Odor identification (OID) is the most common test of human olfaction; test trials typically involve matching a target odor to a correct label among three distractors. However, the underlying processes are poorly understood. Here, we investigated the OID response patterns in a large population-based sample (n=2479) of cognitively intact older adults (aged 60-100 years) from the The Swedish National Study on Aging and Care in Kungsholmen (SNAC-K) where olfaction was assessed by a ‘Sniffin´ Sticks’ OID test with 16 odors. First, we analyzed the pattern of mistakes, and the results show that some distractors are much more commonly selected than others. This suggests that cognitive factors may be present. Second, we conducted an online survey of 1000 older adults, aged 60-90 years, who were asked to imagine and rate the olfactory similarity of the target odors and the three corresponding distractors (e.g. “How similar do apple and mint smell?”). Results show that the imagined similarity of the target-distractor pair could partly predict how often the distractor was mistakenly selected in the SNAC-K experiment. Third, we used data from the Swedish web corpus and Word2Vec neural network algorithm to quantify the semantic association strength between labels of each target odor and its distractors. The semantic distances between target-distractor pairs were found to partly explain the pattern of incorrect responses in the SNAC-K OID test. In sum, our results suggest that OID tests are not just perceptual olfactory assessments, but reflect olfactory cognition and language. Behavioral outcomes are likely influenced by imagining the olfactory qualities corresponding to each odor label and by selecting strongly semantically associated labels. These insights may be harnessed to develop OID tests that are tailored for clinical purposes, for example in predicting the onset of neurological disorders. This research was supported by the Swedish e-science Research Center to PH and JKO.