Content uploaded by Julian Park
Author content
All content in this area was uploaded by Julian Park on Jul 27, 2019
Content may be subject to copyright.
Research Policy 30 (2001) 819–836
Why promising technologies fail: the neglected role of
user innovation during adoption夽
B. Douthwaite a,∗, J.D.H. Keatingeb, J.R. Parka
aDepartment of Agriculture, University of Reading, Earley Gate, P.O. Box 236, Reading RG6 6AT, UK
bInternational Institute for Tropical Agriculture (IITA), Oyo Road, PMB 5320, Ibadan, Nigeria
Received 23 July 1999; received in revised form 8 March 2000; accepted 7 June 2000
Abstract
The paper analyses innovation histories of two agro-mechanical and two seed-based technologies with high and low techno-
logical complexity, introduced into simple and complex farming systems in Asia. The main conclusion, which may be seen as
a hypothesis for further testing, is that, as technology and system complexity increase so does the need for interaction between
the originating R&D team and the key stakeholders (those who will directly gain and lose from the innovation) when the latter
first replicate and use the new technology. This is because a successful technology represents a synthesis of the researcher and
key stakeholder knowledge sets, and creating this synthesis requires more iteration and negotiation as complexity increases.
Instead of assuming a new technology is ‘finished’ when it leaves the research institute, a more effective way of developing
complex technologies is for the R&D team to release them as soon as the key stakeholders will adopt, and then nurture the
technology’s continued development in partnership with the keystakeholders. © 2001 Elsevier Science B.V. All rights reserved.
Keywords: Participatory technology development (PTD); Social construction of technology
1. Introduction
In October 1998, the third CGIAR 1System
Review concluded that: “Investment in the CGIAR
has been the most effective use of official devel-
opment assistance (ODA), bar none. There can be
夽This research presented in this paper was carried out while
the first author was employed at IRRI in collaboration with the
Philippine Rice Research Institute (PhilRice), Muñoz, Nueva Ecija,
Philippines and the University of Agriculture and Forestry (UAF),
Thu Duc, Ho Chi Minh City, Vietnam. The research was funded
by IRRI, the Impact Assessment and Evaluation Group (IAEG)
set up by the Consultative Group on International Agricultural Re-
search (CGIAR) and by the German Government through Deutsche
Gesellschaft für Technische Zusammenarbeit GmbH (GTZ).
∗Corresponding author.
E-mail address: b.douthwaite@cgiar.org (B. Douthwaite).
1Consultative Group on International Agricultural Research.
no long-term agenda for eradicating poverty, ending
hunger, and ensuring sustainable food security with-
out the CGIAR” (CGIAR System Review Secretariat,
1998, p. 1). The impact that the CGIAR system has
achieved since its inception in 1971 has largely been
the result of breeding modern crop varieties that re-
spond better to fertiliser and allow higher yields. In
rice, high yielding varieties (HYVs) developed by the
International Rice Research Institute (IRRI) started
the Green Revolution in Asia.
Despite the huge impact of the Green Revolution
the Chairman of the CGIAR System has said that the
research and development (R&D) approach that devel-
oped and disseminated HYVs needs to be adapted to
better deliver technologies that are adopted by small-
holder farmers in complex farming systems (CGIAR
Secretariat, 1997). The Chairman has said that it is not
0048-7333/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S0048-7333(00)00124-4
820 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
enough for CGIAR centres to do their own research,
they must also develop and transfer new technology
generation and diffusion models. The CGIAR sys-
tem review concluded that: “no institution, however,
successful can base its future purely on past perfor-
mance. Progress and relevance come from building on
past strengths and grappling with past weaknesses”.
CGIAR System Review Secretariat, 1998, p. 1).
Both of these views reflect concerns of the donor
community over lack of demonstrated CGIAR sys-
tem impact on poor smallholder households. These
concerns have grown as the priority of the interna-
tional community has moved from food security to
poverty eradication. This change of focus led in 1996
to the Organisation for Economic Co-operation and
Development (OECD) issuing internationally agreed
development targets, the main one being the erad-
ication of 50% of world poverty by the year 2015
(ODI, 1997). Poverty eradication is now a central part
of the CGIAR system’s mission statement (CGIAR
Secretariat, 1999).
The lack of CGIAR impact on the ‘poorest of the
poor’ has been attributed to the cognitive view of
R&D and technology transfer that many agricultural
scientists assumed and used to plan and manage the
innovation process (Chambers and Jiggins, 1986;
Biggs, 1989; Horton and Prain, 1989; Clark, 1995).
Clark (1995) said that the approach is based on the
view of the technology that sees knowledge flowing
through a pipe-line that has basic research activities at
one end and knowledge embodied as useful products
at the other. The research institute at the beginning of
the pipeline is seen as the single source of innovation
(Biggs, 1989). The innovation is then seen to flow
sequentially down the pipeline, without significant
further innovation, with different participants respon-
sible for different parts of the process (Chambers and
Jiggins, 1986; Biggs, 1989; Horton and Prain, 1989;
Clark, 1995). Farmers are seen to either passively
adopt or not adopt but not significantly adapt the new
technology themselves (Rogers, 1995). The job of
extension workers is, therefore, seen as ‘spreading
the message’ about a ‘finished’ product (Ruthenberg,
1985), not as co-development of new technology.
Research is seen as separate from extension. This
cognitive picture has been called the ‘central source
of innovation’ model (Biggs, 1989) and the transfer of
technology (TOT) approach (Chambers and Jiggins,
Fig. 1. The transfer of technology (TOT) model showing the
traditional mental map of how knowledge in the form of technology
is developed and is transferred to farmers.
1986) summarised in Fig. 1. Interestingly, this view
has a direct parallel in the commercial sector where
it has been assumed that product managers develop
product innovations and users passively adopt (Von
Hippel, 1988). Von Hippel found that the assumption
that product managers develop new product inno-
vations is often wrong and other stakeholders-users
and suppliers-can be major sources of innovation.
He concluded that understanding this can help the
management of innovation.
A year later, Biggs (1989) said very much the same
thing in the context of agriculture when he published
a paper describing the ‘multiple source of innovation’
model. The model explicitly recognises that: “Instead
of simply accepting or rejecting an innovation as a
fixed idea (as the central model assumes), potential
adopters on many occasions are active participants
in the adoption process, struggling to give their own
unique meaning to the innovation as it is applied in
their local context” (Rogers, 1995). Resource poor
farmers (RPFs), upon which the CGIAR is being
encouraged to show greater impact, generally have
more complex and diverse farming systems than
those found in more favourable areas (Chambers and
Jiggins, 1986), requiring more local adaptation to
make new technologies work. The ‘multiple source of
innovation model’ is, therefore, better at describing the
technology generation and adoption process amongst
RPFs in unfavourable production environments.
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 821
In rural development, R&D and extension ap-
proaches that are implicitly or explicitly based upon
the recognition of multiple sources of innovation are
often called participatory. There has been a rapid
growth in participatory approaches over the last
decade. IDB (1996) reviewed participatory approaches
used by seven key international organisations, includ-
ing the Canadian International Development Agency
(CIDA), the Organisation of Economic Co-operation
and Development (OECD), the United Nations
Development Program (UNDP) and US Agency for
International Development (USAID). The review
found that the participation as a principle was impor-
tant in all organisations and all had developed their
own definitions and types of participatory approach
(IDB, 1996). IDB’s own definition of participation is:
“Participation in development can be defined in
broad terms as the PROCESS through which people
with a legitimate interest (stakeholders) influence and
share control over development initiatives, and the
decisions and resources which affect them” (IDB,
1996, p. 2).
