CHAPTER FOUR. CONTENDING PERSPECTIVES ON ORGANISATIONAL

LEARNING

AA metalogue is a conversation about some problematic subject. This conversation should be such that not only do the participants discuss the problem but the structure of the conversation as a whole is also relevant to the same subject.@

(Bateson, 1972:1)

4.1 Introduction

Outside of the large literature on evolutionary theory and epistemology (Cziko and Campbell, 1990) there is an emerging body of social science literature that focuses specifically on organisational learning. The aim of this chapter is to examine some key writers in this field and relate their work to the theory introduced in Chapter Three. This comparison will help identify the strengths and weaknesses of the theory, and further implications for representing and assisting organisational learning.

Two bodies of literature will be examined. The first is the social science literature accessible via the Social Sciences Citation Index (SSCI).Where this literature does have a practical orientation or an empirical basis it is generally focused on business firms and, to a lesser extent, government bodies. Because of the volume and diversity of this literature some selectivity has been necessary in order to make the review task manageable. Attention has been given primarily to writers whose work has been widely cited within the field of organisational learning, an approach consistent with the evolutionary view of learning outlined in Chapter Three. The second section looks at the much smaller body of literature that has been generated within Development Studies, which focuses primarily on the operations of aid organisations and the client organisations they work with. This includes a significant amount of grey literature that is not published in academic journals and noted by the SSCI.

4.2 The Emergence of Organisational Learning

AOrganisational learning@ as a phrase describing a field of concern has only entered into common usage in the social sciences since the late 1980's, though the wider issues of organisational performance have been around for many years (Cyert and March, 1963). An examination of social science journals covered by the Social Sciences Citation Index, by Crossan and Guatto (1996) in 1995, shows that the number of social science journal articles specifically mentioning organisational learning averaged around 5 per year in the 1980's, twice the level of the 1970's. However, from around 1990 the number grew year by year, reaching more than 50 a year by mid-1997. The number of published articles was increasing at a rate faster than the growth of academic publications (Crossan and Guatto, 1996). By the mid-1990's approximately seventy different social science journals had published articles on organisational learning. A similar trend was present in doctoral level research, especially in the USA. Although there were only occasional references in the 1980's, in the early 1990's an average of 19 American thesis abstracts a year referred to organis(z)ational learning.

The area where the growth in interest in organisational learning has been most noticeable is in the publications of books on management themes. In the words of Tom Peters, an internationally known and highly paid management guru (Hucztynski, 1993:5) ALearning organisations have become a hot topic@ (Peters 1992:383). Over the last few years a number of popular management books have been published specifically on the topic of organisational learning, mainly written from a prescriptive perspective and with the interests in mind of managers of businesses who want to improve performance (Wright et al. 1989; Senge, 1990a; Garrat, 1990; Argyris, 1992; Casey, 1993; Marguardt and Reynolds, 1993; Swieringa and Wierdsma, 1993; Burgoyne et al. 1994; Cunningham, 1994; Dixon, 1994; Grundy, 1994; Howard, 1994; Leeuw et al. 1994; Licari, 1994; and others).

The vast majority of this literature is outside the field of development studies. The most common types of organisations examined or referred to are firms, then government bodies. There are only occasional references to aid projects or projects (e.g. Hopkins, 1990). This may be partly an artefact of the various different terminologies being used in the development studies literature. In Development Studies work that relates to organisational learning has been variously labelled as Alearning process approach@ (Korten , 1980; Uphoff, 1992), Aadaptive administration@ (Rondinelli, 1983:89), Alearning from experience@ (Edwards, 1989; Hulme, 1989), Aorganizational learning@ (Edwards, 1997), Ainstitutional learning@ (Bergdall, 1996; Roche, 1995; Slim, 1993), and Abuilding learning systems@ (Bawden, 1992). Much of the literature is not published in journals, but in papers circulated internally, or at conferences and workshops. These features of diversity and informality suggest that discourse in this particular area of Development Studies is at a relatively early and inchoate state. In contrast to the continuous stream of literature reviews on organisational learning outside of Development Studies (see below), so far there has been only one such review within development studies (Edwards, 1997). While there are some well known published writers on the related subject of learning process approaches to development projects, such as Korten (1980) and Rondinelli (1983), their contributions have not informed the body of literature on organisational learning documented in the various SSCI cited reviews of organisational learning (Levitt and March, 1988; Huber, 1991; Easterby-Smith, 1997).

4.3 Influential Writers Outside of Development Studies

The focus of this first section is the SSCI indexed literature that explicitly addresses organisational learning. Table 4.1 below provides a list of all the authors whose work has been cited 50 or more times within the SSCI literature on organisational learning in the 1994-7 period, and who have produced individual papers that have been cited 20 or more times within that same literature. The period up to 1993 has been analysed separately by Crossan and Guatto (1995). The most frequently cited authors for this earlier period are identified in the same table by the brackets after the authors name, which contain their rank order position as of 1993.

Within the list there are some significant differences. The two most cited writers, especially Senge, differ from the rest in that they have explicitly tried to sell their ideas, to companies wanting to improve their performance in the market place. They are part of a wider growth in the proportion of papers focusing on application, as distinct from analysis, synthesis and review noted by Crossan and Guatto (1996). The next two (Huber, 1991; and Levitt and March, 1988) are more academic in orientation. Unlike the first two which have tried to offer a Aunique selling proposition@, differentiating their product from others, the papers by Huber and March are synthesising studies, trying to bring a meaningful order to the diversity of all the previous work done in the field. The next two sets of papers, by March (1991) and that of Nelson and Winter (1982), are both influenced by evolutionary theory. Daft and Weick (1984) are representative of a more interpretative and less systems oriented perspective on organisational learning (e.g. Fiol and Lyles, 1985; Fiol, 1994). The paper by Nonaka is more independent in its development, only giving acknowledgement to one of the other writers on this list (Argyris). It is conspicuous because although only recently published it has been widely cited. Although work by leading organization theorists such as Williamson, Pfeffer, Mintzberg, Simon (Padgett, 1992) and others is referred to in the organisational learning literature it has not been on the same scale as those mentioned above. James March is the only widely recognised organisation theorist who has specialised in this area.

Overcoming Organizational Defences: Chris Argyris

Argyris and Senge are proactive writers on organisational learning. They are advocating views on how organisations can learn more effectively, not simply describing organisational learning as they see it. Their confident advocacy of particular prescriptions (Edmondsen, 1996) is in contrast to the evolutionary view of learning, introduced in Chapter Three, which suggest that the presence and type of learning behaviour that can be found will always be dependent on context, on the local ecology. Argyris partly justifies his prescription for improved learning by assuming that almost all companies are facing a rapidly changing environment and (implicitly) that the primary problem facing companies is not the retention of past lessons, but the acquisition of new knowledge. Even if change is endemic this is questionable. As well as being proactive or simply passive, organisations may also be victims of excessive change. Retaining existing competencies and past knowledge can be difficult when there are high levels of staff turnover (Carley, 1992). Uphoff (1992:11) reports Aan attrition rate of 95 per cent@ over four years amongst the village level workers= employed in the USAID funded Gal Oya irrigation project in Sir Lanka. Even the World Wide Web which is seen as the epitome of change, is based on an invention (hypertext) that was initially developed to cope with the loss of information caused by high levels of staff turnover in a large research institution (Naughton, 1998).

Argyris=s theory of learning is based on one major distinction which has been widely quoted, rediscovered and reformulated in the literature on organisational learning, and one which is often attributed to Bateson (Argyris, 1976; Sutton, 1994; Easterby-Smith, 1997). This is the distinction between single loop and double loop learning. AWhen a thermostat turns the heat on or off, it is acting in keeping with the program of orders given to keep it to the room temperature, let us say, at 68 degrees. This is single loop learning, because the underlying programme is not questioned. The overwhelming amount of learning done in an organisation is single-loop learning because it is designed to identify and correct errors, so that the job gets done and the action remains within stated policy guidelines@ (Argyris, 1992:115-6). In contrast, second order learning involves questioning of the underlying objectives and policies. In much of his discussions of organisational learning he regards the most significant organisational learning as being the capacity to engage in double loop learning. He assumes that many organisations do not have this capacity, and almost axiomatically that more is better (Argyris, 1992). However, while the ability to question and think about other goals and criteria of performance can undoubtedly enable greater flexibility and adaptability by an organisation, Argyris does not give attention to the risks posed by a great diversity of views at that level. In excess such diversity would actually be disabling, preventing any effective joint action.

Argyris=s overall analysis of learning is social-psychological rather than sociological or micro-political. The blockages to learning are what he calls Aorganizational defensive routines". "These consist of all the policies, practices and actions that prevent human beings from having to experience embarrassment or threat and at the same time, prevent them from examining the nature and causes of that embarrassment or threat@ (1994:80).

