groups in our phylogenetic analysis. The tree used simply corresponds to one out of more
than 8000 most parsimonious trees reported by Soltis et al. (1999, 2000), but evaluating
the amount of uncertainty this places on all our estimates is not possible. This simply has
to be looked at on a group-by-group basis. It is worth noting that the great majority of
groups are consistently resolved and receive at least some jackknife support; in particular
the spine of the tree is consistently resolved, and the variability that generates the greater
than 8000 equally shortest trees occurs within terminal groups that are not an issue here
(Soltis et al. 1999, 2000). Relationships within some of the more derived groups such as
the Malpighiales and Lamiales, are resolved in the tree used here, but many of these
receive less than 50 per cent jackknife support (Soltis et al. 1999, 2000). Resolving the
relationships differently within these groups will probably have limited consequences on
the timings for the more inclusive groups (Malpigiales, Lamiales). We have also indicated
on the chronograms (Figures 8.3–8.6) nodes with less than 50 per cent jackknife support.
Future analysesThe analyses conducted are unconstrained, including no fossil-based age constraints; this
permits us to evaluate how the molecular data on their own resolve angiosperm
diversification. One way forward from the current analyses might involve including
constraints, and the type of analyses (NPRS) allows for either minimum- or maximum-age
constraints to be enforced during the analyses. Enforcing constraints affects the actual
analyses and forces the results to fall within the boundaries specified. Although after such
an inclusion we can no longer independently evaluate the fossil-based estimates, such an
approach may provide not only a way of improving the actual estimates, but also a way of
evaluating how evolutionary rates would be resolved, should we accept the current fossil-
based estimates. An analysis including fossil-based constraints would, however, require a
detailed and critical evaluation of the available fossil information, which is clearly beyond
the scope of our current work.
Existing methods for estimating divergence time cannot combine data (necessary if
stochastic errors are to be further reduced) and at the same time take different rate
characteristics into account, much as early likelihood models used for phylogeny
reconstruction were all simple and without such capabilities. Perhaps work such as this
will promote not only an evaluation of the assumptions used in NPRS analyses but also
further developments, so that we can look forward to corresponding improvements in age
estimation analyses over those we have seen in the development of likelihood models for
phylogeny reconstruction. By using the available fossil information as age constraints,
analyses of this kind have the advantage of providing ways to estimate the time of origin for
all the groups lacking a decent fossil record. They also force our estimates into a more
rigorous hierarchical framework, and without such a framework, the full implications of
documenting derived groups from successively older geological deposits become less
clear. The continued incongruence between the results presented here and fossil-based
estimates such as the Turonian Clusiaceae (Crepet and Nixon 1998), the Campanian and
Santonian Actinidiaceae (Keller et al. 1996; Herendeen et al. 1999), the Santonian
Apiaceae/Araliaceae (Herendeen et al. 1999), and the Cenomanian Rosaceae/
Rhamnaceae (Basinger and Dilcher 1984) indicate that we may still underestimate the
164 ANGIOSPERM DIVERGENCE