The CGIAR system will have to be much more
active in developing participatory R&D development
approaches if it is to answer donor concerns about lack
of impact in poor farmers’ fields. Thus, the purpose
of this paper is to facilitate the methodological devel-
opment process by clarifying where traditional R&D
approaches are likely to achieve greater impact, and
why, and where participatory approaches, or a mixture
of both, are likely to be more appropriate.
2. Methodology
The Green Revolution has demonstrated that the
top-down, centrally-controlled TOT approach to inno-
vation has been successful for technology generation
and transfer in relatively simple favourable production
environments (FPEs). However, agricultural resear-
chers have not repeated the success of HYVs with
other types of technology that have been intro-
duced into the same systems, for example, crop and
resource management technologies (Byerlee and
Pingali, 1994). This would suggest that different types
of technology require different research management
approaches even within the same farming system.
But, Kaimowitz et al. (1989) state that CGIAR
centres have used the TOT approach irrespective of
the type of technology transferred.
There is a widely held view that technology can be
usefully thought about in terms of knowledge, and in-
novation thought of as a learning process (Rosenberg,
1982; Mokyr, 1990; Clark, 1995). From this per-
spective the TOT approach can be seen as one that
assumes the researchers can learn enough about a
farming system and then embed this knowledge in a
new technology (hardware) and operating instructions
(software) that is then sufficiently ‘finished’ to require
little or no subsequent local adaptation. Participatory
‘multiple source of innovation’ approaches recognise
that the key stakeholders — those who put the inno-
vation into practice and continue with it — also de-
velop new technology by embedding their knowledge
in it through the adaptations they make. Therefore,
the attribute of the technology that is most likely to
affect the choice between participatory or top-down
approach is the amount of local adaptation expected
by the key stakeholders — the people who are going
to produce, supply and use the new technology. Local
adaptation is part of the “struggling to give their own
unique meaning to the innovation as it is applied
in their local context” that Rogers talks about. It is
not easy. Therefore, the amount of local adaptation
expected is linked to Rogers (1995) concept of tech-
nological complexity — the degree to which an inno-
vation is perceived as relatively difficult to understand
and use.
The discussion above suggests the following
hypotheses:
1. Technologies that require key stakeholders to
change little to make the technology work (i.e.
easy to understand and use technologies intro-
duced into simple farming systems) will have the
fastest impact using a top-down approach.
2. If either the new knowledge requirement or the
system complexity is high then participatory
approaches will be most effective.
Case studies can provide an understanding of the
complex processes that lead to adoption and impact
(GAO, 1987; Yin, 1989; Sechrest et al., 1996). Here,
four case studies are used to describe each of the four
combinations of high and low system complexity and
technology complexity. The case studies are outlined
in Table 1.
822 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
Table 1
Case studies
Knowledge complexity of technology Complexity of system into which technology is introduced
Low High
Low ‘easy-to-employ’ Modern rice varieties (MV) in a favourable
production environment (FPE)
MVs in unfavourable production
environment (UPE)
High ‘hard-to-employ’ ‘Very low cost’ SRR dryer used to substitute
for family labour
Stripper harvester used to
substitute for hired labour
Seed-based technologies, in the form of modern va-
rieties (MV) of rice, provide the two ‘easy-to-employ’
technology case studies. MVs are simple technologies
because nearly all the new knowledge associated with
them is embedded in the seed itself. Farmers need
to learn little to take advantage of MVs because they
can generally be reproduced and grown in similar
ways to traditional varieties. Favourable production
environments (FPEs) generally contain simpler farm-
ing systems compared to less favourable production
environments (LFPEs).
The two ‘hard-to-employ’ technologies are
equipment technologies designed for smallholders.
Equipment technologies are complex compared to
seed-based technology because they require manufac-
turers to learn how to build the machine, and users
often need to acquire new skills to operate the tech-
nology effectively. Moreover, there is large scope to
modify the knowledge embedded in a machine, while
there is no scope for farmers to purposefully change
the genetic make-up of MVs (unless they are trained
in how to do cross-hybridisation).
The two equipment technologies chosen are the
stripper gatherer (SG) harvester adopted in the Philip-
pines and the ‘very low cost’ SRR dryer adopted
in Vietnam. The SG harvester operates in a socio-
economic context made complicated by traditional
labour arrangements designed to redistribute crop to
those without sufficient land to support their own
families. The SRR dryer, on the other hand, operates
in a relatively simple system because it replaces
family labour, not hired labour (Douthwaite, 1999).
Data for both the equipment technology case studies
is reported elsewhere (Douthwaite, 1999) and is based
on actual fieldwork by the first author. The data for the
seed technology case studies comes from published
literature and is as a result more general. The case
studies are constructed and examined to determine:
•the degree to which the invention and innovation
occurred in a traditional or participatory way;
•the amount, quality and sources of innovation that
occurred after first adoption to assess stakeholder
motivations and ability to innovate, and the degree
to which the actual innovation process better fitted
a central- or multiple-source-of-innovation model;
•the degree to which inappropriate choice of inno-
vation models may have affected impact.
Fig. 2 proposes a number of stages in the innovation
process (adapted from personal communication with
R. Yin, 1995). The studies focused on the adaptation
phase, when the key stakeholders first begin to adopt
the technology because, according to Rogers (1995),
this is when most user (as opposed to researcher) in-
novation occurs. The case studies describe the degree
of modification and assess its impact on the ‘fitness’
of the technology. The concept of fitness is borrowed
from evolutionary biology and provides an indication
of the degree to which the attributes of a technology
favour its adoption and use. Rogers (1995) identi-
fied five attributes that explain 49–89% of variation
in adoption rate. One of these attributes is tech-
nological complexity, already discussed. The other
four are:
1. Relative advantage is the degree to which an
innovation is perceived as being better than the
technique it supersedes. Relative advantage is
probably the most important of the five attributes
in explaining adoption rate.
2. Observability is the degree to which the results of
an innovation are visible to others.
3. Trialability is the degree to which an innovation
may be experimented with on a limited basis.
4. Compatibility is the degree to which an innovation
is perceived as consistent with the existing values,
past experiences, and needs of potential adopters.
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 823
Fig. 2. Stages in the innovation process.
Fitness is therefore an average measure of these
attributes.
3. Case studies
3.1. Modern rice varieties in favourable production
environments (simple technology, simple system)
From its foundation in 1962 IRRI was primarily
a rice-breeding institute with the goal of increasing
rice yields with little or no consideration of other
characteristics that rice farmers and consumers might
find important, for example, eating quality. IRRI’s
first major success was with the semi-dwarf variety
IR8, released in 1966 (Holden et al., 1993). This was
followed by a number of ‘finished’ varieties includ-
ing IR20, IR26 and IR36. The high yields and rapid
farmer adoption of the new grain varieties helped
trigger the Green Revolution.
