Talmud for Shevuot 4:4
כָּפְרוּ שְׁנֵיהֶן כְּאַחַת, שְׁנֵיהֶן חַיָּבִין. בָּזֶה אַחַר זֶה, הָרִאשׁוֹן חַיָּב וְהַשֵּׁנִי פָטוּר. כָּפַר אֶחָד וְהוֹדָה אֶחָד, הַכּוֹפֵר חַיָּב. הָיוּ שְׁתֵּי כִתֵּי עֵדִים, כָּפְרָה הָרִאשׁוֹנָה וְאַחַר כָּךְ כָּפְרָה הַשְּׁנִיָּה, שְׁתֵּיהֶם חַיָּבוֹת, מִפְּנֵי שֶׁהָעֵדוּת יְכוֹלָה לְהִתְקַיֵּם בִּשְׁתֵּיהֶן:
If both denied it at the same time [i.e., in the same "speaking span"], both are liable. (If they denied it) one after the other, [i.e., after the "speaking span"], the first is liable, and the second, not. [For since the first denied it (i.e., that he could testify), the second can no longer testify, being only one.] If one of them denied and the other admitted, the denier is liable. If there were two sets of witnesses — If the first denied it and then the second denied it, they are both liable, because the testimony can obtain through both. [The Gemara asks: Why should the first set be liable if there is a second set? What loss did they cause him by their denial? And it answers that our Mishnah is speaking of an instance where the witnesses in the second set were kin through their wives, not being kasher to testify when the second set denied, and their wives were gosesoth (at the point of death). I might think that since the ruling is: "most gosesim die," it is as if they are already dead, and the first set should not be liable, for there is still a second set; we are, therefore, apprised (that this is not so) for now (at the time of the denial) they had not yet died. It is found, then, that only the first set was there at the time of the denial, for which reason they are liable.]