re interrogative verbs + interrogative relators + indefiniteordinals

David Gil gil at EVA.MPG.DE
Fri Mar 30 12:17:56 UTC 2001

Since I find myself in agreement with much but not all of what Matthew
writes, let me try to restate the issues as clearly as I can.

Matthew is right that "whatever generalizations are relevant MUST make
reference to the syntactic category of the interrogative word itself and
not just the properties of the set of possible answers".  So let me
hazard the following two logically independent conjectures pertaining to
the WH word and its answers respectively:

(1)	Grammatical category of WH expression itself:

noun > noun satellite (adj, det, etc) > verb > adposition, affix

(2)	Semantic category of appropriate answer

"lexical meanings" > "grammatical meanings"

(a)  larger cardinality > smaller cardinality
(b)  concrete > abstract

In (1) and (2) above, the ">" symbol means that items to the left are
more likely to occur in languages than items on the right.

Conjecture (1) suggests that NP interrogatives are more common than
adjectival or determiner ones which are more common than verbal ones
which are more common than adpositional or inflectional ones.  Among
other things, this generalization captures Matthew's "impression, having
examined "wh"-questions in over 500 languages, that single words meaning
"how many" are much more common than interrogative verbs, despite the
fact that verbs are generally a more open category than numerals."  It
predicts -- contrary to what might have been entailed by my previous
message -- that there will be languages like English with determiner
interrogatives but no verbal interrogatives even though determiners are
a small closed class while verbs are an open one.  As Matthew correctly
observed, "there is more going on here than just the openness of classes".

But the point of my previous message is that openness nevertheless DOES
matter:  the grammatical category of the WH expression alone, as per
(1), is not enough, and we also have to take into consideration the
range of possible answers, as per (2).  Consider the following paradigm
from English:

(3) 	What does John like?
	appropriate answer:
	THING (concrete or abstract), possibly also other kinds of entities

(4)	Which one does John like?
	appropriate answer:
	THING (concrete)

(5)	Where is John?
	appropriate answer:

(6)	How many does John want?
	appropriate answer:
	QUANTITY (countable)

And so forth.  For all WH words in English, whatever their own
grammatical properties, the set of appropriate answers appears to have
the properties suggested in (2).  Specifically, the set of appropriate
answers is a semantically definable set consisting of meanings of the
sort generally characterized as "lexical" rather than "grammatical".
Accordingly, the set of appropriate answers is generally infinite;
moreover, it *tends* -- though this is obviously not always the case --
to consist of meanings that are more concrete than abstract.

With regard to cardinality, Matthew writes that "David's claim is
probably vacuous since there are probably an unlimited number of
responses to any conceivable class of questions."   But here I beg to
differ:  if indeed it is the case that all attested WH words across
languages allow for an infinitude of possible answers, this is not a
logical necessity but rather an interesting substantive claim on what
constitutes a possible WH word in a language.

To appreciate the empirical force of this, and, more generally, of the
conjecture in (2), consider the impossibility of question expressions
such as the following:

(7)	WHblick does John eat the apple?
	appropriate answer:
Thus, an appropriate answer to (7) would express one of the 12
tense/aspect combinations of English, eg. "John ate the apple" / "John
is eating the apple" / "John will eat the apple" etc.  In other words,
its appropriate answers would be a finite and small number of very
abstract grammatical meanings.  I'd be very surprised to find a language
that had a word like WHblick above.  Or:

(8)	WHbleeck did John buy the apple of Mary
	appropriate answer:

An appropriate answer to (8) would be something like "John bought the
apple for Mary" / "John bought the apple from Mary" / "John bought the
apple with Mary" / "John bought the apple because of Mary", etc.  But
crucially, it would be inappropriate to answer with longer,
compositional expressions such as "John bought the apple to surprise
Mary" / "John bought the apple while crouching a couple of feet behind
Mary", etc.  (If this latter condition were relaxed, there would in fact
be possible cases of WHbleeck, eg. the Riau Indonesian "ngapain" which I
described in a previous message.)

So the generalization seems to be that you can have questions whose
answers convey lexical meanings but not questions whose answers convey
grammatical meanings.  This is what (2) is trying to say.

This seems to me to be right, with one possible exception: yes/no
questions.  Here, the range of possible answers is clearly highly
abstract, and is prototypically limited to exactly two items, yes and
no.  Though in actual fact, one can respond to a yes/no question with
any number of answers along the scale from yes to no, eg. "maybe",
"almost certainly", "with as much likelihood as it will snow sometime
this week", etc.  So maybe yes/no questions aren't a serious
counterexample to (2) after all.  But even if it is a counterexample, it
seems to be the only one.

So if conjectures (1) and (2) are empirically valid, then the question
arises: WHY should things be this way?  Suggestions?


David Gil

Department of Linguistics
Max Planck Institute for Evolutionary Anthropology
Inselstrasse 22, D-04103 Leipzig, Germany

Telephone: 49-341-9952321
Fax: 49-341-9952119
Email: gil at

More information about the Lingtyp mailing list