Wine Dark Sea

Blog

  • Penalty Kicks and Distributed Movement

    Gabriel Dias, graduate student at RPI, has recently modeled the way in which penalty kickers move their bodies as they prepare for a shoot. His findings suggest that there are several “tells” – for example, the angle of the hips, or the position of the planted foot – which predict the ultimate direction of the shot. In the PBS interview that I’ve linked to above, he alludes to the existence of “distributed” movements which show the physical commitment of the kicker to one outcome or another. I hear the word distributed and I immediately think, “integrated physical system,” like a body that is constrained to do certain things because of the way its different parts interact. We see this integration in the competitive world of athletics and the expressive realm of dance. (Perhaps the adjectives here should be reversed?)

    In our analysis of texts, we have also find distributed movements of a sort. We find, that is, that certain types of words tend to move with each other in some genres, and others move away from one another. Does this mean that genre is a physical system like penalty kicking, and that our explanation of these distributed movements – of words rather than points on a body – are themselves grounded in a physical reality? I have myself offered analogies to describe this “commitment of weight” in the process of using words to do certain things: if you want to write a Shakespearean comedy, there are certain things you are likely to do: you will tend to use more first and second person singular pronouns and less description than you would in, say, a history play. If Docuscope is the goalie/keeper, it may need only 30 or 40 lines to decide that the ball is going to go toward comedy rather than history. Other things will be ignored as incidental. If say that tagging a play and watching how its “points” move in a mathematical space is like biotagging a kicker and studying his or her movements, I am proposing an analogy. Like kicking, writing is a behavior. In certain situations (penalty kicking, writing for the stage), some aspects of this behavior are signal or cardinal — position of hips, use of pronouns –while others are inessential, like the curve of the kicker’s index finger. (Actually, given the dynamism of the human body, I would be surprised to find out that there is not, on some level, a connection between finger position and kick.)

    So, does this mean I am advocating an essentially structuralist account of genre? Am I saying that, because language use is a behavior, then writing in a particular genre is also a behavior with certain “tells” that are, in a sense, built into the physical system of writing? I think people who are doing iterative criticism need to have an intelligent answer to this question, complete with an analysis of its underlying analogy. My answer would be that writing fiction in a historically bound literary field does, like penalty kicking, count as a behavior and that such behaviors will exhibit coordination. There is as much connective tissue in language, grammar, plot and audience expectation as there is in the fabric of the human body. But this is not the same thing as saying that there is an essential structure to particular types of writing – that the existence of a tell implies an underlying recipe, essence or structure that is genetically dictating the behavior of the writer.

    Why doesn’t structuralism follow from linguistic integration? First, writing is not like penalty kicking. Dias chose penalty kicking because it is a binary physical outcome. With respect to the standing keeper, the ball goes left or right. Language, on the other hand, is like a flock of birds: it can break any way, 360 degrees, and is doing so dynamically at all times. “Yet the flock shows direction,” you say. “Individual birds may be wobbling left and right, up or down, but there is a recognizable trajectory within the group.” Perhaps there are deterministic ways of saying where this group is going to go next, but I doubt it. The total behavior is distributed, immanent: it has massive integrity as an aggregate, but the existence of that integrity does not imply some non-negotiable locus of control. Another way of saying this, and now I am channeling Whitehead, is to say that the direction of the flock is a continuously unfolding event or “society” of actual occasions. Thus, the penalty kicking example is good for showing entailment and distributed connection in the elements of literary linguistic analysis, but bad as a model for the errant and multiple trajectories of writing.

    The existence of the tell essentially pushes back the timeline of intelligibility of the direction of the ball. A good keeper or student of physiology – like a good literary critic – will know earlier than most what kind of behavior is being exhibited. But unlike a keeper in a football game, the critic is not looking for a binary outcome. Rather, the critic or spectator is comparing the unfolding action onstage to any number of possible theatrical “types” of entertainment and generic conventions. Shakespeare takes five penalty shots at a time, all the time. If you are interested in this aspect of the play – its participation in comic conventions – yes, there will be “signal” or orienting linguistic events at the level of the line which you could consult to predict what he is about to do. But you don’t have to consult the tells and this is not a penalty kick: you already know what is going on and, indeed, are a better judge of the texture and generic tonalities of the play as it unfolds than a keeper who has to wait for the ball to be kicked. (Docuscope really is a keeper; it knows nothing until the event happens.) As we have seen in our research, human beings are massively sensitive to variations and distributed cues in linguistic behavior. We make an astonishing number of connections between the kinds of variation we see among the plays and texts we have encountered. Finding out that there is a linguistic “tell” for comedy doesn’t then mean that comedy essentially or structurally “is” the series of tells we reliably find for it. The “tells” here are a parallel description — and this, after the fact — of a perceptual reality that we render qualitatively and immediately, in our feel for certain types of writing or stories.

