-
49THE IDEAL SCAFFOLDING OF LANGUAGE Phenomenology and the
Cognitive Sciences 3: 4980, 2004.
2004 Kluwer Academic Publishers. Printed in the Netherlands.
The ideal scaffolding of language: Husserls fourth
LogicalInvestigation in the light of cognitive linguistics
PEER F. BUNDGAARDCenter for Semiotic Research, University of
Aarhus, Denmark (E-mail: [email protected])
Abstract. One of the central issues in linguistics is whether or
not language should be con-sidered a self-contained, autonomous
formal system, essentially reducible to the syntacticalgorithms of
meaning construction (as Chomskyan grammar would have it), or a
holistic-functional system serving the means of expressing
pre-organized intentional contents and thusaccessible with respect
to features and structures pertaining to other cognitive subsystems
orto human experience as such (as Cognitive Linguistics would have
it). The latter claim de-pends critically on the existence of
principles governing the composition of semantic con-tents.
Husserls fourth Logical Investigation is well known as a genuine
precursor forChomskyan grammar. However, I will establish the
heterogeneous character of the Investi-gation and show that the
whole first part of it is devoted to the exposition of a semantic
com-binatorial system cognate to the one elaborated within
Cognitive Linguistics. I will thus showhow theoretical results in
linguistics may serve to corroborate and shed light on those
partsof Husserls Fourth Investigation that have traditionally been
dismissed as vague or simplyignored.
Key words: Logical investigations, cognitive linguistics,
grammar, semantic configurationvs. syntactic combination,
linguistic vs. pre-linguistic structure
Hier interessieren uns nur die Bedeutungen (Husserl, 190001, IV,
7)
Preamble
Edmund Husserls Fourth Logical Investigation,1 The Distinction
betweenIndependent and Dependent Meaning and the Idea of Pure
Grammer, con-stitutes a privileged domain for investigating what
the mutual enlightenment(Gallagher 1997) of phenomenology and the
empirical sciences may specifi-cally consist in. At first, this may
sound surprising since the Fourth Investi-gation is not only the
shortest of the six Investigations, it is also likely to bethe one
that scholars have generally paid less attention to. There are a
coupleof good reasons for this. First of all, it has traditionally
been considered a mereapplication on a specific domain (the realm
of linguistic significations) ofinsights laid down in the preceding
Investigation (on the theory of wholes and
-
50 PEER F. BUNDGAARD
parts); the results obtained as to the specific essence of
language therefore seemto have no general import (contrary to the
mereological principles laid downin the Third Investigation); both
these circumstances may have conveyed an,as it were, parenthetical
or simply illustrative flavor to this Investigation.Finally, even
though Husserl very ambitiously and rather categorically setsout to
lay down the essential tenets of a pure logical grammar
underpinninglanguage as such in the vein of la grammaire pure et
raisonne of Port-Royal he never readdressed the issue, nor further
developed his insights and as-sumptions in his following writings,
at least never in terms of a pure mor-phology (Formenlehre) of
language as such. The Idea of Pure Grammarthus occupies a quite
confined and marginalized position in the landscape ofHusserlian
phenomenology.
It is nevertheless the Investigation that has had most pervasive
and mostimmediate import on empirical sciences (I even wonder if it
has any parallelin modern, post-Galilean philosophy). It is hardly
exaggerated to claim thatit provided the tools and the categorical
prerequisites for the systematic de-velopment, if not simply the
birth, of structural linguistics, and thus linguis-tics in the
modem sense tout court. The great Roman Jakobson never ceasedto
repeat that structural linguistics developed in Moscow and Prague
in the1920s through eager readings and passionate discussions of
Husserls Thirdand Fourth Investigations, and the import on
linguistics of their two basicclaims:
It is possible to consider language as such as an abstract
object in its own right (withoutrecurring to historical,
philological, or psychological concepts). It is possible to
describeand characterize the a priori concepts and laws upon which
it rests, its ideal scaffold-ing (Husserl, Fourth Investigation,
14), and thus systematically expose a pure uni-versal grammar.
The laws in question govern the way linguistic parts combine, by
virtue of determinatetypes of relation, into unified meaningful
wholes.
The expression mutual enlightenment does indeed take on a
pregnant sensein this context. Although Husserl in his
Investigations evidently did providethe conceptual and categorical
framework for, for example, R. Jakobsons andPrince Trubetzkoys
studies in phonology, and although he is a recurrent ref-erence in,
among others, K. Bhlers Theory of Language (1934) then, inreturn,
the latter realize, as it were, Husserls a priori claims as regards
thedomain of language. They do so by systematically specifying,
illustrating, anddemonstrating how such concepts as dependence
relations, relations offoundation, in short structure, are
specified in language, and how they sus-
-
51THE IDEAL SCAFFOLDING OF LANGUAGE
tain and govern the unity of complex linguistic objects at a
great many differ-ent levels. (As we will soon see, Husserls Fourth
Investigation is indeed asclose as can be to being blatantly devoid
of examples and elaborate analysesof concrete linguistic
phenomena.)
