GPI-linked hemagglutinin (GPI-HA) of influenza virus was thought to induce hemifusion without pore formation. Cells expressing either HA or GPI-HA were bound to red blood cells, and their fusion was compared by patch-clamp capacitance measurements and fluorescence microscopy. It is now shown that under more optimal fusion conditions than have been used previously, GPI-HA is also able to induce fusion pore formation before lipid dye spread, although with fewer pores formed than those induced by HA. The GPI-HA pores did not enlarge substantially, as determined by the inability of a small aqueous dye to pass through them. The presence of 1,1'-dioctadecyl-3, 3,3',3'-tetramethylindocarbocyanine perchlorate or octadecylrhodamine B in red blood cells significantly increased the probability of pore formation by GPI-HA; the dyes affected pore formation to a much lesser degree for HA. This greater sensitivity of pore formation to lipid composition suggests that lipids are a more abundant component of a GPI-HA fusion pore than of an HA pore. The finding that GPI-HA can induce pores indicates that the ectodomain of HA is responsible for all steps up to the initial membrane merger and that the transmembrane domain, although not absolutely required, ensures reliable pore formation and is essential for pore growth. GPI-HA is the minimal unit identified to date that supports fusion to the point of pore formation.