DOI: 10.1002/jhbs.
21871
ORIGINAL ARTICLE
Bringing radical behaviorism to revolutionary
Brazil and back: Fred Keller's Personalized System
of Instruction and Cold War engineering education
Atsushi Akera
Department of Science and Technology Studies,
Sage 5206, Rensselaer Polytechnic Institute, 110
8th Street, Troy, NY 12180
Substantial portions of this article were drawn
from “Billy Vaughn Koen and the Personalized
System of Instruction in Engineering Education,”
Proceedings of the 2014 ASEE Annual Conference
and Exposition, Indianapolis, IN. 2014,
c Ameri-
can Society for Engineering Education. I am much
indebted to the reviewers of this manuscript
for revealing factual errors and pointing me to
relevant literatures on PSI and Skinnerian behav-
iorism.
J Hist Behav Sci. 2017;1–19. wileyonlinelibrary.com/journal/jhbs
Abstract
This article traces the shifting epistemic commitments of Fred S.
Keller and his behaviorist colleagues during their application of Skin-
nerian radical behaviorism to higher education pedagogy. Building on
prior work by Alexandra Rutherford and her focus on the successive
adaptation of Skinnerian behaviorism during its successive applica-
tions, this study utilizes sociologist of science Karin Knorr Cetina's
concept of epistemic cultures to more precisely trace the changes in
the epistemic commitments of a group of radical behaviorists as they
shifted their focus to applied behavioral analysis. The story revolves
around a self-paced system of instruction known as the Personalized
System of Instruction, or PSI, which utilized behaviorist principles to
accelerate learning within the classroom. Unlike Skinner's entry into
education, and his focus on educational technologies, Keller devel-
oped a mastery-based approach to instruction that utilized gener-
alized reinforcers to cultivate higher-order learning behaviors. As
it happens, the story also unfolds across a rather fantastic politi-
cal terrain: PSI originated in the context of Brazilian revolutionary
history, but circulated widely in the U.S. amidst Cold War concerns
about an engineering manpower(sic) crisis. This study also presents
us with an opportunity to test Knorr Cetina's conjecture about the
possible use of a focus on epistemic cultures in addressing a clas-
sic problem in the sociology of science, namely unpacking the rela-
tionship between knowledge and its social context. Ultimately, how-
ever, this study complements another historical case study in applied
behavioral analysis, where a difference in outcome helps to lay out
the range of possible shifts in the epistemic commitments of radical
behaviorists who entered different domains of application. The case
study also has some practical implications for those creating distance
learning environments today, which are briefly explored in the con-
clusion.

c 2017 Wiley Periodicals, Inc. 1
2 AKERA

This article is about the application of Skinnerian behaviorism to one domain of education via a self-paced system of
instruction known as the Personalized System of Instruction (PSI). PSI was an early example of an inverted classroom
developed by Skinner's close friend and colleague, Fred S. Keller. Paralleling the work on education by other behavior-
ists, including Skinner, PSI was a mastery-based approach to programmed instruction built around study guides and
small study units with tightly specified learning outcomes. Students needed to score 100 percent on each unit test
before advancing to the next unit. In return, every student who completed all of the units earned an A. With PSI, the
grade, or more precisely, progress toward the grade served as a generalized reinforcer. PSI circulated widely during
the 1970s across a wide variety of disciplines. I nevertheless focus on engineering education because the distinct epis-
temic culture of engineering education provides us with a good opportunity to follow Skinnerian radical behaviorism's
epistemic transformation during its epistemic encounter with a field of application.
The basic idea that Skinner's radical behaviorism has been upheld through its applications has been examined by
Alexandra Rutherford in Beyond the Box (2009). Although behaviorism as a whole was challenged by the emergence
of a cognitive paradigm during the 1950s, it continued to gain ascendance during the 1960s and 1970s through its
application in a variety of arenas ranging from clinical psychology and education (Leahey, 2004), to advertising and mil-
itary interrogation (Lemov, 2005). Rutherford specifically traces the efforts of those with a commitment to Skinner's
vision for radical behaviorism (as opposed to the methodological behaviorism practiced by many; see Schneider & Mor-
ris, 1987), and its direct extension into mental asylums, prisons, and eventually even self-help manuals. Drawing on
the expansive definition of technology advanced by Shapin and Schaffer (2011), Rutherford introduces the notion of a
“technology of behaviorism” to account for the broad circulation of Skinner's scientific ideas and methods.
In order to gain a more precise understanding of what occurs during the circulation of a scientific idea, I engage
more directly with Karin Knorr Cetina's (1999) concept of epistemic cultures. Based originally on ethnographic and
ethnomethodological studies of the high-energy physics community, Knorr Cetina advanced this concept specifically
to draw attention to the “machinery” of knowledge production. It makes sense to turn to Knorr Cetina's idea given Skin-
ner's emphasis on apparatus. However, Knorr Cetina's concept also draws attention to the heterogeneous assemblages
of social and institutional arrangements, ideas, symbolic representations, and experimental practices that are also used
to produce and warrant knowledge. This, along with her attention to ethnomethodological detail, offers a way to more
fully unpack how radical behaviorism's epistemic commitments changed as it traveled to new domains of application.
Specifically, Knorr Cetina's concept allows us to examine more fully Rutherford's central claim that the circulation of
radical behaviorism depended on its successive adaptation. Helen Longino's (2013) recent study of the behavioral sci-
ences, which is also informed by Knorr Cetina's work, is also useful for keeping in mind that varied epistemic traditions
can coexist even within a single field of study.
For this article, the most relevant prior work is Nancy Campbell's, Discovering Addiction (2007). In one chapter of this
historical study, Campbell documents the gradual changes in the epistemic commitments of a group of researchers—
behavioral pharmacologists—whose work was substantially influenced by radical behaviorism. Initially, there were
those central to the field's origin who were confident that radical behaviorism could explain everything about the pro-
cess of drug addiction. Yet, as the work progressed, those who extended this line of work softened their epistemic
stance in accepting a more hybridized epistemic position. While I leave the details to the conclusion, the history of PSI
offers an important, complementary case because the story has a different outcome. In addition, PSI provides us with a
rare opportunity to trace the shifting epistemic commitments of a core member of the radical behaviorist community.
As it turns out, PSI's development and diffusion also unfolded across a rather fantastic political terrain. PSI origi-
nated in the context of Brazilian revolutionary history, spread rapidly due to post-Sputnik concerns about an engineer-
ing manpower(sic) crisis, and faltered amidst concerns about stagnating U.S. industrial productivity. The idea that ideo-
logical contexts can shape the sciences is of course a classic problem in the sociology of scientific knowledge (Golinski,
1998). Moreover, the thought that a focus on epistemic cultures can be instrumental to an understanding of the process
by which this occurs has been noted by Knorr Cetina and Longino (Knorr Cetina, 1999, p. 2, as cited in Longino, 2013,
p. 9), and executed by Campbell.1 The focus on context is important in this account, and my references to ideology
are primarily political—the leftist orientation of Brazilian President João Goulart, or the corporatist attitudes toward
labor found in Cold War United States—although disciplinary ideologies could also enter the picture. Indeed, the
AKERA 3

synergy between the liberatory pedagogies that were already emerging in Brazil, and radical behaviorism's commit-
ment to the idea that all organisms learn through a common mechanism of reinforcement was an important factor in
the development of PSI.
This account is told in narrative form to foreground the complex interplay that exists between the political, insti-
tutional, and technical assemblages that constitute epistemic cultures and their successive transformation. Given my
focus on engineering education, I open with the engineering education reform initiatives at the University of Texas;
describe the origins of PSI; document its adoption and adaptation within engineering education; and address some
subtleties of radical behaviorism as applied to PSI; before ending the historic narrative with an account of PSI's decline.
While the analytic issues raised here in the introduction are embedded throughout the narrative, I draw attention to
them in the conclusion and also explore the implications of my findings.

