Aim After colonising isolated islands, many types of plants and animals undergo extreme morphological diversification. In stark contrast, other island taxa change more subtly and converge evolutionarily on intermediate body size (the island rule). This paradox of island evolution has gone undetected for decades, perhaps because each pattern has been investigated separately using different analytical tools. Here, I investigate the paradox of island evolution in the endemic floras of four oceanic archipelagos in the Southwest Pacific. Location Chatham, Kermadec, Lord Howe & Norfolk Islands. Taxon Woody plants. Methods I test whether the sizes of ‘solitary endemics’ (species that evolved anagenetically in the absence of co‐occurring endemic congeners) are correlated with their closest mainland ancestors. I also test whether island–mainland size relationships in 'solitary endemics' differ from isometry, as predicted by the island rule. Next, I test whether the stature of ‘co‐occurring endemics’ (closely related, often cladogenetic species) exhibits exaggerated differentiation. Results The stature of 'solitary endemics' was correlated with their mainland relatives, and two out of the four archipelagos showed evidence of the island rule. On the other hand, the stature of 'co‐occurring endemics' was unrelated to their closest mainland relatives and often highly differentiated. Main conclusions Overall results suggest that the island paradox arises from two distinct evolutionary pathways. Solitary endemics evolve more slowly and often converge on intermediate morphology, while co‐occurring endemics evolve more rapidly and often exhibit highly differentiated morphology. Future work aimed at a better understanding of the mechanisms underpinning the island paradox could help to further unite phylogenetic work on adaptive radiations and macroecological tests of the island rule under a single conceptual umbrella.