Part of the reason for the rapid adoption of IRRI’s
high yielding varieties (HYVs) was the extension
system IRRI developed. It was based on the Ameri-
can agricultural extension model, which is the oldest
diffusion model in the US, and the most successful
public sector one (Rogers, 1995). This model consists
of three components: researchers at universities; ex-
tension specialists based at the same universities; and
extension agents working with farmers at the local
level. These three levels correspond to the transfer of
technology (TOT) model described by Chambers and
Jiggins (1986), shown in Fig. 1.
The TOT model used by IRRI and later by the
CGIAR system differs from the US model in one crit-
ical aspect — researchers and extension specialists do
not work at the same institutes and do not have the
same status. In the CGIAR system research scientists
work at the CGIAR centres not at the NARS institutes
who are responsible for extension. Also CGIAR sci-
entists are generally better qualified and higher status
— ‘internationally’ recruited rather than ‘nationally’
recruited. The most important role that the CGIAR
system has played is to set the tone for the NARS
(Horton, 1997). The tone the CGIAR system has set
has biased the NARS towards research, not exten-
sion, and this is now reflected in national agricultural
research centres that have modelled themselves on
CGIAR centres (e.g. the Philippine Rice Research
Institute, the Cuu Long Rice Research Institute in
Vietnam and the Chinese National Rice Research
Institute that are all modelled on IRRI).
Despite this shortcoming, the TOT model worked
well for the dissemination of HYVs, partly because of
the unusually favourable nature of the technology has
meant that adoption has occurred without there need-
ing to be good links between research and extension.
The attributes of the HYV technology that made its
dissemination so easy were:
•very high relative advantage, giving much higher
yields than traditional varieties in FPEs;
•the requirement for little or no new knowledge to
make it work, and the fact that the seed could be
easily and cheaply reproduced and transported.
824 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
Scientists in national programs used the IRRI breed-
ing lines to generate hundreds of semi-dwarf varieties
more suited to local conditions. Feedback about sus-
ceptibility to diseases and pests reached IRRI. For ex-
ample, it was found that IR8, the first widely adopted
modern variety, was vulnerable to seven major pests
and diseases. IR8 was followed by a succession of va-
rieties that overcame these problems and in 1982 11
million hectares was planted to IR36, the most widely
planted rice variety in history (Holden et al., 1993).
Average rice yields in South and Southeast Asia in
1991–1993 were 83% higher than those in 1964–1966,
the 3 years immediately preceding the introduction
of IR8. Total production rose by 120% while the
land planted to rice increased by only 21% (IRRI,
1993). IRRI and the TOT methodology of working
with NARS undoubtedly was the key factor in this re-
markable increase in rice production. A second major
factor was that, contrary to some opinion, farmers are
very quick to adopt new ideas and technologies that
are of clear immediate benefit to them (Scott, 1998).
3.2. Modern varieties in less favourable production
environments (simple technology, complex system)
TOT approaches, or ‘classical breeding’ methods
as they are also called, have led to very high adoption
rates of HYVs in the relatively homogenous irrigated
ecosystem. Irrigation helps control growing condi-
tions to suit the uniform needs of these pure lines. The
TOT approach has largely failed to deliver benefits to
farmers in less favourable production environments
(LFPEs), or marginal areas, where agro-climatic con-
ditions are far more varied and far less controllable.
LFPEs include areas that are rainfed, have salty soils,
are flood or drought prone, and areas where rice is
grown as an ‘upland’ crop — like wheat or barley
without flooding and ponding of water to control
weeds. In such areas farmers’ priorities tend to be
subsistence in nature, that is, their objective is to guar-
antee sufficient food each year for their families rather
than grow large surpluses for sale, as farmers in FPEs.
Subsistence farmers tend to be poorer in monetary
terms than their commercially-orientated counterparts.
Hence, the concern that CGIAR centre efforts are not
contributing sufficiently to poverty eradication.
Farmers in LFPEs have developed intricate risk
management strategies to ensure they grow sufficient
food and can save seed each year. Since farming
began over 10,000 years ago farmers have been se-
lecting landraces that are suited to the ecological
niches they farm (Holden et al., 1993). This has cre-
ated a huge diversity evident in the estimated 100,000
varieties of rice that existed in Asia. Farmers select
and choose to plant varieties based on a large number
of factors other than yield. For example, a farmer
might choose to plant a drought resistant variety that
competes well with weeds in their higher fields, and
a higher yielding variety in lower ones where water
supply is more assured. Moreover, the landraces they
plant have greater genetic variability than the ‘pure’
cultivars produced by IRRI and national breeding
programs. This means that while some plants may be
susceptible to a certain pest attack, others, with a dif-
ferent genetic make-up are likely to be resistant. This
makes less likely catastrophic loss due to the whole,
genetically pure, crop failing.
Faced with the realisation that few farmers in
marginal areas had adopted modern varieties re-
searchers began looking for alternative breeding and
dissemination approaches to the TOT approach. Two
approaches have emerged: participatory varietal se-
lection (PVS) and participatory plant breeding (PPB),
both of which acknowledge that farmers have rich
local knowledge that makes them better able to select
varieties suitable to their complex and changeable
farming systems than breeders in research stations
(Sperling et al., 1993; Witcombe et al., 1996). In
PVS farmers are allowed to choose finished, or near
finished, products of a conventional plant-breeding
program to evaluate in their own fields. Sperling et al.
(1993) reported that in Rwanda the selections made by
bean farmers involved in PVS outperformed breeders’
choices in farmers’ fields. In PPB farmers themselves
are involved in the breeding process. In one method
of PPB reported by Salazar (1992) the breeder gives
F3 or F4 material to farmers who carry out all further
selection. At F7 or F8 breeders monitor diversity and
select best material for conventional trials.
3.3. The SRR dryer (complex technology, simple
system)
The SRR dryer is one of the cheapest and simplest
mechanical grain dryers ever built. It consists of 3
components—atwo-stage axial fan powered by an
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 825
Fig. 3. The ‘very low cost’ SRR dryer.
electric motor; a heater which can either be an electric
element or a coal stove; and a bamboo mat drying
bin. It was designed to be used by small-area (<1 ha)
Vietnamese farmers growing rice in urban areas or in
rural areas with a good rural electricity supply. The
basic SRR dryer model, shown in Fig. 3, can dry 1 t
in 66 h and cost $90 as of March 1998.
About 665 units had been sold in the Mekong
Delta in the 2.5 years since the first commercial unit
was made in October 1995. One manufacturer, who
was also the main researcher on the SRR dryer de-
velopment project, has manufactured 85% of these.
The main benefit of the SRR dryer found in a July
1997 survey (Douthwaite, 1999) was to reduce the
amount of family labour needed to dry farm produc-
tion to one quarter. Children were found to spend this
time studying, and men and women spent the time
either in money-earning activities or other farm or
household work. Adopters were using the dryers to
dry 75% of their wet season output. This percentage
was significantly higher for poorer farmers with small
farm size.