There are two limitations to his focus on this problem. One is the assumption of pathology, the other limited scale of the analysis. Argyris has described defensive routines in terms that are analogous to the idea of defence mechanisms used by individuals to manage their own anxieties. However, less negatively phrased, they also bear some relationship to the everyday social practices documented by Goffman (1959), whereby people successfully manage their interactions with each other, with minimal cost. Seen from the long term point of view of the organisation they may be maladaptive defences, but from the immediate interests of the participants they may be highly adaptive. Argyris=s bias here may be linked to the fact that his analysis of the behaviour of firms is derived from consultancy work contracted by CEOs wanting to change their organisation=s performance, not contracts with individuals within the organisation who may have more immediate and local interests.

The other limitation is that the analysis is limited to face to face interactions between individuals. Argyris does not give significant attention to the role of conflicts between sub-groups within organisations which may be affecting consensus or disagreement over second order understandings. Nor is there much recognition of individual staff members= often justifiable concerns for their own survival within the organisation. Although defensiveness is explored extensively non-psychological sources of constraint on learning have not been explored. These areas of neglect may reflect the areas of organisational life which an externally contracted consultant such as Argyris feels he cannot hope to change.

Some of Argyris=s analyses of the structure of defensive routines covers similar territory to Bateson's (1972) earlier work on double bind communications, and the psychiatrist Ronald Laing's work on intra-family communication (Laing, 1961). They consist of layers of information: mixed messages, denial of the mixed nature of the message and ruling discussion of the nature of the message out of court (Argyris, 1994). The idea of hierarchy of logical types is present here in Argyris's understanding of face to face communications, but it is not extended further into a wider analysis of organisational learning processes on the scale of whole organisations. Even within his own analysis of inter-personal communications he typically refers only to first and second order learning, and not to the possibility of further levels of abstraction and constraint. This limitation may reflect the limits of individual and organisational awareness encountered during his interventions, an issue which Argyris has touched upon.

Present in much of Argyris=s analysis is a view that much knowledge within organisations is in tacit rather than explicit form. It is captured in his distinction between Aespoused theory@ and Atheory-in-use@ (Argyris, 1992) and built into his strategy for change, described below. His analysis of tacit knowledge is solely in terms of its use as a defence mechanism. ADefensive reasoning occurs when individuals make their premises and inference tacit...@. The functional value of tacit knowledge within organisations has not been explored by Argyris, though this issue has been addressed by other writers on organisational learning, especially Nonaka (1994), who in turn built upon earlier work by Polanyi (1966). For Polanyi knowledge becomes tacit at the individual level when it is consistently applicable and uncontested, however this is seen as an unconscious rather than deliberate process. Given the normal limits to people=s span of attention this process might better be described as an act of economy, not pathology.

Argyris=s strategy for change is essentially psychotherapeutic. Attention is focused on processes that have been routinised and made automatic. Actors' unawareness of the inconsistencies between espoused and theory in use is a key concern of the therapist, the external change agent. This is explored through the use of case studies, video and tape recorders if necessary, usually in a group setting. Here A people strive to make their premises, inferences and conclusions explicit and to subject them to public tests that are genuinely independent@ (1992:263). By bringing differences or contradictions into the public realm in a structured setting it is assumed that they will be resolved more easily than otherwise. As in psychotherapy, the role of the change agent is not to suggest particular resolutions but, to enlarge the range of choice the actors have to change their behaviour. Argyris also offers a complementary form of assistance, a theory of organisational defensive routines, which is intended to enable members of organisations to recognises their presence and thus presumably have some choice about how to manage them in the future.

Argyris=s recent work (1992) has been subject to some criticisms by others (Child, 1994). Firstly, that it is largely repetition of early work and there is no evidence of development. If the theory is reflexive it should help generate some learning about learning. Secondly, A...the root problem of how individual learning relates to, and is to be translated into, organisational learning is not really treated other than in terms of defence mechanisms.@ (Child, 1994:450), i.e. how new learning is prevented. Thirdly, although routines are recognised as a form of embodied and retained knowledge, Argyris=s conception of organisational memory is underdeveloped. Related to this, Child has suggested that a more dynamic view of the role of larger structures needs to be developed. AWhat kinds of organizational policies and practices help to reconcile the simultaneous conditions of diversity and consensus which many authorities agree are fundamental requirements for learning to take place ...This question is partly one of organisational design and amounts to more than just breaking down of defences through the interpersonal and group confrontations and discussions@ (Child, 1994:451).

The Fifth Discipline: Peter Senge

While Senge is a much more recent contributor to the field of organisational learning than Argyris his work has reached a wider level of public awareness much quicker. Unlike the ideas of many other private sector management gurus his views have also permeated quickly into the international NGO sector (Edwards, 1997). In July 1995 I heard his five key ideas being publicly cited in an NGO conference in the United Kingdom on scaling up, by Fazle Abed, the CEO of BRAC, the largest NGO in Bangladesh.

Senge=s most well known publication, a book titled The Fifth Discipline: The Art and Practice of the Learning Organisation (1990a), was a deliberate response to what was perceived to be an emerging market demand by company executives to learn about organisational learning (Galagan, 1991). The nature of that market demand was summarised in an introduction to an article by Senge, in the form of a quote from a CEO that AThe rate at which organisations may learn may be the only sustainable source of competitive advantage@ (Senge, 1990b:7).

Senge has distilled what he has learned from his own consulting work into the form of five Adisciplines@, bodies of practice that have to be learned (Galagan, 1991) These are: Asystems thinking@, Apersonal@ mastery@, Amental models@, Ashared vision@ and Ateam learning@. He acknowledges that each of these disciplines is built upon the work of other writers. Senge=s unique selling proposition, which differentiates his product from others, is the way he has packaged a number of ideas together. Along with the explanation of these concepts are various exercises which can be used to explore them in practice. Harris (1990) has suggested that one function of his book is as an advertisement for his training workshops, a well established marketing approach amongst management gurus (Huczynski, 1993).

Senge differentiates learning organisations from other organisations in terms of their greater adaptability. In a world where Aa full one third of the Fortune >500 industrials listed in 1970 had vanished by 1983" (Senge, 1990a) adaptability is seen as essential. He differentiates learning capacity in terms of two levels, which have been borrowed directly from Argyris, but relabelled as Aadaptive@ and Agenerative@. Adaptive learning is described as coping behaviour whereas generative learning is seen as creativity, and ultimately more valuable. Given this basic analysis, one of his clients posed a provocative question AI talk to people all over the country about learning organisations and the response is very positive@, says William O=Brien, CEO of the Hanover Insurance companies. AIf this type of organisation is so widely preferred, why don=t people create such organisations...@ (Senge 1990a:8). Senge=s response is to blame the competence of corporate leadership and then to prescribe what he calls AThe Leaders New Work@ (Senge, 1990b) which is to implement the five disciplines mentioned above. There is no acknowledgement that there might be some degree of fitness between prevalent forms of organisation (and the relative infrequency of examples of his ideal form) with current market conditions. Although Senge may argue that what is at stake is longer term survival Levinthal and March (1993) have sensibly pointed out that organisations need to survive the present before they can cope with the longer term.

The linchpin of Senge=s five disciplines is the idea of systems thinking, of Aseeing interrelationships rather than things, for seeing patterns of change rather than static snapshots@ (Senge, 1991). The perceived constraint is the linear cause effect model of thinking that he sees as inherent in the very structure of language. His solution is the development of diagram based mental tools that capture the idea of positive and negative feedback processes, and a set of common states and trajectories, called system archetypes, that can be identified. As with Argyris his systems view seems to have originated in a mechanical variant of cybernetics rather than from evolutionary theory (e.g. Ashby, 1956). Here the implicit desired state is homeostasis, whereas evolutionary theory gives more attention to the open ended and emergent dimension of living systems (Jantsch, 1987). It is symptomatic that the seven Aarchetypes@ that Senge focuses on are all problematic states to be avoided. His view of systems theory does not provide a guide to creativity.

Senge=s system discipline does have some more mundane advantages. It enables participants in his workshop to step back from an individualistic interpretation of change, where people may respond unconstructively to organisational problems with blame or by disengagement. Responsibility is both diffused and shared. Systemic views are also achieved by simpler means than system diagrams, simply by withdrawing staff to workshop settings where they are asked to focus on the organisation as a whole, not their day to day tasks.

