This paper presents a classification system for Arabic documents that have multi-type contents. The good classification results for the document contents achieved demonstrate the importance of the research work conducted. The Statistical Classifier for Arabic Documents (SCAD) classifies the document content type to one of the types: text, photographs, geometrics, medium-size font text, and large-size font text. To achieve these objectives (SCAD) is designed and implemented as a group of filters using a statistical concept. SCAD is first trained on samples of document contents of different types. The results obtained from SCAD application on other document samples depend mainly on the filter sizes used.