Co-suppression of host genes and 35S transgenes encoding nitrate reductase was previously reported in transgenic tobacco plants (Nicotiana tabacum cv. Paraguay or Burley) using either a full-length cDNA or fragments devoid of the 3' and/or 5' UTR. Co-suppression was previously shown to affect a limited fraction of the progeny of one transgenic tobacco line homozygous for a single transgene locus, and the phenomenon occurred at each generation. In this work, 38 combinations of transgene loci derived from 13 independent transgenic lines homozygous for a single transgene locus were field-tested under two different conditions in an attempt to determine the corresponding frequencies of co-suppression, i.e. the percentage of plants showing co-suppression. Each of the 13 homozygous lines exhibited a different frequency of co-suppression, ranging from 0% to 57%. High frequencies were found to be associated with transgene loc carrying a high number of copy of the transgene, suggesting a transgene dose effect. Combinations carrying 2 non-allelic transgene loci in a hemizygous state exhibited frequencies of co-suppression between those of each of the 2 transgene loci in a homozygous state, while combinations carrying 2 non-allelic transgene loci in a homozygous state exhibited frequencies of co-suppression higher than the sum of those of the 2 transgene loci alone in a homozygous state, clearly confirming a transgene dose effect. Co-suppression frequencies were increased when the plants were grown initially in vitro, suggesting some environmental effect. The roles of transgene copy number, number of transgene loci and environmental factors are discussed in the light of a threshold hypothesis.