The main purpose of this paper is to search for the origin of -$geyo$- indicating subject-honorification in some Korean dialects. It always follows eo that cannot be recognized as an independent morpheme. In truth, the honorific marker in question is not -$geyo$- but -$eogeyo$- . It can occur only before endings beginning with eo, which is obligatorily deleted by a phonological rule. Before other endings appears --$eu(si)$- , which is used in all the dialects. The first example of - $eogeyo$ - is found in the middle of the 19th century. The forms given below are of the 19th century. In the past tense appears -$eogeyos$-/-$eogeyosi$- (henceforth, the preconsonantal allomorph is presented before the slash, and the prevocalic one after it), which looks as if it had been formed from the sequence of -eogyeo- plus the past-tense marker -eos-/-eosi-. However, as demonstarted in Ko (2000), -$eogeyos$-/-$eogeyosi$- eveloped from -$eo geyos$- just as -$eos/-eosi$- - did from -eo is/isi-. hyeosi- is the honorific form of -$is/-isi-geyos$- . We see here two parallel instances of morphologization. There are two alternatives for the source of -$geyos$- . First, it may have been back-formed from $-eogyeos-/-eogyeosi-$. In other words, it is quite likely that $-eogyeos-/-eogyeosi-$ was reanalyzed as -$eogyeo$- plus$-eos-/-eosi-$-. However, if we accept this interpretation, it will be somewhat difficult to explain why only eo-initial endings can occur after -eogyeo-. Second, $-eogyeo-$ may have been created from $-eo$ gyeosi-eo used equivalently to $-(eu)si -eo in meaning. In other words, it is possible that $-eo$ gyeosi $-e$o was reinterpreted as $-eogyeo-eosyeo$, which is phonologically identical with the former. Originally, $-eosyeo$ is the sequence of $-eo$ plus the particle $-syeo$. Other eo-initial endings occurring after $-eogyeo- are -eo, -eodo, and -eoya$, the last two of which are the sequences of $-eo$ plus the particles $-do and -ya$, respectively. If the second interpretation is right, we can easily understand why only eo-initial endings can occur after $-eogyeo-$. However, $-eo$ gyeosi-eo used equivalently to $-(eu)si-eo$ is found so rarely in documents that we cannot be confident of the second interpretation. It is reserved here to select one from the above alternatives. It is also possible that $-eogyeos-/-eogyeosi-$ and $-eo gyeosi-eo$ interacted to produce $-eogyeo-$. No matter which is the immediate source of -eogyeo-, the ultimate origin of it is -eogyeosi-. There is no reason to search for its origin elsewhere.