Much has been said and written about infant mortality and particularly about its decline. Even statesmen have cited its decrease as suggestive evidence of the benefits which arise from their stewardship. They are, it would seem, oblivious of the fact that the decline is not of recent occurrence but one which has been operating over a period of many years. The facts are easy to appreciate. Fifty years ago the mortality in infancy in England and Wales exceeded 150 per 1,000 births; in 1948 it was 34, an impressive reduction which was particularly observable in the age group 3-12 months. The reduction in the first four weeks of life has been less satisfactory. It may well be asked whether each contributory cause of death in early infancy has declined equally or whether some have declined more than others. If the average annual change in the mortality under 4 weeks during the period 1931-38 is expressed as a percentage of the mean death rate in that period in England and Wales, the statistical picture for certain broad categories is as in Table I. It will be noted that in four instances the mortality has declined, the annual decrement being largest (9-1 per cent.) for convulsions. The death rate in the two remaining groups, congenital malformations and injury at birth, increased, and the increment for the latter was the greater as the values were 1 1 per cent, and 2-6 per cent, respectively. It should be stated, however, that there is a school of expert medical opinion which regards such observed increments as fictitious. Although it recognizes the total neonatal mortality as a reliable figure, it suggests that the constituent causes defy statistical analysis owing to the inability of doctors in general to state, with any degree of precision, the exact cause of death at this period of life. The difficulties of accurate certification of causes of death in the first four weeks of life are well recognized, but it would se m unlikely that medical men are in the habit of exa gerating birth injury as a cause. Indeed, the joi t report of the Royal College of Obstetricians and Gynaecologists and the British Paediatric Association (1949) has suggested a much higher rate of mortality from birth injury than is revealed by the Registrar General's tabulations. It is not the province of a statistician to question the v lidity of medical diagnosis. In the absence of a sudden or startling departure from the normal mortality (apart from epidemic experience), he accepts the declared findings as facts although he is, at the same time, fully conscious of the vagaries of medical opinion and the influence of new preci sions on diagnosis. We have been told, however, tha the increments which have been demonstrated in the mortality from injury at birth may possibly be due to a transfer of deaths from asphyxia, atelec tasis, or convulsions. The death rate from con vulsions has declined, and convulsions in children under four weeks are believed, in many instances, to be the aftermath of an injury during delivery. As a t st we calculated the mortality of a combined group of causes of death, namely convulsions, injury at birth, asphyxia, and atelectasis, over a