If non-genetic heritable information can become associated with a network of cycles built up by genetic evolution, subsequent changes in this information might seem to be largely degenerative involving cycle breakage. Another type of information may however be associated with a network. Suppose that in Figure 1(b) an increase in the intracellular activity of a occurs as an adaptive phenotypic response. If the rate of synthesis of b is kinetically dependent on the activity of a, the activity of b will increase. Thus, in turn, around the cycle, the activities of c and d will also increase, finally reproducing an increased activity of a in the next generation, without the necessity of the original adaptive stimulus. In the model of Figure 1(c) such changes might be transmitted, dispersed or amplified through a network of cycles. In multicellular organisms, cell multiplication or selection among somatic cell lines might also result in increases in the activity of signals within a network of cycles which could be transmitted to future generations. This additional source of heritable information has interesting consequences. First, stochastic development forces acting within female lines of descent might generate heritable variation on which directional selection could act; stabilising selection might also constrain variation between lines. Second, the inheritance of acquired adaptations could occur but unlike the somatic selection hypothesis19 information underlying the changes would not be passed to germline DNA. Third, the strength of causal signals in a network might be amplified over generations causing a phenotypic trend which, as with molecular drive,7 could occur in the face of natural selection. This would require simultaneous changes, resulting perhaps from climatic change, in all lines of descent to keep heritable variation between lines low. The mechanism could account for apparently maladaptive fossil trends.

Phenotypic information

Although the existence of cycles depends on nuclear gene activity cycles provide an additional source of heritable information. How by the criterion of heritable information is this additional information stored? Consider the synthesis of gene product a in Figure 1(b). At this point in time certain changes in magnitude in the activity of a can be transmitted by the cycle. It is convenient therefore to regard information as being stored in the value of the activity of a and transmitted to the next generation by the cycle. This storage is transient as the information is passed to the activity of b then c around the cycle. In multicellular organisms the same principle applies, the information being stored transiently in attributes such as intracellular concentration, cell number, body size and transmitted by a cycle. I propose to call this information 'phenotypic information' to emphasise its association with phenotypic attributes outside the germline, and to distinguish it from other sources of extranuclear information. Phenotypic information may also be nuclear if it is stored partly in the form of more or less stable associations between germline DNA and other macromolecules. It may be noted that reproduction of all information, for example nucleic acid replication, may be interpreted as a cyclic process but that not all cyclic processes store phenotypic information. The distinction between structure, steady states, or cycles may not always be clear cut as may be the case with information stored in the cortex of ciliates or eggs.

 The problems of evidence

The idea of the association of information with cyclic processes is not new and often alluded to in genetic texts. It is the aim of this paper to develop the idea explicitly to the extent now of speculating that phenotypic information may have an important role in evolutionary change.
Classical genetic analysis does not pinpoint all the heritable causes involved in the ontogenesis of a character. For example, some genes may be of such crucial importance for normal development that alternative forms are lethal. Conversely some characters particularly those determining the major morphological features of a species may be invariant. Thus, the existence of substantial genetic variation for a character does not exclude the influence of other heritable causes such as phenotypic information.
It is thought that the evolution of gross morphological differences between species arise from mutations in control genes rather than the evolution of new proteins.29 Consideration of the theoretical properties of phenotypic information suggests that it might play a similar role to control genes. It does however depend on nuclear gene activity and should be regarded as an additional rather than alternative source of information.
As to the question of direct evidence, I suggest that there are few experimental observations which provide clear answers one way or the other. Two main approaches can be considered.
First, attempts could be made to demonstrate heritable changes after physical manipulation or induction of adaptive phenotypic responses. Evidence for the inheritance of acquired characters has been elusive and it is difficult to exclude the genetic effects of selection, inbreeding or segregation.30 Traditional scepticism partly reflects the absence of a theoretical framework that phenotypic information and indeed the somatic selection hypothesis provide. Some more specific predictions can be made for phenotypic information; first the inherited effect may not be identical to the inducing phenotypic alteration, and second the alteration must occur before the phenotypic information is transmitted to the next generation.
A second approach to evidence is through the study of maternal influences. The complex role of the mother in phenomena such as localisation is consistent with the view that heritable information is associated with these maternal properties.5 A cyclic network could be required for setting up this complexity but if the network has a high density of interconnecting 'sufficient' causal links it will have considerable homeostatic power in repairing broken links. Thus although maternal influences may provide information crucial for development they do not necessarily store heritable information. This point is relevant when considering reciprocal phenotypic differences occurring frequently in species crosses. In order to demonstrate extranuclear inheritance it is necessary to replace, by backcrossing, the nuclear genes in the maternal line of one species with those of the other. This has been done in plants,31 but hybrid reproductive problems cause difficulties in animals. Such an experiment may however be able to distinguish between phenotypic information and other types of extranuclear inheritance. Thus, if cycles are broken by foreign nuclear genes in the backcross programme, phenotypic information would be lost but not necessarily restored by reversing the backcross programme and reĀ­inserting the original nuclear genes. Information stored in autonomous replicating structure is less likely to be affected in this way. Nuclear transplantation provides an alternative to backcrossing and an experiment involving two Amoeba species32 provides relevant results. Reciprocal nuclear transplants between species were made into enucleated cytoplasm, and hybrid populations were obtained resembling the nuclear parent biochemically and the cytoplasmic parent in morphological characters such as cell shape. Back transfers of nuclei produced phenotypes different from those of the original parents. These results are consistent with cycle breakage during the whole transfer process. Reciprocal differences have also been observed in interspecific nuclear transplants in multicellular animals, for example frogs and toads.33 Unfortunately, as with crosses, hybrid embryos usually die at an early stage, preventing demonstration of extranuclear inheritance. Reciprocal nuclear transplants have been made between two species of Xenopus and hybrid embryos raised to fertile adult frogs.34 These frogs and their offspring resembled the subspecies providing the nucleus. This suggests that phenotypic information is not involved in determining the subspecific differences, although rare hybrids intermediate between nuclear and cytoplasmic parents were found. Also as maternal effects were not present the cross would not have been chosen a priori in an investigation of extranuclear inheritance. Experiments of this kind do however provide possibilities for corroborating or refuting ideas concerning phenotypic information.
Totipotency of adult cells as observed in plants is on the whole not consistent with an important role for phenotypic information but where it exists in animals totipotency is usually lost after a few egg divisions.