primitive image-schemas

[Rob Freeman]

George and the list(s),

(Which list are we on? Anyone not getting Funknet and Cogling is only getting
half this thread.)

I apologize to Monica for singling her out in my original post. I had no idea
it was central to the definition of image-schemas that they be primitives. It
seems she was merely stating the accepted position in Cognitive Linguistics.

If such primitives are posited I agree with Monica that there is a conflict to
be resolved. If language is subjective, and language reflects cognition, why
should there be universal cognitive primitives (you can make it so by adding
complexity to the theory, but why _should_ it be so?)

It starts to sound like a confusion between "primitive" and "average". Are the
primitives we see really primitives, or are they the average of a lot of
subjective conceptions of the same thing?

I suppose it is inevitable that on some level the topology of the brain will
kick in and you will have a true primitive ("Cogs"!! I like it :-) I would be
surprised if the topology of the brain restricted us to a concept of
"containment" or such, however.

-Rob Freeman

On Wednesday 09 June 2004 00:19, you wrote:
> Hi,
>
> Sorry to be out of touch for so long, but I've
> been working at applying cogling to politics and
> helping to establish the Rockridge Institute
> (www.rockridgeinstitute.org).
>
> As the person who (so far as I know) introduced
> the term "image-schema" into cognitive
> linguistics, I hope I can help clarify the
> discussion so far.
>
> As is natural in science, ideas go through
> various stages as better and better
> understandings of the subject matter are
> achieved. Here are some of the stages in the use
> of "image-schema."
>
> Stage 1: I first used the term "image-schema" to
> refer to the primitive "images" that both Talmy
> and Langacker were discussing in the mid-to-late
> 1970's,  and which  remarkable researchers like
> Casad, Lindner and Brugman expanded on.
>
> Talmy's original discussion (in a talk at
> Berkeley, summer 1975) made the observation that,
> although spatial relations terms seem to have
> quite different meanings in the languages of the
> world,  those meanings can be decomposed into
> primitive "images" that recur across languages.
> Talmy surmised that there was a set of such
> primitives that was universal. His examples
> included containment, source-path-goal, contact,
> encirclement, etc. His idea was that the
> closed-class spatial relations terms (like
> prepositions in English, cases in Finnish,
> postpositions in Japanese) were complex concepts
> made up of such primitives.
>
> Talmy further analyzed the primitives into
> conceptual types: topological, orientational, and
> force-dynamic. I used the term "image-schema" for
> each such primitive in Women, Fire, and Dangerous
> Things, case study 2.
>
> Langacker, starting around the same time, had a
> similar idea about the existence of such spatial
> primitives. He, together with Casad, Lindner, and
> Brugman, have provided extremely rich
> image-schema analyses, breaking down the meanings
> of spatial terms into primitives. These studies
> are among the most sophisticated within cognitive
> linguistics. They are, however, not
> uncontroversial, as alternative decompositions
> into primitives have been proposed. The
> controversy has been about which decompositions
> are cognitively correct, not about whether
> image-schema analysis is right.
>
> Stage 2: Mark Johnson (1987) enriched that idea
> of image-schemas with his discussion of
> phenomenological embodiment --  intuitive
> embodiment subject to introspection. Johnson took
> the idea out of linguistics per se and set it in
> the context of phenomenological introspection: We
> understand our bodies as containers, with an
> inside, a boundary, and an outside. We understand
> every movement we make as having a source, a
> path, and a goal. And so on. For Johnson,
> image-schemas were recurrent patterns of bodily
> experience.
>
> 	Johnson was right, but provided no biological mechanism.
>
> Stage 3: Terry Regier (1995) proposed that image
> schemas were consequences of brain structure and
> that they worked via the mechanism of neural
> computation. The container schema, for example,
> is a constellation of topographic maps that
> jointly compute the schema roles - interior,
> boundary, exterior - given shape input and using
> such mechanisms as outside-to-inside spreading
