Determination of `no observed adverse effect level' (NOAEL) is one of the prime objectives of a repeated subchronic experiment. The data collected in such experiments are huge, and it is a difficult and time-consuming task to determine NOAEL. In such situations, cluster analysis could be used as a valuable tool. In the present paper the determination of NOAEL is demonstrated using the data of a 28-day repeated oral toxicity study carried out in rats. Groups (10/sex/group) of Crj: CD rats were administered the test substance at low, middle, high and top dose levels by gastric intubation daily for 28 days. A concurrent control group was also maintained. In-life measurements included general behavior, body weight, food and water consumption. At termination various hematological and biochemical parameters were determined in the blood of individual animals in all groups. Urinalysis was also carried out at termination. Animals were sacrificed for microscopic and macroscopic findings. Analysis of data showed that 40 measurements (data 1) in the treatment groups were different from the control and 48 (data 2) were not treatment-related. Cluster analysis was carried out separately on data 1 and combined data 1 and data 2. This study revealed that for better judgment of NOAEL performing a cluster analysis on the data which alone showed a significant difference compared to control is more advisable than performing the cluster analysis on all data collected in the study. The advantage of cluster analysis is that quantitative data and qualitative data can be analyzed in annexation. In addition, cluster analysis assists in the result of the univariate analysis and displays NOAEL at once quite obviously.