are written in the two-line system. The fourth oldest source according to Eckhardt is OA with characters "around the middle line"'" (Fig. 1. 4). GE VLAJ KHBA PAZDLJ P LRLNAU 4A" (3H Q0&0d3 06 FPGPBALJVTSFSP RAD YWB:BAJ" JGREVOIAJ) S<BAPSNNAZPITRIIVISJAI" korh z Q9,00305 23791 ov -PM ANI1BA Figure 1. Thorvi Eckhardt's line systems: (1) ZG; (2) AG: (3) KF; (4) 0A.* OGFEVLAJ KCOAIASRKAIJIPVOILNAUJAJ HEITIIROINJ SAT ITZFIZSTILJIIMICJAVA JEPEVIAJODOVPOJIAJTITOVTIVSTJAV = T Figure 2. Eckhardt's lines applicd to the whole alphabet 1 applied Eckhardt's lines (Fig. 2) to the whole proposed alphabet to verify if they properly describe character size and position. The lines confirm Eckhardt's own statement that in ZG and AG “most frequently the seript floats between the lines" (Fig. 2.1 and 2.2 respectively). She does not describe the rules of floating?' although it would be indispensable here to define the different character sizes which allow floating. Fig. 2.3 shows that the significant size difference in KF characters cannot be described by one upper line. Fig. 2.4 reveals that the middle line fals in the middle for five or six characters only. For the others, it is an upper line from which they hang. From Eckhardi's illustrated yet transparent hypothesis, it is obvious that upper, middle or lower lines cannot successfully describe character size or position. It is a useful demonstration of the difficulties inherent in defining the earliest line system. Indeed, from the two-line point of view, the characters from other old Glagolitic sources, like the Sinai Praer Book, the Krk Inscription or the Prague Fragment, also seem inconsistent. KF and OA, with Eckhardt's one line, either upper (Fig. 2.3) or "middle" (it is not really middle, Fig. 2.4), are likewise exceptions to the two-line system. Now we come to an internal reconstruction of the line system in analytical Glagolitic paleography. All the aforementioned characteristics of different line systems in the oldest known sources point toward an unknown older maximal system which would allow for all these variants. Such a maximal system was discovered by the graphic 9 Ibid., 72-73. *9 [bid., 72. 2! Tbid., 75. 287