    I have used the words “signal,” “cardinal” and “orienting” to describe the types of tokens that serve as good landmarks for genre in this alternative descriptive universe. I do not use “essential.” As we work further through this analogy between physical and linguistic behaviors, I think we should adopt Spinoza’s metaphysical position from the Ethics, that there is a parallelism between the twin domains of thought and extended physical beings. Neither has priority. When understood as a species of behavior, theatrical writing or literary production must obviously exhibit certain empirical regularities: it takes place on the fleshy platform of human consciousness and is constrained by the physical limits of our bodies, environment and history. As critic, I would want to insist that no material factor – the practices and limitations of stagecraft, the documented or remembered history of past performances, the politically charged distribution of resources and cultural actors – can be a priori excluded as unexpressable in the behavior that is writing. All constraints are summed and expressed, but in different amounts. But I would also want to insist that– whatever the behavior is that we are tracking – there has to be in place a certain set of agreements to make sense of the “movements” in this system as such. I have to want to count “these types” of words and not those. I have to search for significant coordination of these counted things with respect to “this type of outcome” and not another. Someone has to have the desire to study penalty kicks, for example, or authorship, or genre: behaviors don’t simply want to study themselves.

    The tell is a “sign” that speaks for the kicker, and speaks early. It is a signal event worth attending to if you are a keeper. It is simultaneously an element in a causal sequence, constrained by events prior to it, and a negotiable sign or expression of an intention to do something. It is a physical way of saying, “I mean to kick the ball this way.” The point of the parallelism is that you never get to dump one half of the phenomenon. Leaning to the left, we acknowledge: all physical tells may be redescribed as expressions of an intention, and so tokens of meaning. But inclining to the right, we say: all tokens of meaning are, on some level, also indexes of empirical constraints. The keeper has to dive both ways.

  • Genre Dependence on Character Ideolects? (by Mike Stumpf, UW Undergrad)

    And yet, we know that when human beings are involved, all findings are provisional. Odd.

    Dendrogram displaying various segments from Romeo and Juliet

    To expand on Michael Witmore’s comments in his previous post, it is indeed odd how provisional our results are.  Case in point: I have been examining what John Burrows and Hugh Craig have called the “ideolects” of characters in connection with the plays in which they appear.  I stumbled upon this idea while looking at Shakespeare’s Romeo and Juliet and asking how the language of the title characters may be steering this play towards tragedy or comedy. (This was done as for a panel I presented on with Witmore  and William Blake for a digital salon at UW-Madison.)  Witmore and Blake are themselves working on an analysis of Hamlet without the prince, and the 1 Henry plays/Merry Wives of Windsor without Falstaff: we’re all interested in this kind of “subtraction experiment.” To see my initial findings using this techniques, you can visit my blog, All Is True.

  • Presentation at London Forum for Authorship Studies/Digital Text and Scholarship Seminar

    Jonathan Hope and I presented here in London on a trip arranged by Brian Vickers and Willard McCarty. It was a lovely occasion held in Senate House, attended by some we knew and others we got to know. We began by rolling out paper copies — six feet long scrolls! — of the very large diagram that you saw in the last post. One of the things we have begun to discuss is the ways in which different forces seem to be expressed on various twigs of this dendrogram illustrating relationships among 318 early modern plays. On some twigs, everything that is being grouped together has a common author. On others, the situation is not so clear. Why, for example, aren’t there large groupings of texts written at the same time? (There are some smaller clusters of these.) The principle at work here, when texts are matched in terms of their distance scores on all of Docuscope’s available features (LATs), is that every type of difference present in the population being studies will be expressed in the result. The difficulty is disentangling which type of difference — generational, authorial, generic, company, etc. — is at work in a give grouping.

