A Language Model

OK, folks, here's my 1,000-word-or-so synthesis of How Words Work. Someone please wise me up before the Transcendental Signifier doth smite or kabong me.

On.

Structuralist and post-structuralist poetics descending from Saussure treat semantic meaning as systemic. A familiar demonstration involves any attempt to define a single word, in which one inevitably involves other words.

However, the words we use we do not define.

True, the 26-volume OED fairly bursts with gorgeous definitions, and when I look up the 4 pages or so of entries for the or of, I'm startled to realize that the definitions actually do bear an unusually strong resemblance to my own understanding of these words.

Do I use these definitions to write or speak? No.

The OED's entries are more-than-usually admirable approximations, nothing more. Likewise, when colleagues in discussion refine terms to increase the rigor of a conversation, we may restrict, augment, amend, or disallow associations to one or another word or phrase, but probably few among us would attempt to rigorously defend the idea that we truly define terms.

So, what does happen?

Language seems to involve a dynamic interrelationship between processes described as semantic and syntactic. Neurologists observe that cerebral activity related to language seems to have a dual center, around Wernicke's and Broca's regions, so whatever flaws may exist in the descriptions, language would seem to be in some sense dual or bilateral, and the overall processing perhaps equally describable as an articulated process or related processes.

So, given A Word, my mind summons association which somewhat resembles that of other English speakers. This "Word" might as well be a morpheme or a short phrase, but either way, my mental tracing of that sound, along with the network-constellation of associated experiential traces thereby summoned, comprises what a saussurian or structuralist might describe as signifier and signified, respectively. It's probably worthwhile to explicitly deny that I intend by this any transcendant connection to any universal idea or to anyone else's thought. We humans are similar organisms, of a species, and we hang out together; to some extent we manage to work out what our confreres are thinking; we get by, but not too altogether awfully well.

Any definition of that Word is an entity separate from and disparate from the word itself, an attempt to represent the Word-object-process in language, the better to discuss it. We have wonderful motives to define words, but those definitions do not equal the words themselves any more than Magritte's "Celui n'est pas une pipe" was a pipe, no more than I can pronounce the word table and then brace my coffee on that word while I type.

That said, the ultimate circularity of definition pointed out by Jacques Derrida in his early critiques of structuralist linguistics holds problems for meaning in that it holds problems for definitive meaning, but only in that. This takes nothing away from deconstructive critiques of attempts to linguistically objectify and delimit, but describes differently the possibility of describing and the impossibility of defining, delimiting, or objectifying the world.

Also, much of the complexity of semiotic description comes from the difficulty of matching an understanding of semantics as largely synchronic and the meanings of words as relatively durable with the obviously temporal shifting of meanings within syntax. By eschewing the relevance of defined or definitive meaning, the following simple but admittedly rough description:

Given A Word, my mind summons association which somewhat resembles those of other English speakers. Such associations are material-energetic object-acts independent of whatever process may take place in anyone else's head; no transcendent eidos or anything terribly similar is involved.

Also, whether I describe what's summoned as association or as associations, my phrase is equally misleading, since the event could as rigorously be described as compound or articulated, as process or object.

As I maintain awareness of the Word, the composition of more or less prominent aspects of association remains provisional, pending the relative completion of syntactic and syntagmatic segments.

(I've found it useful to compare this to visual attention, in which I pay more attention to or devote more neural processing to and have more complete understanding about an arc within some degrees of a point central to my binocular focus than I do about events at the periphery of my vision. Similarly, the word dog by itself may summon a ready image, which I may freely alter when it's followed by another syllable, as in dogfish or dog days.)

As Words arrive, whether by speech or by my reading, I take cues internal to the discourse to indicate that segments of discourse have closed. In practice, I may do so because I take the speaker to intend such closure or because I find I can't retain more linguistic elements and must guess as to the speaker or author's drift (see Beckett's How It Is or certain sentences of Faulkner or Proust for different ways this may work). In any event, to whatever extent and in whatever ways I consider previous discourse-segments as closed, I may forget their independent parts or banish them from immediate attention. The process is in some manner intensively recursive.

At whatever point I do this, I make an assembled construct of meaning-up-to-now, and proceed to use that as interpretive context for whatever discourse follows.

A few observations, incompletely systematized ----

In these events, nothing comprises a full stop.

As I decide that segments of language have closed and I combine words, these combinations draw attention to mutually fitting or harmonious aspects of the word-string; aspects that can be mutually imagined or embodied or otherwise incorporated. I'm tempted to suppose that this involves the firing of mutually excitable neuron-patterns, and I do supose that specific spatio-temporal embodiment must be involved, but I'm just supposing.

