Gemara

(R. Chanina bar Papa): "The Meraglim... said 'we cannot... he (the nation there) is stronger Mimenu.'" Do not read 'Mimenu (than us)', rather, 'Mimenu (than He).

Nedarim 37b: Rav holds that the cantillation is mid'Oraisa.

(Rav): "They read in the Sefer Torah of Hash-m" refers to Mikra...

"And they understood the written Torah" refers to the cantillation;

Some say that this refers to the tradition of extra or missing letters.

(R. Yitzchak): The following are traditions from Moshe from Sinai ... words that are read but not written, or written but not read, e.g. "Paras" (in "in his going..."), "Na" (in "forgive...").

Menachos 29b: The Sefer of R. Aba's father-in-law had extra letters. R. Aba said that missing letters mandate Genizah, but extra letters are not a problem.

32b (Rav Yehudah): If a Sefer Torah was written like an Igeres (letter), it is Pasul.

Rishonim

Rashi (35a DH Al Tikri) and Tosfos (Erchin 15b DH Al): "Mimenu" is pronounced the same whether it means 'than us' or 'than him'.

Ibn Ezra (Shemos 1:9) and Teshuvas Rashba (attributed to Ramban, 232): Western Sefarim (which follow the tradition of Eretz Yisrael) have a Dagesh (dot) in the Nun either way. Eastern Sefarim (which follow the tradition of Bavel) have no Dagesh when it means 'than us'.

Rambam (Hilchos Sefer Torah 7:11): If a word that should be written full was written lacking, or vice- versa, or if a Kri-Kesiv (when a tradition teaches that we pronounce a word unlike it is written) was written the way it is read, e.g. he wrote "Yishkavenah" in place of "Yishgalenah", it is Pasul.

Tosfos (Menachos 32b DH Kosvah): If a Sefer Torah like an Igeres, i.e. the scribe was not precise about extra or missing letters, it is Pasul.

Tosfos (Menachos 30a DH Aval): Some explain that we are not concerned for extra letters, therefore whenever they have a doubt they write the word full. This is wrong. We are not concerned for extra letters that were erased.

Teshuvas Rashba (ibid.): We do not add missing letters or remove extra letters based on Medrashim. However, if the Gemara learned a Halachah from it, we do. When texts differ, we follow the majority of Seforim (Maseches Sofrim 6:4).

Poskim

Shulchan Aruch (OC 141:8): Every word that has a Kri and Kesiv is a tradition from Sinai to write it the way it is written in the Torah, and to pronounce it differently. A case occurred in which one read it in front of Gedolei ha'Dor the way it is written. They warned him to read it according to the tradition, and he refused. They excommunicated him and dismissed him (from reading).

Question: Every word of Kri'as ha'Torah must be read from the Sefer Torah. How can we read "Yishkavenah", which is written "Yishgalenah", and other words for which are written unlike the way we pronounce them?

Answer (Rashba 7:361, cited in Beis Yosef DH Kosav ha'Rashba): It is a tradition from Sinai to read the words unlike they are written.

Eshel Avraham (9): If a Sefer Torah is lacking a word that is not read, we do not read from the Sefer. Even a Sefer missing one letter is Pasul. However, if the word is merely covered by wax, one may read from it, since in any case the word is not read from the Sefer.

Kaf ha'Chayim (44): If one read b'Tzibur "Yishgalenah", like it is written, he was not Yotzei.

Shulchan Aruch (143:4): If an error is found in a Sefer Torah, we take out a new one.

Rema: This applies only to a total error. If there are merely extra or missing letters, we do not take out another, for our Seforim are not so precise that we can say that the new one is better than the first.

Darchei Moshe (YD 279:3): The Terumas ha'Deshen (70) says that we never take out another Sefer due to a mistake. I agree regarding extra or missing letters, for none of our Seforim are precise about this, like the Agur says. If it has a mistake not in other Seforim, we take out another, like the Beis Yosef says.

Minchas Yakov (18, cited by R. Akiva Eiger): The Agur holds that even if a Sefer is missing verses, we do not take out another. The Rema rules like him only regarding extra or missing letters.

Magen Avraham (7): This refers to extra or missing Vov'im and Yud'im. It does not depend on whether or not it changes the meaning. Keves and Kesev both mean 'sheep', but it is a different word and we must take out a new Sefer for such a mistake. Maharam Luvlin (90) says that Rachbah in place of Rachbo is a change that mandates another Sefer. Even though one could read it Rachbo even with a Hei at the end, there is a reason for how it is spelled. I say that even without the reason, any error mandates a new Sefer except for missing or extra letters, since we are not proficient about these. Tzemach Tzedek (119) says that if one wrote Migresheihem in place of Migresheihen, we need not take out a new Sefer, since sometimes it is written Migresheihem.

Rebuttal (Eshel Avraham 7): Even if the Torah writes it both ways, presumably there is a Drashah why here it is written this way!

Eshel Avraham (Sof 143): Me'il Tzedakah (54,60) says that if "Lecha" (when it means 'go') was written without the Hei, we do not take out another Sefer. If "Padan Aram" was written 'Padam', it is Pasul. Only a change from masculine to feminine does not mandate taking out another Sefer.

Be'er Heitev (9): Chacham Tzvi (54), Shevus Yakov (2:3) and others say that we take out another Sefer for anything except for missing or extra letters.

Kaf ha'Chayim (33): This opinion takes out another Sefer if Chamishisav is written lacking (without) a Yud, even though it can be read Chamishisav, or Me'en Yevami is written full. Sefardim take out another Sefer for any mistake.

Note: The Chayei Adam (31:34) says so because we learn from the presence or lack of these Yudim. He concludes that we do not take out another, lest also it has mistakes, i.e. missing or extra Yudim or Vovim. We take out another Sefer for mistakes that affect the pronunciation, e.g. the Aleph in "Te'umim." He does not address a silent Yud between a Chirik and a Shva, which causes the Shva to be Na (pronounced).

Shulchan Aruch (YD 275:6): If a word that is written lacking was written full, or vice-versa, or if a Kri-Kesiv was written the way it is read, e.g. he wrote "Yishkavenah" in place of "Yishgalenah", it is Pasul.