I’ve seen this design again and again, but not in relational databases. I usually see it in spreadsheets. I would be completely unsurprised to see a three column Excel sheet with Employee, Skill, and Language headers. (https://kerricklong.com/tmp/demo-4nf-need.jpg) The spreadsheet designer’s intuition is that Skill and Language are lists, and Employee is a section header. Smith might have a few rows with his languages and skills listed, one per cell, in arbitrary top-to-bottom order and with no consideration for how language and skill in a row relate to each other—only to Smith. The Employee column has a single merged cell for Smith, a number of rows tall it needs to be to contain the largest of his lists. When migrating this from a spreadsheet to a database, you lose merged cells as a feature and can no longer rely on order—so you copy the value Smith to each cell that had been merged.