The research that led to the development of the
SRR dryer by the University of Agriculture and Fish-
eries (UAF) began in a traditional way. UAF started
working on in-bin, low temperature drying, the tech-
nique on which the SRR dryer is based, because GTZ
funded a project to do this work. This GTZ project
‘Postharvest Technologies for Rice in the Humid
Tropics’ (GTZ Project) began largely as the result of
the success of a similar GTZ-funded dryer project in
Korea. It was expert driven. UAF had a strong dryer
group and lobbied GTZ to be included in the GTZ
Project. The GTZ Project provided them funds to
evaluate and develop low-temperature dryers, not to
conduct the type of R&D that UAF felt would have
most impact on Vietnamese farmers.
The invention of the SRR dryer was largely the
work of one UAF researcher, Mr. L.V. Bahn. He was
a rice farmer when growing up and by the time he
began working on in-bin drying also owned a small
manufacturing business. His background enabled him
to develop a first prototype of the SRR dryer that
generated a great deal of interest amongst farmers
when first tested at a rice mill in July 1995. This
interest led to a demonstration in a farmer’s house in
September 1995 carried out together with the Binh
Chanh extension office. This farmer, and three others,
ordered SRR dryers. This prompted the decision to
commercialise the technology in October 1995. The
extension officer helped make a radio program about
the dryer that resulted in another 20 orders in 1995.
Mr. Bahn built all the first machines. Other members
of the UAF dryer group also benefited financially
from these sales through a profit sharing scheme
linked to Mr. Bahn’s business.
826 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
UAF’s extension approach was to carry out
‘demo-sales’ for farmers interested enough in the SRR
dryer to come to UAF to find out more information.
Farmers heard about it from the radio program and
other media coverage. If the farmer was interested in
buying a unit, UAF staff transported the dryer to the
farmer’s house. Transport was relatively easy because
the dryer could be loaded onto the back of a motor-
bike. UAF staff installed the dryer and stayed for up
to four days to dry the first batch, during which time
they trained the farmer. If the farmer was satisfied
then he or she paid the full cost of the dryer. If not,
UAF took the dryer back.
The ‘demo-sales’ gave Mr. Bahn and the UAF
dryer team important feedback about the performance
of the early SRR dryers. They quickly realised that
the main constraint to adoption was the requirement
for adopters to have a good electricity supply that
could run the 1 kW heater as well as the 0.5 kW elec-
tric blower. Within 1 year they were supplying coal
stoves to replace the electric heater. The survey of
dryer adopters found that Mr. Bahn had made seven
other non-trivial changes to the SRR dryer to allow
use in areas with poor electricity; or to increase dryer
capacity; or to make the dryer cheaper to build.
In 1996 an owner of an electrical repair shop in
Long An Province, Muoi Dinh, began copying the
SRR dryer and by July 1997 had built about 100 units
while Bahn had built about 560. The same survey
found Dinh had made six modifications, four to allow
use of the dryer in areas with poor electricity supply,
reflecting the conditions in Long An; and two to make
the dryer cheaper to build. All modifications were
rated on a scale of +5 to –5 to reflect their effect on
the ‘fitness’ of the technology. A modification would
be given the highest rating of +5 if it was judged to
be a major improvement to machine fitness. The net
total of Bahn’s modifications were +21, showing his
innovative activity was highly beneficial, while Dinh’s
score was −5 showing his changes had reduced the
fitness of the technology.
The same survey found that users had made 13
significant changes to the operating instructions pro-
vided by UAF. UAF, in designing the operating
instructions, had tried to minimise fuel costs by rec-
ommending the heater was turned off during the day,
at the expense of long drying times. Most farmers,
however, had more than one batch (1 t) to dry at a
time and so reducing drying time was their main
priority. Hence, more than two-thirds ignored the
UAF recommendation and kept the heater on all, or
nearly all, the time. Even if they did turn the heater
off, hardly any used the UAF-recommended drying
strategy, choosing instead to devise versions of their
own, matched to their electrical supply, the initial
moisture content of paddy to be dried, and personal
preference. In adopting a strategy to minimise drying
time, owners were able to reduce the drying time by
39%, or by 42 h for a 22% increase in energy costs
if coal was used, or 37% increase if an electric heater
was used.
While the R&D that led to the SRR dryer was
‘traditional’ in approach as evident in UAF being tied
to evaluate one drying technique that ‘outside’ experts
had deemed appropriate, the subsequent development
and promotion of the SRR dryer can be described as
participatory for a number of reasons:
•the fact that Mr. Bahn was the researcher, manufac-
turer and came from a farming background made
seamless the integration between research, manu-
facturer and end-user needs;
•farmer opinion was sought very soon after the first
prototype was built;
•extension services were included and contributed
from the time of testing the first prototype;
•profit motivation meant the researchers were very
well motivated to improve the technology and
receptive to innovations to the machine and its
operation that improved ‘fitness’, whether they
developed them or not.
3.4. The SG harvester (complex technology,
complex system)
The SG harvester, shown in Fig. 4, consists of a
stripper rotor that spins in the crop as the machine
moves forward and combs, or strips, the grain from the
plants. The rotor throws grain, with some straw, into
a collection container. When full, two people change
the containers and empty the full one. The machine is
operated by a third person, who walks behind the ma-
chine. The mixture of grain and straw emptied from the
container is then threshed and cleaned by the TC800
thresher/cleaner, or any other available thresher. The
latest version of the SG800, the Mark III, costs $2000.
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 827
Fig. 4. The stripper gatherer (SG) harvester.
Together the SG harvester, TC800 thresher/cleaner and
seven people can harvest 4t paddy in 1 day.
About 130 SG harvesters have been sold in the
Philippines in 5 years since the first commercial sale.
Most of the farmers who have bought the machine did
so to reduce the costs and risks associated with organ-
ising, paying and depending on teams of hired harvest
labourers. Most owners use the machine only on
their own farms, which are large (14.4ha) compared
to the Philippine average (2.17 ha (IRRI, 1995a,b)).
Some owners have begun offering contract-hiring ser-
vices using the machines but this has yet to become
commonplace.
Like the SRR dryer, the project that developed the
SG harvester began in a traditional, supply-driven
way. In 1986, Silsoe Research Institute (SRI) in the
UK developed a stripper rotor (Klinner et al., 1987)
that was quickly and successfully commercialised in
the UK. The head of IRRI Agricultural Engineering
Division and the head of the Overseas Division of
SRI both saw potential for the technology in small
Asian rice fields. The first concept prototype of the
SG harvester was built in 1990 and subsequent ver-
sions were tested in farmers’ fields. In 1992, the first
order was placed for a unit which was supplied in
1993. The order was a result of a 1992 press release
that led to newspaper and magazine articles in the
Philippine and international press.
Ropali Trading Company, a large Philippine agricul-
tural machinery manufacturer, negotiated with IRRI
to begin production of the SG harvester in 1993. The
manufacturer bought a prototype from another builder,
tested it, and began building a first batch of 50 units
in November 1993, reassured that the technology was
‘mature’. Unfortunately there were 11 deviations from
the original design that seriously reduced the ‘fitness’
of the SG harvester (see Fig. 4). These included the
fitting of poor quality ground-drive transmissions that
generally broke before harvesting 1 ha; unbalanced
stripper rotors; and collection containers that did not
fit properly and were made from poor quality, under
specification plywood.
The manufacturer invested considerable resources
in advertising and demonstrating the SG harvester.
