Usual disclosures apply
Where do theories come from?
Scientific explanations are often presented as polished, structured products, but they begin as guesses, patterns, analogies, and sometimes simple questions. This post explores my current thinking how theories form: not just individual hypotheses, but the broader frameworks that scientists build, test, and revise. It also considers what a "theory of theory formation" would look like, and why understanding that process matters.
What is a Theory?
A theory is more than a statement of facts. It is also more than a hypothesis. A theory is an organised system that explains and predicts. It organises observations, proposes mechanisms, and produces testable claims. A theory gains credibility not just by fitting existing data but by making new predictions which are then tested and corroborated. A scientific theory must be internally coherent and consistent, general, and open to refutation. Theories are not static; they adapt, evolve, and sometimes get replaced.
A good theory:
Explains more than it assumes.
Coheres with existing knowledge (until it improves on it).
Makes testable predictions.
Can, in principle, be wrong. (Is falsifiable in Popperian terms).
Theories have internal structure. They are not just claims but organised systems. Most have:
Core concepts: key ideas or entities.
Relations: laws, equations, or mechanisms.
Assumptions: simplifications or constraints.
Scope: where and when they apply.
Understanding how these parts develop is part of understanding theory formation. When core concepts shift, the entire theory often shifts with them.
The Evolution of Theory
Successful theories don’t just drop out from the data, and they aren’t purely logical deductions either. They result from feedbacks between observation, intuition, analogy, and reasoning. Sometimes an anomaly sparks a new idea. Other times, an old theory no longer fits and needs replacement.
Theory formation typically follows a loose pattern:
Observation: A pattern, puzzle, or inconsistency emerges.
Hypothesis generation: A possible explanation is sketched.
Testing and adjustment: Evidence is gathered and compared.
Structuring: Successful explanations are systematised into broader models.
This is not linear or guaranteed. It is iterative and contingent. Failed theories teach us just as do successful ones.
Scientific theories are memes and, like memes, evolve. They do not merely accumulate; they reorganise. New theories sometimes extend old ones (e.g., Newton built on and transformed Galileo), but sometimes they replace them entirely (as Galileo et al. supplanted Aristotelian physics). Change may come through:
New evidence.
New methods or instruments.
New perspectives or metaphors.
Internal contradiction.
Progress happens not only when data contradicts existing theory, but when someone sees a better way to think.
Why This Matters
Understanding how we form theories helps us improve science. It clarifies:
Why some ideas get traction and others don’t.
How biases shape what we think is plausible.
Why revolutions in thought often come from outside a field.
How methods and tools limit what questions we even ask.
There is no unified "theory of theory formation" today, but we have parts of it scattered across philosophy, psychology, sociology, and history. A useful synthesis would not try to define one fixed path, but instead offer a toolbox: a set of strategies, processes, and constraints that shape how good theories emerge, and bad ones persist.
From Theory Formation to a General and Unified Theory of Accessibility
Above, we asked: What is a theory of theory formation? We had some notes. Now we take that logic and apply it to a specific domain: accessibility in transport and cities, which I have been involved in for a while.
Accessibility has long been measured, mapped, and modelled, but not always theorised. Some have even identified a “lack of a compelling, robust theoretical foundation for measuring accessibility”, which “results in many conceptual and practical problems in using accessibility in transportation planning and policy analysis.”
What counts as access? How do we explain changes to it? And how should we structure our thinking about it?
Rather than relying on ad hoc metrics or discipline-specific assumptions, we need to ask how a general theory of access is formed and how access can be unified across methods, use cases, and timeframes.
What Is a Theory of Access?
The only reason to locate anywhere is to be near some people, places, and things, be far from others, and possess still others.
A theory of access is not an isochrone. It is not a travel-time matrix. It is a structured explanation of how different people reach the many different things they want and need, and how cities and systems shape that.
In Towards a General Theory of Access, like much of the literature, we define access as the ease of reaching valued opportunities. That requires both:
Opportunities (the things people need),
Transport (the means of getting there),
As well as land use, time constraints, and individual capabilities.
A theory of access explains how these elements interact across space and time. We can write the measure compactly, where the bold notation indicates that these are not single values, but full matrices considering many dimensions of opportunities (O) and general costs (C).
Forming a Theory: From Pieces to System
Much like in general theory formation, access theory starts with:
Observation: Places have unequal, changing, and sometimes surprising patterns of accessibility.
Hypothesis Generation: For instance, “Places with better access to jobs and lower access to competing workers will have shorter commute durations”, or “Networks and land use co-evolve, explained by access,” or “Access to residences and to employment explains transit flows.”1
Model Construction: Measures of access emerge: cumulative, gravity-based, utility-based, time-based.
Evaluation and Revision: Do these measures capture what matters? Are they transferable? Do they generalise?