In any event, nothing in the process seems to be actively delimiting in the sense of a definition, but poetically suggestive. Certain meanings get eliminated, but only in the way that enacting a decision to leave the room eliminates staying in the room.

In consequence, my appraisal of the language elements remains more or less provisional, contingent on the progressive and relative closure of linguistic units, and I willingly reappraise their meaning until I am hit by a blunt object, distracted, or otherwise forgetful.

For instance, certain non-assembled semantic units do cross the "full stop," the grand divide of predication and of sentence. These tend to relate to whatever we consider as noun or scene rather than to what we consider as verb or action. So, for instance, if I write "Jim studied. He passed the test," one assumes that He refers to Jim so that studied and passed relate in a way one might diagram like the two oxygen atoms in diagram of an H2O molecule.

Accordingly, no rigorous distinction between syntax and syntagma exists. Disconnection and connection are complex, partial, and distinctly articulated, like relations between movements of a symphony or sonata.

Accordingly too, lecturers in the academy aside, humans speak something far closer to verse than prose.

The music of speech, the prosody of verse, standard punctuation as prescribed by the MLA, advertising in glossy magazines, and visually sophisticated Websites all differently stylize the logic of segmentation in ways that change syntax, therefore meaning. We interpret each of these in ways that we little understand, with little or no explicit training. That is, one reading a Website aloud will tend to convert its visual cues to aural cues based on an implicit understanding of phrase boundaries. To some extent, this understanding follows lines of custom, but to some extent the reader may reinvent it upon encountering the text and guessing at "the author's intent" or at "what makes sense."

The cues that determine phrase boundaries have to be distinguishable from cues fixedly associated with semantic meaning. So, for instance, two enunciations of the word "father" must share much meaning regardless of, say, their pitch, but pitch that rises relatively towards the end of that word effectively makes the word a question.

Meaning fixedly associated with a Word or signifier may be arbitrary, iconic, or anything on a continuum in-between, but any meaning-relationships malleable by syntax and associated with that word must be arbitrary in something like Saussure's sense, as opposed to iconic or symbolic as Peirce used these.

Some difficulties remain in describing relations between what has been called sign and referent, between a composite Word and what's taken to be a solid and identifiable thing-in-itself. These may be alleviated with the realization that what's taken as a thing-in-itself is itself a neurologically constructed composite that may in some ways less resemble the circumstances of its extra-perceptual provocation than might the verbal rendering that synthesizes sensual elements with remembered and imagined categories and abstractions. For instance, I remain convinced that the chair across the room has weight although I cannot at the moment verify that sensorially.

Among other things, this suggests that discourse involves what might be called an inevitable materiality -- although, once again, materiality and ideality remains a difficult dichotomy to maintain; what this materiality amounts to is sensual palpability. The non-semantic and officially non-syntactic aspects of a material production of writing must and do signal syntax. The undifferentiated font of a book in standard typeface signals a sequential read as easily as the differentiated font-face and color of a Website or glossy magazine may signal distinct relationships of different bodies of text onscreen, and in turn a process of cherrypicking links. If early 21st-century commentators tend to neglect this, it may be because we remain accustomed to a world of uniform letters determined by movable type and typesetting. The technology has shifted, and practice is shifting although conventions and usage take time to adjust. Yet the possibilities -- the inevitability! -- of sensory and social inflection and contexting on rhetoric, narrative, and prosody incumbent on the enhanced plasticity of digital representation is boggling -- demonstrably so, given our partly generational difficulties in absorbing it. On the one hand, one grasps that new forms will be interpreted on-the-fly by global patterns; on the other hand, one must grasp that as-yet-nonexistent social customs will factor into every interpretation, so that the meanings of one's own words have not assembled yet, and leave one's control remaining radically incomplete.

Deleuze & Guattari

What a joy to get back to Deleuze & Guattari after all these years! -- & with more training in semiotics as well.

Let's begin with the infernal rhizome, a concept that D&G redefine and rework throughout 1000 Plateaus. Wikipedia describes a fairly standard botanical rhizome as follows:

In botany, a rhizome is a usually underground, horizontal stem of a plant that often sends out roots and shoots from its nodes, though a number of species of plants have above ground rhizomes or rhizomes that sit at the soil surface including some Iris species. Rhizomes may also be referred to as creeping rootstalks, or rootstocks.