However, after selling 15 machines in 1994 serious
problems began to emerge. These did not stop the
manufacturer building another 20 in April 1994. By
mid-1998 Ropali had sold only another 16 machines
in spite of investing about $7500 in one promotion
campaign alone. The company also spent money in
replacing the faulty transmissions on the machines
it sold and in warranty claims: one owner had his
machine replaced five times.
In hindsight the company received less technical
assistance and training than was necessary. Part of the
reason for this was that the IRRI R&D team who had
been working with the manufacturer stopped after re-
sponsibility for this ‘extension’ work was passed over
to the Philippine Rice Research Institute (PhilRice).
This transfer of responsibility exactly followed the
TOT approach that draws a clear line between research
and extension activities. It took time for the PhilRice
team to learn about the SG harvester and during this
transition period the manufacturer received little help.
In 1995 IRRI and the Philippine Rice Research
Institute (PhilRice) released the Mark II drawings of
828 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
Fig. 5. Net effect of manufacturers’ modifications on the fitness (likelihood of adoption) of the original IRRI design of the Mark II SG
harvester.
the SG harvester. The design included 10 modifica-
tions designed to make the machine more reliable,
robust, 2better able to operate in muddy and soft
field conditions, and easier to operate. Manufacturers
first made several of these changes. The effect of the
changes was to increase the weight of the machine
by 31% which negated the changes made to improve
handling and performance in difficult field conditions.
In 1996 and 1997, nine manufacturers were sur-
veyed for the variations that their machines had from
the Mark II design, which all were building. The
same ranking system was used as with the SRR dryer
described previously. Fig. 5 shows that only one man-
ufacturer had actually improved the ‘fitness’ of the
technology. Morallo Metal Industries developed their
own version of the Mark II design that was almost as
light as the Mark I version but far more reliable. An-
other manufacturer developed an idea for an improved
wheel design that was further developed by IRRI
and proved to give much better mobility in muddy
field conditions. In 1997, IRRI released drawings of
2The first design used relatively expensive Japanese-made bear-
ings, chains and sprockets. Local manufacturers often substituted
these for cheaper, lower quality, lower specification Chinese-made
components. The size of some of the drive train components on
the Mark II SG harvester were increased to give reasonable life
with Chinese-made components.
the Mark III SG harvester based on the Morallo SG
harvester and incorporating the novel wheel.
As with the SRR dryer the development of the SG
harvester was expert driven at the beginning but be-
came participatory when manufacturers started build-
ing it and farmers started buying and using it. Unlike
the SRR dryer, however, researchers were hindered
from fully participating in the early adoption process.
4. Discussion
“The importance of understanding innovation as a
process is that this understanding shapes the way we
try and manage it” (Tidd et al., 1997). Our purpose
in this paper is to assist the development and choice
of approaches to catalyse and manage agricultural
innovations that have a public sector source. Our con-
tribution is to help understanding of the interactions
between management approach, technology type and
farming system type.
4.1. Knowledge and fitness changes in the case
study technologies
Mokyr (1990) defined an invention as an incremen-
tal increase in the total knowledge set of a society.
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 829
Fig. 6. Schematic representations of the change in the knowledge associated with three types of technology during the adaptation phase.
Fig. 6 uses this definition to show a schematic repre-
sentation of the new knowledge, or existing knowledge
used in a novel way, associated with the three case
study technologies. The figure shows how the knowl-
edge changes between the end of the start-up phase
just before key stakeholder adoption begins, and the
beginning of the expansion phase when widespread
adoption of the technology occurs (see Fig. 2). The
figure distinguishes between two types of knowledge:
•Embodied or hardware knowledge — knowledge
that is embedded in the machine or seed itself. For
example, the knowledge needed to build a blower
that works efficiently and delivers the required vol-
ume of air at the necessary pressure, or the knowl-
edge that a particular gene gives resistance to stem
borer.
•Disembodied or software knowledge — knowledge
that is not embedded and has to be socially con-
structed in situ by the people replicating and using
the technology. Software knowledge has two parts:
1. The knowledge required using or growing the
technology.
2. The knowledge required building or repro-
ducing it.
Fig. 6 also shows the extent to which the researchers
and the two key stakeholders (manufacturers and
users) involved in the innovation process contributed
to this knowledge over time.
Bar graph areas in Fig. 6 represent the sum of
knowledge that is new to the system into which the
technology was introduced, and existing knowledge
used in a novel way. All technologies are founded on
what went before so in terms of the total knowledge
associated with a technology the knowledge shown is
the tip of the iceberg. There is no easy way to quantify
new knowledge and Fig. 6 does not attempt to do this.
Rather the figure attempts to give a qualitative picture
of knowledge changes in terms of ‘more’ or ‘less’,
‘increase’ or ‘decrease’. Estimates of the amount and
830 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
sources of knowledge are based on data from the case
studies indicating the extent of the cognitive contribu-
tion of the stakeholders in the innovation process.
Fig. 6 shows important knowledge differences
between ‘simple’ modern seed varieties and the
‘complex’ agricultural machinery. Firstly, modern
rice varieties, whether introduced into favourable
or unfavourable production environments, have the
largest amount of knowledge associated with them,
but nearly all of this is embodied, and nearly all of
this comes from research. For example, IR36, intro-
duced in 1976, had at least partial resistance to seven
of the most serious pest and diseases. Resistance to
one disease — grassy stunt — required researchers
to screen over 17,000 varieties for the resistance
gene (Holden et al., 1993). The knowledge required
to undertake this screening, as well as the ability to
combine resistance genes in a single, high yielding
variety, is part of the knowledge shown in Fig. 6(a).
The fact that genes that gave this resistance were
available to IRRI scientists to find was due to natural
and farmer selection. This resistance can be thought
of as existing knowledge. However, IRRI researchers
discovered and applied it in novel ways so research
is shown as the source of the innovation. Fig. 6(a)
shows that this embodied knowledge does not change
once the MV is released. This is because MVs breed
pure — there is not genetic diversity in their progeny
on which farmers can select, unlike with landraces.
In contrast, Fig. 6(b) and (c) shows that the embod-
ied knowledge in equipment technologies had mul-
tiple sources of innovation and they changed a great
deal after first release. These modifications represent
experiential learning cycles carried out on the part of
the farmers and manufacturers as they used and built
the equipment. Increases in researcher innovation after
the release of the technology are also the result of re-
searchers learning more about the technology in actual
usage. Fig. 6(b) and (c) show more of the hardware
knowledge originating from researchers and manufac-
turers than from end-users reflecting the finding that
manufacturers and researchers, not users, made the
majority of modifications that were incorporated in
design after commercialisation. Fig. 6(b) shows that
when first commercialised the SG harvester, being
a far more complex technology, contained a greater
amount of knowledge than the SRR dryer. Most of this
knowledge originated from research. In contrast the
SRR dryer was developed by a research team that also
had experience manufacturing dryers and had farming
backgrounds, so a greater proportion of knowledge
came from these sources, at least indirectly.
Compared to the ‘easy to use’ MVs, Fig. 6(b) and
(c) show more ‘software’ knowledge associated with
the ‘hard to employ’ technologies at the end of the
adaptation phase, with more sources of innovation.
