Although it is easy to see much development of computer-based systems as technology driven, there is a strong sense in which the information world was ready and impatiently waiting for the computer. (This comment is not in any way negated by the remark above about the slowness of take-up -- what the information world needed was the practical devices for particular tasks, and it took computers a while to reach that status.) There are several pointers to this state of readiness, but I will emphasise just one: the idea of co-ordinate searching.

Post-co-ordination, or co-ordinate searching, was one of the great inventions of information retrieval. We now recognise the principle simply as Boolean AND, thereby placing its invention well outside the field of IR. But the principle was well established in IR long before it was equated with the Boolean operation, and indeed long before we had computers to do it on. In fact much of the pre-computer mechanization work was stimulated by the desire to allow post-coordination, together with the more mundane twin motivations of reducing clerical tasks and sources of error.¹

The most important predecessor technology (or group of technologies) was the punched card. The variety of punched-card methods was well represented in the Journal [1] [2] [6] [7]. Fairthorne [10] constructed an informational model of clerical systems based on punched card systems. Garfield [12] gave a practical account of the preparation of indexes using Hollerith cards. Ranganathan and Perry [8] included a discussion of punched cards in a paper on the principles of classification and coding for different purposes. A couple of reviews by Dyson attested to the rate of development of punched-card techniques [11]. In 1956, MacKay reviewed two Russian publications concerning a new system based on punched cards [14]. There was also a curious example of a paper which is pure probability theory, with no citations to any IR literature, about superimposed coding (which was used in some punched-card systems) [15].

There were, of course, several very different types of punched cards. Herman Hollerith's system was invented in the 1890s, for the analysis of census data (a major component of which was cross-tabulation, which is the statistical equivalent of post-coordination). Machine-sorted Hollerith-type cards were used for information retrieval by a number of large information departments, though they tended to be regarded as `elaborate and expensive' [1]. Other types of cards were optical coincidence or peek-a-boo cards (first invented in the 1910s, and the most obvious response to the need for post-coordination), and edge-notched (hand-sorted) cards. For some purposes in IR, punched paper tape was also useful (see e.g. Wilson [34]).

It is worth noting that the distinction between item cards such as edge-notched (one card per item) and feature cards such as peek-a-boo (one card per index term) transferred directly into computer-based IR systems as the distinction between serial searching and inverted files. Inverted files rapidly won out (so that, by the time Dolby discussed programming languages for IR systems in 1971 [65], he assumed inverted files), but the transfer of the discussion itself is another indication of the readiness for computing technology indicated above.

The last paper on punched cards to be published in the Journal did not appear until 1975 (Jolley [86]). Further, in 1983 Vickers [112] mentions a firm (medium sized, technology based, manufacturing) that is still using peek-a-boo cards.

Another class of mechanized system was microform-based systems. Although these have little connection in general with computer-based systems, Shaw's account of the Rapid Selector [5] included some comments which could transfer directly to computers, on the advantages of electronic selection (speed, space, multiple access points).