a, To analyse the response of derived lines to selection, a single representative WS genotype was obtained from each replicate population (set of eight microcosms) from both ancestral and derived lines. To obtain the representative set of derived types under the CE regime, SM colonies were collected from the end of phase I at generation 10 and pooled. The pooled sample was used to found phase II, at the end of which a single WS-type representative of each replicate was selected as described for the baseline (see below). This yielded 14 such types, one representing each replicate. To obtain the set of baseline types representing the ancestral state, SBW25 was used to found phase II. At the end of the 3-day period lines were harvested and plated. The single most abundant WS type from the most densely populated plate was selected as representative of that replicate. A third set of representative WS genotypes was obtained from lines evolved under the CP regime. This was as for the CE regime, but instead of pooling SM at the end of phase I, WS were collected and pooled. A single dominant WS type was chosen from each replicate. b, Cell- and line-fitness assays. Lines founded by representative WS genotypes (from ancestral and derived lineages) were competed against a marked (blue colonies) reference strain (SM and WS, for the CE and CP regimes, respectively). Use of the marked reference strain allowed the competitive performance of all ancestral and derived types to be assessed against a single common genotype. Cell fitness is the total number of cells in the mat after phase I, whereas line fitness is the proportion of evolved ‘offspring’ mats relative to a marked reference strain. In total 2,472 microcosms were assayed (three replicate assays per line).