1 BILLY VAUGHN KOEN AND THE FOCUS ON TEACHING EFFECTIVENESS

AT UT AUSTIN

PSI circulated broadly in engineering education largely due to the efforts of a nuclear engineer at the University of
Texas at Austin, Billy Vaughn Koen. A native Texan, Koen was born in Graham, TX to two committed teachers. While
Koen was still in grade school, his father changed his career from teaching to photojournalism and moved the family
to Austin where he became a staff photographer for the University of Texas. Being part of UT's public relations office,
Koen's father became friends with many senior faculty members, including John J. McKetta, the chair of the Chemical
Engineering Department and later Dean of the College of Engineering (Koen, 2011a, segment 3–7; Koen, 2011b).2
Koen enrolled at the University of Texas to study Chemical Engineering, under McKetta's influence. Having excelled
as a student, he moved on to a doctoral program in nuclear engineering at MIT. There he pursued cutting edge work
on reactor kinetics that utilized statistics and computational modeling. His MIT doctorate, which he received in 1968,
provided Koen with three options for a career: A position at Los Alamos, a job at General Atomics, or a teaching position
back at UT Austin. Koen embraced the role of teacher (Koen, 2011a, seg. 6–7).
It was no accident that Koen returned to Texas. In 1963, McKetta became the dean of engineering and undertook
a two-pronged strategy for elevating his college's standing. On the one hand, conversations about the future of oil
production—peak oil in Texas would occur in 1972—brought the state to set its sights on nuclear energy (Railroad Com-
mission of Texas, n.d.). UT Austin was one of the schools that won support for a teaching and demonstration reactor,
and McKetta needed a faculty that could leverage this support into a strong research program in nuclear engineer-
ing. But in the wake of Sputnik, there were also broad concerns about an engineering “manpower(sic)” crisis. While the
state had set up its own land-grant type institution, engineering education in Texas remained underdeveloped; what
strengths that existed were in areas such as petroleum engineering and agricultural engineering. Seeking to reorient
his program to meet the state's evolving workforce needs (which included a growing defense industry because of the
political pressure to distribute defense spending), McKetta, with the support of General Dynamics, had already estab-
lished a Bureau of Engineering Teaching, a colloquium on effective teaching, and an annual prize awarded to the best
engineering teacher. Koen was someone McKetta could entrust with both his initiatives.3
Being a home-grown candidate, and one who hailed from the Nuclear Engineering program at MIT, Koen
approached his interview with considerable hubris. By his own recollection, Koen made three promises to the search
committee. First, he promised to make an internationally known contribution to the field of nuclear engineering. Sec-
ond, he offered to make an equally important contribution to engineering education that would bring credit to his
department and to his university. Third, he assured the committee members that he would make an even broader con-
tribution that would “change the way the world sees the human species.” This was work that would earn him the highest
award of the Liberal Education Division of the American Society for Engineering Education (ASEE; see “Sterling Olmsted
Award”). Here we focus only on the productive tension between Koen's interest in teaching and research (Koen, 2011a,
seg. 19).
4 AKERA

In this respect, the most important thing to note is that, having promised the moon, Koen, like so many PhD stu-
dents, received no real training as a teacher while in graduate school. Undaunted, Koen applied the same heuristic (and
hence epistemic practice) he used during his studies of nuclear engineering, which was to canvass the state of the art
in the discipline within which he wished to claim some expertise. It was through this process, and through information
provided by an MIT classmate, that Koen discovered Keller's PSI (Koen, 2011a, seg. 21).

2 PSI'S ORIGIN IN REVOLUTIONARY BRAZIL

At the time that he developed PSI, Keller had just rotated off of a term as the head of the Psychology Department at
Columbia University. Back in the late 1920s and early 1930s, Keller was part of a group of young Turks in Harvard's
Psychology Department, headed up by Skinner, who as PhD students dared to challenge the introspective approach
in American psychology. While this break has been exaggerated, there were those in Harvard's psychology depart-
ment who continued to draw on the introspective approach of Wundt and Titchener. Riding on the crest of the rise of
the experimental sciences, early behaviorists—Pavlov, Thorndike, and the like—came to occupy a more agnostic posi-
tion with respect to the mind, and with Watson, the original, radical behaviorist stance that there was no dividing line
between humans and animals. Watson called for human psychology to be pursued as a branch of the natural sciences
(Schneider & Morris, 1987). It was Skinner who then gave full articulation to this position and backed it with experi-
mental data in his 1938 seminal text, The Behavior of Organisms: An Experimental Analysis (Bjork, 1993; Costall, 2006;
Leahey, 2004; Rutherford, 2009; Wiener, 1995).
This was, however, work that was completed 20–30 years earlier. During the intervening decades, Keller built up
an experimental laboratory for radical behaviorism at Columbia, trained a large cadre of graduate students, and with
the ascent of behaviorism as a whole, saw the placement of his students in various prominent institutions and fields of
application. While not one to harbor strong institutional ambitions, Keller came to head his department in 1959. Yet
despite these accomplishments, Keller continued to operate in Skinner's shadow, especially as Skinner gained further
notoriety through his more visionary work published after his return to Harvard. Still, Keller was approaching retire-
ment having built Columbia into one of the country's most significant shops for graduate training in radical behaviorism
(Keller, 2009, pp. 160–245).
But then in 1959 Keller received a rather unusual invitation. During the 1950s and 1960s, Brazil was experiencing
considerable political unrest, fueled by the pressures of modernization and accelerated development. Following the
suicide of President Getúilo Vargas in 1954, Brazil was thrown into a period of turmoil marked by the quick succession
of presidents. However, there was a period of relative calm between 1956 and 1961 when the progressive vision put
forward by Juscelino Kubitschek—a promise to deliver “fifty years of progress in five”—ignited the country's imagina-
tion, bringing with it a measure of economic prosperity, and hope. This hope was based on uniting the nation, which was
symbolized by Kubitschek's commitment to a new capital city, Brasilia, and its shining new university. It also called for
the modernization of Brazilian universities outside of the capital (Burns, 1980, chaps. 6–7).
It was under Kubitschek that the Director of the Faculty of Philosophy, Sciences, and Letters at the University of
São Paulo, one of the leading national universities, invited Keller to come to his university to help modernize its Psy-
chology Department. Unfortunately for Keller, by the time he arrived in 1961, the director had been ousted as a result
of internal differences. As an indication of the broader turmoil that was affecting Brazilian society, Kubitschek's suc-
cessor, Jânio Quadros, left the presidency after just seven months in office. In what historians regard to have been a
political miscalculation, Quadros tendered a resignation hoping that this would instead bring him broader powers. The
ascent of João Goulart, the Vice President and member of Brazil's Labour Party, represented a distinctly leftist turn
in the government. This was the era during which the United States was struggling with Fidel Castro, and was caught
meddling with Brazil's presidential succession. These events unfolded during Keller's year in São Paulo. On his arrival,
it was already evident that he was at the mercy of a skeptical if not hostile faculty that remained unconvinced that an
outsider, and an American at that, could or should transform their academic programs (Keller, 1974a, p. 143ff; Cirino,
Miranda, & de Cruz, 2012).
AKERA 5

Unwilling to return to Columbia having made the journey, Keller began working with Dr. Carolina Martuscelli Bori,
one of the younger, more sympathetic faculty members, along with her graduate student, Rudolpho Azzi, an experi-
enced teacher. If intellectual conservatism reigned in Harvard's Psychology Department during the 1930s, Latin Amer-
ica's peripheral positioning produced a temporal elision that extended this conservatism two decades longer so that
neither behaviorism nor experimental psychology had developed a solid footing at the University of São Paulo. As such,
Keller could simply replicate his work at Columbia. Drawing on the animal handling capabilities—and the epistemic cul-
ture of experimentation—that were already present in an adjoining physiology lab (much as Skinner had done), the
group assembled a modern research laboratory for experimental psychology built around classic Skinner boxes. The
group then used their meager resources to assemble a teaching laboratory, this through the considerable efforts of
Azzi. In the process, the group acquired the capacity to tinker with devices in the same manner by which Skinner earned
his fame. Once the doors to the institution were opened, reproducing the epistemic culture of an ascendant tradition
in the behavioral sciences was a relatively straightforward process. The teaching laboratory also made it possible to
develop an introductory course in psychology built around behavioral experiments (Cirino et al., 2012; Matos, 1998).
There was also a progressive, gender dimension to the institutional arrangements—a crucial part of Knorr Cetina's
heterogeneous assemblages—that helped propagate radical behaviorism's epistemic traditions. Women had only made
limited inroads into Brazilian academe, but new doors opened especially under Goulart. As indicative of this trend, one
of the key figures who helped spread radical behaviorism in Brazil, Maria Amelia Matos, got her start as an undergrad-
uate, then graduate assistant in Bori and Azzi's teaching laboratories in São Paulo (Matos, 1998). Youthful enthusiasm,
and a new willingness to recognize youth, carried not only Matos, but Bori and Azzi, as the liberalization of Brazilian
academe enabled them to enter the domain of radical behaviorist research.
Bori's reforms were deemed a success, and she along with Azzi were invited in 1963 by Goulart's chief of staff, Darcy
Ribeiro, to take their work to the University of Brasilia and to establish a new psychology department there. Keller and
his former PhD student, Gilmour Sherman, were invited to join them as advisors. With the university's doors yet to
open, the group had time to plan. It was in the course of designing a new curriculum from scratch that the group came
up with PSI. Pressed to be experimental in form as well as content, the group began considering how their knowledge of
radical behaviorism could be injected into the classroom itself. Keller was no doubt also motivated by the fact that Skin-
ner himself had been writing about education; he may have also known other radical behaviorists who were working in
this area.4
As it turns out, Keller had done some relevant work earlier. Amidst the crucible of World War II, Keller was called
upon to use reinforcement theory to help speed up the training of Morse Code to military officers. In a rather ele-
gant paper entitled, “The Phantom Plateau,” written about a decade later, Keller drew together the observations that
had enabled him to contribute to this wartime training program. Quantitative studies of Morse Code training regimes
extended back to the nineteenth century, namely to the heyday of telegraphy. Most of these studies revealed a classic
plateau, where trainees would fail to reach a standard minimum rate of transmission and reception required for certi-
fication, but would achieve certification at a later date after regaining an ascending learning curve. Those interpreting
the curve speculated that students attained a certain mental state (becoming strong enough at deciphering words to
begin deciphering common phrases). However, a number of subsequent experiments produced instances exhibiting no
such plateau. Under different training regimes, some operators could improve their speed on a continuous basis, with
a better overall learning curve (Keller, 1958).
Analyzing the different training regimes through radical behaviorism's epistemic lens, Keller determined that the
problem lay not with the student, but with the curriculum. Training regimens that produced the plateau effect did so
because of a static regimen that students quickly mastered, and further improvement had to wait until they discovered
a way to learn that was not part of the regimen. Drawing somewhat on the principles of operant conditioning, Keller
speculated that a training method that focused on speeding up single-letter (not word) recognition through instan-
taneous feedback could be used to produce continuous performance improvements. In what became known as the
“code-voice method,” the trainee had to strive to beat a trainer in calling out the correct letter in timed intervals that
were made progressively shorter based on a 95 percent accuracy criterion. Keller initially wrote up the results during
the war, but used the postwar period to give full articulation to his ideas, as backed by new experimental data. For the
6 AKERA

