U&iittiif^i.
HD28
.M414
-3t
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
Beyond Persistence:
understanding the commitment of pioneers in
emerging fields of science and technology
Michael A. Rappa
Massachusetts Institute of
Koenraad Debackcre
Gent
Rijksuniversiteit
Technology
December 1991
Sloan
WP # 3388-92
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
1
Massachusetts Institute of Technology
BEYOND PERSISTENCE:
UNDERSTANDING THE COMMITMENT OF PIONEERS IN
EMERGING FIELDS OF SCIENCE AND TECHNOLOGY
Koennad Debackere
Michael A. Rappa
Massachusetts Institute of
Technology
Rijksuniversiteit
December 1 99
1991
Sloan
Gent
WP # 3388-92
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
Alfred P. Sloan School of
Management
Technology
50 Memorial Drive, E52-538
02139-4307
Cambridge,
Massachusetts Institute of
MA
M.I.T IJ3P
\^AR
9 1992
1
Beyond Persistence:
Understanding the Commitment of pioneers in
Emerging Fields of Science and Technology
Michael A.
RAPPA and Koenraad DEBACKERE'''
Massachusetts Institute of Technology
December 1991
abstract
This paper examines pioneering scientists
and
and
the factors that influence
an area of research, and compares their
experience with those scientists who follow in their footsteps once the field
becomes legitimated. Evidence is presented from an international survey of
more than seven-hundred researchers working in the field of neural
their choice
persistence in
networks. After controlling for age
of the
and
professional experience, the results
group are indeed different in both
enter and their persistence to remain in the field. In
analysis suggest that pioneers as a
their motivations to
generaL compared with other
scientists,
pioneers are more deeply influenced
by the intrinsk intellectual appeal of the field and are
social dynamics of the research community.
less
influenced by the
introduction
Sociological
investigations
of scientific and technological
communities have
contributed a great deal to our understanding of the research and development process.
area of investigation at the genesis of the research process,
is
why
the question of
choose to pursue the topic of research they do, embracing certain problems
while disregarding others as inconsequential
—
any explanation of the emergence of nascent
scientists
important
the so-called "problem of problem choice"
(Zuckerman, 1978; Gieryn, 1978; and Ziman, 1987). The answer to
to
as
One
fields
this
question
of science and technology,
as
is
pivotal
it
lies at
the very heart of the matter. For whatever reason, by choosing to apply their energy to one
set
of problems and not another, scientists can thereby steer the frontier of knowledge
new and
different direaions.
A common
between
'
approach to unraveling the problem of problem choice
factors
Michael Rappa
Debackere
in
is
is
to distinguish
influencing scientists' decisions that are internal and external to the
an assisunt professor
in
the behavioral and policy sciences area of the
a Fulbright post-doctoral fellow at
Gent, Belgium.
is
MIT
and
MIT
Sloan School. Koenraad
a research associate at the Vlerick School, Rijksuniversitcii
—
community (Cole and Cole, 1973; Zuckerman,
scientific
science include both the cognitive
and the
social.
1978). Factors internal to
Recognizing the influence of the cognitive
side of the equation
on problem choice, Zuckerman (1978) points
scientists' theoretical
commitments and the
their
judgment
in
as "the art
Good
of finding a solution. Scientists use
determining problems that are intellectually interesting and yet
tractable given the state of current
of science
feasibility
to the importance of
knowledge and technique. Recalling Medawar's theme
of the soluble," Zuckerman quotes:
study the most important problems they think they can
scientists
solve. It
is,
after
all,
their professional business to solve
merely grapple with them.
The
problems, not
spectacle of a scientist locked in
with the forces of ignorance is not an inspiring one
the scientist is routed (Medawar, 1967).
if,
in the
combat
outcome,
Ultimately, scientists seek solutions to problems they perceive as significant to
"science"
is
it
and that have a good chance of being found, preferably sooner than
that informs their
holds that
it
is
judgment of what
is
indeed significant?
The
the cognitive struaure of the
what
field.
By
virtue
from the chaff Problems
of their training,
at the periphery,
is
But what
argument
wonhy
of
problems of central imponance
scientists are
a theoretical fi-amework that guides their research agenda, allowing
them
of reasoning, the problem of problem choice
is
less relevant.
to
equipped with
to sort the
wheat
problems whose answers are unlikely
contribute to the core cognitive struaure, are perceived as
this line
internalist
the logic of the scientific process itself that defines
research; that science continually generates interesting
later.
to
As a consequence of
purely one of intellect.
Suspicious of the self-contained nature of such an explanation, sociologists have sought
to infuse an appreciation of the social
dynamics of research communities, thereby making
any investigation of problem choice behavior, well, a
bit
more problematic. Careful
consideration must be given to the social interaction of researchers in influencing their
choice of problems, including the effect of the reward system of science,
and
social stratification.
nature,
While
Merton (1938) and
than being
made
is
true that, as
others have
shown
Medawar
claims, scientists grapple with
that they also grapple with each other. Rather
an ainight cognitive realm, where "good"
in
important problems
it
are, the
communal norms,
problem choice decisions of
scientists
know what
scientists are influenced
the
by peer
perceptions, competition for rewards, and the opinions of those scientists in a position of
authority.