This is because users needed to learn much more to
use equipment technologies than grow MVs: operators
needed to learn new skills to successfully operate the
SG harvester and SRR dryer and farmers needed to
reorganise their harvest and drying arrangements. In
contrast farmers can profitably grow MVs in the same
way they grew their traditional varieties and so have
comparatively little to learn.
In Fig. 6(b), knowledge to operate the equipment
technologies is shown as all researcher knowledge at
the first release stage which was passed on to opera-
tors through training materials and first-hand instruc-
tion. Fig. 6(b) shows that this researcher knowledge
actually decreased at the beginning of the success-
ful adoption stage as a result of operators learning
their own ways of better organising themselves and
operating the machines in their conditions.
Equipment technology requires manufacturers to
learn to build it resulting in an additional source
of innovation that MVs do not have. Manufacturers
needed to learn how to build the equipment technolo-
gies using their (often limited) workshop assets, set
up quality control procedures, build jigs and fixtures,
learn where the cheapest sources of non-standard raw
materials exist, etc. This is represented in Fig. 6(b)
and (c) as a large increase in manufacturer knowl-
edge between the first-release and successful adop-
tion stages. In contrast there is no new reproduction
knowledge needed with MVs because farmers can
grow and save seed of MVs in much the same way
as they did with traditional varieties.
Fig. 7 shows the ‘fitness’ of the case study technolo-
gies and how this changed after first adoption. Fig. 7
identifies two fitness levels: (1) good enough to be
adopted by innovative farmers; and (2) good enough
for widespread adoption.
Fig. 7 shows that MVs introduced into FPEs are
highly fit, which gives rise to high adoption rates. The
attributes of the technology (fitness) improve slightly
with time as farmers adapt crop management to their
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 831
Fig. 7. ‘Fitness’ trajectory of case study technologies.
local conditions. The increase rate is small because
the scope for learning and innovation is relatively
little compared to the equipment technologies. Then,
as pests and diseases evolve to by-pass the genetic re-
sistance the fitness falls off and a new MV is needed
to take its place. The situation is different with a MV
introduced into a LFPE. Here, if a suitable variety
is found at all (and few are) then Fig. 7 shows that
the MV has lower fitness than in a FPE, reflecting
the finding that farmers apply less fertiliser and so
achieve a lower return in more marginal areas. How-
ever, lower fitness implies lower adoption rate which
in turn means slower breakdown of resistance, hence
much more prolonged fall-off in fitness.
The fitness trajectories of the complex technologies
are very different to those of the simple ones. This re-
flects the much larger scope for innovation that existed
for the complex technologies, and the result this inno-
vation had on fitness, both beneficial and detrimental.
Fig. 7 shows that neither complex technology was
suitable for widespread adoption when first released.
Both technologies were initially bought by farmers
who were drawn to adopt the technology after hearing
about it in the media. These farmers’ characteristics
closely match Rogers (1995) definition of an inno-
vative adopter, namely, people who are venturesome,
are drawn by the technical challenges posed by new
technology, are prepared to take the risk and have the
resources to do so.
Fig. 7 shows that the SG harvester was less fit than
the SRR dryer when it was first introduced. This was
partly because the machine itself was much more com-
plex. 3This additional complexity also explains why
the fitness of the technology fell when manufactur-
ers began building it because there was more to learn
and more scope to make mistakes. Also any changes
they made to the design were more likely to cause
unforeseen and detrimental results (see Fig. 5). The
fitness of the SG harvester finally began increasing
when Morallo Metal Industries developed their ver-
sion of the machine, ARC Manufacturing developed
3The SG harvester has four belt drives, two chain drives and
five shafts while the SRR dryer has none.
832 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
an improved wheel and IRRI put both improvements
together and promulgated them as the Mark III design.
The SG harvester R&D team at IRRI speeded up the
evolution of the technology by playing this selection
and promulgation role and may have made the differ-
ence between success and failure. They were uniquely
qualified to play the role through their knowledge of
the technology and their motivation to see it succeed.
Low initial fitness of the SG harvester was also
due to the higher complexity of the system in which
the machine had to work, and the system clashes
that resulted. One example of a system clash was the
negative response of manual labourers who felt threat-
ened by the introduction of the machine. Labourers
threatened or took sanctions against farmers using
the harvester which reduced the attraction of farm-
ers using the machine. One sanction was to refuse
to harvest when the SG harvester was unable due to
difficult field or crop conditions. In contrast the SRR
dryer was bought to reduce family labour, just like
families in developed countries might buy washing
machines. Such technologies cause less social conflict
than those that displace hired labour.
In contrast to the SG harvester, Fig. 7 shows that
the fitness of the SRR dryer increased after first com-
mercialisation. A key reason for this was that the
main researcher was also the manufacturer so there
were no problems in transferring knowledge about the
technology. UAF, in allowing a researcher to become
a manufacturer, contributed to the success of the in-
novation. Unwittingly, UAF was following successful
innovation management practice identified in indus-
try in the 1960s. In the conclusions from a survey
of factors affecting the success of ten key innova-
tions in Europe and the USA, Layton et al. (1972)
wrote: “Our studies showed that an important factor
in successful innovation is the transfer of information
from development to production; and here we found
no substitute for the movement of people, including
highly qualified ones” p. 3.
The increase in fitness of the SRR dryer came
about when UAF staff, who were both researching
and manufacturing the dryer, learned more about
the performance of the technology in actual farm
conditions and made modifications accordingly. Im-
provements also resulted from user innovations. Like
IRRI, UAF played an important role in discovering
and promulgating these.
Fig. 7 shows that the rate of increase in the fitness
of the SRR dryer began to fall 1 year after first com-
mercialisation. This was when the manufacturer in
Long An Province started making and selling inferior
quality machines that also did not work as well.
Another difference between the SRR dryer and the
SG harvester is that the fitness of the former became
good enough for widespread adoption 6 months af-
ter first commercialisation, once the innovation of
using a coal stove instead of an electric heater was
made. Although the knowledge associated with the
SG harvester is greater than the SRR dryer 5 years
after commercialisation, it is still not fit enough for
widespread adoption. Further improvements to the
performance of the machine, or demand-side changes,
are required before sales of the technology will begin
to increase substantially. These changes are likely to
occur if economic growth continues in the Philip-
pines and with it shortages of manual harvest labour
continue to increase.
5. Testing the hypotheses
The hypotheses were set-up so that testing them
would help achieve the purpose of the paper, namely,
gaining a greater understanding of the interaction
between management approach, technology type and
the system into which the technology is introduced.
The results of testing the hypotheses will give an in-
dication of the most effective mix of top-down and
participatory management approaches.
Hypothesis 1. Technologies that require key stake-
holders to change little to make the technology work
(i.e. easy to understand and use technologies intro-
duced into simple farming systems) will have the
fastest impact using a top-down approach.
