2.2 Bound morphemes. . . . 43 2.3 Relations . . . . 49 2.4 Linking. . . . 58 2.5 Naive grammar. . . . 66
Our goal is to develop a semantic theory that is equally suitable for the lexical ma-terial (words) and for the larger constructions (sentences) put together from these. In 2.1 we begin with the system of lexical categories that are in generative grammar rou-tinely used as preterminals mediating between syntax and the lexicon. Morphology is discussed in2.2, where subdirect composition is introduced. This notion is further de-veloped in 2.3, where the geometric view is expanded from the standard word vectors and the voronoids introduced in Chapter 1 to include non-vectorial elements that ex-press binary relations. These eigenspace techniques receive further use in 2.4, where some crucial relational devices of syntactic theory,thematic relations, deep cases, and k¯arakasare addressed. How much of syntax can be reconstructed with these is discussed in2.5.
2.1 Lexical categories and subcategories
Whether a universal system of lexical categories exists is still a widely debated question.
Bloomfield,1933, and more recently Kaufman,2009argued that certain languages like Tagalog have only one category. But the notion that there are at least three major cate-gories that are universal, nouns, verbs, and adjectives, has been broadly defended (Baker, 2003; Chung,2012; Haspelmath,2021).4lang subdivides verbs into two categories:
intransitive U and transitive V; retaining the standard N for noun; A for adjective; and also uses D for aDverb; and G for Grammatical formative.
While this rough categorization has proven useful for seeking bindings in the original 4 and in other languages, there is no theoretical claim associated to these categories,
nei-© The Author(s) 2023 39
A. Kornai, Vector Semantics, Cognitive Technologies, https://doi.org/10.1007/978-981-19-5607-2_2
40 2 From morphology to syntax
ther the universal claim that all languages would manifest these categories (or at least, or at most, these), nor the (four)language-particular claim that these categories are some-how necessary/sufficient for capturing the data. In fact, 4lang is a semantic system, and it says remarkably little about the system of lexical categories and subcategories, be they defined by morphological or syntactic cooccurrences. If anything, our findings lend support to the thesis of Wierzbicka,2000that cross-linguistic identification of lexical categories is to be achieved via prototypes rather than by abstract class meanings.
To the extent that none of the six lexical categories U,V,N,A,D,G is ever referred to by any definition or rule,4lang holds fast to the autonomy of syntax thesis (by and large, one can think of the system as operating on categoryless roots). The categories are listed with each entry only to help the English-speaking user distinguish between e.g.
cook fo3l coquitur gotowac1_sie1 822 U get heat
cook fo3z coquo gotowac1 825 V
=agt make <food>, ins_ heat
cook szaka1cs coquus kucharz 2152 N person, <profession>, make food
Here, and in most cases, the other three languages actually manifest the distinction either morphologically or in the choice of stem, but readers familiar with the largely (perhaps fully?) universal distinction between intransitive and transitive verbs and nouns can read off the distinction between the three senses ofcookfrom the 7th column containing the categoriesU, V, andN.
This small example already displays some of the vexing problems of morphology that we need to consider here. First, whichever these three we take as basic, in English it would require phonologically null affixes to obtain the other two. Second, using these six categories creates a lot of ambiguity where there doesn’t seem to be any, e.g. between verbal and nominal interpretation of English noun-verbs such asdivorce, suggesting that six categories are too many. Third, there are obvious meaning distinctions e.g. between agent, action, and abstract nouns that share the category N: compare acook‘the person who does the cooking’ to ashoot which means ‘preparing a segment of a film’ rather than ‘the person who does the shooting’ oraddition‘mathematical operation’ toaddition
‘an extra room that is added to a building’. This widespread phenomenon would suggest that six categories are too few.
In regards to zero affization,4langrefrains from stating the categorial signature of elements even when it is obvious, e.g. that-ize(for which see2.2) is NÑV, producing a transitive verb from a nominal base (see Lieber,1992that in productive uses the resulting verb must be transitive). In informal contexts, where human readability is a concern, we go a step further and feel free to enhance English paraphrases by category-changing nonzero formatives such as be, that, a/an, the, to, -ly . . . with the goal of making the English syntax come out right.
