Roethlisberger and Dickson describe the first thirteen periods of the relay assembly test room in detail and the rest in summary; Stephen Jones’s 1992 table, built from the company’s records, gives all twenty-four. The two are combined here. Output figures are the weekly averages per operator that François Geoffroy tabulates from the 1939 book, rounded as he gives them.

The periods

• Period 1, April 25 to May 10, 1927. The five operators still in the regular department, their output taken from the payroll records without their knowledge. About 2,400 relays a week each.
• Period 2, May 10 to June 11. Moved into the test room on the afternoon of May 10. A 48-hour week: 7:30 to 12:00 and 12:45 to 5:00, Monday to Friday, and 7:30 to 12:00 on Saturday. No other change. About 2,400.
• Period 3, June 13 to August 6. Pay changed from a group piece rate shared among about a hundred assemblers to one shared among the five. About 2,500. On August 2 four of the five were called into the office for talking.
• Period 4, August 8 to September 10. Two rest pauses of five minutes. Output up.
• Period 5, September 12 to October 8. Two rest pauses of ten minutes. Output up sharply.
• Period 6, October 10 to November 5. Six rest pauses of five minutes. About 2,400.
• Period 7, November 7, 1927, to January 21, 1928. A fifteen-minute pause in the morning with a lunch the company provided, and ten minutes in the afternoon. About 2,500. The “talking problem” came to a head; from November 17 the operators were made to call out their assembling difficulties; Operators 1A and 2A were interviewed and told off.
• Period 8, January 23 to March 10, 1928. Work stopped at 4:30 instead of 5:00, the operators’ own choice. On January 25 Operators 1A and 2A were returned to the department and replaced by Operators 1 and 2, born in 1909 and 1907, chosen by the foreman for comparable rates and output. Defective parts in this period. Output up sharply.
• Period 9, March 12 to April 7. Work stopped at 4:00. About 2,900.
• Period 10, April 9 to June 30. Back to 5:00. About 2,800. Mayo’s first measurements in the room, April 26.
• Period 11, July 2 to September 1. No work on Saturday mornings, with the morning’s pay kept. Daily output up, weekly output down.
• Period 12, September 3 to November 24, 1928. Back to the original 48 hours with no rests and no lunch. The operators disliked it, and on September 24 the observer found them holding output down so that the rests would come back; he promised them the rests would return. Hourly rates fell for four of the five; weekly output reached a new high, about 2,900.
• Period 13, November 26, 1928, to June 29, 1929. The rests of Period 7 restored, without the lunch. About 3,000. This is where the detailed account in the 1939 book stops.
• Periods 14 to 23, July 1929 to February 1932. Operator 5 replaced for a time by Operator 5A and then restored; defective parts until the spring of 1930; seating changed and changed back; the working week shortened as the Depression deepened; the rests finally removed. Jones’s table marks the main changes.
• Period 24, March 1, 1932, to February 8, 1933, by Franke and Kaul’s count. All five operators laid off and replaced by more senior workers who had never assembled relays. Jones counts the study as 270 weeks to June 1932.

The five hypotheses

By 1929 the investigators were, in the words of the 1939 book, “puzzled”: “The general upward trend in output independent of any particular change in rest pauses or shorter working hours was astonishing.” They considered five explanations. First, the test room’s better light, its fans, the chute, and the fewer types of relay its operators had to make; this was never convincing to most of them, though Operator 5, who had the most types to make, had improved least. Second, that the rests had relieved cumulative fatigue, so that the operators could stand Period 12’s full hours; the fall in hourly rates in that period supported it. Third, that the rests, the pay and the small group had relieved monotony. Fourth, that the pay change of Period 3 had given the operators a direct wage incentive for the first time. Fifth, that the change in supervision, the removal of what the book calls the operators’ “apprehension of authority,” was the thing, and that “social factors were the major circumstances limiting output.” The book’s five hundred remaining pages follow the fifth.

Three readings of the same tape

Parsons, 1974. H. McIlvaine Parsons, an industrial psychologist, argued in Science that the operators had known their output at every moment, from the counters and the clicking of the tape, and had been paid on it from Period 3: the rising output was “a combination of information feedback and financial reward,” the ordinary working of consequences on behaviour, and the Hawthorne effect was better defined as “the confounding that occurs if experimenters fail to realize how the consequences of subjects’ performance affect what subjects do.”

Franke and Kaul, 1978. Richard Franke of Worcester Polytechnic Institute and James Kaul borrowed the test room’s records from the plant between November 1976 and May 1977 and put copies on microfilm at the University of Wisconsin at Milwaukee and at Worcester. Over twenty-three periods they regressed the group’s hourly and weekly output, and the repair time that measured its quality, on twelve variables: hours, rest time, the small-group pay, the replacement of the two operators, the Depression from October 24, 1929, the defective parts of periods 14 and 15, and the rest. The replacement, which they called managerial discipline, explained 56 per cent of the variance in weekly output; the Depression and rest time 8 per cent each; the pay change 1 per cent; hours 24 per cent. Quality was explained by the defective parts. Their conclusion: “Imposition of managerial discipline, economic adversity, and quality of raw materials provide most explanation, obviating the need to draw upon less clearly definable human relations mechanisms.” Milton Bloombaum in 1983 argued that the result rested on the algebra of the method and on confounded variables, and reached conclusions closer to the original researchers’.

Jones, 1992. Stephen Jones of McMaster University used the same microfilm for the individual weekly output of all eight women over 270 weeks, allowing for the experimental variables, for the replacements, and for the operators’ influence on each other, and tested the Hawthorne effect under two definitions, a narrow one in which output jumps after each experimental change and a broad one in which it drifts up for weeks afterwards. “Whether the Hawthorne effect is defined in a narrow or a broad sense, the conclusion is the same: the original Hawthorne studies contain little clear evidence of a Hawthorne effect.” The one definition his data could not test was that the whole five years had been a single experiment, in which case there was one observation and no control. Levitt and List, who found daily output for the first thirteen periods beside the lighting graphs, reported in 2009 that their results were “in the spirit of Jones”: little support for the effect when the data are read properly, and the appearance of it when they are not.

The people, so far as the record names them

The 1939 book gives the operators numbers, not names. Theresa Layman Zajac, whose pay cheque of August 13, 1927, is in the Baker Library, said in 1976 that she “had no idea there would be so much happening and so many people watching us.” Jennie Sirchio, by Edwin Gale’s account, was the best of the five and was kept on in an office for a few months after the room closed. Operator 1A married during Period 7. Operator 2A’s first medical examination found a slightly anaemic condition; the book says the investigators made no serious attempt to understand her, and that a later inquiry would have found interesting results in her case. Homer Hibarger, the observer, was succeeded by Donald Chipman. George Pennock, the superintendent, wrote to Dugald Jackson in 1930 that the researchers had grown sceptical of proving anything about human behaviour under varying conditions. Mayo wrote to Pennock in October 1929 that the change they were working toward would be humane rather than mechanical.