So, plants that normally reproduce by shoots form rhizomes. The most familiar example might be the fungal mats that send up mushrooms or toadstools after a rain. D&G use both the synchronic form of the irregularly interwoven mat and the chronological process of the exploring radicle-tip to describe various linguistic, semiotic, artistic, social, and cognitive forms.They oppose rhizome to what they call not linear but arborial form, their prime example being Chomsky's branching analysis of sentences. Anti-Oedipus and particularly 1,000 Plateaus became popular in the USA in part as theorists scrabbled to adjust to the formal demands of epublication.

OUR STORY (in 1,000 Plateaus, assuming one reads the book in sequence, which D&G don't particularly recommend) starts as D&G, apparently leery of the sequentiality of their own hardcopy publication, launch an attack on certain assumptions of books.

A book has neither object nor subject; it is made of variously formed matters, and very different dates and speeds. To attribute the book to a subject is to overlook this working of amatters, and the exteriority of their relations. It is to fabricate a beneficient God to explain geological movements. In a book, as in all things, there are lines of articulartion or segmentarity, strata and territories; but also linees of flight, movements of deterritorialization and destratification (3).

D&G's destratification resembles other uses of image to qualify abstract distinctions; the authors mean to undercut certain hierarchical relations of ideas. Their deterritorialization is a little odder, but pertains to the destabilization of ideas by something rather like Derridean difféance. D&G, strangely and tellingly, don't mention Derrida anywhere even when their argument follows "Structure, Sign, and Play" point by point.

On time --

Thought lags behind nature. Even the book as a natural reality is a taproot, with its pivotal spine and surrounding leaves. But the book as a spiritual reality, the Tree or Root as an image, endlessly develops the law of the One that becomes two, then of the two that become four . . . . Binary logic is the spiritual reality of the root-tree. Even a discipline as 'advanced' as linguistics retains the root-tree as its fundamental image, and thus remains wedded to classical reflection (for exapmple, Chomsky and his grammatical trees, which begin at point S and proceed by dichotomy (5).

As one tries to put these ideas into practice in electronic literature, it becomes less obvious how Chomsky's analysis and its binary distinctions is arboreal or linear as opposed to rhizomatic or global. A rhizome itself is, in a sense, made of lines and intersections of lines, and the proceeding live radical proceeds teleologically. Since one must speak one word after another, natural language proceeds temporally. Consequently, most if not all language involves constant binary opposition at least in terms of temporal comparison of linguistic elements. Each word or sound must be before or after something else, and we find that significant. Also, as the various structuralisms and formalisms have pointed out in their various ways, semantics involves differentiation; for every a, one must understand a more-or-less articulated non-a. However, the very branching of Chomsky's system -- arboreal, if you will -- does provide for a synchronic differentiation and comparison of linguistic elements that ultimately involves synchronic interrelationships between more than two entities, and non-oppositional differentiations. That is, the binary differentiations must be ultimately taken as elements of a larger set of interrelationships, as a combinations of edges in a picture, in which every edge comprises a binary division, can be taken to describe events that are otherwise non-binary (take a diagram of a triangle as perhaps the simplest example).

One might be tempted to dismiss D&G's criticism of Chomsky on this point alone. This would be premature, however. Their criticism of Chomsky, after all, is not exactly that he might be incorrect. They describe his work at various points as a "tracing," as opposed to a "map." This word tracing, again, is not self-explanatory, but a simple image might be the kinds of mapping one gets from AAA when one asks them to map out a route. The initial map, which would correspond to what D&G call map, shows the layout of the land in general without respect to any particular voyage. The AAA counselor will "route you out" by taking a highlighter and marking along a convenient route from where the AAA member starts and finishes. The route is then complete, in a sense, in that one has a personalized plot or model of how one might proceed. It's not exactly linear, but contains a line of travel along with the surrounding areas that we might take as corresponding to immediate associations.

Following this out, one finds that D&G have a truly remarkable model here. What one gleans is that Chomsky's model corresponds significantly to linguistic structures, and renders significant information and questions -- all that Chomsky himself ever claimed for it or, one suspects, would allow any other theory -- but that it does not describe the entire event available for description; it only describes one abstractable operation.

Again, one might counter that this is fair enough without diminishing Syntactic Structures or what follows in the slightest. And, one might continue, the same criticism could be made of D&G's own argument, which equally follows a single logical chain, and that one's read through it is as temporal as any other linguistic process. Yet D&G do leave abundant evidence that they have indeed recognized these objections and have engaged themselves in response. 1,000 Plateaus itself is laid out in such a way that their own arborescent arguments intersect richly, creating complex parallelisms as one proceeds between chapters.

