Infectious HPIV3 was produced by the intracellular coexpression of four plasmid-borne cDNAs. These separately encoded a complete HPIV3 genome (negative-sense), the HPIV3 nucleocapsid protein N, the phosphoprotein P, and the polymerase protein L. The cDNA-encoded HPIV3 genome differed from the JS wildtype (wt) strain of HPIV3 used in its construction by seven point mutations: four of these are silent mutations in the HN or L gene coding regions that serve as markers of a cDNA-derived virus, two were introduced to create an amino acid substitution that ablates an epitope recognized by the HN-specific monoclonal neutralizing antibody 423/6, and the remaining point mutation results in an incidental amino acid substitution in the HN protein at amino acid position 263. The four plasmids were transfected into HEp-2 cell monolayers and their expression was driven by T7 RNA polymerase supplied by a vaccinia virus recombinant. The titer of virus present in the harvested transfection supernatant was low (<5 PFU/ml), and the recovered recombinant virus (rJS) retained each of the seven mutations present in the cDNA from which it was derived. Despite the introduced and incidental mutations, rJS retained the wt phenotypes as regards replication at elevated temperaturein vitroand efficient replication in the upper and lower respiratory tract of hamsters. rJS was also recovered from a cDNA encoding a complete antigenome (positive-sense) with slightly greater efficiency than from the negative-sense construct. The ability to generate infectious HPIV3 from cDNA should greatly enhance our ability to develop new live-attenuated parainfluenza virus vaccines, including chimeric PIV1 and PIV2 vaccines, and to understand the genetic basis of attenuation of PIV3 candidate vaccines.