What does a bootstrap value of 95 percent mean?

It means the clade appeared in 95 percent of resampled pseudo-replicate analyses; high values indicate the result is robust to resampling of the characters, though they are not exact statistical probabilities.

Why do Bayesian posterior probabilities differ from bootstrap values?

They are computed differently, with posterior probabilities reflecting the proportion of sampled trees containing a clade; for the same data they are frequently higher than the corresponding bootstrap support.

Tree Support and Confidence

Support measures quantify how strongly the data back individual branches of an estimated tree, so that strong and weak parts of a phylogeny can be distinguished.

Definition

Tree support is a quantitative estimate of confidence in a given clade or branch of a phylogenetic tree, derived from resampling the data or from the posterior distribution of trees.

Scope

This topic covers resampling-based support such as the nonparametric bootstrap and jackknife, decay (Bremer) support under parsimony, Bayesian posterior probabilities, and the interpretation and known biases of these measures, including the meaning of a particular support value.

Core questions

How do resampling methods estimate confidence in a clade?
What is the difference between bootstrap values and Bayesian posterior probabilities?
How should a given support value be interpreted?
What biases affect support measures?

Key theories

Nonparametric bootstrap: Characters are resampled with replacement to build pseudo-replicate datasets, and the frequency with which a clade recurs across replicates is reported as its bootstrap support.
Interpretation of bootstrap values: Empirical and simulation studies show bootstrap proportions are conservative estimates of reliability under realistic conditions, informing the convention that high values indicate well-supported clades.

Clinical relevance

Support values tell researchers which inferred relationships are trustworthy enough to act on, an essential safeguard when phylogenies guide outbreak attribution, source tracing, or conservation decisions.

History

Felsenstein's 1985 adaptation of the bootstrap to phylogenies introduced a practical confidence measure that became near-universal; subsequent empirical tests and the rise of Bayesian posterior probabilities refined how systematists report and interpret branch support.

Debates

Bootstrap versus Bayesian posterior probabilities: Bayesian posterior probabilities are often higher than bootstrap values for the same data, and there is ongoing discussion about which is better calibrated and how each should be interpreted.

Key figures

Joseph Felsenstein
David Hillis
James Bull

Seminal works

felsenstein1985
hillis1993
felsenstein2004

Frequently asked questions

What does a bootstrap value of 95 percent mean?: It means the clade appeared in 95 percent of resampled pseudo-replicate analyses; high values indicate the result is robust to resampling of the characters, though they are not exact statistical probabilities.
Why do Bayesian posterior probabilities differ from bootstrap values?: They are computed differently, with posterior probabilities reflecting the proportion of sampled trees containing a clade; for the same data they are frequently higher than the corresponding bootstrap support.