Incrementalism¶

Version: v2 · 2026-09-01 · History
Prime #: 1488
Origin domain: Public Administration & Policy
Also from: Political Science, Organizational & Management Science
Aliases: Incremental approach, Incrementalist strategy, Successive limited comparison
Related primes: gradualism, Iteration, adaptive planning, continuous improvement, muddling through

Core Idea¶

Incrementalism is a strategy for changing a system by moving from its present state through a succession of bounded commitments rather than attempting one comprehensive, fully specified jump. Each step alters only a limited part of the state or moves it only a limited distance. The actor evaluates the consequences that become visible after the step, retains what remains acceptable, and chooses the next adjustment from the new position. The recognition invariant is therefore not smallness alone. It is the coupled architecture of a current-state anchor, a bounded change, a local comparison, a committed intermediate state, and a next choice informed by what the preceding change exposed.

In public-policy analysis, Charles E. Lindblom described decision making by successive limited comparisons as an alternative to a synoptic ideal that would enumerate all values, alternatives, and consequences before choosing. His account emphasizes restricted margins, intertwined means and ends, repeated comparison with existing policy, and serial correction rather than exhaustive optimization.^[1] Later work with David Braybrooke developed disjointed incrementalism as a strategy adapted to limited analysis, plural values, distributed attention, and politically fragmented decision settings.^[2] These sources supply one historically important formulation, but the Prime is not confined to government. Its identity survives wherever bounded changes create information and preserve an actionable intermediate state.

The step must be consequential. Merely thinking about a small option, subdividing a plan on paper, or describing a smooth trend is not yet incrementalism. A genuine increment changes the operative state, resource allocation, design, rule, or practice enough that later choice begins from a different baseline. This commitment can be reversible, but reversibility is not required. Some increments create path dependence, sunk cost, or institutional precedent. Incrementalism manages that risk by limiting exposure per step and keeping feedback close enough to influence what follows; it does not promise that every step can be costlessly undone.

The strategy is also selective about comparison. Instead of reopening the entire option space after every observation, an incremental process usually compares a small neighborhood of alternatives near the current state. That restriction saves analytical and coordination cost, makes disagreement more tractable, and allows action despite incomplete models. It also produces the canonical failure boundary: important remote alternatives, regime changes, interacting constraints, or accumulating harms can remain invisible because the search never leaves the local neighborhood. Incrementalism is therefore a governing search-and-commitment strategy, not a claim that local steps are always optimal.

Logical incrementalism in organizational strategy adds intentional direction without requiring a fixed master plan. James Brian Quinn described executives building strategy through staged commitments, experiments, coalition formation, timing, and learning, so that broad purpose can coexist with evolving means.^[3] This distinguishes incrementalism from unstructured drift. A process can have a durable orientation, explicit constraints, and milestone criteria while allowing the path and even intermediate goals to change as evidence and stakeholder alignment develop.

The Prime contains both reactive and designed forms. Reactive incrementalism responds to the next salient problem from the inherited status quo. Designed incrementalism deliberately creates small batches, pilots, provisional rules, staged investments, or limited-scope releases to buy information before larger commitment. The two share the same bounded-delta architecture, but differ in whether the sequence, measurement, and escalation rules are planned. That distinction matters when evaluating competence: a process is not well designed merely because its changes happen to be small.

Structural Signature¶

Operative present state. A current policy, design, allocation, routine, agreement, or configuration supplies the baseline from which change is measured.
Directional concern. A problem, opportunity, aspiration, or broad orientation motivates movement without necessarily specifying a complete end state.
Bounded delta. The next commitment changes a limited scope, magnitude, duration, population, resource share, or component set relative to the baseline.
Local option neighborhood. Attention is concentrated on a manageable set of alternatives adjacent to the present state rather than the full theoretical possibility space.
Intermediate commitment. The step is enacted or otherwise made operative, producing a new state from which later action proceeds.
Consequence channel. Observation, performance, stakeholder response, conflict, cost, error, or environmental reaction exposes information unavailable or too costly before commitment.
Retention and correction rule. The actor preserves acceptable parts, repairs or reverses unacceptable parts where possible, and determines whether to continue, redirect, pause, or escalate.
Serial dependency. The feasible and attractive next steps depend on the sequence already taken; history is part of the state rather than disposable background.
Escalation boundary. Conditions identify when local adjustment is no longer adequate and a wider redesign, discontinuous intervention, or comprehensive analysis is required.
Recognition test. Ask whether an enacted bounded change creates an informative intermediate state and whether that information materially governs the next commitment. If either link is absent, the case may be gradual change, arbitrary drift, decomposition, or a fixed phased plan rather than incrementalism.

