The effect of the DNA sequence of a template on the progress of DNA polymerase alpha was determined at the resolution of single nucleotides by using single-stranded, circular phi X174 DNA as a template and unique phi X 174 ONA fragments terminally labeled at their 3'-ends as primers. Therefore, the amount of radioactivity observed was always proportional to the number and not the length of nascent DNA chains. Extension of a primer by alpha-polymerase revealed that 3'-ends of nascent DNA chains accumulated at specific arrest sites consisting of GC-rich sequences of 1-8 bases distributed nonuniformly along the template with intervening sequences of 0-140 bases. The precise location and composition of these sites were determined by DNA sequencing methods, but a consensus sequence was not detected. However, the same pattern of arrest sites recognized by DNA polymerase alpha from CV-1 cells also arrested alpha-polymerase from HeLa cells, calf thymus tissue, and phage T4 DNA polymerase. An exception was DNA polymerase I from Escherichia coli which was insensitive to arrest signals. Analysis of recombinant DNA templates containing phi X174 DNA inserted into M13 DNA revealed that arrest sites were defined exclusively by those sequences within 24 bases upstream and 140 bases downstream of the arrest point; long range interactions between template sequences were not involved. One of the strongest arrest sites was located in the first two nucleotides at the base of the stem on the primer-proximal side of a stable hairpin structure, suggesting that such structures arrest the polymerase. However, only 28% of all arrest site nucleotides were found in this position; the rest were up to 25 bases upstream from any computer-predicted hairpin. Therefore, all arrest sites cannot be defined by secondary structure alone, although proximity to secondary structures may amplify normal variations in the rate of DNA elongation caused by primary sequences.