Not only
are scientists guided
by theory
in
choosing problems, they are
influenced by each other's problem choices and opinions regarding what
with the opinions of some
scientists
within a research community.
is
important
having more weight than others, given their stature
not social influences reinforce or distort the cognitive influences on
^X'l^cther or
scientists' decisions
undoubtedly
is
a critical question for the sociology
clearly, the potential for distortion exists.
fields
is
The
of science. But
periodic exponential growth of scientific
well-documented (Crane, 1972; Menard, 1971) and not always easy
solely in terms
of the underlying theoretical importance of
tendency for "fashionable" problem areas to emerge
Hagstrom, 1965). Fashion
scientists,
is
contemptible insofar
guided more by perceptions of what
When
to science.
is
as
it
Rather, there
a field.
in science (Barber,
to explain
is
a
1990; Crane, 1969;
represents a situation
popular, mistake what
where
truly important
is
popularity and importance do not coincide, scientists are given the choice
of going along with their peers or breaking from the pack and reconciling themselves to
less
glamorous problem areas that are
relatively starved for recognition
and funds.
Attention to faaors external to science that might nonetheless influence the problem
choices of scientists was precipitated by an awareness of the tremendous industrial and
military implications of science,
The
called "big science."
size
and the attendant economic cost of what Price (1963)
and importance of
frontier research in
science necessitates a greater degree of administrative planning
and control, which
consequendy places heavy constraints upon the freedom of individual
their
own
scientists to
choose
research agenda. At the very least, scientists' perceptions of important problems
are influenced
by military and corporate imperatives;
have assumed control over the decision process
problem choice an organizational
In sorting out the
kind of
some branches of
scientist:
those scientists
problem of problem choice,
initiate
field
extreme, such imperatives
and have thus made the problem of
issue to be resolved in
one who pioneers a new
who
itself
at the
it is
committee.
interesting to focus
and continue working
choose to enter a new
term (Rappa and Garud, 1991). In
field,
his
in a field
of research before
Among
only a small portion will stay with
know
that the struggle for acceptance
is
likely to
be arduous,
but they are not prepared for how long the process actually takes, nor for
the amount of misunderstanding and misconstrual that inevitably ensues.
Tenacity
is
it
is
as
widely
the scientists
it
for the long
study of pioneering scientists in systematic zoology,
Hull (1988) observes:
Scientists
a particular
of research. Pioneers are defined here
perceived as significant, or perhaps even legitimate by their peers.
who might
on
as essential as originality in science....
Anyone promoting
a
nonstandard view must be prepared to write the same paper, fight the
good fight, answer the same criticisms over and over again.... Failure in
science is more often a function of the lack of resolve than anything else.
Thus, by virtue of their unconventional problem choices and unrelenting determination,
may
pioneering scientists
Who
are pioneers
ultimately lead the
and what
is
it
way
in creating a
that informs their vision of
what
is
new
research specialty.
significant research?
Are
they influenced in ways that are different from their non-pioneering colleagues in making
their
problem choices?
from
criticism
Why
their peers?
do they
Given
persist in
working
Although the study of emergent
scientists
is
problem
fields
is
of science and technology has occupied a
in the literature, empirical
research
limited, particularly with respect to their
at
understanding factors that influence the
development but do not single-out pioneers
scientific
of science and
worthy of carefril examination.
on the nature of pioneering
problem choice behavior. For
example, in their account of the emergence of eight different
(1976) aim
areas in the face of
their central role to progress at the frontier
technology, the problem choice behavior of pioneers
prominent position
in
Lemaine,
fields,
rate, direction,
et al.,
and content of
for systematic analysis.
Even
general studies in the sociology of science, pioneers have been curiously absent
in
more
among
the
various kinds of scientists that have merited special attention. Instead, research has been
focused on highly productive scientists (Long, 1978; Allison and Long, 1990), highly
influential
scientists
scientists,
who make major
among whom Nobel
it
may
as
determined by citation patterns (Cole,
1968), "marginal"
contributions (Gieryn and Hirsh, 1983), and scientific
laureates figure prominently
elites,
(Zuckerman, 1967 and 1977). Although
very well be the case, pioneers are not necessarily the most produaive, nor the most
influential,
nor marginal, nor the most prominent of
scientists.
Thus, a focused
investigation of pioneering scientists appears warranted.
PIONEERS IN NEURAL
The
NETWORK RESEARCH
present study was undertaken in order to understand in greater detail
what
influence pioneers in their choice of a research agenda and their persistence in the
seleaed
as
fields
one
networks holds no special significance other than the
opportunity to do so presented
field,
is
We
of science and technology that we are studying concurrently. The
decision to examine neural
network
field.
the basis for this examination the field of neural networks research, which
of nearly a dozen
factors
we concluded
it
itself first.
After a preliminary investigation of the neural
would be
interesting to
with a primary focus on pioneering researchers.
conduct a comprehensive study
A
of the
neural network
human
a type
By using
brain.
certain features that
is
make
it
example, a neural network
of information processing system that
a biological
unique
is
in
model
form and
in
its
is
inspired by models
design, a neural network system has
fianction
from conventional computers. For
not programmed in the usual sense, but rather
it
trained
is
with data. Tliis implies that the computational performance of a neural network improves
with experience:
as
it
more and more information
processes
becomes increasingly more accurate
in
performing a
in
response. Another feature
its
parallelism in processing a task. Unlike a normal
computer with
is
its
task,
it
degree of
a single or small
number
of sophisticated central processing units, a neural network has a very large number of
simple processing elements that operate simultaneously on a computational problem.
