The Patron Saint of Superheroes

Last year I posted a sequence of five posts exploring how comics visualize plot. The first harmonized Freytag’s plot pyramid, Todorov’ equilibrium circle, and Cohn’s panel types; the second used that combined structure to define closure through implied plot content; and the third and fourth tested it on abstract comics.

I’ve since been in conversation with Neil Cohn (who has to be the most generous email correspondent ever), working out differences between his visual narrative grammar and my proposed structure. In short, mine (balance, disruption, imbalance, climax, new balance) exists in a comics viewer’s mind, while his panel types (Orienter, Establisher, Initial, Prolongation, Peak, Release) exist on paper (which then of course is experienced in the viewer’s mind too). That means there’s no such thing as, for example, an invisible or undrawn Initial, but there is a mental experience of a parallel disruption in a viewer’s understanding of the partially drawn, partially implied plot event. Or at least that’s my hypothesis. To make my claims worthwhile, they need some theoretical grounding.

Another way to put it, narrative panel types are event structure manifested. Cohn’s Initial-Peak-Release (typically) aligns with his mentor Ray Jackendoff’s event structure Preparation-Head-Coda. But while Cohn is looking literally at panels, “events” (as Jeffrey Zacks and Barbara Tversky explain) “are in the minds of the beholders.” Each is “a segment of time at a given location that is conceived by an observer to have a beginning and an end.” Since my aim is to explain closure through a theory of undrawn comics content, Cohn’s grammar panels aren’t useful because they apply only to drawn content.

I thought Jackendoff would be useful, but I found that his Prep-Head-Coda structure didn’t always align with my own event structure. For him an “event” consists of “a Head (the main action), with an optional Preparation (things that have to be done before the Head can be begun) and an optional Coda (things that are done to restore the status quo ante).” That’s similar to Todorov’s three phrase equilibrium-equilibrium-equilibrium plot structure (and Todorov’s fourth step is an “attempt of restore equilibrium”), but Jackendoff’s example didn’t line up.

He subdivides “making coffee” into twenty units, including “put water in the machine.” He explains:

“The Head consists of actually pouring water into the machine from the pot. But in order to do this, one must first measure water into the pot–the Preparation–which in turn is organized into Preparation plus Head, and each of these has further organization. And once one has poured the water into the machine, one must replace the empty pot in the machine–the Coda.”

So it seems “Head” sometimes refers to the culmination of a “main action” (which parallels Cohn’s Peak) and sometimes it is the “main action” (which parallels Cohn’s entire visual narrative grammar).

So now I’m going back a couple more decades to a series of experiments and papers by Darren Newtson in the 70s. Newtson divides actions into “a series of cognitively discrete units” by identifying unit boundaries he terms “breakpoints,” which are boundaries between larger, multi-unit sections and are also units themselves. Breakpoints are key because they serve as “points of definition” for an overall action, where “each breakpoint is a point of reference, or comparison, from both the preceding and following breakpoints” creating an experience of “continuity … between successive action units.” He even divides a video into still frames that he calls “an almost comic-strip summary of ongoing action sequences.”

His approach harmonizes with Cohn’s narrative grammar (because he’s using manifested images) and plot structure (because he’s also talking about unseen and so implicit behavior inferred from manifested images).  This also suggests why Cohn’s Peaks are so significant because “action units are defined by a type of information uniquely available at breakpoints.” This is true primarily because “breakpoints are selected [and here I would say “created” in the case of comics] on the basis of a meaningful change having occurred relative to the preceding breakpoint.”

Newtson also reveals something about Todorov. Todorov identifies five plot points, but only three broad units: equilibrium, disequilibrium, equilibrium. They are separated by a disruption and (altering his language a little) a restoration. That’s an event structure (and so not necessarily manifested as comics panels), but in Cohn’s narrative grammar they correspond with Initial and Peak. Viewing them as breakpoints reveals how much less important other panel types are because they are one of many available “nonbreakpoints” (Newtson’s term) within the equilibrium, disequilibrium, equilibrium spans. That accounts for why Cohn can divide the first phase into Orienter and Establisher, with multiple Prolongations in the middle phase, and (presumably) multiple Release panels in the third. Only Initials and Peaks then are unique and defining.

