We estimated American black bear (Ursus americanus) abundance across the Upper Peninsula of Michigan (41,984 km²) during 8 years between 1990–2004 using tetracycline-laden baits and hunter harvest in mark–recapture analyses with the Lincoln–Petersen estimator. We marked 179–251 bears during each of 8 summer marking sessions using a mean bait density of 1 bait/69 km². We examined teeth collected annually during bear registration for tetracycline marks in the year of marking and through 6 years post marking. From harvest samples, we recovered 6–25% of bears marked in the year of marking. Annual proportion of harvested bears with tetracycline marks from a given marking session through 6 years post marking ranged from 0.0–5.6%. Males with tetracycline marks were more prevalent in the harvest as were bears 3–4-years old. Placement of baits in aspen (Populus spp.) vegetation type and presence of recent bear activity increased bait consumption and consequently the number of bears marked. Simulation modeling suggested that observed rates of bears ingesting multiple baits, number of bears harvested, and proportion of teeth examined did not affect population estimates, whereas tetracycline detection rates, ingestion of baits by cubs, and marks by external tetracycline sources did. Additionally, legal restrictions against harvest of females with cubs resulted in a negative population bias. Bear population estimates generally increased from 1990–2000, then apparently declined in 2002 before increasing again in 2004, possibly influenced by natural food availability. Incorporating recoveries over multiple hunting seasons generally resulted in increasing population estimates with decreasing confidence intervals. We recommend using ≥2 years of recapture data, including the year of marking, to estimate black bear population size. Population estimates of bears in Michigan's Upper Peninsula using tetracycline can be improved by increasing bait density and bear ingestion rates of baits, and decreasing ingestion rates by cubs. Development of a correction factor for tetracycline marks from external sources would also increase precision of estimates.