Notes Toward a Non-linear Prosody of Space

I. Linear Metrics

1. Background: Dimensions of inequality among syllables

The fundamental axiom of prosody in English is that all syllables are not created equal. (This is not the case in all languages; in Japanese, apparently, syllables are considered equal enough, metrically, that they are simply counted rather than being distinguished by structural patterns.) By plotting the time sequence of how these inequalities occur one obtains metrics -- a double plural, meaning a multitude of metrics. The most well known is the stress degree metric, which forms most of the discussion below, but there are several others. (1) Pitch Degrees. Although I have not analyzed this in detail, there is a prosody at work in many American communities, most notably in a style of reading in the black community, in which I believe pitch degrees play a more important role than stress degrees. This forms the basis of an incredibly rich prosody, a different prosody than the classical one, which must be analyzed on its own terms. Whereas stress is a linguistic property, pitch is a purely acoustical property; thus a pitch degree prosody is more absolutely musical -- in a literal sense -- than a stress degree prosody. A pitch degree prosody is more free to use an absolute musical sense of time and structure; the sound structure balance is different -- not all phonemes carry pitch. Where a pitch degree prosody and a stress degree prosody are present at the same time, incredibly subtle effects are possible. Pitch degree prosody deserves extensive analysis in its own right.

(2) Vowel Position Degrees. Again, my impressions here are based on intuition rather than detailed analysis. My understanding of this goes back to a somewhat delphic comment by Robert Duncan. Duncan spoke often of "tone leading vowels" in talking about Pound. Not understanding what he meant by this, but wanting to know more, at a reading once I asked him where in Pound's writings I could find the discussion of tone leading vowels. To my total astonishment, he replied that it wasn't anywhere; Pound had used the term in letters to Duncan, simply assuming that Duncan would understand what he meant; Duncan was left to figure it out for himself. Hearing this I simply couldn't resist: "Well, what is it??" Duncan said two things, one of which was quite straightforward, the other of which was extremely obscure. He explained that when a diphthong (a glide from one "pure vowel" to another) occurs, the "leading tone", i.e. the pure vowel at which the glide begins, plays a special role in terms of later reinforcement. So far so good. Then he said: "When you hear a sound, it's reinforced when you hear it again. But it can also be reinforced when you don't hear it again." Just as Duncan was left to figure out on his own what Pound meant by "tone leading vowels", I felt strongly that I should simply accept this remark as a gift and work on my own to figure out what it meant; I didn't ask anything else.

I certainly don't wish to be in the business of interpreting Robert Duncan, even less interpreting Duncan interpreting Pound; but the train of thought that started with Duncan's remark has ended up with the idea of a vowel position degree metric. Vowels are sometimes diagrammed as a trapezoid which plots the position of the tongue in the mouth. I believe (again this is based on intuition, not analysis) that the possible vowel positions tend to cluster into just a few groupings, and that the plot of which grouping occurs at a given moment forms a metric. This metric is probably superimposed on the stress degree metric, again giving a very fluid effect. (I don't know if this has anything at all to do with "tone leading vowels", but I certainly do hear something like this frequently when I read or listen to Pound.)

2. The Classical Stress-Degree Metric Stereotype

Before giving my own approach to the stress-degree metric, let me briefly sketch the stereotype most people have for how this works. If you want to argue that the stereotype is not accurate in scanning this or that poet, I would hardly disagree; but I think this stereotype does have a lot to do with attitudes towards metrics. Hopefully, by presenting a different approach, the stereotype can be broken.

There exist -- a priori, in advance of any particular poems -- a collection of abstract patterns of stress degrees. These patterns may be called templates. They tend to have names, e.g. iamb ('|), trochee (|'), anapest (|'') etc. (I'm using ' for an unstressed syllable, | for a stress.) Scanning is an activity consisting of pattern-matching template instances against the words of the poem. In the most formalized case, we say a line "scans" if repeated instances of the same template, end to end, match against the words of the line. There is a tendency to speak of poets "writing in" a meter, which means choosing a template in advance and writing lines which will scan to that template.