> (Ramanujan). He argued that orientational schemas
> could be learned on the basis of orientational
> sensitive cells.  Regier also showed how the
> learning of the meanings of complex terms could
> be done without negative input.  In short, under
> the Regier proposal, image-schemas are primitive
> cognitive/imagistics structures computed by brain
> structures, at least some of which are innate in
> all human beingsd and therefore universal (e.g.,
> Container, Contact.)
> However , the words and morphemes that express
> spatial concepts are complex - made up of
> multiple schemas bound together.
>
> 	Regier's book was an oversimplification
> done to get the computations to work most simply.
> His model also was not designed to account for
> image-schematic inferences (e,g., if you're in a
> container, you're not out of it)..
>
> Stage 4: Srini Narayanan (1997) proposed that (a)
> the aspectual schemas are neural structures in
> the premotor cortex, that (b) aspectual
> understanding is a matter of simulation, that (c)
> aspectual inferences were arrived at by neural
> computation over those neural structures, that
> (d) metaphorical mappings are neural circuits
> linking different ("source" and "target") brain
> areas, and that (e) metaphorical inference arises
> from the computational combination of source and
> target neural computations via the brain
> circuitry that constitutes metaphor.
>
> 	Narayanan added aspectual schemas and
> showed how they could do inferences via neural
> computation. Narayanan's model was made to fit
> the theory of neural simulation, based on
> findings that the imagination of perception and
> action uses the same neural substrate as
> perception and actions themselves.
>
> 	Though there is no running computational
> model of Regier's and Narayanan's systems work
> together, one can imagine various ways of
> combining their systems to produce inferences for
> primitive image-schemas.
>
> Stage 5: The Cog Theory of Image-schemas. In  the
> course of writing The Brain's Concept's" with
> Vittorio Gallese, I noted that Srini's aspectual
> schemas are (a) located in "secondary" brain
> structures (cf. Gallese  and Lakoff, submitted to
> Cognitive Neurospyschology and present in 2003 at
> ICL ) with (b) neural connections to more
> "primary" brain areas (closer to effectors and
> sensors); and that (c) they compute the semantics
> of grammatical elements. I applied the term "cog"
> to all  such cases. I then observed that,
> thinking in Regier's terms, all image-schemas
> could be understood as cogs.
>
> Under the cog theory, (1) image-schemas are
> computed specific neural circuits used in
> sensory-motor operations, (2) those sensory-motor
> operations are multi-modal (cf. Lakoff and
> Gallese), (3) image-schemas are multi-modal and
> not located in any one module, (4) different
> image-schemas are computed by circuitry in
> different parts of the brain, and (5) the
> circuitry operates over neural clusters, not
> individual neurons.
>
> This theory explains why image-schemas are
> examples of what Talmy calls "ception" - neutral
> between perception and action, common to both,
> yet used for conception.
>
> The Cog Theory of image-schemas is vague, and necessarily so.
>
> It is informed by what we have learned from
> Regier and Naryanan about the kinds of neural
> structures that could carry out the necessary
> computations. But it makes no claims about the
> exact neural circuits that do the job, because
> not enough is known. But, it is only by via the
> Cog Hypothesis that neuroscientists could even
> imagine looking for circuits of the right kind.
>
> Stage 6: ECG: Embodied Construction Grammar. ECG
> uses a notation motivated by brain studies and
> neural computation that can be used for
> characterizing linguistic structures precisely.
> ECG elements reduce to neural clusters and to
> circuitry that can do the appropriate
> computation. In ECG, only the parameter structure
> of image-schemas is notated - the topological,
> orientational, and force-dynamic structure is
> factored out and reduced to a formalism that can
> be used in language understanding systems. The
> formalism, though lacking the imagistic and
> force-dynamic content, has the utility of
> permitting precise accounts of image-schemas.
>
> =46or linguists, the big advantage of the cog
> theory is that we can go ahead using ECG
> formalism with a reasonable guess as to how the
