The first Journal of Documentation paper to include the word `computer' in its title did not appear until 1964 [36]; in the previous year, a paper described some experiments with a computer program [32]. The first description of an operational system appeared in 1966 [42]. However, readers would have been aware much earlier, from a number of mentions in book reviews and reports of meetings, of the approaching transition from punched cards and other forms of mechanization to computers, despite some apparent scepticism. In 1948, Fairthorne reviewed a book by Vannevar Bush (containing the famous essay, `As we may think'); the review was affectionate but quite critical of the essay [3]. In the same year, the perception of change was reinforced by the Royal Society Scientific Information Conference, whose proceedings were reviewed in 1949 by Lancaster-Jones [4]; she quotes the

...desirability of a full-scale experiment in a scientific library with mechanical methods of indexing and selection.

Another article by Fairthorne discussed various devices that might be useful, including robots to carry books around, and OCR devices [9].

In 1957, Anthony reported [17] on (and reviewed [18] the proceedings of) a symposium on `Systems for Information Retrieval' at Western Reserve University. He made an interesting 3-way classification of systems into (a) manual, (b) machine, and (c) computer systems. Of the last category, he said:

All those demonstrated were experimental systems which are not likely to produce any practical result for 5-10 years...cost will be high...only national documentation centres...

A Perry and Kent book describing in detail the system at Western Reserve was reviewed by Claridge [23]. Hanson reviewed a National Science Foundation report on current research with the main emphasis on machine methods [22].

A reader of the reviews would also be aware that the idea of machine translation was beginning to emerge [13] [21].²

At the same time, some authors were (according to predisposition) either embracing the possibilities or worrying about whether computers really are going to help. Vickery reviewed another book by Perry and colleagues, which discussed at length their ideas on information retrieval in the new machine age [16]. On the other hand Farradane, in a review of a book by Taube [19], was critical of mechanistic approaches to retrieval; the following year (1959) he presented a paper at a conference (reported by Risk [20]), discussing the problem of appropriate logical structures for machine IR (in the subsequent discussion, Garfield defended machine methods). Farradane published a full paper in the Journal in 1961 [25].

In the same year, Fairthorne reviewed at length a report by Bar-Hillel, and was critical of Bar-Hillel's position that machines could do little for information retrieval [24]; in 1963 he was very critical of a book by Taube, which again suggested that the possibilities for making machines do anything very clever were limited [30].

On the other hand, Vickery reviewed (favourably) a book by Fairthorne [27], in which Fairthorne suggested that the benefits of work on automation might for some years come more from the ideas it generates (about the nature of the processes involved) than from actual, practical automation activities. Indeed, already in 1960, a book by Maron (reviewed by Vaswani [28]) suggested the use of statistical methods in indexing. Fairthorne [29] also reviewed a U.S. report describing mathematical research relating to information selection; the report was intended for mathematicians, but the reviewer attempted to interpret it for documentalists. He specifically picked out the idea of Boolean logic for searching; but added the rider that the mathematical papers tend to ignore the problem that `documentation is inherently imprecise'.