Nevertheless, partisanship persisted: fake news warnings help briefly, but bias returns with time

Cognitive Research: Principles and Implications

Table 3 Predictors of increased number of false items believed to be true

	Poisson regression model predicting count of false items rated as true at end
	Coefficient	Standard error	p value
Warning condition
Warning-After	0.128	0.108	0.238
Warning-During	0.087	0.110	0.428
Order condition
Memory items first	0.086	0.087	0.322
Personal variables
Male gender (relative to female)	0.031	0.090	0.732
Ideology (higher is more conservative)	− 0.010	0.027	0.718
Interest in politics	− 0.034	0.044	0.434
Conspiratorial disposition**	0.108	0.035	0.002
Online news variables
Social media usage	0.015	0.043	0.717
Trust in social media	− 0.015	0.051	0.765
Trust in online news	− 0.021	0.054	0.702
True news stories rated as true***	0.074	0.022	0.001
Constant	− 0.500	0.312	0.109

Items with *** were statistically significant at p < 0.001, ** are significant at p < 0.01, and others were not significant (p > 0.05). The reference categories were Warning Before for Warning condition, Bias awareness questions first for Order condition, and female for Gender (with other genders excluded due to sample size)