Four researchers observed arraignment proceedings and collected data for 1,000 cases in Manhattan criminal court over the 6-month period. Data were collected only for cases that were continued past arraignment, since no release decision was involved in cases disposed at arraignment. The release-on-recognizance (ROR) decisions and the amount of bail set were analyzed separately in order to determine whether different factors were involved in theses two aspects of the arraignment decision. In addition to identifying the factors that affected the ROR decisions and bail amounts, this report also presents data and information on case and defendant characteristics and attorney requests and arguments. The major finding was that the assistant district attorney's (ADA's) bail request was by far the strongest predictor of both the judge's ROR decision and the amount of bail set. The bail request was a better predictor than charge severity alone and remained the strongest predictor for both ROR and bail amount even in the absence of much variation in charge severity. The defendant's criminal record, particularly a history of failure-to-appear in prior cases, was significantly more important for the release decision than for the bail amount. Overall, judicial variability explained only a small part of the variation in ROR for both misdemeanor and felony cases. It explained much more of the variation in bail amounts. 23 tables, 2 figures, 55 references, and appended supplementary methodological information and exhibits