The second of Senge=s disciplines is Apersonal mastery@, the institutionalisation of continuous learning at the personal level. It involves Acontinually clarifying and deepening our personal vision, of focusing our energies, of developing patience, and of seeing reality objectively@ (Senge 1991:40). Senge=s almost evangelical vision is one of staff of organisations being empowered to develop their full human potential. The constraints on this form of empowerment within the structural confines of organisations are not explored. Senge does not think that personal mastery can be commanded, rather it is a matter of such an approach being modelled by significant others, such as CEOs. This discipline contributes little to a theory of organisational learning.

The third discipline involves mental models. This requires Asurfacing, testing and improving our internal pictures of how the world works@ (1990a:14) Here Senge has borrowed extensively from Argyris, using the ideas of espoused theories, theories in use and defensive routines. Change comes about by making the existing models public, especially those that are widely shared within an organisation. Only then can internal and external contradictions inherent in the current model be dealt with. As with Argyris, this approach makes the unjustified assumption that public-ising knowledge, and it=s contradictions, within organisations will always be functional. It assumes internal contradictions can be resolved. The functional limits to the value of diversity are not examined. This response is understandable if it is assumed that organisations are inherently constraining and that the role of external agents is to help empower those within. In both Argyris and Senge=s analysis of defensiveness the background metaphor seems to be one of organisations as psychic prisons (Morgan, 1986). Within this framework, the external agent of change is the hero.

The fourth discipline is shared vision, that Awhich binds people around a common destiny. A genuine vision causes people to do things because they want to, not because they have to@ (1991:40). In some respects this seems similar to the development of new mental models. However the emphasis is on developing consensus, not exploring and testing. It is also about a higher level and more inclusive conceptualisation of what the organisation is doing. A positive effect on motivation is assumed to come about because the vision that is developed is not just that of the CEO, but all the staff. Senge uses the example of a hologram AWhen you add up the pieces of a hologram something interesting happens. The image becomes more intense, more lifelike. When more people come to share a vision, the vision becomes more real in the sense of a mental reality that people can truly imagine achieving@ (Senge 1990a:13). However this contradicts more common experience with documents such as mission statements which are developed to incorporate everyones= views, the more acceptable they are to a wider range of people the more bland and meaningless they become. One useful function of such a shared vision, blurred as it may be, would be to provide a sense of territory within which staff of an organisation felt they could safely experiment with more practically oriented mental models.

The fifth discipline is team learning. Senge argues that team learning is central because Ateams, not individuals, are the fundamental learning unit in modern organizations; unless the team can learn, the organization cannot learn@ (Senge, 1991:40). There is some foundation for Senge=s focus on teams in the more descriptive and analytic literature on organisational learning, especially that comparing the role of teams versus hierarchy (Romme (1996). However this literature also sees a positive role for hierarchy, in stabilising and refining the learning that is accomplished in teams. Senge=s focus on teams reflects his pre-occupation with new learning rather than the retention and exploitation of skills and knowledge accumulated in the past.

Senge=s strategy for improving learning via teams is to address the type of communication that takes place between team members. Dialogue is distinguished from discussion. The former is characterised by the suspension of judgement and normal social defences, and is seen as the key to team creativity. The use of such group processes has an established history dating back to the 1940's in the USA (Banner and Gagne, 1995). Limiting the exercise of critical thinking and critical behaviour in such environments encourages diversity in terms of how people think and talk about an issue. New solutions and ways of responding to a problem can develop. The problem that then arises, in the use of groups for both therapy and organisational development processes, is how to sustain these interpersonal processes in more day to day contexts. Senge does not address the question of how teams should relate to formal organisational structures, especially hierarchy. This question addresses learning at a truly organisational level, not simply learning by and between individuals.

Unlike Argyris, Senge cannot be criticised for developing his work in isolation. His work is based on the synthesis of other peoples work. The area where criticism has been made is the nature of his evidence that the application of his five disciplines produces valuable results. Harris (1990) has pointed out that where Senge has cited real world evidence of his ideas he relies heavily on the practice of three corporations, more specifically three CEOs, who have been closely involved with the MIT organisational learning programme for several years. From the point of view of theory building, as distinct from marketing, a major weakness is the lack of integration of the five disciplines into a unified theory of organisational learning (Hawkins, 1994).

Learning about Organisational Learning: Huber

Huber=s (1991) paper AOrganizational Learning: The Contributing Processes and The Literatures@ is one of a continuing series of review papers on organisational learning produced by academics over the last 15 years (Hedberg, 1981; Shrivastava 1983; Fiol and Lyles, 1985; Levitt and March, 1988; Huber, 1991, Dodgson, 1993; Hawkins, 1994; Nicolini and Meznar, 1995; Miller, 1996, Easterby-Smith, 1997). There seems to have been plenty of effort put in to try to learn about organisational learning.

In his introduction Huber dismisses the suggestion by some writers that the concept of learning in organisations should be limited to that which is intentional, or necessarily improves effectiveness, or necessarily changes behaviour. This view is largely consistent with an evolutionary view of learning. Variation is the basis of the evolutionary algorithm, and this can be generated both intentionally and unintentionally. The only notion of effectiveness built into evolutionary theory is a minimalist one, of survival and proliferation. Any more refined conceptions of effectiveness are themselves objects subject to evolution, and their prevalence will be dependent on local conditions.

Huber=s own definition of learning unfortunately does not build upon earlier theories of organisational learning, and is problematic in itself. AAn entity learns if, through its processing of information, the range of its potential behaviours is changed@ ...@an organization learns if any of its units acquire knowledge that it recognizes as potentially useful to the organization@ (Huber, 1991:89). There is no explanation given as to how a Apotential behaviour@ or Apotentially useful@ information could possibly be identified. This definition stands in contrast with the simplicity of an evolutionary view of learning as the selective retention of information, and with information being a difference that makes a difference.

Huber differentiates various attributes of learning. He distinguishes Abreadth@ of learning, involving the number of Acomponents@ in an organisation that obtain an item of knowledge. Then there is Aelaborateness@ in the form of the variety of the interpretations of this knowledge within the organization. Then there is Athoroughness@ of learning when Amore organizational units develop uniform comprehensions of the various interpretations@ of this knowledge (Huber, 1991:90). The difference with the evolutionary model is that this second level of learning is not about developing a common body of interpretation, as is also argued by many other writers on organisational learning (e.g. Fiol and Lyles, 1985; Daft and Weick, 1984), but the development of a common understanding of the variety that exists. There does not seem to be any functional value in this type of interpretation. It privileges diversity alone without regard for the need for effective joint action, for some order.

Huber summarises all previous work on organisational learning under four organising constructs which stand alone from any earlier or wider theory of organisational learning: knowledge acquisition, information distribution, information interpretation and organizational memory. Most attention is given to differentiating various forms of knowledge acquisition which by their labels are equated with learning. Information distribution, the nature of the flow of information within organisation is given only brief attention, although it is the structure of relationships between people that differentiate organisations from crowds, and presumably their learning as entities. Memory is treated separately and briefly, with most of the discussion focused on computers. In computers memory is located separately from processing units. Information input (acquisition) is also separate from processing (interpretation). Overall, Huber=s analytic structure seems strongly influenced by the metaphor of computer design. It is one that is hard to relate to the structure of organisations, where all staff clearly have a capacity to perceive, interpret and remember. People in organisations process information in parallel rather than serially, as is the case with most computers.

Within the discussion of knowledge acquisition Huber identifies five different forms of learning and eight other related processes, on the basis of his literature review. This proliferation of types can also be seen in other recent reviews of the literature (Miner and Mezias, 1996 (4 types), Miller, 1996 (6 types). This diversity is problematic because there is no explanation of the functional relationships between these types of learning. They are typologies, but not theories. In contrast, the evolutionary view of learning has one core conception of the process: learning as a selection process. Within this basic concept there is some differentiation, recognising different direction and frequencies of learning and different logical types of learning. These will be revisited later in this chapter.

Huber ends his review of the organisational learning literature with a series of conclusions, most of which are negative. As above, he notes the numerous and varied forms of learning that have been observed and sees this as a problem, especially in the light of the fact that Athere is little in the way of substantiated theory concerning organisational learning@ (Huber, 1991:107). Somewhat ironically, he also concludes that researchers have not made significant use of previous research to design or interpret their own research and there is little sign of cross-fertilisation or synthesis of work done by different research groups. The one exception he notes is the work of James March, whose work is discussed below.

Faced with these problems Huber then proceeds to do an ecological analysis of the nature of learning between organisational learning researchers, in contrast to the earlier contents of his article which gives no attention to ecologies of learning at all. He notes the competition involved in Ascience-making@, as well as the cooperation. There are incentives for developing intellectual products which can be differentiated from those of others. In turn, success is seen as an incentive to further specialisation, and specialisation can lead to isolation from the work of others. One possibility he sees is that Aas the landscape of research on organisational learning becomes more densely populated, much of what an investigator might do might have been done, and so the investigator is compelled to do work closely adjacent to and interfacing with the work of others.@ (Huber, 1991:108). In these conditions more attention will have to be paid to prevailing norms and procedures about intellectual property, and thus he argues, to more explicitly synthetic work.

