In the last thirty years, research on differential sentencing practices became one of the dominant thrusts of academic interest in criminal justice studies. This was mainly because several reforms had been added to the various sentencing structures in the United States, one of which was the adoption of guideline based sentencing strategies. Although several studies were conducted regarding the impact of these guidelines, these were provided by a limited number of sources. For instance, the majority of this research was conducted on data collected by the United States Sentencing Commission and the Pennsylvania Sentencing Commission. This study attempted to further the discussion on sentencing practices by examining data from a new source, the Arkansas Office of Courts. This study first examined the judge's decision to imprison and jail the defendant using logistic regression, and second, for those individuals incarcerated, OLS and negative binomial regression analyses were conducted to explore potential disparities in the length of prison and jail sentences. The results of this analysis in Arkansas illustrated striking comparisons to studies that were conducted on other guideline structures. The legally relevant variables were the greatest predictors of whether or not an offender was sentenced to prison. Extralegal variables, on the other hand, had negligible effects. Implications for policy are also discussed. © 2006 Elsevier Ltd. All rights reserved.