Psychologists were the first to discover learning curves (see Mazur & Hastie, 1978, for review). These researchers focused on the behavior of individuals. Psychologists found that the time individuals took to perform a task and the number of errors they made decreased at a decreasing rate as experience was gained with the task (Ebbinghaus, 1885/1964; Thorndike, 1898). For example, Thurstone (1919) found that the learning-curve pattern shown in Fig. 1.1 characterized the performance of students as they progressed through a typing course.
Researchers working in the individual psychological tradition often fit their data to an exponential rather than a power function, as is customary in organizational learning-curve analysis. There is evidence, however, that power functions may fit individual learning data better than exponential functions (Newell & Rosenbloom, 1981). Further, Delaney, Reder, Staszewski, and Ritter (1998) found that the fit of power functions could be improved by plotting learning curves separately for each problem-solving strategy individuals used. When estimating learning-curve data, one should be sensitive to the choice of the appropriate functional form. Issues that arise in estimating learning rates are discussed later in this chapter and in Chap.3.
A recent trend in research on individual learning is to analyze individual learning in an organizational context. For example, Singh, Ton, and Youn (2011) analyzed individual learning in open source software projects. Kim, Krishnan, and Argote (2012) studied individual learning of employees of a computing call center. Mukhopadhyay, Singh, and Kim (2011) analyzed learning curves of participants in a physician-enabled referral system.
Additionally, learning curves have been found at the group level of analysis. For example, in their studies of the effects of various communication networks on the performance of groups, Guetzkow and Simon (1955) and Leavitt (1951) found that the errors made by groups and the time groups took to complete tasks decreased at a decreasing rate as groups gained experience. Similarly, in their analysis of the effect of planning on group performance, Shure, Rogers, Larsen, and Tassone (1962) found that group performance followed a learning curve.
Learning curves have also been found at the organizational (e.g., Wright, 1936) and industry levels of analysis (e.g., Sheshinski, 1967). Several researchers distin- guish among learning curves, progress curves, and experience curves as a function of the level of analysis. According to Dutton and Thomas (1984), the term “learning curve” is frequently used to describe labor learning at the level of an individual employee or a production process. The term “progress curve” is often used to describe learning at the level of the firm. Experience curves are used to describe learning at the level of an industry. These distinctions, however, are not made uni- versally in the literature. I use the term learning curve to encompass these related phenomena and specify the level of analysis of the phenomenon.
The focus of this monograph is on learning at the organizational (or organiza- tional subunit) level of analysis. Relevant work at the group and at the interorgani- zational levels of analysis is incorporated when it has implications for learning at the level of the organization. Research on group learning provides many of the micro underpinnings of organizational learning. These micro underpinnings of organizational learning are developed in Chap. 5. Research on interorganizational or population-level learning (Miner & Haunschild, 1995) provides the macro con- text in which organizational learning takes place and also has implications for how one organization learns from another. The implications of interorganizational learn- ing for organizational learning and productivity are discussed in Chap. 6.
Wright published an early documentation of a learning curve at the organiza- tional level of analysis in 1936. Wright (1936) reported that the amount of labor it took to build an aircraft decreased at a decreasing rate as the total number of aircraft produced, cumulative output, increased. Dutton, Thomas, and Butler (1984) noted that Rohrbach had reported that the same pattern characterized the production of aircraft in Germany in the 1920s. Alchian (1963) found that the learning-curve pat- tern characterized the production of a variety of aircraft.
Many studies were conducted after the publication of Wright’s classic paper that investigated whether the learning-curve pattern characterized the manufacture of other products as well. Dutton et al. (1984) provided an excellent review and inter- pretation of research on organizational learning curves through the mid-1980s (see also Dutton & Thomas, 1984; Yelle, 1979). Lapré and Nembhard (2010) provided a more recent review and integration. A few of the particularly important early studies on organizational learning are highlighted here.
One early study compared rates of learning in different types of production work. Hirsch (1952) found that improvements in unit labor costs associated with cumula- tive output were greater in assembly than in machining work. Hirsch’s finding has been interpreted as providing evidence that learning curves are steeper in labor- intensive than in machine-intensive industries. This finding has received mixed sup- port in subsequent studies (cf. Adler & Clark, 1991). In a similar vein, Baloff (1966, 1971) found that the tendency for learning curves to “plateau” or level off was greater in machine-intensive than in labor-intensive industries.
Although early work on learning curves focused on industries that manufactured discrete products such as planes, trains and automobiles, learning curves have also been found in continuous process industries. For example, Hirschmann (1964) found that petroleum refining followed a learning curve. His finding has important implications because it suggests that learning curves are not due solely to “labor learning” because labor played a relatively minor role in these settings but rather depend on modifications in the organization and its technology as well (see also Baloff, 1966). Dutton et al. (1984) described how these findings on the presence of learning curves in continuous process industries were important in contradicting the prevailing and misguided view that learning curves could be explained mainly by learning on the part of direct production workers.
Productivity, of course, has been found to depend on other factors besides cumulative output. For example, Preston and Keachie (1964) found that unit labor costs depended on the rate of output as well as on the amount of cumulative output. An organization that tries to increase its rate of production dramatically may experi- ence productivity problems independent of learning that are reflected in high unit costs. Preston and Keachie’s (1964) work showed the importance of including changes in the rate of output as well as cumulative output in assessing learning rates.
Similarly, Rapping (1965) demonstrated the importance of controlling for addi- tional factors such as economies of scale in assessing learning rates. Rapping (1965) showed that the production of Liberty Ships during World War II followed a learn- ing-curve pattern when inputs of labor and capital were taken into account in the analysis. Although the effects of labor and capital were significant, the effect of cumulative output remained highly significant when these additional factors were included in models of productivity. Thus, Rapping (1965) demonstrated very con- vincingly that productivity gains associated with cumulative output were not due to increased inputs of labor or capital or to increasing exploitation of economies of scale. Although these factors were important, evidence of learning remained strong when they were taken into account.
Much of the work on organizational learning curves between the publication of Wright’s influential piece in the 1930s and the early 1980s focused on studying the phenomenon in different industries. Research during this period also focused on specifying the functional form of the relationship between the unit cost of production and cumulative output (Yelle, 1979). Although attempts were made to identify fac- tors underlying the learning curve, empirical evidence on the importance of various factors was very limited (Dutton & Thomas, 1984).
Source: Argote Linda (2013), Organizational Learning: Creating, Retaining and Transferring Knowledge, Springer; 2nd ed. 2013 edition.