What It Is Not¶

Not gradualism alone. Gradualism describes a slow tempo or preference for gradual change. Incrementalism additionally specifies how successive bounded choices are anchored, evaluated, and used to select what comes next.
Not Iteration as generic repetition. Iteration repeats an operation or loop, often under a fixed rule. Incrementalism can use different operations at each step and is defined by locally bounded commitment from the inherited state.
Not Progressive Refinement from Core Model. That Prime begins from an explicit tractable model and adds corrections or detail while monitoring the model's validity. Incrementalism can operate on institutions, budgets, designs, or agreements without a formal core model or expansion parameter.
Not Continuity versus Rupture. That Prime classifies the character of a transition. Incrementalism is an action strategy that produces and learns through a sequence; a series of increments can eventually enable a rupture in aggregate outcome.
Not a fixed stage-gate plan. A large program divided into predetermined phases is phased implementation if evidence from one phase cannot change the content, direction, or continuation of later phases.
Not procrastination. Deferring a difficult choice without making an informative bounded commitment generates delay, not an increment.
Not random drift. Uncoordinated small changes may accumulate, but they are not incrementalism unless a decision process relates current state, local change, consequences, and subsequent choice.
Not automatically conservative. The method often privileges the status quo, yet a directed sequence of small commitments can produce large cumulative transformation or prepare a deliberate threshold crossing.
Not automatically reversible. Pilot scope and modular design can improve reversibility, but statutes, infrastructure, hiring, ecological intervention, and public commitments may create durable effects at each step.
Not a universal optimum. Local comparison economizes on cognition and coordination while risking local lock-in, hidden coupling, slow response, and failure to address emergencies or indivisible changes.

Broad Use¶

Public policy and administration. A legislature, agency, or local government can adjust eligibility, funding, enforcement, or service scope relative to an existing arrangement, observe administrative and political consequences, and revise the next measure. Lindblom's canonical analysis arose here, where values conflict, information is incomplete, and many actors possess veto or implementation power.^[1] Literal recurrence requires serial policy commitments and consequence-sensitive revision; the mere fact that annual statutes differ slightly does not establish the strategy.

Organizational strategy. Leaders can build strategic direction through experiments, selective resource commitments, coalition development, and timed issue resolution rather than publish one fully determined plan. Logical incrementalism preserves a broad orientation while allowing means and timing to emerge through learning and negotiation.^[3] The structure is not an excuse for secrecy or inconsistency: roles, constraints, and review points still need to be inspectable.

Software and digital services. A team can release a bounded feature or migration slice, observe defects and use patterns, preserve the working remainder, and alter the next slice. This qualifies when production evidence changes later design. A backlog executed unchanged in small tickets is decomposition or small-batch delivery, not necessarily incrementalism, because the feedback-to-next-choice link may be absent.

Engineering and product design. Designers can modify one subsystem or controlled parameter, test the altered artifact, and use measured behavior to select the next change. The method is especially valuable when models omit manufacturing, human-use, or environmental interactions. Safety and coupled-system constraints can forbid naïve one-variable changes; boundedness must be established against system consequences, not merely component count.

Budgeting and resource allocation. An organization can begin from the prior allocation and negotiate limited additions or subtractions. This reduces decision cost and stabilizes expectations, while also importing inherited inequities and obsolete line items. Incremental budgeting is a literal instance only when the previous allocation is the operative anchor and local changes form the primary choice space.

Negotiation and coalition building. Parties can make limited agreements, implement confidence-building measures, and use compliance or conflict evidence to determine whether broader agreement is feasible. Each agreement changes expectations and outside options, so sequence matters. Token concessions with no learning or change in the feasible next set are not enough.

Scientific and operational practice. A laboratory, field program, or service operation can introduce a bounded procedural revision, measure its effect, and decide whether to retain or modify it. Ethical review, preregistration, safety controls, and evidential standards remain domain requirements; the Prime describes the change architecture and does not license uncontrolled experimentation.

