But though this reply has a formal finality, in my view it evades the underlying question that is being asked. True, life could not have evolved in the absence of a steady stream of energy from the sun -- a kind of energy wind on the earth. But if there were no more to the mechanism of molecular evolution than this, we should still be at a loss to understand how more and more complex molecules cam to establish themselves. All the energy wind can do, in itself, is to increase the range and frequency of variation around the average state: that is, to stimulate the formation of more complex molecular arrangements. But most of these variant arrangements fall back to the norm almost at once, by the usual thermodynamic processes of degradation; so that it remains to be explained why they do not all do so, and how instead some complex arrangements establish themselves, and become the base for further complexity in their turn.

It is therefore relevant to discuss the Second Law, which is usually interpreted to mean that all constituent parts of a system must fall progressively to their simplest states. But this interpretation quite misunderstands the character of statistical laws in general in nonequilibrium states. The Second Law describes the final equilibrium state of a system; if we are to apply it, as here, to stable states which are far from equilibrium, we must interpret it and formulate it differently. In these conditions, the Second Law of Thermodynamics becomes a physical law only if there is added to it the condition that there are no preferred states of configurations.

In itself, the Second Law merely enumerates all the configurations which a system could take up, and it remarks that the largest number in this count are average or featureless. Therefore, if there are no preferred configurations (that is, no hidden stabilities in the system on the way to equilibrium), we must expect that any special feature we find is exceptional and temporary, and will revert to the average in the long run. This is a true theorem in combinatorial arithmetic, and (like other statistical laws) a fair guess at the behavior of long runs. But it tells us little about the natural world which, in the years since the Second Law seemed exciting, has turned out to be full of preferred configurations and hidden stabilities, even at the most basic and inanimate level of atomic structure.

The Second Law describes the statistics of a system around equilibrium whose configurations are all equal, and it makes the obvious remark that chance can only make such a system fluctuate around its average. There are no stable states in such a system, and there is therefore no stratum that can establish itself; the system stays around its average only by a principle of indifference, because numerically the most configurations are bunched around the average.

But if there are hidden relations in the system on the way to equilibrium which cause some configurations to be stable, the statistics are changed. The preferred configurations may be unimaginably rare; nevertheless, they present another level around which the system can bunch, and there is now a countercurrent or tug-of-war within the system between this level and the average. Since the average has no inherent stability, the preferred stable configuration will capture members of the system often enough to change the distribution; and, in the end, the system will be established at this level as a new average. In this way, local systems of a fair size can climb up from one level of stability to the next, even though the configuration at the higher level is rare. When the higher level becomes the new average, the climb is repeated to the next higher level of stability; and so on up the level of strata.

When there are hidden levels of stability, one above another, as there are in our universe, it follows that the direction of time is given by the evolutionary process that climbs them one by one. Indeed, if this were not so, it would have been impossible to conceive how the features that we remark could have arisen. We should have to posit a miraculous beginning to time at which the features (and we among them) were created ready-made, and left to fall apart ever since into a tohubohu of individual particles.

Time in the large, open time, takes its direction from the evolutionary processes which mark and scale it. So it is pointless to ask why evolution has a fixed direction in time, and to draw conclusions from the speculation. It is evolution, physical and biological, that gives time its direction; and no mystical explanation is required when there is nothing to explain. The progression from simple to complex, the building up of stratified stability, is the necessary character of evolution from which time takes its direction. And it is not a forward direction in the sense of a thrust towards the future, a headed arrow. What evolution does is to give the arrow of time a barb which stops it from running backwards; and once it has this barb, the chance play of errors will take it forward of itself.