The metaphorical use of rhizome and the rhizomatic structure of the work vary in each tracing; the chapters thus inform one another.For instance, the critique of Chomsky returns in various guises:

Chomsky's grammaticallity, the categorical S symbol that dominates every sentence, is more fundamentally a marker of power than a syntactic marker; you will construct grammatically correct sentences, you will divide each statement into a noun phrase and a verb phase (first dichotomy) (7).

This follows Derrida's politicization of what the L=A=N=G=U=A=G=E school folks (Bernstein, Hejinian, Silliman and others) call closure. (The discussion of the pharmakoi in Of Grammatology is probably still the signal text here).

A discussion of the implications of Chomsky's diagrams on power politics seems odd without a discussion of Chomsky's own anarco-syndicalist politics, but D&G probably don't mean to question the sincerety of Chomsky's commitment, but the interrelation of his analysis to alternate models of language, including its viability for use with apparently contradictory models. Part of the clue for this comes down the page, and the form in which one compares illustrates the point D&G seem to be circling around.

A semiotic chain is like a tuber agglomerating very diverse acts, not only linguistic, but also perceptive, mimetic gestural and cognitive; there is no language in itself, nor are there any linguistic universals, only a throng of dialects, patois, slangs, and specialized languages (7).

This describes what Chomsky calls I-language. But on an individual or intra-personal level, one may see that sequences of sentences may apply to each other, in ways that change the realtime arrangement of syntactic and semantic elements. Since I've just gotten done reading DH Lawrence's piece on Classic American Literature, it's an example that comes freely to mind. In his usual high dudgeon, DHL syncopates sentences and paragraphs. He uses both fragments and one-sentence paragraphs freely, and does so when neither would meaningfully stand alone without the preceeding or succeeding discourse. Now, not only does his prose achieve a high degree of closure in terms of its sense, but he gets downright didactic. By contrast, D&G themselves hold fairly closely to long-line arborescent sentences with full and elegant predications; and these they nest, moreover, in full and elegant paragraphs with traditionally sequential, intra-referential and closed forms. The paragraphs have identifiable topic sentences which succeeding discourse qualifies and explicates, with reasonable conclusions and transitional elements woven from the repetition of key terms. Complexity aside, the structure does not violate the spirit of a freshman essay.

The structure of the book as a whole is another matter, but even within any given chapter, reading D&G gives a far more inquisitive and open-ended experience than reading Lawrence. In both cases, one may see that elements break across the syntactic line of predication to form other connections. Again and again, through various fragments and fragmentary paragraphs, Lawrence omits elements that readers may guess from his parallelism. In doing so, he accomplishes several things: readers reassemble something very like Lawrence's ideas as their own. Within this process, Lawrence manages to suggest that we ourselves find his constructions self-evident and natural, since they seem to spring from the self-evident music of his language. And of course in so doing his language embodies the mental leaps and lunges of a person in passionate argument.

In D&G's more temporally extended sentences a converse pattern occurs.Repeated but subtly differentiated parallel modifiers give an impression of undefined and undefinable retracings of reference that finally suggest a body of representation neither defined nor fully embodied by their text. In terms of paragraphing, this amounts to interpretive operations in which one clearly plays modifiers in succeeding sentences against one another to some degree independently of their relationships to their respective predicative units. This action, of course, relates quite directly to their complaints with the dominant S-trunk of Chomsky's trees, and their analysis of this as a tracing. The predicative function surely exists, if perhaps not exactly as described. Yet the very action that embodies that predicative activity, the recursion involved in parsing linguistic units, involves other comparisons and connections that may be equally or even more important or more active.

One consequence of all this is the recognition that any linguistic model that attempts to analyze syntactic activity apart from its dynamic interrelations with semantic shifts will in some measure run aground of its own limitations. Listeners (and, to a lesser extent, readers) determine phrase boundaries by information internal to the language they encounter. Some of this information is musical or prosodic, but some of it is semantic as well. At the same time, the meaning of semantic elements is regularly deferred pending syntactic or even syntagmatic closure. This leaves us in the neighborhood of Derrida's difféance again, but also with Dickinson's many "inflections of the pen" that she insisted could be simultaneously multiple while vocal inflections had to commit themselves. (This appears in the letters, but I'm still chasing the exact quote).