I shall not comment any further on the intellectually passionate
affinitiesbetween Husserls Third and Fourth Investigations on the
one hand and struc-tural linguistics on the other. Remarkable texts
have already been written onthis issue.2 In the present paper I
will rather claim the actuality of the two In-vestigations with
respect to a different and much more recent research pro-gram in
linguistics, so-called Cognitive Linguistics (also known as
CognitiveSemantics or Cognitive Grammar). Before coming to grips
with the issue, letus take a brief look at this linguistic research
program.
Cognitive Linguistics comprises works from authors such as M.
Johnson(1987), G. Lakoff (1987), R. Langacker (19871991), Lakoff
and Johnson(1980, 2000), and L. Talmy (2000). All the above authors
call into questioncertain fundamental presuppositions in modem
linguistics, namely: (1) thatlanguage is an autonomous,
self-contained system whose formal nature canand should be studied
strictly in its own terms; (2) that the core property oflanguage is
its syntactic armature; (3) that the characterization of the
rulesgoverning the combination of syntactic categories is to be
considered an es-sential characterization of language as such (cf.
Langacker 1991, chapter 1);and finally that (4) such
characterization lays bare a pure, universal linguistica priori,
prior to any empirically given language and grammar.
In contradistinction to this assertion of the immanent autonomy
or modularityof the linguistic system, cognitive linguists hold a
holistic-functional view onlanguage according to which (1) language
is a cognitive subsystem charac-terized by its essential relations
to other cognitive subsystems (the visual one,notably); (2)
language is not adequately describable in solely syntactic termsand
without reference to cognitive processing in general; (3) language
isgrounded on and serves the purpose of expressing
pre-linguistically organ-ized conceptual material; thus, language
cannot be reduced to any purely lin-guistic a priori, or a
deep-structure linguistic competence, nor to any innatealgorithm
governing the combination of symbols into complex
linguisticsignifications. The task of Cognitive Linguistics
consists then in providingsystematic evidence for the linguistic
subsystems dependency on other cog-nitive subsystems, in
elucidating the nature of the pre-linguistic structures andmental
processing on which it is founded, and, crucially, in showing how
thelatter are grammatically specified in language.
Though the above presentation is very summary and deliberately
vague(certain essential assumptions will be illustrated and
corroborated in later
-
52 PEER F. BUNDGAARD
sections), it does, at least, suggest the existence of
determinate affinities be-tween crucial assumptions in Cognitive
Linguistics and well-known tenets ofHusserlian phenomenology. Among
these, the most fundamental one seemsobviously to be the shared
claim that predicative structure is rooted in ante-predicative
structure, or that linguistically articulated signification is not
ex-haustively describable in its own, grammatical terms, but is
tributary to specificmeaning conferring and meaning fulfilling acts
and the latters essential struc-ture. It is therefore no surprise
that cognitive linguists have explicitly acknowl-edged their debts
to phenomenology; yet rarely, if ever, directly to Husserl;rather,
indirectly, via M. Merleau-Ponty.3
A natural way of approaching the historical and notional
affinities betweenHusserls phenomenology of language and Cognitive
Linguistics would there-fore be to trace back the essential claims
of the latter to those investigationswhere Husserl explicitly
develops hypotheses on the essentially pre-linguis-tic basis of
linguistic signification and acts of signification (say, the First
Logi-cal Investigation or Experience and Judgment). Here, however,
my approachwill be somewhat different: I will instead take the bull
by the horns and con-sider the Fourth Logical Investigation
entirely devoted to determining thespecific essence of language and
compare its results with fundamental hy-potheses within recent
Cognitive Linguistics.
Here is how I will proceed, and why I will do so. The Fourth
Investigationis first and foremost known for, if not establishing
then at least asserting theexistence of a purely autonomous and
essentially syntactic grammatical a pri-ori. It aims to lay down
the laws governing the internal unity and formal con-gruity of
linguistic parts qua tokens of syntactic categories; that is to
say,independently of what language is used to symbolize or mean,
and independ-ently of the organisms (embodied human beings) that
use this symbolic vehi-cle to express their experiences. The claim
could hardly be more opposed tothe basic assumptions of Cognitive
Linguistics, and is, indeed as many schol-ars have observed the
theoretical source itself of Chomskyanism in both lin-guistics and
cognitive science.4
Now, in the first part of this paper, I will show that the above
is only par-tially true. My claim is what only very few scholars
have observed andexplicitly commented on (I only know of one:
Benoist (1999, 2001)) (thatHusserl in the Fourth Investigation does
not abstract one, but two grammati-cal a priori, which are not only
essentially different, but also, in the frame-work of his analysis,
mutually incompatible: a semantic-mereological apriori (valid for
the coherent configuration of linguistic parts qua
partialsignifications in a meaningful whole) and a genuinely
syntactic a priori (validfor the linear combination of linguistic
parts qua tokens of syntactic catego-
-
53THE IDEAL SCAFFOLDING OF LANGUAGE
ries notwithstanding their specific semantics). The first a
priori is devel-oped in the first nine paragraphs of the
Investigation, the second in the remain-ing. The first approach has
as such5 been ignored, neglected, dismissed asvague or considered
as a mere preliminary exercise that only serves the func-tion of
introducing to the fundamental hypotheses of the last section.
