Over the last two weeks, we’ve been looking at visualizations of Piers in space (with maps and regions) and Piers copying in time. Of course, neither of these slices of data exists exclusive of the other, so this week we are going to put both space and time into animated graphics to help us to visualize the movement of Piers through both space and time.
In this map, we have a time-lapse animation of when Piers manuscripts appeared and where. If you click on the map itself, it should bring you to a live graph in which each of these data points successively appears on the maps with a kind of ripple effect helping you to keep track of recently added data points as new ones appear. As with any slice of data visualized, this map too has its limitations.
In order to use the animation feature of CartoDB, it was not possible to maintain the geographic regions (or Polygons) we were using in either our dialect region maps or our GeoJSON from last week. To animate, each manuscript must be represented by a single data point with latitude and longitude coordinates. But, our data is simply not that accurate. Instead, to make use of the animation feature, this graph contains points that have been located within the geographic region delineated on the other maps.
This means that we are indeed looking at arbitrarily placed markers inside the regions, but that doesn’t make them inaccurate. What it makes them is limited. The points have a margin of error that is really rather large, depending upon how narrowly we can locate what region or county or portion of a county a scribe’s dialect points to.
What this map does allow us to do, though, is add back in the manuscripts that are entirely unlocatable in the more precise map (on the right). Manuscripts with dialects that cannot be located appear as the data points lined up to the right of the Island of Britain (yes, in the middle of the North Sea–no, they weren’t made there).
The second limitation in this graph is how time is conveyed. In this particular graph, manuscripts are entered with date ranges.
This means I would either have had to choose a “start date” (earliest possible date the MS was made) or the “end date” (latest possible date it was made) for it to pop up. Since not all date ranges for MSS were equal, this would make interpreting the map’s information somewhat difficult (and possibly the data would be so skewed as to be meaningless).
Instead, this data is presented in a “draw order” (in the far right column) that is assigned to MSS based on all the different categories (and sometimes dates) assigned to them in Schmidt’s Parallel-text Edition. Thus, all the manuscripts from 1390-1400 appear at once , then all the manuscripts dated to ca. 1400 appear at once, and so on and so forth.
This, unlike the bar graphs from earlier, does not require us to lump together things with “ca. 1425,” “1425-1450,” and “1427” dates all together. Instead, they can all be drawn with everything else that belongs to the same date group.
The downside of this method, however, is that it tends to make it look like a whole group of manuscripts dated to the same quarter century all appeared at once. We can, of course, control for this and get manuscript data points to appear in an order that is proportional to their appearance in history, but it requires a second, supplementary map.
In this map we have the same considerations being made for space, but now we are projecting the emergence of manuscript phenomena into less precise times in order to better reflect how and when manuscripts may have appeared over the 160 years of copying history they exhibit.
This time, to account for time, manuscripts were assigned a specific date within their given date range. Choosing this date was much like choosing the point in space to locate the manuscript’s representative data point. It is located within the known date range (or on a given date, if known) and within the known dialect region.
What that means for this map is that while it is not perfectly accurate, what it allows us to see is less the specifics of when and where individual manuscripts were made and instead see a projection of how the existing corpus may have possibly come to be in the state in which we find it today.
This visualization is about representing a complex phenomenon unfolding in history. To keep the realism of historical copying in tact (with its manuscripts that emerge intermittently, take time to travel between copy centers, and take time just to be copied), we actually make our data less accurate.
Wha?! LESS ACCURATE DATA?? On Purpose?? WHY?
Because manuscript copying itself, and Piers in particular, was a messy business. The more we try to codify the data and stick it into nice, neat categories, the less the data represents a real world phenomenon. What it represents is a very narrow slice of data about a real world phenomenon. Which is sometimes good, and necessary. But it is also sometimes too narrow.
So, what we have here is data that is negotiating a kind of middle ground between narrow, accurate data and messy, complex real world phenomenon. Think of it as the scenes of a documentary that are re-shot with actors to re-enact real-world scenarios. The facts are all the same, but lots and lots of little components are left open to the actors’ and director’s interpretation, particularly with reference to the story that is being told. Think of this second graph in particular as a dramatization of the facts of Piers manuscript copying. It represents the best facts as we know them, with some interpretation to make the action more realistic, and in the service of a larger narrative of interpreting the corpus as it came into being in both space and time.