The main use of an uncoded corpus is searching for a particular word or sequence of words. Concordancing software is designed especially to allow searches of this type, for example, to examine frequencies of words or collocations, or to find examples of certain words or structures. However, many linguistic investigations – including most of the analyses in this book – are not possible if we are restricted to simply searching for words. Even structures that conform to fairly regular morphological or syntactic patterns are not easy to study based on word searches.

Suppose, for example, that you wanted to investigate the use of passive voice. With an uncoded corpus, you might start by searching for any form of be plus a word ending in -en. This search pattern would find many passives – such as was eaten and is taken – but it would miss the passives ending in -ed (e.g., been carried, was kicked). Even if you expanded your search to include these passives, you would still miss all the irregular passives, of which there are many – e.g., shown, torn, built, kept, sold, meant, brought. Each ends with a different sequence of letters, making patterned searches impossible. In addition, you would miss instances that have intervening adverbs (e.g., was completely eaten). Conversely, this search pattern would fit structures that are not passives, such as was green, is red.