The snails Cepaea nemoralis and C. hortensis are remarkable for an extensive and stable polymorphism involving the colour and banding of the shell. It was formerly thought that the variation in frequency of the different morphs between populations was random. Cain & Sheppard, for C. nemoralis and Clarke, for C. hortensis, have shown that in many English colonies visual selection by thrushes, and no doubt other predators, strongly influences the frequencies of the morphs, the more conspicuous on a given background being more heavily predated. In consequence populations tend to match their backgrounds, but remain polymorphic. In some districts of high chalk downland, this correspondence with background does not occur. The predominance of a few morphs irrespective of habitat and background characterizes areas vastly larger than that of a panmictic population. Such a constancy of morph frequencies over a large and diverse area in spite of visual selection we call an area effect. The principal district we have studied is the Marlborough Downs, where in an area of several square kilometres there are no five-banded C. nemoralis although in a contiguous area they predominate. Part of the non-five-banded area has a vast excess of browns, and another part of yellows. The form spread-banded and the cross-product ratio of pink and yellow to unbanded and banded also show such effects. In some places the morph frequencies change with extraordinary abruptness over 100 to 300 m. The area effects are not due to differential incidence of visual predation, nor, since they bear no relation to variation of habitat, to differences in its direction. In only two subareas do we think that visual selection is affecting morph frequencies. The observed frequency distributions cannot be accounted for by sampling drift (`genetic drift') at the present day since the numbers involved are far too large and the frequencies too constant over large areas. In the few populations that have been observed for up to 10 years, no major changes of frequency have been found. The probability of a reduction to a few isolated populations because of ploughing up or drought in the last 200 years and subsequent drift and expansion is shown by the known agricultural history of the district to be slight. Restriction by spread of C. hortensis is also unlikely. A few colonies with restricted variation which might seem to show the action of drift or the founder effect are only extreme examples of local tendencies. Moreover, subfossil material from just off the south-western corner of the district strongly suggests that the area effects seen there have been in existence since Neolithic times. A survey of another district of high downland (Lambourn Downs) has shown a similar state of affairs to that on the Marlborough Downs, with a large area characterized by excess of yellow and mid-banded, and adjacent to it localities in which visual selection is effective. Observations from various other places on and off the Chalk also indicate that area effects are frequent on the Chalk, but that away from it visual selection is the principal agent determining local variation in gene frequencies. There is good evidence that the pigmentation of the body, which is apparently multifactorially controlled, also shows area effects; and part of the correspondence between body colour and background shade reported by Cain & Sheppard may be due to them. The evidence available for C. hortensis suggests that this species also shows area effects in shell characters. Yellow, pink or brown may predominate in C. nemoralis, but area effects in banding seem due mostly to the excess or defect of the modifier M^3 which reduces the five-banded phenotype to the form with only the middle band. It seems clear that the area effects are caused by some form of selection, but the topography, geology and vegetation of the Marlborough Downs gives no clue to what this could be for banding. Brown is known to be common only in the northern half of the range of nemoralis, and hortensis to extend much farther north than does nemoralis. A study of the distributions of the two species and of the brown morph on the Marlborough Downs suggests that local features of topography of open downland may produce localized climatic conditions influencing the relative distributions of the species and the abundance of brown. The abruptness of change of gene frequencies in both colour and banding might be caused by the change-over from one balanced gene complex to another requiring very different frequencies. Examination of Fisher's equation for stability of a polymorphism maintained by heterosis (the most likely condition in these species) shows that in the districts where visual selection is effective in altering gene frequencies in nemoralis, the heterozygote advantage can only be of the order of a few percent, and that local differences of a few percent in the selective disadvantages of the homozygotes concerned could well produce area effects as marked as those we have observed. For biological purposes it is essential to recognize the difference between changes in gene frequency caused by selection and those produced by the effects of sampling error. `Genetic drift' has been generally used to refer to the latter, but Sewall Wright uses it for all apparently random changes, whatever their cause, and perhaps for all changes in gene frequency; we therefore use sampling drift for the effects of sampling error. Surveys based only on the observations of frequencies and population size in widely scattered populations do not allow one to distinguish between the effects of selection that varies in direction and intensity from place to place (although more or less constant in time) and those of sampling drift. In general it is exceedingly difficult to identify the result of sampling drift in the wild except in certain situations. Casual collecting over such a district as the Marlborough Downs might well give the impression that sampling drift was effective there, but a more intensive survey shows the contrary.