Linguistic corpora are increasingly being used to study many features of language variation. The use of corpora for this purpose raises an important methodological issue: can a single corpus be reliably used as the basis of studies examining many different language phenomena? This article describes a study of variation in the use of pseudotitles based on the various regional components of the International Corpus of English (ICE)—components containing comparable samples of spoken and written English totaling one million words and taken from the many regional varieties of English that have developed worldwide. The study demonstrates that although certain kinds of language variation can be validly studied in a corpus such as ICE, other kinds of variation require different kinds of corpora. The final section describes how to determine whether a given corpus is suitable for a particular linguistic analysis and how future corpora can be designed to best facilitate the study of language variation.