open letter to andrea lunsford 3

Arnold Zwicky zwicky at CSLI.STANFORD.EDU
Thu Jun 12 16:39:52 UTC 2003

        Before I leave Harper's and Follett for the murkier world of
Pence, I'd like to float one further proposal that's hinted at by the
wording in a few of the handbooks (Follett: "A noun in the possessive
case... is seldom a competent antecedent of a pronoun", with an
in-Mary's-book illustration of the problem), though I don't think any
of them actually comes right out with it.  This is the idea that
finding an antecedent for a personal pronoun is just a matter of
finding the first noun before the pronoun.  On this view, prenominal
possessive antecedents (like the "Mary" of "Mary's father") would
always be out because in backtracking from the pronoun you'd always
get to the head noun (here, "father") first.

        Such a proposal would illustrate the elevating of a
proscription not only to a rule of grammar (Just Say No), but to a
*necessary* rule of grammar: what I'll call the It Has To Be So
tactic, in which a usageist maintains that some principle
*necessitates* a proscription.

        It Has To Be So arguments have been given against usages that
are innovative (I can't resist mentioning Richard Grant White's
chapter-long rant against the spread of the analytic progressive
passive, as in "The bridge is being built" - one of the great
entertainments of 19th-century usageism), or are associated with
certain social groups (by region, class, race, ethnicity, and the
like), or are characteristic of informal language, or of spoken rather
than written language - anything that is not established formal
written standard language (which is privileged as the True Language).

        It Has To Be So arguments always appear in masquerade, either
as arguments "from logic" or as arguments from such communicative
values as clarity, brevity, and avoiding both ambiguity and
redundancy.  Nevertheless, they all embody crucial unexpressed
assumptions about the nature of syntax, semantics, pragmatics, and the
relationships among these.  (R.G. White maintained that the analytic
progressive passive is logically incoherent, because it has two
occurrences of the verb _be_ in the same clause; this was a
self-evident absurdity to him, so he didn't explore the matter
further.)  They deny that what counts as established formal written
standard language is just a historical accident and assert instead
that efwsl is as it is because it is the best, or in fact the only,
possible world.

        I'm not going to rehearse here the responses of professional
linguists to this edifice of fuzzy thinking; there's a lot to be said,
though much of it is wearyingly familiar.  I do, however, want to
point out just how risky It Has To Be So arguments are.

        They are obviously risky from a practical point of view:
telling people that the language they learned from childhood and use
as their medium of communication in daily life is intrinsically
illogical, incoherent, ineffective, etc. is a bad strategy for a
teacher of efwsl.  It's just so likely to produce reactions of anger,
shame, or rejection in the students.

        But they It Has To Be So arguments are also desperately
vulnerable as *arguments*, because they depend on the validity of the
principles they appeal to, yet in my experience these never survive
critical scrutiny.

        Let's go back to antecedent finding and the PAP.  The relevant
background principles include a claim that readers (and, presumably,
hearers) find the antecedent for a pronoun by backtracking from that
pronoun; a claim that readers/hearers consider only the most recent
candidate; and the claim that this candidate is a noun.  The first two
claims are preposterous, and the third is significantly inexact.
(This might all seem to you like beating a never-living horse, but
bear with me.  I'm going to introduce some important ideas that I'll
need later on.)

There's no reason to think that a pronoun generally causes readers to
scan back in text for an antecedent, and for spoken language there's
not a shred of evidence in favor of backtracking through a verbatim
store of previous discourse.  Instead, what happens is that
expressions establish discourse referents in the reader's/hearer's
mind (in particular, an NP usually establishes some entity as a
discourse referent) and the unfolding context guides the reader/hearer
in ranking such referents as more or less salient, so that these
referents are more or less available as antecedents when a suitable
pronoun appears.