These
features allow
it
to
perform certain tasks that otherwise might be very
difficult
using existing computer technology. Neural networks are also referred to as conneaionist
systems, adaptive systems, or neurocomputers. For further details, refer to the recent report
by
DARPA
(1988).
Neural networks have a long history of development, stretching back to theoretical
explanations of the brain and cognitive processes proposed during the 1940s. In the early
years, researchers formulated
they then used to explore
and elaborated upon
phenomena such
random networks. By the 1960s
as
basic
adaptive stimulus-response relations in
perceptron was considered a watershed, but
criticism
intelligence.
became seen
at the
from researchers more interested
The
as
implement neural networks,
there were several efforts to
the most notable being the single-layer "perceptron."
for
models of neural computing that
Among
neural network researchers the
same time
in
it
served as a lightning rod
the burgeoning field of anificial
idea of neural networks, as exemplified by the perceptron, quickly
almost antithetical to the symbolic reasoning principles of
intelligence. Critical analysis
to proclaim that the
researchers to waste
of the perceptron led some highly respeaed AI researchers
concept was fundamentally flawed, and
much
artificial
on the
effort
subject.
may have
antagonists of neural networks
By
inappropriate for
as such,
casting doubt as to
its
legitimacy,
effectively dissuaded other researchers
from
entering the field in larger numbers (Minsky and Papen, 1988).
The
controversy surrounding neural networks notwithstanding,
the early 1970s with perhaps no
field.
Undeterred
in their belief
more than
light
by researchers
hundred
researchers
worldwide
in the
of the potential of neural networks, their persistence over
the next decade eventually paid-off.
new
a few
work continued during
By
in a variety
the 1980s, neural networks began to be viewed in a
of disciplines, such that the
field
soon achieved
a
position of legitimacy within the scientific
A
community.
professional society for neural
network researchers was formed, specialized journals and books were published, and the
a scries of international conferences were held.
first in
exaaly
why
While
it
is
difficult to explain
the perception of the field changed so dramatically, at least four important
technical events can be discerned: (1) the evolution of the single-layer perceptron into a
multi-layer system;
(2)
the rapid development of related technologies that enabled
researchers to develop, simulate,
and diagnose neural networks of greater sophistication;
(3) significant progress in theoretical
understanding of neuro-biologicaJ processes; and
(4)
the contributions of researchers pursuing the idea of parallel distributed processing, the so-
PDP-group. In
called
light
of these developments,
as well as others, interest in the field
became widespread, such that the number of researchers working on neural networks
expanded
rapidly.
By the end of
from a few hundred to
several
the decade the size of the field swelled in
thousand researchers worldwide.
TTie evolution of the neural network research
typical
for
of emerging
new
fields
fields in
some of
its
community
is
not unusual and
sociological characteristics.
to lack widespread acceptance for long periods,
controversy, other times simply being ignored by researchers. But
fields
fields
we have examined
so
far.
It is
may
fairly
even be
common
sometimes attracting
when
tend to grow rapidly. This panern has occurred, to greater or
of the dozen
membership
they do catch on,
each
lesser extent, in
Given the recent experience within the
field
of
neural networks, this case presents us with an excellent opportunity to examine in great
detail the
who
behavior of pioneering researchers relative to large numbers of researchers
follow in their footsteps.
METHOD
Through an
conference proceedings for the two-year period from 1988 to 1989,
we
we were
From
Given the scope of the research community,
questionnaire was determined to be the most appropriate
method of
activities, (b) their decision to
begin working on neural networks,
might lead them to cease their neural network research
interaaion with the
rest
in favor
a survey
investigation.
twelve-page questionnaire in English was sent to researchers inquiring about
(d) their
this
able to determine the exact address for each of 2,037 researchers in
thirty-five different countries.
network
more
identified
than 3,000 researchers worldwide working on the subject of neural networks.
material,
and
analysis of published sources, including books, journal articles,
(a) their
(c)
A
neural
faaors that
of another problem
of the neural network research community, and
area,
(e) their
demographic
Additional
arising
characteristics.
tests
among
The
were conduaed
in
those respondents for
questionnaire was pretested in the United States.
Europe
whom
to reduce potential interprctational difficulties
English
Since there were thirty-seven researchers with
is
a
second language.
more than one address during
the time
period considered, a total of 2,074 questionnaires were mailed in February 1990. After the
third
week of data
on computer
collection,
we mailed
a follow-up lener
and posted e-mail messages
network researchers of the survey.
bulletin boards to alert neural
questionnaires, 162 were returned as undelivered by the post office.
seven researchers with
questionnaires.
later,
more than one address were represented
Of the
None of
in
2,074
the thirty-
the undelivered
At the completion of the data collection period approximately ninety days
720 of the 1875 questionnaires presumed
to be delivered
returned, yielding a final response rate of 38.4 percent.
Some of the
were completed and
faaors that
may have
affected the response rate include: the length of the questionnaire, the global scope of the
survey,
and the
institutional mobility of researchers.
DATA
Validity checks
In order to rule out apparent self-seleaion biases,
made
to
demographic comparisons were
determine whether the survey population departs significantly from the
respondent sample. In particular, first-order comparisons between the two groups were
made with
respect to the geographic location of a respondent's institutional affiliation,
type of institutional affiliation, and disciplinary background.
