I'm told

[bwald]

----------------------------Original message----------------------------
With regard to the issue of the emergence of the indirect passive in English,
Cynthia Allen notes:
 
> The collapse of the accusative/dative distinction
>did not immediately result in the introduction of the indirect passive.
>The loss of the accusative/dative distinction happened in most dialects by
>the early part of the 13th century, but no genuine indirect passives are to
>be found at this time.  There followed a period of about 125 years in which
>although the recipient and the were unmarked, they could occur in either
>order-either 'he gave the king a gift' (the only order possible in ModE) or
>'he gave a gift the king'.
 
It should be added that there remains a residue of this process in some
(British) English dialects.  That is, with a few verbs, esp. the
"arch-dative" verb 'give' (but also "show"), the variation in object order
has continued into the twentieth century (and even has a coherent
geographic distribution according to the British dialect atlas reflecting
the rural mid 20th c.)  It seems to be largely if not exclusively limited
to the inanimate pronominal theme "it", as well as to certain verbs, e.g.,
"give/show it me" (maybe even the recipient has to be a pronoun).
 
I have long been interested in this issue, and also why the areal
distribution is what it is, and why IO-DO order has been the trend for
fixed order in most dialects.   Cynthia mentions that the last problem is
related to pragmatic features of topicality in discourse which favors the
animate= indirect object as more often topical and given than the inanimate
= direct object, so the ordering is related to grammaticalisation of
pragmatic information structure of given before new in specific contexts.
This is no doubt involved, but the issue is quite complex and has a number
of features I do not sufficiently understand (yet), such as how and whether
the survival of DO-IO order is constrained by choice of verb, e.g.,
grammatical formations like "cook it me" for "cook me it" (some "gimme it"
speakers will even balk at "cook me it").  To my knowledge, such historical
details remain to be resolved.
 
(NB.  As I have observed in print, the effect of inanimacy on exophoric
deictic uses of pronouns is striking.  Thus, "look at HIM" or any other
animate, but NOT "look at IT" -- instead "look at THAT".  For "them" it
matters whether animate or inanimate.  "look at THEM" (animate) but "look
at THOSE" (inanimate).  This seems to be on the cusp between pragmatics and
grammaticalisation, having to do with the greater informational load of
demonstratives compared to pronouns and the pragmatic relation of such
things to animacy status.  This is related to the probable preferential
survival of "give it me" to "give them me", and, I think, plays an indirect
but important role in the geographical distribution of "give it me", but my
account would be digressive here.  Suffice it to say that "it" shuns
stress, except in metalinguistic contexts, more than any other pronoun,
with the possible exception of "them" in inanimate contexts, cf. what
happened to etymological "it" in Scandinavian.)
 
Without checking, I think I have noted some cases like "cook it me" in
Middle English texts, but I am not clear on how and when "Verb IO DO" =
"Verb DO *FOR* IO (as opposed to ...*TO* IO) arose, or what constraints
there are on choice of verb in such constructions, cf. "build me it", but,
I think NOT "manufacture me it"(?), etc.  Hence, it is not clear to me
whether "benefactive" IO-DO constructions should be historically
distinguished from "dative" IO-DO constructions, but I think so.  (Uh, in
context, by IO I mean where the *unmarked* Indirect Object).
 
In any case, the lexical angle occurred to me because of some facts in some
Bantu languages, where historically passivisations akin to English indirect
passivisation are commonplace, and in some languages have even evolved to
exclude direct (theme) passivisation when there is more than one "object".
This is relevant to Cynthia's comment:
 
>.... (late 14th century), the indirect passives appear
>because grammatically, the recipient is indistinguishable from the theme.
 
The key word is "grammatically".  C's passage implies equation of
grammatical object with unmarked argument, hence accusative/dative where
these are no longer overtly distinguished.  Nevertheless, why should
speakers reinterpret non-themes as grammatical objects (and thus passivise
them), absence of marking notwithstanding?  I would suggest that the
correlation between accusative marking and "theme" as a semantic concept
was never invariant, even in Old English, and was very much dependent on
the *lexical* verb (either for completely formal reasons, as far as
speakers were concerned, or dependent on the particular meaning of the
verb, e.g., "use" takes an instrument not a theme as object, etc etc) so
that there was already notable reliance on accusative marking as an
arbitrary license for passivisation.  With the loss of the distinction
between accusative and dative marking the situation was greatly
exacerbated, so that one line of development was to ignore role, previously
not often reliably marked anyway, and allow passivisation to eventually
expand to all unmarked objects.  Historical English scholars are quick to
point out that early English word order allowed "impersonal passives",
e.g., "the boy-DAT was given a book-ACC", where it is actually still "book"
that is passivised, to be reinterpreted as "the boy-NOM ..."  (German still
approximates English indirect passivisation this way.)  So that is usually
offered as another factor occasioned by case loss/merger.  It is, I think,
not an inevitable development, since, as I noted above, there seems to be a
lexical dimension to the survival of the unmarked DO-IO order (where it
survives) according to the verb, and also to benefactive (unmarked) IO-DO
order according to the verb.  The latter, in particular, contrasts with the
other line of development in which prepositional marking becomes mandatory
with some verbs (and at least preferred in such additional cases as "he
gave YOU to me" rather than "he gave me YOU", where inanimacy does not
mitigate).
 
