Mean of posterior probabilities of Bayesian inference character state evolution of successive iterations (9,000,000) by RJ MCMC (see text) for six characters.

The 95% confidence intervals of the posterior probabilities were all less than��0.004. In bold character state evolution as traced in MacClade optimization (Fig. 2). Particular points of disagreement between Bayesian and the MacClade optimization are underlined. Node codes as in Fig. 2A.Note: Values in the table reflect estimates based on the averaging over 1000 Bayesian tree.aLeaf shape: linear/linear-lanceolate to elliptic/ovate-lanceolate/ovate.bLabdanum secretion: 0�C5/5�C10/10�C15% per unit leaf dry weight.cLeaf pubescence: glabre/subglabre/dense tomentum.dSoil: silicolous/calcicolous/serpentin.eInsolation conditions: helioxerophyllous/subheliophyllous/subsciophyllous/submesophyllous.fEnvironment: bush/scrub/Quercus suber-ilex& Pinus woodlands/Quercus suber-ilex woodlands/Quercus pyrenaica-faginea & Pinus woodlands/Pinus halepensis-pinea & Quercus suber woodlands/Abies cephalonica woodlands.