Readers familiar with prosody terminology may note with some surprise that so far I have not mentioned the word 'foot'. Unfortunately, the word 'foot' is deeply ambiguous: It has at least two wildly different uses in prosody. Because I believe this plays a crucial role in the widespread prejudice against and misunderstanding of metrics by many people, I want to explain this in detail, and will propose a different terminology to avoid the problem. Meanings of the word 'foot': (1) 'Foot' is often used to designate a type of template. (E.g. one speaks of "an iamb".) As such it is part of an a priori scheme, given in advance of the poem. (2) 'Foot' is used to designate the group of syllables in the actual poem which is matched by one of the templates. In this usage, a foot is a unit of meter: just above the syllable, which is the atomic unit for metrics, and below the metric line. (Of course this is not meant to deny that a foot can contain just one syllable or a line one foot.) I.e. here foot is a living breathing part of the poem, with a specific tangible identity in a metrical structure.

Many different strains in poetics, such as the Open Forms of the Black Mountain poets and Improvisational Composition of the beats, have led in directions away from the a priori, away from choosing a fixed form given in advance. Because the concept of 'foot' is so closely linked with a system of a priori templates, this has led to a sad neglect of metrics generally, and to interest in the concept of mapping the metrical structure of living breathing poems specifically. To avoid this problem, in what follows I will use the term measure for a unit of meter intermediate between the syllable and the line where we make no assumption in advance for what shape a measure may take.

3. Bonding Strength

I wish to propose yet another dimension along which syllables can vary: bonding strength. Actually bonding strength applies not to syllables but to the boundaries between adjacent syllables. Before going into the details of how this works, it is important to note two important points about this methodology. (1) The methodology applies against the sound of the poem. In some cases this may be difficult to infer from a printed text absent a live or recorded reading by the poet. This idea will be quite threatening to those academics who believe that all there is to know about a poem is contained in the printed text. However, there is nothing terribly original in the idea that prosody means looking at the structure of the sound: it was Ezra Pound who said if you want to know about the prosody, open up your ears and listen to the sound it makes. (2) The methodology -- and I emphasize that it is a methodology, not a theory! -- is ruthlessly empirical. It works by examining a recitation and trying to answer the question: what is the metrical structure of this recitation. It may or may not yield useful results for any given recitation. It is not designed to answer what may be a fallacious question -- what meter did the poet "write in" -- but simply to discover whatever metrical structure happens to be there. Because poets build a voice, there is a reasonable chance that in many cases conclusions can be drawn about metrical structure, but perhaps not.

Each syllable has a bonding strength for the syllable before it and after it. By bonding strength is meant: (artificially) inject a pause at the syllable boundary in question, and then judge how natural the pause is against the recitation. There will be different degrees of naturalness -- different degrees of resistance to the injection of pause -- at different syllable boundaries. We say a syllable boundary has a high bonding strength if an injected pause is extremely unnatural compared to the recitation; where the pause is natural (or already there, of course) bonding strength is low. I will call a measure a string of syllables bounded on either side by low bonding strength and having only high bonding strength in any internal syllable boundaries. It is this empirical, overtly sound-based concept of measure which I wish to use as a replacement for usage (2) of 'foot' above.

Of course this will work with any recitation; it does not have to be poetry. I propose that the speech rhythm of English prose tends toward what I will call The Standard Measure, defined by the following rules: (1) A measure has only one major stress; (2) the stress tends to come at the end of the measure, but (3) if there are only unstressed syllables following the major stress out to the end of a natural grammatical boundary, those syllables will be incorporated into the measure. (You can think of this as an attempt to reformulate the classical idea that English is "predominantly iambic" while institutionalizing the counter-examples.) The concept of "grammatical boundary" is really the same concept of bonding strength applied to syntax. Although I haven't analyzed this in detail, a simple explanation would relate grammatical bonding strength to the distance that must be traveled in a parse tree to get the closest common antecedent to two consecutive words.

