The research examined the problem-solving performance of 16 sixthgrade students from a depressed area. The students were first taught the mechanics of how to use a computer-based teletype to solve arithmetic word problems. Following the initial instruction set, a series of 100 word was presented to the students. The solutions of these were analyzed to determine the variables related to problem difficulty. A linear regression analysis showed that a word problem is difficult to solve if (a) it is of a different type from the problem that preceded it, (6) its solution requires a large number of different operations, (c) its surface structure is complex, (d) it has a large number of words, or (e) it requires a conversion of units. There exists a great diversity of approaches to the investigation of human problem solving, and a wide range of materials, techniques, and problems has been used for such study. Subjects have been required to solve anagrams, matchstick problems, water-jar problems, pendulum problems, concept-identification problems, analogy problems, number-series problems, or arithmetical word problems, to name but a few. Although several theoretical formulations have been offered and many facts have been discovered, there is still no single adequate theory into which they can be integrated. In addition, there is little analysis of why arithmetic word problems, specifically, are difficult for students. We know that students have difficulty in solving word problems. The present study was an attempt to find out why. It is an attempt to explore the notion that in solving a set of word problems, certain items are more difficult to solve than others, and to understand what structural variables cause some word to be hard and others to be easy.