Dear WG list,

At last I've found time to join in the discussion of Ancient Greek, which
I've been longing to do since Chet started it off two weeks ago. It seems
to me that he's right: the Greek middle data are a challenge for any theory
of morphology because they straddle the division between inflection and
derivation in a rather uncomfortable way. 

On the one hand they are clearly part of the inflectional system, because
the morphological reflexes of 'middle' are mixed up, inextricably, with
those for clear inflectional categories such as tense and subject-agreement.

But on the other hand, they are clearly derivational in as much as their
valency and semantics are exceptional. This can't be handled as mere
inflectional morphology by the morphological system that he and I wrote
about in our Lingua paper because there's no mechanism by which an
inflection can override a lexeme's default characteristics because neither
isa the other: e.g. WANT:past isa both WANT and past, but past does not isa
WANT, nor vice versa, so WANT:past can only inherit from them both if
they're 100% compatible (i.e. by simple unification, not by default
inheritance). That's why I don't think we can use Jasper's suggestion:
"Unexceptional middles must inherit from two categories: their lexeme and
the verbal ('inflectional') category MIDDLE, which helps  decide their
form, valency, meaning, etc." 

In contrast, derivation works by relating two different lexemes to one
another, and allows them to have conflicting properties. E.g. SPEAKER is
the 'N>V' (noun-derived-from-a-verb) of SPEAK, with similarities and
differences that are controlled by a default 'N>V' relationship which adds
-er to the stem, turns it into a noun, changes the sense from an action to
the person who does the action, etc etc. This is obviously the machinery
that's needed for Greek middles, but it also seems right for ordinary
passives (in English or any other language). I'll come back to passives below.

So, how do we ride these two horses that are going (apparently) in opposite
directions? I suggest that the answer may, in fact, be simple: we treat
them as *both* an inflection and a derivation. Here's how:

1. We recognise Middle as an inflection, alongside Past, etc. The
morphological rules will refer to various combinations of Middle and other
inflectional categories, but that's normal in inflectional languages like
Greek - the inflectional realisation for Past interacts with other
inflections for subject-person etc. By default a verb is not middle, but
Middle is one of the available inflections. But the only characteristics
inheritable from Middle are morphological; there's no syntax or semantics
because middles have exceptional syntax and semantics, and as I said before
this would conflict irresolubly with that of the lexeme.

2. We also recognise Middle-of as a derivational relationship which links a
verb lexeme V to a distinct lexeme, its middle M. One of the
characteristics of M is that it isa (the inflection) Middle, but it has
other characteristics, semantic and syntactic, which override those of V.
That's the *regular, default* entry for the Middle-of relationship.
However, given this possibility of overriding it's not surprising that some
of these derived lexemes become permanent and acquire even more irregular
meanings, such as the ones that Chet lists (e.g. peitho: I persuade to vs
its middle, peithomai: I obey). 

I think the same machinery is needed for passives in English: inflections
in terms of their morphology (even more clearly than in Greek, because by
default their inflection is the same as that for past tense) and
derivations in terms of their linking. As Chet points out, passives are
clearly derived just by derivation in some languages (e.g. in Beja, there's
a passive morpheme which combines with the active stem, and then you add
all the usual inflections to this derived passive stem). 

Derivations would also be useful in handling the Spanish data Joe supplied
(ir   = go vs irse = leave), though there I don't think there's any
interaction with inflection, because I still think (pace lots of people)
that reflexives contain a reflexive clitic pronoun, not an inflection. As
Joe says, SE 'creates a new lexeme', whose distinctive characteristics
include (a) the need for a reflexive pronoun and (b) its distinctive meaning. 

What I'm suggesting on a more general level is that derivation and
inflection are logically quite different - one allows overriding, the other
doesn't; one is a relationship, the other is a word-type. But there's no
reason why they shouldn't interact, which produces complicated arrangements
like Greek middles. This is different from Jasper's interesting
counter-suggestion: 

"I'm suggesting that perhaps the distinction between inflection and 
derivation is not a property of the grammar. You can provide a list 
of properties commonly associated with inflections and a similar 
list for derivations, but examples like these show some infl. and 
some deriv. properties. Then you have to say for each case (bit of 
data) which properties from each list are applicable. Then you can 
do away with the lists because they don't buy you anything (you've 
just separately enumerated the properties of the category in hand 
anyway)."

What do youse (I like that!) think?

Dick


Richard (= Dick) Hudson

Phonetics and Linguistics, University College London, 
Gower Street, London WC1E  6BT.
+44(0)171 419 3152; fax +44(0)171 383 4108;
http://www.phon.ucl.ac.uk/home/dick/home.htm


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%