latter, Keller employed a single-subject experimental design that was not uncommon to radical behaviorist research.5
By demonstrating that Anne Simmons, “an 18-year-old high-school graduate who was working to earn money for col-
lege expenses” could produce a learning curve that far exceeded military officers trained under other regimes, Keller
could use a single instance to blast apart prior understandings of the phenomenon (Keller, 1943, 1958, p. 8; Keller,
1974a, p. 146).
While this work was based more or less on a straightforward extension of radical behaviorism's experimental prac-
tices to human subjects—at most, it veered from operant conditioning by focusing on response speed rather than
assembling complex behaviors—it nevertheless laid an important foundation for the work in Brazil. What the group
in Brazil retained from Keller's early work was a focus on continuous learning and reinforcement, and mastery-based
education—100 percent versus 95 percent in the case of PSI. However, a challenge lay in their desire to teach not a
narrow skill, but a broad subject. Drawing on the broader epistemic traditions of radical behaviorism, Keller and his
colleagues determined that, like any good experiment, it was necessary to have a well-defined system of contingencies
designed to produce desired behaviors. This implied the concise teaching units, with carefully specified learning objec-
tives. One area where the group had to relax their epistemic commitments—and therefore extend radical behaviorism's
epistemic repertoire—was with the practical difficulty of providing reinforcement for each and every discrete learning
event. This left the group with the task of coming up with a more general system of reinforcements that would moti-
vate continuous learning despite the complex content of a general course in experimental psychology (Keller, 1974a, p.
147).
Since PSI was initially developed as a laboratory-based course, one such reinforcement could be found in the lab
itself. After all, the successful replication of a known experiment, such as a rat learning to press a lever, could be as
reinforcing to the student as it was for the rat. However, as reformulated to work with text-based courses, the basic
principles of PSI were laid out as follows:

“1 The go-at-your-own-pace feature, which permits a student to move through the course at a speed commensurate
with his(sic) ability and other demands upon his time.
“2 The unit-perfection requirement for advance, which lets the student go ahead to new material only after demon-
strating mastery of that which preceded [it].
“3 The use of lectures and demonstrations as vehicles of motivation rather than sources of critical information.
“4 The related stress upon the written word in teacher-student communication.
“5 The use of proctors, which permits repeated testing, immediate scoring, almost unavoidable tutoring, and a marked
enhancement of the personal-social aspect of the educational process.” (Keller, 1968, p. 83)

At the heart of the “Keller Plan,” as it was also called, was a self-paced system of instruction built around concise
teaching modules with specified learning outcomes, and an inverted system of instruction that utilized written study
guides and tutoring, as opposed to lectures. There was also a “readiness test” and a requirement that students to score
100 percent before they could advance to the next unit. Like mice on a treadmill, or rats in a maze, students would
strive to learn the right answers, retaking the test as often as they needed to complete all of the units, for which they
earned an “A.” The method worked well. In fact it worked all too well, for as many as 80 percent of the students earned
an “A” after having spent as much as twice the amount of time on this class as their other classes. Occurring before the
era of rampant grade inflation, this would generate, outside of Brasilia, considerable suspicion and envy on the part of
other faculty (Keller, 1968).
The revolutionary moment in Brazil and the institutional setting of the University of Brasilia clearly created a space
conducive to pedagogic thinking and experimentation. The very architecture of Brasilia, with its high-modern Soviet-
style housing units, spoke of a grand social experiment, an attempt to promote social harmony and to meliorate the
social ills of a highly stratified society. In the Northeast of Brazil, Paulo Friere was developing his liberative pedagogies
in an effort to combat adult illiteracy. Keller and his Brazilian colleagues may not have been entirely swayed by Goulart's
leftist agenda—Brazilian academic institutions remained elite institutions—nevertheless PSI was attractive because it
AKERA 7

was potentially democratizing to its core. The underlying idea, if still tainted with a twinge of Social Darwinist thought,
was that anyone who put in the time and effort could master any subject, regardless of social background or innate
ability (Burns, 1980, pp. 459–462, 487–504).6 PSI did require Keller and his colleagues to step away from controlled
experiments and rely on a student's progress toward the A as a generalized reinforcer. But this was not inconsistent
with the way in which radical behaviorism was being extended into other fields, including most notably Ayllon and
Azrin's work on token economies (Rutherford, 2009, chap. 3).
However, the University of Brasilia's leftist leanings also brought the work on PSI to a close. While the first two
courses in the psychology sequence ran for two semesters, the university was shut down and its faculty dispersed to
other institutions following a 1964 military coup. Sensing political turmoil, Keller and Sherman had already left for a
position at Arizona State University. Still, having long operated in Skinner's shadow, Keller felt that, quite late in his
career, he had finally contributed something truly original to the world of teaching, if not scholarship. At Arizona State,
Keller and Sherman continued to utilize and work on PSI, albeit under less ideal circumstances (Keller, 1974a, p. 149ff).

3 TRANSPORTING PSI INTO ENGINEERING EDUCATION

Koen came across published accounts of Keller's methods, including his famous invited address, “Good-bye-teacher…,”
delivered in 1967 before the Society for the Teaching of Psychology (Division 2) of the American Psychological Asso-
ciation. Never one to be constrained by disciplinary boundaries, Koen immediately decided to apply Keller's method
to his senior-year nuclear engineering course. On the other hand, neither one to do things haphazardly, Koen taught
the course in a conventional manner during his first year back at UT Austin while he prepared the course for PSI in the
second. This gave him a clear point of comparison. As the method promised, PSI generated astounding results. Seventy-
two percent of the students preferred his course to a traditional lecture course. Eighty-eight percent reported putting
in more effort than any of their other courses, and the same number of students reported that they looked forward to
this course more than anything else they were taking. And whereas only 20 percent of the students earned an A in the
earlier non-PSI version of the course (28 percent earned “B”s and 35 percent earned “C”s), 70 percent of the students
earned an A in the PSI version of the course (Koen, 2011a, seg. 30; Koen, 1970, 1971).7
The College's Bureau of Engineering Teaching, led by James E. Stice, who was Koen's senior by perhaps six years,
took an immediate interest in Koen's success. Other instructors, especially in Mechanical Engineering, soon began
using the method. PSI, given its reported successes, also began garnering national attention, being discussed widely
in the American Society for Engineering Education's (ASEE) flagship journal, Engineering Education, and other publica-
tions. Soon there were national conferences and at least one ASEE summer teaching institute dedicated to PSI and
other forms of individualized instruction. While PSI was utilized in other fields, including psychology and teacher edu-
cation, the method circulated quite early and rapidly within engineering education (Hoberock, Koen, Roth, & Wagner,
1972; Keller, 1974a, p. 153; Keller & Koen, 1976; Stice, 1971).
There was a specific logic to PSI's uptake among engineering educators. Ever since Sputnik, there were perpetual
concerns about an engineering “manpower” crisis, and although Sputnik might have been expected to augment engi-
neering enrollments, the dominant national emphasis on science resulted in an actual decline in enrollments during the
latter part of the 1950s (Teare, 1960–1961, p. 10f). Although this was corrected by the mid-1960s, concerns about
enrollment remained deeply embedded within engineering educators’ ideas and practices surrounding educational
reform. To these educators, one of the major contributing factors was the problem of retention: at the time, some-
thing approaching 40 percent of those who entered engineering left the field before completing their studies. This was
largely due to the fact that Engineering was a tough major with a highly compacted curriculum made “necessary” by
the fact that universities continued to try to offer a professional degree in an ever more esoteric field in just four (and
occasionally five) undergraduate years. Engineering's Cold War turn toward “engineering science,” which did not align
well with an incoming student's image of engineering, exacerbated matters (Foecke, 1964–1965, p. 191; Lachney &
Nieusma, 2015).
8 AKERA

