Text this: Feature identification learning both shapes and is shaped by spatial object-similarity representations