Subjects randomly surveyed within view of a church gave more conservative responses than those surveyed within view of a secular school. Katherine Harmon reports