Recognition of handwritten text has gained great importance due to the wide spread of the pen-based applications and interface devices. Such handwritten text can be captured as images and stored for later accesses. However, to facilitate easy access to such collection of handwritten text images, automatic indexing of the individual documents is required. An essential step in the process of indexing arabic handwritten image documents is "segmentation", which is the process of dividing the arabic image document into individual words. This paper presents a segmentation system for arabic handwritten image documents.The segmentation of arabic handwritten image documents into words is performed in two steps. First, the arabic written image document is divided into line images using a horizontal projection method to detect the baselines and borders of a line image. Second, each line image is ddivided into word images using a ercital projection method to detect the words' separations. Each word image is considered the smallest unit in a image document. Keywords: Word recognition, arabic handwriting, handwritten document segmentation, pen-based technology.