Density is a significant structural feature of populations of organisations and one that has been explored by organisational ecologists researching founding and mortality rates in organisations, but not organisational learning as such (Amburgey and Rao 1996). It has also been noted as an important factor by journalists analysing the development of local economies such as Silicon Valley (Bronson, 1998). In Chapter Six some evidence will be presented of density effects on organisational learning within NGOs in Bangladesh. The density effects suggested by Huber are in effect a form of feedback loop between micro and macro-level learning processes. Success at the micro-level leads to macro-level changes, which in turn affect further micro-level changes.

Huber=s final conclusion, which he argues is of singular importance, is that AWith very few exceptions, work on organisational learning has not led to research-based guidelines for increasing the effectiveness of organisational learning.@ (Huber, 1991:108). The theory of organisational learning that is developed in this thesis does address this task.

The Ambivalence of Learning: James March

James March is the second most widely cited writer on organisational learning and the only one amongst those listed on Table 4.1 having two papers on organisational learning which have both been widely cited within the organisational learning literature (Levitt and March, 1988; March 1991). Although his work on organisational learning is influenced by evolutionary theory (March 1994) it also bears the influence of a range of past concerns: information processing, coalitions, organized anarchy, new institutionalism, bounded rationality (Padgett, 1992). His views on organisational learning are an extension of his earlier concept of the Abounded rationality@ of individual agents (Padgett, 1992), and contrasts with models of organisations that stress calculative rationality (Levinthal and March 1993).

AMagic would be nice, but it is not easy to find@ is a concluding statement in a recent paper by March (Levinthal and March 1993) which expresses his sceptical view of organisations, and the possibility of constructive interventions. His writings emphasise the limited, ambivalent, and sometimes paradoxical nature of organisational learning. These are not easily translatable into marketable imperatives, as is the case with Senge or Argyris. March=s views pose a challenge on the theoretical and practical level. How can the limitations and ambiguity of learning be managed, within an evolutionary framework? And given such a sceptical perspective what is the value of March=s suggestions for interventions into the process of organisational learning? Both of these questions are examined below.

Limitation and ambivalence are seen to arise from three sources: A...partly from inadequacies of human cognitive habits, partly from features of organisations, partly from characteristics of the structure of experience.@ (Levitt and March, 1988:335). In the case of the latter, organisations are simultaneously faced with Acomplexity@, Apaucity@ and Aredundancy@ of experience. This is a more differentiated version of the problematic introduced at the beginning of this thesis, diversity and the problem of how to manage it. Although March explores these in terms of their implications for how organisations should respond, they are not followed up in his subsequent writing.

In the 1988 review paper March briefly summarises the literature on the limitations of human beings as observers, statisticians and analysts of causation (Levitt and March, 1988:323). The emphasis on the bounded rationality of members of organisations is not problematic for an evolutionary theory of organisational learning. Variation is the basis of future learning, and can arise from intentional and unintentional action. The combination of these two sources is in fact likely to generate a more useful range of variations than one alone. Purposeful variations are likely to involve more functional sub-units of knowledge or practice. It should be noted here that while evolution does not require purposeful action it is not correct to say, as some claim, that evolution requires the absence of purpose (Economist, 1998).

The third cause mentioned by March are features of organisations. Many of the problems of learning described by March are problems because of one general feature: the hierarchical structure of learning within organisations. There are interactions between agents at different levels, and between agents within the same levels, which complicate the evaluation of achievements at any one location. What is good at one level, one location or one time is not necessarily good for another. Three examples taken from the 1988 review paper are summarised below. One concerns the structure of routines, another the structure of interpretations and another the structure of agents within organisations.

March has defined organisational learning as Aroutine-based, history-dependent, and target-oriented@ (Levitt and March 1988). Routines are defined rather broadly as Athe forms, rules, procedures, conventions, strategies and technologies around which organisations are constructed and through which they operate@ They can be seen as rules, and sets of inter-connected rules, of the kind described as the basis of Burn=s evolutionary theory, introduced in Chapter Three. Routinisation involves the progressive reduction in error through repeated practice. The extensive literature on learning curves found in practices across a range of industries is cited by March as evidence of the value of repetition itself, as a basis for effective learning. However, what March calls Acompetency traps@ can arise when repeated experience leads to skills developing to such an extent that later movement to better techniques can be seen as too difficult because of the re-learning costs (including the lost investment made in past learning). March has also called this an example of Athe myopia of learning@ (Levinthal and March 1993). Such (subsequently defined) inefficient sub-routines can become embedded in larger more efficient routines which survive and are replicated on a large scale. The classic example being the QWERTY keyboard design carried over from early typewriter design to computer keyboards (Levitt and March 1988). March points out that Acompetency traps result in organizational histories for which broad functional or efficiency explanations are often inadequate@ (Levitt and March, 1988:323). The reason is that functional value is typically decided on a local and immediate basis.

Another form of ambivalence can be seen in the hierarchical structure of interpretations prevalent within organisations. According to March, organisations spend time developing collective understandings of events. Not only are some events remembered but they are framed in a particular way. Although March recognises that changes in these frames constitute second order learning as described by Argyris and others he emphasises the resilience and potency of what has been learned at this secondary level. The problem is that Astories, paradigms and beliefs are conserved in the face of considerable potential disconfirmation...@ (Levitt and March, 1988:324). The utility of particular interpretative frames is often more difficult to disconfirm than perceptions of events they frame. Furthermore Awhat is learned appears to be influenced less by history than by the frames applied to that history...@ (Levitt and March, 1988). There is in practice a complicating inter-dependency between observation and interpretation, between levels of learning. Examples of such relationships will be evident in the analysis of the results of CCDB=s participatory monitoring system, discussed in Chapter Eight.

The already complicated process of interpreting the meaning of events, such as success, by individuals is compounded by interaction effects when the agents concerned are not seen as isolated units but as part of an Aecology of learning@, to use March=s phrase. When such agents are located in the same organisational structure AConflict and decision advocacy within putatively rational decision processes lead to inflated expectations....New organizational leaders are inclined to define previous outcomes more negatively than are the leaders who preceded them...Different sub-groups in an organisation often have different targets and evaluate the same outcome differently.@ (Levitt and March, 1988:325).

March himself gives some recognition to the hierarchial nature of learning: A...learning takes place at several nested levels. In such multi-level learning, organisations learn simultaneously both the discriminate between routines and to refine the routines by learning within them@ (Levitt and March 1988:322). Learning is also happening in parallel within particular levels: Aorganisations are collections of sub-units learning in an environment that consists largely of other collections of learning sub-units@ (Levitt and March 1988:331). In his 1994 paper on AThe Evolution of Evolution@ referred to in Chapter Three he takes this view further A...units are nested in space and in time. Firms are nested in industries which, in turn, are nested in societies. The short run future is nested in the long-run future@ (March, 1994:46). Both temporal and spatial forms of nesting will be evident in the analysis of CCDB=s structure and routines in Chapter Seven. The design of the participatory monitoring system, detailed in Chapter Eight, makes explicit use of the same features.

In the same paper March outlines his views of the key changes in evolutionary theory. One was a move Afrom an emphasis on using evolutionary theories to predict history to an emphasis on the engineering of history.@ (March, 1994:39). In the process of discussing this change he argues that there are three broad kinds of interventions possible:

• A...altering the possibilities for transmission, retention and retrieval of the lessons of history.@
• A...altering the structure of interactions among units of evolution@
• A...managing the exploration/exploitation balance@

The first of these focuses on the retention component of the variation-selection-retention algorithm. Examples given by Levitt and March (1988) are the invention of the printing press and the construction of computer data bases. Although such developments can dramatically expand the total of what can be learned (as in retained) there are also risks that their widespread use can discourage use of other information (e.g. oral testimony). In the case of CCDB part of the design of the participatory monitoring system involves the formal documentation of staff knowledge about events in their locations which had not been previously subject to regular written documentation. Organisational memory of those events was enhanced.

The second intervention is less explicitly dealt with in March=s recent papers. In the 1988 review March offers only a brief comment on the role of organisational structure: AOrganisations facing complex uncertainties rely on informally shared understandings more than do organisations dealing with simpler, more stable environments (Ouchi, 1980).@ (Levitt and March 1988:327). The significance of unpredictability is consistent with the theory of learning introduced in Chapter Three. What is needed is a further development of this conception of the difference between formal and informal understandings and their relationship to hierarchies and teams, which Senge and Argyris argue are critical to new learning.