First, a
geographic comparison was made clustering the respondent sample and the
survey population into four regions: the Americas, Europe, the Far East, and the Middle
East.
Of the 720
respondents, 63 percent reside in the Americas
(all
but a few percent
in
the U.S.), 25 percent in Europe, ten percent in the Far East, and about two percent in the
Middle East (mainly
Israel).
These percentages coincide almost perfealy with the survey
population (X^=5.24, d.f=3,
smallest category (the
A
second
test
Middle
n.s.).
East)
Similar results were achieved
from the
when omitting
the
test.
compared the resp>ondent sample and survey population with
respect to
the institutional affiliation. Respondents were classified into three categories: universities,
commercial firms, or other types of institutions (mostly government ftinded laboratories
that are not university-based).
Among
the
720 respondents, 452 (63 percent)
are affiliated
with academic laboratories, 177 (25 percent) are employed in commercial firms, while 91
(12 percent) arc engaged in other types of institutions.
statistically significant
comparison
reveals that
no
departure exists between the respondent sample and the survey
population (x^=5.6l, d.f.=2,
A
A
n.s.).
final, albeit less precise, test
compared the
background of the sample
disciplinary
respondents with those of the survey population. Although respondents indicated their
disciplinary backgrounds, for the survey population
disciplines ft-om their postal address
when
we were only
able to infer researchers'
a departmental affiliation
was provided
(e.g.,
researchers belonging to electrical engineering departments were classified as electrical
engineers).
for
Upon
careful
inspeaion of the survey population, the disciplinary background
about 1,500 researchers were found. Using
significant difference
disciplines
this data,
when comparing respondents with
we were unable
and engineering
(7 percent),
find a
the survey population.
most represented among respondents include
percent), physical sciences (19 percent),
to
engineering (36
electrical
computer science (18 percent),
The
biological sciences
mathematics (7 percent), and psychology and cognitive science
(5 percent).
Sample
characteristics
The 720 respondents
are
employed
in
220
different
academic institutions, 101
commercial firms, and 62 other (mostly governmental) types of organizations. There are
89 respondents
who
report
more than one employer. The
large majority of respondents (82
percent) hold a doaorate or are in the process of obtaining one.
Only 16 percent
holding a master's as their highest academic degree, and just two percent are
a bachelor's degree.
The
distribution of academic degrees
respondents' current positions of employment.
The
is
at
report
the level of
further reflected in the
majority of them are university faculty
(38 percent) or hold a scientific post (32 percent), ranging from staff scientist to chief
About 17 percent of
scientist.
the respondents are students, virtually
all
of
whom
are in
doaoral programs. Regarding other employment positions, only seven percent of the
respondents report themselves
The remaining few
The
in
as
being an engineer, while just
percent of respondents
average respondent
is
37
fall
into several other
years old (s.d.= 9.2 years).
five
percent are managers.
employment
categories.
The respondent sample
ranges
age from 22 years for the youngest to 69 years for the oldest, with the youngest quanile
between ages 22 and 30 and the oldest quartile between ages 43 and 69. The median age
35
years.
For the 702 respondents
who
specified the year they began
is
working on neural
networks, the average length of time in the
length of time in the field
field
is
The median
6.2 years, (s.d.= 6.1 years).
four years, with the range between one and 40 years.
is
A
variable measuring the length of a respondent's professional experience (not necessarily in
the field of neural networks) was created based on the time elapsed since graduating with
the highest degree.
years).
The
average length of professional experience
The median number of years
graduates) to
46
years.
is
six,
The quanilc with
with a range from zero
is
8.7 years (s.d.= 8.8
(i.e.,
students and recent
the least professional experience has between zero
and one year of exf>erience; the one with the most experience has between 14 and 46
years.
Idmtifying pioneers
Since the focal point of this study
is
on understanding the problem choice motives of
pioneering researchers and their persistence in the
pioneers are
is
a critical issue.
be classified as a pioneer
is
Our
field,
identifying exactly
based upon
marked by a period of rapid growth
when
in
a respondent enters the field of neural
is
and books, and an international conference.
of the establishment of neural networks
of research within the broader
scientific
community. Thus,
for separating pioneering researchers
from the
rest
field
community
membership during the 1980s and by the advent of
a professional society, specialized journals
as indicators
the
determination of who in the respondent sample should
networks. As previously discussed, the evolution of the neural network
These events serve
who
as a legitimate
a point of demarcation
of the sample should
lie
somewhere
in
shown
in
this period.
The cumulative
Figure
1,
illustrates
the change over time in participation in the
cumulative growth pattern
onwards.
which
distribution of the entry year for each respondent,
it
is
clear that the field
The number of respondents
the 1960s and 1970s.
is
field.
From
this
grew most rapidly from about 1984
entering the field each year was
About 25 percent of the respondents entered
much
lower during
the field prior to 1984,
whereas about 75 percent of them entered from 1984 to 1990. Both the historical overview
of neural network research and the entry pattern illustrated
point of demarcation for identifying pioneers
During the period from 1981
As
a
1
suggest that the
somewhere between 1980 and 1985.
to 1984, a total of 91 respondents entered the field.
consequence of our analysis of
the 165 respondents (25 percent)
the remaining
is
in Figure
who
historical events in the field,
to classify
entered the field by 1983 as pioneers, while placing
537 (75 percent) who entered
group provides us with a means
we decided
after
1983 into the control group. The control
for contrasting the responses
of pioneers.
