American Dialect Quiz

For each answer you give, the model compares how common that answer is in a given part of the country with how common it is nationally. Answers that are distinctive somewhere carry weight; answers most people give everywhere carry almost none. Evidence from all your answers is combined, and cells with little underlying survey data are held back rather than allowed to spike. In the smart quiz, each next question is the one expected to sharpen that picture the most, and a nameless certainty score updates after every answer.

• Source. The publicly rendered Harvard Dialect Survey legacy result layers hosted by the Cambridge Online Survey of World Englishes. Retrieved —. This project uses no New York Times data or assets.
• What the numbers are. The source layers are raster maps of anti-aliased response markers. Marker coverage is converted into estimated counts on a 228 × 102 grid, giving 506 answer grids across 122 questions. These are reconstructed aggregate counts, not individual survey responses — and no response coordinates are stored or shipped.
• What the map means. It shows where your combination of answers is comparatively common among survey respondents. It is not a statement about your ethnicity, identity, birthplace, or current address, and it is not a calibrated probability that you live anywhere in particular.
• Known gaps. Some questions have live survey categories that the local question catalog cannot represent. Those responses stay in the denominator but are never assigned to a named answer, so no category is inflated. Seven questions carry too much of that residual to be useful and are left out of the smart quiz.
• Licence. Source data is CC BY-NC-SA 3.0; attribution and share-alike terms apply to derived data. Confirm licensing before any commercial use. Data version zoom9-v1.