I'm told

[bwald]

----------------------------Original message----------------------------
I was very pleased to receive Cynthia's more extended comments on the
indirect passive and related phenomena in the history of English.
Actually, while I was waiting I became intrigued with the general
relationship between the indirect and impersonal passives, not only in the
history of English, but in Eurafrasian (Nostratic?) ACC(usative) languages
in general, e.g., early I-E and Semitic (Classical Arabic and OT Hebrew),
in German, Uralic and Altaic.  The whole nature of both passivisation and
impersonal (non-subject, non-passive) varies in interesting ways across
these contiguous case-marking families, leading to insight about what's
common and what's distinctive about any particular line of evolution, of
course including Middle English.
This informs my following comments on Cynthia's message.
 
I was encouraged that Cynthia confirmed my basic understanding of various
developments in the history of English.  Indeed, I have read Denison and
some of Cynthia's earlier writings (among others), but I do not always
trust my memory, and there's also always the possibility that an author has
changed their (!) mind about something for some reason or other.
Therefore, it is entirely appropriate for Cynthia to refer me to some of
her writings, implying that she feels they express her current ideas to the
extent that she understands my interests.  The only disappointment I had is
that she did not respond to my interest in the historical status of the XY
= Y FOR X construction, as in "cook me it" = "cook it for me".  I mentioned
that I thought I had noted ME examples of the type "cook it me", but I
hadn't studied the problem closely enough to determine whether this remnant
of historical X-DAT Y-ACC case-marking and word order has become (or always
was) distinct from the type which has become "XY = Y TO X", e.g., "give me
it = give it to me".  As far as I understand the literature, this is not an
issue.  It is not singled out one way or another.  So, I would assume that
analysts don't feel there is a basis to make a distinction between "X TO Y"
and "X FOR Y" in relation to "XY".  But I would not take it for granted.
So I'm looking for explicit (dis)confirmation of that assumption.  The
difference could be in the relation of the two prepositions to thematic
role (or whatever you want to call it) and/or in the degree of independence
of that role from any particular lexical verb.  That remains a concern for
me, because of the level of comparison I want to make for syntactic
constructions associated with ACC case-marking in Eurafrasian languages.
The issue involves ACC and inevitably the ACC/NOM! relationship (and thus,
for the moment, excludes the ergative case-marking languages that confront
Eurafrasia).  The objective is to explore whether the development of the ME
indirect passive was indeed more likely than not under the circumstances
(hence, what ARE the circumstances? cf. Semitic also develops NOM-marked
indirect passivisation), or whether the linguistic (as opposed to, say,
social) conditions were ever sufficient to favor this direction of change
over others.
 
Meanwhile, Cynthia's following comments particularly interested me.
 
>... when I was talking about 'bare NP objects' I
>was referring to nominal, rather than pronominal objects.....
 
I think that can be taken for granted.  And it occurred to me that
lexical(ly filled) NPs (or whatever you want to call them) establish the
lead for pronouns, despite the fact that the pronouns maintain a NOM/ACC
(ACC = older ACC/DAT) distinction throughout the period of development of
the indirect passive.  The latter is a fact that is quite striking, though
we generally take it for granted because it is so familiar to us.  It is
all the more striking, I suggested, in view of the much longer time period
for which (pronominal) pre-verbal ACCs (from older DATs) remained with
specific NON-passivised verbs.  Something about the passive encouraged a
much more rapid completion of the change.  Some would relate this to the
idea that change in the nature of "grammatical subject" in English
(corresponding somewhat to changes in the nature of "grammatical object")
began to occur much earlier than "grammaticalisation" of word order to
preserve distinctions once indicated by (NOM/ACC) case-marking.  The
implication is that ("long") before word order became a prominent
role-expressing device, case-marking continued to play a role (in directing
NPs to "subjecthood" and "objecthood"), perhaps because of pronominal
case-marking, where pronouns are generally more common than lexical NPs in
relevant constructions in discourse.
 
Cynthia mentions the following progression:
 
>1....a pronominal object still always preceded a nominal one,
>and when both were pronouns, the normal order was Th REC...
 
That is a striking fact in view of the commonality of the order IO DO among
languages.  Th REC is DO IO.  I might understand the underlying principle
if Th was more likely to be pronominal than REC, but I do not think that is
the case, i.e.,  I do not think it is based directly on an
"information-status" order (cntr. PRO NP).  I can only note that the same
order occurred historically for the French object clitics, and still does
for third persons (but was otherwise reversed to a  "topicality" order 1,2
> 3).  My vague notion about why Th REC order in such cases is that DAT
came to be considered a more "marked" case than ACC in the history of
Romance (I think more generally than French) and fit into a "markedness"
order similar to the information-status order.  Some might just call it a
"heaviness (of information)" order (the DAT marker is also phonologically
more prominent than the ACC, and remains so in current French, lui vs.
le/la).
 
