Historically a gap has existed between the study of the one-dimensional organization of hereditary information in genes, and of the three-dimensional organization of macromolecules in biological structures. In this article we describe progress in closing this gap through the genetic and biochemical analysis of the assembly of the icosahedral shells of spherical viruses, a class of subcellular structures whose subunit organization is relatively well understood. The genes specifying the proteins required for capsid assembly have been identified for many bacterial viruses. By using mutants defective in these genes, it has been possible to identify intermediates in shell morphogenesis and DNA condensation, and to unravel the different levels of the genetic control of macromolecular assembly processes. In general, a precursor shell or procapsid is first constructed, and the DNA is subsequently coiled within it. The construction of a closed shell poses as difficult a problem for a virus as for an architect. In the well-studied bacteriophage P22 of Salmonella typhimurium, the construction of the procapsid requires the interaction of about 200 molecules of the gene-8 scaffolding protein with 420 molecules of the gene-5 coat protein, forming a double-shelled structure with the scaffolding protein on the inside. Once completed, procapsids undergo substantial alteration in the course of encapsulating the viral DNA. In P22, the initiation of DNA packaging triggers the exit of all of the scaffolding molecules from within the capsid, probably through the coat-protein lattice. These released molecules are re-utilized, interacting with newly synthesized coat subunits to form further procapsids. Thus, the scaffolding protein functions catalytically in capsid assembly. All of the well-studied DNA phages require a scaffolding protein species for procapsid assembly, though their properties vary. Purified coat and scaffolding subunits by themselves show little tendency to polymerize, and are stable as monomers in solution. Upon mixing together under the appropriate conditions, however, the proteins copolymerize into double shells. Their interaction with each other appears to be critical for efficient assembly; this interaction probably occurs on the edges of growing shells, and not among subunits in solution. We have termed this kind of process, which we previously described in T4 tail morphogenesis, self-regulated assembly. The subunits are synthesized in a nonreactive form and are activated, not in solution, but upon incorporation into the growing substrate structure. A number of further transformations of the capsid subunits occur only within the organized structure and not as free subunits. Thus, aspects of the genetic information controlling the assembly process are not fully expressed at the level of the properties of protein subunits, but become manifest only through interactions with other proteins, or at a higher level, after completion of the correct organized structure.