        As for the proposal that you pick the absolutely most recent
noun: even if there were no such things as possessive antecedents, the
claim is just false, and this is a truth universally acknowledged.
The handbook advice not to let an antecedent get too distant from its
pronoun, ceteris paribus, is a good caution - because distant
potential antecedents are likely to be very much backgrounded, or even
beyond the horizon - but that's a very big ceteris paribus.

        Finally, the expressions that establish discourse referents
aren't single nouns (unless these happen to constitute NPs on their
own), but *noun phrases*.  Even if you *were* backtracking from the
"her" of "Mary's father admires her", searching for an antecedent, you
wouldn't get to a potential antecedent "father" first; you'd get to
two potential antecedents, "Mary" and "Mary's father", at the same
time, when you hit the word "Mary".  So the third assumption, that
antecedents of pronouns are single nouns - this is an assumption about
the syntax-semantics interface - is seriously flawed.

        To recap: the PAP would follow as a logical necessity from
suitably strong assumptions about antecedent finding, but these
assumptions themselves are crap.

        I believe that this case study provides a general model for
demolishing It Has To Be So arguments: look at the background
assumptions, and they'll always turn out to be at best dubious.  This
is just what we'd expect by modus tollens, since the predictions that
follow from the assumptions are empirically false.

        But this was a hypothetical instance of It Has To Be So in an
effective-writing context.  So far as I know, no one has actually
espoused a backtracking-to-most-recent-noun principle.  However, other
It Has To Be So arguments have been wielded in favor of the PAP, and
widely, at least since Pence.  I have to say that I was more than a
little surprised to discover such arguments in the world of
effective-writing advice, since they are just *so* hard-edged,

        There are two symptoms of It Has To Be So arguments in a
formulation of the PAP: Adjective Function, a reference to possessives
"being modifiers", "functioning as adjectives", or even simply *being*
adjectives; and Implicitness, a reference to antecedents that are
"(merely) implied", "implicit", "imagined", or "hidden" in
possessives, perhaps with an explicit contrast to antecedents that are
"(actually) named".  These are two sides of the same coin.  Some
formulations of the PAP allude to Adjective Function, some to
Implicitness, some to both (and, of course, some just enunciate the
PAP without justification; it's a natural law).

        Follett invokes Adjective Function, in a modifier I elided
when quoting him above:
        A noun in the possessive case, being functionally
          an adjective, is seldom a competent antecedent of a pronoun.
Pence's formulation (Pence & Emery 1963, p. 222) at first seems
innocuous, and is also hedged by "as a rule":
        A pronoun should not, as a rule, refer to a noun used as
          a possessive modifier.
(Like Follett, Pence gives exactly one example, and once again it's of
the in-Mary's-book sort: "In Winston Churchill's public addresses he
has given...").  The operative word here is "modifier", since earlier
in the book, in the chapter on "functions of the parts of speech in a
sentence" we are told that among the uses of nouns and pronouns come
"possessives as adjectival modifiers" (p. 34).  So, not just as
modifiers, but as *adjectival* modifiers.

        Pence at least makes a distinction between the adjective as a
category and adjectival function.  Various later handbooks get this
wrong; they seem not to distinguish syntactic categories and syntactic
functions, as if roughly a century of linguists' work on the
fundamental concepts of syntax had just passed them by.  In any case,
the hedged formulations of Follett and Pence give way, within a few
decades, to rigid formulations in which the PAP is said to be
*necessitated* by the nature of possessives.  It Has To Be So.

        The unexamined assumptions here are that possessives are
adjectivals, perhaps even adjectives (Adjective Function); that
consequently possessives do not actually instantiate referring
expressions, but merely allude to entities, in the same way that the
adjective (and adjectival) "Mexican" doesn't actually instantiate the
referring expression "Mexico", but merely alludes to the entity Mexico
(Implicitness); and that it is individual noun words, and not NPs that
refer to entities and serve as antecedents for pronominals.  (A few
manuals manage to get the NP thing right - Martha Kolln's _Rhetorical
Grammar_ (4th ed., 2003) p. 262 - is an honorable exception.)