>2. The first few true examples of 'he was given a book' are found later on
>in the 14thC.
 
This shows an increasing association of NOM with "subject" -- in the case
of pronouns.  It is probably difficult to securely establish a priority
between this change and a shift of agreement for the passivised verb when
the subject is a lexical NP and thus indistinguishable for "oblique" and
NOM.  The issue is where "the boys was given a book" > "the boys WERE given
a book" fits in to "them was given a book" > "they were given a book".  I
think NOM presupposes verb agreement, but unmarked preverbal NPs do not.
(I could anticipate attestation of "them were given books", but not "they
was..." in the relevant period.)
 
>3. By the early 16thC, there are no more examples in my texts of ProTh NREC
>order, although TH REC is the only order found when both objects are
>pronouns.... He gave it him but *he gave it the king.
 
IO-DO order expands but is resisted by PRO-PRO, originally motivated by the
higher "markedess" of DAT over ACC.  IO-DO may be associated with the
higher of DAT/IO as more favorable than DO to both humans (topicality) and
pronominal reference (information status).  (I am aware of unrelated
languages in which IO-DO order generalised, conversely, to allow NP-PRO.)
 
>4. I US, Australian and some varieties of British English, the order REC TH
>has completely generalised so that it is the only possible order even with
>two pronouns.
 
There are even details here, if "they cooked it me" is less widespread than
"they gave it me". Where the REC TH order becomes totally general, the
markedness of IO focus increases, since it is preferably further marked by
a preposition when it is stressed, i.e., "they cooked it *for* ME" rather
than "they cooked ME it", etc.
 
Cynthia goes on to say:
 
> What I actually said was simply that the REC TH order (with two
>nominals) became progressively more dominant.
 
I'm interested in the how's and why's.
 
Cynthia suggests:
 
>...the position of the old (nominal) IO became identical to the position
>of a DO of a >mono-transitive verb: directly postverbal.  The explanation
>that I suggest for why a
>language-learner would make such an equation crucially involves a level of
>grammatical relations, as opposed to semantic relations, and assumes that
>listeners like to use processing strategies that make use of either
>case-marking or grammatical relations.  My idea is that once the order of
>two nominal objects became fixed, the language-learner came up with this
>processing strategy:
 
The progress of word order as a pervasive strategy to the exclusion of
case-marking continues to interest everyone concerned with English
historical syntax.  It took rather a long time for the IO-DO word order to
completely take over PRO-PRO contexts.  It suggests that case-marking has
not been given up without a protracted struggle, as we still know.  In a
general way, the consequences of the former case-marking on English (as
I-E) grammar were so pervasive that it is taking the evolution of *most*
varieties of English a great deal of time to undo them all, and they seem
to implicate a great deal of grammatical detail.  I'm not really interested
in predicting here, but the trends up to now are quite compelling. (NB.
such late developments as pronominal compound objects like "him and I"
further weaken the link between case-marking and grammatical role).
 
>.....  I think that the reason why 'He was given a
>book to' is not favored in English is simply that if you are going to make
>a recipient be the subject, you might as well make it a core role (the
>object) to start with, instead of making it an oblique.
 
That grammatical analysis need not conflict with the notion that there is a
grammatical conflict between passivising an "object" and giving it the
focus that a preposition establishes.  Sometimes the conflict cannot be
avoided, where passivisation is favored for some reason, but a case-marked
strategy has not survived, e.g., with locatives ("the bed was eaten
potatoes chips IN").  It remains interesting to contemplate why non-marking
has survived for certain DAT and benefactives (perhaps even further
developed for the latter).  It does not seem to be only an arbitrary
lexical matter (esp in the case of benefactives).
 
>... Certainly, what
>Denison refers to as the 'complex prepostional passive' (e.g. he was taken
>advantage of) seems to have entered the language a bit later than the
>'indirect' passive, but not that much longer-there are convincing examples
>from the early 15th C.
 
If it is securely "later", it shows accommodation of the prepositional
object to other objects, all as unmarked for case.  In any case, it
strengthens the association of passivisation (and topicality) with NOM, to
the extent that even PRO objects of prepositions are shifted to NOM
marking.  The stranding of prepositions is involved in such passivisation,
but has a distinct history as a grammatical strategy in English.  cf.
Scandinavian strands prepositions, but stops short of the indirect passive.
Case rearrangement is more radical in English.
 
Cynthia notes that:
 
>...the reality seems to
>be that the Experiencer of *please* was always unambiguously the object
>except in this construction, where it starts showing up as a nominative
>subject in the early 16thC.  ...No ambiguity-driven syntactic reanalysis
>can have been
>involved for two reasons (1) the 'reanalysis' takes place much too late and
>(2) there was no model for a syntactic reanalysis; i.e. sentences like 'the
>king will do as the king pleases' are not found until 'the king will do as
>he pleases' are also found; before this, 'the king will do as it pleases
>the king (or him)' is the only possibility.
 
