In a theoretically fascinating review of Elizur Bar-Asher Siegel’s Introduction to the Grammar of Jewish Babylonian Aramaic, (Münster: Ugarit-Verlag, 2013), Yeshiva University biblicist and linguist Aaron Koller describes the radically different possible starting points for studying the language of the Bavli–and sometimes, how we can even know what it says:
Bar-Asher Siegal argues…that there are no really reliable manuscripts for the Bavli…Whereas [Matthew] Morgenstern contends that there are such texts, but they are texts not seriously studied by Kutscher – namely, the Early Eastern manuscripts – Bar-Asher Siegal’s claim is that there are no manuscripts that can be utilized as the “primary texts” due to their exceptional reliability. What that means in practical terms is that each and every linguistic phenomenon, from orthography to morphology to syntax, has to be investigated thoroughly and independently in all of the manuscripts…
[By contrast with Morgernstern] Eljakim Wajsberg has said that there are some “actually good” manuscripts of the Bavli, but apparently is confident in only two: a Yemenite manuscript in Oxford of Sukkah, and a Geniza fragment of Bava Metsia‘.
This would mean that every truly significant linguistic problem in the Bavli must be studied in somewhat the same way as Dennis Pardee describes Ugaritic: a hermeneutic circle between the philology, which raises questions that encourage us to reconsider the manuscripts, and the manuscript readings, which can help us rethink the philology. The circle need not be vicious: the point is rather that we need to make arguments rather than simply point to forms or readings.
Among the fascinating problems relevant to the study of the Hebrew Bible and its language, two further issues stand out: the question of internal linguistic diversity, and the minimal nature of verbal marking, which does not specify tense, mood and aspect in the same way as some Indo-European languages seem to.
On the first, Koller notes:
One of the theoretical points made by the author which constantly accompany the analysis is that the text of the Bavli is a problematic witness for [Jewish Babylonian Aramaic]. This is for two reasons. First, the Bavli seems to reflect different dialects…
but second and more fundamentally,
the Bavli may not reflect JBA because scribes often try to mask developments within colloquial language in their written texts. When manuscripts differ between a more archaic and a later form, it often cannot be known whether the text originally had the older form and this was later mistakenly updated, or whether it original reflected the newer form and was then mistakenly “corrected” to the older form.
Here he cites a striking example from Bar-Asher Siegal
If this situation reflects the actual forms of JBA, then clearly the two phenomena could not reflect one stage in one language, since either /y/ elided or its morphological role was reanalyzed. Thus, the various forms should either reflect different historical stages or two dialects. This is another example where JBA regularly reflects more than one linguistic system.
Second, the fact that after centuries of study scholars still do not agree on such basic points about the verbal system as whether the prefix-form of the verb marks indicative or irrealis should provoke some rethinking:
Here Bar-Asher Siegal parts ways with most older presentations, and denies that the prefix conjugation expresses the irrealis mood; he also argues that the verb הו”י, in different forms, serves to mark the tense of imperfective verbs as past or future.
To my mind, the problem and the most promising direction for its solution is summed up in the incisive statements of Holger Gzella, “Some general remarks on interactions between aspect, modality, and evidentiality in Biblical Hebrew” (Folia Orientalia, 2012:225-232, in line with earlier comments of Ed Greenstein on Ugaritic). Verbal forms in languages with lighter verbal morphology like the classical Semitic ones should not be straightjacketed into the categories of old-fashioned school grammar, but rather may map onto broader functional ranges where tense, mood and aspect intersect in various ways, rather than simple one-to-one functions.
…because if not I’m just gonna keep adding to it ‘ti it is.
 Joosten 2002’s arguments, based on the theory of Kuryłowicz 1973 that tense is the first axis of opposition in all verbal systems, would imply that archaic Hebrew could not have had a preterite-imperfective-anterior system. But it is based on the extremely powerful assumption that no system lacking morphological tense-marking can have morphological aspect-marking. There are at least two serious objections to this assumption. The first is that it does not consider widely-known counterevidence, such as the best-documented early stage of an ancient Semitic language, Old Babylonian, in which a preterite iprus ~ durative iparras ~ perfect iptaras opposition is widely recognized (e.g. Huenergard and Woods 2004). Second, it proposes to confirm Kuryłowicz’s theory with an examination of the very system under dispute (2002:52), which runs the risk of an obscurum per obscurius explanation, since two centuries of scholarly debate have not brought agreement on this patter (for an important example of the continuing debate see Pardee 2012 with n23). At the risk of attempting to cut a large and old Gordian knot with a single quick blow, I would question how scholars as careful as Joosten and Pardee, with precisely the same data set and the benefit of 200 years of scholarship from Gesenius to Driver, can continue come to flatly opposite conclusions about the entire CBH prose verbal system. And I would suggest that the solution lies in the fact that ancient Semitic deixis, both spatial (cf. the frequent lack of a morphological distal-proximal (“this” vs. “that”) distinction, e.g. in Hebrew, Aramaic and Akkadian demonstrative pronouns) and verbal, tends to rely more heavily on implicature than the Indo-European languages from which all of these scholars from Gesenius and Driver to Joosten and Pardee derive their categories. The knot would then lie in the unavoidable pragmatics of implicature: precisely because of their linguistic and analytical competence, these scholars will necessarily produce comprehensive and consistent readings of the (from an Indo-European viewpoint) always morphologically under-indicated tense and aspect of CBH.