This paper aims to clarify what the underlying form of the ``sais-sori`` (epenthetic sound) of Korean is and by experimenting phonetically, to find traces of the ``sais-sori`` that do not reveal the phonetic values on the surface. The underlying form of the ``sais-sori`` of Korean is represented as /s/ (/ㅅ/) in i) the reception of foreign languages that take [t] as the final sound, ii) the cognition of speakers who are non-proficient in orthography, and iii) the reference to the phonetic values of the ``sais-sori`` which is presumed in Middle Korean. Since /s/, as the ``sais-sori`` of Korean, is realized as a surface form by applying the existing rules in Korean phonology such as the unreleased rule, the automatic tensification rule and the consonant cluster simplification rule, it does not complicate the Korean grammer. The experiment for revealing the traces of the ``sais-sori`` of Korean is briefly accomplished by measuring the utterances of three Korean native speakers. Here it is anticipated that the length of the fortis achieved in the ``voiced sound(including vowels) # lenis consonant`` environment and in the ``obstruent # lenis consonant`` sequence of compound nouns in which the ``sais-sori`` is considered to be inserted, will appear to be longer than the length of another fortis unrelated to the ``sais-sori``. However, the former turned out to be the longest while the latter did not show any difference from the length of the fortis unrelated to the ``sais-sori``. I think in the former case, the ``sais-sori`` is being reflected as the length of the fortis, whereas the latter case is due to the application of the consonant cluster simplifiction rule. Hence we can tell that the current spelling system of adding the /s/ when the preceding segment ends in a vowel is perfectly adequate.