A set of eight closely related plasmid constructs carrying C1857-controlled recombinant genes has been used as a model to study plasmid stability in Escherichia coli, in the absence of antibiotic selection. Plasmid loss rates and relative interdivision times of plasmid-bearing cells and plasmid-free cells have been analyzed throughout prolonged cultures. Whereas the calculated plasmid loss rates are not consistent for a given plasmid and set of conditions, the relative growth fitness of plasmid-bearing cells is highly reproducible. In the absence of gene expression, plasmid maintenance is influenced by the length of the cloned segment, the growth temperature, and the plasmid copy number, but not by the plasmid size. At high, inducing temperatures, the effects of the metabolic burden are eclipsed by the toxicity exhibited by the different proteins produced, which is determined by structural features. Despite the multifactorial nature of the negative pressures acting independently on plasmid-bearing cells, the relative cell fitness in a mixed cell population is very reproducible for a given vector, resulting in a monotonous spread of the plasmid-free cells in recombinant cultures.