The figure given by log. likelihood must be in relation to the maximum amount proposed by the programme under Maximum possible likelihood. If these two figures are similar, then this is an initial guarantee that the theory fits and that the rule is actually in place.

The second test -c2-, seen earlier in the U&D, is used in many statistical analyses. The amount required to find out whether the results fit the theory is that following the letter p; the closer it is to 0, the more reliable the rule.

Finally, the scattergram of the data presents these results as a diagram. If the points of convergence between the theoretical model and the real data follow the line of the diagram, the confidence level of the analysis is very high.

It is clear from this brief introduction that, in order to obtain an analysis with a high confidence level using the Goldvarb programme, we need to present the Input of application of the initial rule and one of the tests guaranteeing the likelihood of the theoretical model constructed from the iteration of independent variables. Subsequently, we can move on to interpret the data, which is always related to the figure, 0.5. Thus, we can determine when a factor favours the presence of the initial rule (in this case, maintenance of the solution [a]) and when it does not facilitate application of the initial rule, as seen with factors o (of the variable 1) and 4, 5 and 6 (of the variable 3).

Finally, it is important to bear in mind that, although probabilistic analysis enables us to observe future tendencies, it does not provide an exhaustive analysis of all possible combinations of factors. It is therefore often interesting to complete the analysis by cross tabulating factors to detail the connections of the percentages obtained on each independent factor according to the different crossovers of variables. One of the applications of the above programme allows the creation of cross-tables of factors using the cross tabulation instruction, once the condition, cell and results files have been created. Thus, Goldvarb permits analysis of different features using the same initial classification of data.

5. Goldvarb in variationist sociolinguistics

Variationism has taken significant steps towards adapting a method capable of providing rigorous analyses and reliable linguistic interpretations, and the Goldvarb programme is one of the results.

As we have seen, this programme enables us to obtain descriptive statistics and to find out the inference statistics of variable phenomena very accurately. However, we need to be aware beforehand of the factors susceptible to variation (which can arise from phonetic/phonological, morphological, lexical, semantic or syntactic elements of grammar) and, above all, we need to know how to find the linguistic and extralinguistic factors that can explain, in one way or another, the different variable phenomena. To reach these intuitions, we need to carry out prior qualitative analyses on the variation and nature of the diverse phenomena to be analysed.

While finding the variable and explanatory factors of variation is important, knowing how to combine these adequately in order to obtain the most representative results of the variable reality is no less relevant. This means that we need to look and, eventually, know how to find the best combination of explanatory factors which is very often reached by re-doing the analysis time and time again.

In short, the Goldvarb programme allows us to explain any process of variation, so long as the initial theory is good and is accompanied by a series of factors capable of bearing it out.

6. Bibliography

CARRERA-SABATÉ, J. (1999) L’Alternança a/e al Segrià. Doctoral thesis. University of Barcelona. Barcelona.

CARRERA-SABATÉ, J. (2001) La normativització del català modifica els hàbits fonètics dels parlants? Llengua i literatura, 12: 175-199.

CEDERGREN, H. J.; SANKOFF, D. (1974) Variable rules: Performance as a statistical reflection of competence. Language. 50: 333-355.

KAY, P.; McDANIEL, C. (1979) On the logic of variable rules. Language in Society. 8: 151-187.

LABOV, W. (1969) Contraction, Deletion, and Inherent Variability of the English Copula. Language. 45: 715-762.

LABOV, W. (1994) Principles of linguistic change. Internal factors. Blackwell. Cambridge.

LÓPEZ MORALES, H. (1989) Sociolingüística. Gredos. Madrid.

MORENO, F. (1994) Status quaestionis: sociolingüística, estadística e informática. Lingüística. 6: 95-154.

SANKOFF, G. (1988) Variable Rules. U. Ammon; N. Dittmar; K. J. Mattheier (eds.) Sociolinguistics. An international handbook of the science of language and society. Walter de Gruyter. Berlin & New York: 984-997.

Josefina Carrera-Sabaté
jcarrera@filcat.udl.es
carrera@lincat.ub.es

University of Lleida
University of Barcelona