Semantic Features as a Cause of Tensification in Korean Sub-compounds

Nominal compounds of ‘N1 + N2’ in Korean can be classified into the following three major categories: co-compound, sub-compound, and fusion. Among these three major categories, insertion of /t/ in the compounding process and subsequent tensification are found only in sub-compounds. This peculiar phenomenon of /t/-insertion which causes, in turn, tensification in sub-compounds has been long controversial because linguists have not been able to expect in which phonological environment of sub-compounding insertion of /t/ takes place. In this paper, I explore a phonological rule which makes it possible to expect the phonological environments of sub-compounding that allow insertion of /t/ and automatic tensification of the subsequent consonant in the onset of N2. In this process, I show that semantic feature(s) between two combined roots should be considered as one of the important structural descriptions in phonology.


Introduction
The phonological phenomenon of tensification in Korean compound words has long been one of the most confusing topics to explain by rules. Many attempts such as , Jung (1980), and Ahn (1985) among others, have been done to explain the phenomenon through purely phonological perspectives, with no satisfactory results. Hyman (1978)"s idea based on the concept of "boundary change" does not provide why tensification takes place only for some cases of sub-compounds in Korean. Unlike previous studies, My research begins with the idea that probably the difficulty in explaining the tensification phenomenon of Korean compounds may be due to some more complicate internal features hidden and working in the process of tensification, and the features may not purely phonological. In realizing this idea, one very critical issue is if there are some non-phonological features involved in this process, then is it admittable for us to capture them into the field of phonology.
This issue of abstractness in the so-called science of phonology maybe justified by the many theories in phonology employing the abstract levels just before coming into the surface forms of representation in phonology, too. According to pilot studies on this topic, some semantic featuresor semantic relationsbetween N1 and N2, which combine to form a compound, seem to be working very positively than other semantic ones, and in this paper I am trying to find them out to explain the tensification phenomenon in the sub-compounds in Korean in the form of semantics unified in phonology.

Discussion І
When there is a nominal compound of the structure "N(oun)1 + N(oun)2" and the coda of N1 is filled with a voiced segment, voicing of the following obstruent of N2 is quite generally expected instead of tensing. Interestingly enough, however, there are some Korean nominal compound words in which tensification of the obstruent in the onset of N2 is observed as often as voicing. The following are the data I will discuss here.
(1) and (2)  "bed" As we see from the data of co-compound and fusion in (1) and (2) respectively, they do not show tensification in surface forms; while the data under (3.c) to (3.e) show tensification in sub-compounding. Comparison of the phonological environments of the data in terms of [ tensification] is shown in the following: As we see from the description of (4), it is very hard in which phonological environment tensification takes place. The same phonological environment causes tensification in some data of [+Tensification] by crossing over the word boundary #, but not in others of [-Tensification]. Among others, , applying the idea of "boundary", suggested the following rule to explain this phenomenon.
However, his data were quite limited and we can easily find data that are not explainable by his rule of (5). For example, in the following data, tensification does not occur though the phonological environments meet with the condition of (5): (1a) ma # so  maso, horse and cow, (1b) pɔm # kaɨl  pɔmkaɨl, spring and autumn, (2b) kaŋ # san  kaŋsan country, (3.a.1) ssal # sul  ssalsul ricewine, etc. Jung (1980) tries to explain the same phenomenon by way of following two rules: Jung covers broader and more data than those in this paper: it goes beyond nominal compound of N1 + N2 which is the main focus of this paper. Jung explains tensification in sub-compound ranging from 3.c to 3.e such as "sɔnt'ɨŋ' back of the hand (3.c.), mulk"ɔki fish (3.d.), ʤamʤ"ari bed (3.e.), and more data such as pɔmp'aram spring wind, sulʤ'ip a bar, tɔnʤ'umǝni a (coin) purse, etc. (Of course, the new data introduced by Jung can be explained by my theory which will be developed and depicted soon.) However, what about the data from (1) to (3b) which do not show any tensification? Neither is possible by (4) of Jung (1980), nor are the following counter data, which have the same phonological environments as Jung"s (1980) data 9 . Consider: The existence of data such as those from (1) to (3.b) plus the counter data shown in (5) is strong evidence of the insufficiency in explanatory power of Jung's set of rules.
Another approach to this topic by Ahn (1985) is different from Kim and Jung in that Ahn is based on morphological explanation rather than phonological one. Trying to unify two separate rules offered by Jung which are shown above as (6), he suggests a morphological level on which / t /-insertion occurs.
Ahn assumes four levels in the morphological component: S1 for sub-compounding, S2 for co-compounding, S3 for derivation, and S4 for inflection and case-marking. He further assumes that /t/-insertion takes place only on the first level, that is, S1. By limiting the level of /t/-insertion to S1, which is assumed for sub-compounding, Ahn explains tensification of (7a,b&c) such as pɔmp'aram spring wind, sulc'ip a bar and tɔnʤ'umǝni a (coin) purse as well as non-tensification of co-compounding such as (1.a) maso horse and cow' and (1.b) pɔmkaɨl spring and fall, etc. Ahn, however, cannot deal with the counter examples against Jung which are shown as (9) in the following (The following is a repeat of part of (7)). Consider: The data shown in (9) belong to sub-compound words, so they are supposed to be subject to Ahn"s rule of (8). Again, however, it cannot explain why the data (9) do not show the expected process, nor can we include the data (9.a',b',& c') in the co-compound category.
The discussion so far leads us to conclude that as far as purely phonological approach is concerned, we cannot explain /t/-insertion and subsequent tensification phenomena which occur irregularly inside compound words because the same phonological environment(s) cause different phonological results.

