The three possible stemmata that could be reconstructed from the example data in Table 6 (all the possible ways of forming two pairs from four manuscripts). Reconstructed readings of hypothetical nodes are shown in italics (the ACCTRAN, DELTRAN and MINF algorithms all give identical reconstructions). We show all edges with equal length to illustrate the topology of the stemmata, but label each edge with the number of changes of reading occurring along it. The sum of these numbers is the total number of changes of reading required over the whole stemma ((A) 6, (B) 8, (C) 7). The most parsimonious stemma is therefore (A). Edges along which no changes occur can be collapsed. Thus (B) has all the extant manuscripts forming a star, and (C) includes two extant manuscripts, one of which is directly descended from the other.