It is much easier to determine the grammatical affinity of a language than to read it, and the place of Babylonian in the family of languages was definitely fixed before much progress was made in the work of translation. From the time that Grotefend’s attention was first directed to cuneiform research, he endeavoured to include the second and third columns, no less than the first, within his sphere of inquiry. But he achieved very little success. In his Essay published by Heeren (1824) we find that he had already singled out the groups in the third column that corresponded to Cyrus, Hystaspes, Darius and Xerxes. The Babylonian, unlike the Persian, has no sign to mark the division of the words, and the difficulty attending their separation was at first very great. The process was facilitated when it was recognised that each line begins and ends with a word: that is to say, a word is never divided and carried over from one line to another. In 1837, Grotefend successfully divided eight lines of the Elvend inscription, with only a slight mistake. He also divided the B inscription of Darius nearly correctly: the exception being that at the end of his first line he seems to treat three words as one.[771] Nor was the difficulty confined to the separation of the words only. Some of the signs are so long that they were at first mistaken for two or more letters. Thus the sign for _ar_ in the word for Xerxes was treated by Grotefend as _th_ and _r_;[772] and Löwenstern divided the sign for _gi_ into _r_ and _s_, which continued for a long time to be a source of trouble. In 1840 Grotefend gives a Table to show the transliteration of his four royal names. He reads Cyrus ‘Kho · re · s’ for ‘Ku · ra · as’; Hystaspes ‘Wi · scht · as · p’ for ‘Us · ta · as · pa’; Darius ‘Da · r · ha · a · wesch’ for ‘Da · ri · ya · a · vus’; Xerxes ‘Kh · sch · ah · th · r · sch’ for ‘Hi · si’ar · si.’[773] We have here sixteen different signs with their values attached; and of these only three (_as_, _da_, _a_) are absolutely correct; though the others give the consonantal values. These values appeared substantially in his tract of 1837, with the addition of ‘wo-hu’ for the signs that read ‘rabu.’[774] Hincks, writing in December 1846, makes the very liberal admission that perhaps Grotefend knew the values of ten cursive characters correctly and of ten others approximately.[775]