This study provides estimates of practice variation for primary physician services in Norway. Nearly all previous studies on practice variation have encompassed mainly specialist and hospital services. Thus we provide evidence for treatment variation for a health care sector where information about treatment variation is relatively scarce. Our analyses were carried out on a large set of data from individual physician practices. We estimated the extent of practice variation according to type of diagnosis and type of treatment option. The main finding is that variation in clinical practice between physicians is an important determinant of expenditure for primary physician services. Depending on the type of diagnosis, physician-specific effects explained 47-66% of the variation in expenditure for laboratory tests, 59-66% of the variation in expenditure for consultations lasting over 20min and 49-61% of the variation in expenditure for specific procedures. The age and gender of the patients explained only about 1% of the variation in expenditure. This finding supports the results from previous studies, which have shown that patients' age and gender are inadequate when used as risk adjusters. We also identified physicians who moved their practice from one municipality to another, i.e. they changed their patient population completely. Our results showed that those physicians who moved did not change practice style after they moved, i.e. that physician's style of practice is stabile. This indicates that practice style reflects a deeply rooted behaviour with respect to how to practice medicine.