In Chapter Three a distinction was made between hierarchies and heterarchies. It was suggested that the process of learning involves a transition from heterarchy towards hierarchy. The transition involves a selection process, whereby many links are tried out, most are abandoned, a few are retained and their relative weight may differ. There is a process of simplification and stabilisation, a transition form generalist to specialist structures. Teams can be seen as a specific class of heterarchies. A group of people randomly connected to each other (by one communication link per person) will take the form of a heterarchy. There will be different levels of connectedness. Some people will be connected to many others; some will be connected to very few. Those with many connections to others will find they share many connections with others who are well connected, but also have some connections that are not shared. The important difference about teams is that a newly formed team is likely to take the form of an unstable heterarchy. The prevalence of links (in the form of communications) to particular members may vary from moment to moment. But over time leaders may emerge, and communication links may become more stable and structured.

In another recent paper March (Levinthal and March 1993) has explored what appears to be another dimension of the environment, other than unpredictability, as a factor effecting organisational structure. He points out that some environmental problems can seen as Adecomposable@ and others less so. Decomposability is the possibility of Acarving nature at the joints@ (Dennet, 1995:37). Other well known writers on organisational learning have also emphasised the importance of this feature of an organisations environment (Daft and Weick, 1984:287). Romme (1996) has argued that teams are the best structure for dealing with problems which are highly interconnected and least decomposable, and hierarchies for those which are more so.

This analysis neglects the significance of time. The use of either hierarchy or teams can be seen as an anticipatory interpretation of the nature of the problem. Is the problem decomposable on the basis of existing knowledge or not?. Only if a team persists over time as the main structure used to address a problem could it be said to reflect the complexity of the problem being addressed. However, where problems are complex but static, specialised structures of knowledge do seem to emerge over time. For example, specialisms in the field of physics. On the other hand, it is perhaps not surprising that one body of specialist knowledge has not accumulated so visibly in the social sciences, where the subject matter is much more changeable.

The third intervention suggested by March (1994) involves identifying the appropriate balance between what he calls exploration and exploitation. Exploitation is the use of past knowledge, exploration is the development of new knowledge (March, 1991). In March=s words intervention involves A...manipulating the level of risk taking or the salience of diversity relative to unity...@ (March, 1994:45). In Campbell=s terms what is involved is the balance of emphasis given to variation versus selection. The use of quality control procedures in industry is a variation reducing practice aimed at maximising the value of an organisation=s current knowledge (Winter, 1994). On the other hand, the users of participatory methods in development programmes generally seek to increase diversity of knowledge within organisations (Holland and Blackburn, 1998). The participatory monitoring system described in Chapter Eight enhances variation in knowledge available at the field level, but uses the existing organisational hierarchy to impose a series of selection processes which subsequently reduces that diversity to a level manageable by CCDB senior staff.

March argues that the relative prevalence of exploration versus exploitation is sensitive to the rate of change in the environment. This is consistent with the theory of learning introduced in Chapter Three. However, when he examines the risks and returns to organisations the evidence as a whole suggests that there are internal incentives that provide a strong bias towards exploitation and away from exploration. ACompared to returns from exploitation, returns from exploration are systematically less certain, more remote in time, and more organizationally distant from the locus of action and adaptation@ So much so that Athese tendencies to increase exploitation and reduce exploration make adaptive processes (i.e. exploitation based organisational learning) potentially self-destructive@ (March, 1991:73). This view will help explain the inward looking and conservative nature of learning that is identified within Bangladeshi NGOs discussed in Chapters Six and Seven. According to March=s analysis, if exploration exists on any significant scale this must be due to incentives stemming from the nature of the surrounding environment. This view is consistent with the trend towards ecologically based theories of learning in both humans and animals in the second half of this century (Johnston and Pietrewicz, 1985).

What is neglected from March=s recent analysis of organisational learning is attention to where learning is taking place. Particular organisational structures, and routines carried out within those structures, are a selection from a much wider set of possibilities. In any location within an organisation not all events can be documented or analysed, and choices have to be made. Knowledge will by necessity become specialised in some areas and neglected in others. Where attention in the form of limited staff time and resources is devoted must be a matter of some significance. Even where resources are concentrated choices will also need to be made between the degree of exploitation versus exploration that will be involved. Although the overall balance of exploitation versus exploration in an organisation may be influenced by a CEO, it is also likely that this parameter will be tuned differently in different areas of an organisation=s operations. For example, a Finance Unit being more exploitation oriented versus a Research Unit being more oriented to exploration.

An Evolutionary Theory of Economic Change: Nelson and Winter

Nelson and Winter=s (1982) An Evolutionary Theory of Economic Change is now seen as one of the key texts in contemporary evolutionary economics (Andersen, 1994), though the roots of the latter in early American institutionalism are acknowledged (Hodgson, 1994). In adopting an evolutionary view of economic change Nelson and Winter (1982) have sought to overcome some basic problems with traditional economic theory: firms are assumed not to make mistakes or be less than perfect, variation between firms is not significant and change is not endemic. Within Nelson and Winter=s framework firms are seen as valuable integrated packages of competencies, but not perfect nor all identical. It is the way they manage knowledge which distinguishes and explains them as a form of organisation This view contrasts with a more contractual explanation of the firm which emphasises the successful management of the costs of interactions between agents (Hodgson, 1994). The competency view allows room for creativity and not just conservation of a static pool of resources.

The core of their view of competency is the idea of a routine, an interpersonal process that they see as the organisational equivalent of an individual=s skills: A...it is firms, not people that work for firms, that know how to make gasoline, automobiles and computers@ (Nelson and Winter, 1982:86). The structure of the firm stabilises and preserves these bodies of knowledge and they are in turn the source of competitive advantage that enables firms to survive and grow. Nelson and Winter=s theory assumes a multi-level learning process, one involving the variation, selection and retention of routines within the firm and variation, selection and retention of firms within industries. These are the two aspects of their theory which are relevant here: the idea of routines as a core element and the structure of the selection processes involved.

Routines as described by Nelson and Winter are now widely acknowledged as important elements of organisational learning (Huber, 1991; Miller, 1996). They are the one part of the organisational learning process where March feels the evidence of effective learning is least ambiguous (Levitt and March 1988). Nelson and Winter go further and Amake the case for >routines= as a fundamental unit of analysis in the evolutionary approach to organizations@ (Winter, 1990:271). Routines as genes is seen as a useful metaphor, and this usage is compared favourably to the idea of Amemes@, a term suggested by Dawkins (1976), for use in discussions of cultural evolution. The main problem with this approach is one of observation, how to identify routines as entities. Their main characteristic is the fact that they are repetitive activities and they preserve their identity across repetition. While routines are recognisable in these terms there is no proposed procedure for unambiguously distinguishing between routines, although they may appear to operate on very difficult scales. Difficulties are compounded when it is noted that many individual skills and organisational routines have considerable elements of tacit knowledge, much of the nature of what exists is not publicly visible (Winter, 1994).

Nelson and Winter=s advocacy on behalf of routines is part of a larger debate in evolutionary theory about what are the units of selection that are the meat of the evolutionary process (Dennet,1995) . In this thesis a different approach to the units of selection problem has been taken to that proposed by Nelson and Winter. It is argued in this thesis that there are no basic units in the sense of entities. That which is subject to selection is difference, as described in the discussion of Bateson=s ideas in Chapter Three. There are potentially a large number of differences between various routines, but some of those and not others will come to be seen as the most important within a particular organisational setting. There are in fact competing classificatory schemes, both at the individual and organisational level. This view gives observers a significant interpretive role in the evolutionary process that Nelson and Winter describe, a role that others such as Daft and Weick (1984) have emphasised in their view of organisational learning. It is also one that was introduced in Chapter Three as an essential part of the process of cultural evolution, of which organisational learning is one part.

Winter (1986:174-5) also recognises that AOrganizational routines form quasi-hierarchical structures@ and point out that Athe hierarchy of routines generally parallels the hierarchy of authority in an organization. In an organisation=s formal system of authority, the power to authorise departures from existing routines, as well as responsibility for investigating and implementing changes in routines, typically resides in individuals of higher rank than those responsible for the execution of routines@. This is the organisational equivalent of Bateson=s hierarchy of recursiveness introduced in Chapter Three. Winter also points out that as the focus moves up Athe hierarchy the subtlety and complexity of the individual skills being exercised typically rises@. This is understandable because at the highest level, in the model, agents would be dealing with the most abstract information (differences about differences about differences...). They would also be dealing with the least frequent events, changes in parameters of routines, rather than the management of events processed by those routines.