(NB also alternative marking in English, as in such commonly discussed
cases as "he sprayed the wall with paint" and "he sprayed paint all over
the wall".  This development seems to be in line with the hesitation with
which English decides between the prepositional vs. unmarking of case
relations.)
 
Interestingly, English shows further development from indirect
passivisation to prepositional passivisation, as in: this bed was eaten
potato chips in, etc.  (I think I have the chronological order right, if it
is really clear.  Does the preference for "the boy was given a book" over
"the boy was given a book TO" have anything to do with historical order of
stabilisation of the two types of passive?)  So, eventually (in fact,
fairly quickly) passivisation transcended whether role is marked or not.
Again, this does not seem to have been inevitable, just possible
(obviously, since it occurred).  Maybe Cynthia can comment further on the
historical relationship between these two processes in English and whether
there was anything else favoring the generalisation of passsivisation,
which I suspect was not an inevitable one.
 
The lexical dimension continues to intrigue me.  Embedded in a quite
different grammar (but with many points of similarity) is the lexical
difference between two verbs "give" in Umbundu, an Angolan Bantu language.
One verb, iha, allows either theme or recipient passivisation, so that
either argument can function as unmarked objects of that verb.  This verb
can be traced back to Proto and Pre-Bantu, and allows recipient
passivisation in ALL Bantu languages (which retain passivisation as a
productive process).  In contrast, another verb, eca, has the same meaning
but only allows theme passivisation, and the recipient must be
prepositionally marked in all contexts.  I don't know where it came from,
only that it is underived in Umbundu.  The verbs seem to differ only in
this lexical property.  Although iha reflects the Proto/Pre-Bantu
situation, eca behaves like derived verbs in benefactive and dative
contexts, allowing only the theme to passivise.  Such constraints on
passivisation are quite general in West Bantu, but contrast with East
Bantu, where either both passivisations are allowed or the opposite
constraint on passivisation (favoring "indirect") has developed for various
reasons.
 
Back to English, retention of unmarked DO-IO order with *some* verbs in
*some* varieties of English seems to be lexically determined.  My
impression, needing further investigation, is that they are limited to
relatively high frequency monosyllabic (Germanic) verbs which *imply a
non-theme argument* (whatever that turns out to mean), e.g., "give",
"show", maybe "send".  Problematic to me is how "do", "make", "cook" and
other *benefactively* used verbs without prepositional marking historically
fit into this pattern.  Unlike the other type, they do not *lexically*
imply a non-theme argument.  Cynthia may be able to shed further light on
the history of this benefactive pattern, which seems to be syntactic rather
than lexical, as in "do me that favour" (?? "do me it"//????? "do it me" --
for "give it me" dialects).
 
Cynthia continues with respect to cross-linguistic generalisations:
 
>I am not competent to comment on the Japanese facts, but will only say that
>I am not claiming that it is impossible for a language to have an indirect
>passive without formal identity of the recipient and the theme-only that in
>a language like English in which only direct objects passivized originally,
>it is possible for the indirect passive to arise when former indirect
>objects are reanalysed as direct.
 
Bantu evolution supports this claim, which amounts to saying that unmarked
arguments are "potential" objects (perversely, this even applies to
post-posed subjects in at least one Bantu language -- though in such a case
it is not reflected in passivisation, of course, but in other grammatical
processes.)   At the same time, Bantu evolution does not support IE
evolution in concern for preserving explicit marking of case
relations/thematic roles.  I have argued elsewhere that large areas of
Bantu allowed "case" ambiguity to arise by preferrably grammaticalising
topicality hierarchies (e.g., human > animal > inanimate).  This goes
against conventional wisdom concerning the role of case relations in
constraining grammatical change. Thus, what may play a role in the
direction of English change (i.e., grammatical compensation for weakening
of inflectional case marking) is not to be taken for granted for all
languages (esp those like Bantu, and Niger-Congo more generally, which do
not have a history of case inflection but have intricate topic/focus
ordering strategies).  Apart from that, as Cynthia and many other careful
investigators of the history of English have noted, reactions to weakening
or loss of case-marking are much more complex and historically much more
sluggish (according to the record) than theories glorifying "preservation"
of case-relations (as an "essential" feature of "language") would lead one
to suppose.