In the study, 64 older adults (60-74; average 70) and 64 college students were compared on a word recognition task. Both groups first took a vocabulary test, on which they performed similarly. They were then presented with 12 lists of 15 semantically related words. For example, one list could have words associated with "sleep," such as "bed," "rest," "awake," "tired" and "night" — but not the word “sleep”. They were not told they would be tested on their memory of these, rather they were asked to rate each word for pleasantness.

They then engaged in a five-minute filler task (a Sudoku) before a short text was read to them. For some, the text had to do with age-related declines in memory. These participants were told the experiment had to do with memory. For others, the text concerned language-processing research. These were told the experiment had to do with language processing and verbal ability.

They were then given a recognition test containing 36 of the studied words, 48 words unrelated to the studied words, and 12 words related to the studied words (e.g. “sleep”). After recording whether or not they had seen each word before, they also rated their confidence in that answer on an 8-point scale. Finally, they were given a lexical decision task to independently assess stereotype activation.

While young adults showed no effects from the stereotype manipulation, older adults were much more likely to falsely recognize related words that had not been studied if they had heard the text on memory. Those who heard the text on language were no more likely than the young adults to falsely recognize related words.

Note that there is always quite a high level of false recognition of such items: young adults, and older adults in the low-threat condition falsely recognized around half of the related lures, compared to around 10% of unrelated words. But in the high-threat condition, older adults falsely recognized 71% of the related words.

Moreover, older adults’ confidence was also affected. While young adults’ confidence in their false memories was unaffected by threat condition, older adults in the high-threat condition were more confident of their false memories than older adults in the low-threat condition.

The idea that older adults were affected by negative stereotypes about aging was supported by the results of the lexical decision task, which found that, in the high-threat condition, older adults responded more quickly to words associated with negative stereotypes than to neutral words (indicating that they were more accessible). Young adults did not show this difference.