Studying the Influence of Films on the Style of Roger Ebert

The role of the critic has always existed in a limbo. They are simultaneously the audience for one text and the creator of their own. They produce and consume in equal measure. Here we will try and explore these two roles and attempt to estimate to influence of one role over the other.

So we will examine the work of Roger Ebert. Ebert is both an example of the critic as a creator and advocate for the argument that the critic is fundamentally an audience member.

Ebert has a vast portfolio of work and any original creative work of his has been smothered by the success of his movie reviews. Most of Ebert’s reviews were written for a newspaper. The intention was that people would quickly read them to decide whether or not they would go to the cinema. They were made to be quick and easily consumed.

These reviews proved so popular that they were collected in books and eventually, decades after many were written, published online. Clearly there was a considerable interest in reading his reviews for their entertainment value.

However this takes us away from the critic as an audience member. In an interview with George Hickenlooper, published in Hichenlooper’s collection of interviews Reel Conversations Candid Interviews with Films Foremost Directors and Critics Ebert himself said that

“The notion that the critic is supposed to be able to direct or produce or act or write is completely apart from the function of a critic, which is to be an ideal member of the audience.” (Hickenlooper, 365 – 366).

Clearly Ebert, and popular film criticism as a whole, is a prime example of the consuming and producing critic but we still need to better understand this discourse. We can study the effect of being a member of the audience affected him by studying the style of his reviews and how they were changed by the films he watched.

We will now use computational stylistics to test out a hypothesis. The hypothesis is that the film being reviewed would have a noticeable impact on the style of the writer and that factors such as director, genre and star rating of the film would all contribute to this influence on style.

If we can gather evidence to support this hypothesis then we will have hard data to back up the claim that the critic as a consumer and the critic as a producer are both integrally linked to one another.

I also thought that testing this hypothesis would open up further avenues of study. As I have discussed in an earlier blog post, Marshall McLuhan wrote that “the Medium is the message”. Basically that a text could be studied by the effect it had on the audience. So applying computational stylistics to film criticism may allow us to explore films themselves. The more we understand the effect on the audience than the more we understand the films.

Therefore the tests I ran could determine whether or not a larger more extensive study of the subject would be worth doing.

The first test was to see if individual film critics had their own unique style. It was possible that the common function of the writing, as a consumer guide, would create a homogenous writing style and overwhelm any individual writing style. I took four different professional movie critics, Ann Hornaday, Christy Lemire, Michael Phillips and Roger Ebert, these critics were chosen simply for their substantial and accessible portfolio, and I took four reviews from each and made a simple cluster analysis.

Cluster Analysis of Critics

Cluster Analysis of Critics

Hornaday and Lemire seem to be quite similar but the results clearly show that both Ebert and Philips have distinctive writing styles. This corroborates with Ebert’s reviews being written for their entertainment purposes and not just as a guide for consumers. This result was predictable but essential for a thorough and satisfying study.

The next goal was to test the extent that the film being reviewed would influence the writers. I chose four films that all four critics reviewed, Argo, Skyfall, Tower Heist and the Hunger Games, and compared the reviews that they had written about these films. We have already established that the writers have their own style. If the reviews for the same films shared more similarities in style than the reviews of different films than this would clearly indicate influence from those films on the critic’s style.

Cluster Analysis

Cluster Analysis

The results first appear to be quite significant. You can see here that the movies have all been grouped together by style. However this would suggest that the movie being reviewed is more influential on the style of the review than the reviewer themselves and this is clearly preposterous.

Studying the frequency tables showed that words such as Argo, hunger and games were occurring with great frequency and that this influenced the style greatly. The word cloud found at this link provides greater insight into the possible problems.

The most often reoccurring words include Affleck, Argo, Bond, games, heist, hunger, Katniss, Skyfall and tower as well as many other film titles and the names of main characters, cast and crew. The problem here is that each film review contains a summary of the plot. The reviews themselves are also quite short. The average word count for these sixteen reviews is 729.5 words per review. The reviews all have two to four paragraphs, which can account for almost half the total word count, dedicated to summarising the plot.

