The Conceptual Dimension

Why this exists#

A frame's definition in ordinary FrameNet is prose: a paragraph describing a scene, plus a list of frame elements. That prose is doing a lot of work at once — it states a force-dynamic structure, a role inventory, a coreference constraint, a temporal ordering, a default for what happens when something goes unmentioned, a distinction from a neighboring frame — and none of those facts is separately addressable. Nothing can be checked against anything else, because nothing beyond the frame elements is actually recorded as structure; it's all just paragraph.

The conceptual dimension exists to fix that, from inside the frame apparatus itself rather than by translating the lexicon into some external formalism. It is a layer of highly schematic frames — a small, fixed inventory of abstract structures like containment, part-whole composition, classification, force and transformation — that ordinary lexical frames are read against. A frame doesn't just have a prose gloss anymore; it is grounded in a schema, with its own frame elements explicitly bound to that schema's elements. Once that binding exists, a great deal that used to live only in the gloss becomes structure a validator can actually check.

This is also how FNBr makes Cognitive Linguistics' own working vocabulary — image schemas, construal, profiling, perspective — genuinely computable inside the resource, rather than something that only survives as description in a paper. A schema is a database object. A construal is a specific selection over that object's elements. A claim a cognitive linguist would make about a word can be recorded and checked, not just asserted.

Two primitives, and nothing else#

Everything in this layer is built from exactly two primitive notions.

A region is any concept, modeled as a location in a conceptual space — a temperature, an idea, a social role, a physical object, anything at all. A region's location is its meaning. This is broader than the everyday sense of "region" as a spatial area; here it covers absolutely any concept a frame might be about.

A force is manifested between two regions, and the general shape that manifestation takes — two regions and a tie between them — is itself a schema, with two slots for the regions it relates. A force is never recorded as an object in its own right; it's known only by its effects, the way physics treats force. All of a force's actual content lives in which specific relation names the tie between the two regions.

A force does one of two things to a region:

Something that looks static is force in equilibrium, not the absence of force — and an equilibrium is usually itself what an earlier transformation left behind, so the two effects aren't sealed off from each other.

This layer is topological and relational, never metric. It records that regions are organized, and how — ordered, grouped, composed, moved from one position to another — but never how far apart two regions are. Wherever an actual distance genuinely matters, that belongs to a separate representation this layer can point into, not to the schemas themselves.

Profile over base: the one pattern that recurs everywhere#

To profile something is to bring part of a structure into the foreground — the base — while the rest stays present, just out of view. Backgrounded material is never gone; that's exactly what lets a profile support inference about things it never mentions directly. A janela quebrou ("the window broke") backgrounds whatever broke it, but still entails that something did.

This single pattern shows up at every level of the conceptual dimension:

A profile is always a set, so it can overlap with other profiles and it never forces a single either/or choice the way a plain classification would. It can also be ordered — the highest-ranked element in a profile is the trajector, the figure the rest is read against. A genuinely symmetric relation is simply left unordered.

One frame can be read against more than one schema at once, each with its own profile. Entrar ("to enter") is read as a traversal and as a boundary crossing at the same time — neither schema alone captures both — so it's bound to both, and each binding contributes its own share of what the word means. Conflating several construals into one word is what lexicalization normally does; a framework that only allowed one grounding per frame would be forcing a modeler to throw away everything but a single winning reading.

Beyond words: what "evokes" a schema#

A lexical unit is the most common thing that evokes a frame, but it isn't the only thing that can. FNBr also annotates images, video, and deixis, and those annotations evoke frames too — a pointing gesture, an annotated region of a photograph, a conventionalized visual form. Wherever this document set talks about a lexical unit's profile, the same mechanism is available to any evoking element that can occur across more than one scene: a word can appear in a different sentence, and a gesture in a different video, but one particular region of one particular photograph cannot — so the first two can carry a profile of their own, and a one-off annotation stays a plain annotation instead.

The real object of study is the network of concepts a person invokes to understand a sentence or a scene, and a word is one way into that network — not the only one.

How this relates to the other two dimensions#

FNBr classifies a frame along three independent axes, and it's worth being clear about which question each one is actually answering.

The ontological and conceptual dimensions carve up the same territory from two different, independently-motivated warrants, and in practice their answers line up closely — a frame in the ontological dimension's eventive family is, as a rule, grounded in the conceptual dimension's own family of transformation schemas. That convergence is genuinely useful: two independently-recorded facts that are expected to agree let a real disagreement between them work as a signal worth investigating, in a way that a single fact checked against itself never could. But the two stay separate on purpose, because "what kind of thing is this" and "what structure does it rest on, and which part is in view" are two different questions — the second one is a profile, and a profile has to be read against structure, which is exactly what the conceptual dimension supplies and the ontological dimension does not.

The domain dimension is orthogonal to both: what a frame is about — cooking, law, football — correlates with neither what kind of thing it is nor what structure it rests on.

The rest of this document set works through the conceptual dimension's own machinery in order: the schema inventory itself, how a frame grounds in a schema, how a lexical unit profiles one, profiling in full — incorporation, the internal/external split, and the database it lives in, and the microframes that name the forces a schema's own elements stand in.