See René Thom, Structural Stability and Morphogenesis, trans. D.H. Fowler, foreword by C.H. Waddington, Reading, Massachusetts: W.A. Benjamin, 1975, pp. 280-283 
FINALITY IN BIOLOGY

summary of extract

A. Finality and Optimality

 "When a biologist finds an organ or behavior that is obviously well adapted, his first concern is to ignore this adaptive character and to emphasize the factors immediately responsible for the process.  For example, in the well-known study of the orientation of leaves toward light, he isolates a substance, an auxin, produced by light rays, which inhibits the growth of tissues.  The immediate mechanism of the process is then explained perfectly, and usually, for a biologist, that is sufficient.  But if we, goaded by an understandable feeling of intellectual dissatisfaction, ask him how it comes about that the process is so obviously beneficial to the plant's metabolism, he will certainly invoke a principle of natural selection: plants in which an accidental mutation established this process enjoyed an advantage that eliminated those without it through selection.  This lazy and entirely unverifiable answer is at present the only interpretation of biological finality, even though the process presents a challenge worthy of further explanation.

go back to summary 
 "The mathematician von Neumannnote commented that the evolution of a system can be described in classical mechanics in two ways: either by local differential equations, for example, Lagrange's or Hamilton's equations, or by a global variational principle, like Maupertuis' principle of least action; and these two descriptions are equivalent, even though one seems mechanistic and locally deterministic, whereas the other appears to be finalistic.  The same is probably true in biology: every epigenetic or homeostatic process is susceptible of a double interpretation, deterministic and finalistic.  We must not forget that the essential object of study in biology is not the isolated individual but the continuous form in space-time joining parents to descendants (the regulation figure); more precisely, when two or more species have some functional interaction between each other, such as predation or being an auxiliary in the fertilization process, etc., it is necessary to consider the total figure in space-time, the union of all the forms associated with each species.  Then, for each adaptive process, we canprobably find a function S of the local biological state expressing in some way the local complexity of the state with respect to the process considered, and the configuration will evolve between two times t0 and t1, (e.g., the parent at age A and the descendant at the same age) in such a way as to minimize the global complexity [formula given].  In this way the minimum complexity and hence the most economical adaptation of the process will be realized.  Natural selection is one factor in this evolution, but I myself think that internal mechanisms of Lamarckian character also act in the same direction.  However, in contrast with classical mechanics, we should not expect that this evolution will be differentiable, or even continuous, on the individual level because the global continuous configuration must conform to the boundary conditions of a system restricted by spatial reproduction in a given chemical and ecological context.  Hence there will be not a continuous deformation but a finite chain of relatively well-determined, subtly interrelated, local processes (or chreods); even the variation of the global figure can introduce qualitative discontinuities into the structure of this chain - this is called mutation.  The effect of the global variational principle will be too weak for the local mechanisms to show no random fluctuations, and only the resultant of these local variations will finally be oriented by the variational principle.  Although the teleological nature of organs and behavior in living beings will be immediately apparent to us (with reference to what we ourselves are and to our own behavior as human animals), their deterministic and mechanistic nature will escape our attention because it operates on a very long time scale and has a statistical character inherent in evolution, and its decisive factors (the influence of metabolism on the statistic of mutations) are probably very tenuous.  Let me be more precise. 
go back to summary 
B. Chance and mutations

 "One of the dogmas of present-day biology is the strictly random (if this means anything) nature of mutations; however, it seems to me that this dogma contradicts the mechanical principle of action and reaction: of two possible mutations m and m', the one with the better effect on metabolism (i.e., the one that minimises the production of entropy) must have a greater probability of happening.  In the classical diagram of information theory,

source  -»  channel  -»  receptor
it is clear that the source has an effect on the receptor; therefore the receptor has an inverse effect on the source, usually unobservable because the energy of the source is very large with respect to the interaction energy.  This is certainly not the case, however, of nucleic acid, where the binding energy is much less than the energies of the metabolism.  One might object that here the receptor is an open system, in the language of thermodynamics; it is possible that DNA has a directing action on the metabolism not requiring the introduction of a large interaction energy.  In systems in catastrophe, a very slight variation in the initial conditions can cause large modification of the final state, and the interaction of the DNA chromosome could give rise to very small initial variations amplified later to large effects, a situation similar to that of a point determining the route of a train whereas the train has no effect on the point.  But this comparison is specious, as are all examples taken from human technology; they can occur only in a state of zero metabolism.  The effect of a signalman altering slightly the points under a moving train is disastrous, whereas it seems that most spontaneous mutations occur in interphase, during full metabolic activity.  The breakages and displacements of chromosomes observed in metaphase are only the visible results of earlier metabolic accidents in the interphase which have upset the course of the anaphasal catastrophe. 
go back to summary 
 Most mutations are attributed to chemical modifications in the DNA sequence in nucleotides, due to errors in the duplication process of DNA.  I am reluctant to subscribe to the current belief that a point mutation, affecting just one nucleotide, is sufficient to inhibit the activity of a gene; this seems to me to repeat on another plane the error of the morphologists who believed that the destruction of one neuron in the brain would stop the process of thinking.  To suppose the strict validity, without some random noise, of the genetic code amounts to making the basic regulation mechanism of the cell fully dependent on a process in a state of permanent catastrophe.  Even if life is only a tissue of  catastrophes, as is often said, we must take into account that these catastrophes are constrained by the global stability of the process and are not the more-or-less hazardous game of mad molecular combination.  Even adopting the anthropomorphic point of view that there is a mechanism for reading the DNA that is perturbed by errors, might we not push this anthropomorphism to its full extent and admit that the errors are oriented, as in Freudian psychology, by the "unconscious" needs and desires of the ambient metabolism ?  It seems difficult to avoid the conclusion that the metabolism has an effect, probably very weak, which in the long run can dominate the statistic of mutations, and the long-term results of the effect explain the variational principle of minimum complexity and the increasing adaptation of biological processes leading to finality.
GO BACK TO THE MAIN PAGE ON CHANCE