Purpose: This study assessed the difference in possible risk factors based on differences of case definition for shoulder disorders in a working population. Methods: A cross-sectional study was carried out among 698 workers in 12 manufacturing and hospital/service facilities. Shoulder symptoms and physical examinations were used to define shoulder disorders. Information on individual biomechanical exposures, personal characteristics, psychosocial workload, physical component score (pcs12) and mental component score (mcs12), departmental work organizational factors were also collected. Logistic regressions, using the group with no shoulder symptoms and no physical findings as reference (n = 514), were conducted to identify the variables related to the disorder criteria of: shoulder symptoms with no physical findings (n = 78), physical findings without symptoms (n = 54) and both symptoms and positive physical findings (n = 52). All of the multivariate models were adjusted for age and sex. Results: For those with shoulder symptoms but no physical findings, female [odds ratio (OR) 1.78, 95% confidence interval (95% CI) 1.45-5.13], pcs12 above median (OR 2.33, 95%CI 1.29-4.23), mcs12 above median (OR 1.95, 95% CI 1.06-3.57) and high social support (OR 1.19, 95% CI 1.07-1.33) were related factors. For those with no symptoms but positive physical findings, age over 40 (OR 2.10, 95%CI 1.08-4.06), good general health (OR 0.73, 95% CI 0.59-0.93), machine/line work pace (OR 9.8, 95% CI 1.18-81.94), and work rotation (OR 2.3, 95% CI 1.18-4.57) were related factors. For those with both symptoms and positive physical findings, good general health (OR 0.66, 95% CI 0.54-0.80) and pcs12 above median (OR 2.54, 95% CI 1.32-4.91) were related factors. Conclusion: The results suggested that case definition was critical in understanding of work related shoulder disorders.