The Flanderization of Homer Simpson

Flanderization” refers to the notion that as a show progresses, certain character traits often become exaggerated or become one of the sole defining features of the character. In the Simpsons fandom, the “stupidification” of Homer Simpsons is an oft-cited example.

Now, this got me curious to whether a) this was indeed the case and b) if so, how we might quantify Homer becoming less intelligent, i.e., Flanderizing. As a huge Simpsons fan, I intuited that Homer has indeed become less intelligent across the seasons. As a language acquisitionist by trade (alas, if only my extensive Simpsons trivia could get food onto the table for me and my cats….mostly my cats), I wondered if I could capture Homer’s Flanderization in his language — that is, could different linguistic measures reveal that over time, Homer has become less and less intelligent?

For more information on the scripts and data set themselves, please feel free to visit the github repo I’ve set up. Essentially, what I ended up doing was scraping the FXX Simpsons World website for the available transcripts/scripts of all available episodes of the Simpsons (it appears that it is available only up to Season 26). After that, I performed the following analyses of Homer’s dialogue:

Mean length of utterance (MLU)

MLU is a common measure used in language acquisition research that is thought to reflect children’s linguistic complexity and productivity. It is computed by first calculating the number of morphemes (i.e., the smallest meaningful unit) produced by a child in a speech sample and then dividing that by the total number of utterances produced. This may not be completely appropriate to use for adult speech, but just as a first pass analysis, may nonetheless provide some insight. I ended up calculating the MLU for each of the Simpsons family member. I examined the other family members because they can serve as controls. For example, we don’t typically talk about Lisa’s Flanderization in terms of her intelligence. Thus, we might expect Lisa’s linguistic complexity to stay relatively the same across seasons. It might be also interesting to just compare Homer’s linguistic complexity to the other family members to investigate whether he is, at least linguistically speaking, less intelligent.

plotting-mlu-1

As one can see, there actually isn’t a very big change in Homer’s MLU — there appears to be a small dip between Seasons 9-13, but it generally hovers between 6 to 7 morphemes per  utterance. He also appears to cluster around the other family member’s MLU as well.

Other measures of linguistic complexity

I had previously mentioned that MLU may not be an entirely appropriate measure for adults. Thus, I calculated four additional linguistic measures:

  1. Flesch reading-ease Score indicates how difficult it is to understand a passage or text. The lower the score, the more difficult/complex it is. To learn more, see this wikipedia entry
  2. Type-Token ratio, where higher TTR indicates larger amount of lexical variation. It is calculated by dividing the number of unique words (types) by the total number of words (tokens).
  3. Average Word Length
  4. Average number of syllables per word

I created these four measures for each of the family members. Here we present Homer’s analysis first:

homer-complexity-graph-1

Again, we don’t see huge changes across seasons. If anything, it looks as if Homer appears a little more linguistically complex across time. For example, his Flesch-reading score actually decreases from 94.98 to 90.4. Recall for the Flesch-reading score, the lower the score the more complex the text / passage is. Additionally, his lexical diversity and word length both seem to slightly increase.

Finally, here is how Homer compares to the other family members (for each members individual plots, please visit the github repo):

complexity-all-members-1

Similar to the MLU measure, Homer appears to pattern similarly to the other family members in the four measures (with the exception of maybe Lisa, who appears to have, for example, a slightly higher range of type-token ratio than Homer).

So is this all to say that the intuition of Homer’s Flanderization may be incorrect? I don’t believe this is the case. There are many other ways in which we might capture changes in Homer’s intelligence across the seasons. For example, does he become less rational or possess less world knowledge across the seasons? With his language, as I’ve measured it at least, Homer does not appear to have dramatically changed throughout the years. There are probably also additional language measures that can better capture Homer’s change across the years — but I suspect that these are a little harder to automatize, for now.