Researchers at the University of Aberdeen found exaggerated symptoms and feelings were expressed in the first instance, but that disappears when people are asked a second and third time.
The findings could help improve the way surveys and trials are conducted, as well as aid understanding of which diagnosis and treatments patients should be given.
Gertraud Stadler, from the university’s health psychology group, said: “This upward bias in self-assessments could have implications for research and practice across disciplines.
“Everything from assessments before and after treatment, over eligibility in randomised controlled trials, policy-guiding symptom prevalence, to national happiness indices could be affected.
“Because many political and health surveys only ask questions at one point in time, researchers have no way of seeing this initial elevation bias and get the mistaken impression of genuinely elevated emotions and symptoms.
“Just as doctors take blood pressure twice, we may need to check self-reports twice to get valid assessments.”
Researchers have long noticed a “puzzling” decrease when people are repeatedly asked about their physical symptoms and feelings.
It had been assumed it was the subsequent answers that were biased, with people down-playing their symptoms upon repeated questioning.
But this latest set of experiments suggests the opposite is true - that it is the first answer that is “upwardly biased”.
To study this decrease, the research team conducted four experiments asking participants at multiple times about their distress, physical symptoms and energy level.
Among the participants were recent law school graduates preparing for a career-deciding exam, pre-med students gearing up to take exams and college students over the course of an academic year.
The results across the four experiments showed people overstated their initial feelings and physical symptoms, but that after being assessed a second and a third time this “initial elevation bias” disappeared.
The study has been published in the journal Proceedings of the National Academy of Sciences of the USA, and also included researchers from US universities - New York University, Columbia University, Purdue University, Adelphi University, UCLA, Iowa State University, Arizona State University, and the University of Texas at Austin.