Studies on Sterility and Prenatal Mortality in Wild Rabbits: I. The Reliability of Estimates of Prenatal Mortality Based on Counts of Corpora Lutea, Implantation Sites and Embryos

Sterility in a female mammal may result from failure to produce fertilized ova, or from subsequent death of the embryos. Although only the latter falls strictly under the heading of prenatal mortality, both are included in these studies. It has been shown already (Brambell, 1942, 1944) that in the wild rabbit by far the greater part of the wastage of ova is due to the death of the embryos at some stage after fertilization. This may be due in part to the fact that in the rabbit ovulation is not spontaneous, the animal remaining in persistent oestrus until copulation occurs or the breeding season ends. Once copulation occurs in an oestrous rabbit ovulation ensues approximately 12 hr. thereafter. Since very few oestrous rabbits fail to produce ova the proportion that do not become pregnant is small. Female rabbits that are not in oestrus, unlike the majority of other mammals, will copulate freely (Brambell, 1944), not only during pregnancy and pseudopregnancy, but even during the non-breeding season.

Although these special considerations apply only to the rabbit and a few other species, it appears that even in species that ovulate spontaneously, ovulation and fertilization rarely fail to occur at each oestrus, at least in the wild state during the effective breeding season, and that sterility, when it occurs, is due mainly to the subsequent death of the embryos. It seems probable that much of the sterility in farm-stock, and even in man, that is commonly attributed to failure to achieve fertilization, likewise is due, in fact, to the death and reabsorption or abortion of the embryos. It is only when abortion of relatively advanced embryos occurs that prenatal mortality can be recognized readily in the living animal. If the dead embryos are reabsorbed in situ, or if they are aborted at an early stage of development, the mortality is difficult or impossible to detect, except at autopsy. Moreover, if the mortality occurs early in development, before or soon after implantation, it may result in no interruption of the oestrous cycle, and then the fact that the animal was pregnant, not pseudopregnant, can be recognized only by dissection. The death of whole litters between the eleventh and fifteenth days of gestation, that has been shown to occur commonly in the wild rabbit (Brambell, 1942, 1944; Brambell & Mills, 1944), and their subsequent reabsorption, results in the animals coming into oestrus again at or soon after the time when they would have done so had they been pseudopregnant ; and the degree of development of the mammary glands at this time is comparable also. If this phenomenon occurred in tame rabbits the animals affected would be classed almost certainly as pseudopregnancies, unless the uteri were examined while reabsorption was proceeding.

It is apparent, from these considerations, that a thorough investigation of prenatal mortality, in at least one suitably selected species, is an essential preliminary to the study of sterility in mammals generally. Such an investigation should be on a large scale and should lead to the elaboration of suitable techniques, which could subsequently be adapted to other species. It should yield much information concerning the fundamental problems involved in the detection and analysis of prenatal mortality in mammals.

These studies have been undertaken with the objectives named above. They represent a continuation and extension of the preliminary work already reported (Brambell, 1942, 1944). They are based on extensive new data, but the earlier data have been reanalysed and are included also. The reason for their inclusion is that, in conjunction with the new data, they have yielded more information, rendering it possible to amplify and to correct the interpretations previously placed upon them.

The rabbit provides very favourable material for such studies because, in the first place, it can be investigated in the wild state under natural conditions and, at the same time, can be kept conveniently as an experimental animal; secondly, it is polytocous; thirdly, it has a short gestation period; and, fourthly, it is available in quantity and is cheap. Monotocous animals provide much more difficult material because, once the single embryo has died and has been removed, the mother is no longer pregnant. In such animals the mortality at early stages of development can be discovered only if examination takes place during the brief interval between the death and the removal of the embryos. At later stages the dead embryos are removed by abortion as a rule, and then the involution of the uterus is more rapid than if reabsorption had occurred, with a consequent decrease in the period during which the loss can be recognized at autopsy. Polytocous animals that lose the whole litter simultaneously present the same difficulty as monotocous species, in that they drop out of the category of pregnancies, but even so, in them the dead embryos often are reabsorbed in situ and the mortality is detectable at autopsy, unless the animal was near full-term. The death of some only of the embryos in a polytocous animal seldom terminates pregnancy as the dead embryos are reabsorbed without interfering seriously with the course of development of the remainder. In such cases, since the corpora lutea in the ovaries and the sites in the uteri, where the dead embryos were reabsorbed, are remarkably persistent structures, often remaining identifiable until parturition, as in the rabbit, the total mortality of ova throughout the period of pregnancy up to the time of examination can be determined from comparison of counts of the corpora lutea, implantation sites and embryos. At any given stage of pregnancy the discrepancy between the number of living embryos in the uteri and of corpora lutea in the ovaries, except for certain qualifications that will be considered in this paper, is equivalent to the total loss of ova from the time of ovulation until the animal is examined. Further, once implantation has occurred, the implantation sites provide a means of distinguishing between the loss of ova that occurred before implantation and that which occurred after implantation. The excess in the number of corpora lutea over the number of implantation sites is equivalent to the number of ova, fertilized or unfertilized, lost before implantation in surviving litters. The excess in the number of implantation sites over the number of living embryos is equivalent to the loss after implantation.

