Identifying Shot Scale with Artificial Intelligence
Shot Scale in Classical Hollywood: A Historical Analysis
After a long hiatus due to starting a new job, this post is the fourth in a series. Subsequent posts will be released every other Sunday. The code for this project can be found on my GitHub. The Google Colab notebook used to analyze the data can be found here.
The “scale” of an image is integral to how we think about movies. Words like “close-up” or “establishing shot” are used so habitually that it’s hard to imagine life without them. But how has shot scale changed over time? Wouldn’t it be nice to have some idea of its history?
Because the scale of a frame is highly correlated to the size of the most prominent face contained within it, I ran a facial detection algorithm over a representative sample of more than 4,000 Hollywood movies produced between 1915 and 1970. Taking an average for each film, and then averaging those by year, can show us how shot scale changes historically.
Film scholars have generally argued that the arrival of sound (circa 1927–1928) severely limited the cinematic possibilities developed during the silent era (notably lampooned by Singin’ in the Rain [1952]).
In particular, the realities of early sound recording meant the camera had to be largely stationary, constraining the meticulous camera movements of the late silent era. While not always mentioned explicitly, the enormous close-ups were also lost. If the arrival of sound really did affect film style, specifically the frequency of close-ups, we should be able to see it in the data, which I’ve plotted below.
Just looking at the plot tells a story. We see a remarkable rise in shot scale followed by a sharp decline. The shift is so abrupt, dropping by about half in a period of only three years, that it cries out for an explanation. Since this decline occurred after the introduction of sound, the data sure make a compelling case that the arrival of sound had a strong effect on shot scale.
Perhaps more interesting than the sudden drop in relative face size is that it remained around this level for most of what we call “classical” Hollywood. It perhaps counters claims made by many scholars that the effects of the transition to sound temporary and basically resolved by the mid-1930s. The data suggests that, at least regarding shot scale, the changes brought on by the arrival of sound were much more lasting than we may have thought. It was not until 1970, over 40 years after the first sound film, that shot scale again reached the level of the late silent era.
This rather straightforward analysis demonstrates the possibilities that AI has for media analytics. In this case, the data helps confirm the transition to sound as the single most impactful event on shot scale during “classical” Hollywood. It empirically supports the consensus view that the transition to sound led to drastic changes in film style.
Shot Scale and the Demise of Old Hollywood
The effect of sound is remarkably intense, but it’s not the only significant development. In some ways, the rapid increase in face size after 1960 is the more interesting phenomenon. Because the 1960s were a time of great upheaval for Hollywood — the “studio system” had already begun disintegrating in the 1950s — the increase in shot scale can get lost in the chaos. However, the intensity of the transformation is striking. Face size nearly doubled during the 1960s, a sea change analogous to the arrival of sound. Rarely has the scale of this change been acknowledged, however, possibly because the change occurred gradually enough to go relatively unnoticed.
What could possibly explain such a rapid increase in shot scale? Answering that question is well beyond the scope of this post. Some scholars have already proposed explanations that may be relevant here, but few address the problem directly.
Widescreen
According to some, the introduction of widescreen led to more close-ups. More room on the sides of the image meant more space to fill. Also, the lateral space meant tight close-ups were “balanced” by the negative space of the more spacious screen. In this argument, the new widescreen technology disrupted aesthetic norms, thus prompting a shift in film style.
As we can see in the data, if the above argument is correct, the aesthetic transformation lagged behind the transition to widescreen, which began in 1953 and proceeded rapidly. Most films were using the new widescreen technology just a few years later. It was not until 1960, however, that face size started to increase. Because of the delay, it’s difficult to interpret any causal relationship between widescreen and increasing shot scale.
However, there may be an intriguing way to test this theory. We could compare widescreen to “flat” films during the same time period. If widescreen genuinely drove the increase in shot scale, we would expect significant differences between the shot scale of wide vs. standard films. If no such difference exists, that would seriously undermine the widescreen theory. I may explore this in a later post.
Television
Another common explanation for the increasing frequency of close-ups toward the end of the studio system identifies television as the culprit. The reduced frame of the TV, the thinking goes, required more close-ups, which then influenced film style. The chart above also undermines this explanation since television was already a powerful force by the mid-1950s. If it did have an effect, there is again a “lag time” within the relationship. It may be possible to test this empirically. We would need a representative sample of television over the same time period to compare against, however, but that is a project of its own.
Knowing how shot scale changed after 1970 would be incredibly helpful. It might give us insight into why earlier changes occurred. But again, this is its own project. For reasons I’ve discussed earlier, it is much harder to collect films after 1970 because of copyrights. Older films are more easily attainable because they are more frequently in the public domain, or no one cares about the copyright.
Next Steps
As we’ve seen, the chart above tells us how shot scale changed but not why. What factors drove those changes? I’ve reviewed a few possible explanations in this post, but I’ll need to dive deeper into the data to evaluate them. Subsequent posts will examine shot scale in relation to factors like genre, widescreen, color, production company, and director. Closely parsing the data will hopefully reveal which factors are most responsible for historical changes in shot scale. But it can also tell us other things, such as which directors frequently used close-ups and which did not.
Previous Posts
