Divisive clustering is a top-down form of hierarchical clustering that begins with all data points in one cluster and recursively splits them into smaller clusters. It contrasts with agglomerative clustering, which builds clusters bottom-up by merging smaller groups.