In this section we investigate the dynamics of memes traveling through movie history and the findings we were able to make by visualizing this model of culture.
The visualizations are based on data from IMDB (The Internet Movie database) which is an Online movie community with 42 million users. IMDB collects all kinds of information about movies, one part is a collection of references between movies. In the database these references are stored under the menu point ‘connections’. There are nine different reference types (alternate language version of, edited from, features, follows, references, remake of, spin off from, spoofs & version of ). For the visualization we focused on the category ‘reference‘. The dataset contains 119,135 such connections from 42,571 movies. To be able to responsively visualize the data in a web browser we selected the 3,000 most connected movies by in-degree and their 10,000 connections to each other.
To give a better impression what these references look like here two examples for the movie ‘Star Wars’ from 1977: The robot C3PO was modeled after the robot from the movie Metropolis (1927). C3PO used a line from the movie Seven Samurai from 1954: „It seems we are made to suffer. It‘s our lot in life”. But not only Star Wars referenced other movies they also got referenced in many occasions. The line „May the Force be with you.” was referenced by movies like Barney Miller: Quo Vadis? (1978), The Big Fix (1978), Sledge Hammer!: They Shoot Hammers, Don‘t They? (1986), Beverly Hills Cop II (1987) and many others. We mapped this copying, transforming & combining of ideas over time in a network structure. While positioning the nodes of the graph on the y-axis on a time-line from 1900 to 2015 the x-axis has different versions like degree (number of connections), modularity (closeness of the nodes to each other) or genres the movies are ordered into. The different visualizations can be filtered by year, degree and by the characteristics of each graphic like genre or modularity. These graphics give a summery over the inspiration in the history of cinema.
One well visible artefact of the created ‘culture.graphs’ is the color gradient within the graphics from the bottom in blue to the red top.
The dense red area at the top of the graphic which starts around the 1980’s is know in film studies as the postmodern cinema. An era in which movies strongly cite the style and stories of past movies (Hill, J., 1998, Kramer, P., 1998, Distelmeyer, J., 2008).
The three area charts compare the number of movies created for each year (in blue), the years in which movies got referenced (in yellow) and in red the years in which movies referenced other movies. We can see that while movie production approximately doubled over the last 50 years, the references made quadrupled in the same time period. While the area charts show the statistical view over time in a very clear way, the ‘culture.graphs’ connect the macro view the mirco view and show the references of each movie that give rise to the larger pattern of the graphic. It lets the viewer ‘zoom in’ and explore how the phenomena arises and which movies play a key role in this development.
While the rise of the postmodern cinema is a dynamic process that got produced on the mirco level but only exist as a phenomena on the macro level there are also patterns that can be found and analyzed on the level of each movie.
One difference that can be seen between the individual movies is the citation patterns they have. The movie ‘The Wizard of Oz’ got released in 1939 and for 30 years until the beginning of the 1970’s the movie nearly did not received any references. In comparison to Alfred Hitchcocks ‘Psycho’ which got released in 1960 and got referenced right after its release. But even with the delay of 30 years ‘The Wizard of Oz’ got more references than any other movie, accept Star Wars. Figure 8. 6 shows the references of the most cited movie in the IMDB, Star Wars. One interesting pattern that can be seen is the ‘burstiness’ of the movie. Around 1990 there is a gap of 4 years in which ‘Star Wars’ did revived much less references than before or after. This is a well studied phenomenon in network theory (Barabási, A.-L., 2005, Barabási, A.-L., 2010.). Interestingly ‘Star Wars: Episode V - The Empire Strikes Back’ shows a similar pattern just not as defines as in the first movie. This could point to the assumptions that the ‘burstiness’ has something to do with the series of ‘Star Wars’ movies.
Other patterns question the validity of the dataset. The movie ‘Cannibal Holocaust’ for example has a large number of references. This might not only be based on the high number of references the director made but also influenced by the strong fan base behind this movie which wrote down every video case that is visible in the movie. The meaning of the word ‘reference’ is very much up to the creators of this lists.
The first question we ask ourself about movie genres was about the rise and the fall of different genres. The area chart in figure 8.17 shows the number of movies in each genres over the last 100 years. The different genres are ordered on the y Axis by the number of movies produced per genre. Such a view shows the very short time-span of ‘film-noir’ between the 1940 and 1960, the era of ‘Western’ movies in the 1930 to 1950 as well as certain peaks of genres like ‘Horror’ around 1990 or ‘mystery’ between 1930 and 1950. What this visualization does not show is how ideas got adapted from this different genres. Only because many ‘mystery’ movies were made between the 1930 and 1950 does not mean that this movies were influential and had impact when it comes to the adaptation of ideas. The first exploration to make this idea exchange visible was by adding a filtering method for the two graphs ‘by connectedness’ and ‘by community’. Once filtered the movies by ‘mystery’ in the ‘by connectedness’ visualization three movies were standing out in 1940 (Rebecca) and 1941 (The Maltese Falcon & Citizen Kane). Interestingly while the total number of mystery movies goes down after the 1950s there were still very high referenced movies in the time after 1950. Like ‘Vertigo’ (1958), ‘Psycho’ (1960), ‘Rosemary’s Baby’ (1968), ‘2001: A Space Odyssey’ (1968) or ‘The Shining’ (1980).