        The result is that violations of the PAP are predicted, from
first principles, to be as bad as "anaphoric island" violations like
        I like Mexican food, but it's a terrible place.
(Indeed, PAP violations are often lumped together with anaphoric
island violations of various sorts.)  Putting aside the objection that
even anaphoric island violations can be argued to be
semantic/pragmatic in nature, not syntactic, the background
assumptions from which the PAP is predicted are all flawed: in
particular, possessives are not adjectivals (their syntactic function
is that of a determiner; this is easy to show) and certainly not
adjectives (they belong to the category NP, not AdjP); and possessives
routinely serve as referring expressions.  There is no basis for an It
Has To Be So argument for the PAP, and that's a damn good thing.

        Now, Andrea, I am the bearer of further bad news about L&C's
formulation of the PAP, which I repeat here:
        Though an adjective or possessive may clearly imply a
          noun antecedent, it does not serve as a clear antecedent.
In light of my discussion of It Just Has To Be claims, we can see two
new problems here.  The lesser of the two is the reference to nouns,
rather than NPs, as antecedents.  The greater problem is the lumping
of possessives together with adjectives (suggesting Adjective
Function), along with the oblique allusion to Implicitness in "may
clearly imply".  Goodness knows what students make of this
proscription; I'd expect a student who actually paid attention to what
the manual says to ask questions like the following: What is "clearly
imply" contrasted with?  Isn't "Welty" actually *in* "Welty's story"?
And what does it mean to say that a pronoun implies a *word*?  (This
is very far from ordinary-language uses of the verb "imply", where
what is implied is a proposition, or sometimes an entity.  There's
some special technical use of "imply" here, but its meaning isn't
clear from context.)

        This formulation has the whiff of an It Just Has To Be support
for the PAP, in the hints at Adjective Function and Implicitness, but
it has no explicit justification.  Presumably, the L&C formulation
is the descendant of other handbooks' more explicit statements; much
of the content got lost as the proscription was handed down from
generation to generation.  Did you and Connors even think about what
your formulation meant, beyond its applicability to in-Mary's-book

        A really smart student is going to be puzzled indeed when they
check out "possessive nouns" in L&C's index and are directed to page
687, in the section "for multilingual writers" written by Franklin
E. Horowitz (rather than you and Connors).  Horowitz gets a lot of
things right, or almost right.  He correctly classifies "possessive
nouns and noun phrases" as a type of *determiner*, not adjective; but
that, of course, makes your lumping possessives together with
adjectives mysterious.  Horowitz also almost gets it right with his
reference to "possessive nouns and noun phrases".  The technical
concept he wants here is NP, embracing single-word NPs as well as
multi-word ones, but he continues to use a non-technical sense of
"phrase", for multi-word expressions, and so has to describe a general
class by means of a disjunction.  (There's absolutely no reason to
avoid technical terms here.  A handbook on grammar and usage is going
to be jam-packed with technical terminology.  The trick is to pick
the right terms and to use them consistently.)

        Horowitz isn't perfect.  Possessive pronouns are listed as
determiners, but in a separate entry from possessive NPs; the entry is
unlabeled, and merely enumerates the possessive pronouns, so that it
fails to connect them to other possessive NPs.  And, though this is
surely not Horowitz's fault, the index entry is to "possessive nouns",
not "possessive nouns and noun phrases" (or, better still, "possessive
noun phrases").

