A certain parallelism in the sentences or groups of characters has also been discovered. Attention was first called to this by me in the work referred to, but is more fully explained by Dr. P. Schellhas in his paper entitled "Die Mayahandschrift der königlichen Bibliothek zu Dresden." It will readily be understood from a single illustration. Take for example the lower division of Plate XV of the Manuscript Troano (see Study Ms. Troano). Omitting from consideration the numerals and the day column at the left, there are here two short columns on the left and two on the right over the animal figures, and three longer columns between. As explained in the work referred to, the short columns are to be read as lines from left to right and the longer columns separately, from the top downward. There are, in all, five groups or sentences, each containing four compound characters. Representing these by letters, repeating those which indicate similar characters, and arranging as in the plate, the result is as follows: ___________________________________________ | | | | | | | _b_ _a_ | _h_ | _l_ | _m_ | _w_ _a_ | | | | | | | | _r_ _n_ | _a_ | _a_ | _a_ | _r_ _s_ | |____________| | | |____________| | _r_ | _r_ | _r_ | | | | | | _p_ | _k_ | _t_ | |_____|_____|_____| In this case the characters represented by _a_ and _r_ are repeated in each group and in the same relation to the other characters. It is apparent, therefore, that each group is to be read separately, and, as each repeats in part what is given in the others, it is more than probable that they are simply short formulas to be repeated in certain religious ceremonies. This parallelism, though not always so apparent as in the case presented, is nevertheless found running through all the codices. The advantage to the attempts at decipherment which results from this fact is evident, as it will often justify the restoration of blurred or obliterated characters, and, what is of still more importance, will enable the investigator to test his conclusions by comparing the different characters and pictures with which they are associated. Although it appears to be well settled that, as a rule, the writing, when in lines, is to be read from left to right--the lines following each other downward and the columns to be read from the top downward, but the groups, as before explained, to be read separately--it does not follow that the _groups_ succeed one another from left to right. This has generally been taken for granted, but there are some reasons to doubt the correctness of this conclusion as regards a number of plates and possibly one entire codex.