The second most produced genre ‘Comedy’ is not represented by any of the highly cited movies. The most cited ‘Comedy’ movie is ‘Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb’ from 1964 with 147 references. The second least represented genre when it comes to highly referential movies is ‘Romance’ with the movie ‘Casablanca’ from 1942 with 321 connections. This indicates that the number of produced movies within a genre does not effect the influential ‘blockbusters’ of that genre.
The graph ‘by community’ was much harder to interpret than the arrangement by number of connections (‘by connectedness’). While genres often created communities (visible when circles are arranged in one line vertically) there were also many movies who did not stick to this pattern. Who were standing on their own unconnected to the rest of the circles. The visualization indicated this and the interactive version lets the viewer explore the movies that are within one community. One can find movies series that are located on the same line like the British ‘Carry On’ franchise consisting out of 31 movies between 1958 and 1992 or the a series of ‘Tarzan’ movies between 1939 and 1955 in the action genre. Many of these clusters become visible but the interactions between different genres still stayed hidden.
The visualization ‘by community’ leaded to the question how movie genres pass on ideas in between each other. How interconnected or separated are genres from each other? In other words is the human categorization of movies into genres reflected in the citation patterns of memes between them?
Do genres stay on there own and only adapt ideas from their own surroundings? Or are there highly connected and reference each other? And if so which genres are connected to each other? This questions could not be answered by visualizing only single genres as we did in the ‘by connectedness’ and ‘by community’ graphs. The visualization to see the connections must show interaction between different genres.
To answer some of the questions we developed the visualization ‘by genre’. Here the interaction between different genres becomes visible on a macro scale. Such a view does not show the micro and the macro perspective in one view but to get an overview of genre diffusion it was the right level of aggregation.
The first observation that can be made is that the interaction in meme transmission between genres is high. If this would not be the case there would be a diagonal red line going through the visualization. But in reality genres are highly connected to each other. The genre who references the most is ‘Comedy’ this might have to do with movies like ‘Scary Movie’ which are known to reference many movies and reference out of a really wide range of movies. While many genres reference and get referenced to a similar degree there are some disbalances. ‘Film-Noir’ nearly did not referenced any movies but ‘Film-Noir’ got referenced by many genres. Same can be said about ‘Western’ both genres which had their period centuries before the post-modern cinema. On the contrary ‘Documentary’ referenced many movies but does not get referenced many times. Also by looking at individual combinations interesting patterns appear. ‘Action’ is referencing ‘Adventure’ 1125 times while ‘Adventure’ only references ‘Action’ 757 times. Even when there are much more action movies produced than adventure movies. Adventure seams to be more influential in idea transmission.
An interesting question that we adapted from biology deals with the idea of ‘fitness landscapes’. In biology the term is used to understand the relationship between genotypes and reproductive success (Kauffman, S.A. & Johnsen, S., 1991, Gavrilets, S., 2004). The basic idea is that once an Organism found a successful path to survive and reproduce others will follow this path and it becomes a ‘peak’ in the landscape of all possible Organisms. This peaks might become what we than call a species.
We adapted this model to our cultural data of movie references by looking at the genre interactions over time. Are there clusters or ‘fitness landscapes’ visible that are not created by one movie but by a group of movies in a very specific time-frame?
Can we find patterns of genre connections over time that show certain eras in which entire genres inclined in the idea adaptation?
The image above shows the 7 most highly connected genres and their interactions over time. The idea of adapting ‘fitness landscapes’ of cultural inspiration patterns over time is highly speculative. The here presented findings are only indications towards an interesting field of study.
One pattern that can be seen are areas of genre interactions over time in very specific time frames that can not be produced by a single movie but must rather be ‘movements’ or trends within the idea exchange throughout genres. Multiple of such dense areas are visible in different genre combinations. The image below shows the idea exchange between crime movies and their adaptation in ‘comedy’ and ‘action’ movies. Two areas show clusters over multiple years at the end of the 1970 and at the end of 1990. The ‘crime’ movies produced in these times were much more influential than ‘crime’ movies from other times.
A similar pattern can also be seen in the genre ‘horror’. While the horizontal lines which are visible for example in 1931 are most probably produced by single highly influential movies in this case by the movie ‘Frankenstein’. The dense areas on the other hand indicate towards trends of the entire genre. The strongest trends can be seen between the 1980 and 1990 were multiple genres adapted ideas from horror movies.