Ternary complexes containing RNA polymerase, DNA and nascent RNA are intermediates in all RNA syntheses and are the targets of cellular factors that regulate RNA chain elongation and termination. Hence, elucidation of the structure and properties of these complexes is essential for understanding the catalytic and regulatory properties of the enzyme. We have described methods to prepare ternary complexes halted at defined positions along the DNA template, using specific dinucleotides to prime chain initiation along with limited subsets of the NTP substrates. Study of these static, halted complexes may provide information about the structure and properties of the transient elongation intermediates involved in transcription, although there is no necessary direct relationship between the two. Using specific halted complexes as precursors, we have walked the RNA polymerase along its template, producing defined ternary complexes at unique sites along two different transcription units. These complexes differ significantly from one another in many biochemical properties, in dramatic contrast to the properties expected from models that postulate a monotonous structure for elongation intermediates. These differences include variations in complex mobility during electrophoresis in non-denaturing polyacrylamide gels, in thermal stability and in stability to dissociation. Some halted complexes lose the ability to resume elongation when presented with the missing substrates. These "dead end" complexes must represent metastable structures in which elongation is blocked, and demonstrate clearly that not all halted complexes can be considered true intermediates in elongation. Other halted complexes rapidly cleave the nascent RNA seven nucleotides from the 3' terminus, in an unexpected and unusual biochemical reaction. These differences in properties among complexes bearing transcripts that differ by only one or a few nucleotides suggest that they have distinct structures. These differences must be due, at least in part, to differences in the template sequence and the length of the transcript. The results raise important questions as to the actual mechanism of transcription elongation, and suggest that it is a much more complex process than previously assumed.