There has been much speculation about Trump’s mental facilities from the press. Is it possible our President is in the beginning stages of Alzheimers or another form of dementia?
The seemingly stark difference in speaking style between a younger Trump and today’s Trump have been mentioned many times.
Take a look:
But, is there a way to actually try to measure quantitatively any changes in speech over the 40+ years Trump has been giving public interviews? The answer is yes, but the results may surprise you.
Measuring Trump’s Speech Sophistication Over Time
We began by collecting accurate transcripts from more than 20 individual Trump interviews (not pre-written speeches), roughly one from every two years from 1980-Present. We then used TextStat and LexicalRichness, two python libraries to generate 17 lexical analysis metrics, each providing different insights into changes in Trump’s speech patterns over time. We measured:
- Avg Syllable – Average number of syllables per word in given text sample.
- Eeading ease – A measures how complex a text is. The lower the score, the more difficult the text is to read. The Flesch readability score uses the average length of your sentences (measured by the number of words) and the average number of syllables per word in an equation to calculate the reading ease.
- Gunning Fog – An index that estimates the years of formal education a person needs to understand the text on the first reading. Uses syllables per word.
- Smog – Another index that estimates the years of formal education a person needs to understand the text on the first reading. Uses syllables per word.
- Readability – Another index that estimates the years of formal education a person needs to understand the text on the first reading. Uses characters per word.
- Coleman Liau – Another index that estimates the years of formal education a person needs to understand the text on the first reading. Uses characters per word.
- Linsear – A readability index developed by the US military specifically designed to calculate the United States grade level of a text sample based on sentence length and the number of words used that have three or more syllables.
- Dale Chall – A readability test that provides a numeric gauge of the comprehension difficulty that readers come upon when reading a text. It uses a list of 3000 words that groups of fourth-grade American students could reliably understand, considering any word not on that list to be difficult.
- Unique Terms – The raw number of unique terms in a given text sample.
- Type Token Ratio – A measure of lexical richness, or variety in vocabulary. Type-token ration (TTR),tells you if the subject use the same words over and over, or does s/he use a variety of different words to communicate? A type-token ratio (TTR) is the total number of UNIQUE words (types) divided by the total number of words (tokens) in a given segment of language.
- Root TTR – Alternative, possibly more accurate TTR measurement.
- Corrected TTR – Alternative, possibly more accurate TTR measurement.
- Mean Segmental TTR – Alternative, possibly more accurate TTR measurement.
- Moving Avg TTR – Alternative, possibly more accurate TTR measurement.
- Textual Lexical Diversity – A measurement of the breadth and variety of the vocabulary used in a piece of writing.
- HHD – Another measure of lexical diversity
These metrics were calculated for each 1,000 word interview sample, and then grouped into two buckets: 1980-2002 (early Trump), and 2003-Present(late Trump). The difference (change) was calculated for each metric with the results shown below:
Trump – Average Syllables Per Word – A Measure of Vocabulary Decline
Though Trump has been shown to have among the lowest number of average syllables per word of any recent president, having vocabulary without many longer more sophisticated words is not an indication of dementia. Perhaps surprisingly, Trump’s avg number of syllables per word, did not decline in a statistically significant way over the 40 years interview text of samples we looked at.
Trump – Unique Words – A Measure of Vocabulary Size
Surprisingly, research has shown at least a weak correlation between mild cognitive impairment, as is seen in early dementia or Alzheimer’s, with increases in the overall size of spoken vocabulary, as measured by unique words in spoken transcripts. Below we see the average number of unique word used in a 1,000 word sample of text taken from Trump interview transcripts. As you can see there is considerable fluctuation, but overall, we measured an 5% decline when comparing the averages for young and old Trump( 1980-2002 vs. 2002-present. ) While not conclusive, it’s an interesting result running counter to arguments for Trump having early evidence of dementia or Alzheimer’s. In those with early cognitive impairment, it is thought that individuals:
“may be engaging in behavior to conceal their deteriorating cognition, thereby leading to a temporary increase in their active spoken vocabulary. ” Aramaki et. all
Trump – Lexical Diversity – A Measure of Vocabulary Usage
Lexical diversity is an aspect of measuring lexical richness and is a ratio measure of unique words vs. all words used in a sample. Studies have shown that decreases in lexical diversity are positively correlated with cognitive decline. In this case, we see this evidence reflected in the data from Trump’s interview transcripts over time. Here we see an 8% decline when comparing the averages for young and old Trump( 1980-2002 vs. 2002-present. )
Additional Trump Analysis – Readability
Most of the additional lexical analysis metrics we measured were associated with “readability.” These metrics are listed above in the second paragraph. There are many different approaches to measuring the readability or grade level of text samples, but on a whole, we found that Trump’s speech declined on nearly ever one.
Comparing Trump to Bernie Sanders
Given there seem to be some indicators of cognitive decline on some measures, we thought it would be interesting to do a similar analysis of Bernie Sanders, a similarly aged candidate, also with 40+ years of interview transcripts available. Here is what we found.
Sanders – Average Syllables per Word Used Over Time
Similarly to Trump, we see a decline of about 3%, when comparing young and old Sanders( 1980-2002 vs. 2002-present.) Just as with Trump, we would actually expect to see a significant increase over time with this metric. As mentioned, previous studies have found a correlation between increases in average syllables per word and mild cognitive decline.
Sanders – Unique Terms Over Time
With Sanders, we see an even greater change in number of unique terms used over time. In this case, there was a 7% decrease when comparing young and old Sanders( 1980-2002 vs. 2002-present.) According to research, an increase in unique terms is counterintuitivly correllated with cognitive decline. So in this case, as with Trump, we don’t see evidence for cognitive decline based on this metric.
Sanders – Lexical Diversity Over Time
Similarly to Trump, we saw lexical diversity for Bernie Sanders decline almost 9% when grouping young (1980-2002) and old (2003-present) Sanders.
Additional Sanders Analysis – Readability
As with Trump, we also looked at the 14 other lexical analysis scores available to us with the two Python libraries mentioned at the start of the article. Interestingly, we didn’t see nearly across the board declines on these other metrics with Sanders as we did with Trump. The chart below again compares young vs. old Sanders, and looks at the percent change between the readability metrics found for these two groupings of interview text transcripts.
While it may seem “obvious” to some of us that Trump isnt the man he used to be in terms of cognitive or speaking ability, sussing out proof of this decline is harder than one might imagine. Although there are some clues, they are far from conclusive. By comparing Trump with Sanders, we see similarities in changes over time, perhaps with Trump’s changes being mildly more pronounced. Whether this means both Trump and Sanders have declining cognitive abilities is an open question. For now, it appears there isn’t enough of a change over time to find anything resembling conclusive evidence for Trump having dementia, Alzheimer’s or other cognitive decline.