But rather than deal with these issues head on, in a convoluted logic that was itself integral to the epistemic habits
of engineers—preferring to be blind to politics but confident in their ability to improve on any system—engineering
educators determined that higher retention rates could contribute as much, if not more than enhanced recruitment in
addressing the workforce crisis. More efficient and effective teaching methods would improve retention by allowing
a greater proportion of engineering students to complete their studies. It was for this reason that the Bureau of Engi-
neering Teaching, which was apparently the first program of its kind among engineering schools, placed such emphasis
on teaching effectiveness, and was institutionally in a position to notice an initiative like Koen's. From the Bureau's
point of view, any new pedagogic experiment that accelerated learning, and allowed 70 percent of the students to
obtain an “A,” was highly welcome—although the bimodal grade distribution, and the high number of incompletes and
withdrawals, should have also been a concern from the point of view of retention (“A history of the teaching improve-
ment program. In College of Engineering,” 1967).
On the national scene, it is clear that the experimental dimension of PSI also contributed to its uptake by engineering
educators. The very fact that student progress was measured using unit tests that occurred with high frequency, some-
thing inherited from radical behaviorism's experimental practice, meant that PSI generated gobs of data. And while
engineers were not the only educators to have an affinity to data—statistical analysis was well established within edu-
cational research—the fact that engineers were familiar with statistics (and therefore not just Koen, who used statistics
to study nuclear safety) meant that engineering educators could quickly learn to contribute to this kind of research.
Improvising on their own epistemic repertoire, there were even those who experimented with their course design in
an attempt to optimize learning outcomes (Koen, 2011a, seg. 59; Kulik & Kulik, 1975).
ASEE had in fact already established an Educational Research and Methods Division (ERM) several years earlier.
However, this simply meant that there was a group that provided fertile soil for PSI's reception. After all, the learning
outcomes data generated by PSI helped to establish that there could be a statistical foundation for engineering edu-
cation research, complete with standard deviations. PSI was also backed by what, to engineering educators, was the
incontrovertibly rigorous theories of behaviorist research. Here, there was a clear and fresh amalgamation of epistemic
practices and traditions, if still a loosely formulated one. For the engineering educators who were beginning to cast
their pedagogic work as research, they were able to strengthen their epistemic claims by adopting the ideas, practices,
and even symbolic representations found within the broader “technology” of radical behaviorism. As an example, the
time-dependent records of student progress, which were visually similar to the data produced by cumulative recorders
used during animal experiments, were used to demonstrate PSI's efficacy—although the fact that PSI employed gener-
alized reinforcers meant that there was no standard measure of what was being learned with each trial. Nevertheless,
PSI was a haven for those who were beginning to seek recognition for research in engineering education as well as the
discipline in which they were trained (Kulik & Kulik, 1975; Rutherford, 2009).
Keller also took direct interest in Koen's work. While the method's efficacy was being documented in Psychology and
other fields, there was a real appeal to applying PSI to a field like Engineering where there was well-defined content,
and hence where learning outcomes could be more precisely measured. Although the two were separated by more than
a generation, Keller and Koen became fast friends. For Koen, Keller was an inspirational mentor, a renowned authority
in behaviorist psychology who was now making a mark in educational psychology. For Keller, Koen was an unexpected
and faithful disciple. Insofar as Koen saw himself as a nuclear engineer and educator, rather than an experimental psy-
chologist, Keller could count on Koen to practice PSI in its purest form. Indeed, in a manner not unfamiliar to radical
behaviorism as a whole, there was a great deal of enthusiasm, and perhaps even something like an evangelical atti-
tude toward PSI, one that demanded faithful adherence to its basic tenets. It behaved—in Knorr Cetina's words—as
one of those epistemic cultures that strove to “curl up upon themselves” (1999, p. 2), where adherence to practice and
doctrine was an expectation. Keller and Koen would joke about those who would try to “improve” the method, only to
encounter disappointment. They referred to such efforts as “SLI” or “Something-Like-It” versions of PSI (Koen, 2011a,
seg. 37).
To be fair, PSI was not always successful when not rigorously implemented. One of Koen's mechanical engineering
colleagues, Lawrence Hoberock, applied PSI to a more traditional course on kinetics and dynamics. His students did
demonstrate accelerated learning during the first part of the semester. However, Hoberock, who remained a bit too
AKERA 9

wedded to course content, packed too much into each unit and made the tests too difficult to ensure continued success.
Students began losing wind, with only 18 percent of the students actually making it to the end. This made it necessary
to modify the grading structure so that this cohort of students could continue on with their studies. Hoberock had, in
fact, cheated.8 He offered supplemental lectures to cover ground that he did not think the students could learn on their
own, and this was based on the fact that he had tried to cram too much into the course. Like so many other engineering
educators, Hoberock considered this a necessary evil. However, Thorndike's Law of Effect operated as a governing
law in that once the frequency of success fell below a certain level, students, like rats, could become discouraged and
give up—or, as a radical behaviorist might put it, fail to emit the desired behaviors due to an inadequate reinforcement
schedule—thus negating the positive effects of reinforcement (Hoberock, 1975).

4 THE BEHAVIORIST FOUNDATIONS OF PSI

There are some additional subtleties regarding the behaviorist foundations of PSI that need to be addressed before
reaching the end of this story. We can begin by taking note of some of the paradoxes inherent to PSI. The method
was cast as a form of liberative pedagogy, and yet many would regard behaviorism as a field built around notions of
control. One only needs to read Rebecca Lemov's account of some very different strands in behaviorist thought and
application—from suggestive advertising to military interrogation and brainwashing—to realize just how far this could
go (Lemov, 2005). PSI was also born within Brazil's radical social vision, and yet it was made to serve the much more con-
servative aims of a Cold War state, albeit within a field—engineering—that continued to serve as a vehicle for upward
social mobility even during the 1960s and 1970s. The PSI classroom also inverted, at least on the surface, the tradi-
tional authority between students and faculty, as student proctors who were often just a year or two ahead of those
they tested ran the entire course. Yet, the instructor-qua-experimentalist remained in control of the entire learning
situation, much like the military junta that manipulated Brazilian politics from the shadows.
Besides, what exactly was being reinforced with PSI? And what were the reinforcers that operated within a PSI
classroom?
Being committed to radical behaviorism, Keller offered one answer to this question in The Keller Plan Handbook,
which he wrote with Gilmour Sherman in 1974. Placed in tabular form, Keller listed the behaviors and reinforcers of
PSI as follows (offered here in somewhat reduced form):

Behaviors

• Study behavior, including “patterns of response to stimulation from textual material” that form the association of
ideas.
• Supportive behaviors, such as attending school, carrying out instructions, and moving from place to place (within a
PSI classroom). Routine acts he(sic) must perform weeks on end in connection with every [PSI] course.
• Collateral behaviors, such as laboratory exercises and field trips.

Reinforcers (for students)9

• Understanding, “those little explosions of satisfaction or relief from tension that come with understanding.”

• Getting ahead, unit by unit, within PSI.

• Small work units that ensure more frequent reinforcement.
• Immediate grading that ensures the immediacy of reinforcement.
• Attention, approval, and even affection of the proctor.
• Token reward of the “A” (Keller, 1974b, pp. 53–55).
10 AKERA

Aside from the fact that there appears to be some conflation between reinforcers and the schedule of reinforce-
ment, this offers one answer to the learning behaviors and reinforcers designed into PSI.
A close look at the data generated by PSI courses provides another window into the behavioral dynamics of a PSI
classroom. Using Koen's own course as an example, we note again that 88 percent of the students put more effort
into this course than any other course. And while still an exceptionally high figure, only 72 percent of the students
indicated that they liked the course better than a lecture course, even as the same, higher percentage of students (88
percent) said they looked forward to this course more than any other course. This suggests that Koen, Keller, and their
cohort were conditioning students to love PSI and its activities at least as much as, if not more than the course content.
Keller's list from above confirms this point, for most of the behaviors he lists are behaviors associated with the class
itself and not, or at least only incidentally behaviors (e.g., studying and test taking) associated with the acquisition and
demonstration of knowledge (Koen, 1970, p. 735).
Also relevant is Keller and Sherman's reference to tokens. Token economies were being broadly implemented during
the early 1970s, including within classroom environments (Rutherford, 2009, p. 72). In PSI, the grade, or even progress
toward a grade never served formally as tokens in the way that Teodoro Ayllon and Nathan Azrin developed them in
their classic study at Anna State Hospital.10 Nevertheless, both PSI and token economies used generalized reinforcers
to bring about broad changes in the subject population's behavior. Even Koen may have known about this work: Seeking
to create a better setting in which to achieve the benefits of PSI, he appealed to the College of Engineering's adminis-
trators to set up a dedicated PSI teaching facility—although unlike an asylum or prison, his students would have been
free to enter and leave (Lamb, 1975).
A further understanding of how reinforcement unfolded within a PSI classroom requires us to venture into Skinner's
ideas about education. As is well known, after World War II Skinner left his early work on experimental analysis to
turn to his more visionary work on the wider implications of radical behaviorism. This included work on intentional
communities, verbal behavior, and education. He had in fact written about teaching as early as 1954, initially based
on observations of his daughter's schooling, and this work culminated in a text, The Technology of Teaching (1968), that
paralleled Keller's work with PSI.11
However, the difference between Skinner and Keller's approach is instructive. Unable to extract himself from the
reductive habits of the Cold War experimental sciences, Skinner had used back of the envelope calculations to estimate
that some 25,000 reinforcements were required to cover a basic subject such as arithmetic. Relying in other ways on his
own epistemic repertoire, Skinner turned to programmed instruction and mechanical aids to teaching as a promising
avenue for improving U.S. public education (Skinner, 1968, p. 17).
By contrast, Keller, Sherman, and their Brazilian colleagues recognized that PSI could not be about reinforcing dis-
crete learning events. As experienced teachers they all understood, at least intuitively, that their focus had to be on
higher order learning behaviors—knowing how to read a text, follow instructions, engage in laboratory exercises, and
other forms of “supportive” and “collateral” behaviors. There were also other behaviors, such as knowing not to pro-
crastinate, or knowing when to ask for help from the proctor, which remained unnoticed or were at the very least
absent from Keller and Sherman's descriptions of PSI. These were nevertheless relevant behaviors that the “good”
students presumably had already developed before entering college, which could be reinforced, remolded, and accen-
tuated through a system of reinforcement.
In this respect, it is important to realize that although PSI was cast as a form of individualized instruction, PSI func-
tioned in an elaborate eusocial environment made up of proctors, graduate assistants, and student peers. While we
can only imagine all that remains unstated in Keller and Sherman's text, reinforcement occurred through a deep and
rich social process in which “attention, approval, and even affection” by the proctor served only as a starting point. It is
probably safe to assume that what made PSI such an exciting, robust learning environment were the rich opportunities
for students and their senior peers to model, mentor, share, and compete with one another in a complex environment
of mutual reinforcement.
As a nuclear engineer, Koen no doubt remained partially wedded to course content like his mechanical engineer-
ing colleague. Nevertheless, for Koen, and not just his radical behaviorist colleagues, the true goal of PSI was to cul-
tivate sound learning habits and strategies conducive to lifelong learning. This was in fact something engineering
AKERA 11

