Data | From ‘therapy’ to ‘psammophile’, the evolution of Spelling Bees

The Scripps National Spelling Bee has been picking tougher winning words to keep up with the expanding vocabulary of its participants. Here is a look at how tough this competition that has been dominated by Indian-origin winners is.

On June 1, 14-year-old Indian American Dev Shah won the Scripps Spelling Bee by correctly spelling ‘psammophile’, meaning a plant or animal that thrives in sand. Since 1925, there has been 94 editions of the Spelling Bee. The first one saw nine participants; this year saw more than 200. Here we look at how this prestigious U.S. competition has evolved over the years.

With a silent ‘p’ and a double ‘m’, the 2023 winning word is tricky. But how about ‘therapy’? It was the winning word in 1940. Earlier editions of the Spelling Bee had other easier words like ‘milieu’, ‘psychiatry’, ‘knack’ and ‘interning.’

From ‘therapy’ to ‘psammophile’, there seems to be a trend of the winning word in the competition becoming increasingly obscure.

Using Google Books Ngram Viewer’s frequency ratings, we tested how frequently the winning words of various editions of the Spelling Bee were being used in books published around that period. Ngram Viewer analyses a corpus of 5.2 million books published between 1800 and 2019 and plots the frequency of words found in this corpus.

To arrive at the frequency, we took a 20-year period (10 years before and 10 years after the contest year) and calculated the average of how frequently the winning word was found in books published in those years.

For example, ‘Psammophile’, the 2023 winning word, has an average frequency of 0.0000000606% between 2008 to 2019 within books published in that period. That means, out of all the 1-gram words (words with no space) available in Google Books between 2008 and 2019, 0.0000000606% is ‘psammophile.’ The average frequency of all the Spelling Bee winning words (taken at year of contest) is 0.0000557%.

Compare this with ‘therapy’, which had a frequency of 0.0017% or even ‘interlocutory’ at 0.000156%.