Status
Established concepts with heterogeneous disciplinary meanings. BSI conceptual extension under investigation.
Terminological origin
Coherence derives through French cohérence from Latin cohaerentia, related to cohaerēre, “to stick together” or “to hold together”. The etymology already suggests a relational idea: coherence concerns how elements hold together rather than a property possessed by an isolated element.
This etymological observation should not be mistaken for a theoretical definition. Modern uses of coherence and incoherence vary substantially across disciplines.
Concise definition
In its broadest established sense, coherence concerns the way in which multiple elements form an intelligible or appropriately connected whole.
It should not be confused with truth, logical consistency, completeness, absence of ambiguity or agreement.
BSI investigates a more specific problem: semantic coherence when heterogeneous representations must be interpreted, transformed or used together across human and artificial systems.
Conceptual scope
Coherence is fundamentally relational in many established uses, but the nature of the relevant relations differs across disciplines.
In discourse, coherence concerns meaningful relations between textual or conversational units.
In epistemology, coherence has been analysed in terms of relations of support among beliefs, propositions, memories or testimonies. Different coherentist theories have appealed to logical consistency, probabilistic support, inferential connections and systemic integration. There is therefore no single accepted epistemological measure of coherence.
In logic, by contrast, the central formal notions are usually consistency and inconsistency. These must not be treated as synonyms for coherence and incoherence.
In ontology engineering, coherence and incoherence can also acquire specialized technical meanings connected with class satisfiability. These usages should likewise be distinguished from discourse coherence and from the BSI proposal.
Across disciplines
Linguistics and discourse
Discourse coherence concerns the intelligible organization of discourse and the relations connecting its parts. Frameworks such as Rhetorical Structure Theory (RST) analyse relations between discourse segments rather than reducing coherence to the absence of contradiction.
Consequently, a discourse may coherently represent conflicting propositions:
Authority A states P.
Authority B states not-P.
The presence of conflict does not by itself make the discourse incoherent. What matters is how that conflict is represented and related to the surrounding discourse.
Epistemology
Coherentist theories of epistemic justification provide an important precedent for relational conceptions of coherence. A belief may derive justification from the way it fits with or receives support from other beliefs.
Historical and contemporary accounts differ significantly. A. C. Ewing connected coherence partly with consistency and logical consequence; C. I. Lewis explored probabilistic mutual support; Laurence BonJour proposed a multidimensional account involving consistency, inferential connections, integration and the absence of unexplained anomalies.
This literature establishes an important limitation on any BSI originality claim:
The relational character of coherence is not itself a BSI innovation.
It also establishes another crucial distinction. A coherence theory of epistemic justification is not necessarily a coherence theory of truth. Contemporary coherentists may regard coherence as relevant to justification without claiming that coherence is what makes a proposition true.
Logic
In logic, inconsistency concerns formal relations among propositions. Classical inconsistency and the associated principle of explosion must be distinguished from semantic or discursive incoherence.
Paraconsistent logics are specifically designed to permit non-explosive reasoning in the presence of inconsistency. They do not thereby provide a general theory of semantic coherence.
Ontology engineering
Ontology engineering provides another specialized use. An ontology may contain unsatisfiable classes, giving rise in parts of the literature to the description incoherent, while questions concerning whether an ontology as a whole has a model concern consistency.
It should not be generalized automatically to semantic incoherence in discourse or artificial intelligence.
Distinctions
The following distinctions are central to the BSI Lexicon:
Coherence ≠ truth.
A representation may be coherent and false.
Coherence ≠ consistency.
Contradictory propositions may be represented coherently when their relation, source, scope or status is adequately preserved.
Coherence ≠ completeness.
An incomplete representation may remain coherent.
Coherence ≠ unambiguity.
An unresolved ambiguity may be represented coherently when the ambiguity itself is preserved.
Coherence ≠ agreement.
Disagreement can be part of a coherent representation.
Incoherence ≠ factual error.
A false statement is not, for that reason alone, semantically incoherent.
Incoherence ≠ information loss.
Summarization and abstraction may legitimately remove information.
Incoherence ≠ semantic drift.
Semantic change is a process that may or may not result in incoherence.
Incoherence ≠ invalid inference.
An invalid inference and semantic incoherence may occur together, but they characterize different aspects of a failure.
These distinctions are not merely terminological. They delimit the phenomenon that BSI is attempting to investigate.
BSI Perspective
Contemporary artificial intelligence systems increasingly operate across heterogeneous representations.
