Deceitful Pumbeditans and Conditional Probability (full article)
Here is my Jewish Link article for the week (also in flipdocs):
In Avoda Zara 70a, some thieves came to Pumbedita and, in the course of their thievery, opened barrels of wine. Rava, a fourth-generation Babylonian Amora located in Pumbedita who eventually led the Pumbeditan academy, declared the wine permitted. The reasoning was that most of the thieves (in Pumbedita) were Jewish. A similar incident happened generations earlier in Nehardea, and first-generation Amora Shmuel permitted the wine, though apparently for a different reason.
Might we relate this to Chullin 127? There, Rav Giddel quoted the first-generation Amora, Rav: If a Nereshian kisses you (on the mouth), count your teeth (to make sure he hasn’t stolen them). If a Nehar Pekodian travels with you, it’s because of your nice cloak (that he wants to steal). If a Pumbeditan travels with you, change your lodging place (lest he rob your room). In Bava Batra 46b, fourth-generation Abaya explains to Rava how the swindlers in Pumbedita operate, with an example of a (presumably Jewish) swindling craftsman who crafts a legally valid claim to avoid returning the cloak he was given to repair.
Conditional Probability
This connection is not as compelling as it initially seems. In Chullin, the genre of Rav’s statement might be humorous stereotype instead of statistical fact with halachic implications. Further, while perhaps implicit, Rav never specifically noted that these untrustworthy city residents were Jewish. In Bava Batra,
Furthermore, there is a difference between P(A | B), that is, the conditional probability of event A given condition B; and P(B | A), the probability of event B given condition A. Let us make up some numbers for a town, say Pumbedita. There are 100 total residents, with 80 Jewish and 20 gentile. There are 10 thieves in the town, of whom 6 are Jewish and 4 are gentile. We might consider different probabilities. For instance, P(thief), the probability of a resident being a thief, is the simple frequency, that is #thieves / # residents, or 10 / 100, which is 10%. Meanwhile, P(jewish) is 80% and P(gentile) is 20%. Moving to conditional probabilities, P(thief | jewish), the probability of someone being a thief given that they are Jewish is the frequency of Jewish thieves, that #jewish thieves / #jews, or 6 / 80, or 7.5%. P(thief | gentile) is #gentile thieves / #gentiles, which is 4 / 20 = 20%.
Reversing the event and the condition, P(jewish | thief) represents the probability that a randomly selected thief is Jewish. That is #jewish thieves / #thieves, 6 / 10, or 60%. Meanwhile, P(gentile | thief) is #gentile thieves / # thieves, 4 / 10, or 40%. These probabilities add up to 100%. It is possible to convert one conditional probability into another, calculating P(A | B) given P(B | A), P(A), and P(B), via Bayes’ Rule, but we mustn’t confuse one for the other. Using my made up numbers, a greater percentage of gentiles are thieves, but a greater percentage of thieves are Jewish.
Here, the probability that within Pumbeditan residents, a given Jew is a thief, P(thief | jewish pumbeditan) is the subject of Chullin. The probability that a thief is a Jewish, P(jewish pumbeditan | thief) is the subject of Avodah Zarah. Put more simply, just because the gemara explained Rava / Shmuel as saying that most (Pumbeditan) thieves are Jewish, that does not bear on whether most (Pumbeditan) Jews are thieves.
The Thieves Arrived
Rav Steinsaltz renders Rava’s reasoning as “Most of the thieves [in Pumbedita] are Jews”, adding Pumbeditan residency. Slightly better, Artscroll renders it as “Most thieves [in the Pumbedita area] are Jews”. This is partly guided by Tosafot on Bava Batra 55b, s.v. רבי אליעזר מטהר. There, they contrast Rava’s Pumbeditan case with Shmuel’s Nehardean case, explaining that only Rava could rely on רוב גנבי ישראל נינהו, the majority of thieves being Jews, given the gemaras in Chullin and Bava Batra. This was not the situation in Nehardea.
Aside from my objections above regarding conditional probability, Rava’s case is introduced with הָנְהוּ גַּנָּבֵי דְּסָלְקִי לְפוּמְבְּדִיתָא, that certain thieves came to Pumbedita, with the implication that these are not Pumbeditan residents. Perhaps that’s why Artscroll mentioned the broader Pumbeditan area. Still, why would the character of Pumbeditan residents come into play, within Tosafot’s thought?
Further, while both our printed texts and Tosafot had רוב גנבי, “the majority of thieves”, the manuscripts (Munich 95, JTS 15, Paris 1337) have approximately רוב שיירי or רוב שוורי. In JTS 15, the original was שוארי but some scribal hand crossed it out and replaced it with גנבי. Consider that in the preceding context, it was גנבי / thieves who arrived, and also that Rava שְׁרֵי / permitted the wine, so either word could potentially influence a girsa change. However, there’s a common phenomena of arcane words in the Talmud being replaced by glosses or more frequent, better known synonyms, so that’s what likely happened here.
Jastrow renders this word, מְשַׁוְּורֵי, as freebooter, or rover, connecting it to jumping. Sokoloff similarly renders שוורי as robbers, contrasting it to a Syriac where it means “to attack”. However, they can only point to our very sugya in Avodah Zarah as the singular occurrence, a hapax legomenon, so it strikes me that their translations are speculative.
I would render מְשַׁוְּורֵי or whatever its word form as “member of a caravan”. שְׁיָירָא as caravan is well attested in Tannaitic and Amoraic texts, such as a brayta on Bava Kamma 116b about a caravan attacked by a troop of bandits in the desert. But, what relevance does this have to majorities?
A Mishnah on Ketubot 14b relates an incident of a young girl raped when descending to fill a jug from a spring. The identity of the rapist was unknown but his lineage had halachic repercussions. Based on a majority of residents, they assumed the rapist was someone of unflawed lineage. On Ketubot 15a, Rav Nachman, quoting Rav, explained to his student Rava that this incident happened during the time of the wagons of Tzippori – that is, when a contingent of men of unflawed lineage happened there. (Rashi explains that the time of the wagons is market day, when סיעות ושיירות, caravans from elsewhere arrived there.) That way, there was both the majority of residents and the majority of caravan members to rely upon.
Here in Avodah Zarah, the selfsame Rava could channel this idea of valid caravan members. Since they somehow knew (perhaps based on timing) that the thieves came to Pumbedita, they arrived by caravan, and most caravan participants coming to Pumbedita were Jews. Therefore, the opened wine would not be yayin nesech. This explanation fits with the typical meaning of שיירא instead of inventing a new term; is consistent with Rava as decisor; and fits with the thieves arriving at Pumbedita – which is why I find it compelling.