Personal and everyday problem solving. A person can alter one routine, observe whether the burden and benefit are sustainable, and then keep, revise, or abandon the change. The case remains structural only when the steps are deliberate enough to create usable evidence. Habitual small adjustments without a consequence-sensitive decision link are ordinary adaptation rather than incrementalism.

Institutional reform. A complex institution can test limited jurisdiction, duration, or participant scope before wider adoption. Sunset clauses, pilots, staged authority, and modular implementation can make intermediate states legible. Yet temporary status alone does not qualify: the pilot must have a decision-relevant evaluation path, and affected parties must not be treated as costless test material.

Across these substrates, the nouns vary but the verbs travel: anchor, bound, commit, observe, retain, compare, redirect, and escalate. The Prime is warranted because these verbs retain their literal decision role rather than serving as metaphorical similarities among otherwise unrelated practices.

Clarity¶

A clear account states the baseline, the dimension along which the step is bounded, the authority that commits it, and the evidence that can alter the next step. Calling a change small without naming the relevant exposure is inadequate. A one-line code edit can affect every user; a modest statutory amendment can restructure incentives; a geographically narrow ecological intervention can have irreversible spillovers. Boundedness concerns credible consequence and commitment scope, not word count or visible effort.

The end-state question must also be explicit. Some incremental processes pursue a known target through adaptive routing. Others hold only a direction or problem frame. Still others explore whether a goal should be retained. These forms can all qualify, but they support different judgments. Progress toward a fixed target can be measured by remaining distance; open-ended incrementalism needs constraints, review criteria, and stopping or escalation rules to avoid converting responsiveness into aimlessness.

Feedback means decision-relevant consequence, not any observation after a step. Metrics can be delayed, confounded, strategically manipulated, or too narrow to reveal distributional harm. Stakeholder acceptance can reflect power rather than merit. The process must state what evidence is considered, whose experience enters, and how contradictory signals are resolved. Incrementalism economizes on prediction by learning from action, but it does not solve the epistemology of that learning.

Sequence dependence deserves separate notation. Let the operative state after step t be S_t and the committed change be delta_t. The transition S_(t+1) = F(S_t, delta_t, environment_t) can alter both the system and the menu from which delta_(t+1) is chosen. This is not a claim that all cases have a numerical state equation. It makes the historical dependency visible: returning to an earlier policy or design may be impossible because expectations, capabilities, and external conditions changed during the sequence.

Finally, the label should not be used evaluatively by itself. Incremental can describe a prudent learning strategy, a politically feasible compromise, a locally trapped search, an erosion of safeguards, or an inadequate response to urgent harm. Evaluation requires the objective, affected population, feasible alternatives, time horizon, reversibility, cumulative exposure, and escalation performance.

Manages Complexity¶

Incrementalism compresses a high-dimensional change problem into a local decision neighborhood. Instead of forecasting every interaction across a complete transformation, the actor chooses a bounded move and lets the resulting state reveal some of the hidden response surface. This converts part of a prediction problem into a sequence of observation-and-choice problems. The reduction can be decisive when the world model is incomplete, stakeholder preferences are distributed, or implementation capacity develops only through use.

The method also limits simultaneous coordination. A comprehensive plan may require agreement about distant consequences and value trade-offs before anyone can act. A local step can permit agreement on the next move while leaving later disputes open. That is a structural economy, not merely political timidity. It lowers the amount of common knowledge and commitment needed at any one time, though it may increase the total number of decisions and the chance that provisional arrangements become entrenched.

Information is purchased with exposure. An enacted step reveals operational facts that simulation or debate might miss, but affected people and systems bear the consequences. Small scope, reversibility, monitoring, and stop rules can reduce exposure; they cannot erase it. In high-consequence domains, a bounded trial still requires substantive safety, consent, legal, and ethical safeguards. Incrementalism supplies no exemption from them.

The method deliberately discards global optimality guarantees. A sequence can settle into a local optimum, preserve inherited categories, or make a remote alternative progressively harder to reach. Braybrooke and Lindblom's account treats limited analysis as a response to constrained capacity, not a theorem that local comparisons find the best possible policy.^[2] A mature incremental process therefore periodically widens the search, challenges the baseline, and asks whether the cumulative path still serves the problem that initiated it.