> formalism will eventually be fleshed out, but
> with out having to know about the neuroscience
> details.
>
> Given the rapid development of the theory of
> image-schemas, it is understandable that there
> should be conflicting understandings.
>
> Given this as background, some comments are in order.
>
> Here's what Monica's conference announcement said:
>
> In cognitive linguistics, image schemas are
> pre-linguistic cognitive structures, arising from
> universal aspects of how the human body interacts
> with its environment, both physical and social,
> and existing largely outside of conscious
> awareness. It follows that image schemas are the
> same for everyone, regardless of the language a
> person speaks. In contrast, the idea of
> linguistic relativity maintains that language
> influences thought. The goal of the workshop is
> to scrutinize assumptions surrounding
> image-schemas and linguistic relativity in an
> attempt to elucidate (and resolve) the conflict
> between the two research areas.
>
> =85   The fragment "In cognitive linguistics, image
> schemas are pre-linguistic cognitive structures,"
> fits all stages.
>
> =85   The sentence, "In cognitive linguistics,
> image schemas are pre-linguistic cognitive
> structures, arising from universal aspects of how
> the human body interacts with its environment,
> both physical and social, and existing largely
> outside of conscious awareness," fits Johnson's
> 1987 characterization, but doesn't mention the
> more recent neurally-based accounts (though it is
> consistent with them).
>
> =85   The sentence, "It follows that image schemas
> are the same for everyone, regardless of the
> language a person speaks," fits all stages.
>
> =85   The sentence, "In contrast, the idea of
> linguistic relativity maintains that language
> influences thought" contains two mistakes.
>
> The most relevant mistake is contained in the
> phrase "in contrast." Every version of the theory
> of image schemas is consistent with the idea that
> language influences thought. The reason is this:
> Image-schemas are conceptual in nature, part of
> the sensory-motor system (see Gallese & Lakoff).
> They are part of the semantic pole in Langacker's
> sense. Language requires a linking of the
> phonological pole with the semantic pole. A
> spatial relations term is phonological in nature.
> The term (morpheme, word, or phrase) when paired
> with its meaning is  part of language.
>
>   Now each spatial relations term, in general, is
> paired not with one meaning but with a radial
> category of meanings (see Women, Fire and
> Dangerous Things, case study 2) and the writings
> of Langacker, Casad, Lindner, Brugman, and
> others). Such radial categories differ from
> language to language, and each language form a
> different system of how  spatial relations terms
> are related to their categories of meanings.
>
> (a) Since this system of pairs of forms with
> meaning-categories is language particular,
> (b) since it works unconsciously,
> (c) since the elements of the meanings include
> complex image-schemas  (complexes of primitive
> image-schemas used automatically as complexes in
> thinking), and
> (d) since image-schemas are embodied conceptual
> universals, it follows that both
> (e) "language infuences thought"  and
> (f) image-schemas are universal.
>
> Not only does the universality of image-schemas
> not conflict with the idea of linguistic
> relativity, these two ideas have consistently
> been discuss together in the cognitive linguistic
> literature: See chapter 18 of Women, Fire, and
> Dangerous Things. In that chapter, I discussed
> the second mistake in Monica's sentence beginning
> "in contrast," namely that linguistic relativity
> is the idea that language influences thought.
> That is only one small part of Whorf's idea. For
> the rest, see the long discussion of Whorf in
> Women, Fire.
>
> I hope that this false dichotomy does not persist.
>
> =46or example, Wally Chafe, in this discussion has
> written "thought involves imagery and emotions."
> He is of course correct. But that has no bearing
> on whether image-schemas are universal and
> computed by brain structures that are pretty much
> the same (computationally equivalent) for
> everyone. If anyone is aware of the complexities
> of word meanings and the differences in thought
> across languages, Wally is. But that again has no
> bearing on the question.
>
> I hope this helps.
>
> George