educators had long insisted was essential to engineering professional identities, but could rarely find a way to reduce
to practice.12 PSI did so. In yet another instance of epistemic hybridization, PSI's validity was warranted by associating
its educational outcomes with a specific profession's articulation of educational aspirations.
By 1974, Keller and Sherman had moved from Arizona State to Western Michigan, and then to Georgetown Univer-
sity, where they set up a Center for Personalized Instruction. They understood that such a center ought to delve into
the psychological foundations of PSI. By that point there were more than 300 papers, articles, and research reports
on PSI. However, most of this literature focused on simply documenting PSI's efficacy. In an essay entitled, “PSI Today,”
Sherman directed those who were investigating PSI to begin taking the positive findings of PSI as a given, and “direct
experimentation toward an analysis of the system itself” (Sherman, 1974, p. 62).
Yet if measured by the standards of behaviorist research, the work that followed was disappointing. Sherman him-
self, instead of citing behaviorist research on the psychological foundations of PSI, named as an exemplary study a
two-year research project funded by the Alfred P. Sloan Foundation… and carried out by Koen and his colleagues at
the University of Texas. A look at the research questions that Koen's team listed in their proposal demonstrates little
attempt to understand the psychological basis for PSI's success. The questions they asked—about efficacy, grade distri-
bution, retention, faculty development, cost, attrition, and scalability—all dealt with operational issues. Only a couple
of the research questions, such as the “procrastination problem,” were suggestive of the underlying behavioral dynam-
ics. And it remained unclear whether a group of engineering educators was qualified to pursue such work (Sherman,
1974, pp. 62f).
By 1974, Keller was an emeritus professor. However, Sherman was presumably young enough to carry out new psy-
chological studies. While asking why a particular line of research did not occur necessarily entails considerable specu-
lation, this question remains central to the history of radical behaviorism and its applications. Indeed, in the historiog-
raphy of behaviorism as a whole, there exists a curious apologetic in which it is often noted that behaviorism continued
to survive through its applications. Although those who continue in the tradition of radical behaviorism may deny that
this has happened, the latent concern indexically referenced by those employing this rhetorical trope is the worry that
the theoretical foundations of behaviorism had encountered serious limitations, irrespective of the continued efficacy
of behaviorism—radical or otherwise—in their various domains of application.
I continue to tread on dangerous terrain by entering this conversation with no formal training in psychology. But
it would seem to me that the contemporary interest in instructional efficacy meant that there were opportunities for
probing more deeply into the psychological mechanisms behind PSI's success. For instance, if one were willing to break
with the epistemic traditions of radical behaviorism, one might utilize ethnographic observation to study the interac-
tion between students and proctors, or the manner in which constructive behaviors for studying and test taking were
shared (both circulated and reinforced) among the students. It might have also been interesting to observe the student
reactions when data about their progress were posted on a chart visible to the other students; the different reactions
might have provided important clues as to the bimodal distribution in grades. Carrying out other such observations
could have generated a more nuanced list of the behaviors and reinforcers that operated within a PSI classroom. More-
over, if one were to break even further from the epistemic traditions of radical behaviorism, researchers could have
interviewed the students themselves and analyzed their articulations about the motivational machinery of PSI.13
In addition, there were opportunities to analyze data that were not so far removed from the epistemic traditions of
radical behaviorism. For instance, the time-dependent data on individual progress could have been analyzed in ways
that were not all that different from the work that was done with animal experiments. Moreover, this data could have
been aggregated to produce comparative data on rates of learning. Such work could have led to further observations
about the benefits of mastery-based teaching strategies. Yet, from the historical records, it does not appear that such
work took place.14
There are several possible reasons why those who studied PSI failed to move this direction. First, being trained
in a laboratory-based discipline, the radical behaviorists who were involved with PSI may have had a hard time see-
ing PSI as a possible site for observation, since the subjects played such an active part in constructing the behavioral
dynamics within the classroom. It is interesting in this context that Keller's list of behaviors and reinforcers were based
mainly on his design of PSI and not, it would seem, what he might have observed in the classroom. (To be fair, few who
12 AKERA

contributed to educational research at the time were studying learning in social versus individual terms [Haggis, 2009].)
Second, once radical behaviorists brought their ideas into the domain of educational research, and developed a peda-
gogic method that was as efficacious as PSI, it may have seemed more important just to focus on diffusion. Sherman's
criticism notwithstanding, it was precisely the work that validated the efficacy of PSI that contributed most to its cir-
culation. For better and for worse, this also fit comfortably within the epistemic traditions of educational research
(Lagemann, 2002). It is indicative that the two projects at his Center that Sherman described in his essay both had
to do with circulation, not analysis; they were the publication of the PSI Newsletter, and an initiative that allowed the
Center to function as an information clearinghouse for PSI workshops, conferences, and publications (Sherman, 1974,
p. 63).
Radical behaviorism as a whole had unfolded with great enthusiasm and fervor. Skinner's terminology, which tried
assiduously to avoid references to internal mental states, spilled over beyond a basic theoretical commitment to both
define and limit the scope of the research questions and observational practices deemed appropriate by radical behav-
iorism's most serious adherents. For those bound by the ideology of radical behaviorism, it may have remained difficult
to break from the shackles of their core epistemic practices.
I suspect all of these factors were at work, and furthermore, that the adherence to doctrine was not inconsequential.
This was hardly the only reason for PSI's downfall—the method was simply too taxing on all but the most committed
instructors. Still, without a robust scientific understanding of PSI's efficacy, it remained difficult to control the outcomes
or justify the resources that a proper implementation of PSI required. As we shall see in the final section below, this left
PSI vulnerable to the personal and administrative costs of the method that ultimately worked to limit its circulation.

5 THE CHALLENGE OF SUSTAINING A REFORM EFFORT

As PSI's implementation spread to faculty members unfamiliar with the psychological foundations of the method,
reports from the field began to grow more mixed. In 1974, the University of Texas’ El Paso campus and New Mex-
ico State University held a joint symposium on PSI. The symposium was set up to help promote the method, but the
organizers had to admit that the general tenor of the meeting was “not one of enthusiastic optimism, but mild disap-
pointment and discouragement” (Carver, 1974, p. 448f). The principal organizer concluded that existing articles about
PSI were too positive, setting up false expectations. Instructors who were hoping for a miracle found the intense work
required to set up a PSI course to be a real “eye opener.” And those who launched their course without having care-
fully prepared their course in advance set themselves up for disaster, as they struggled to produce study guides, chock
full of errors, which impinged negatively on the student's learning experience—and hence reinforcement. Also misun-
derstanding the fifth principle of PSI about student-teacher interaction, which was really meant to apply only to the
proctors, many instructors embraced PSI expecting more rewarding interactions, only to find that they were spending
too much time “writing study guides and exams to work closely with [the] students” (Carver, 1974, p. 449).
Compounding the difficulties was the epistemic habit that many engineering educators had—having been born of
a culture of invention—of trying to engineer a better solution. For instance, John T. Sears from West Virginia Univer-
sity tried to apply PSI to his junior-level thermodynamics course, despite having no proctor. Also wedded to the idea
of a grade distribution, Sears only guaranteed the students a C for finishing all of the units. Additional work, evalu-
ated using traditional methods, was used to assign students a higher grade. Sears also felt it necessary to give students
“careful criticism” of their homework, which meant that they received feedback after some delay. This ran against the
principle of providing instant feedback, which was considered essential from the point of view of operant conditioning.
Because there were no lectures and no proctors, Sears also repurposed class time, during which they were supposed
to be preparing for the unit tests, with design exercises that added an additional learning objective to an already over-
burdened course. When students began falling behind, Sears instituted specified deadlines for every two weeks during
the semester (Sears, 1971). This was a clear instance of what Koen and Keller worried about, a compromised, “SLI”
implementation of PSI that could only discredit the method (Koen, 2011a, seg. 37; Koen, 1974).
AKERA 13