This is not to say that "the standard measure" will in fact always occur. Robert Creeley, for instance, is well known for having many non-standard measure breaks in his poems, and many other poets use non-standard measure breaks quite prominently. Detailed analysis can produce some interesting results. For instance, in some of Creeley's poems, lines are predominantly either one or two measures; interestingly in the two-measure lines the internal measure boundary is a standard measure break. (I.e. the celebrated Creeley line-break is exactly that, a line-break, not a measure- break.) The non-standard measure boundaries are easy to hear, but the internal standard measure boundaries do not stand out so prominently. (Of course you would "hear their absence" as a lifeless regularity.) In these poems one could say that the grammatical structure goes from the middle of one line to the middle of the next; but the line endings are articulated sonically by the prominent non-standard measure boundaries. This is a familiar metrical structure going back hundreds of years, all the way to Anglo-Saxon poetry.

II. Non-linear Metrics

Prosody is traditionally the study of poetic sound structure as mapped in time. Because of the inherent linearity of time, this poses an obvious problem for non-linear poetry. Shall we say that prosody only applies within the lexia, thus punting the problem completely? Some may take this view, but I find it personally distasteful. It exempts from prosody details of hypertext structure which I think clearly need to be considered. In Michael Joyce's work, the names of links are so clearly musical it takes one's breath away: they are part of the prosody. Given that I am on record as advocating taking hypertext into the fine structure of language, thereby fragmenting the lexia, I simply can't accept leaving prosody as an inherently linear concept that applies only inside of lexia.

It would be nice if what follows were as well worked out as what preceded, but at this point I have only questions and some launching points for a view of non-linear prosody.

1. Linear prosody as a prosody of space

There are some concepts in prosody that are so overtly time-based that one simply has to give up on them in non-linear poetry. Isochrony -- the tendency of major stresses to fall in an even musical beat -- is one such concept. There may be isochrony within the lexia, but given that outside the lexia there is no concept of time, in non-linear prosody one can only treat isochrony locally. However: In a one-dimensional structure, time and space are very nearly the same thing. What may appear to be a time-based concept may in fact be a space-based concept. The concept of bonding strength, as articulated above, occurs between adjacent syllables. 'Adjacent' is clearly a spatial concept, not a time-based concept. I defined bonding strength as the resistance by a recitation to the injection of an artificial pause -- time language again -- but one could just as well speak of injecting space into the poem as time. Thus:

2. Spatial Bonding Strength

The concept of bonding strength works in any kind of topology. One may speak of the bonding strength of two adjacent units as their resistance to the injection of artificial space. The replacement for the concept of measure above is a spatial clustering. This has some interesting consequences for "traditional" hypertext rhetoric. Whereas the classical hypertext link is typically discussed using travel vocabulary, a spatial prosody would ask the question:

What is the bonding strength THROUGH the link?

(This has an interesting resonance with the Kaplan/Moulthrop concept of hypertext warping space.) It may be objected that in trying to assess bonding strength through the link, one is reversing the direction of the arrow (assuming a one-directional link, and recalling Ted Nelson's caution that all links should really be bidirectional). Well, having no problem measuring bonding strength "against" the arrow of time in linear poetry, I have no problem doing the same thing "against" the arrow of the link. Travel may be the right vocabulary to use, but one can also speak of the attraction of two nodes for one another; yes, there may be an asymmetry in link direction, but there is always an asymmetry between "located here" and "could be located there". In discussing hypertext there is an overt tendency to discuss following links based on similarity to where one is; but perhaps not! Perhaps one wants to take a link based on dissimilarity to where one is! (Robert Duncan: "A sound is reinforced when it is not heard again ...")

Prosody may form an overt basis for following links. I find myself following links in Michael Joyce's work based on sound structure all the time. Perhaps I'm "not supposed to do that", but I do, and without having asked him about this I think Michael wouldn't mind that I do.

Prosody must work THROUGH links, not just inside "the lexia" (if one has lexia ...)

I speak about "links" here only because that is the most familiar form of hypertext structure; this discussion generalizes to conjunctive structures as well.

