For counting somatic cells in milk (''cell counting'') as part of the EC/EU- and national regulations standardization is indispensable; this includes, in particular, the comparability of methods. In the present study flow cytometry (Somacount) is compared with the fluorescence optical system (Fossomatic 360), as well as with optical count (Breed) as reference method, the aim being to establish, to what extent the new method yields comparable results. Precision was calculated from multiple counts. In the range below 500 (in 1000 cells/ml) the s.d. of repeated countings does not exceed 10, the variation coefficient in the range between 250 and 1.000 is 2 %, and beyond this range 1.6 %. After pretreatment using the preservatives potassium dichromate and Bronopol the samples keep for more than 48 h. With a limitation in time the same results in terms of shelf life are obtained when sodium azide and boric acid are used. The shelf life of the ''raw milk method'' (cold storage at 6-degrees-C) does not lead to temporal stability. The measuring results between preservation methods are comaparable, when potassium dichromate, boric acid or Bronopol were used. With the ''raw milk method'' slightly lower values are obtained. Preservation with sodium azide leads to markedly lower values. This is in agreement with the experience gained with other instruments. In the normal measuring range the carry over error of the apparatus from sample to sample is 1-2 %, and, hence, of no importance. No problems were encountered with the temporal stability of the measuring method for 4 days using potassium dichromate. With plus/minus 2-3 % relative differences (means of repeated countings) are within the range of general accuracy. Using 407 quarter foremilk samples Somacount and Fossomatic were compared; with untransformed scale a correlation coefficient of 0.998 was found and after log transformation 0.917. In units of the log scale the residual s.d. is 0.06 in the range between 40 (in 1000 cells/ml) and 2.500, which corresponds to a factor of 1.14. This means, the relative error is 14 %. As to the counting level a 1:1 agreement is found. The optical method's missing accuracy is disadvantageous for comparisons. Considered all in all, however, there are no substantial differences in the counting level compared with the Somacount. A first international collaborative trial involving 17 laboratories with altogether 26 sections was performed according to the IDF standard 148:1991. In this intercomparison trial 10 milks samples in the range between 129 and 1.086 (in 1000 cells/ml) were used. A first evaluation showed a satisfactory repeatability (r = 26-103). Reproducibility was poor (R = 62-541) for the 10 milks (= cell levels). This was, however, solely due to the inexperience of one laboratory only. After exclusion of this laboratory good intercomparison test results were obtained (r = 26-95, R = 14-89). The IDF-targets for cv(r) and cv(R) are clearly observed. With a cell level of 428 cv(r) is 4.1 % (target = 4-5 %) and cv(R) 8.1% (target = 10-12%). Hence, the Somacount instrument itself and the participating laboratories fulfil, after adequate training, the targets of the IDF standard 148:1998. Good Laboratory Practice (calibration of apparatus) is supported by reference materials, which are treated using thermo-chemical methods and keep for several months (''Kiel standards''). Microscopically controlled numbers of cells are available in biological matrix for controlling the laboratory equipment. There exist no systematic differences between the values counted in the standard samples prior to (raw milk) and after treatment.