There's a misunderstanding here.  The reanalysis is not too late, given
that we are dealing with lexical diffusion of DAT > SUBJ for experiencers,
and the OED (Onions, I guess) notes the innovation in "please" as
corresponding to "like".  It cites 1500 Dunbar (Northern) "your melody he
pleases [= likes] not til hear" (in modern spelling).  The equivalent use
of "like" (current use) is cited as early as 1200, but the archaic
"inverted" use (from DAT of experiencer) continues as late as 1616 in Ben
Johnson: "if this play does not LIKE [= please], the Devil's in it" and
even *1784* in Cowper: "they...howl and war as LIKES [= pleases] them".
 
Cynthia suggests:
>...I think that we have to look at the semantics of
>the construction here.  This experiencer is in fact more agentive than the
>experiencer of please usually is, because the construction is explicitly
>saying that I am in change of my own pleasure.
 
This is generally true, at least currently.   Generally, "like" does seem
to have less of a "volitional" element in contexts that readily come to
mind than "please".  I'm not sure this is lexical rather than pragmatic.
("please" < "if you PLEASE" < "if it please you",  as a politeness marker
in requests appeals to volition similar to "you wanna..." = "would you
please..."; "would you like to..." is used as a suggestion or invitation,
not a request).
 
She continues:
>So I think the experiencer
>started to get treated as a possible subject because it was similar
>semantically to a typical agentive subject, not because there was any
>confusion as to its grammatical relation .
 
To me, what experiencers and agents have most in common is that they are
extremely favorable to humans (and other animates).  The association of
experiencers with NOM and/or subject is quite common in I-E, although DAT,
significantly, was more favorable to experiencers in earlier I-E languages
than it became later.   (Hence DAT > NOM or DAT/NOM variation with lexical
conditioning occurs more widely and earlier in I-E than any of the other
developments discussed here.  There is plenty of precedent for the shift in
"please", but usually not so faint-hearted).  Sense verbs like "feel",
"smell" and  "taste"  (and other verbs with those meanings in other I-E
languages) have agentive/experiencer capability, e.g.,  "I smelled a
flower" (cntr. "sniff" is more biased to agent subj); "I felt a flower" vs.
"I felt a pain", etc.  NOM is not strictly agentive in NOM/ACC languages;
that is how they differ from ergative languages.  (Ironically, Halliday
playfully used "ergative" with respect to English for developing further
intransitive uses of transitive verbs, e.g., "the newspaper sold quickly",
rather than the reverse development, e.g., "they disappeared the subversive
newspaper".)
 
>.... I analyse the Experiencer of the impersonal 'methinks'
>etc. as a subject, despite the object case.  Lexical case marking to
>subjects was retained because it was clearly distinct from the normal
>(structural) case for subjects.
 
I agree that this is a useful analysis.  It shows the separation of
"subject" from NOM as a case marker.  I believe a similar analysis is
sometimes proposed for German, where a dummy subject 'es' (3s neuter) is
required for impersonal passives like "es wird gevochten"  (it fight-PASS)
"there's fighting going on" (more lit. "there's being fought") vs.
"indirect" passives like ihm wird geholfen (him-DAT help-PASS) "he's being
helped", where the need for a dummy SUBJECT is obviated by the preposed DAT
filling the SUBJECT position.  This suggests that "subject" analysis of
preposed DAT in OE, as with verbs like 'deman', could have preceded the
shift to NOM marking.  (Of course, English did not develop the impersonal
passive in the same way as German, and the emergent construction "there's
[[fight]V-ing]N..." is involved with the distinctive and striking
grammatical development of the English gerund.)
 
Cynthia also notes:
We still get impersonal uses like this
>until the early 16thC, although less and less frequently.  Then they are
>reduced to fixed expressions like 'methink(s)'.
 
So I assume that she fixes "please" with EXP/AGENTIVE subject at a period
when "methinks" was already fixed as an expression.  The loss of such
expressions suggests that they were still analysed grammatically in some
way which disfavored their later survival (e.g., as non-NOM "subjects").  I
already noted how "like" holds on with EXP OBJECT well beyond the early
16th c.
(Yes, "methinks" is not historically descended from surviving "think", but
as a fixed expression it is quite equivalent to "I think", which can thus
be viewed as replacing it.)
 
>I'm afraid that I will have tried the patience of any histlingers who have
>bothered to read this far and hereby declare my intention of ceasing to
>repeat what I have already said in print-for the time being, at least!
 
I am less defensive about running on, but I think that there is nothing for
you to apologise for here.  You are informing readers who are interested
but not familiar with the facts and your views.  It is appropriate to use
HIST.LING in this way, just as the question about "I'm told" encourages.
Thanks for your replies.  -- Benji