In 1950 the Bank of America asked SRI to assess the possibility of
developing electronic computers that could take over the labor-intensive
banking tasks of handling checks and balancing accounts. The creation of
branch offices and the rapidly increasing number of checks being used by a
growing clientele threatened to overwhelm the existing manual processing
and record keeping. At that time no large-scale electronic machine for any
bank was under development. Existing computers were used mostly for
scientific calculations. They were unreliable, and had extremely limited
input and output capability. In spite of these facts, SRI's feasibility
study, issued in May 1951, was sufficiently encouraging for the Bank of
America to authorize a major multi-year development effort.
We now take for granted the many ways that computers assist individuals
and businesses. The forty-year-old project briefly described here,
provided a vision of what business could expect from the application of
data-processing machines, and illustrates how and why some of the key
capabilities were invented, including bookkeeping, optical character
recognition (OCR or scanning), and robotic document sorting. The automated
teller machine (ATM) is the natural descendant of this work, and
illustrates the progression away from paper checks toward all electronic
The Banking Problems
The problem posed by the bank for solution by machine included all
accounting that normally attends many thousand commercial checking
accounts of a bank. Such a machine must be able to keep record of deposits
and withdrawals for each client, make current-balance information
available at an instant's notice, watch for overdrafts, stop payments, and
held funds. It must be able to provide, on a strict schedule, periodic
statements of the account along with the accumulated checks. The machine
must not only handle all necessary arithmetic but also handle the paper
documents in whatever physical condition they exist after passage through
many hands. All machine operations must, furthermore, be as exact as
banking accounting calls for and be in constant step with hourly, daily,
and monthly routines of the banking system.
Vacuum Tubes vs. Transistors
In the early 1950s transistors were beginning to replace vacuum tubes in
electronics applications. The urge to use transistors to achieve a design
requiring less power, generating less heat, and with the obvious advantage
of smaller volume, was nearly irresistible. In fact, only after parallel
developments using transistors and tubes had been carried for several
months did circuit designers reluctantly admit that the state of the art
in production of transistors left little hope that a sufficient number of
reliable units would be available as needed. Transistors would have to
wait until the commercial unit, called Mark II.
MICR -- Magnetic Ink Character Reading
The most difficult challenge to be solved was enabling the machine to read
the necessary information from checks, deposit slips, and other routine
documents. Techniques involving photo-electric scanning of characters and
codes printed in a variety of colors and fluorescent inks were examined.
All displayed a common fault. Such schemes were successful when tested in
the laboratory under controlled conditions. However, when required to scan
material on which the characters were overprinted with cancellation
stamps, endorsements, and the like, the error and reject rates rose to
Breakthrough on this problem came with the development of techniques for
reading magnetically characters printed in a black ink containing
particles of a magnetizable oxide. Because the reading element is
sensitive only the magnetized ink, subsequent overprinting or visual
obliteration has no effect on the machine's ability to read. The magnetic
technique has the additional advantage that if two checks go through
together--for example, if they are stapled, the magnetic head, by reading
through the top check, senses the second check. The machine thus rejects
such a "double."
The techniques of machine reading of characters printed in magnetic
ink--both in a code of bars and Arabic numbers--were demonstrated in July
1956, to the Bank Management Committee of the American Bankers'
Association. The recommendation of that committee led to the adoption of
the principle of magnetic-ink character reading or MICR as a standard for
all member banks. You can find the strange looking MICR numbers at the
bottom of your checks today.
Handling Paper at High Speed
Another challenge was the necessity of physically handling thousands of
pieces of paper daily in normal routine work. Any real solution to the
general problem of better handling of the Bank's commercial checking
accounts would have to begin with the (then) "impossible"
assignment of providing fully automatic equipment than can pick up,
transport, read, and sort a wide variety of sizes and thicknesses of paper
checks at high speed. Nor could it choke on checks creased, torn, or
Development of apparatus for handling paper accurately and at high speed
proceeded in parallel with that for character reading. An
electro-pneumatic machine was designed, built, and tested. Stacks of
checks were fed to the machine, which removed one check at a time, read a
number, and transported it to the appropriate bin according to the digit
read. Very reliable sorting speeds of 600 per minute were achieved and
laboratory devices handled checks at rates of over 3000 per minute.
ERMA Makes its Debut
In September 1955, SRI gave a public and press demonstration of the
prototype electronic accounting machine, by then designated ERMA (for
Electronic Recording Method of Accounting). Its performance proved the
soundness of the concepts and workability of the electrical and mechanical
The true test of ERMA began in the Fall of 1956. This was to process the
accounts of a branch bank in the same manner--and the same pace--that
would be required if it were in use as a central accounting facility
serving that branch and others. These day-by-day three-month-long tests
proved ERMA's ability to perform all accounting routines, and in
synchronism with the records kept by the bank.
ERMA is Commercialized
In April 1956, the Bank of America announced that General Electric
Corporation had been selected to manufacture production models. ERMA Mark
II was designed around solid-state logic elements (i.e. transistors) and
magnetic core memory. Numeric data input was read automatically from the
original documents using the MICR method. SRI contributed to General
Electric's development effort with consultation on character reading and
paper-handling techniques and assistance with the detailed programming of
the operational steps to be followed by the new equipment.
In 1959, General Electric delivered the first 32 ERMA computing systems to
the Bank of America. ERMA served as the Bank's accounting computer and
check handling system until 1970.
Console of General Electric's commercial ERMA.