94 DIFFICULTIES AND SOURCES OF ERROR. [CH. 



see in them serious warnings against dogmatic conclusions or 

 expressions of opinion on imperfect data and insufficient evidence. 

 The description of a fragment of a handle of a Wedgewood 

 teapot as a curious form of Calamite^ and similar instances of 

 unusual determinations need not detain us as examples of 

 instructive errors. The late Prof Williamson has on more than 

 one occasion expressed himself in no undecided manner as to 

 the futility of attempting to determine specific forms among 

 fossil plants, without the aid of internal structure^ ; and even in 

 the case of well-preserved petrifactions he always refused to 

 commit himself to definite specific diagnoses. In his remarks in 

 this connection, Williamson no doubt allowed himself to express 

 a much needed warning in too sweeping language. It is one of 

 the most serious drawbacks in palaeobotanical researches that 

 in the majority of cases the specimens of plants are both 

 fragmentary and without any trace of internal structure. 

 Specimens in which the anatomical characters have been pre- 

 served necessarily possess far greater value from the botanist's 

 point of view than thqse in which no such petrifaction has 

 occurred. On the other hand, however, it is perfectly possible 

 with due care to obtain trustworthy and valuable results from 

 the examination of structureless casts and impressions. In 

 dealing with the less promising forms of plant fossils, there is 

 in the first place the danger of trusting to superficial resemblance. 

 Hundreds of fossil plants have been described under the names 

 of existing genera on the strength of a supposed agreement in 

 external form ; but such determinations are very frequently not 

 only valueless but dangerously misleading. Unless the evi- 

 dence is of the best, it is a serious mistake to make use 

 of recent generic designations. If we consider the difficulties 

 which would attend an attempt to determine the leaves, 

 fragments of stems and other detached portions of various 

 recent genera, we can better appreciate the greater probability 

 of error in the case of imperfectly preserved fossil fragments. 



The portions of stems represented in figures 20 and 21, ex- 

 hibit a fairly close resemblance to one another ; in the absence 



' An example referred to by Carruthers (71) p. 444. 

 2 WilUamson (71) p. 507. 



