Ancient DNA (aDNA) sequences, especially those of human origin, are notoriously difficult to analyze due to molecular damage and exogenous DNA contamination. Relatively few systematic studies have focused on this problem. Here we investigate the extent and origin of human DNA contamination in the most frequently used sources for aDNA studies, that is, bones and teeth from museum collections. To distinguish contaminant DNA from authentic DNA we extracted DNA from dog (Canis familiaris) specimens. We monitored the presence of a 148-bp human-specific and a 152-bp dog-specific mitochondrial DNA (mtDNA) fragment in DNA extracts as well as in negative controls. The total number of human and dog template molecules were quantified using real-time polymerase chain reaction (PCR), and the sequences were characterized by amplicon cloning and sequencing. Although standard precautions to avoid contamination were taken, we found that all samples from the 29 dog specimens contained human DNA, often at levels exceeding the amount of authentic ancient dog DNA. The level of contaminating human DNA was also significantly higher in the dog extracts than in the negative controls, and an experimental setup indicated that this was not caused by the carrier effect. This suggests that the contaminating human DNA mainly originated from the dog bones rather than from laboratory procedures. When cloned, fragments within a contaminated PCR product generally displayed several different sequences, although one haplotype was often found in majority. This leads us to believe that recognized criteria for authenticating aDNA cannot separate contamination from ancient human DNA the way they are presently used.