Surely these are empirical questions, and surely part of the problem is
the IPA assumption that a sound either "is" one thing or another.  We
can debate these issues in terms of impressionistic listening and
proprioception all we want, and that is probably useful for generating
ideas and hypotheses.  But ultimately the problem is that speech
sounds aren't going to fit neatly into the boxes provided by the IPA
chart and (more importantly) are going to vary from occasion to occasion
in ways that require quantitative/continuous description rather than
qualitative/categorical.

Just the other day Jim Scobbie and I were looking at ultrasound records of
post-vocalic dark L in Scottish English and it seemed pretty clear that
there is WITHIN-speaker variation in the amount of alveolar contact, with
zero contact a definite option.  However, in all cases we could clearly
see a kind of concave tongue shape, with backing/raising of the tongue
body AND a raising of the tongue tip toward the alveolar ridge (but not
necessarily reaching it).  So we did seem to be looking at a configuration
that would create lateral airflow and the appropriate spectral patterns,
even in the absence of actual alveolar contact.  Possibly some
Browman-&-Goldstein-type concepts will help here (gestural reduction,
etc.), but it seems pointless to debate whether the sound "is" alveolar or
velar.

I'm not suggesting that "dark l" is necessarily the same in Scottish
English and American English, and I do agree with the sugestion that some
North American speakers have an /l/ phoneme that sounds very much as if it
involves mostly dorsal constriction of some sort.  (I have two
acquaintances who realise /l/ like this, both male, educated, and
Northeastern, and neither with Eastern European background, for what it's
worth.)  All I'm saying is that with issues like this we are reaching the
limits of the scientific usefulness of traditional IPA classification (and
the limits of what traditional trained-ear methods can achieve) and are
moving into the realm where only quantitative instrumental investigation
will yield the answers to our questions.

Bob Ladd