        Things could be worse.  Kolln leaves the world of (potentially
ambiguous) Mary's-mother and (potentially too backgrounded)
in-Mary's-book examples entirely, and concentrates instead on examples
of the Mary's-father type, which are usually unproblematic for anyone
who hasn't been bashed on the head with a usage manual, things like
        The neighbor's front porch is covered with trash,
          but he refuses to clean it up.
(She gives four examples of this type, plus one of a subtype of
anaphoric island violation, anaphora to the first N in an N+N
compound: "Last summer I didn't get to a single baseball game, even
though it's my favorite sport."  It's hard for me to find fault with
even this example, except on theoretical grounds.)  She gives an
Adjective Funcion justification for her version of the PAP - the
antecedent of the pronoun "is not a complete noun phrase; it's only a
noun modifier" - and then connects this justification to effective
writing by baldly asserting that adjective function produces "a
problem of fuzziness that could easily cause a blip in the reader's
comprehension."  I wonder what students make of all this; do they
just abandon their intuitive feel for their own language, or do they
put the rule down as one of those arbitrary and inexplicable things
you have to memorize for the test?

        (Kolln does give one example that strikes me as poor writing,
*out of context*: "My sister's boyfriend works for a meat-packing
plant.  She's a vegetarian."  But this is wobbly because the first
sentence is clearly *about* the boyfriend, not the sister.
Salience/foregrounding/topicality strikes again.  And, as usual, if
you adjust the context, the example improves enormously: "At least
one person in my family has a conflict between belief and emotional
attachment.  My sister's boyfriend works for a meat-packing plant,
yet she's a vegetarian."  What's so distressing, to me at any rate,
is that Kolln, who is, like L&C, so much concerned with context,
the flow of discourse, rhetorical effects, and so on, should fall
back on rigid and theoretically based constraints at the sentence

        One more.  This is Wilma R. Ebbitt & David R. Ebbitt, _Index
to English_, 5th ed. (NY: Oxford University Press, 1990), p. 217 on
"clear reference", enumerating "situations in which reference lacks
the precision that college writing demands."  The third entry in
the list is "When the pronoun refers to a noun used as a possessive
or as an adjective", an allusion to Adjective Function.  *But*...
there are only two illustrations of the rule, and neither of them
involves a personal pronoun with a possessive antecedent.  The second
is an anaphoric island violation of the N+N compound type:
        Nancy longed for a chinchilla coat, though she
          wouldn't have dreamed of killing one.
The first involves a *relative* pronoun problem, of a sort that turns
out to be quite different from any of the possessive antecedent types
we've looked at so far (though a few other manuals castigate examples
of this sort, which are hard to find in writing, though they occur
occasionally in speech):
        Bill was skipping stones across the swimming hole.
          One cut open a young girl's head who was swimming under
The problem here is not so much with pronoun antecedents but with the
conditions under which a relative clause ("who was swimming under
water") can be extraposed away from its head ("a young girl").  An
interesting issue, but only distantly connected to possessive
antecedents for personal pronouns.

        Once again, I wonder what students make of this.  If a
Mary's-father or an in-Mary's-book example appeared on the exam, would
they understand that the Ebbitt & Ebbitt rule was supposed to cover
it?  How could they possibly?

        But enough.  I'm dismayed to see how often the handbooks,
yours included, fail to take context and discourse organization into
consideration, how often they reach for Just Say No and It Has To Be
So accounts of syntactic phenomena (working from genuine cases of
inept writing, but failing to examine the actual practice of elite
writers, and disregarding the scholarship of works like MWDEU), and
how little they seem to know about fundamental notions of syntax (I'm
not talking about the details of formal theoretical frameworks, but
the sort of basic technical vocabulary and observations assembled in
Huddleston & Pullum's CGEL) and basic semantics/pragmatics.  The
advice the handbooks offer could be much more comprehensible, and much
more useful, if their writers would only look beyond the insular
traditions of usage teaching.


P.S.  I see I've failed to provide an example of a possessive
antecedent for a personal pronoun from the works of Jane Austen, who
serves linguists as a sort of gold standard for elite writing; if it's
in Austen, you can pretty much bet it's standard English.  I'm sure
there are examples, but I'll leave it as an exercise for other readers
of this letter to find them.

More information about the Ads-l mailing list