However, this is undoubtedly a simplified model. Staff at any level in organisations such as CCDB usually have more to do than simply monitor and tune the parameters of the work of those below them. They can have other tasks (routines) specific to their own position and appropriate to the scale of aggregation available at that position. They generate content as well as regulate process. In the case of CCDB, Project Officers write their own monthly reports about developments within their project area, as well as supervise their junior staff.

Not only are routines nested, but contents are as well. As March (1994) has noted, accounts of events from the lowest levels are aggregated within accounts produced at higher levels. Here there is another example of the trade-offs between different forms of learning, of the type emphasised by March (1991). The large scale aggregation of accounts takes time, and encourages some limitation on the frequency with which such events take place. Smaller scale aggregations allow greater frequency. Examples of how such trade-offs are managed will be given in the description of the development of CCDB=s participatory monitoring system, in Chapter Eight.

Nelson and Winter=s description of a hierarchy of routines is an idealised model. Variations from that norm may have implications for the type of learning that can take place. Nelson and Winter=s description seems to assume that power to control parameters of routines will necessarily be decentralised. In practice managers of organisations may keep much of that power to themselves. The difference between these two possibilities can be seen in terms of degree of specialisation, of the organisation accumulated differentiated knowledge in different locations, versus it all being held by one key individual. There is also another possible extreme, power may be delegated so completely to what would be otherwise called field staff that the knowledge they hold may become inaccessible to the chief executive. Some centrally funded but locally implemented research projects carried out by UK NGOs have experienced this problem (Davies, 1997a).

Organisations as Interpretation Systems: Daft and Weick

Daft and Weick have written extensively on organisational learning over the past two decades, focusing particularly on the role of interpretation. In their most cited paper Daft and Weick (1984) make four working assumptions about the nature of organisations and how they are designed and function. The first is that Aorganisations are open social systems that process information from the environment...relevant to their survival@ (Daft and Weick, 1984:285). This will be evident in the case with CCDB, as described in Chapter Seven. What is not acknowledged in the Daft and Weick paper is that this process is undertaken by individual staff within the organisation who are also mindful of their own survival within that organisation and who view the internal environment as equally, if not more important. Views of the external environment are often mediated by those of the internal environment.

Secondly, it is assumed that A...the organisational interpretation process is something more than what occurs by individuals. Organisations have cognitive systems and memories@, enabling the preservation of knowledge despite the turnover of staff. AReaching convergence characterises the act of organising...the thread of coherence among managers is what characterises organizational interpretations@ (Daft and Weick, 1984:285). The prevalence of views is seen as an outcome of the process of organisational learning. The cognitive systems leading to these outcomes are the processes of interaction between staff members, both formal and informal. This is consistent with the analysis of organisational learning developed in this thesis.

The third assumption is that AWhen one speaks of organisational interpretation one really means interpretation by a relatively small group at the top of the organizational hierarchy@ (Daft and Weick, 1984:285). Daft and Weick emphasis that at lower levels all knowledge is partial. AOrganisations can be conceptualised as a series of nested systems, and each sub-sector may deal with a different external sector.@ (Daft and Weick, 1984:285). The CCDB case study will suggest a slightly different interpretation. While the top of the hierarchy does have an all-inclusive view of the organisation it is at the necessary cost of leaving behind much of the detailed knowledge held by lower units. In practice all units have partial knowledge.

AThe fourth assumption is that organizations differ systematically in the mode or process by which they interpret the environment. Organizations develop specific ways to know the environment@ (Daft and Weick, 1984:286). Given the diversity of organisations that exists this view seems almost commonsensical. The question then is what are the main differences in the ways in which organisations interpret their environment, and what accounts for these differences. Daft and Weick do offer a view, detailed below.

Daft and Weick see the process of interpretation in similar terms to this thesis. There is a surplus of experience to make sense of, Aan ocean of events@. Organisations are selective, Aattending to some [events], ignoring most of them@ (Daft and Weick, 1984:286). Interpretation is a process both of active sense making by individuals and a social process of enrolling others in these constructions. Learning is also seen in similar terms to this thesis, but within a more limited focus. AOrganizational learning is defined as the process by which knowledge about action outcome relationships between the organization and the environment is developed@ Action outcome relationships being a form of if-then statements, a provisional rule relating to external events.

Daft and Weick use the concept of a feedback loop experienced in individual learning to explain the refinement of this knowledge over time. Concepts are developed and then tested in practice, repeatedly. AOrganizational interpretation is analogous to learning a new skill by an individual@ (Daft and Weick, 1984:286) In this thesis the concept of a feedback loop has not been central in the explanation of organisational learning. Experience of practice within organisations and in sectors of organisations is much more mediated, and less direct than is the case within individuals. In addition, publicly voiced representations of experience often have to reconcile multiple conflicting constraints, not just one as implied by a feedback loop. In organisations, and even more so in less structured sectors of organisations, it may be more appropriate to think in terms of interpretations of experiences being located in network of connections rather than a single loop. This view is consistent with the ideas of teams, heterarchy and hierarchy as important differences in organisational structures.

Daft and Weick have proposed that there are two important differences in how organisations view their environments. They are: A(1) management=s belief about the analyzability of the external environment, and (2) the extent to which the organization intrudes into the environment to understand it@ (Daft and Weick, 1984:287). In the first case the environment may seem concrete, with events and processes being hard and measurable, or at the other extreme, the results of inquiries may be more ambivalent in meaning, reflecting the influence of the inquiry as much as the external world. In the second case, an organisation may actively seek information about the environment, or be more passive and accept whatever information the environment makes available. There is some recognition by Daft and Weick that choices made between these perspectives, are themselves learned behaviour, based on prior experience.

What differences do these differences make? Daft and Weick combined the two differences together to generate provide four different Ainterpretation modes@: Undirected Viewing, Conditioned Viewing, Enacting, and Discovering. They argues that each particular combination leads an organisation to adopt a particular form of strategy formulation and decision making process, which others have noted in organisations. They argue that this compensates for the fact that while A..one of the widely held tenets in organisation theory is that the external environment will influence organization structure and design...The paradox is that research into the environment-structure relationship gives scant attention to interpretation@ (Daft and Weick, 1984:292).

However Daft and Weick did not complete their task. If the two differences are important, then they might be expected to make a difference, not only to behaviour within an organisation, but also to the organisations relationship to its environment. In a particular industry or sector, certain Amodes of interpretation@(e.g. active/analysable) might be expected to generate more benefit to an organisation than others. If they did not, they must be of less significance and interest in terms of the scale of explanation they can offer to organisation theorists. Daft and Weick cite examples of conditioned and undirected viewing being found in companies within the same industry (clothing manufacture) and active and passive approaches co-existing in others (listed companies relationships with shareholders). While some diversity might be expected in any environment Daft and Weick do not make any suggestions, for further exploration, about the type of conditions where one particular interpretative schema might be expected to be more prevalent than others.

There are other basic differences in the way organisations have learned to interpret their environment which are more significant. These are when and where organisational attention should be specialised. Here linkages can be made without difficulty between activities within organisations and events within their environments. For example, ethical investment funds who want to invest in appropriate businesses make use of specialist research units to investigate the behaviour of firms they may be interested in. In the process they have to decide how much money they should invest in such research. This decision has some relationship to Daft and Weick=s active/passive distinction. However, it seems more reasonable to view perceived analysability as a secondary concern, coming into play only in those areas where it is recognised that attention must be focused. Analysability cannot be assessed in areas that are not being attended to in the first place. However, it is conceivable that analysability may have some feedback effects on subsequent willingness to maintain attention in a particular direction.

A Dynamic Theory of Knowledge Creation: Nonaka

As detailed in Table 4.1, Nonaka=s (1994) paper has been widely cited, given how recently it has been published. While Nonaka makes use of Bateson=s concept of information his theory of organisational learning is based on a variant of evolutionary epistemology that emphasises self-organisation, called Aautopoiesis@ (Maturana and Varela, 1980) , and includes some elements of phenomenology, particularly the emphasis on the role of intention and prior knowledge in perception.

A central idea in Nonaka=s paper is that of tacit knowledge, a concept that has been an important part of other theories examined above, especially that of Argyris. Citing Polanyi (1996:4) he notes that AWe can know more than we can tell@. This view is consistent with the evolutionary epistemology introduced in Chapter Three, where it is argued that the structure of organisms embodies what they have learned, how they have been in-formed by their environment (but they cannot necessarily tell us about it).

Nonaka=s argument is that organisational knowledge is created through a continuous dialogue between tacit and explicit knowledge. Four modes of knowledge creation in organisations are identified: socialisation (tacit to tacit), externalisation (tacit to explicit), combination (explicit to explicit) and internalisation (explicit to tacit). It is suggested that there is an ideal balance between these processes but that this is often not achieved. The proposed ideal involves a form of cycling through each of these states. His analysis of learning in terms of these changes is heavily influenced by a prescriptive perspective, of how these changes can be affected. There is little in the way of description of how they are managed in day today life.