Complexity is managed well when the cost of each new observation is proportionate, the state remains diagnosable, and the escalation boundary is credible. It is managed badly when many small changes interact faster than anyone can evaluate them, metrics lag behind commitment, or no actor retains authority to halt accumulation. Then the label incremental can conceal aggregate transformation beyond the process's monitoring capacity.

Abstract Reasoning¶

Type the present state: identify what is already operative, inherited, funded, enforced, deployed, or socially expected.
Name the directional concern without pretending that it supplies a complete terminal design.
Define the next delta on consequence-bearing dimensions such as scope, duration, population, capital, reversibility, or coupling.
Enumerate the nearby alternatives actually considered and disclose important remote alternatives excluded by the local search frame.
Predict immediate consequences sufficiently to satisfy safety and feasibility constraints before treating action as a source of evidence.
Commit the bounded step through the relevant authority and record which parts are provisional, durable, or difficult to reverse.
Observe outcomes through multiple channels and separate system response from measurement artifacts, implementation variance, and strategic behavior.
Compare the observed state with the prior baseline and with the explicit continuation, pause, reversal, and escalation criteria.
Retain useful components while preventing sunk cost or precedent from silently becoming evidence that continuation is justified.
Choose the next step from the new state, recognizing that the sequence has altered feasibility, expectations, and the comparison set.
Periodically conduct a nonlocal review so alternatives outside the immediate neighborhood are not permanently excluded by habit.
Report conclusions conditionally: this sequence improved or worsened these outcomes under these observations and constraints, not incrementalism is universally superior.

Knowledge Transfer¶

Transfer begins with roles, not with the adjective small. In policy, the baseline may be a statute and the delta an amendment. In software, the baseline is a deployed service and the delta a release slice. In engineering, the baseline is a tested design and the delta a controlled modification. In budgeting, the baseline is the inherited allocation and the delta a negotiated adjustment. Each qualifies only when the new state and its evidence govern what comes next.

Vocabulary travels with useful precision. Baseline means the currently operative state, not necessarily a neutral measurement reference. Delta means a bounded committed difference, not necessarily a numerical derivative. Feedback means observed consequences routed back into choice. Retention means preserved parts of the new state. Escalation means widening the scope or changing the mode of intervention when local correction fails. These translations preserve the mechanism across substrates.

The transfer fails when the example supplies only gradual time. Geological erosion, biological growth, and demographic change may proceed through small increments without an actor selecting each change from feedback. They instantiate accumulation, adaptation, or continuous dynamics, not necessarily incrementalism. Likewise, a manufacturing line executing a fully fixed sequence of small operations lacks adaptive next-step choice even though its units are incremental.

Domain safeguards do not travel automatically. A reversible feature flag is not equivalent to a reversible law; a software test population is not equivalent to human research subjects; an engineering parameter sweep is not equivalent to experimenting on public infrastructure. Transfer recognizes the decision architecture while revalidating authority, evidence, reversibility, liability, ethics, and time horizon locally.

A particularly useful transfer is the distinction between planned sequence and adaptive sequence. Teams in many domains confuse decomposition with learning. If later steps are fixed and evidence cannot change them, breaking work into pieces may reduce batch size but does not instantiate the full Prime. This diagnostic helps preserve the concept's identity instead of making incrementalism a synonym for anything done in more than one step.

Examples¶

Policy adjustment. A city changes one transit route's frequency for a defined period, measures crowding, reliability, access, and displaced demand, then uses the results to revise the next set of routes. Baseline, bounded scope, operative commitment, multiple consequence channels, and evidence-sensitive continuation are visible. If the city had already committed to the full network plan and used the trial only for publicity, the sequence would be phased implementation rather than incrementalism.^[4]
Logical strategy formation. An organization enters a new service area through a limited partnership, learns which capabilities and stakeholder relationships are decisive, and commits additional resources only after that evidence changes its strategic map. Broad direction persists, but the pathway and coalition emerge through staged commitments. This maps to Quinn's logical incrementalism rather than to accidental drift.^[3]
Software migration. A team redirects a small, monitored traffic fraction to a new subsystem, compares correctness and latency, fixes a discovered state-consistency problem, and changes the next migration slice. The traffic boundary limits exposure and the observation changes the sequence. A predetermined ten-part rewrite with no decision gate is not the same structure.
Budgeting failure. An agency rolls last year's allocations forward and changes each line slightly, even though its mission and cost structure shifted. This is incremental budgeting, but the example reveals the method's status-quo anchoring failure: local comparison makes zero-based alternatives invisible. Recognition does not imply endorsement.
Safety boundary. A chemical plant cannot justify an unvalidated control change by calling it a small experiment. If coupled hazards make even narrow scope capable of catastrophic propagation, the delta is not consequence-bounded. Incrementalism begins only after the domain's engineering and regulatory controls establish a safe test envelope.
Cumulative transformation. A sequence of modest zoning amendments, infrastructure investments, and permit rules changes development incentives over many years. Each step is locally intelligible, while their cumulative effect can reorganize the city. The case shows why incremental action and large aggregate transformation are compatible and why periodic whole-path review is necessary.

