How did Rava pronounce Hebrew?

The Gemara Berakhot 15b discusses the requirement to enunciate words carefully when reciting qeriat shema. In doing so we are unwittingly given some interesting linguistic evidence about Hebrew pronunciation at the time of the 'amora Rava (3rd-4th c.).

To whit:

תני רב עובדיה קמיה דרבא ולמדתם שיהא למודך תם שיתן ריוח בין הדבקים עני רבא בתריה כגון על לבבך על לבבכם בכל לבבך בכל לבבכם עשב בשדך ואבדתם מהרה הכנף פתיל אתכם מארץ

R. Obadiah recited in the presence of Raba: 'And ye shall teach them': as much as to say thy teaching must be faultless by making a pause 'between the joints'. For instance, said Raba, supplementing his words 'Al lebabeka [upon thy heart], 'al lebabekem [upon your heart], Bekol lebabeka [with all thy heart], bekol lebabekem [with all your heart], 'esebh be-sadeka [grass in thy field], wa-'abaddetem meherah [and ye shall perish speedily], ha-kanaph pesil [the corner a thread], etthkem me-erez [you from the land]. (trans. Soncino)

The plain meaning is probably that all these words must be pronounced carefully, with a space between them, so that the words remain distinct from one another. After all, if you pronounce words that end and begin with the same sound it is easy to mash them into one word. The issue becomes even more acute when dealing with certain kind of words. For example, if one says אלקיכם אמת quickly one ends up with something like אלקיכמת, which can sound like אלקיכם מת--a definite no-no!

Therefore we must take extra care not to allow words which tend to combine into one from doing so.

Rava's eight paired examples are interesting.

They are

1. 'Al lebabeka
2. 'al lebabekem
3. Bekol lebabeka
4. bekol lebabekem
5. 'esebh be-sadeka
6. wa-'abaddetem meherah
7. ha-kanaph pesil
8. etthkem me-erez

The first four present no questions. L + L = LL. The 6th and 8th present no question. M + M = M & Ms MM.

But what exactly is the danger of the 5th and 7th? How can the sound of an undageshed bet slur into a dageshed one? (The v of 'voice' should't be able to slur into the b of 'boys.') And how can an undageshed peh slur into a dageshed one? (The f of feather shouldn't be able to slur into the the p of 'pleasant.')

Okay, obviously these sounds are related. In fact, the soft forms are the spirantization of the hard ones. When we wonder about a v and a b it isn't as if we are wondering about how a z and an l manage to slur into each other. Nevertheless it is clear that the relationship between Rava's pairs in his 6th and 8th example is different from the pairs in the other six, involving like ls and like ms.

So what gives?

The answer might well be as follows:

1. For Rava there was no distinction between the bhet and bet.
   
2. Or there was a slighter distinction than the one we know of.
   
3. Or these words did not then have the distinction which appears in the later masoretic codices and indicated by a dagesh which would indicate that the underlying rules which establish when letters are pronounced hard or soft were not yet firmly established.

Source: Werner Weinberg 'Observations About the Pronunciation of Hebrew in Rabbinic Sources,' Hebrew Union College Annual Issue: 56 (1985) p.117

Is it possible that the same distinctions existed then?