It is generally assumed that fusogenic peptides would require a certain conformation, which triggers or participates in the rate-determining step of membrane fusion. Previous structure analyses of the viral fusion peptide from gp41 of HIV-1 have yielded contradictory results, showing either an alpha-helical or a beta-stranded conformation under different conditions. To find out whether either of these conformations is relevant in the actual fusion process, we have placed sterically demanding substitutions into the fusion peptide FP23 to prevent or partially inhibit folding and self-assembly. A single substitution of either D- or L-CF(3)-phenylglycine was introduced in different positions of the sequence, and the capability of these peptide analogues to fuse large unilamellar vesicles was monitored by lipid mixing and dynamic light scattering. If fusion proceeds via a beta-stranded oligomer, then the D- and L-epimers are expected to differ systematically in their activity, since the D-epimers should be unable to form beta-structures due to sterical hindrance. If an alpha-helical conformation is relevant for fusion, then the D-epimers would be slightly disfavoured compared to the L-forms, hence a small systematic difference in fusion activity should be observed. Interestingly, we find that (1) all D- and L-epimers are fusogenically active, though to different extents compared to the wild type, and--most importantly--(ii) there is no systematic preference for either the D- or L-forms. We therefore suggest that a well-structured alpha-helical peptide conformation or a beta-stranded oligomeric assembly can be excluded as the rate-determining state. Instead, fusion appears to involve conformationally disordered peptides with a pronounced structural plasticity.