Many writers have expressed dissatisfaction with population censuses as a basis for research in the social sciences. The chief defects result from the varying size and form of the administrative units for which data are provided and from the frequent boundary changes of these areal units. Inadequate classification of the data, the delay between enumeration and publication, and the omission of certain topics of current importance also contribute to the general imperfection of the census for research purposes. Maximum flexibility both in presentation of the data and in the units of area to which they refer could be achieved by (a) allocating to every individual and every dwelling a unique locational reference, such as a twelve-figure National Grid reference; (b) storing the coded records of individuals and dwellings as a census data bank. For the whole population of Scotland these procedures would require the equivalent of twelve million standard punched cards, a perfectly manageable amount of material even with present storage techniques. Requests could be made from individuals or establishments for a specific set of data for any size or shape of area defined by National Grid co-ordinates. Since the records represent the first stage in processing the enumeration schedules, data for research could be obtained without delay while they are still representative of the current social and economic situation. Tabulations and summaries of proven worth could be prepared for administrators as a normal second-stage procedure. The attachment of a co-ordinate value to each data-file would make possible automatic data-processing with output both of numerical tables and graphical illustrations. When fears of the loss of confidentiality are overcome, the banking of data at individual level would pave the way for advances in the field of record linkage. IN THE PAST few decades increasing demands have been made on the population census (hereafter called the census) for research in the social sciences. It has been the unhappy experience of most investigators that, in order to reconcile unsuitable classifications and disparate areal units with the problem on hand, they have to adopt lengthy and often indirect procedures. In some cases research workers and planners have had to abandon the census in favour of other sources of demographic data, considerably more restricted in content but nevertheless more flexible in use, for example, electoral registers and private surveys. With the approach of the 1971 Census and with the recent revolution in methods of handling data, it is timely to consider previous defects and the means of their elimination. T. Hagerstrand, G. M. Stubbs and E. C. Willatts are among writers who have wrestled with the census and who criticize in particular the attachment of census data to an administrative framework.1 J. N. Jackson and C. M. Law discuss problems associated with classification and with definition, while A. J. Hunt, R. J. Johnston and H. R. Jones are among workers who have forsaken the census for electoral registers.2 Glasgow Corporation Planning Department (see Hunt and H. A. Moisley3) has also utilized electoral registers to obtain population figures for street blocks and has spent much time in surveys and in collecting data from other authorities, even though much of the required material on housing and sanitation is included in the census; for the census data cannot be related to precise and meaningful spatial units. Delay in publication of census tables and paucity of data, particularly on migration, are emphasized by Jones and in the report on Depopulation in Mid-Wales.4 Varied examples of the pitfalls to be avoided in population research are set out by G. C. Dickinson, who also reviews the general inadequacies of the census for thematic mapping.5 The personal experience of the author began with a study of the occupational structure of