10
100.0
75.0
n
c
Ji
B
3
3
z
£.
50.0
t>
25.0
"
c.
•a
n
-3
FIGURE
1:
Cumulative number of survey respondents entering the
field of neural networks
Pioneers
We
and
and cumulative percent.
their decision to enter the field
began our analysis of pioneers by examining respondents' motives for
choosing neural networks
Respondents were asked to
as a
problem area of research
rate the degree
(see
Appendix
initially
for core items).
of importance of numerous factors that might
have influenced their decision to work on neural networks.
The
faaors reflea the diversity
of potential internal and external influences on researchers that have been discussed
previously in the literature.
A
first-order analysis based
differences between pioneers
enter the field in
all
on
t-test
comparisons points to
and the control group with respea
but two often items (see Table
1).
The
statistically significant
to their motivations to
strength of the differences
between pioneers and the control group were further scrutinized using two control variables
in
an analysis of covariance procedure. Because
influenced a respondent's decision to enter the
we
are interested in the factors that
field,
we used
control variables that
refleaed the state of reality at the time the decision was made.
The
first
entering the
covariate concerns the respondent's professional experience at the time of
field.
The
average
number of
before staning neural networks research
is
years of professional experience for pioneers
2.3 years (N = 156),
compared with 4.9
years
11
12
(N = 518) for the control group
disproportionate
number of
academic degree; that
is,
(pdifr.<-001)- ^^
is
particularly interesting to note that a
pioneers entered the field prior to receiving their highest
while they were students.
percent) entered before graduation and another
1 1
A
total
of 98 of die
1
56 pioneers (63
entered the same year in which they
graduated. In comparison, 200 of the 518 respondents (39 percent) in the control group
entered before graduation and another 41 entered the year of graduation. This distribution
is
significantly different across both groups
The second
(x2=29.6 with d.f.=2 and p<.001).
covariate concerns the respondent's age at the time of entering the field.
average, pioneers are four years younger in age
respondents in the control group.
The
when beginning work on
among
Pearson correlation coefficient
The
results
show
neural networks than
average age at entry for pioneers
(N = 160), compared with 31.8 years (N = 532)
(r
is
=.85)
is
significant (p<.001).
of the analysis of covariance with each covariate are shown in Table
that the first-order differences reported in Table
control variables. Thus,
The main
result
we
1
27.8 years
group (pdifF.<001)- The
for the control
the control variables
On
2.
The
data
are sustained in the case of both
find further support for treating pioneers as a distina group.
from the
analysis of entry items
is
that substantial differences between
pioneers and the control group can be found, despite the apparent differences in their age
and professional experience. In other words, the
analysis enables us to
of whether or not the first-order differences indicated in Table
motives of researchers
who
are attracted to the field
once
1
answer the question
reflea changes in the
has gained legitimacy, or
it
merely changes in the demographic characteristics of researchers. The analysis of
covariance clearly shows that the faaors influencing pioneers in entering the field cannot be
explained solely in terms of differences in age or professional experience relative to
respondents in the control group.
The
support
Table
proposition that pioneers are a distinct group of individuals received ftirther
when examining two
2).
One
on the entry
set
sets
of questions concerning the charaaeristics of the
of questions pertains to the influence of the
field's size
field (see
and growth
rate
decision, while the other set pertains to the financial cost of doing research in
neural networks and the
amount of specialized knowledge
appears that perceptions about
on the decisions of pioneers
community
to
size
enter the
required to
and growth
field.
Once
rate
work
in the field. It
have a different influence
again, statistically significant
differences remain even after controlling for the respondent's age
experience at the time of entry. Using the same analytical approach,
and professional
we
investigated the
13
14
influence of financial and intelleaual requirements on the respondent's entry decision. As
shown
in
Table
no difference could be found between groups with respect to these
2,
factors.
For
all
of the analysis so
who were doaoral
found
far
reponed, we repeated the procedures omitting respondents
students or recent graduates at the time of the survey.
in the original analysis persist
The
with one deviation ("potential for peer recognition"
non-significant, with p=.053), thereby ruling-out the possibility that a large
students in the control group
may
in
number of
work environment when entering the
field
terms of the suppon and encouragement of colleagues and supervisors
there was one at the time).
The
is
adversely affea the response pattern.
In a final series of questions, the resp)ondent's
was examined
differences
results
of the first-order
t-tests are
shown
in
Table
3.
(if
The
data indicate no difference between pioneers and the control group insofar as the awareness
of colleagues or supervisors
encouragement
is
concerned. Although there
fi-om sujservisors, this difference
is
a difference in terms of
is
no longer
statistically significant
students and recent graduates are omitted from the analysis.
With respea
encouragement from within the respondent's organization, there
is
when
to collegial
no difference between
the two groups. However, encouragement from colleagues outside the respondent's
organization
In addition,
tests,
statistically significant,
is
when
even after omitting students and recent graduates.
controlling for age and professional experience in analysis of covariance
the difference remains statistically significant. Respondents in the control group
more encouragement from
report receiving
colleagues outside their organization than
pioneers do.
The
persistence
ofpioneers
Careful examination of the historical development of the neural network community,
as well as
other
persist in their
fields,
shows that not
commitment
while important,
is
all
researchers
to a field. Therefore,
— including those who
by
itself,
enter early
the decision to enter a field,
not sufficient for understanding pioneering behavior. Given the
proposition that pioneers enter early and stay with
we
it,
investigated the influence of
various internal and external factors that might lead a respondent to cease working on
neural networks (see the Appendix).
