Getting off the GoldVarb Standard: Introducing Rbrul for Mixed‐Effects Variable Rule Analysis

Abstract

The variable rule program is one of the predominant data analysis tools used in sociolinguistics, employed successfully for over three decades to quantitatively assess the influence of multiple factors on linguistic variables. However, its most popular current version, GoldVarb, lacks flexibility and also isolates its users from the wider community of quantitative linguists. A new version of the variable rule program, Rbrul, attempts to resolve these concerns, and with mixed‐effects modelling also addresses a more serious problem whereby GoldVarb overestimates the significance of effects. Rbrul's superior performance is demonstrated on both simulated and real data sets.

The status of s in Dominican Spanish

“…The decision to enter the individual study participants into the model as a random term was motivated by our observation that certain individuals appear to us to produce more intrusives than do others. By fitting the data to a mixed effects rather than a fixed effects model, we are more confident that the significant trends that we observe are predictive and do not just reflect the behavior of one or two participants in our sample (Johnson, 2009;Drager and Hay, 2012).…”

“…Below, we present the descriptive statistics on the overall rates of retention, intrusion, and deletion in the data set that includes all the observations (section 6). Next, we show the results of the mixed-effect logistic regression (Johnson, 2009), which analyzes the contribution of gender, literacy, speaker, and the linguistic internal factors to the probability that lexical-s will be retained rather than deleted. In a second analysis, in section 6.2, we describe the distribution of intrusive-s and discuss the correlation between an individual's use of lexical-s versus his or her use of intrusive-s.…”

“…For the analysis of s-retention, we removed the observations of intrusive-s (n = 82). Mixed effects logistic regression models were fitted using Rbrul (Johnson, 2009) in the statistical package R (R Development Core Team, 2008) to examine the factors that are most predictive of the retention of historical-s versus its deletion. For the analysis of historical-s retention, the factors of position (with three levels) and following segment type (with seven levels) were combined into a single 'super' factor with twelve levels.…”

“…In addition to the linguistic-internal and external predictors set out above, each participant was entered as a random intercept in the statistical models as was each unique word token (see Johnson, 2009;Tagliamonte and Baayen, 2012). This allowed us to test whether only one or two of the participants were responsible for the tokens of overt s expression or whether there were trends among our groups once the effect of the individual was accounted for.…”

Sobre puntos de la gramática particularmente resistentes al cambio: cuatro (largos) siglos en la evolución de las relativas oblicuas de lugar. Un análisis variacionista

“…Para ello utilizamos el programa Rbrul (Johnson 2009), con el que obtenemos una relación de factores seleccionados como significativos, cuando todos se analizan al mismo tiempo, así como las potenciales interacciones entre sí. Además, Rbrul permite agrupar los casos según factores individuales aleatorios, como, en este caso, el escritor o los antecedentes con que se relacionan los relativos.…”