Structural Tensions¶

Learning versus exposure. Acting reveals information that prediction cannot, but the step imposes real consequences before uncertainty is resolved. Diagnostic: Is the information value proportionate to the bounded and ethically governed exposure?
Local tractability versus global blindness. Restricting the comparison set makes choice feasible while excluding remote alternatives and system interactions. Diagnostic: When was the last credible nonlocal review of the option space?
Stability versus inherited injustice. Starting from the current state preserves coordination and expectations while treating historical allocations as the default. Diagnostic: Which features persist only because the baseline, rather than a defensible criterion, grants them priority?
Reversibility versus path dependence. Small steps can be easier to reverse, yet commitment changes capabilities, expectations, legal status, and future options. Diagnostic: What would restoration actually require after stakeholders adapt to the new state?
Direction versus opportunism. Flexibility allows learning, but every locally convenient step can dissolve strategic coherence. Diagnostic: Which durable purpose or constraint makes the current delta part of one intelligible trajectory?
Responsiveness versus metric capture. Frequent evaluation can correct quickly, but narrow or short-lag measures can steer the sequence away from hard-to-observe values. Diagnostic: Which affected outcomes arrive too late or resist the chosen measurement channel?
Bounded step versus cumulative load. Each increment can pass a local threshold while the sequence crosses a system threshold. Diagnostic: Is aggregate exposure tracked across steps rather than reset after each local approval?
Political feasibility versus problem adequacy. Limited agreement can unlock action, but an indivisible or urgent problem may require coordinated discontinuity. Diagnostic: Does local adjustment reduce the governing harm fast enough, or merely postpone a necessary mode change?

Structural–Framed Character¶

Incrementalism grades mixed-framed on the structural–framed spectrum (aggregate 0.6) — the most framed of its grading cohort, and the three graders were unanimous on every criterion. The architecture itself is describable neutrally: a baseline anchor, a bounded delta, local comparison of nearby alternatives, serial correction over rounds. Nothing in that loop praises or condemns; evaluative weight reads zero.

Everything else pulls toward the framed pole. The prime's origin is squarely institutional: Lindblom's "successive limited comparisons" arose inside public administration and policy analysis, a discipline whose subject matter is human institutions (a full point). Its definition is constitutively agential — an actor who evaluates acceptability, makes commitments, and chooses from an inherited state, over policies, allocations, and agreements; strip the deciding agent away and what remains is merely small-step change, not incrementalism (a full point on practice-boundedness). The policy lexicon partially travels with it — "muddling through," "disjointed incrementalism" — and applying the prime to, say, a biological or algorithmic process half-recasts that process in decision-strategy terms rather than simply labeling structure already there (half points on both). The verdict: a genuinely framed-leaning prime whose core loop is portable but whose identity is anchored in the practice of deciding.

Substrate Independence¶

Domain breadth — 0.995. Public statutes, organizational resource commitments, software deployments, engineered designs, budgets, and negotiated agreements use different materials and authorities, yet all can begin from an operative state, enact a bounded delta, observe consequences, and choose the next delta. The recurrence crosses institutional, technical, economic, and personal substrates.

Structural abstraction — 0.994. The identity survives replacement of every domain noun. It does not depend on code, money, law, physical geometry, or individual psychology. It does depend on serial state, bounded commitment, feedback, and next-step selection; removing any of these turns the case into a neighboring abstraction.

Vocabulary travel — 0.992. Baseline, increment, local comparison, commitment, feedback, retention, correction, path dependence, and escalation remain operative across domains. Their units and evidential standards change, but their relational role does not.

