After reading the Manovich and Douglass piece “Visualizing Temporal Patterns in Visual Media” (and Julieta’s blog post) the data set of 35 historical paintings struck me as odd. I understand that the authors chose paintings to exemplify a data representation, but this seemed to bring a major issue of temporal representation to my mind.
It seems like if you had an agenda it could be easy to use a temporal graph to prove your point very easily by choosing the data that you use and excluding other factors. Manovich and Douglass make a point of explaining different historical events in the USA, Russian and France to contextualize the graph on movie clip lengths, because they wanted to help put the data into historical context, but an author would not need to do this. It seems that putting temporal data into a graph and comparing different items over time can almost be a way to take groups (or art, movies, etc.) out of their historical context because without a thorough explanation of the data, the context is not easily read in the graphs. Overall, when time is included as a factor for graphical explanation (which strives to be authoritative and objective) there are many factors outside of the data chosen, mainly historical context, that need to be included in an explanation of the data otherwise the graph can be misleading.
To me this all ties into the concept of choosing a data set. Temporal representations in general seem to be problematic with large data sets. For example, if we want to create a graph to show patterns in 19th and 20th century art it would be very difficult to find a fair and balanced data set. One could make a graph or chart that would be very interesting, but from a theoretical standpoint it would have many issues. For example, would every piece of art of every country need to be found? What about groups of people who primarily make sculptures rather than paintings? What political constraints do artists have in one country that is different than others? And many more questions.
However, if we chose a small dataset, many of these issues are resolved. For example, if one took only the movies of Woody Allen and Martin Scorsese, it would be easy to objectively compare their careers and find patterns within their careers. With only ~50 movies each it would be much less difficult to create a problem free dataset and a researcher would be able to compare how often they make black and white vs. color movies, how long their cuts are, and if that changed over time, etc.