Examining the
issue
investigate pioneers in terms of their present attitudes
of jxrsistence also enabled us to
— something
that
is
not possible
15
>
i 1-
<
16
when
inquiring about past decisions to enter the
the entry and exit motive questions, which
The
results
of the first-order
t-test
Thus, there
field.
imporunt
is
When
a complementarity to
to the design of the study.
comparisons are shown
Table
in
The
4.
and the control group
indicate statistically significant differences between pioneers
items but one.
is
omitting students and recent graduates fi-om the analysis, the
data
for
all
results
remain unchanged. Moreover, an analysis of covariance using the respondent's age and
professional experience provides fiirther evidence confirming the first-order differences
between groups, with the only exceptions being the "overcrowding of the
lesser extent, the "increased cost
It
of research."
(see
Table
field" and, to a
5).
should be noted that, contrary to the previous analysis of covariance
age and professional experience at the time the respondent entered the
field,
respondent's present age and professional experience as covariates.
The
pioneers
is
The
control group (pdifT.<.001).
pioneers
is
compared with 35.1
presently 42.3 years {s.d.= 8.7),
average
presently 13.4 (s.d.= 9.7),
number of
compared with 7.2
respondent
Table
5.
A
is
to the subject
first-order
does serve
as a
average age of
years (s.d.= 7.9) for the control
is
is
.87 (p<.001).
working
in
a rather difficult question
rudimentary indicator of
how committed
of neural networks. The findings confirm the
Mann-Whitney non-parametric
is
willing to continue
the field, given the current rate of progress. Admittedly, this
it
use the
years of professional experience for
Another question addresses how long the respondent
and answer, but
we now
years (s.d.= 8.3) for the
group (pdiff.<001). The Pearson correlation coefficient between covariates
to pose
which we used
in
test reveals that
the
results reported in
pioneers repon a
willingness to persist longer with their neural network research agenda than the respondents
in
the control group: a
median of
five-to-ten years as
opposed
to two-to-five years
(p<.001). Funhermore, controlling for age and professional experience,
Mann-Whitney
comparisons were repeated by grouping the respondents into "cohorts" based on age and
experience (Table 6). Each cohort consists of 20 percent of the respondent sample.
results
of this analysis show that
and the control group
is
it
is
of each cohort, the difii:rence between pioneers
statistically significant,
professional experience (p=.068).
cohon,
in the case
When
with the exception of
COHORTI
percent, with similar results.
based on
examining the median for both groups within each
apparent that pioneers intend to continue longer in the
colleagues in the control group do.
The
The 20
field
than their
percent cohort criterion was changed to 25
17
18
19
20
A
final
question asks respondents to indicate whether neural networks
research interest, a major interest, one of several interests, a
interest in
keeping informed. From
this question,
stature of neural networks research to other
it
is
minor
statistically significant
(p=.0012).
test
an
possible to ascertain the relative
problem areas
A Mann-Whitney
their only
interest, or just
for pioneers
group. Table 7 shows that pioneers more frequently report neural networks
or major research interest.
is
indicates
and the control
as their
that this
primary
difference
is
21
which of the two approaches
is
used and no matter what year
is
chosen to identify pioneers,
the overall results remain the same. All of the discriminant functions are statistically
significant
(p<.001).
consistently
obuin
Regardless of the year used to create the pioneer group,
a Wilks'
X of about 0.88 on
we
the entry motive items and about 0.89 on
the exit motive items. This suggests that the choice of the cut-off year
robust across the
is
relevant time interval.
DISCUSSION
The
forgoing analysis provides some insight into the character of pioneers and
from colleagues
differ
in their decision to enter
respect to the choice of neural networks as their
and
persist in a field
problem
how
they
of research. With
area, pioneers are
more deeply
influenced by the intrinsic appeal of the subjea matter than are respondents in the control
group.
The
pioneers rate
less
"intellectually compelling nature of the field"
more highly than the control group (p<.001).
is
in
In the
faa the only item that
same
vein, pioneers are
influenced than the control group by a lack of other topics (p<.001)
neural networks. This result shows clear
suppon
for the internalist
when choosing
argument that
sees
cognitive influences as the predominant faaor in a researcher's choice of problem area.
Pioneers are far
less
when deciding
concerned with what
to
enter the field.
is
happening
in the rest
of the research community
For example, pioneers rate
as
significantly less
imponant: "positive opinions of leading researchers" (p<.001), "success of other
researchers" (p<.001),
and "potential
for peer recognition" (p<.05). Pioneers are also less
influenced by the size and growth rate of the field (in both cases, p<.001). Furthermore,
they are
less
influenced by practical maners, such
and the potential
as
the availability of funding {p<.001)
for financial rewards (p<.01).
