This project was concerned-with teaching algebra novices, all girls aged 13 or 14 years, to solve algebra word problems using an electronic spreadsheet. It was based on the realisation that a spreadsheet cell provides a suitable cognitive model for an algebraic variable and that the manipulation of a spreadsheet is essentially based on the construction of algebraic expressions. The main objectives were to test the effectiveness of spreadsheet use on the ability to construct algebraic expressions and to examine the effect of manipulating problem contexts (abstract vs. concrete) on this ability. Other objectives were to determine the relationship between general numerical ability, attitude to mathematics, attitude to computers and the experimental treatments. The particular skill taught was the construction of algebraic expressions to represent relational propositions from verbally stated problems. Problems from current textbooks and examination papers (Intermediate Certificate Syllabus B) were used in the instruction. A pretest - posttest control group design was used. Seventy three volunteers were recruited and received approximately eight hours of Instruction in a reasonably natural school setting. There were two treatment groups. One group worked on abstract (numerical) problems and the other group worked on mathematically identical problems set in concrete contexts which were familiar and relevant. Both treatment groups made considerable gains between pretest and posttest. The abstract group performed significantly better than the concrete group on the total posttest (p < .01), on its abstract subsection (p < .01) and on its concrete subsection (p < .05). Attitude to mathematics was also found to have a significant Interaction with the treatment (p <. 05). Those with a positive attitude to mathematics learned more from abstract problems, but the difference was much less for those with a negative attitude. Neither numerical ability or attitude to computers had any significant effect.