>[dialects or single language]
>How does one decide what
> picture to imagine?
Some further thoughts:
My own feeling is that it is a question of how secure our reconstruction is.
We can reconstruct a period with dialects fairly securely for many items,
whereas before that (if there is a before) we are less secure for more
items, and earlier even less secure, and so on.
Are the oddities of Hittite (and Tocharian?) sufficiently explained by
saying they separated early? Or are they better explained by saying they
come from a slightly different dialect group, with different uses of shared
resources?
If we think of a single, unified, common ancestor, then we have to accept
loss of certain features in some dialects, (such as aorist -s- in Germanic)
after the period of separation. If we allow dialects, then the absence of
these features can be taken further back in time.
If we assume a single unified common language, then innovations (such as the
augment) must occur after the separation, so they cannot be from "PIE"
resources - PIE no longer exists. If we allow dialects, then forms such as
*h1e- make sense.
Peter