The case study of the early MVs developed by IRRI
supports the hypothesis. It shows that the top-down
TOT approach developed by IRRI and the CGIAR sys-
tem has been extremely effective at developing HYVs
and transferring them to FPEs in many countries, who
have then transferred them to farmers. The nature of
the technology transfer system allowed impact to be
achieved quickly because IRRI transferred breeding
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 833
lines to many NARS who were able to adapt and ex-
tend the technology to their farmers. It is not true to
say that the TOT approach was not participatory —
the approach did engender the active participation of
NARS plant breeders. It would be more accurate to
describe it as not being farmer-participatory. In prac-
tice this meant neither CGIAR nor NARS scientists
saw farmer learning and innovation as important in
the extension and final impact of the technology. To a
large extent they were probably right given the nature
of MVs. Most of the knowledge associated with a
MV is ‘locked’ into the seed and cannot be altered
by farmers. Growing rice is not a new technology so
there was very little farmer learning necessary to make
MVs work. The only scope for farmer innovation with
MVs is in crop management practices, and while this
has certainly taken place, no references have been
found of work that has looked for this innovation and
assessed its impact on technology performance and
adoption rate. Most impact studies carried out have
simply looked at adoption or non-adoption of MVs,
not farmer modifications to the technology package
(personal communication with G. Castillo, 1999).
While the TOT approach has been very effective
with MVs in FPEs, it is conceivable that greater
participation of farmers during breeding would have
resulted in varieties with even higher fitness. For
example, the early IR varieties would have been
improved if they had tasted better.
Hypothesis 2. If either the new knowledge requi-
rement or the system complexity is high then partici-
patory approaches will be most effective.
The mechanical technology case studies showed the
importance of the participation of the key stakehold-
ers and researchers during the adaptation phase (early
adoption). As Fig. 7 shows neither technology was fit
enough for widespread adoption and only improved
with key stakeholder and researcher learning and in-
novation. The assumption of the TOT approach in the
development of the SG harvester restricted the partic-
ipation of the original R&D team from facilitating the
learning of the key stakeholders, and also restricted
the R&D team learning more about the technology
and further improving it. As a result, key stakeholder
knowledge gaps lasted longer than they should have,
more mistakes were made, and innovations to solve
technical shortcomings took longer than they might.
Together, these factors contributed to the fitness of
the SG harvester deteriorating after release, and the
length of time before it recovered.
The invention and subsequent development and
diffusion of both the SG harvester and SRR dryer
involved participatory and non-participatory ap-
proaches. By definition technologies that qualify
as macro-inventions when first introduced require
non-participatory ‘expert driven’ early R&D. Mokyr
(1990) defined a macro-invention as a new technology
without clear-cut parentage that represents a distinct
break from previous technique in a given society.
Mokyr also went on to say that macro-inventions
represent a challenge to the status quo and an at-
tack by an individual on a constraint that everyone
else has taken for granted. Farmer-participatory ap-
proaches that make use of local knowledge rather
than challenge it are more likely to be useful for
micro-inventions. Micro-inventions are innovative
changes to macro-inventions to get them to work
better. Both the SG harvester and SRR dryer were
macro-inventions with respect to the systems into
which they were introduced. Therefore, one can pre-
dict that the participation of the key stakeholders was
likely to be most useful when the concept had already
been established. In practice this point represents the
beginning of the adaptation phase when the key stake-
holders gain a ‘stake’ in the technology by starting to
build it and own it.
Modern varieties are not macro-inventions when
first introduced into a farming system so one can pre-
dict that farmer participation can be of use from the
beginning of the R&D process. This is borne out as
seen previously by recent work in which farmers do
most of the breeding (Salazar, 1992; Witcombe, 1996).
6. Implications for R&D planning and
implementation
The tests of the hypotheses above suggests that
effective management of invention and innovation
requires a mix of top-down and participatory ap-
proaches whatever the complexity of the technology
or system. The criteria for choosing the amount of
R&D team participation in the early adoption pro-
cess is the amount of key stakeholder learning that
834 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
will be required in implementing the new technol-
ogy, and the scope for innovation after adoption. The
more learning necessary, and the greater the scope
for innovation, then the more the R&D team should
participate in the adaptation phase.
Key stakeholder participation will be of most help
in the development phase if the new technology is a
micro-invention rather than a macro-invention. Key
stakeholder participation should be sought as soon as
a workable prototype of a macro-invention has been
developed. The R&D team should attempt to ‘perfect’
a new technology as much as possible before release to
reduce the key stakeholders’ risks of adoption. How-
ever, given the length of project cycles the R&D team
will probably have no more than 3 years to demon-
strate the viability of the innovation through actual
farmer adoption. Early key stakeholder participation
reduces the risk of the development of inappropri-
ate technology and is more parsimonious (Chambers
and Jiggins, 1986) because the key stakeholders
embody the necessary local knowledge themselves
instead of researchers trying to capture it through ex-
pensive surveys. Furthermore, early key stakeholder
participation increases the potential sources of inno-
vation and so increases the rate at which the fitness
of the technology will increase, if the process is
properly managed.
While the amount of key stakeholder learning
and innovation expected during adoption determines
the mix between researcher-driven and participatory
approaches, the type of learning and innovation ex-
pected determines the actual type of participatory or
non-participatory approach used. The case studies
showed that participatory varietal selection (PVS) is
not suitable for equipment technology. This is because
farmers are not inventive in PVS, in that they do not
invent the novelty that distinguishes one variety from
another, rather they select novelty that is generated by
crossing, genetic diversity, and from the interaction
of genotype with the environment. In contrast, with
equipment and most other types of technology benefi-
cial novelties are usually the result of micro-inventions
of the stakeholders. This means that managing par-
ticipatory technology development (PTD) in other
technologies is likely to be harder than managing PVS
because the need for invention makes the process less
predictable and requires people who are inventive as
well as good at selection and promulgation.
Other characteristics of seed technology that makes
management of the innovation process particularly
easy are its high trialability and observability att-
ributes. Farmers can easily experiment with new
varieties because seed is cheap to produce, easily
transported, can be given to farmers without instruc-
tions (Witcombe, 1996) and farmers can then grow
their own seed and pass it on to their neighbours who
have seen its benefits for themselves.
7. Conclusions
This paper has taken the view that an invention
represents an increment in the total technological
knowledge of a society or farming system. New
knowledge associated with a new technology is em-
bedded in the hardware (seed or machine) and the
software, i.e. the knowledge to replicate and use the
technology. The paper has shown that the adoption
rate and impact of some new technologies depends
upon the motivation and success of the people who
directly benefit from an innovation — the key stake-
holders — in learning about the technology and their
ability to make it cheaper and more profitable to
replicate and use. It also depends upon researchers —
the group who developed the first prototype and who
have scientific knowledge of the technology — work-
ing in partnership with the key stakeholders to help
impart scientific knowledge, learn more about the per-
formance of the technology in real conditions them-
selves, and to make improvements of their own. The
importance of these learning and modification pro-
cesses on adoption rate and eventual impact depends
on the amount of new knowledge that needs to be
learnt by the researchers and key stakeholders.Inthe
past technology generation and transfer approaches
in public sector agricultural research have been based
on a model developed for the transfer of very simple
technologies — high yielding crop varieties (HYVs)
— into simple systems. Seed technology, particularly
HYVs of closed pollinated species, is unusually sim-
ple because the knowledge embedded in the hardware
(the seed) cannot be altered by the user. Furthermore,
HYVs can be grown and reproduced in similar ways
to farmers’ traditional varieties. Taken together this
means that farmers need to learn very little, and can
modify very little, when adopting HYVs. Models for
B. Douthwaite et al. / Research Policy 30 (2001) 819–836 835
generating and transferring HYVs can, therefore, be
simple. One resulting organisational simplification is
the separation of research from extension. However,
this simplification is not appropriate for the research
and dissemination of more complex technologies
where more learning and more adaptation is possible
and required. This paper has shown that with some
technologies, e.g. agricultural equipment, modifica-
tions made by the key stakeholders can result in rapid
changes in the fitness of the technology that can be ei-
ther beneficial or detrimental. Deterioration of fitness
due to mistakes as a result of knowledge gaps of the
researchers and key stakeholders are likely during the
initial adoption phase. Research planners and man-
agers need to be aware of this and act accordingly.
