Phonological Correspondence as a Tool for Historical Analysis: An Algorithm for Word Alignment

Introduction In historical linguistics research it is often necessary to compare words from di#erent languages or di#erent stages of a single language in order to reconstruct sound changes, or simply in order to determine whether the words are cognate. One stage of this process is aligning the words with one another. For example, Spanish vosotros ([ ############### ]) and French vous ([vu]) should align as follows: v u as opposed to, for example: v u The algorithm to be discussed here attempts to automate this process and find a best phonological correspondence between words. 2 The Program The underpinnings of this algorithm were originally written as part of a word-comparison algorithm that would ignore various types of phonological changes in order to split words up into morphemes (Albro 1997). Covington's (1996) article led to the realization that the algorithm might be used as an improvement (or at least a di#erent take) on his method of aligning words. In its origina