Since all the subsequent statistical work is based on counts of corpora lutea, implantation sites and embryos it is necessary first to inquire into the accuracy of these. The newly ruptured follicle gradually develops into a corpus luteum, which attains its maximum size during mid-pregnancy. Thereafter it remains constant in size until it begins to regress rapidly a few days before parturition, at which time in the rabbit it is still quite large and easily visible to the naked eye. After parturition it regresses more rapidly and soon disappears. If there is any tendency to overlook corpora lutea when counting, it might be expected that the incidence of such errors would be greater at the beginning and end of pregnancy, and least at mid-pregnancy, during the period of maximum development.

Similarly, implantation sites undergo regression after the death of the embryos. They do so relatively rapidly in animals that have ceased to be pregnant. The rate of reabsorption of the embryos and regression of the sites in such animals, in which the death of all the embryos at known stages of development has been induced experimentally, will be the subject of another paper that is in preparation. It is possible, indeed probable, that the rate of regression is retarded in animals that, owing to the survival of some of the embryos, continue to be pregnant, but it is certainly necessary to investigate whether, in such cases, the sites of embryos that have died soon after implantation persist until full-term or whether counts made near the end of pregnancy are liable to error through the disappearance of such sites.

Finally, although counts of embryos after implantation can be relied on, since the embryos are unmistakable at that time and become increasingly conspicuous thereafter, counts made before implantation, because of the technique and the small size of the ova or early embryos, are liable to serious error, and it is desirable to know if they are of any significance. Before implantation the ova or early embryos can be recovered only either by the perfusion of the Fallopian tubes or uteri as the case may be, or by the much more laborious method of serially sectioning these organs. The labour involved in the latter process is prohibitive in large samples of an animal as big as the rabbit. Although the proportion of ova and unimplanted embryos that can be recovered by perfusion in rabbits is high, it is difficult for technical reasons that have been outlined previously (Brambell, 1942) to recover them all, and impossible to be certain of having done so in every case. Moreover, unfertilized ova are particularly difficult to recover once they have traversed the Fallopian tube and have entered the uterus. Probably unfertilized ova are quickly eliminated from the uterine tract and, when all the ova are unfertilized, their passage through the Fallopian tubes may be accelerated. Consequently the proportion of such ova recovered is likely to be substantially less than the true proportion.

The errors introduced into the counts of corpora lutea, implantation sites and embryos through these causes are dealt with in this paper. A section is devoted also to showing that the loss of ova before implantation is unconnected with that occurring after implantation. It is convenient to introduce this here, since the subsequent statistical work, dealing with these two fractions of the total prenatal loss respectively, falls naturally into two parts.

The technique originally employed has been described in detail already (Brambell, 1942, 1944) and it is only necessary to mention here those respects in which it has been modified and improved. First, during the earlier work, the uteri of visibly pregnant animals were fixed whole and the embryos dissected out and examined subsequently. This method was laborious and was liable to result in distortion of small embryos by pressure resulting from contraction of the uterine wall during fixation. Consequently during 1943 and subsequent years the uteri of visibly pregnant animals, after removal, were pinned out in a dissecting dish, opened under saline, and examined under a binocular dissecting microscope before fixation. This procedure permitted of more critical examination of the embryos and embryonic membranes, with a consequent increase in precision in determining whether they were developing or were in process of reabsorption.

Secondly, during and after 1944, whole mounts were made of tubal and early uterine ova, recovered by transfusion, instead of serial sections. They were mounted direct from water in gum chloral containing borax-carmine. The gum chloral was made up on the usual formula:

One part of borax-carmine was added to four parts of gum-chloral. The stain was made up by dissolving 4 g. borax and 3 g. carmine in 100 c.c. of hot distilled water and filtering. Mounts made with this medium were permanent, and the ova were stained adequately for determining the stage of development after the mount had been in the slide-drying oven for a few hours.

Both these modifications of technique resulted in a substantial saving of time and enabled the total prenatal mortality in a sample to be determined within a few hours of having received it. Since both involved the use of fresh material, they did lengthen somewhat the routine to which a sample had to be subjected immediately it was received ; a serious consideration when the usual sample of thirty to forty pregnant females was received late in the evening, involving several hours’ work for two people.