Pioneers appear to follow
more
closely the classic textbook explanation
of problem
choice in science. In contrast to the control group, they are drawn to the field by their
and
intellectual interests
are rather oblivious to the social forces at play in the research
community. Moreover, even though pioneers
field,
are
predominantly students when entering the
the first-order results remain after controlling for age and professional experience
with one exception:
eliminating the weak significance of "opportunity to build
a
company." Thus we can rule-out the presumable youth and professional inexperience of
students as a plausible explanation for the differences
motives of pioneers.
that
is
at the heart
It is
we
find in the
problem choice
not simply a case of youthful idealism, but rather something
of pioneering behavior.
else
22
While the evidence with rcspca
to the respondent's decision to enter the field strongly
faa that
favors a view of pioneers as a distinct group, the
years prior to the survey
the field
may
is
an important consideration.
distort perceptions of
what has
The
this decision
several
passage of time since entering
faa influenced
in
was made
a pioneer's decision to
work on neural networks. Are pioneers simply remembering things
Undoubtedly, the element of time makes pioneering behavior
understand. However, by examining the persistence of pioneers,
all
we
the
differently?
more
are able to
difficult to
complement
the analysis of problem choice with an analysis of situational perceptions that are
contemporaneous with the survey period
for
all
respondents.
TTie uniqueness of pioneers as a group can be seen equally as clear in their response to the
importance of faaors that might lead them to leave the
are different
from the control group
statistical difference
field
in every respect
of neural networks. Pioneers
but one. Although there
with respect to the importance of diminishing intelleauaJ challenge,
pioneers consistently
show
concern regarding
less
all
other factors. Even so, pioneers seem
to be motivated by the intrinsic qualities of neural networks as an intelleaual endeavor
to be
no
is
single-minded in their determination. Neither slow progress
(p<.001), nor rapid progress in alternative
has as
fields,
does to the control group (p<.001). Regarding
influenced by the lack of interest
among
in solving
much importance
communal
factors,
to
and
problems
them
pioneers are
as
it
less
other researchers (p<.001) or the unfavorable
opinions of leading researchers (p<.001). Furthermore, the lack of funding, lack of financial
rewards, negative opinions of supervisors and the discontinuance of neural networks at the
respondent's organization are each significantly
p.<.001).
Here
less
important to pioneers
(in all cases,
again, as with the decision to enter the field, the results are not substantively
changed when considering the respondent's age and length of professional experience.
Pioneers have a passion for their chosen subjea, which
remain in the
field.
Not
only
is
their interest less
pioneera also intend to endure in the
field for a
is
also
diminished by a number of faaors, but
longer period of time.
cohorts and controlling for age and professional experience,
cohorts, in
networks.
all
we
networks, considering
it
divided into
find that pioneering
field
of neural
more focused on
further shows that most pioneers are
neural
their major, if not their only, research interest.
picture of pioneers that emerges fi-om the data
is
choosing an area of research, pioneers are more influenced by their
is
When
but one case, express a longer term commitment to the
The evidence
The broad
born-out in their intent to
intelleaually interesting and
less
fairly consistent.
own
When
perceptions of what
concerned with the aaions and opinions of others,
as
23
well as the practical realities of funding
and rewards. Likewise, with respect
persistence in the field, pioneers are equally single-minded in their perspective.
to their
While the
evidence seemingly supports the internalist view of the primacy of cognitive influences in
the problem choices of pioneering scientists, the findings also have implications for
understanding the motives of
evidence suggests that
the
community, play
The
with
who
group that
this
enter a field once
social forces,
communities
pioneers
is
legitimated.
The
both internal and external to
of the neural networks research community arc compelling, fUrther work
To
this
end, two other research
are currently being examined. Second, understanding the subtle behavior of
quite difficult in the context of a cross-seaional survey such as this one. Instead,
is
required
is
a long term study of pioneering that follows the decision process of
researchers over time.
initial
is
problem choice.
a larger role in
required to test the external validity of these findings.
what
it
exploratory nature of this investigation must be emphasized. First, although the
results in the case
is
it is
scientists
Given the
large
number of respondents,
it is
feasible to transform the
survey into a longitudinal design by periodically following-up with the respondents
to see if they
remain
study. Nonetheless,
in the field.
it
is
We
imponant
have already
commenced work on such
to recognize the limitations
a follow-up
of survey methods
in
examining pioneering behavior.
CONCLUSION
Pioneering researchers play a central role in the emergence of
technology.
enter
and
The
results
of our analysis show that pioneers are unique
chosen
persist in their
colleagues, pioneers are
found
intelleaually compelling
research
new
and
community and by
to
less
be more influenced by their
of science and
in their
motivations to
When compared
of neural network research.
field
fields
own
with their
perceptions of what
is
influenced by the opinions and actions of others in the
the availability of funding and rewards. After controlling for
age and professional experience, these differences continue to hold, even though most
pioneers in the sample were students
The
results
examine
when
they entered the
field.
of this study point to several avenues for funher research.
One
avenue
is
to
institutional differences in pioneering scientists and, in particular, the differences
that exist
between academic and commercial laboratories. Another avenue
the characteristics of students
dissimilar
who
pioneer a
new
from that of their professorial colleagues.
field
is
to investigate
and how they are similar or
REFERENCES
AND
ALLISON, PAUL D.
SCOTT LONG. 1990. "Depanmental
J.
Produaivity," American
Barber, Bernard.
COLE,
JONATHAN
Sociology
Effects
on
Scientific
Sociological Revinu. Vol. 55. pp. 469-78.
1990. Social
StuMes of Science. London: Transaaion Publishers.
1968. "Pattetns of IntellcCTual Influence in Scientific Research,"
R.
of Education. Vol. 43. pp. 377-403.
COLE. JONATHAN R. AND STEPHEN COLE. 1973. Social Stratification in
University of Chicago Press.
