The researchers then used statistical modeling to measure the effect of the tweets they saw for certain treatments, like trust and confidence in elections, support for democratic norms or support for political violence or democracy, said Katie Clayton, a political science researcher at Stanford University and the study’s lead author. They looked at whether there were changes over time by comparing the responses across different waves of the study.

The tweets sometimes reduced the belief among Trump supporters in a peaceful transfer of power, said Brendan Nyhan, a Dartmouth College political science professor and study co-author. They generally didn’t increase support for political violence, according to the study.