This example also points, if indirectly, to the broader institutional arrangements that limited PSI's circulation. Even at
UT Austin, Koen struggled with the difficulties associated with working within the constraints of an academic calendar.
The fact that all courses started and ended at a specified date made it impossible to operate with what one proponent
labeled, “learning as a constant and time [spent on a course] as the variable.” PSI's proponents regarded this as the
ideal way to run a PSI course, given the different backgrounds, innate abilities, and intrinsic interests of each student
(Flammer, 1971). Meanwhile, the “incompletes” that Koen and his colleagues frequently assigned to students as a way
of getting around this problem drew unwanted attention from the registrar, who eventually prohibited the practice. In
addition to the envy that PSI's grade distribution generated, having other faculty within a department who were not
committed to the method meant that the opportunity for integration, which was especially important for a mastery-
based educational method, remained limited.15 Soon, other vocal critics joined the fray, including those troubled by
the significant effort students placed in a PSI course, supposedly at the expense of their other courses. All this cast
additional doubt on PSI as a potential solution to the broader challenges of engineering education (Gessler, 1974; Koen,
2011a, seg. 35; Robertson & Crowe, 1975).
All in all, PSI flashed across the horizon during a brief period between the late 1960s and early 1970s when student
demands, faculty interest, campus environment, and administrative expectations aligned to create an institutional con-
text conducive to an emphasis on teaching—at least across a certain range of U.S. universities. But two oil crises and
the resulting decade of economic “stagflation” during the 1970s brought engineering educators to shift their attention
to a perceived crisis in U.S. industrial productivity. Given the intense work required to set up a PSI course, few faculty
members from research universities (and from universities with high teaching loads) were willing (or able) to give PSI a
try. The situation only grew worse during the 1980s, as the concern about productivity transformed into a broader dis-
course about “national competitiveness.” Engineering research received greater attention as the necessary means for
transitioning to a “high tech” economy. At UT Austin, as elsewhere, it became less fashionable for engineering faculty
to express a strong interest in teaching (Buskist, Cush, & DeGrandpre, 1991; Grant & Spencer, 2003).
PSI did continue to be employed at UT Austin, having secured the institutional support necessary to propagate its
epistemic culture.16 Nevertheless, the fate of the method at UT Austin remains indicative of broader, national trends.
Given the inflated OPEC oil prices, Texas actually weathered the recessionary period of the 1970s quite well, and had
made major investments in higher education amidst significant economic and demographic growth.17 However, the
collapse of the OPEC cartel and the resulting drop in oil prices between 1981 and 1983 had a devastating effect on
the Texas economy. A 26 percent cut in the state's higher education budget was seriously considered for the 1986–
1987 fiscal biennium (Bennett, 1985). A shift toward research was already occurring by the late 1970s (in no small
part because Texas had reached peak oil in 1972), but it was this crisis that prompted the state to place even greater
emphasis on research, and a high tech economy modeled after Silicon Valley and Route 128. It was said that educated
minds would become “the oil and gas” of Texas’ future economy (White, 1985; “Texas past and future,” 1981).
The state's most concerted bid to enter the high tech era occurred through its successful bid to bring the Microelec-
tronics and Computer Technology Corporation (MCC) to Austin. MCC was the nation's first research consortium, said
to be the U.S. response to the Japanese “Fifth Generation Project” in computer architecture and artificial intelligence.
The firm was headed up by none other than a former director of the National Security Agency, Admiral Bobby Ray
Inman (Ret.). The proposal affected the University of Texas directly, since a key commitment to the winning bid, beyond
the $20 million building erected to house this private corporation, was a promise to upgrade UT Austin's research pro-
file by creating 30 new faculty lines, especially in fields such as electrical engineering and microelectronics (MCC, 1991;
“MCC commitments for Department of Electrical Engineering,” n.d.).
Koen's own career reflected this turn of events. While Koen was tenured early, in just three years, it took him longer
to attain the rank of Full Professor. This was partly because of time spent away from UT Austin—Koen spent several
years at the French Atomic Energy Commission's main research laboratory in Saclay, France, where he pursued work
on reactor safety that earned him international recognition.18 However, the tenure standards at Austin were sup-
posedly such that he needed to build his vitae through work performed at the university. It is also worth noting that
Koen's vitae contains a mixture of research publications in nuclear engineering and those pertaining to PSI and educa-
tional research. In the end, it was probably his teaching that earned Koen his promotion. McKetta's successor, Ernest F.
14 AKERA

Gloyna, placed him on the list of candidates for promotion to Full Professor in 1982, the year Koen won a prestigious
statewide teaching award. Then, amidst some backlash generated by the university's heavy emphasis on research, UT
Austin's Vice President for Academic Affairs placed Koen at the top of the list of candidates for promotion from the
College of Engineering. While PSI continues to be used at UT Austin in somewhat modified form (much to Koen's cha-
grin), even Koen came to think carefully about how to advise young, untenured faculty who were interested in teaching,
but had to meet the university's evolving tenure bar (Koen, 2011a, seg. 47, 72).

6 CONCLUSION

On the surface, this is a story about the rise and fall of a pedagogic experiment. Yet, embedded within it is also a study
of the epistemic shifts that accompanied radical behaviorism's extension into one field of application. Here in the con-
clusion, we can review these changes within those who developed PSI, and those who adopted it for engineering edu-
cation. I continue to draw on Knorr Cetina's notion of epistemic cultures, focusing not only on scientific ideas, but the
heterogeneous assemblages that include the social and institutional relations, machinery, practices, and representa-
tions that are used to produce and warrant knowledge. I also consider the significance of broader ideological contexts
in shaping this epistemic change, and close with some observations about the contemporary implications of this study.
We can more clearly see the epistemic changes that occurred with PSI through a comparison with the behavioral
pharmacologists described by Campbell in Discovering Addiction. In the beginning, the radical behaviorists described by
Campbell entered the field with the youthful confidence that “Skinnerian behaviorism seemed to explain everything
about the way the world worked” (Campbell, 2007, p. 179, quoting Roland Griffith). Behavioral pharmacologists ini-
tially ignored the work that had taken place at the Addiction Research Center (ARC) in Lexington, KY because this
work relied on self-reported human subjects data and as well as statistics for such things as measuring the severity of
withdrawal. The behavioral turn in drug addiction research did make significant contributions to the field by introduc-
ing, for example, the idea of drug receptors. Nevertheless, as this work progressed, there were those without a deep
commitment to radical behaviorism who began drifting toward an instrumental and methodological stance towards
behaviorism by relying, for instance, on the old ARC data to bolster their findings. Understood in terms of epistemic
cultures, this meant accepting an epistemologically hybridized form of explanation. Close and frequent encounters
with experimental subjects also brought some behavioral pharmacologists to view drug addiction and readdiction as
something that could be better explained through the history of an individual's drug experiences and their accrued atti-
tudes toward drug use, reintroducing ideas about “mind” that had been exorcised during the early, radical behaviorist
interpretations of addiction (Campbell, 2007, chap. 7).
To an extent, Keller and Sherman followed a similar trajectory. Political developments in Latin America created
an opportunity to bring radical behaviorism to Brazil, where the call for modernization and a preexisting physiology
lab made it possible to replicate the epistemic culture of radical behaviorism in a foreign context. Radical behavior-
ism's application to teaching nevertheless necessitated some accommodation. Beginning with the minor adjustments
that occurred with Keller's “code voice method,” Keller, and his Brazilian colleagues developed PSI in a way that broke
from radical behaviorism's early experimental traditions, especially through the use of generalized reinforcers. Those
seeking to warrant PSI's efficacy also turned to self-reported statistical data, which in their case arrived fully formed
through the well-established epistemic traditions of educational research.
Meanwhile, the Cold War engineering workforce crisis partly remade engineering education into a domain of exper-
imental knowledge. And the focus on student retention, lifelong learning, and teaching effectiveness offered a highly
receptive context for PSI. Radical behaviorism and its experimental practices were therefore quickly absorbed into
the epistemic culture of engineering education, especially by a group of educators who sought to cast their pedagogic
efforts as research.
Yet, there were also limits to this epistemic adaptation. As discussed above, Keller and Sherman's actions belied
a continued commitment to some of the core principles of radical behaviorism, especially its observational practices.
AKERA 15

