An early step in nuclear lamin polymerization is the longitudinal, head-to-tail association of lamin dimers through the highly conserved end segments of their rod domains. Lamin fragments lacking the carboxy-terminal tail domain (tailless lamins) form long, thin protofilaments more readily than full-length lamins do. By morphology alone it cannot be ascertained whether tailless lamin protofilaments also arise through head-to-tail association of dimers. Here, we studied by transmission electron microscopy which types of interaction are important for formation of protofilaments by a tailless lamin fragment derived from Drosophila lamin Dm(0). First, we measured the lengths of tailless lamin filaments shorter than 200 nm. Many particles were longer than one and shorter than two dimers. Second, we labeled filaments formed by tailless lamin with a Fab fragment derived from a monoclonal antibody binding to an epitope residing in the amino-terminal head domain. The pattern of Fab fragments bound to lamin filaments clearly showed that most tailless lamin dimers are not arranged in a head-to-tail fashion. Third, we mutated tailless lamin using two point mutations known to inhibit head-to-tail association of full-length lamin. Thus mutated tailless lamin still formed filaments. We interpret these data according to models involving lateral associations between dimers of cytoplasmic intermediate filament proteins, using the lengths of short tailless lamin filaments as constraints. The data also demonstrate that a segment of at most 41 amino acids of the lamin tail domain is sufficient to bring about the change of mainly lateral to mainly longitudinal assembly.