What is the variable rule program?

When conducting a quantitative analysis regarding the variation of a certain variable, there are a number of things one must consider. Why does a feature change, when does it change, and what causes it to change etc.? Only when the analyst has an informed understanding of the various contextual factors that favor one form over another can he or she achieve scientifically well-grounded results. However, in regards to determining which factors have an impact on one's choice process, and ascertaining their respective influential strengths, each finding must be backed up with statistical significance. Obviously this calls for a series of calculations based on various, extremely complex mathematical processes, which would exceed the average human's mathematical competence. This is where the variable rule program comes in handy. It is considered “the key statistical tool" (Tagliamonte, Chapter 20: 388) in variationist research , as it provides the variationist with access to detailed data information, facilitating the otherwise painstaking process of extracting, coding and interpreting the often vast sets of data. It is basically a logistic regression package conceived for predicting the probability of the occurrence of a feature based on the presence of certain factors, by fitting gathered data-samples to a logit curve and modeling the effects these multiple factors have (cf. Tagliamonte 2006: 128ff.).

The beginnings of the method of variable rule analysis can be traced back to William Labov, who, in the 1960s, was the first to discover the underlying, systematic choice process speakers go through while formulating an utterance. The method's mathematical implementation, however, was fostered by Henrietta Cedergren and David Sankoff in 1974, who also provided the program's earliest description. Since then, the variable rule program has been repeatedly improved and, as already mentioned, is now considered “one of the most appropriate methods available for conducting statistical analysis on natural speech” (Tagliamonte 2006: 129) by many variationists.

This chapter introduces the newest version of the variable rule program, Goldvarb X, which is an updated version of its predecessors Goldvarb 2.0 and Goldvarb 2001. It has been re-written in C++ and runs in OS X and has been upgraded to operate in Macintosh, Windows or Linux. Also, it comes equipped with several user interface features, which are especially practical for cross-reference searches and other complex applications.

Goldvarb X can be downloaded free of charge from http://individual.utoronto.ca/tagliamonte/goldvarb.htm. This site also provides a manual, which was originally written for Goldvarb 2.0, but works for Goldvarb X likewise. The manual, however, is not intended as a “technical instruction to the statistical workings of the program nor as a guide to the use of multivariate analysis”, which is why further literature is recommended for an in depth understanding of the variable rule program. For a list of more references with standard descriptions of the variable rule program go to Standard descriptions of the variable rule program.

The online-manual leads through a number of steps, one of which is entitled § Getting started. In this step, you will be asked to upload a tokenfile into the program. According to the manual, the program comes with an imaginary sample tokenfile entitled “Nepean.Tok”, which provides information on plural variation in informal speech amongst adolescent gang members in the area of Nepean, which is part of the regional municipality of Ottawa-Carleton. However, if this is not the case go to http://nlp.stanford.edu/~manning/courses/ling236/handouts/Nepean-mac.tok or http://nlp.stanford.edu/~manning/courses/ling236/handouts/Nepean-win.tkn where you will find copies of the sample Nepean files.

Created with the Personal Edition of HelpNDoc: Full-featured Documentation generator