These results might not be outright dismissed. There are still enough function words irrelevant to the plot that can be used for stylistics. However it does mean that the results are compromised and that without significant cleaning being done to the texts we cannot determine how much of the style is based on common content rather than common style.

From this point on all of the reviews studied were written by Roger Ebert. By focusing on a single writer we will not encounter the difficulties of the last test. We can also use a single critic to better isolate films that have a single factor in common. The rests of the tests were designed to study the effects of director, film genre and star rating on the style of the review.

To study the influence of directors on the style six directors were chosen. These directors, Martin Scorsese, Alfred Hitchcock, Spike Lee, Kathryn Bigelow, Quentin Tarantino and Takeshi Kitano, were selected for several reasons. They are all well known for a distinct directorial style, they are all well respected and, as a group, they represent a diverse range of backgrounds with regards to nationality, ethnicity and gender.  Therefore it is the directors themselves that are the most significant difference between these films.

Cluster Analysis of Directors

Cluster Analysis of Directors

You can see here that the style of the review was not largely affected by the director. This outcome is of particular with regards to Ebert because he was a well-known proponent of the auteur theory. If directors did not influence his style than it is likely they would not impact any critic’s style.

However the films that are being paired often have vastly different plots such as Inglorious Basterds and Strange Days, or Do the Right Thing and Psycho. This does indicate that there is an influence on the writing when repeating names and film titles are not corrupting the results.

After this we would study the impact of genre on the reviews. The next test split sixteen films into four different genre groups; with these genres being romance, action, horror and family.

Cluster Analysis of Genre

The results showed that most of the film reviews were not grouped together according to genre. In fact only two film reviews were paired together that shared a genre.

In fact more of the film reviews paired together are for films that could not be more different. The Raid: Redemption is a brutal Indonesion martial arts film while One Day is a saccharine British romance. There are similar disparities between Sinister and Ponyo as well as the Back Up Plan and I know what you Did Last Summer.

However these sixteen films can also be split into four other groups according the number of stars they received from Roger Ebert. One film from each genre got one star; another from the genre got two stars and so on.

Cluster Analysis of Star Rating

Cluster Analysis of Star Rating

When the films are split like this the results show that there is some influence from the film on the review.

There are two conclusions that can be taken from the results of this test.

The first is that the mood of the writer is a major influence on their style. Film reviews are written in a brief time period shortly after the film itself is watched. If the film put Ebert in a good mood, which would certainly happen if he enjoyed it, than this could influence his writing style in a very particular way.

The second is that Ebert may also have changed his style on purpose because he felt better films deserved better reviews. A great film, such as one that got four stars, would deserve an intelligent thoughtful review. However a bad film, such as one with only one star, wouldn’t need this treatment and could be filled with jokes about the film’s poor quality. This conclusion is supported by the cluster analysis as it shows that it is the one and four star reviews that have the most distinct styles.

Looking at the study as a whole we can come to the following conclusions.

Film reviews do not readily lend themselves to computational stylistics. The effect of pollution of plot summaries can be seen above. Applying stylistics to a film review can only be done if the text is carefully cleaned beforehand or if the test only uses a single reviewer. This makes the work slow, tedious and limiting.

There is evidence for the hypothesis although it is not significant. The only contributing factor that could be isolated is the influence of the film’s ratings.

There is certainly a case for the critic as an overlap of dichotomies. This paradox offers the potential for a greater understanding of literature but it seems this fusion will remain something of a mystery for the time being.



McLuhan, Marshall. Understanding Media: The Extensions of Man. New York: McGraw Hill, 1964.

Hickenlooper, George. Reel Conversations: Candid Interviews with Films Foremost Directors and Critics. USA: Carol Publishing, 1991.

Reviews taken from

Cluster Analysis created using R and R Studio.

Word Cloud created using Wordle.


Leave a comment

Filed under Uncategorized

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s