Diagnostic travel — 0.992. In any substrate, ask what state is inherited, how the delta is bounded, what changed operatively, what consequences were observed, how they affected the next choice, and when local change must yield to a wider intervention. The same test distinguishes incrementalism from phased execution and gradual drift.

Intervention travel — 0.991. Designers can reduce step scope, improve reversibility, shorten feedback delay, diversify measurement, record cumulative exposure, widen option review, and install escalation criteria in policy, engineering, organizational, and software processes. These interventions alter the same mechanism rather than merely sharing a theme.

Substrate-removal test. Remove government and the architecture remains in product development. Remove technology and it remains in budgeting and negotiation. Remove formal organizations and it remains in deliberate personal problem solving. Remove any particular temporal scale and the structure persists as long as consequences arrive soon enough to govern the next commitment.

Residual after catalog subtraction. Iteration supplies repetition, Transformation supplies state change, Feedback supplies return of consequences, and Continuity versus Rupture supplies a tempo frame. None requires the conjunction of a status-quo anchor, bounded local option set, operative intermediate commitment, consequence-sensitive next choice, and serially altered feasibility. Progressive Refinement from Core Model additionally requires an explicit core model and successive representational correction. Incrementalism retains a stable autonomous residual after each subtraction.

The combined score is 0.993. The remaining framing dependence concerns legitimate bounds, values, and authority, not substrate dependence of the mechanism. The Prime classification is therefore supported above the operational 0.99 threshold.

Relationships to Other Abstractions¶

Current abstraction Incrementalism Prime

Parents (1) — more general patterns this builds on

Incrementalism is a kind of Transformation Prime

The accepted reference-grade review places Incrementalism under Transformation because the child instantiates or depends on the parent's broader structure while retaining its own constitutive identity.

Children (2) — more specific cases that build on this

Animal protectionism Domain-specific is a kind of Incrementalism

The proposed strict upward parent is prime:incrementalism.
Continuous design Domain-specific is a kind of Incrementalism

The proposed strict upward parent is prime:incrementalism.

Hierarchy path (1) — routes to 1 parentless root

Incrementalism → Transformation → Function (Mapping)

Neighborhood in Abstraction Space¶

Incrementalism sits in a sparse region of abstraction space (66^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.

Family — Event Phases & Staged Recovery (10 primes)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-10

Not to Be Confused With¶

Gradualism. A preference or pattern of slow change; it need not use feedback from committed intermediate states to choose subsequent changes.
Iteration. Repetition of a step or operation; it can be fixed, computational, and independent of locally bounded policy or design commitment.
Progressive Refinement from Core Model. Adds detail or correction to an explicit baseline model; incrementalism can change the world directly without a formal core representation.
Continuous improvement. A management family aimed at ongoing process improvement; it often instantiates incrementalism but adds organizational routines, quality methods, and performance goals.
Agile development. A family of software and product practices that can implement incrementalism, but ceremonies or short sprints alone do not establish consequence-sensitive next-step choice.
Pilot or experiment. One bounded trial can supply an increment, but an isolated trial without serial retention and next-choice logic is not the whole Prime.
Path dependence. Explains how sequence constrains outcomes; incrementalism is a strategy that creates and manages such a sequence rather than the dependency phenomenon itself.
Conservatism. Incrementalism can privilege the status quo, but it is a change architecture rather than a political ideology and can cumulate into radical outcomes.

The prospective workspace queue contains one strict upward edge to prime:transformation. No live DAG mutation is authorized.

Solution Archetypes¶

No catalogued solution archetypes reference this prime yet.

References¶

[1] Charles E. Lindblom, 'The Science of Muddling Through,' Public Administration Review 19(2), 79–88 (1959), DOI 10.2307/973677. registry ↩a ↩b

[2] David Braybrooke and Charles E. Lindblom, A Strategy of Decision: Policy Evaluation as a Social Process, Free Press, 1963. registry ↩a ↩b

[3] James Brian Quinn, Strategies for Change: Logical Incrementalism, Richard D. Irwin, 1980, ISBN 978-0-256-02682-5. registry ↩a ↩b ↩c

[4] Charles E. Lindblom, The Intelligence of Democracy: Decision Making Through Mutual Adjustment, Free Press, 1965. registry ↩