alpha-satellite DNA [8]. However, the association of CENP-A with neocentromeres lacking detectable alpha-satellite DNA, and the lack of CENP-A association with alpha-satellite-rich inactive centromeres of dicentric chromosomes together suggest that CENP-A association with kinetochores is unlikely to be determined solely by DNA sequence recognition. We speculate that CENP-A binding could be a consequence of epigenetic tagging of mammalian centromeres.