    One thing we spent some time discussing yesterday was three clusters in which Jonson’s plays appear. Here they are below:

    All of Jonson’s masques are clustered at the bottom of the diagram (except Cynthia’s Revels, which is clustered in the middle). These are possibly the most distinct items in the entire corpus we are currently working with. Notice how far right the cluster extends before joining with the rest of the diagram: this indicates its dissimilarity with other clusters. But notice too that, within this cluster (as Jonathan pointed out yesterday), there is also a lot of variation. Not only are Jonson’s masques very different from the rest of Renaissance drama (including several interludes), but they are quite different from one another. It’s like a galaxy that is far away from all of the others, but whose stars are themselves quite spread out.

    So, what about the other two clusters? We decided to profile all three and came up with some interesting findings. First, the masques. After performing PCA and then rating the clusters on the different components, we found several that were quite good at isolating the items on particular twigs. (This is not a scientific procedure, but it is our first attempt.) With the masques, we found that the language is high in StandardsPositive, StandardsNegative, and ReportingStates. Here’s an exemplary passage, with both StandardsPositive and StandardsNegative in green, and Reporting States in purple:

    Masques describe what you are seeing or have just seen in a comparatively static fashion, hence the reporting states. As Brian Vickers pointed out in the question period, the genre of encomium deals with praise and blame, which are the words that are being picked up in the positive and negative standards.

    Compare this, now, to the profile of some of Jonson’s other comedies: Poetaster, Volpone, and the other items in the top group. These items are characterized by OralElement (yellow), Question (blue), Intensity (orange), and Person Property (purple):

    Here we see a pattern we also saw in Shakespearean comedy: a lot of items associated with one to one interaction. The OralElement here marks the bustle of persons whose social function is marked (PersonProperty) and who are mixing in a state where contact must be established or maintained. Some of the satirical force of the scene is bundled into the intensity strings, which show the emphatic nature of certain social performances that are mannered and so open to mockery. We noticed these intensity strings in Middleton as well, which makes us suspect that a combination of PersonProperty strings and intensity might be a feature of City Comedy. Something to check out in the future.

    What makes this top cluster different from the second? Different author? No. Different genre? Not really, at least, not according to the ones we recognize critically. And note too that there are multiple authors on this middle cluster: Chapman, Jonson and Fletcher. Perhaps we should be thinking in terms of modes instead of genres: is there a different mode of storytelling, dramaturgy, or conducting comic business here? When we use PCA to characterize this cluster and compare the results with those that characters the one at top, we find similarity and difference. What’s similar is the OralElement (yellow), Question (blue), and PersonProperty strings (purple):

    But we now see strings associated with TimeShift (scarlet), which indicate that a person is marking the difference between two temporal frames (then/now, now/future), and here seems to be associated with figuring out what someone might do or bring about in the present or near future. Here they are anticipatory, looking at what is to come from the standpoint of the present. (In Shakespeare’s late plays, by contrast, we found that action from the past is frequently narrated from the standpoint of the present.) The other thing that is different in this cluster is something that we would never see, because it is not there. The plays in this cluster lack something:

    These purple strings, which are classed as ReportingStates. They are tokens that occur frequently in this text — look at how many of them are in this play, which is from the second cluster — but as a whole the plays in this group lack these strings with respect to the larger population of early modern drama (whereas the top group did not). This kind of relative difference between generally quite frequent items is one that you could probably only grasp with the aid of statistics. We hypothesize that these strings are allowing the actors to report action that has taken place offstage in the past, keeping attention focused on the present which is hurtling forward in time. Should this be its own subgenre of Jonson that includes Fletcher and Chapman? Would it be worth naming a grouping like this? Another question for further study.

    We received some terrific comments and questions. To our comment that the first Principal Component for this population does seem to track a broad and evolving temporal shift (plays score lower on the component as time goes on), Richard Proudfoot asked if there was more variation in the very early plays in our collection. This is indeed the case, and he followed with the point that we have an uncertain grip on this earlier population because little of it survives. Other explanations for wider variation in the pre-1590 items: English as a language is more fluid prior to 1600, as Jonathan pointed out. It may also be the case that the genre system itself has not stabilized because the professional theater is still gaining its footing in London.

