The hypothesis of functionally adaptive diversification of wood structure in the course of evolution as advanced by Carlquist is critically tested for vessel member length and type of perforation plate. The functional significance of within-tree variation in vessel member morphology is discussed first, and a criticism of Philipson & Butterfield’s theory for explaining classical length-on-age curves for fusiform initial length is given. At the genus and family level examples are cited which contradict the hypothesis that specialization of the perforation plate has a strong selective advantage for the occupation of more xeric environments or for high conductive rates. Other examples, however, seem to support this. The functional interpretation of correlations between vessel member length and ecological conditions is criticised. A comparison is made between whole woody floras or vegetation types with respect to the proportion of genera with scalariform perforation plates (Tables 1 and 2, Fig. 2). In tropical lowland rain forests the percentage of genera with scalariform plates is rather low. It strongly increases for tropical montane forests and for temperate to arctic floras. It is lowest in seasonally dry to arid regions. These trends support Carlquist’s idea that scalariform plates are, in general, only successful for conditions which require low rates of water conduction. In addition, the role of temperature is stressed. The total evidence presented in this paper is discussed in terms of random ‘patio ludens’ evolution which has to a considerable extent been canalized through selective pressures by environmental factors.