Nonaka emphasises that his theory explains the process of knowledge creation in organisations, unlike others that see organisational learning in more reactive and homeostatic terms. In his view A...the articulation of tacit perspectives...is a key factor in the creation of new knowledge@ (Nonaka, 1994:16). However, it seems that it is more appropriate to see articulation as part of a process of making new and better use of existing knowledge. That which is tacit has already been learned, the person and organisation has been in-formed. This seems to be given some recognition at the beginning of the paper when reference is made to four modes of knowledge conversion, rather than creation. The model seems to give minimal attention to how an organisation=s body of knowledge is informed by experience of the outside world. The theory itself mirrors the same problem, there are no acknowledged antecedents for this four phase cyclic view. The same learning problem seems evident when Nonaka subsequently claims that the principles he elaborates A..have a more general application to any organisation, either economic or social, private or public, manufacturing or service, in the coming age despite their field of activities as well as geographical and cultural location@ (Nonaka, 1994:34).

While the existence of active learning (the generation of new knowledge) has been described in this thesis as dependent on the nature of the environment (especially its unpredictability), and one rationed by the costs involved, the process described by Nonaka is described as an inherently unconstrained and expansive process. AThe interactions between tacit knowledge and explicit knowledge will tend to become larger and faster in speed as more actors in and around the organisation become involved@ (Nonaka, 1994:20). The isolation of this process, and the theory itself, seems to reflect the solipsism inherent in the background theory of autopoiesis.

Despite these problems Nonaka=s four stage cyclic theory is developed as a means of enhancing organisations= abilities to accumulate new knowledge. As with Argyris and Senge, the conservation of past knowledge is not a major concern. Within the new learning agenda it is thus not surprising to find that self-organising teams are seen as the most appropriate organisational form. In addition to arguing the merits of an idealised group based learning process Nonaka also makes two proposals for enhancing organisational learning, focusing on management style and organizational design. These are called the Amiddle-up-down@ management model and the Ahypertext organisation@.

The first proposal is yet another Anew model of management@ (Nonaka, 1994:30). This advocates recognition of the special role for middle managers in organisational learning, as brokers, mediators and catalysts between the different conceptualisations of the world that are associated with positions at the top and bottom of organizational hierarchies. These differences involve abstract/concrete, small/large scale and short/long term views of events. It is of value to this thesis because it reinforces the idea of organisational structures as large scale conceptual structures. Nonaka does add value to this view when he emphasises middle managers role in bridging top management dreams with field staff reality. There is a parallel in the artificial neural network model referred to earlier in Chapter Three. The earliest work in this field was not as successful as expected. It was only with the introduction of a middle level layer of nodes that the practical capacities of such systems was dramatically expanded (Aleksander and Morton, 1991).

The second proposal is actually another attempt to integrate contrasting bodies of knowledge within an organisation into a single model of organisational structure, albeit a normative one. Three levels are conceived. The first is a Aknowledge-base layer@ which consists of tacit knowledge, associated with organisational culture and procedures and other forms of what could be considered common knowledge. The second is the Abusiness-system layer where normal routine operations are carried out by a formal, hierarchical, bureaucratic organisation@ (Nonaka, 1994;33). The third layer is the Aproject-system layer...where multiple self-organising project teams create knowledge@. Nonaka points out that ANon-hierarchical or >heterarchical= self-organizing activities of teams are indispensable to generate new knowledge...on the other hand, a hierarchical division of labour is more efficient and effective for implementation, exploitation and accumulation [retaining] of new knowledge@ (Nonaka, 1994:32). There is some correspondence here with the idea of a continuum of structures developed earlier in this chapter. However, Nonaka simply equates, teams with heterarchies without distinguishing between them, in terms of stability of their structure, as I have suggested above.

What is new in Nonaka=s framework is the Aknowledge-base layer@. It refers to knowledge which has already been acquired, but its ownership is not limited to specific locations within the organisation, unlike that held in the forms of various specific routines and structures. In Chapter Seven knowledge about relative status differences between staff within CCDB is identified as one form of such common knowledge. Although not mentioned by Nonaka, it is this form of knowledge which connects organisational knowledge with the wider body of knowledge held in the culture at large. Common knowledge is a form of proliferation, one expression of learning defined in evolutionary terms. It is in effect what the organisation has learned most thoroughly and for that reason is a form of knowledge that would probably best survive any radical dismemberment or restructuring of an organisation.

Nonaka argues that good organisational design should enable a quick and efficient switch to take place between hierarchical and heterarchical forms of learning. While this is plausible, it shares a weakness with March=s analysis referred to earlier. That is, important decisions have to be made by organisations about not just how best to learn, but where to learn. Although Nonaka gives some recognition to the role of hierarchy in learning (above) there is a risk in misconceiving the role of teams. As has been emphasised by March and others (Levitt and March, 1988), a substantial amount of learning also takes place within day to day routines located within existing hierarchical structures. The prevalence of different structures (teams and hierarchies) reflects the nature of the problems being solved, and interpretations of those problems, not simple choices of whether to learn or not.

Nonaka=s overall theory is contained in two dimensional view of organisational learning. One dimension, called the epistemological, covers the range from tacit to explicit knowledge. The idea of tacit knowledge is not itself differentiated. The second he describes as the ontological and concerns the social structure of knowledge, involving different groupings of people. The main example of the latter is the three level view of structure given above. What is needed is further differentiation in terms of direction, where those structures develop, or not.

4.4 An Interim Summary: Attributes of Organisational Learning

Building on the initial analysis in Chapter Three, five main attributes of organisation=s learning behaviour can be identified. They vary in the extent to which they have been given attention by the writers on organisational learning reviewed above.

1. Frequency of learning: Individuals and organisations update their knowledge about different parts of their environment with varying frequencies. Because of the costs involved, frequency is likely to be greater in the case of those events seen as the most important, but this depends on the expected speed of change of those events. Organisational routines can be ordered in terms of a temporal hierarchy, from the very slowly iterated to the very frequently iterated. Nelson and Winter give the most attention to this feature, though the alignment of organisational hierarchies with the observation of events on different temporal scales is recognised by March and Nonaka.

2. Direction of learning: Individuals and organisations are selective in what they learn. Attention is focused in some directions and not others. Specialised knowledge develops in those areas where attention is frequently directed. This is evident in the structure of individual=s category systems and in organisational structures. Little attention is given to this feature by any of the writers above. Along with frequency of learning, the direction of learning will be a central concern in the analysis of organisational learning within CCDB, in Chapter Seven.

3. Depth of learning: Individuals can develop interpretations of events which contain multiple levels of logical types of information, connecting many concrete observations with very abstract distinctions. Within organisations the parameters of some routines can be controlled by other over-arching routines. This is the most widely recognised attribute of learning behaviour, mentioned in almost all reviews of organisational learning, and the writers above.

4. Scale of learning: Individuals can learn about events taking place on different scales, ranging from the very small to the very large. Larger scale events are more likely to contain internal diversity, and unpredictability, and can therefore be more difficult to learn about. Organisations are able to learn about events on larger scales than individuals. They can manage more diversity. The process of managing diversity involves both dis-aggregation (specialisation) and aggregation. Having multiple levels of learning (depth) enables diversity to be aggregated on a large scale. Aggregation of experience can take place on an atemporal (geographic or demographic) and a temporal scale. Ambitions for aggregation on one scale will constrain those on the other, because resources are limited, both within individuals and organisations. Perhaps because it may seem so self-evident, scale of learning is not an attribute that is frequently mentioned by writers on organisational learning.

5. Openness of learning: Much of what has been described above are ways of describing what has been already been learned. But learning is a process open to the future. The degree of openness is evident in the extent to which people and organisations seek confirmation versus novelty, favour exploitation over exploration. This will be evident in the form of variation in particular practices, and awareness of that variation. It will also be evident in the relative importance given to the use of structures such as teams versus hierarchy. Openness is a feature of learning that has been recognised in both prescriptive and descriptive analyses of learning (e.g. Argyris, Senge, March).

There is a sixth feature of learning behaviour which is less easily observed. It has been argued in this thesis that organisational learning is homologous with individual learning, and in turn, with the wider processes of evolution. Although organisations are of a different logical type to individuals they are made up of individuals. A significant commonality of process should be expected. However, there are differences between individuals, organisations and populations of organisations that may make a difference to the nature of learning that takes place at these different levels. These are in the density of the component parts and the stability of their relationships with each other (Jantsch 1987). Huber=s speculative analysis of ecologies of learning is the only one of the organisational learning theorists examined above, that has touched on this attribute and its possible consequences. Its use will be explored in the analysis of Bangladesh NGO sector in Chapter Six.