    Erica Fudge asked another interesting question: some of the comic strings associated with interaction and comedy (we showed our Shakespeare comedy results) reminded her of the writing in Montaigne. What, she asked, is the relationship between skepticism and comedy, and would we be interested in tracing the presence of something like a skeptical inclination across prose writing and drama. This is a very good question. I would hope that we could study, with these techniques, something like the “sentence level intellectual culture” of the period, one that extends across genres like drama and the essay. Like most of our presentations, we left with more questions and ideas about future experiments. This work seems to us to be provisional in a way that other humanities research is not. You get an idea, talk about it with others, try it, and then decide to try something else. Academic papers at humanities conferences, on the other hand, usually present findings with an air of categorical certainty. And yet, we know that when human beings are involved, all findings are provisional. Odd.

  • Docuscope Goes Live on Shakespeare Quarterly Open Peer Review

    Jonathan Hope and I have written a new piece that we submitted to the special issue of Shakespeare Quarterly on “Shakespeare and New Media.” The essay cleared the first stage of editorial review, and is now posted at MediaCommons for general comment and critique prior to final editorial evaluation. Please visit the essay here and make your views known. The abstract and title are as follows:

    “The Hundredth Psalm to the Tune of ‘Green Sleeves’”: Digital Approaches Shakespeare’s Language of Genre
    In this essay, we explore the underlying linguistic matrix of Shakespeare’s dramatic genres using multivariate statistics and a text tagging device known as Docuscope, a hand-curated corpus of several million English words (and strings of words) that have been sorted into grammatical, semantic and rhetorical categories. Taking Heminges and Condell’s designations of the Folio plays as comedies, histories and tragedies as our starting point, we offer a portrait of Shakespearean genre at the level of the sentence, showing how an identification of frequently iterated combinations of words (either in their presence or absence) can allow us to appreciate the integrity and fluidity of Shakespeare’s genres in new ways. Calling this approach “iterative criticism,” we situate our critical practice in the context of both Shakespearean criticism and more general protocols of reading in the humanities, concluding with a genre map of Shakespeare’s plays in the context of 282 other early modern plays.

    As the last line suggests, we have now managed–with the help of Martin Mueller at Northwestern–to produce an analysis of 282 plays from the TCP database alongside the Moby Shakespeare written between 1519 and 1659. I think this is the first visualization of its kind purporting to treat 150 years with of Renaissance drama, which itself feels like something of a hurdle overcome. Here it is:

    Dendrogram Produced using Ward’s clustering method on scaled data using 99 LATs to profile 318 plays written between 1519-1659, color coded by genre and separating out the works of Shakespeare as a category of their own: Red=Comedy, Blue=Interlude, Green=History, Cyan=Tragedy, Purple=Tragicomedy, Orange=Masque, Gold=Shakespeare. The item names follow the protocol: (genre)-(date)-(author)-(title).

    Two points to make here, although there could be many more. First, this diagram was constructed using scaled data, which means that the “mile away” linguistic markers of similarity and dissimilarity are being balanced with markers whose variation is less visible from a distance. Variables with large standard deviations are not dominating with respect to those with smaller ones. Note then that most of Shakespeare’s works cluster together here, comedies, tragedies and late plays all on the same twig. When I tried this analysis using non-scaled data, these genres split up and Shakespeare’s comedies clustered together with Jonson’s, suggesting that Ward’s clustering procedure on unscaled data is better for picking up genre differences, while the same procedure conducted on scaled data (as is the case here) is more sensitive to authorship. (For an earlier analysis of Shakespeare’s plays only using scaled data with Ward’s clustering technique, see this.) This finding should be tested in other contexts and with other data sets, but it is interesting, since it suggests that authorship becomes legible when fluctuations in variables that contain lots of tokens (say, Description) are coordinated with those that have many fewer tokens. It may be this “adding a dash of something” that pulls the author as such to the fore in an analysis.