But this becomes a problem when each model becomes its own isolated micro-theory. That’s the issue we took on in Unifying Access.
Fission and Fusion
Accessibility metrics have proliferated, but often without a consistent conceptual backbone. As Unifying Access shows, most access measures reduce to two components:
Opportunities available (supply-side),
Impedance to reaching them (cost-side).
Yet the relevant research communities developed separate literatures for place-based and person-based access, for potential and revealed access, and so on: often without reconciling them.
This is the equivalent of physics having separate equations for rocks and feathers falling.2 What we propose is a framework that reveals the underlying structure of all access measures. That is, they unify rather than redefine.3
The literature has coalesced around four key elements that must be included in any theory of access:
People – with goals, constraints, and means.
Land Use or Opportunities – which determines the spatial distribution of activities.
Transport – the system that mediates interaction.
Time – as all access is temporally bounded.
These elements form the conceptual primitives. A general theory then explains how they relate: how people make decisions, how systems evolve, how policies intervene, and how inequalities emerge.
This mirrors theory formation more broadly: identify the core components, define relationships, generate predictions, and remain open to revision.
By treating access as a proper theory, and not just a metric, we gain clarity:
Unification: We can compare and translate across modes, scales, and contexts.
Generality: We can apply the same conceptual structure in Minneapolis or Manila, for work access or hospital access.
Revisability: We can adapt to new evidence, technologies, or needs, just as good theories must evolve.
Rather than reinvent the wheel for each policy use case, we build on a shared foundation.
TLDR
A theory of theory formation gives us tools to assess and build better theories in any field. When applied to accessibility, it shows us how fragmented thinking can be replaced with structural clarity and conceptual economy.
We want to move the field forward not by multiplying metrics, but by showing the shared logic beneath them. A theory of access, like any good theory, simplifies in order to explain, connects what looks disparate, and stays open to challenge.
That is what it means to theorise access.
FIN
Notes
The Role of Evidence and Hypotheses
Access theories aren’t just abstract. They are used to justify infrastructure, frame equity debates, and allocate funding. But evidence in access studies is often over-indexed on measurability rather than explanatory value.
Hypotheses like “more transit stops increase access” may hold only under certain land use or institutional conditions. Too many transit stops may in fact decrease overall access to destinations, because the people on-board are delayed more than the people and boarding reduce their access/egress time to the stop.
Models that show high transit access is associated with high transit mode share now doesn’t necessarily imply causality, or that an increase in transit access will increase transit mode share.
Just as in general theory formation, theory must guide what we measure, not the other way around. We need evidence to test structure, not just support assumptions.
Which students of History of Science will note, it did from the days of Aristotle through to time of Galileo. Heavier things were thought to fall faster. (Obviously not everyone was so feather-headed).
On Lumpers and Splitters
One can know more and more about less and less until one knows everything about nothing. Or
One can know less and less about more and more until one knows nothing about everything.
The first are professors. The second are pundits.
Lumpers and splitters exist everywhere. Some people categorise the world into broad, unifying principles — lumpers. Others insist on distinctions at every turn — splitters. Both approaches have their uses. The lumpers create grand theories, synthesise knowledge, and find common patterns. The splitters specialise, identify the crucial nuances, and ensure we don’t miss the details that matter.
I visit a new city. I can look for similarities or differences. It’s more fun to look for differences, but the similarities are huge. The road networks, land use patterns, transport systems, governance structures—variations on a theme. Yet, the way intersections are managed, the type of transit vehicles, the built form of neighbourhoods, these differences make a city feel distinct. Splitters might catalog each unique feature and develop a taxonomy of urban form. Lumpers might argue that all cities are basically the same and follow predictable laws.
Academia favours splitters. Specialisation is the path to tenure. We carve out niches, define subfields, and create ever more precise classifications. The system rewards deep knowledge within narrow boundaries. But this comes at a cost. The synthetic — the effort to bring ideas together across disciplines — often takes a back seat. Hence we get a plethora of access measures.
In science and engineering, we train reductionists. Break a system into its components, understand each piece, and reassemble. This approach works well for machines and physics, but less so for complex, emergent systems like cities, societies, and economies. The incentives to synthesise are weak. Journals tend to publish novel findings, not grand unifications. Funding flows to discrete problems, not cross-cutting integration. Yet, many of our biggest challenges: climate change and urbanisation require synthesis.
Perhaps it’s time to balance the splitters with more lumpers within science and engineering. To reward not just those who discover something new, but also those who connect the dots. To train the next generation to see not only the trees, but also the forest.
After all, the real world doesn’t care about our categories.
Keep reading with a 7-day free trial
Subscribe to Transportist to keep reading this post and get 7 days of free access to the full post archives.