Direct, uninterpreted records of the subject of your research project. Thus, a primary source can be almost anything, depending on the subject and purpose of your research.
Historical liturgical studies attempts to answer three questions with reference to past communal worship:
1) What did people do when they worshiped?
2) What did people say when they worshiped?
3) What did people think what they did and said meant when they worshiped?
Thus primary sources attempting to answer these questions could be broken down into two major categories: non-written and written.
Non-written primary sources for historical liturgical studies include:
architectural spaces in which worship took place (e.g., baptisteries, churches, shrines...) and their surface adornment (e.g., wall painting, mosaics, stone carving, bronze cast doors, etc.)
artifacts employed in worship (e.g., enameled reliquaries, fabric for vesture, glasswork for illumination, icons, ivory situlas, metal chalices and patens, woodwork croziers, etc.)
Written primary sources for historical liturgical studies include two major sub-categories:
written documents to be used in worship (e.g., collections of prayers [libelli missarum, sacramentary, missal], collections of readings [marked Bibles, capitularies, lectionaries], collections of preached texts [sermons], collections of sung texts [antiphonaries, hymnals], collections of directions [ordines romani]
written documents about worship (e.g., church orders, travel diaries, liturgical commentaries)