@grimalkina I saw this in the wild just last week. Looked for existing literature on a scale some colleagues had used and the key publication (not by the survey developer) basically said, the one reverse coded item here sucks and wrecks the scale’s psychometric properties. Sure enough, that was true in the dataset I’d been handed, too. I do suspect some failure on the part of the scale’s authors to interrogate whether their construct was actually unidimensional. I think if they’d stopped to ask themselves that, they’d have written a scale with more than one such item.