COLE, STEPHEN, RUBIN,
L AND JONATHAN R
Science," Scientific American.
CRANE, DIANA. 1969. "Fashion
Science.
Chicago:
COLE. 1977. "Peer Review and the Suppon of
Oaober. pp. 34-41.
in Science:
Does
it
Exist?" Social Problems. Vol. 16. pp.
433-41.
CRANE, DIANA. 1972.
Invisible Colleges: Diffiision
of Knowledge in
Scientific
Communities.
Chicago: University of Chicago Press.
DARPA 1988. Neural Network
GIERYN,
THOMAS
F.
Study. Fairfax, Virginia:
AND RICHARD
F.
AFCEA
International Press.
HIRSH. 1983. "Marginality and Innovation
in
Science," Social Studies of Science. Vol. 13. pp. 87-106.
GIERYN,
THOMAS
1978.
F.
"Problem Retention and Problem Change
Sociobgical Inquiry. Vol. 48, pp. 96-1
HagstrOM, WARREN
HARGENS. LOWELL
L.
Science,"
in
5.
O. 1965. The Scientific Community.
New York:
Basic Books.
AND DIANE
Stratification in Science."
Hull, Davtd
1
H. FELMLEE. 1984. "Structural Determinants of
American Sociological Review. Vol. 49. pp. 685-97.
An
Evolutionary Account of the Social
Conceptual Development of Science. Chicago: University of Chicago Press.
L.
1988. Science as a Process:
LEMAINE, GERARD, ROY MACLEOD, MICHAEL MULKAY, AND PETER WEINGART
Perspectives on the Emergence of Scientific Disciplines. The Hague: Mouton
LONG,
SCOTT. 1978. "Productivity and Academic Position
American Sociobgical Review. Vol. 43. pp. 889-908.
J.
MEDAWAR.
P.B.
1967. The Art of the Soluble. London:
MENARD, HENRY W.
MERTON, ROBERT
New York:
1971. Science: Growth
K. 1938. Science,
Harper
&
Row
1976.
Company.
in the Scientific Career,"
& Company.
and Change. Cambridge: Harvard Univ.
Technobgy and
[1970].
Methuen
(eds.).
&
and
Press.
Society in Seventeenth Century England.
25
MINSKY, MARVIN AND SEYMOUR PaPERT. 1988. Perceptrons:
Computational Geometry. Expanded Edition. Cambridge, Mass.:
PRICE,
DEREK DE SOLLA. 1963. LittU
Science,
Big Science.
New
An
Introduction
MIT
to
Press.
York: Columbia University
Press.
MICHAEL A. AND Raghu Garud. 1991. "Using the Literature in the Study of
Emerging Fields of Science and Technology: An Examination of Researcher
Rappa,
Contribution Spans,"
ZIMAN.JOHN.
1987.
MIT
working paper, Quly).
"The Problem of 'Problem Choice,'" Minerva. Vol.
25.
Numbers
1-2.
pp. 92-106.
ZUCKERMAN, HARRIET.
1967. "Nobel Laureates in Science: Panerns of Productivity,
Collaboration, and Authorship," American Sociological Review. Vol. 32. pp. 391-403
ZUCKERAUN, HARRIET. 1977.
York: The Free Press.
ZUCKERMAN, HARRIET.
Scientific Elite:
Nobel Laureates in the United
States.
New
1978. "Theory Choice and Problem Choice in Science,"
Sociological Inquiry, Vol. 48, pp. 65-95.
26
APPENDIX
Items inquiring into the respondent's decision
How
to
enter the field
important were each of the following factors in influencing your
networks
your research agenda (respondents were asked
important) -scale with midpoint 4 (somewhat important)}.
in
to circle
decision to include neural
initial
on a 1 (not at all important)
to
7
(very
compelling nature of neural networks
1.
intellectually
2.
lack of other promising research topics
3.
availability
4.
potential for financial rewards
5.
potential for recognition by peers
6.
dissatisfaction
7.
positive opinions
8.
successes of other researchers with neural networks
9.
opportunity to build a commercial enterprise
10.
opportunity to solve an important societal problem
of funding for neural networks research
with a previous research agenda
of leading researchers
Items inquiring into
How
of neural networks:
what might lead
in the field
the respondent to leave the field
important would each of the following
network
activities (respondents
midpoint
4 (somewhat
were asked
to circle
factors be in diminishing
on a
I (not at all
1.
slow progress in solving technical problems in neural networks
2.
lack of funding for your neural
3.
diminished interest
4.
rapid progress in alternative areas of research
5.
opinions of leading researchers unfavorable to neural networks
6.
negative opinion of your supervisor
7.
discontinuance of neural net
8.
lack of financial rewards
9.
diminished intellectual challenge of neural network research
10.
increased financial cost of conducting neural network research
11.
overcrowding
12.
difficulty in
in
among
network research
other researchers in neural networks
terms of the
(if
any) toward neural networks
activities at
number of
your organization
neural network researchers
keeping up with new developments in neural networks
55'1*3
038
your current interest
important)
important))-.
of neural networks:
to
7
in neural
(very important) -scale with
MIT LIBRARIES DUPl
3
TDflO
0D75b7fl3
4
Date Due
HOVl?
m
BAJou
ii^i'u
Lib-26-67
_MIT LIBRARIES
iiiiiiiiiiiiiii
3
IQflO
0Q7Sb7a3
M