The accumulation of data rendered it desirable to record them on perforated index-cards which could be sorted mechanically, so as to reduce the labour of analysis. The system adopted was the Cope-Chat Paramount sorting system, recommended to us by Dr C. Elton, of the Bureau of Animal Populations, Oxford, and it has proved entirely satisfactory. The card employed was designed in 1943, when the work had reached a sufficiently advanced stage to form a clear idea of what was required. The design, which is reproduced in Fig. 1, has proved efficient and may be of use to others. All the data have been entered on these cards, without which analysis of so much material would have been so laborious as to be almost impracticable.

The material consists of 7137 wild rabbits obtained between 7 February 1941 and 30 May 1945, all of which were examined, dissected and recorded. These form eleven series according to the year and locality. Males were only included in those series intended to provide information regarding the duration of the breeding season. One intersex and two hermaphrodites were obtained. The localities, dates and sex distribution of each series are given in Table 1.

Particulars of series 01, 0 and 1 have been published already. However, since the various series differ from each other in certain significant respects, such as the type of country from which the rabbit came, the methods by which they were obtained, the size and frequency of the samples aimed at, and the special treatments to which some were subjected as the result of experience or to throw light on some particular aspect of the problem, the characteristics of each series are set down here for comparison, since they affect the purpose for which the data can be used legitimately.

Series 01 (specimens labelled R1–355). Obtained on the Vaynol Estate, near Bangor, in north Caernarvonshire, consisting of a park, with woods, coverts and some arable land, near the sea. The majority were shot or caught by dogs, after flushing from cover or bolting with ferrets. They were received in good condition on the day they were killed or the following morning. A weekly sample of about forty rabbits of both sexes was aimed at but not always achieved.

Series 0 (specimens labelled R356–1697). Obtained around Brynkir and on the Lleyn Peninsula in south Caernarvonshire except for a single sample of twenty-one rabbits on 23 October from the Vaynol Estate in north Caernarvonshire. The country is hilly, intersected by rivers and near the sea. It consists of farmland, mainly lowland grass, with stretches of rough ground, covered with bracken and gorse, and some patches of woodland and of bog. The fences are mainly old rotten banks, often with hedges. Near the sea there are areas of dunes. The whole area is very heavily infested with rabbits. Practically all the material was trapped by the trappers of the War Agricultural Committee. Some of it was in excellent condition when examined, but some was not and a few carcasses were so bad when received that they were useless. Weekly samples of forty rabbits of both sexes were desired, but the numbers obtained often fell below the target.

Series 1 (specimens labelled R1698–2487). Obtained from the same locality and in the same way as series 0. Weekly samples of about forty rabbits of both sexes were continued until 3 February. Thereafter the collection of males was discontinued and a weekly sample of forty females was the target.

The uterine tracts of all females of series 01, 0 and 1 were preserved. The embryos in visible stages of pregnancy were not dissected out and examined until after fixation. In series 01, o and that part of series 1 collected during January and February, the mammary glands were classified simply according to whether or not they contained milk. Thereafter they were classified according to whether (a) they were definitely suckling, (b) they had milk in the glands but were not certainly suckling, or (c) they had no milk in the glands. Tubal and early uterine eggs were embedded and serially sectioned.

Series 2 (specimens labelled AR 1–208). Obtained in the northern part of the island of Anglesey, within a radius of 5 miles of Llanerchymedd. This is low-lying country consisting mainly of grassland, interspersed with stretches of bog and rough ground covered with bracken and gorse. Old rotten banks fence the small fields and the population of rabbits is heavy. All the rabbits were trapped and they were received in fairly good condition. The object was to obtain a small sample of female rabbits during the height of the breeding season to determine whether or not whole litters died and were reabsorbed after implantation. Weekly samples of up to forty rabbits were examined, until over 200 had been obtained. All the swellings in visible stages of pregnancy were opened before fixation under saline and full notes made of dead and reabsorbing embryos and, wherever possible, the order in which they had died according to their position in the uterus recorded. No attempt was made to obtain by transfusion tubal or early uterine ova and hence 0–6-day stages of pregnancy were not identified.

Series 3. Obtained in south Anglesey within a radius of 5 miles of Capel-Mawr, near Trefdraeth. This is low-lying farmland, mainly consisting of grass, with some arable and rough land. It is divided up into small fields by banks and hedges. It adjoins extensive areas of marsh to the east and of dunes to the south and carries a heavy population of rabbits. The majority of the rabbits were trapped, but some were netted, and some ferreted. They were received in very good condition, many of them being still warm and few more than 12 hr. dead. A weekly sample of forty females was the target. The full routine of examination, described at the beginning of this section, was employed for this sample and for the succeeding five series obtained in 1944. Non-pregnant uteri and those from which early stages had been recovered by transfusion were not preserved as a rule, nor were the uteri and embryos of many of the very late stages. This omission, which applies equally to series 4–8, was unfortunate, for it prevented estimates of the proportion of whole litters lost after implantation being made subsequently from the proportions of recently pregnant uteri in which reabsorption or parturition had occurred. Such uteri can be distinguished histologically. This series was obtained at the same time as series 4 and the treatment was similar, hence the two are strictly comparable.