For engineering educators, meanwhile, a range of institutional constraints (such as an academic semester, or lack of
support for course development), along with a disciplinary compulsion to improve upon known solutions when faced
with new constraints, prevented PSI from reaching a stable form that functioned reliably in engineering education.
Perhaps the main difference between PSI and behavioral pharmacology was the simple fact that the latter work
took place in the context of funded research, which demanded the work to remain valid as psychological research. By
contrast, the radical behaviorists and engineering educators who turned to PSI did so largely in instrumental terms,
the former as an application of their ideas, the latter as a means of student retention, effective instruction, and the mild
hope that their work would be recognized in promotion and tenure decisions. However, this meant that there was noth-
ing in this disciplinary encounter that prompted either party to delve more deeply into the psychological foundations
of learning. As such, PSI remained, at best, a trading zone instrumental to the interests and epistemic traditions of the
different participants, not a vital arena for interdisciplinary research capable of generating its own research questions
and operating under a more fully hybridized epistemic culture. For how many of behaviorism's other applications might
this have been the case?19
This story of PSI also makes a rich series of connections between historical context and scientific content. Clearly,
the broader, political history of the Cold War era facilitated, and ultimately hindered the development and diffusion
of PSI. On the other hand, I place no particular emphasis on a specific type of interaction between context and con-
tent. While the politics of Brazilian modernization was instrumental to the diffusion of radical behaviorism, it was the
more direct affinity between Goulart's leftist agenda and behaviorism's commitment to the idea that all organisms learn
through similar mechanisms for reinforcement that facilitated radical behaviorism's entry into undergraduate peda-
gogy. This latter, political association nevertheless did not persist during PSI's diffusion into U.S. engineering education,
even though radical behaviorism's core ideas regarding learning remained embedded within the package of ideas and
methods that made up PSI. Nor was a particular type of context given privileged explanatory standing. The epistemic
cultures of engineering education and educational research were as important a “context” for absorbing and extending
PSI as were Cold War politics. This is consistent with more recent directions in Science and Technology Studies, which
grants no privileged ontological standing to the different forms of context that are said to shape scientific ideas (Hack-
ing, 1999; Latour, 1993); a focus on epistemic cultures and their heterogeneous assemblage in fact serves as a vehicle
for more fluid accounts of mutual influence across different domains of knowledge, where the weight of different kinds
of contexts emerges out of the broader flows of history, and historical narratives.
However, in addressing the practical implications of this study, the relative dearth of the “flow” between undergrad-
uate education and the psychological foundations of learning is something that educators and behavioral psycholo-
gists may want to reconsider. Especially given the emphasis that universities and accreditation agencies are again plac-
ing on the quality of undergraduate education, there is reason to ask whether it is sufficient for educational research
to remain of the superficial sort criticized by Ellen Lagemann in An Elusive Science: The Troubling History of Education
Research (2002). It is interesting to note that when I describe this story to engineering educators who are struggling
with distance learning and “flipped classrooms” today, they immediately recognize the close parallels between PSI and
their own work: both function by breaking learning up into small, discrete units, and rely on immediate feedback to
motivate students.20 Upon further discussion, it becomes clear that the main difference between present-day distance
learning environments and PSI is the sense of isolation so often felt in the former, and the robust social interactions
in the latter. This kind of thinking could lead to new ideas about how to remake learning environments, both distance
learning and otherwise.
Likewise, for behaviorists, advances in cognitive science, neuroscience, and learning theory appear to offer new
opportunities for a more integrative view of human behavior, and through it, more precise accounts of the behavioral
dynamics that operate within learning environments. Whether or not any of the speculations I offered earlier prove
useful—I am not qualified to assess this—there are surely opportunities to do more in terms of exploring the psycholog-
ical foundations of learning, especially when it comes to behavioral approaches to enhanced learning, as was the case
with PSI, as with other mastery-based approaches to learning. What this account of PSI suggests is that such explo-
rations might require a more flexible stance toward one's epistemic commitments, and furthermore, that we are in
need of institutional arrangements for this kind of work. Extant surveys of educational and learning research already
16 AKERA

point to the promises of this kind of interdisciplinary convergence (Haggis, 2009; Jonasse & Land, 2000; Mitchell &
McConnell, 2012). It would be great to see such work unfold.

ENDNOTES
1
Specifically, Campbell demonstrates how Charles Schuster's scientific and political commitments were influenced by his
exposure to, and participation in drug culture while performing in Jazz clubs prior to his scientific training. This led him to
view drug addiction as a conditioned behavior, as opposed to a moral failing (Campbell, 2007, chap. 7).
2 Unless otherwise indicated, biographic information on Billy Vaughn Koen is obtained from an oral history interview in the
author's possession, Koen (2011a).
3
See esp. A history of the teaching improvement program. In Engineering Teaching Effectiveness Colloquia 1966-1967, and other
documents in UT College of Engineering Records, 1923–2000 (hereafter, CoE Records), Box CDL3-A12. University of Texas
Archives (hereafter, UT Archives).
4
Notably, Teodoro Ayllon started his work on behavioral psychology in an educational setting—the campus preschool at the
University of Houston (Rutherford, 2009, pp. 65f, 72).
5
As noted by one of my reviewers, single subject experiments in radical behaviorism are often a misnomer. The reviewer
notes that research articles in applied behavioral analysis are generally published with no fewer than three subjects,
and that they are designed with within-subject replication, citing Johnston & Pennypacker (2008). For Keller's, a sin-
gle subject experiment, along with an inexplicit reference to age and gender politics, was sufficient to unseat prevailing
assumptions.
6 There is no ready academic source in English on the history of Brasilia or its university, but relevant information may be found

on the História de Brazília entry at the Brazilian Wikipedia site. Online. http://pt.wikipedia.org/wiki/História_de_Brasília
(Accessed February 8, 2014).
7 Under PSI, Bs and Cs were often awarded to students who completed some specified portion of the units.

8
I note that this is my inference, not Koen's.
9
Keller, Sherman, and Koen all recognized that reinforcement existed for other actors in the system including instructors and
administrators.
10
One of the reviewers also pointed out the difference between the way generalized reinforcers operate in PSI and within
token economies. It is, however, worth considering the extent to which grades do operate as contrived reinforcers within
our society.
11
Personally, I found Derek Blackman's (1995) description of the subtleties of Skinner's position to be informative.
12
This position is found in any of a number of major investigations carried out by the American Society for Engineer-
ing Education, and its precursor, the Society for the Promotion of Engineering Education (SPEE). See for example SPEE
(1930).
13
In again assisting me with the many subtleties of behaviorism, one of the reviewers pointed out that even research in the
radical behaviorist tradition did not always exclude subjective measurements, especially when addressing the “social valid-
ity” of their work. See Wolf (1978). This piece by Wolf includes a reference to how those working on PSI relied on student
self-reported data on how much students liked a “Keller-type PSI” course (p. 206).
14 A contemporary review of the work on PSI can be found in Kulik and Kulik (1975); and Kulik, Kulik, and Cohen (1979). A
more recent review of subsequent work on PSI may be found in Grant and Spencer (2003).
15 The challenges of working within an existing administrative structure are noted by Koen (2011a, seg.37) as well as by Hobe-

rock (1979). They are also reported more dramatically by Heimbach (1971), and tabulated by Hess, Edwards, Greenspoon,
Heydinger, and Soares (1976).
16 A review of the ongoing work on PSI may be found in Grant and Spencer (2003). There are also variant approaches, such
as interteaching (Boyce & Hineline, 2002), which are organized around behaviorist principles, and drew specific inspiration
from PSI.
17 A record of the shifting emphasis within UT Austin's College of Engineering can be found in the annual reports and planning

documents in CoE Records, CDL3-A13. UT Archives.

18
Koen was at Saclay, France between 1971 and 1972, and again between 1976 and 1977. There he developed a canonical
method for assessing reactor safety that utilized artificial intelligence and an extension of the computational methods he
utilized while at MIT (Koen, 2011a, seg. 51).
19
Here we are drawing a distinction between trading zones and a fully mature domain of interdisciplinary research that goes
beyond a trading zone. See Galison (1997), chap. 9.
AKERA 17

20
Indeed, those who have continued to use and study PSI have explicitly recognized the connection between PSI and distance
learning. See, for example, Grant and Spencer (2003).

