A comprehensive survey of teacher stress, job satisfaction and career commitment among 710 full‐time primary school teachers was undertaken by Borg, Riding & Falzon (1991) in the Mediterranean islands of Malta and Gozo. A principal components analysis of a 20‐item sources of teacher stress inventory had suggested four distinct dimensions which were labelled: Pupil Misbehaviour, Time/Resource Difficulties, Professional Recognition Needs, and Poor Relationships, respectively. To check on the validity of the Borg et al. factor solution, the group of 710 teachers was randomly split into two separate samples. Exploratory factor analysis was carried out on the data from Sample 1 (N=335), while Sample 2 (N=375) provided the cross‐validational data for a LISREL confirmatory factor analysis. Results supported the proposed dimensionality of the sources of teacher stress (measurement model), along with evidence of an additional teacher stress factor (Workload). Consequently, structural modelling of the ‘causal relationships’ between the various latent variables and self‐reported stress was undertaken on the combined samples (N=710). Although both non‐recursive and recursive models incorporating Poor Colleague Relations as a mediating variable were tested for their goodness‐of‐fit, a simple regression model provided the most parsimonious fit to the empirical data, wherein Workload and Student Misbehaviour accounted for most of the variance in predicting teaching stress.