Several seed extraction procedures, used for detection of Clavibacter michiganensis ssp. michiganensis ( Cmm ) in naturally infected and artificially infested tomato seed lots were evaluated. Extraction methods that included grinding the seeds were significantly better at detecting the pathogen in three different seed lots than methods that used only soaking. The detection threshold of Cmm in relation to seed sample size was determined by adding naturally infected seeds into samples of three different sizes. Cmm was detected by agar plating assay, on three media (CNS, mSCM, D 2 ANX), and by direct PCR from seeds and Bio-PCR (bacteria cultured on agar media prior to PCR). In samples of 10 000 seeds containing one infected seed, Cmm could be detected only by Bio-PCR and in only one replicate out of five. In samples containing five or 10 infected seeds per 10 000 seeds, three of five and five of five replicates, respectively, were detected by the three detection methods. In samples of 5000 seeds, one infected seed could be detected in all five replicates only after adding a concentration step. A high correlation ( R 2 = 0·9448) between artificially infested seeds and the disease incidence was found. Seed lots infested with less than 58 colony-forming units (CFU) per g did not cause disease under glasshouse conditions, whereas lots with about 1000 CFU g − 1 caused disease in 78 plants out of 2000.

The coliform agar produced by Merck was tested for rapid diagnosis of Erwinia amylovora (the causal agent of fire blight) in pear blossoms. The medium enabled the diagnosis to be completed within 36 h. Diagnoses performed with the medium were confirmed by the BIOLOG and the fatty-acid profile methods. The diagnostic medium was used to determine the spatial distribution of colonized blossoms in the orchards and it was found that E. amylovora may be distributed both in clusters and at random. These findings were used in the development of a statistical model for sampling blossoms in the orchard. The model determines the number of trees to be sampled in the orchard and the number of blossoms be taken from each tree, which would enable the true colonization incidence of blossoms in the orchard to be estimated at desired levels of accuracy and confidence. Parameters included in the model are: the total number of trees in the orchard (T), the number of trees to be sampled in the orchard (t), the number of blossoms to be sampled from each tree (n), the true colonization incidence of blossoms (pi), a coefficient of aggregation (rho), the required level of confidence (1 - alpha), and the required level of accuracy (L). Sensitivity analyses revealed that the parameter governing sample size is the required level of accuracy. Sampling of 20 blossoms from each of several hundred trees is required to achieve an accuracy of +/-1%, but only a few single trees are needed for an accuracy level of +/-10%. A sampling procedure then was developed, validated with an independent data set, and found to be accurate. It was concluded that sampling of pear blossoms and estimation of the incidence of blossom colonization by E. amylovora could improve fire blight management, but not in all cases.