The usefulness of various suggested species demarcation criteria was compared in attempts to determine the taxonomic status of ten new tombusvirus isolates. Five of them (Lim 1, 2, 3, 5 and 6) were obtained from different sources of commercially grown statice (Limonium sinuatum), two (Gyp 1 and 2) from different sources of commercially grown Gypsophila paniculata and three from water samples, i.e. from a small river (Schunter) in Northern Germany, from a brook (near Dossenheim) in Southern Germany and from the groundwater in a Limonium production glasshouse in the Netherlands (Lim 4). The immunoelectron microscopical decoration test allowed a quick preliminary assignment of various isolates to several known tombusviruses. A more precise analysis of the relationships was achieved by comparing the deduced amino acid sequences of the coat proteins. Sequence as well as serological data suggested that eight of the isolates should be classified as strains or variants of either Carnation Italian ringspot virus, Grapevine Algerian latent virus, Petunia asteroid mosaic virus or Sikte waterborne virus, respectively, whereas the 9th isolate (Lim 2) appears to represent a distinct new tombusvirus species. The case of the 10th isolate (Lim 5) illustrates the classification problems experienced when the properties of a virus place it close to the more or less arbitrary man-made borderline between virus species and virus strains. The coat protein gene sequences were also determined for some viruses for which these data had not yet been available, i.e. Neckar river virus, Sikte waterborne virus and Eggplant mottled crinkle virus. The sequences of the coat protein gene and also of ORF 1 of the latter virus proved to be almost identical to the corresponding genome regions of the recently described Pear latent virus, which for priority reasons should be renamed. Criteria which have been suggested in addition to serology and sequence comparisons for tombusvirus species demarcation, i.e. differences in natural and in experimental host ranges, in cytopathological features and in coat protein size, appear to be of little value for the classification of new tombusviruses.