Looking again at my own system, I’ve already literalized the two breakpoints by drawing them as actual breaks in the preceding line angles:I now want to revise out “climax” (which for Freytag has a different meaning anyway) and use “restoration” instead–though that’s not quite the right term either, because it implies a return (Todorov’s circle) to the the original state, which is wrong. Similarly for Newtson, the second breakpoint is the same as the first, so they’re both disruptions (points of defining change) of a preceding status quo. But that doesn’t quite fit his own analysis that “Behavior is composed of coherent units with beginnings and endings.” A beginning and ending requires two breakpoints, a first and second relative to each other, making each distinct.

Notice how this analysis matches McCloud’s original closure example for an action-to-action panel transition:

These are the two action-defining breakpoints of a larger implied plot structure that could include multiple other panels. The same is true of McCloud’s original moment-to-moment example:

Though here there’s some ambiguity. Both images could be understood as what I term “balance,” the first establishing a status quo (eyes open), and the second the new status quo (eyes closed). If the action is “She closed her eyes,” the first image is necessary to contrast the action-defining change in the second. Technically, the disruption would be implied (her eyes began to close), making the second either a restoration (if we understand this to be the exact moment that they fully shut) or a balance (if this is one of several moments in which here eyes remained closed). The second image could also revise the first into a disruption if it is understood retroactively to mean: “Her eyes were beginning to close” (even if visually imperceptible). We could similarly debate which panel types from Cohn’s visual grammar best fit, but it’s the underlying (and so undrawn) event structure that allows viewers to relate them at all.

McCloud’s original subject-to-subject example is interesting too. The first image alone implies an entire event: “She won the race.” The second image could be understood to take place at the same moment, and so a second manifestation of the same breakpoint. But the learned logic of panel progression implies a movement forward in time too, so the second image becomes a restoration in which the holder of the stopwatch is no longer timing the winning racer. If so the overall action is: “She recorded the racer’s time,” making the first image a disruption (of the status quo of the stopwatch clicking away), followed by the time-keeper responding by moving her thumb until it achieves it’s action-defining position in the second image. If so, both images are separate breakpoints.

So while an infinite number of images could be drawn between breakpoints (because of the Zeno effect), those possible images are constrained by event structure. That means closure is event structure. The only undrawn story content that readers experience through the juxtaposition of two or more images is the minimal content required by a reader’s mental construction of the partially drawn event. While anything could have occurred in the ambiguous lapse of time between any of the above paired images, we understand the images as breakpoints that define a discrete event. Any story content that is not part of that discrete event is not implied and so did not happen–at least not via closure. An artist can always reveal later that something previously unimplied did happen, retconning content that reveals an earlier lie of omission.

Zacks and Tversky discuss this generally as “schema theory”:

“recognizing an event as an instance of a category consists of matching it to a schema stored in memory. Understanding what is going on consists of matching features of the perceptual world to variables in the schema. In ongoing perception, missing information is filled in by references to the patterns of intercorrelation captured by the schema, leading to a fluid interplay of bottom-up and top-down processing.”

Bottom-up means from the eyes to the brain and so applies to the actual images of a comic, and top-down means from long-term memory which is where event structure lives. I’m still vacillating about my own terminology–is my system “plot structure” or “event structure”?–but I think that covers the basics. If I’m right, we have a method for constraining closure and so understanding McCloud’s “invisible art.” I’m tentatively calling it:

Event closure.

Tags: Barbara Tversky, Darren Newtson, Jeffrey Zacks, Neil Cohn, Ray Jackendoff, Todorov