<< >> TOC
As noted above, natural kind terms such as child and proper
names can be freely converted into one another, as the examples in (16) show:
a. Sit down, Child! (common noun -> proper name)
b. John is a liar. I tell you, a John is not to be trusted! (proper name -> common noun)
The conversion possibilities between prototypical common nouns (CN) and proper
names (PN) pose a problem for any assumption of distinct categorial status,
semantic or syntactic. In languages with overt determiners (like English,
German, French, Spanish etc.), conversion is exacted by simply adding or
removing a determiner. As I will discuss below, namehood thus seems to be simply
a pragmatically determined interpretation of overt syntactic structure, rather
than a covertly supplied property of particular lexical items. If it was such a
property, e.g. a feature like [+/- proper name], the semantics would need to
specify two mechanisms: one to override a [+ proper name] marking for cases of
PN-to-CN conversion and one to supply this feature for cases of CN-to-PN
conversion. A unitary treatment seems preferable in any case.
question that presents itself is whether one of the two categories <e> or
<<e, t>, t> is better suited for free conversion than the other. In
order for <e> to become a set, it needs to be type-wrapped, i.e. a
mechanism needs to apply that takes the information of the input type and
reinterprets it as a member or subset of the output type. Thus, it is a
non-trivial operation. If <<e, t>, t> serves as the base type, on
the other hand, “conversion” to something that looks like <e> is truly
trivial: after all, a set with cardinality one differs only in type from
<e>, not in content.
Levi-Strauss argued from a conceptual perspective that proper names and common
nouns form a continuum with proper names representing an extreme case, he may
have been right on the mark in set-theoretic terms: a set with cardinality one
is arguably the smallest set that allows contentful interpretation. The
theoretically smallest set is obviously the empty one, but my hunch at this
point is that the empty set is not interpretable, but rather leads to a
computational error condition: all “referential” nominals seem to either
denote a set with a cardinality of at least one or at least require such a
set for comparison or evaluation purposes (cf. the discussion of specificity
below). The assumption that all nominals denote sets is thus computationally
In this context,
I would like to bring up an important detail that the literature surprisingly
seems to have missed so far:
whenever proper names are discussed, first names are used in the examples. Of
course, most Johns will also have a last name, e.g. Smith. The reason why this
obvious fact has been ignored for the most part is, I speculate, that last names
are trivially proper names, but also clearly serve as class terms. While the
explicit natural kind use of a first name (cf. 16b) does seem to require some
unusual computation, the same is not true of last names, as can be seen in (17):
(17) Yesterday, I met a Smith, namely John. I had often heard people talk about the Smiths, but I had never met one myself.
The proponents of names as individual terms could of course
counter that what we use as last names are, in fact, natural kind terms, but
that it is first names that the term ‘proper name’ refers to. However, there
is evidence that there really is no conceptual difference between Smith and John
(cf. Levi-Strauss 1962), as the following dialogs show:
A: I met John yesterday.
B: Which one?
A: John Smith.
A: I met Smith yesterday.
B: Which one?
A: John Smith.
In both cases, speaker A mentions a proper name, obviously
intending to denote one particular discourse referent. Speaker B, however, makes
it clear that at least in the domain of discourse accessible to her, the proper
name denotes a set. Thus, A has to narrow down the search space by adding
another identifying proper name. The set relations can thus be represented as
Thus, both first
and last names can possibly serve to identify a superset or a subset. While last
names are more likely to have superset status in most discourse situations, they
do not seem to have an inherent quality that makes them more suitable for set
descriptions than first names. Of course, a similar argument holds for first
names: there is no inherent reason other than their common usage for their
increased likelihood to denote a subset of a larger set described by a last
name. In fact, certain sociolects show a preference for using last names for
vocative purposes. In these cases, dialogs like (19) are probably the rule
rather than the exception. Of course, the whole issue of first vs. last name is
in itself highly culture specific.
For the rest of
this paper, I will follow common usage and concentrate on first names, since it
is my goal to tickle out the characteristics of the extreme case of the nominal
<< >> TOC
© Philipp Strazny 1998
 In making my point on set relations between first and last names I draw upon Levi-Strauss (1957), but his work can hardly be considered mainstream in semantic theory.