Pure category-changing, be it performed by zero or non-zero affixes, is modeled by change in the head (first, distinguished element) of a definition: consider
2.1 Lexical categories and subcategories 41
official hivalatalos publicus oficjalny 1065 A at authority
official tisztviselo3 officialis urze1dnik 2398 N person, has authority
When this process is regular, as with agentive-er, the suffix morpheme is given the same status in4langas any free-standing word would:
-er -o1 -tor/-trix -ac1/ic1 3627 G
stem_-er is_a =agt, "__-er" mark_ stem_
which will in turn yield definitions such as
buyer vevo3 emptor kupujac1y 3628 N
=agt, buy, -er/3627
renter be1rlo3 conductor dzierz1awca 3632 N
=agt, rent, -er/3627
seller elado1 venditor sprzedaja1cy 3629 N
=agt sell, -er/3627
We started with affixation because this area (which we will discuss in a more systematic fashion in 2.2) offers laboratory-pure examples of change of syntactic category with-out change of meaning and, in the case of zero affixation, withwith-out change in form. We have seen that4langdoesn’t offer a full account of such phenomena. Needless to say, traditional lexicographic practice is no better off, with definitions often led by vague cat-egory phrasesused to, of or about, to be, someone who, relating to, done as, a way of, according to, to make, something that, a type of, the process of, . . . which contribute very little beyond a hint for the syntactic type – we have already discussed a rather striking example, the verbbe, in1.4.
Traditionally, the idea of lexical categories, a notion that we would reconstruct in a purely syntactic means, seeS19:4.2and6.3, come hand in hand with the idea ofclass meanings, ‘the meaning common to all forms belonging to the same form class’. This is a highly contentious idea, of which Bloomfield,1933(Sec 16.2) has this to say:
The school grammar tells us, for instance, that a noun is ‘the name of a person, place, or thing’. This definition presupposes more philosophical and scientific knowledge than the human race can command, and implies, further, that the form-classes of a language agree with the classifications that would be made by a philosopher or scientist. Is fire, for instance, a thing? For over a century, physi-cists have believed it to be an action or process rather than a thing: under this view, the verb burn is more appropriate than the noun fire. Our language sup-plies the adjective hot, the noun heat, and the verb to heat, for what physicists believe to be a movement of particles in a body. (. . . ) Class meanings, like all other meanings, elude the linguist’s power of definition, and in general do not coincide with the meanings of strictly defined technical terms. To accept defini-tions of meaning, which at best are makeshifts, in place of an identification in formal terms, is to abandon scientific discourse.
42 2 From morphology to syntax
On the other hand, there is something greatly appealing about the idea of conceptual parts of speech, and a type theory of one sort or another lies at the heart of many mod-ern developments in philosophy, mathematics, and computer science. Here we compare 4langto the ontological classification developed by Jackendoff,1983, where eight ma-jor categories are distinguished: Thing, Event, State, Action, Place, Path, Property, and Amount.
ThingsIn1.4we definedobjectby the cluster of properties that physical objects have:
thing, <has colour>, has shape, has weight, <has surface>, has position, <lack life>. Jackendoff takes a more cognitively inspired view and singles out individuated entities within the visual field as the central aspect of his definition. However, it seems clear that he would assignkidney, liver, lung, . . . and other objects which rarely occur in the visual field (indeed, they are characteristically hidden, internal) as Things.
On the one hand, it is very clear that Bloomfield is right, we still don’t have sufficient philosophical and scientific knowledge to formally define what a Thing must be. On the other hand, it is also clear that people will show remarkable inter-annotator agreement if we ask them whether some X is an object or not, very much including the fact that Jackendoff’s ‘Thing’ and theobjectof4langdesignate pretty much the same class of entities.
We also grant full ontological status to abstract nominals like happiness or equi-lateral triangle. Since Jackendoff is committed to the same kind of theory that we are propounding, where meanings are concepts, ideas, things in the head (see6.3), confer-ring first class citizenship on abstract nominals is not a problem for him, but in other settings the task is highly nontrivial (see Zalta,1983for a well worked out proposal, and Moltmann,2013for a different approach).
Events, ActionsWhile we have much to say about events, actions, and event structure in3.2 and6.1, we don’t have separate ontological categories for events or actions as such. Rather, we usemattersas a convenient term that covers both Things, Events, and Actions in the Jackendovian sense, and we steer clear of the philosophical issues of whether events exist (occur) ‘in the world’.
Contemporary philosophical theory is near unanimous in granting existence (onto-logical status) to Things, but far from unanimous in the treatment ofeventsor actions.
An important consequence of our parsimonious stance is that we see nominalization pro-cesses as purely morphosyntactic, with no corresponding change in meaning: we treat breatheandbreath, divorce(V) anddivorce(N) as alike.
States, Properties Again, philosophers are near-unanimous that sensory qualia exist, and we follow suit here in granting them ontological status. However, the distinction between these two classes is too subtle. Even the core examples, emotional states like anger, fear, sorrow, joy, . . . are trivial to view as properties, and core properties likered, smelly, triangular, . . . are in turn easily conceptualized as Things. In fact, States like feelings are generally treated as belonging in a well-circumscribed subclass of Things,