REFERENCES
A history of the teaching improvement program. In College of Engineering, Engineering teaching effectiveness colloquia, 1966–67
(pp. 9–15). UT College of Engineering Records, 1923–2000 (hereafter, CoE Records), Box CDL3/A12, University of Texas
Archives, Briscoe Center for American History, University of Texas at Austin (hereafter, UT Archives).
Bennett, S. (1985, June). The state of our state. Texas business. Un-paginated copy in RG 100 Legislature, AC 1989/76, Box 2/17.
Texas State Archives, Austin, TX.
Bjork, D. W. (1993). B. F. Skinner: A life. New York, NY: Basic Books.
Blackman, D. (1995). B. F. Skinner (1904-1990). In R. Fuller (Ed.), Seven pioneers of psychology: Behaviour and mind (pp. 109–132).
London: Routledge.
Boyce, T. E., & Hineline, P. N. (2002). Interteaching: A strategy for enhancing the user-friendliness of behavioral arrangements
in college classroom. Behavior Analyst, 25, 215–226.
Burns, E. B. (1980). A history of Brazil (2nd ed.). New York, NY: Columbia University Press.
Buskist, W., Cush, D., & DeGrandpre, R. J. (1991). The life and times of PSI. Journal of Behavioral Education, 1, 215–234.
Campbell, N. (2007). Discovering addiction: The science and politics of substance abuse research. Ann Arbor: University of Michigan
Press.
Carver, W. L., Jr. (1974). A realistic appraisal of first efforts at self-paced instruction. Engineering Education, 64, 448–450.
Cirino S. D., Miranda R. L., & de Cruz R. N. (2012). The beginnings of behavior analysis laboratories in Brazil: A pedagogical view.
History of Psychology, 15, 263–272.
Costall, A. (2006). “Introspectionism” and the mythical origins of scientific psychology. Consciousness and Cognition, 15, 634–
654.
Flammer, G. H. (1971). Learning as the constant and time as the variable. Engineering Education, 61, 511–514.
Foecke, H. (1964—1965). Engineering degrees (1963-64) and enrollments (Fall 1964) in institutions with one or more ECPD-
accredited curricula. Journal of Engineering Education, 55, 190–203.
Galison, P. (1997). Image and logic: A material culture of microphysics. Chicago, IL: University of Chicago Press.
Gessler J. (1974). SPI: good-bye education? Reader comment. Engineering Education, 65, 252–255.
Golinski, J. (1998). Making natural knowledge. Cambridge: Cambridge University Press.
Grant, L. K., & Spencer, R. E. (2003). The personalized system of instruction: Review and applications to distance education.
International Review of Research in Open and Distance Learning, 4(2), 1–17.
Hacking, I. (1999). The social construction of what? Cambridge, MA: Harvard University Press.
Haggis, T. (2009). Student learning research: A broader view. In M. Tight, K. H. Mok, J. Huisman, & C. Morphew (Eds.), The
Routledge international handbook of higher education (pp. 23–35). New York: Routledge.
Heimbach, C. L. (1971). Personalized instruction in mechanical engineering. Engineering Education, 61, 506–507.
Hess, J., Edwards K. A., Greenspoon J., Heydinger R. B., & Soares C. (1976). Major problems identified in the use of the PSI method.
In Keller & Koen (Eds., 1976), topic 5, n.p.
Hoberock, L. L. (1975) The professor's viewpoint. Engineering Education, 66, 167–168.
Hoberock, L. L. (1979). To PSI and back. Engineering Education, 69, 399–401.
Hoberock, L. L., Koen B. V., Roth C. H., & Wagner G. R. (1972, February). Theory of PSI evaluated for engineering education.
IEEE Transactions on Education, 15, 25–29.
Johnson, J. M., & Pennypacker, H. S. (2008). Strategies and tactics of behavioral research (3rd ed.). New York: Routledge/Taylor &
Francis.
Jonassen, D. H., & Land, S. M. (Eds.). (2000). Theoretical foundations of learning environments. Mahwah, NJ: Lawrence Erlbaum
Associates.
Keller, F. S. (1943). Studies in international Morse code: I. A new method of teaching code reception. Journal of Applied Psychol-
ogy, 27, 407–415.
Keller, F. S. (1958). The phantom plateau. Journal of the Experimental Analysis of Behavior, 1, 1–13.
Keller, F. S. (1968). Good-bye teacher. Journal of Applied Behavior Analysis, 1, 79–89.
18 AKERA

Keller, F. S. (1974a). An international venture in behavior modification. In F. S. Keller & E. Ribes-Inesta (Eds.), Behavior modifica-
tion: Applications to education (pp. 143–155). New York, NY: Academic Press.
Keller, F. S. (1974b). PSI and reinforcement theory. In Keller & Sherman (1974) (pp. 50–58).
Keller, F. S. (2009). At my own pace. Cornwall-on-Hudson, NY: Sloan Publishing.
Keller, F. S., & Koen, B. V. (Eds.). (1976). The personalized system of instruction: State of the art 1976. Austin, TX: Engineering
Institutes, University of Texas at Austin.
Keller, F. S., & Sherman J. G. (1974). The Keller Plan handbook. Menlo Park, CA: W. A. Benjamin.
Knorr Cetina, K. (1999). Epistemic cultures: How the sciences make knowledge. Cambridge, MA: Harvard University Press.
Koen, B. V. (1970) Self-paced instruction for engineering students. Engineering Education, 60, 735–736
Koen, B. V. (1971). Self-paced instruction in engineering: A case study. IEEE Transactions on Education, 14(1), 13–20.
Koen, B. V. (1974). The Keller plan. In L. P. Greyson & J. M. Biedenbach (Eds.), Individualized instruction in engineering education
(pp. 37–42). Washington, DC: American Society for Engineering Education.
Koen, B. V. (2011a). Oral history interview by author 17 July 2012, Brunswick, ME. MP3 audio-file, numerically indexed.
Koen, B. V. (2011b). Résumé, updated 1 September 2011. In author's possession.
Koen, B. V., & Schmidt, K. J. (2001). The professor and the media laboratory: A case study in web-based course creation. Frontiers
in Education Conference, Session F3F.
Kulik, J. A., & Kulik C. C. (1975). Effectiveness of the personalized system of instruction. Engineering Education, 66, 228–
231.
Kulik, J. A., Kulik, C. C., & Cohen, P. (1979). A meta-analysis of outcome studies of Keller's personalized system of instruction.
American Psychologist, 34, 307–318.
Lachney, M. M., & Nieusma, D. (2015). Engineering bait-and-switch: K-12 recruitment strategies meet university curricula and
culture. Paper ID #13481. Proceedings of the ASEE Annual Conference and Exposition, 26.616. 1–16.
Lagemann, E. C. (2002). An elusive science: The troubling history of education research. Chicago, IL: University of Chicago Press.
Lamb, P. (1975). The chairman's viewpoint. Engineering Education, 66, 166–167.
Latour, B. (1993). We have never been modern. Cambridge, MA: Harvard University Press.
Leahey, T. H. (2004). A history of psychology: Main currents in psychological thought (6th ed.). Englewood Cliffs, NJ: Prentice-Hall.
Lemov, R. (2005). World as laboratory: Experiments with mice, mazes, and men. New York, NY: Hill and Wang.
Longino, H. (2013) Studying human behavior: How scientists investigate aggression and sexuality. Chicago, IL: University of Chicago
Press.
Matos, M. A. (1998). Contingências para a análise comportamental no Brasil, Psicologia: USP, 9, 89–100.
MCC. (1991). Executive summary: the Texas incentive for Austin. RG 301 [Office of the Governor], 1991/141 Mark White
Records, Box 6, Fld: MCC / Adm. Inman's Statement on Location in Austin. Texas State Archives, Austin, TX.
MCC commitments for Department of Electrical Engineering. (n.d.). CoE Records, Box CDL3-A2, Fld: Micro Electronics & Tech-
nology Corp. UT Archives.
Mitchell, A. W., & McConnell, J. R. (2012). A historical review of Contemporary Educational Psychology from 1995 to 2010.
Contemporary Educational Psychology, 37(2), 136–147.
Railroad Commission of Texas. (n.d.). History of Texas initial crude oil, annual production and producing wells, web. Retrieved
from http://www.rrc.state.tx.us/data/production/oilwellcounts.php
Roberson, J. A., & Crowe C. T. (1975). Is self-paced instruction really worth it? Engineering Education, 65, 761–
764.
Rutherford, A. (2009). Beyond the box: Skinner's technology of behaviour from laboratory to life, 1950s-1970s. Toronto: University
of Toronto Press.
Schneider, S. M., & Morris, E. K. (1987). History of the term radical behaviorism: from Watson to Skinner. Behavior Analyst, 10,
27–39.
Sears, J. T. (1971). Developing intellectual skills in a self-paced course. Engineering Education, 61, 515–516.
Shapin, S. & Schaffer, S. (2011). Leviathan and the air-pump: Hobbes, Boyle, and the experimental life. Princeton, NJ: Princeton
University Press.
Sherman, J. G. (1974). PSI today. Keller & Sherman (1974), 59–64.
Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. New York, NY: Appleton-Century-Crofts.
AKERA 19

Skinner, B. F. (1968). The technology of teaching. New York, NY: Appleton-Century-Crofts.

Society for the Promotion of Engineering Education (SPEE). (1930). Report of the investigation of engineering education, 1923–
1929. Volume I. Pittsburgh, PA: SPEE.
Sterling Olmsted Award. (2016). Web. Retrieved from https://www.calvin.edu/academic/engineering/aseeled/olmsted/
Stice, J. E. (Ed.). (1971). The personalized system of instruction (PSI): The Keller Plan applied in engineering education. Bureau of
Engineering Teaching, Bulletin No. 4., CoE Records, Box CDL3-A12. UT Archives.
Teare, B. R., Jr. (1960–1961). Wanted: Leadership in engineering education. Journal of Engineering Education, 51, 8–13
Texas past and future: A survey. (1981). Office of the Governor, Texas 2000 Project. CoE Records, Box CDL3-A14. UT Archives.
White, M. (1985). Governor's opening address, Select Committee on Higher Education. Meeting. Transcript, 14 October 1985.
Select Committee on Higher Education Records, Box 1, Folder 1. Texas State Archives, Austin, TX.
Wiener, D. (1995). B. F. Skinner: Benign anarchist. Boston: Allyn and Bacon.
Wolf, M. M. (1978). Social validity: The case for subjective measurement or how applied behavior analysis is finding its heart.
Journal of Applied Behavior Analysis, 11, 203–214.

How to cite this article: Akera, A. Bringing radical behaviorism to revolutionary Brazil and back: Fred Keller's
Personalized System of Instruction and Cold War engineering education. J Hist Behav Sci. 2017;00:1–19.
https://doi.org/10.1002/jhbs.21871