Serum samples from 361 pigs (194 fattening pigs and 167 sows) were examined by means of two commercial ELISAs (Svanovir; Svanova Biotech and Salmotype; Labor Diagnostik) used for the serological diagnosis of salmonellosis in pigs; 211 of the samples came from farms of known bacteriological status and the other 150 were collected randomly from 60 farms of unknown status. The ELISAs were done according to the manufacturers’ directions and the samples were categorised accordingly. The results were compared by using a linear regression analysis and by the calculation of Kappa values. To try to improve the agreement between the tests, the raw optical densities (ODs) were transformed to sample/positive (S/P) ratios by using the positive control as a reference, and cut-off values for these S/P ratios were calculated by means of a receiver operating characteristic (ROC) analysis. All but two of the known infected farms were recognised as such by both tests. However, the correlation of the raw ODs for individual pigs was poor (r=0·546) and had a Kappa value for the results categorised according to the manufacturers’ recommendations of 0·191. On some farms the correlation was high (r=0·97) but on others it was low (r=0·05) with no apparent reason for the difference. The S/P ratios did not improve the agreement (Kappa=0·25).