Software inspections are an efficient method to detect faults in software artefacts. In order to estimate the fault content remaining after inspections, a method called capture–recapture has been introduced. Most research published in fault content estimations for software inspections has focused upon point estimations. However, confidence intervals provide more information of the estimation results and are thus preferable. This paper replicates a capture–recapture study and investigates confidence intervals for capture–recapture estimators using data sets from two recently conducted software inspection experiments. Furthermore, a discussion of practical application of capture–recapture with confidence intervals is provided. In capture–recapture, used for software inspection, most research papers have reported Mh-JK to be the best estimator, but only one study has investigated its subestimators. In addition, confidence intervals based on the log-normal distribution have not been evaluated before with software inspection data. These two investigations together with a discussion provide the main contribution of this paper. The result confirms the conclusions of the replicated study and shows when confidence intervals for capture–recapture estimators can be trusted.