Papers in the Biological Sciences


Date of this Version



Bio/Technology (November 1989) 7(11): 1,134-1,139.


Copyright 1989, Nature Publishing Group. Used by permission.


There are significant differences if the distribution of weedy characteristics among weeds, normal plants, and crops. The world’s most serious weeds possess on the average 10 or 11 of these characters, a random collection of British plants have an average seven of the traits, and crop plants only five. For the average crop to become as “weedy” as the average weed, it would need to acquire five weedy traits. Even using the unlikely assumption that those traits are single loci in which a dominant mutation would provide the weedy character, this would require the simultaneous acquisition of five gene substitutions. Since the probability of multiple mutations is generally the joint probability of single mutations, the probability of changing the average crop to the average weed is (10-5), or 10-10. Even in the most numerous crop plants (perhaps 18 billion maize individuals are grown annually) this is not very probable. Since most of the crops listed are purchased from seed suppliers and not allowed to propagate, the plants will not gradually add weedy traits. Perennial and self-seeding crops, while more able to accumulate mutants, are generally grown in much smaller numbers. The probability of joint occurrence of new alleles producing significantly weedy plants from most crops is low.

There are several important qualifications to this finding. First, the mean result is only a mean. There is much less difference between the extreme individuals of the different groups. For example, among the weeds, Cirsium arvense (Canada thistle) infests 27 crops in 37 countries but appears to have only six of 12 weedy characteristics while, among the crops, tomatoes (Lycopersicon esculentum) have seven of 13 weedy characteristics, making them “weedier” on this measure than the thistle.

In addition, six of the 20 crop plants (30 percent) have weedy races, and nine of the 37 weeds (24.3 percent) are actively cultivated somewhere—indicating that the two categories actively exchange members. Even if a crop becomes a weed, only because cultivation is discontinued and not through evolution of weediness, a genetically engineered crop will still become a genetically engineered weed.

The recent emergence of a seriously weedy race of millet (Panicum miliaceum) in Wisconsin and Minnesota after 200-300 years of cultivation in North America without weed problems emphasizes how much we do not understand about weed evolution. Until such events can be anticipated, there will be an ongoing risk of weeds derived from genetically engineered crops.

This analysis should not be interpreted as a quick fix to problems of the new technology, but rather as directions for case-by-case problem solving. Plants with very low weediness and no weedy relatives are unlikely to be the source of weed populations in the future any more than they have been in the past (e.g., maize, pineapple). Plants with high inherent weediness and/or weedy relatives (oats, sunflowers) will, on the other hand, require serious scrutiny if we are to avoid additional problems. Moreover, study of the causes of weed success can suggest methods of modifying crop plants to reduce the risk of weed evolution. For example, infertile plants will have much less risk of producing weeds than fertile plants, due to lack of recombination, gene exchange, and propagules. Other approaches can also be suggested: poor seed longevity, careful management of vegetative reproduction, or dependence on cultivation practices, e.g. a trace mineral or soil disturbance for survival. To some degree, such dependencies already exist and could be exploited.