Discussion П
Before we get into the answer to the question at hand, we need to discuss Hyman"s (1978: 443-447) suggestion concerning "boundary change" first. He suggests two grammatical boundaries within a word: one is "#" which is for internal word boundary and the other is "+" which is for morpheme boundary. Phonological rules are supposed not to be able to apply crossing over the internal word boundary "#" and they can"t influence a segment belonging to another word. Whereas, they can cross the "+" boundary. This means that phonological sound change crossing over a word boundary which is depicted here as consonant tensification in the onset of the following root, should be preceded by the application of some rule or condition which allows crossing-over or weakening of # boundary. This idea as well as the relative strength of each boundary in morphology is expressed in the following: (10) Boundary Change (Hyman 1978) ║ → ## → # → + ∅ Secondly, it is necessary to assume that semantic features such as Agentive, Patientive, Genitive, Locative (Time), etc., which are used to distinguish among sub-compound words, must be considered as part of the structural descriptions.
With these two conditions in mind, we can expect that the internal structure of the sub-compound words which show /t/-insertion and subsequent tensification should be [ # ____ + ____ # ] after boundary change in (10). Therefore, sub-compound words with the internal structure [ ## N1 # SR* # N2 ## ] will experience weakening of internal word boundary "#" to "+" only when SR* (Semantic Relations) is one of those such as Genitive, Locative(Time), and Beneficiary. Recall that data only from (3c,d,&e) of sub-compounds are showing tensification, and the Semantic Relations -or Semantic Features -of N1 and N2 of these data are Genitive, Locative(Time), and Beneficiary..
After weakening of the internal word boundary, /t/-insertion and subsequent tensification of the first segment on the onset of of N2 will follow. The process of boundary weakening may be described as (11) (SR* =Semantic Relation) Concerning the inserted sound, there have been many explanations. However, based on the following data, I will follow the explanation of /t/-insertion which was suggested by Kim-Renaud (1975), Yoo (1964), and Heo (1984)  All the data in (12) belong to sub-compounds with semantic relations of Genitive (12a&b), Locative (12d) and Beneficiary (12c). For each of the data of (12) which have N1 as an open syllable, we have two realizations: one with /t/-insertion and the other without /t/-insertion, but both with tensification. The data in (12) and those from (3c) to (3e) show two important similarities: (1) all of them belong to the category of sub-compounds. (2) All of them belong to the semantic relations of {Genitive, Locative (Time) and Beneficiary}. In addition to these similarities, the observation that there could be two realizations of each sub-compound word in (12) and one of two realizations has unreleased [t>] in the coda of N1, leads us to the assumption that probably all the sub-compound data showing the same tensification phenomenon may experience /t/-insertion after boundary weakening which is caused by the intervening proper semantic relations. The rule for /t/-insertion we can obtain from the observation above should be as follows: Based on the two important rules we have developed so far and some other rules which have already explored for the explanation of Korean data, let"s see how the data in (3c -e) and (12) could be explained. (14) is a repeat of some data of (3c-e) and (12)