Spinach chloroplast DNA polymerase was shown to copurify with a 3' to 5' exonuclease activity during DEAE-cellulose, hydroxylapatite, and heparin-agarose column chromatography. In addition, both activities comigrated during nondenaturing polyacrylamide gel electrophoresis and cosedimented through a glycerol gradient with an apparent molecular weight of 105,000. However, two forms of exonuclease activity were detected following velocity sedimentation analysis. Form I constituted approximately 35% of the exonuclease activity and was associated with the DNA polymerase, whereas the remaining activity (form II) was free of DNA polymerase and exhibited a molecular weight of approximately 26,500. Resedimentation of form I exonuclease generated both DNA polymerase associated and DNA polymerase unassociated forms of the exonuclease, suggesting that polymerase/exonuclease dissociation occurred. The exonuclease activity (form I) was somewhat resistant to inhibition by N-ethylmaleimide, whereas the DNA polymerase activity was extremely sensitive. Using in situ detection following SDS-polyacrylamide activity gel electrophoresis, both form I and II exonucleases were shown to reside in a similar, if not identical, polypeptide of approximately 20,000 molecular weight. Both form I and II exonucleases were equally inhibited by NaCl and required 7.5 mM MgCl2 for optimal activity. The 3' to 5' exonuclease excised deoxyribonucleoside 5'-monophosphates from both 3'-terminally matched and 3'-terminally mismatched primer termini. In general, the exonuclease preferred to hydrolyze mismatched 3'-terminal nucleotides as determined from the Vmax/Km ratios for all 16 possible combinations of matched and mismatched terminal base pairs. These results suggest that the 3' to 5' exonuclease may be involved in proofreading errors made by chloroplast DNA polymerase.