Saying that the Simpsons has declined in quality over time is probably not particularly controversial. A question we might want to ask, then, is what aspects or changes in the show gives us this sense of decline and whether we can quantify it in some way. I think some fans and critics of the show have attributed some of the decline to the show’s reliance on zanier plot lines and less on the grounded, more touching stories that were present in the earlier decade of the show. (Some examples from the classic era like Lisa’s Substitute, Mother Simpson, Duffless, A Fish Called Selma, for example, come to mind). Here are some relevant quotes from commentators/critics:
“Episodes that once would have ended with Homer and Marge bicycling into the sunset (perhaps while Bart gagged in the background) now end with Homer blowing a tranquilizer dart into Marge’s neck. The show’s still funny, but it hasn’t been touching in years.”
“The central tragedy of The Simpsons is that it has gone from commanding attention to merely being attention seeking. It began by proving that cartoon characters don’t have to be caricatures; they can be invested with real emotions. Now the show has in essence fermented into a limp parody of itself.”
“...The Simpsons is not the show it was – in every sense of the phrase. Take an episode at random from the mid 1990s and compare it to one from a decade later. Not only would the decline in quality be painfully evident, but only in the names, appearances and voices of the principal characters would you find any kind of continuity. Come to that, you’d be lucky to find continuity itself in the later example – let alone the immaculate narrative structure, emotional depth or roundedness of character that would invariably grace its older relative. “
A way to evaluate these claims is to potentially examine whether the general tone or sentiment of the Simpsons has indeed changed over time. That is, do the earlier seasons have quantifiably more of a sweet, touching tone (what we might think of as more of a “positive” sentiment) than the new seasons? To answer this question, I performed a sentiment analysis on the Simpsons corpus that I had collected previously for my Flanderization of Homer analysis.
To perform the analysis, I used the neat R package
Tidytext (Silge and Robinson, 2016). A sentiment score was essentially calculated for each line of dialogue by adding up the sentiment score assigned to each word in the sentence. The sentiment score was determined for each word by checking it against a dictionary that contained a list of words that had pre-existing scores, ranging from -5 (negative sentiment) to 5 (positive sentiment). These types of dictionaries are usually complied via methods such as crowd-sourcing (e.g., asking for judgments via Amazon Mechanical Turk) or from the authors’ own coding. And while the one I used scored words from -5 to 5, others can simply score words in a binary fashion (e.g., yes/no “positive”). For more information on how I performed the analysis (e.g., managing instances of negation), please visit my github repo.
So what did we find regarding the change in sentiment over the course of the 26 seasons of the Simpsons? Here’s the plot of the average sentiment score for each season. The red dotted lines indicate the season in which there was a change in showrunner.
As we can see, overall, the show has a pretty positive sentiment (e.g., it never falls below zero). However, there does seem to be a steep drop in positive sentiment after season 3. I’m not quite sure yet what might account for this change. Initially I had wondered if it was because a change of showrunners, but the switch to Al Jean and Mike Reiss (from Matt Groening, James L. Brooks, and Sam Simon) occurs between season 2 and 3. If other people might know more about Simpsons history that might account for such a change (e.g., large exodus of writers), I’m open to ideas!
Going back to our original question, is there a discernible difference in sentiment between the earlier and newer seasons of the Simpsons? Let’s just say that we define the earlier, classic era of the Simpsons as seasons 1-9 and the newer, modern era of the Simpsons as seasons (13-26).¹ If we look at the average sentiment score for these two eras, there doesn’t seem to be a large difference (56.35 for the classic era and 48.88 for the modern era). We can also see in both eras, the amount of positive sentiment can be somewhat erratic season to season. Aside from the dips in sentiment in season 18 and season 22, both eras have a similar range in their sentiment scores. Lastly, if I were to run a simple statistical test (i.e., Mann-Whitney U Test) on the sentiment score for the two eras, we actually don’t find any significant difference between the two eras. So overall, in looking at just the positive/negative sentiment across the seasons, it does not seem to suggest that the later seasons are drastically less “positive” than the earlier seasons.
However, this type of sentiment analysis is not perfect and can still miss more subtle emotions expressed in a text. Moreover, humor is still hard to capture and quantify; it might be that if we focused on the sentiments of just the jokes/gags, we can better reveal the differences. This is all to say that, while broadly speaking the positive/negative sentiment does not seem to have changed much across the seasons, this does not suggest that fans’ and critics’ critique that the Simpsons has become zanier and less grounded is inaccurate. Rather, we just need finer tools to be able to better quantify this intuition.
¹ I am aware what season actually marks the end of the classic, golden age of the Simpsons can be somewhat contentious but I have chosen the cut off at season 9 given that I feel that while season 10 had some episodes that feel like “classic” Simpsons (e.g., Lard of the Dance), I feel like the number of those types of episodes are fewer. Additionally, given that there is some argument to when the classic era ended, I chose season 13 as the start of the modern era, since I think it’s far enough from the seasons that people might still consider “classic.”
Silge, J., & Robinson, D. (2016). Tidytext: Text mining and analysis using tidy data principles in R. The Journal of Open Source Software, 1 (3). doi:10.21105/joss.00037.