DEVELOPMENT OF MATHEMATICAL THOUGHT. 705 



mathematical thought. Up to that time " one would 

 have said that a continuous function is essentially cap- 

 able of being represented by a curve, and that a curve 

 has always a tangent. Such reasoning has no mathe- 

 matical value whatever; it is founded on intuition, or 

 rather on a visible representation. But such representa- 

 tion is crude and misleading. We think we can figure 

 to ourselves a curve without thickness ; but we only 

 figure a stroke of small thickness. In like manner we 

 see the tangent as a straight baud of small thickness, 

 and when we say that it touches the curve, we wish 

 merely to say that these two bands coincide without 

 crossing. If that is what we call a curve and a tangent, 

 it is clear that every curve has a tangent ; but this has 

 nothing to do with the theory of functions. We see to 

 what error we are led by a foolish confidence in what 

 we take to be visual evidence. By the discovery of this 

 striking example Weierstrass has accordingly given us a 

 useful reminder, and has taught us better to appreciate 

 the faultless and purely arithmetical methods with which 

 he more than any one has enriched our science." 1 



" metaphysics and theory of the 

 fundamental conceptions in mathe- 

 matics : quantity, limit, argument, 

 and function " (Tubingen). This 

 work touches the borderland of 

 mathematics and philosophy, as 

 does the same author's posthumous 

 work ' Uber die Grundlagen der 

 Erkenntniss in den exacten Wissen- 

 schaften' (Tubingen, 1890), and will 

 occupy us in another place. 



1 M. Poincar^ in the 'Acta 

 Mathematical vol. xxii., " L'ceuvre 

 mathe'matique de Weierstrass," p. 

 5. The " test-case " referred to in 

 the text consisted in the publica- 



VOL. II. 



tion by Weierstrass (in the year 

 1872, 'Trans. Berlin Academy,' re- 

 printed in Weierstrass's ' Math. 

 Werke,' vol. ii. p. 71) of the proof 

 of the existence of a continuous 

 function which nowhere possessed 

 a definite (finite or infinite) differ- 

 ential coefficient. This example 

 cleared up a point brought into 

 prominence by lliemann in his 

 posthumously (1867) published 

 Inaugural Dissertation of 1854 

 ('Werke,' p. 213). The question 

 had already, following on Rie- 

 mann's suggestions, been dis- 

 cussed by Hermann Hankel in a 



2 Y 