Series 4. Obtained from the same areas of south Caernarvonshire and in the same way as series 0 and 1. A weekly sample of forty females over a period of a year was the target. The treatment was the same as in series 3.

Series 5 and 6. Obtained in Norfolk within a radius of 15 miles of Norwich. The area was mainly grass and arable lowland, with banks and hedges bordering the fields. Some of the material came from Foxley Wood, of some 500 acres extent. Most of the animals were trapped but many were ferreted and shot. None of the animals was more than 24 hr. dead when received, and the majority less than 12 hr., so their condition was good. The treatment was the same as in series 3 and 4. The two series differ only in that in series 5 it was intended to secure a random sample of approximately thirty adult females per week, whereas series 6 consisted only of females presumed to be pregnant because of the presence of recent corpora lutea in the ovaries. The aim was to secure a sample of approximately 500 females as quickly as possible during the height of the breeding season, to determine the incidence of prenatal mortality and to compare it with that experienced in North Wales.

Series 7 and 8. Obtained in Dumfriesshire and Kirkcudbrightshire within 15 miles to the south and to the west of the town of Dumfries respectively. The country is hilly, frequently rising above 600 ft., consisting mainly of grass and arable land, with hedges bordering the fields. The animals were mostly trapped, but a few were shot. All were received in very good condition within a few hours of death. The treatment and aim was the same as in series 5 and 6, and the target was reached in a shorter time. Series 8, like series 6, consisted of animals selected because they were believed to be pregnant.

Series 9. Obtained in south Caernarvonshire from the same area and in the same way as series 0, 1 and 4. This sample was designed to supplement the information already obtained regarding the breeding season in Caernarvonshire, to provide material for embryological investigation, and to elucidate certain points regarding the incidence of prenatal mortality. Since it was necessary to keep some of the swellings in each visible pregnancy intact for subsequent sectioning, it was not possible to determine by dissection the full number of embryos that were dead and reabsorbing. A special point was made of preserving all non-pregnant and early pregnant uteri, so as to determine histologically which were post-partum and which post-reabsorption. Great care was taken also to recover, as far as possible, all tubal and uterine ova, so as to estimate the proportion of ova lost through not being fertilized.

The corpora lutea in rabbits attain a size of 2 · 5–3 ·0 mm. in diameter during pregnancy. They are then comparatively large and conspicuous structures which can be easily identified and counted with the naked eye in fresh ovaries. They develop slowly and only attain their maximum size at mid-pregnancy, as can be seen from the growth curve given in Fig. 2 and the data of the mean size of the corpora lutea in each of seventy-four pairs of ovaries given in Table 2. Thereafter they do not decrease in size appreciably until the last 2 or 3 days of gestation, and they are still easily visible at the time of parturition. They retrogress rapidly after parturition, or during the later stages of the reabsorption of a litter that has died in utero, and then they soon cease to be countable. Even the newly ruptured follicles and the corpora lutea in early stages of development appear to be easy to count macroscopically, especially as the rupture point is marked by a small pimple on the surface of the ovary at this time. In practice difficulty is experienced occasionally when the rupture point of a young corpus luteum is close to the mesovarium, where it may be overlooked unless the Fallopian tube is completely reflected so as to expose the whole surface of the ovary or when two fully formed corpora lutea have developed so close together as to appear to be fused. Sometimes the corpora lutea are difficult to count in the ovaries of a young animal which has recently attained puberty and ovulated for the first time, probably because the ovaries are so translucent at this time that degenerating follicles within the ovaries appear as opaque masses which can be confused with developing corpora lutea. Sometimes one or more abnormally small corpora lutea are present in a pair of ovaries together with several larger ones which are all subequal in size and which they resemble in colour and texture. It is then difficult to decide whether these small ones are corpora lutea vera, which should be counted with the others, or whether they are corpora lutea atretica formed from unruptured and immature follicles, which should not be counted. The decision in such cases is necessarily somewhat arbitrary, unless the rupture points happen to be clearly visible.

