The Oxford English Dictionary is a valuable resource for diachronic study; however, it faces several challenges that can be traced back to its non-digital dictionary origins, namely the diasystematic labelling of frequency information. Lexicographic indications of historical frequency of use, register and obsolescence are well known to be fallible, but we believe this can be addressed via in-depth corpus-driven study. Given that obsolescence and low frequency remain under-researched areas of usage-based linguistics, this chapter aims to contribute to redressing this by providing a detailed account of low frequency words in the oed. We focus on a small sample of relatively infrequent -some adjectives listed in the oed. This chapter aims to test a corpus-driven methodology using Google Books Standard and Google Books Advanced (Davies 2014) for collecting more accurate information regarding historical obsolescence and frequencies of usage. Although the paradox of corpus analysis is that absence of usage does not equate with proof of obsolescence, we seek to test how historical corpora can be used to help to address the improvement (and systematising) of frequency and obsolescence indicators. Results show that infrequency and obsolescence are scalar, reversable, heterogenous and dependent on contextual factors. This investigation suggests that frequency indicators need to be specifically related to text type, genre and period as well as competitor frequencies.