Abstract The purpose of the present study was to establish culture conditions for human uterine cervical epithelial cells on permeable support and to determine how it affects cervical cell differentiation. Human ectocervical epithelial cells (hECE), HPV-16 immortalized hECE cells (ECE16-1) and Caski cells were grown on collagen-coated filters. Culture conditions, density of cells in culture and expression of epithelial and cervical-cell phenotypic markers were determined and compared in cells grown on filter and on solid support. Compared with the latter, cultures on filter had a higher cell density, hECE cells stratified to 5–12 cell layers compared to 1–3 on solid support, and cells of all three types expressed intercellular tight junctions. The cytokeratin profiles revealed differences between the three cell types as well as differences within the same cell species when grown on filter, compared to solid support. Of particular importance was the finding of a higher expression of K-13 in hECE grown on filter compared to solid support; K-13 is a marker of ectocervical cell differentiation. The cytokeratin profiles of the cultured hECE, ECE16-1 and Caski cells resembled those of ectocervical, squamous metaplastic and endocervical epithelia, respectively. hECE and ECE16-1 expressed involucrin protein, the level of which in both was higher in cells grown on filter compared to solid support. Polarization of the cultures was determined by morphology (stratification of hECE cells, expression of pseudomicrovilli in the apical cell membrane), selective apical vs. basolateral secretion of [ 35S]methionine- and [ 35S]cys-teine-, [ 3H]fucose- and [ 14C]glucosamine-labeled mole-cules, and positive short-circuit current (I sc) under voltage-clamp conditions. Confluency of the cultures was determined by measuring transepithelial unidirectional fluxes of inert molecules with different molecular weights (MWs) through the paracellular pathway, and by measuring transepithelial conductance. The results indicated transepithelial permeability of 7–22 · 10 −6 cm · sec −1, which was 5–100 fold smaller compared to blank inserts, with a cut-off MW of 40–70 kDa for hECE and Caski cells. Transepithelial conductance ranged 18.5 to 51.5 mS · cm −2, indicating a leaky but confluent epithelia. Collectively the results indicate the epithelial nature of the cells and their improved differentiation when grown on filter support; hECE is a model for ectocervical epithelium while ECE16-1 and Caski express phenotypic characteristics of squamous metaplastic cervical epithelium and endocervical epithelium, respectively.