School of Information Management & Systems
Previously School of Library & Information Studies

Michael Buckland, Professor.

Emanuel Goldberg and his Statistical Machine, 1927

Abstract   Emanuel Goldberg's "Statistical machine", patent applied for in 1927, appears to the have been the first functioning document retrieval system to use electronics. Priority is for this technology has been wrongly ascribed to Vannevar Bush.

Note   The following first appeared as M. Buckland. "Zeiss Ikon's "Statistical Machine" in Zeiss Historica Journal vol. 17, no. 1 (Spring 1995): 6-7. For a more detailed account see M. Buckland, Emanuel Godleberg and his Knowledge Machine (Libraries Unlimited, 2006).

Zeiss Ikon pioneered in the automation of the storage and retrieval of business records. In 1931, well before digital computers, Emanuel Goldberg of Zeiss Ikon was demonstrating what may have been the first document retrieval system using electronics.

Emanuel Goldberg was another example of academic talent recruited by Zeiss. Born in Moscow in 1881, he went to Germany for graduate study and stayed. The versatile Goldberg had made significant contributions to galvanizing, photochemistry, photography, and color printing by 1917 when he left his professorship in Leipzig to join Zeiss. Initially a technical advisor at Jena, he became director of research at Ica, the Zeiss subsidiary in Dresden. In 1926 when some Zeiss subsidiaries were merged to form Zeiss Ikon, Goldberg was appointed the principal director.

In the 1920s the practice of microfilming business records was becoming common. Banks, in particular, found that they could reduce fraud by microfilming the checks they handled before returning the cancelled originals to customers. From his extensive work on movie camera and projector design, Goldberg was familiar with the engineering problems of handling 35 mm film. The problem with microfilmed business records was in finding and retrieving particular records promptly when needed from long spools of film storing images of thousands of records in no particular order. One could create indexes to the microfilmed images, but that was tedious and only yielded an address for the record, not the record itself.

Goldberg was interested in electronics. He became actively involved in the development of television technology and, as a hobby, he enjoyed rebuilding radios. In particular, photoelectric cells were central to sound movie technology, a development of great importance for Zeiss Ikon.

To solve the problem of retrieving individual records from spools of microfilm, Goldberg used movie projector technology handle the microfilm and a photoelectric cell to do pattern recognition in finding the right record. The principal application was expected to be retrieving accounting and sales data, so the new device was called a "Statistical Machine".

The central idea is shown in Figure 1. When the documents are microfilmed, they are also indexed. Whatever feature is likely to be used for retrieval (e.g. amount of check, account number, sales area) is represented by a code. One could use letters or numbers, but patterns of opaque dots were simpler to implement. The index code(s) would be photographed alongside each document, either to one side of the images (like a movie sound track) or underneath, as in Figure 1.

In Goldberg's basic design a "search card" is created and placed between a light source and the film. The search card blocks all light from the light source except for a pattern of very small beams defining the code that is to be sought. Beyond the film is a photocell. As the film containing images of documents moves through the machine, some of the light that passes through the search card will pass through the film and reach the photocell where it generates a low voltage electrical current. But when opaque dots on the film coincide exactly with the pattern of light beams defined by the search card, all light is blocked and no light reaches the photocell. When no light reaches the photocell, the flow of low voltage electricity falters. Circuitry detects the loss of current and signals that the desired document has been found.

A modified movie projector was used, with suitable lenses to focus the beams of light on to the film, as shown in the lower half of Figure 2. Another mechanism was needed to project the image of the selected document for viewing or copying as shown in the upper half of Figure 2.

In August 1931 the 8th International Congress of Photography was held in Dresden and Goldberg described and demonstrated the Statistical Machine at one of the technical sessions. It might have received more attention if the delegates had not been preoccupied by Goldberg's presentation of the proposal, eventually adopted, for an international standard for film speeds and, indeed, by his widely reported demonstration of sound movie technology. In October 1931, Goldberg was the Taylor Lecturer at the Royal Photographic Society in London and demonstrated the Statistical Machine as part of his lecture, receiving "prolonged acclamation".

Goldberg himself clearly attached great significance to the Statistical Machine. The German patent (No. 670,190, issued Dec. 22, 1938) was issued jointly to Zeiss Ikon and himself. He is said to have negotiated an agreement with Zeiss Ikon that he would receive a share of any royalties from it.

Two prototypes are said to have been built and he described a variant design in which one could dial the code number, like a dialling a telephone number, instead of having to make a search card. But in 1933, Nazis kidnapped Goldberg from the Zeiss Ikon offices. He was released, but he had to leave Germany.

In the USA, IBM promptly acquired rights to the U.S. patent (No. 1,838,389, issued Dec. 29, 1931). Separately, at MIT, a similar device was developed by Vannevar Bush in the late 1930s with support from Eastman Kodak and from National Cash Register. It was called the Microfilm Rapid Selector and was to be rapid because the film moved continuously without a movie gate. Bush's patent application was denied because the patent examiner identified the Zeiss Ikon Statistical Machine as "prior art". Bush also supervised the development, in great secrecy, of a variant version—the Comparator—to help Navy cryptanalysts to break enemy codes. It was designed to locate occurences of particular characters in an encrypted message.

After World War II, the MIT-designed Microfilm Rapid Selector was revived by Engineering Research Associates. The development of retrieval machines combining microfilm and photocell continued, but with little success, into the early 1960s when they were overtaken by digital computers.

The Statistical Machine was not developed into a Zeiss Ikon product and few traces of it seem to remain. It was, however, a pioneering achievement and, apparently, the first of its kind.

Further reading.
M. Buckland, Emanuel Goldberg and his Knowledge Machine (Libraries Unlimited, 2006). The May 1992 issue of the Journal of the American Society for Information Science (vol. 43, no. 4) contains an article on the Zeiss Ikon statistical machine (pp. 284-294) and also a translation of Goldberg's original description of it at the 1931 congress in Dresden (pp. 295-298).
For an excellent, detailed account of the similar machinery developed in the USA for cryptanalysis and for document retrieval see Colin Burke Information and secrecy: Vannevar Bush, ULTRA, and the other Memex (Scarecrow Press, 1994).
Go to Goldberg,   History,   Concepts,   Information Selection, or Michael Buckland's Home page