Empirical studies of spreadsheet programming have commonly shown high over all error rates but much less attention has been paid to reasons for these errors. One often mentioned cause for errors is the low conceptual level of spreadsheet systems. By changing the conceptual level of spreadsheet system, we wanted to study whether this will produce different type of errors compared to traditional spreadsheet systems. In this paper we present an empirical study comparing the traditional spreadsheet calculation paradigm with the structured spreadsheet calculation paradigm that utilizes goals, plans and spreadsheet data structures in computation. The results show that the error behavior of novice spreadsheet users is systematically different between paradigms. Introduction Traditional spreadsheet calculation systems provide a fast method to express data and to make calculations, provided that there is some natural way to lay all the data on one or more twodimensional sheets. Spreadsheet cal...

Programming knowledge can be characterized in the form of goals and plans that describe what must be achieved and how this is done. We have conducted interviews of spreadsheet users and analyzed spreadsheet applications qualitatively. The analysis resulted in a set of basic programming goals and plans describing spreadsheet programming knowledge. This paper introduces a model for spreadsheet programming knowledge description and uses it to present the results of our analysis.