4.5 Organisational Learning Within Development Studies

The body of literature on organisational learning in non-government organisations is very small when compared to that examined above. In a search through the International Development Abstracts only five articles about learning and aid organisations were identified. As indicated at the beginning of this chapter, part of the problem is that many of the papers that do exist are in the form of grey literature, not published in journals, but circulated internally and at workshops and conferences (e.g. Slim, 1993; Birch, 1996; Britton, 1995 ). There is also a wider variation of terms used to describe organisational learning related processes, making the boundaries of the field less distinct and searches more difficult.

There are three strands of writing within the sphere of Development Studies which do relate to organisational learning. The oldest is that concerning process approach to development projects (Mosse et al. 1998). This includes the work by Korten (1980), Rondinelli (1983) and Uphoff (1986, 1992) on learning process approaches to development projects and, less directly, work on process monitoring (Mosse et al. 1998). A second strand is the literature on evaluation of development projects, some of which contains some explicit references to organisational learning (Forss et al. 1994; Marsden and Oakley, 1990). More recently, a series of papers have been produced by NGO staff specifically on organisational learning within NGOs (Slim, 1993; Birch, 1996; Britton, 1995; Howes and Roche, 1996; Edwards, 1997). Each of these three strands will be examined in turn.

4.5.1 Process approaches

A Learning Process Approach: David Korten

The concept of a learning process approach to development projects is strongly associated with the name of David Korten (Mosse et al. 1998), especially his widely quoted 1980 paper on a learning process approach to community organisation and rural development (Korten, 1980). In that paper he reviews the lessons learned from five AAsian Success Stories@ (one cooperative, three NGOs, and one para-statal). Summarising them all Korten says Athey had achieved a high degree of fit between program design, beneficiary needs, and the capacities of the assisting organisation@ (Korten, 1980:496). This definition, couched in terms of fitness, is close to that used within an evolutionary perspective.

However Korten goes a step further and specifies the types of fitness which he thinks is important in all cases. They are: (a) between beneficiaries needs and organisational resources, (b) between the means by which beneficiaries are able to define and communicate their needs and the processes by which the organisation makes decisions, (c) between the task requirements of the program and the distinctive competence of the assisting organisation. In practice this is simplistic and a more multi-dimensional idea of fitness is necessary, to recognise the possibility of other key actors in an NGO=s environment. In Bangladesh the future of BRAC=s programmes (one of Korten=s five success stories) was subsequently threatened by fundamentalist groups, despite BRAC=s services being in demand by beneficiaries and BRAC having the resources to meet their needs (Holloway, 1994).

There are other features of Korten=s analysis which are consistent with an evolutionary perspective. He argues that AThe spontaneous replication BRAC is observing is probably the strongest available indicator that its program is truly meeting felt needs...@ (Korten, 1980:490). In Chapter Six attention will be given to nature of imitation and replication taking place within the NGO sector in Bangladesh, and the form of learning that it signifies.

Korten differentiates Alearning organisations@ from others using three characteristics: (a) their response to error, (b) the role of peoples= participation in planning, (c) how knowledge is linked to action. According to Korten, learning organisations Aembrace error@, openly Adiscuss their own errors, what they have learned from them and the corrective actions they are attempting@ (Korten, 1980:498). Korten cites BRAC=s discovery in 1979 that Aaccess to consumption credit in times of crisis is more important to most poor families than access to production credit. BRAC is re-examining its credit programme accordingly@ (Korten, 1980:498). While this is important, it is useful to take the idea of learning beyond changes in individuals= understanding, to include changes made in organisational structures. Using Bateson=s idea of information as a difference that makes a difference we can ask what difference did this difference in knowledge make, how did it inform BRAC? Seventeen years later Montgomery (1996: 110) reports that consumption loans were Aintroduced in 1992 - but [were] provided extremely rarely due to staff perceptions of high risk and the prolonged application process which makes this facility inappropriate for any urgent contingencies@. This incident illustrates another fitness requirement not noted by Korten. Services provided have to meet the personal needs of staff and not just fit with the resources and services available. As will be shown in Chapter Seven the personal interests of staff can have a major impact on how NGOs respond to their beneficiaries.

The value of participatory planning is reasonably self-evident, in terms of information requirements in development projects. Project beneficiaries possess specific local knowledge unavailable to managers of large projects. Ultimately, it is beneficiaries= interpretations of project initiatives which will mediate any effects on their lives. Since Korten=s 1980 paper PRA methods have been widely used as a means of accessing local knowledge and judgments about development activities. While the epistemological problems associated with their use have been analysed in detail (Mosse et al. 1994), experiences of NGOs such as ActionAid indicate that even bigger problems lie in the management of that information when it subsequently passes up through organisational hierarchies. There is a massive loss of information about important local differences, even when they seem to be identified by field staff with some degree of accuracy (Davies, 1997a). Improved participatory approaches to learning have to address not only events at the interface with beneficiaries but also the internal dynamics of organisations.

Korten=s third characteristic, concerning the linkage between knowledge and action, emphasises the need for short and quick feedback loops for effective learning. These are found when organisations are first established and founders are in close contact with fieldwork. They are also present in close knit teams where there is not yet a specialised set of roles (Korten, 1980:499). As with Nonaka and others noted above, Korten sees teams as the medium through which new learning takes place, and specialisation of roles is seen as an outgrowth of that learning process. His concern is with what might be called premature specialisation at the early stages of project development. Related problems have been identified in the use of artificial neural networks. Slowing down the process of learning can produce solutions that have wider applicability. The challenge here is how to translate this knowledge into useful guidelines for organisational learning. Limiting the flow of funds into a new project is one method, though Korten recognises that this is not easy.

Looking at the success stories Korten argues that there are three stages in a learning process approach. These are learning to be effective, to be efficient, and to expand. Although this idea has later been used in the literature on NGOs (Edwards and Hulme, 1992:100) there is little evidence to support it as an accurate description of how organisations learn. The extensive literature on learning curves in manufacturing indicates that cost reductions are, not surprisingly, directly associated with improvements in effectiveness of a process, whether the value added is labour or capital (Henderson, 1980). They are not disassociated in time. Secondly, the shape of the learning curves described by Korten bears no relationship to that which has been found in this literature on learning curves. They are rough sketches at best. Thirdly, as the example of consumption loans given above shows, successful cases such as BRAC can expand dramatically without resolving basic issues concerning the effectiveness of their services. The whole notion of linear stages is contradictory to contemporary evolutionary theory and the view of organisational learning developed in this thesis so far.

Korten identifies two potential barriers to learning process approaches. One is Athe bureaucratic imperative to move large amounts of money@ (Korten, 1980:502) when in fact a learning process approach requires only small amounts at the beginning, for pilot projects. The second is the nature of established programming procedures, especially planning and budgeting requirements. These information demands require project management to Ain effect act as if it knew what they were doing before there was an opportunity for learning to occur@. In the case of NGOs both of these constraints may be accentuated or mitigated by the behaviour of their donors. These potential influences will be examined in the analysis of learning within CCDB.

Development projects as policy experiments: Rondinelli

Rondinelli's (1983) book of the same name is a critique of planning and administrative methods as applied to large scale development projects up to the early 1980's. His book provides an exhaustive examination of the reasons why control oriented, top down, long range, detailed planning of development projects does not work. His critique is of interest because of its very strength. If the methods and their attendant assumptions are so flawed then a question which is relevant to this thesis is why are they so prevalent and persistent. Without attending to this problem, theorising about organisational learning in these settings runs the risk of committing the same error that Rondinelli has accused may users of planning based approaches, too much emphasis on the apparent self-contained rationality of methods, and not enough on the context of their actual use.

Although he does not explore this question in detail a number of causes are identified. Rondinelli (1983:29) suggests that A..perhaps the greatest impetus to national planning [in developing countries] was the insistence of international aid agencies that grants and loans be made in conformance with coherent plans for national development@. The methods adopted were based on the models that were available at the time, including project planning techniques used in the construction industry. The promotion of these methods had some degree of fit with the requirements with the prevalent political ideologies of the time. National level planning was seen as appropriate because the public sector was seen as the dominant force for development in capital-scarce countries. Other less explicit needs were also met. Referring to experts Rondinelli argues ATheir tools became their power....the experts dependency on measurement is very real. Measurements and quantitative analysis are the basis of knowledge which differentiates them and therefore, a basis of their social power" (Rondinelli, 1983:6). The survival and proliferation of these methods was associated with a multi-dimensional form of fitness, at the psy