The Field Guide › Paper
Across five studies, Boothby, Cooney, Sandstrom and Clark find that after an ordinary conversation, people reliably underestimate how much their conversation partner liked them and enjoyed their company, a mistake they name the liking gap. It cannot be that both partners in a dyad are correct in feeling less liked than they actually are, which is what establishes this as a systematic error rather than mere modesty: trained coders watching the same conversations back could accurately judge how much two people liked each other, meaning the liking was genuinely being signalled, but the people inside the conversation failed to register it in themselves. The mechanism the authors trace is a mismatch of attention rather than a lack of information. After talking, people's most salient thoughts about their own performance are self-critical, replaying the sentence that landed awkwardly or the pause that went on a beat too long, while they have no equivalent access to their partner's inner monologue, only to that partner's outward behaviour, which is generally warm, engaged and perfectly fine. People then use their own harsh internal verdict as a stand-in for their partner's opinion, and the harshness travels across but the warmth of the actual signal does not. The gap is not a product of one bad exchange: it held across strangers, dorm-mates followed for a full academic year, and members of the public at conversation workshops, and it was larger in shyer people, which speaks against the alternative explanation that people were simply being modest rather than genuinely mistaken. For a corpus about psychological safety this is a precise account of the private cost of every ordinary interaction that precedes the high-stakes ones. If people default to reading themselves as less liked than they are, the felt risk of speaking up, disagreeing or asking a question is inflated by a bias that operates before any actual evaluation has taken place, which pairs directly with prospect theory's account of loss aversion in the voice calculus: the anticipated cost of a misstep is measured against a self-assessment that is already running low. Its limits are that it measures general liking and enjoyment in largely dyadic, first-meeting or early-relationship contexts rather than the specific, higher-stakes judgements of a workplace hierarchy, so the size of the effect in a performance review or a team meeting is this paper's plausible extension rather than its direct finding. (Text drawn from the 2018 Psychological Science paper, 29(11), pp. 1742-1756.)