They retrieve, select, summarize, translate, classify, combine and transform information originating in different conceptual, disciplinary, normative, temporal and epistemic frameworks.
The resulting problem is not exhausted by asking whether individual statements are factually correct or whether a set of propositions is logically consistent.
A system may combine individually acceptable representations while silently altering the relations under which those representations can be interpreted together.
For example, a clinical recommendation and a legal prohibition may both be accurately retrieved. Yet summarizing them as two expert opinions that “disagree” changes the relation between them: clinical recommendation and legal permissibility belong to different normative structures.
Similarly, an effectiveness rate established according to one endpoint cannot automatically be transferred to another endpoint merely because both are described using the word effective.
This motivates the BSI research question:
What happens to semantic coherence when heterogeneous representations must be transformed, combined and used together?
BSI Working Proposal
Provisional definition, not an established disciplinary definition
Semantic coherence is a relational property whereby representations used together preserve the distinctions and relations necessary for their adequate joint interpretation within the relevant framework.
Correspondingly:
Semantic incoherence is a relational failure in which representations remain individually interpretable but are combined, transformed or related in a way that inadequately alters, suppresses or transfers distinctions or relations necessary for their joint interpretation within the relevant framework.
The qualification individually interpretable is important.
The BSI proposal is particularly concerned with situations in which the components do not obviously fail when considered separately. The failure emerges through their relation.
This gives a provisional diagnostic question:
If the individual representations remain acceptable, does the relation constructed between them remain interpretively legitimate?
A relational rather than eliminative approach
The BSI proposal does not assume that greater coherence always means eliminating more differences.
A coherent system may need to preserve:
contradiction;
uncertainty;
ambiguity;
disciplinary differences;
normative conflict;
or disagreement between authorities.
Removing such differences may sometimes make a representation superficially simpler while making it semantically less faithful.
Hence:
Coherence should not be equated with homogenization.
This point is especially important for artificial intelligence systems expected to reason across plural and heterogeneous sources.
Transformation and preservation
Semantic preservation does not imply identity.
A technical statement may be legitimately simplified for a non-specialist audience. A representation may change vocabulary, level of abstraction and structure while preserving what is relevant to the intended interpretation.
Conversely, a very small linguistic change may substantially alter modality, authority, scope or measurement conditions.
The relevant question is therefore not:
Did the representation change?
but:
What distinctions and relations changed, and what interpretations became possible or impossible as a result?
This question connects semantic coherence to another major BSI research concern: provenance.
Coherence and provenance
Provenance concerns more than identifying a source. It can include the history of entities, agents, activities, derivations and transformations through which information has passed.
BSI investigates an additional question:
Can the trajectory of information reveal not only where a representation came from and how it was transformed, but also whether the relations necessary for its interpretation survived those transformations?
This can be summarized by a broader BSI research question:
What happened to meaning along the way?
Research status
The BSI proposal remains under investigation.
Current work is examining, among other questions:
whether the relevant framework can be characterized without merely duplicating established notions such as context, ontology, conceptual framework or interpretive framework;
whether semantic incoherence can be operationally distinguished from semantic loss, semantic drift and invalid inference;
whether different forms of transformation share identifiable interpretive constraints;
and whether the distinction between intra-framework and inter-framework incoherence is sufficiently robust to support formal or computational analysis.
The definition above should therefore be understood as a working research proposal rather than a completed theory of semantic incoherence.
Related concepts
Consistency / Inconsistency
Semantic Engineering
Provenance
Semantic Drift
Epistemic Authority
Normative Authority
Ontology
Context
Interpretive Framework
QualComputing
References
BonJour, Laurence. The Structure of Empirical Knowledge. Harvard University Press, 1985.
Ewing, A. C. Idealism: A Critical Survey. Methuen, 1934.
Lewis, C. I. An Analysis of Knowledge and Valuation. Open Court, 1946.
Mann, William C., and Sandra A. Thompson. “Rhetorical Structure Theory: Toward a Functional Theory of Text Organization.” Text, 1988.
Stanford Encyclopedia of Philosophy. “Coherentist Theories of Epistemic Justification.” Substantive revision 2025.
Stanford Encyclopedia of Philosophy. “The Coherence Theory of Truth.”
Priest, Graham, Koji Tanaka, and Zach Weber. “Paraconsistent Logic.” The Stanford Encyclopedia of Philosophy. Substantive revision 2026.
World Wide Web Consortium (W3C). OWL 2 Web Ontology Language: Direct Semantics. W3C Recommendation, 2009.