Mutations in the fliK gene of Salmonella typhimurium commonly cause failure to terminate hook assembly and initiate filament assembly (polyhook phenotype). Polyhook mutants give rise to pseudorevertants which are still defective in hook termination but have recovered the ability to assemble filament (polyhook-filament phenotype). The polyhook mutations have been found to be either frameshift or nonsense, resulting in truncation of the C terminus of FliK. Intragenic suppressors of frameshift mutations were found to be ones that restored the original frame (and therefore the C-terminal sequence), but in most cases with substantial loss of natural sequence and sometimes the introduction of artificial sequence; in no cases did intragenic suppression occur when significant disruption remained within the C-terminal region. By use of a novel PCR protocol, in-frame deletions affecting the N-terminal and central regions of FliK were constructed and the resulting phenotypes were examined. Small deletions resulted in almost normal hook length control and almost wild-type swarming. Larger deletions resulted in loss of control of hook length and poor swarming. The largest deletions severely affected filament assembly as well as hook length control. Extragenic suppressors map to an unlinked gene, flhB, which encodes an integral membrane protein (T. Hirano, S. Yamaguchi, K. Oosawa, and S.-I. Aizawa, J. Bacteriol. 176:5439-5449, 1994; K. Kutsukake, T. Minamino, and T. Yokoseki, J. Bacteriol. 176:7625-7629, 1994). They were either point mutations in the C-terminal cytoplasmic region of FlhB or frameshift or nonsense mutations close to the C terminus. The processes of hook and filament assembly and the roles of FliK and FlhB in these processes are discussed in light of these and other available data. We suggest that FliK measures hook length and, at the appropriate point, sends a signal to FlhB to switch the substrate specificity of export from hook protein to late proteins such as flagellin.