3. Agencies

By "agency" I mean a unit of doing. Because there is an overt structure resulting from hypertext linkage, we have a bit of a tendency to focus on that structure rather than the structure of what the reader does. Consider the familiar node/link/lexia model. A reader follows a series of links, then -- for whatever reason -- decides to backtrack. That series of nodes is a unit of doing. Unfortunately, most of the software does not treat it that way. For instance I can't save the series. Perhaps I can save my entire history, but I typically can't mark it to denote the way I think the links clump together. Perhaps the best I may be able to do is save my history to a text file and then import it into some tool where I can annotate it.

Doing has boundaries, has a shape, has units. The structure of doing unfolds against the skeleton of the hypertext structure, but it may not be the same. The hypertext structure may appear to be disjunctive, but the reader can do conjunctive things with it anyway. (E.g.: "These are the links I really like, they seem to work together ...) In a hypertext with conventional lexia, there is a concept of bonding strength among lexia.

When I speak of agency, I am not speaking of history. Agency and history are related but are not the same. The structure of history is relentlessly linear: "this is the sequence of what happened ..."; the structure of agency is likely not to be linear. There will be clusters of doing that include the same event. Some agencies will stand out clearly in the reader's mind; these may clump together even though they are isolated in time by navigations that don't stand out in the reader's mind as a clear agency: agencies don't always happen.

4. The Granularity of Agency

We have, perhaps, come far afield of prosody. I am propounding a view of linear prosody as clustering in one-dimensional space; from this one generalizes to clustering in a space of arbitrary topology. The granularity of this clustering as it traditionally affects prosody is very fine: down to the level of the syllable. A hypertext prosody granularity of space can become this fine only when hypertext is taken into the fine structure of language. I am always (it seems) advocating for this, but with few takers at present. With a more "traditional" lexia, there is a "granularity boundary" concerning units of prosody that are inside the lexia vs. those that transcend the lexia. Is the poetic lexia a stanza? Is it a line? Is it yet larger than the stanza? Even I would balk at taking hypertext inside the word. The words are given to us, by and large; it does not seem reasonable to me to intervene in that natural process with an external administration of hypertext structure. It should be noted that in the discussion of linear prosody above, the "standard" speech rhythm of English does not interpose measure boundaries inside the word either. If I will not intervene inside the word with hypertext structure, and the natural rhythm of English does not intervene inside the word, that leaves the entire measure in a hypertext context likely to be linear. At its finest granularity, hypertext structure relates measures.

In the large, the structure of agency extends to the session. While surely multiple sessions will associate in the mind of the reader, just as there is a natural limit at the measure boundary -- even though measures contain still smaller units, namely syllables -- there is a natural boundary at the large end of the granularity of agency at the session. You have a well defined cut in doing when you sign off, surely. As obvious as this sounds, there is still a great deal of work to be done in understanding what the identity of the session should be. Hypertexts tend often to be large; it can take many sessions before the reader even begins to get a feel for how a particular hypertext "is supposed to work". Because the number of paths is genuinely infinite, the reader may have no help at all in deciding when a session should end.

Shall we say as writers, overtly and explicitly, that we want the reader to compose a structure of agency? Shall we say that even if the writer does not want the reader to do this, readers will want to anyway?

It seems the prosody of hypertext needs new terminology. I am comfortable with the terms syllable/word/measure/lexia, but up from there the familiar terminology breaks down completely.

5. Multiplicity of Agency

There are clearly multiple concurrent agencies in a multiuser hypertext that is being used by several users at once. An interesting rhetorical question arises from this: do we need to discuss multiplicity of agency in the context of a single reader reading a single hypertext? Put differently this is the question: How does a single mind apprehend a network? Must one simulate multiple concurrent agencies in one's own mind? (It is interesting to note in this regard that some theorists of mind -- notably Dennett -- have proposed that despite the illusion that the mind is a singular stream of agency, in fact the mind is a multiplicity of concurrent agencies.)

Many in the hypertext community will surely object that in this whole discussion so far, I have spoken as if reader and writer are different people -- i.e. the hypertext is "closed". It seems reasonable to me to entertain the idea that writing and reading are separate agencies even if done by the same person in a single session. Writing-as-reading simply becomes yet another dimension to the structure of agency.

Concluding Question:

How do we -- or do we at all -- code the hypertext for the structure of agency?

Jim Rosenberg's home page o Poetics