A decision-makin g model is presented in which (a) the apparent weight of evidence, to a juror deciding on a case, varies randomly over the population of potential jurors; (b) jurors decide (against) the plaintiff or defendant if apparent weight is less (greater) than a decision criterion, which varies little over the population but may vary over cases, instructions to jurors, etc.; and (c) decision confidence increases as absolute difference between apparent weight and decision criterion increases, this increase being taken as linear most of our presentation. We assume that the distribution of apparent weight is based on the exponential distribution and obtain estimates of mean apparent weight and decision criterion from four civil cases and two criminal cases. The effects of jury deliberation, jury size, the definition of reasonable doubt, the introduction as evidence of a previous record of conviction, and instructions to disregard such evidence are described in terms of the parameter estimates. Finally, the theoretical relationship between decision frequency and confidence is discussed, and a measure is given of the difference between the subpopulations of jurors voting for and against. It is shown that this difference measure equals the area under a confidence-rating operating characteristic curve.