    I’d like also to offer another observation here about the fact that so many Shakespeare plays are hanging together (as are Shirley’s and Middleton’s), remaining agnostic for the time being about whether it is authorship or genre that is producing these clusterings. The majority of Shakespeare’s plays are clustering on a twig that contains mostly comedies. So when compared with 282 other items written between 1519-1659, Shakespeare’s plays look for the most part like plays that Harbage (in the Annals of English Drama) classed as comedies as opposed to some other genre. (Martin tells me that he followed Harbage for the most part, but made some guesses himself about genre designations based on title page information and common sense.) The thing to remember here is that an individual genre may cluster in different ways depending upon the larger population in which it is situated. That is, a fuller collection of texts from the period–not just the ones that Martin was able to modernize so that we could run a test on them–might show new subdivisions that end up splitting the Shakespeare block into a number of smaller splinters. (Or it may not: this may be a stabilized portrait, more or less.) The best way to understand more about the groupings themselves is to begin looking at them with the help of PCA and other techniques we’ve been using already. That’s where we’re headed next.

  • Early and Late Plato II: The Apology and The Timaeus

    In the previous post we were examining three dimensional clusterings of the Platonic dialogues as rated on scaled Principal Components 1, 2 and 5, a technique that allowed us to see the early Platonic dialogues (as defined by Vlastos) standing apart from the middle and later ones. Vlastos’ claim, we remember, was that these early dialogues represent the historical Socrates, whose technique of argumentation was elenctic. Socrates used this technique to draw out the implications of an opponent’s views until those views collapsed under their own contradictions.

    The translator of these dialogues, Jowett, would have had to preserve at least some of the linguistic “footings” required for such a dialogical structure in the early dialogues, and it was my contention in the previous post that Docuscope would detect these footings because they are exactly what a translator must preserve. Perhaps a more provocative claim, which I would like to advance now, is that the irony which attends this elenctic method — while not itself visible to Docuscope — might also require certain reliable linguistic pivots. In keeping with our analogy of the body of a dancer, certain upper body moves like the ironic twist in which Socrates seems to be asking a question for the sake of clarification but is actually pushing his interlocutor into deeper confusion, require a lower body stance that can support the weight of the move. If we could define this lower body stance, we would not be defining Socratic irony itself, but rather its linguistic correlates. (At some point, the analogy will break down, since language is not a “weight bearing system”: but it does support gestures and turns, so let’s see how far we can go with it.)

    What is it exactly that is happening in these early dialogues that Docuscope and principal component Analysis are able to see from afar? Here is a scree plot which rates the power of the principal components as they are derived sequentially, from most powerful to least:

    Scree Plot for Principal Components Derived from Cluster Docuscope Data on Jowett Translations of the Platonic Corpus

    The first two principal components are shown here to be quite powerful: together they account for almost 54% of the variation in the entire corpus. When we rate all of the dialogues on just these first two components, we get the following bubble plot:

    Graph of Platonic Dialogues Scored on Principal Components 1 and 2

    I have highlighted the upper left quadrant, where almost all of the dialogues that Vlastos identified as “early” are clustering. Their presence in this quadrant means that they score low on PC1 and high on PC2. PC1 might be described as an anti-early component, because it powerfully discriminates against early dialogues. PC2, on the other hand, might be described as a pro-early component, since its highly loaded variables are more frequently used in early dialogues. We can literally see the sorting power of these two components here, but it can also be quantified by the Tukey text, which was applied to both principal components, the results being available here and here. Note that the Apology is one of the most strongly “early” dialogues by these measures, whereas the Timaeus is one of the least early. We will pay closer attention to these two items as a way of exemplifying the differences that Docuscope sees between the two types of items.

    Before making the comparison, let’s look at the variables that are most powerfully loaded on these components and so are most responsible for discriminating the early/non-early difference. We do this either by consulting the loadings of our variables on the two principal components or by looking at a biplot which arrays those variables in two dimensions, exactly the two that were used to produce the bubble plot above. First the loadings scores (reported as eigenvectors) and then the loadings biplot:

    Loadings of Cluster Scores for PC1 and PC2
    Loadings Biplot for PC1 and PC2

    The loadings biplot (lower diagram) is a two dimensional image of the loadings scores (upper diagram), showing how these variables behave with respect to one another in the entire corpus. Clusters of words that oppose each other by 180 degrees — for example, [Public_Values] and [Special_Referencing] — tend not to co-occur with one another in the same text. Here we are interested in what makes a particular text cluster in the upper left-hand quadrant, so we are looking for vectors (red arrows) that extend furthest to the left and to the top of the diagram. Vectors extending to the left are: Reasoning, Interactivity, Directing Action, Interior Mind and First Person.  (These are the clusters that have significant negative loadings on the first column in the top diagram: if an item scores high on words contained in these clusters, it will be “punished” for that abundance and pushed to the left of the plot, as the red dots are above.) Note that we can also use our 180 rule to say something about items that are far left in the bubble plot as well: they must lack items contained in the clusters that are positively loaded on PC1, which are Narrating, Description, and Time Orientation.

