Catastrophe theory is introduced as a sort of merger of Whitney's theory of singularities of mappings and Poincaré's qualitative theory of dynamical systems. First Whitney. A surface is projected onto a plane. Somewhere the surface is folded, so that the inverse of the projection is multi-valued. Now, the plane may represent the possible values of the control parameters of a dynamic system, and the surface the possible states of the system. Moving continuously in the plane across the boundary between a single-valued and a multi-valued region may cause a jump on the surface to one of the other sheets--i.e. a small external change causes the system's state to change drastically: a "catastrophe". Poincaré's bifurcation theory of dynamical systems may now be perceived similarly on a metalevel where the systems themselves are points in a space--again an infinitesimal move in the system space may cause drastic changes of the system's equilibria. This type of geometric thinking may then be used in applications--but only sober ones, mind you: elasticity, optics, etc. Back in the old days Thom quite successfully pushed his catastrophe theory on gullible non-mathematicians. Arnol'd states his own view on those matters clearly and repeatedly throughout the book. From the preface: "Neither in 1965 nor later was I ever able to understand a word of Thom's own talks on catastrophes. He once described them to me as 'bla, bla, bla'". Arnol'd instead prefers the mathematical meat and potatoes of catastrophe theory: the theory of singularities. If only one were as enthusiastic as Arnol'd about singularity classification theorems then this would be very interesting indeed.