The practice adopted was to count carefully the corpora lutea in the ovaries first, then to count the uterine swellings, if these had developed, or to dissect out the tract and transfuse out early stages. If the number of embryos in either tube or uterus exceeded the number of corpora lutea in the corresponding ovary, the luteal count was very carefully checked. If, as was common, the discrepancy still could not be resolved, the ovaries were fixed, subsequently cut with a safety-razor blade into freehand sections about 1 mm. thick, and the corpora lutea counted under the dissecting microscope. All corpora lutea more than a day old could be identified readily by this method, so that it was found to be unnecessary to imbed and make microtome sections of the ovaries. Some of the discrepancies were resolved by this means but others were found to be real.

Although the macroscopic counts of corpora lutea made at the time of dissection appear to the observer to admit of a high degree of precision, notwithstanding the occasional difficulties experienced, which are enumerated above, it is desirable to know the magnitude of the experimental error. This is necessary if reliance is to be placed on estimates of prenatal mortality based on counts of corpora lutea. If it is supposed that the precision of identification of corpora lutea depends on their size then it follows that the risk of omitting some from the counts would be greater before they have attained their full development and after they have begun to regress. To test this hypothesis all the data of the size distribution of the sets of corpora lutea in pregnant animals, grouped according to the stage of pregnancy, have been examined and are given in Table 3, together with the mean size of litter and its standard error in each group. The differences of the means and the standard errors of the differences are:

The first and the last of these are significant, the difference being more than twice its standard error in each case, while the second is scarcely significant. The maximum difference, between the means of groups 16–20 and 26–32, is 0·405 ±0·093, and is thus more than four times its standard error. Regarding Table 3, it is difficult to account for the mean number of corpora lutea in a set being less at the beginning and end of gestation than in mid-pregnancy other than by errors of omission in the counts. It will be seen that the maximum difference between the means represents a discrepancy of 6·8% of corpora lutea counted. Since the error must be due in the main to the omission of corpora lutea from the counts, rather than to overcounts, it follows that estimates of the prenatal loss will be underestimates by the amount of the error.

The macroscopic counts of 180 pairs of ovaries were checked by microscopic counts made from freehand sections, to determine directly the error in the original counts. The ovaries in which the counts were to be checked were all derived from the Caernarvonshire material and the majority from series 01, 0 and r. They were selected only as regards age, so that thirty pairs of ovaries were checked from each age group. Ovaries in which the counts had been checked previously because the number of embryos in a uterus appeared to exceed the number of corpora lutea in the corresponding ovary were excluded. The results are given in Table 4, in which the divergence of the original macroscopic count from the microscopic check count is shown. The microscopic counts were made with great care and there was little probability of any corpora lutea having been overlooked. Sometimes an arbitrary decision had to be made, as in the macroscopic counts, as to whether a corpus luteum should be included in the count or classed as a corpus luteum atreticum and excluded ; in most cases the corpora lutea atretica could be readily identified by their much smaller size and by the frequent presence of a central cavity, sometimes containing a blood clot. It will be seen that the original counts diverged from the check counts by 6·4% of corpora lutea, affecting 26 · 6% of litters. Instances of the original counts exceeding the check counts were much fewer than errors of omission in the original counts and were confined to the first half of the gestation period. The net error was a deficiency of 4 · 5 % of corpora lutea in the original counts which is of the order required to account for the differences between the mean numbers of corpora lutea in a set in each age group, as shown in Table 3. This error, however, was distributed throughout all the age groups and not confined to the beginning and end of gestation, as would be required to compensate for the smaller mean numbers of corpora lutea at these periods as given in Table 3, but the sample of thirty pairs of ovaries in each age group checked is too small to attach statistical significance to the age distribution.

It is well known that corpora lutea atretica are formed from unruptured follicles during pregnancy in the ovaries of several mammals, including the baboon (Zuckerman & Parkes, 1932), the bank-vole (Brambell & Rowlands, 1936) and the mare (Cole, Howell & Hart, 1931). (For full discussion of the relevant literature see Brambell, in press.) These are smaller than the functional corpora lutea, they have no rupture points, since they are formed from unruptured follicles, and remains of the follicular antrum often persist as a cavity in the centre. Abnormally small corpora lutea, clearly differing in size from the other corpora lutea of pregnancy in the same pair of ovaries, were observed in some of the rabbits and were taken to be corpora lutea atretica. They were found to be present in twenty-nine of the 180 pairs of ovaries examined microscopically. There were thirty-three of these small corpora lutea, two being present in each of four pairs of ovaries, thus amounting to 1 · 7% of all corpora lutea in the 180 pairs of ovaries examined. They were at least as numerous in the first half, as they were during the second half, of pregnancy, and their histological appearance closely resembled that of the functional corpora lutea of pregnancy. These facts suggest that they are formed only at the beginning of pregnancy at the same time as the normal corpora lutea. Many had a central cavity containing some blood, but the thickened luteanized walls distinguished them from blood follicles. Some were situated so deep in the cortex that they must have originated from follicles which could not have ruptured. In most cases the number of embryos present was equal to or less than the number of normal-sized corpora lutea, but in four instances the number of embryos was greater, suggesting that in these at least the small corpora lutea had been formed from follicles that had ruptured and liberated functional ova. The majority of these small corpora lutea fell within the size limits of 1 · 4–2 · 0 mm. in diameter and therefore corresponded in size with mature follicles about to ovulate.