    Similarly, with PC2, we are looking for the tall vectors heading upward: Emotion, Public Values and Topical Flow. Having tokens that were counted under these clusters will push an item up in the diagram, as will lacking items from the negatively loaded clusters: Directing Readers, Elaborating, Special Referencing. Note that Topical Flow (which is often populated by third person pronoun use) is loaded positively for the second principal component, but also positively for the first, which makes it fork upward and to the right. This means that an item scoring high on Topical Flow tokens will probably lack some of the items to the far left and contain items to the far right, which may discourage that item’s appearance in our “early” quadrant unless there are differences in these other variables.

    I have discussed some of these clusters in earlier posts about Shakespeare, so my main focus here will not be on elaborating the contents of the clusters. Rather, I want to use these loadings to zero in on specific words in exemplary passages from the early and later dialogues to see what is captured and then leave it to readers to say what these particular tokens are doing. Looking at our bubble plot above, the two dialogues that exemplify these opposing linguistic trends — in translation — are the Apology and the Timaeus.

    Here are two passages from the Apology that exemplify “earliness” in the Platonic corpus, if we agree that the clustering above seems compelling. Note that these are screenshots from Docuscope in which the clusters that are doing the work of pushing the texts up and to the left are turned on or color coded. I have not turned on the clusters that are absent, since these will be exemplified in the Timaeus:

    I think these passages are certainly illustrative of the elenctic method described by Vlastos, although it ought to be said that the high amount of dialogical interaction here — one that was a hallmark of comedy in Shakespearean drama — is sometimes implied by Socrates rather than really enacted by both speakers. That is, Socrates sometimes simulates a dialogue that is not really happening (“to him I may fairly answer”), and this procedure actually multiplies the Interaction strings (sky blue) beyond what might be the case in actual interaction. Note too that Docuscope is seeing lots of Public Values words, words that gesture toward communally sanctioned values, in this earlier style: demigods, heroes, fairly, mistaken, good for, doing right, disgrace. These values must be cited in elenctic exchange because they are the topic of conversation (people have opinions about them), but such implied communality may also coerce assent from an interlocutor for reasons that extend beyond mere shame at self-contradiction. We see, too, more emotionally charged words (in orange); the occasional Topical Flow token (their); and some Reason tokens (if he, thus, may, do not).

    Now look at a passage from the Timaeus, which does the things that items in the early quadrant (on the whole) cannot do:

    This is cosmogeny, not dialogue, which is why we have a number of Narrative strings (the year when, then, the night, overtaken the, as they) and Description strings (orbit, the moon, stars, sun, wanderings, motion, swiftness). Special Referencing here is picking up a lot of abstract references (dark purple) such as animals, measure, relative, the whole, nature, variety and degrees. The slightly lighter purple, Reporting strings, are complimenting the Narrative tokens: having, completion, After this, came into being, received, to the end that, created. This should not be surprising since the two vectors for these clusters were almost overlapping in the loadings biplot above.

    Whereas the Apology is staging a dialogue (real or implied), the Timaeus is creating a world and pacing that act of creation (through narrative) with a set of abstract terms that can be referenced in conversation. Indeed, one of the burdens of this kind of world-making, I think, is that the abstractions must be folded in with the concrete descriptions in equal measure so that the passage is something more than a Georgic description of a natural scene or a praise poem to nature. Note too that there is absolutely no irony in this passage from the Timaeus. That is not because Docuscope has a category that allows it to discern irony in its local environs and so rule out such an effect in the Timaeus: only a human being can make such a discrimination, by virtue of being able to look beyond the simple mentioning of words to assess their use. (For Docuscope, all counted words are mentionings of words whose single use has been classed a priori in the categories assigned to them.)

    And yet, even in translation, Docuscope may be identifying the linguistic footings of irony: a necessary but not sufficient condition for its use.