Does size beat everything? Why philology still matters in corpus-based linguistics. Some thoughts based on older varieties of Spanish.

This paper asks whether the sheer size of a dataset can compensate for a lack of philological scrutiny in corpus-based linguistics. Drawing on a case study of Old Spanish de-nominals – bare partitives and putative partitive articles – it compares the results obtained from philologically curated data with those from larger, uncurated datasets retrieved from CORDE and the Corpus del Español. Quantitatively, the absence of philological control proves relatively harmless; qualitatively, however, it is decisive: the uncurated data appear to yield more than a dozen partitive articles, yet every one dissolves, upon collation with its source, into a corpus error. Enlarging the data in this way thus moves the results of the analysis further from those obtained on the basis of the philologically curated data, rather than closer to them. Consequently, the paper defends that no statistical threshold can substitute for case-by-case philological inspection. Moreover, it argues that the case for philological rigour is ultimately not only methodological but also ethical – a matter addressed here, through a sympathetic appropriation of Kabatek (2011), under the notion of linguistic dignity.