It may be concluded from these observations that corpora lutea atretica are formed in the wild rabbit at the time of ovulation from not more than 2% of the mature follicles, which have failed to rupture, and that none are formed subsequently during pregnancy. These corpora lutea can be distinguished as a rule by their small size, although, very occasionally, such abnormally small corpora lutea may be formed from ruptured follicles.

It has been shown by Heape (1905) that maturing follicles which undergo degeneration frequently give rise to blood follicles through rupture of the thecal blood vessels and extravasation of blood into the antrum. Only maturing follicles appear to degenerate in this way. The maturing follicles in oestrous rabbits that are prevented from mating, and which therefore do not undergo the final rapid growth that occurs in the 12 hr. interval between copulation and ovulation, degenerate after a few days and are replaced by others, as has been shown by Hill & White (1933) and confirmed by Smelser, Walton & Whetham (1934). Many of these, according to Heape (1905), Hammond (1925) and Hill & White (1933), give rise to blood follicles, which are relatively persistent structures that are gradually absorbed in a similar manner to a blood blister in the skin. Consequently such blood follicles are present as a rule in the ovaries of unmated oestrous rabbits. The corpora lutea atretica referred to above, appear to arise, on the other hand, from maturing follicles which have undergone the final growth that occurs after copulation but which have failed to ovulate. They have then undergone the luteal transformation at the same time as the ruptured follicles of the same generation, but their affinity to blood follicles is shown frequently by the presence of some blood in the persistent remains of the undischarged antrum. True blood follicles do occur in the ovaries of wild rabbits, but they are not nearly so common as in tame rabbits, presumably because in the wild state rabbits copulate very soon after they come into oestrus. This suggests that blood follicles arise mainly, if not exclusively, from maturing follicles which retrogress before the time of ovulation.

It follows that the percentage of litters in which the apparent loss has been reduced by polyovuly is 1 · 32–0 · 83 = 0 · 49. Thus the error introduced into the estimates of prenatal mortality from this cause is small and for most purposes may be neglected.

The eighteen litters showing polyovuly included sixteen showing an excess of one each, and two showing an excess of two each. Four other Etters were observed in which the number of unimplanted embryos recovered by perfusion of the Fallopian tubes or uteri exceeded the number of corpora lutea in the ovaries by one each, but these were not included in the calculation, which was confined to Etters with implanted embryos, because the experimental error introduced by failure to recover all the embryos by perfusion in stages less than 7 days’ pregnant is too great. Since the mean size of Etter in the sample of 2179 Etters available was 5 · 76 and polyovuly occurs in 1 · 32% of the Etters, it must result from 0 · 23% of folEcles which ovulate.

The passage of ova liberated from one ovary to the opposite uterus in wild rabbits was noted previously (Brambell, 1944). Since the two uterine horns have separate cervical canals transmigration of ova cannot take place by way of the cervix in rabbits. Therefore any that occurs must be by way of the peritoneal cavity. There were, in all, fifty-five litters in which the number of embryos on one side exceeded the number of corpora lutea in the corresponding ovary, though the total number of embryos on the two sides did not exceed the total number of corpora lutea in the pair of ovaries. Of these, eleven were in animals less than 7 days’ pregnant, and forty-four were in animals at least 7 days’ pregnant. The latter were derived from a sample of 2179 litters with implanted embryos, of which they represent 2 · 02%. In these there was evidence of the migration of two ova in each of two animals only, and of one ovum in each of the remaining forty-two animals, or 0 · 37 % of ova ovulated.

The migration was from right to left in twenty-eight instances and from left to right in twenty-seven, so that the probabilities of migration occurring in either direction appear to be equal.

The only alternative explanation to transmigration that can be offered to explain these data would be the occurrence of polyovuly on one side accompanied by preimplantation loss on the other. Since it has been calculated that polyovuly accompanied by pre-implantation loss occurs in only 0 · 49 % of litters, and since the gain and loss would fall on opposite sides respectively in half of the cases, this explanation would account for less than one-eighth of the number observed.