A large part of the value of D&L's work here comes in their appreciation of the intrinsic retention of sensorial and even sensual or substantially alinguistic or metalinguistic materia prima within linguistic and semiotic constructs. Their constructs, accordingly, become useful in evaluating textual materiality. For instance, the Derridean analysis of linguistic structure involves an eternally elusive definition. In D&G, words do not mean because they're defined. So, for instance, one might consider an isolated word, lonelier than a character without its play, as involving a web of associative aural, visual, kinetic, even olfactory experiential trace. One should not consider this as "direct" experience that might later be modified or organized, but as tracings of previous considerations traced, organized, compared, constructed in various ways -- "associations" we call it. These associations retain similar aspects, but may allow for direct contradictions as well. Thus, one understands a relationship between a kitchen table and other tables, yet one would not attempt to eat from a table of figures or a table of contents.

Chomsky and others have enabled useful analysis by limiting some analyses to the relationships of complete, atomistic or monadistic words at a predicative or sentence level. Yet in some ways words must also be understood as extensively and radically composite, as well as fuzzy in definition. So in response to a given text, one or another aspect of a word may be prominent in my attention, whereas another aspect may be receded. I may consider a table of contents as an abstractly flat entity upon or along which contents can be "laid out" in an otherwise undifferentiated or parallel context for comparison and examination. The of contents tells me what aspects of table should occupy my attention, but the other aspects of the table association-complex ride close, somehow.

I sorely feel the need for more extensive neurological descriptions of such phenomena. For present I will satisfy myself by saying that the process appears similar to that of visual attention, in which I am "thinking about" a particular region that appears to me as central to my gaze. And if someone asks me a question about something at the periphery of my vision, I quickly turn my eyes towards that point to gather it in binocular vision, but also to focus a greater portion of cerebral processing on the matter. Similarly, congruent associations reinforced by syntax seem to "bring to mind" aspects of words within the syntactic chain.

Of course, since each usage of this sort further informs future associations with each word, we have thereby the extension of derridean play. I would have welcomed D&G's specifically citing Derrida where they clearly responded to his work, but even without that the description is quite useful. Most of the backhanded reference to Derrida comes in the chapter-essay "On Several Regimes of Signs." Here's a slice:

There is a simple general formula for the signifying regime of the sign (the signifying sigh): every sign refers to another sign, and only to another sign, ad infinitum. That is why, at the limit, one can forgo the notion of the sign, for what is retained is not principally the sign's relation to a state of things it designates, or to an entity itsignfifes, but only the formal relation fo sign to sign insofar as it defines a so-called signifying chain. The limitlessness of signifieance replaces the sign. When denotation (here, designation and signification taken together) is assumed to be part of connotation, one is wholly within this signfiying regime of the sign. Not much attention is paid to indexes, in other words, the territorial states of things constituting the designatable. Not much attention is paid to icons [for this, see Peirce] that is, operations of reterritorialization constituting the signifiable. Thus th sign has already attained a high degre of relative deterritorialization; it is thought of as a symbol [Peirce as opposed to Saussure again] in a constant moveement of referral from sign to sign.

OK, that's a mouthful, and not the least because D&G casually track Peirce's distinctions through a discussion traditionally based on the quite different distinctions proceeding from de Saussure. D&G insist on what amounts to an evocative significance of the sign beyond its existance in a system of words per se. Now, the implied criticism of Derrida is probably not altogether fair, since they don't discuss the distinctions or resemblances between spoken and written or even thought language and what has been translated as "arche-writing." Descriptions of arche-writing in Derrida's "Freud and the Scene of Writing" clearly ground Derrida's conception of the play of signification as grounded outside of areas usually circumscribed by the words langage or parole. Derrida clearly wishes to deal with neural activity as semiotic or representational in some sense; as opposed to an "out-there," or some species of noumena.

In this sense, Derrida insists on describing the process as play of differences in a relatively continuous system of signification that must operate temporally, must change, and therefore can never stabilize. D&G describe the same activity as interplay between "regimes of signs" or "planes of signifiance" or "plateaux" or "planes" of a rhizome or "tracings" that are mutually dependent in interpretation. So they might call the special terminology of Saussurean linguistics as a "regime of signs" for instance, but they consider that this cannot isolate itself successfully from Peirce's sign-system or from one or another individual's daily French or even English.

More later, ye gods.


Deleuze, Gilles and Felix Guattari. 1,000 Plateaus Brian Massumi. Trans. Minneapolis: Minnesota, 2003.