When properly managed the early release of a new
technology can be a parsimonious and rapid way to
generate improvements resulting from a synthesis of
scientific and local knowledge. These improvements
in fitness make widespread adoption and impact more
likely. However, if badly managed the result can be
technology failure. Good management involves ensur-
ing the R&D team participates in the early adoption
process to facilitate key stakeholder learning while
at the same time making innovative improvements
themselves.
The stage at which key stakeholder participation
can be effective depends upon the extent to which the
technology is seen as an improvement of an existing
technique (a micro-invention) or a new technique that
challenges existing beliefs and cultural practices (a
macro-invention). In the latter key stakeholder partic-
ipation is likely to be effective only when a working
prototype has been produced.
References
Biggs, S.D., 1989. A multiple source model of innovation of
agricultural research and technology promotion. ODI Agricul-
tural Administration Network Paper 6, Overseas Development
Institute, London, England.
Byerlee, D., Pingali, P.L., 1994. Rice research in Asia: fulfilments
and frustrations. In: Paper Presented at the XXII Conference
of the International Association of Agricultural Economists,
Harare, Zimbabwe, 22–29 August 1994.
CGIAR (Consultative Group on International Agricultural Rese-
arch) Secretariat, 1997. Chairman’s opening statement “Meeting
the challenges of a changing world”, 27 October 1997.
Downloaded from the internet from http://www.worldbank.org/
html/cgiar/publications/icw97/icw97.html, May 1998, WA,
USA.
CGIAR (Consultative Group on International Agricultural
Research) System Review Secretariat, 1998. Third system
review of the Consultative Group on International Agricultural
Research (CGIAR), 8 October. Published on the internet at
http://www. cgreview.worldbank.org/cgrevrep.htm, May 1999,
WA, USA.
CGIAR (Consultative Group on International Agricultural
Research) Secretariat, 1999. What is the CGIAR? Published
on the internet at http://www.cgiar.org/whatis.htm, May 1999,
WA, USA.
Chambers, R., Jiggins, J., 1986. Agricultural research for resource
poor farmers: a parsimonious paradigm. Discussion Paper
220, Institute of Development Studies, University of Sussex,
Brighton, England.
Clark, N., 1995. Interactive nature of knowledge systems: some
implications for the third World. Science and Policy 22,
249–258.
Douthwaite, B., 1999. Equipment evolution: case studies of
changes in rice postharvest technologies in the Philippines
and Vietnam. Ph.D. Thesis, University of Reading, Reading,
England.
GAO (General Accounts Office), 1987. Case study evaluation.
Program Evaluation and Methodology Division, Transfer Paper
9, GAO, Washington DC, USA.
Holden, J., Peacock, J., Williams, T., 1993. Genes, Crops and the
Environment. Cambridge University Press, Cambridge, UK.
Horton, D., 1997. Disciplinary roots and branches of evaluation:
some lessons from agricultural research. Discussion Paper No.
96–7. ISNAR, The Hague, The Netherlands.
Horton D., Prain, G., 1989. Beyond FSR: new challenges for
social scientists in agricultural R&D. Quarterly Journal of
International Agriculture 3/4.
IRRI (International Rice Research Institute), 1993. IRRI rice
almanac. IRRI, Los Baños, Philippines.
IRRI (International Rice Research Institute), 1995a. IRRI rice
almanac. IRRI, Los Baños, Philippines.
IRRI (International Rice Research Institute), 1995b. World rice
statistics 1993–1994. IRRI, Los Baños, Philippines.
Kaimowitz, D., Snyder, M., Engel, P., 1989. A conceptual
framework for studying the links between agricultural research
and technology transfer in developing countries. Linkage
Theme Paper No. 1, International Service for National
Agricultural Research (ISNAR), The Hague, The Netherlands.
Klinner, W.E., Neale, M.A., Arnold, R.E., Geikie, A.A., Hobson,
R.N., 1987. A new concept in combine harvester headers.
Journal of Agricultural Engineering Research 38 (1), 37–45.
Layton, C., Harlow, C., de Hoghton, C., 1972. Ten innovations.
An International Study on Technology Development and Use
of Qualified Scientists and Engineers in Ten Industries. George
Allen & Unwin, London.
Mokyr, J., 1990, The Lever of Riches: Technological Creativity
and Economic Progress. Oxford University Press, Oxford, UK.
ODI (Overseas Development Institute), 1997. Global hunger
and food security after the world food summit. Briefing
paper 1997 (1) February. Published on the internet at
http://www.oneworld.org/odi/briefing/1 97.html.
836 B. Douthwaite et al. / Research Policy 30 (2001) 819–836
Rogers, E.M., 1995. Diffusion of Innovations, 3rd Edition. Review
Edition of Communication of Innovations, The Free Press,
New York, USA
Rosenberg, N., 1982. Inside the Black Box: Technology and
Economics. Cambridge University Press, Cambridge, UK.
Ruthenberg, H., 1985. In: Jahnke, H.E. (Ed.), Innovation Policy for
Small Farmers in the Tropics. Clarendon Press, Oxford, UK.
Salazar, R., 1992. MASIPAG: alternative community rice breeding
in the Philippines. Appropriate Technology 18, 20–21.
Scott, James C., 1998. Seeing Like a State: How Certain Schemes
to Improve the Human Condition have Failed. Yale University
Press, New Haven, USA.
Sechrest, L., Stewart, M., Stickle, T.R., Sidani, S., 1996. Effective
and persuasive case studies. Jaguar Graphics, Tucson, AZ, USA.
Sperling, L., Loevinshohn, M.E., Natbormvra, B., 1993. Rethinking
the farmers role in plantbreeding: local bean experts and
on-station selection in Rwanda. Experimental Agriculture 29,
509–519.
Tidd, J., Bessant, J., Pavitt, K., 1997. Managing Innovation: Inte-
grating Technological, Market and Organizational Change.
Wiley, Chichester, UK.
Von Hippel, E., 1988. The Sources of Innovation. Oxford
University Press, Oxford, UK.
Witcombe, J.R., 1996. Participatory approaches to plant breeding
and selection. Biotechnology and Development Monitor 29,
2–6.
Witcombe, J.R., Josi, A., Josi, K.D., Sthapit, B.R., 1996.
Farmer participatory crop improvement. I. Varietal selection
and breeding methods and their impact on biodiversity.
Experimental Agriculture 32 (4), 427–444.
Yin, R.K., 1989. Case Study Research: Design and Methods.
Sage, Beverley Hills, CA, USA.