Implantation of the blastocyst of the rabbit is of the central type and attachment to the uterine was takes place early on the seventh day of gestation. Thereafter the blastocyst expands rapidly, and the swellings on the anti-mesometrial side of the uterus that mark the implantation sites can be clearly distinguished macroscopicaUy by the middle of the seventh day. They are large and conspicuous by the middle of the eighth day and increase in size continuously thereafter, as development proceeds, until parturition. There is, therefore, no possibility of omitting to count swefiings that contain living embryos after the seventh day. Moreover no precise information is available regarding the persistence during pregnancy of swellings in which the embryos have died. Such swellings certainly begin to regress at once and, equally certainly, such regressing swellings can be distinguished amongst those containing developing embryos in a proportion of uteri at all stages after implantation, including those approaching full-term. The problem is whether a swelling in which the embryo has died soon after implantation can regress sufficiently to become indistinguishable before parturition. Experiments are being carried out to determine the rate of regression when all the embryos in utero are killed simultaneously at a given stage of development, and the results will be published elsewhere shortly. These experiments show that the swellings that contained embryos killed at and days are still visible at least at and days respectively. It is probable that the rate of regression is more rapid when the whole litter is destroyed, for then the animal comes into oestrus again soon, even before the swellings have disappeared, than it is when only some of the embryos are affected and pregnancy is maintained. It is hoped to investigate experimentally this problem at a later date. Meanwhile the possibility remains that if some embryos die sufficiently soon after implantation their placental sites may disappear completely before the remaining embryos are born. If placental sites do disappear during gestation then the mean number of placental sites observed in the uteri at successive stages of pregnancy should decline, since no more can be formed. All the data of the distribution of the number of placental sites, both those containing living embryos and those regressing, in litters, grouped in four successive stages, are given in Table 5, together with the mean size of litter and its standard error in each group. The differences of the means of successive groups and the standard errors of the differences are given also. It will be seen that the differences between the 11–15- and the 16–20-day groups, and between the 21–25- and 26–32-day groups are significant, being more than twice and thrice their standard errors respectively. Thus there is a significant falling-off in the mean size of litter after the maximum in the 16–20-day group which can be accounted for by assuming the disappearance of some reabsorption sites. The significance of the initial rise in the mean size of litter from the 11–15- to the 16–20-day group is very difficult to understand, and no satisfactory suggestion as to its meaning can be offered. Comparison of Table 5 with Table 2 shows that the successive changes in the mean numbers of corpora lutea and of placental sites during pregnancy closely parallel each other. This is brought out more clearly in Fig. 3. Assuming that this is due to a similar rate of disappearance, then the decline in the number of placental sites will tend to compensate the decline in the number of corpora lutea when both are used to estimate the prenatal loss. Such estimates, based on the difference between the numbers of corpora lutea and of placental sites, irrespective of whether the latter contain developing or regressing embryos, are estimates of the loss before implantation. Therefore estimates of the loss suffered before implantation in the surviving litters will tend to be unaffected by the disappearance of corpora lutea and of placental sites. Estimates of the loss suffered after implantation are based on the difference between the number of developing embryos and of placental sites, and estimates of the total loss during pregnancy on the difference between the number of developing embryos and of corpora lutea. Since, by definition, developing embryos cannot disappear until parturition, both these estimates will be affected by the disappearance of placental sites and of corpora lutea respectively, being underestimates by the amount of this.

It has been indicated that the prenatal loss suffered before implantation can be distinguished from that suffered after implantation in any visibly pregnant animal by comparison of the number of placental sites with the number of corpora lutea and of developing embryos respectively. The problem of whether the loss suffered after implantation is connected with that suffered before implantation, or is independent of it can be tested. It might be supposed that some or all of the factors which contribute to loss before implantation would continue to operate after implantation and to occasion further loss then ; in other words, that an animal which has suffered loss before implantation would be predisposed to suffer further loss after implantation. If this were so, a greater proportion of those animals which had suffered loss, than of those which had suffered no loss before implantation, would suffer loss after implantation. If this were not so, the litters that suffered loss after implantation would be distributed between those that had suffered loss before implantation, and those that had not, in proportion to the relative numbers in these two groups. The available data from the various series are given in Table 6. The expected numbers showing loss before and after implantation, calculated on the assumption that the loss after implantation is independent of that before implantation, is given in the right-hand column. The observed values diverge from the expected values for the whole sample by only 6 · 5. Testing this, x²= 0 ·48,,n = 1 and P = 0· 5. Therefore it must be assumed that the loss after implantation is independent of that before implantation. The expectation expressed as a percentage is represented graphically in Fig. 4 in the form of a square 10 × 10, and the deviation of the observed from the expected values is shown. Further, it can be seen from Table 6 that the observed do not diverge significantly from the expected values in any series, taken alone, the values of P ranging from 0·2 to 0·95.

