Table 1: Relationships between patterns in trait-space use (Figures 2(a)2(d)), the GAP model, and statistical tests implemented for analysis of attacked trait spaces.

Pattern in trait-space useHypothesisAttacked trait spaces marginal? 1Attacked trait spaces restricted?1Attacked trait spaces distant?

Monophagous, attacked trait space marginal and adjacent to alternative host (Figure 2(a))Adjacency favours host shifting (adjacent errors hypothesis)Ancestral host: marginal Novel host: marginal but rare2Attacked spaces close2

Oligophagous, attacked trait spaces marginal and adjacent (Figure 2(b))Adjacency persists after host shifting (adjacent oligophagy hypothesis)Ancestral host: marginal Novel host: marginalAttacked spaces close

Nascent host races, attacked trait spaces marginal and distant (Figure 2(c))Distance permits, and is also symptomatic of, genetic isolation (trait distance-divergence hypothesis)Ancestral host: marginal Novel host: marginalAttacked spaces distant

Pair of monophagous species, attacked trait spaces central on each host (Figure 2(d))Other isolating mechanisms reduce importance of trait-space distance (distance relaxation hypothesis)Ancestral host: not marginal
Novel host: not marginal
Ancestral host: restricted (central) or nonselective Novel host: restricted (central) or nonselectiveAttacked spaces neither close nor distant

1More strictly, only restricted or marginal trait-space use along the PC axis (or axes) defining the difference between available hosts is directly relevant to the GAP model.
2But statistical detection is difficult because attack on the novel host is rare.