In a paper recently published in the journal Nature Human Behavior, researchers from the National Institutes of Health and the University of Chicago use computer-modeling to show that some words are more memorable than others. This research could establish a universal baseline for research in memory since the greatest challenge in this research is that memory is subjective.
During a small word-pairing experiment, published in March 2020 to learn more about how neural networks form and retrieve memories, the researchers found that some words were remembered more often among all participants, with the most ubiquitously remembered words being successfully recalled on average seven times more frequently than the least often remembered words. In a more extensive word-pairing study of over 2,600+ healthy individuals conducted later in the spring, researchers found the same group affiliation for some words.
These observations inspired Dr. Xie, a post-doctoral fellow at the NIH’s Institute of Neurological Disorders and Stroke and author of the most recent study, to write a computer model that would match the results of the spring 2020 studies, as a way of working backward to map neural networks. The computer model was fed the same words as participants in previous word-matching studies. It organized the words by frequency in the English language, “concreteness” of the definition (a sock is a physical item vs. a dream), and how similar the meaning is to other words.
The researchers found that when the computer organized the words by how similar the definition is to other words, the computer correctly predicted the words that were most frequently recalled in previous studies. The relatedness of words reinforces neural networks that connect word and definition, making the memory easier to access.