This research emerged from an effort to develop a theory of computer anxiety relevant to beginning adult learners in a range of university disciplines. To this end, the first step was to design and refine an instrument which was reliable and valid for this population. The development of the "a priori" model of computer anxiety used in this study followed this procedure: a pilot study with first year teacher trainees (N=101) using CARS (Computer Anxiety Rating Scale); an inspection of items in other anxiety instruments from previous research; a literature search in anxiety theory, leading to substantitive hypotheses of factors likely to contribute to computer anxiety; and generation of factors (constructs) and items considered to have face validity for these factors. The instrument consisted of 111 items divided into four parts or subscales: (A) Gaining Initial Computing Skills; (B) Sense of Control; (C) Computer Self-Concept; and (D) State Anxiety in Computing Situations. The validity of the model was tested on a sample of 794 undergraduate students from four disciplines of a university in the western region of metropolitan Sydney, Australia. Evidence suggested that there may be 10 factors underlying anxiety towards computers for beginning users. However, a more parsimonious model was identified using CFA (confirmatory factor analysis) analyses of the data. In particular, the results supported the existence of one second (higher) order factor for each of Parts A and D. As for Parts B and C, there was strong support for one substantive factor in each case, with negatively worded items forming a separate "method factor." A summary is provided of first- and second-order factors for each section of the instrument. Four tables illustrate data. (Contains 15 references.) (MAS)