The independence of the loss before and after implantation has been tested also for each age group from implantation to full-term. The results are shown in Table 7, and the observed do not differ significantly from the expected values in any group, P = 0·2–0·8. There are therefore no grounds for supposing that any fraction of the loss after implantation occurring at a particular stage of development is connected with the pre-implantation loss.

Finally, the distribution of those litters in which all the implanted embryos had died and were in process of regression in relation to loss before implantation was examined. The results given in Table 8 show that the loss of whole litters after implantation is also independent of the loss before implantation, P=0·7–0·8.

Thus no relation has been discovered between the loss before and after implantation respectively, and it is legitimate, therefore, to treat them as independent and to examine them separately. This separation of the prenatal loss into two independent fractions had not been recognized when the earlier reports (Brambell, 1942, 1944) which dealt only with the total loss up to the time of examination, were published. The separation of these two fractions has thrown light on several problems that were previously obscure.

1. The technical problems underlying the estimation and analysis of the total prenatal mortality in a mammal are reviewed, with particular reference to the rabbit.

2. The material consisted of 7137 wild rabbits, of which 5089 were females. These comprise eleven series, according to the year, locality and treatment, particulars of which are given.

3. The experimental error in counts of corpora lutea was investigated and the technique critically examined. The changes in size of the corpus luteum during pregnancy were measured and a growth curve constructed. The mean number of corpora lutea in a litter, as determined from macroscopic counts, was found to be significantly lower at the beginning and end of pregnancy than in the middle. If this is due to omissions in the counts it represents a maximum error of 6·8% of corpora lutea. Microscopic counts of the corpora lutea in freehand sections of 180 pairs of ovaries of pregnant animals were made as a control. These revealed a total error in the original counts of 6·4% of the corpora lutea, affecting 25·6% of the litters, but this was distributed evenly throughout gestation.

4. Corpora lutea atretica formed from unruptured follicles are present in 16% of the pairs of ovaries, and comprise less than 2% of all corpora lutea. They are formed at the same time as the normal corpora lutea, and there is no evidence that any are formed subsequently during pregnancy in the rabbit.

5. It is estimated that 0·23 % of the follicles which ovulate produce two embryos through the liberation of two ova.

6. Transperitoneal migration of 0·37% of ova, affecting 2% of the litters, was found to occur. The probability of migration either from left to right or from right to left appeared to be equal.

7. Significant changes in the mean number of implantation sites in the uteri counted at successive stages of pregnancy were observed, and particularly a decline at the end of gestation. They are of the same order as those in the mean number of corpora lutea, which they will tend to compensate so far as estimates of loss of ova before implantation are concerned. It is by no means clear that these are due to experimental error in the counts.

8. Estimates of the loss of embryos before implantation are based on the difference between the number of corpora lutea in the ovaries and of implantation sites in the uteri, and estimates of the loss after implantation on the difference between the number of implantation sites and of developing embryos. No significant relation has been found between the loss before and after implantation respectively, and it is concluded that the loss suffered after implantation is distributed independently of that which occurs before implantation.

The work forms part of a larger scheme of research on prenatal mortality being conducted in this department and financed by a grant from the Agricultural Research Council, which we have pleasure in acknowledging.

We are much indebted to Prof. D. M. S. Watson, F.R.S., for encouragement and advice during the course of the work, to Dr A. S. Parkes, F.R.S., Mr Charles Elton, F.R.S., and Mr G. P. Wells for their interest in the work and for advice on certain aspects of it.

We have pleasure in acknowledging the kind offices of Sir Michael Duff, Bart., and of the Caernarvonshire and Norfolk War Agricultural Executive Committees in facilitating the supply of material. We are indebted to the Norfolk War Agricultural Executive Committee and to the Executive Officer, Mr J. C. Mann, to Mr S. Culpin and his staff for providing laboratory accommodation for the work at Norfolk, and to the Animal Health Division of the Department of Agriculture for Scotland, and the Divisional Inspector, Mr G. A. Sangster, M.R.C.V.S., for laboratory accommodation for the work at Dumfries and for much other assistance. We have pleasure also in acknowledging the ready co-operation in providing material of Mr Gwilym Roberts and Mr Charles Ffoulkes, Pest Officers Caernarvonshire W.A.E.C., Mrs Matthews, Pest Officer Anglesey W.A.E.C., Captain Palmer, Pest Officer Norfolk W.A.E.C., Mr Munn, Norfolk W.A.E.C., Mr S. Isbister, Department of Agriculture for Scotland, Mr Charles Parker, Head Keeper, Vaynol. We would also like to acknowledge the assistance of the trappers, too numerous to mention, who obtained the material and took a keen interest in maintaining suitable and adequate supplies.

Finally, the work owes much to the skilled technical assistance of Mr R. A. Lansdowne and also to the assistance of Mr W. Holland, both of whom we wish to thank.