The minor transitions in hierarchical evolution and the question of a directional bias

One of the most salient and certain trends in the history of life is the increase in hierarchical structure in organisms. Figure 1 shows the approximate trajectory of the trend. The left-most data point marks the ®rst known prokaryotic cells, the next is the origin of the ®rst eukaryotic cell from symbiotic associations of prokaryotic cells, the third is the advent of multicellular individuals from clones of eukaryotic cells, and the last is the origin of the ®rst colonial individual from associations of multicellular entities. The increase is hierarchical in that in each transition, a new level of nesting of entities within entities arose. The existence of such a hierarchical trend has been widely acknowledged (e.g. Spencer, 1900, 1904; Needham, 1943; Stebbins, 1969; Wimsatt, 1976, 1994; Corning, 1983; Salthe, 1985, 1993; Buss, 1987; Bonner, 1988; Maynard Smith, 1988; Swenson & Turvey, 1991; Maynard Smith & SzathmaÂry, 1995, 1999; SzathmaÂry & Maynard Smith, 1995; McShea, 1996a; Pettersson, 1996; Heylighen, 1999; Stewart, 2000; Wright, 2000; Knoll & Bambach, 2000). It is also ± to my knowledge ± unchallenged. What is not widely agreed upon is the mechanism or dynamics of the trend, the pattern of change among lineages that accounts for it. One possibility is that change among lineages is biased, so that increases are more probable than decreases, as shown in Fig. 2(A). Such a mechanism would be consistent with the suggestion that hierarchical increases are blocked ordinarily by design limitations, by what Sterelny & Grif®ths (1999) call `right walls' on the hierarchy scale, but that very occasionally these walls are breached, producing an organism with greater hierarchical structure (Stebbins, 1969; Sterelny & Grif®ths, 1999; Sterelny, 1999; Knoll & Bambach, 2000). In most versions of this mechanism, breaching of the walls is understood to be irreversible, ratchet-like, or nearly so (e.g. Maynard Smith & SzathmaÂry, 1995), which is equivalent to a strong directional bias. (The term bias is sometimes associated with error, and thus the phrase `directional bias' might