One of the aims of this paper is to rehabilitate the first
approach. Initially,this implies that I will have to take exactly
that detour through the realm ofmeaning that Bar-Hillel (1957, p.
369) considered superfluous in order to char-acterize and better
understand Husserls precise aim at the beginning of his
In-vestigation, and in order to show why this approach is
fundamentally incom-patible with the syntactic approach developed
in the last part of the Investigation.Next, it consists in
demonstrating the richness and first and foremost, the
actualrelevance of his semantic analyses, both as regards the study
of languageand as regards the relation between the structure of
language and cognition.
The latter task is at present much easier to fulfill: for almost
thirty years,cognitive linguists and among them most particularly
Leonard Talmy haveindeed provided minute analyses and systematic
evidence corroborating theexistence of the very kind of structures
underpinning the unity of linguisticsignifications that Husserl set
out to explore in his Fourth Investigation. I willtherefore expose
some of the results of this research, notably Talmys workson the
both formal and semantic distinction between lexical and
grammaticalelements of language, which shows striking analogies to
Husserls distinctionbetween categorematic and syncategorematic
expressions and independent anddependent meanings.
In his Investigation, Husserl did inaugurate a fundamental idea
shared byboth Chomskyan grammar and Cognitive Linguistics: the
study of languagetells us a lot about the mind. Yet, I believe that
his point is much more cog-nate to the latter than to the former:
if language reveals anything essential aboutthe mind, it does so
not because the mind is structured like a language, buton the
contrary because language, to the extent that it expresses and
articu-lates what the mind has in mind, is structured like the
mind.
The semantic-mereological a priori: how do partial
significationsconfigure into consistent wholes?
The Fourth Investigation is most particularly known for its
distinction betweensenselessness (Unsinn) and nonsense (Widersinn);
that is to say, between se-quences of words that have no intrinsic
meaning (for instance, formerly arewalking or), and sequences of
words that do make sense whether or not they
-
54 PEER F. BUNDGAARD
are true, formally consistent, or refer to any possible object
of whatever sort(like stubborn quantifiers or electric complex
numbers). The import of thisdistinction on linguistics at the turn
of the century (as well as in the followingdecades) can hardly be
overestimated: it serves the crucial purpose of clearlydelimiting
the genuine object of a pure logical grammar, that is to say:
lin-guistics in the modern sense (whose aim is to lay down the
universal charac-teristics of language). Indeed, however
fundamental, logical laws establishtruth (or falsity) only on the
basis of intrinsically meaningful propositions.Before it can be
considered true or false, a proposition has to make sense,
andcorrelatively it does make sense even if it is flagrantly false.
Therefore, therules governing the compounding of partial
significations and the rules gov-erning the latters formal
consistency are not the same; rather, Husserl claims,there exist a
priori laws, in the realm of pure significations, that establish
thedistinction between what makes sense and what does not make
sense, and theselaws of sense [. . .] direct logic to the
abstractly possible forms of meaning,whose objective value it then
becomes its first task to determine. This logicdoes by setting up
the wholly different laws which distinguish a formally con-sistent
from a formally inconsistent, i.e. absurd, sense (71/334).
The task of the Fourth Investigation (and linguistics in
general) is, in otherwords, to define the laws governing the
compounding of significations, theunity of meaning, i.e., the a
priori patterns in which meanings belonging todifferent semantic
categories can be united to form one meaning, instead ofproducing
chaotic nonsense (49/295). These laws would then make out a
puremorphology (Formenlehre) of significations, in
contradistinction to the purelogical theory of validity.
Yet, however important and rich in consequences it is, this
distinction ismerely evoked in the Introduction, and thoroughly
developed only at a ratherlate stage of the Fourth Investigation
(12 ff.). Until then, the approach is ofa quite different
character. Although the issue is the same according to
whichprinciples do significations combine into a meaningful whole?
it is not ad-dressed in terms of laws or a priori patterns
governing the linear combina-tion of linguistic parts into
well-formed sequences of syntactic categories. Infact, no such
syntactic considerations are displayed in the first paragraphs
ofthe Investigation, and Husserls argument is certainly not guided
by the dis-tinction between Unsinn and Widersinn. The guiding
question is rather thefollowing: given that a complex linguistic
object (say, a sentence, but a com-pound would do just as well) is
made out of signifying parts, how then, accord-ing to which
principles, do these parts combine so as to constitute a
genuinesignificant whole, and not merely a multiplicity of
significations? The tasktherefore consists in determining the types
of relations between parts that
-
55THE IDEAL SCAFFOLDING OF LANGUAGE
warrant their unity. To fulfill this task, Husserl applies the
tools already elabo-rated in the Third Investigation and proceeds
to examine whether linguisticparts are all of the same type, or
whether an essential distinction should bedrawn between them: the
distinction between independent and dependentcontents, the one
Husserl following C. Stumpf found to hold generally be-tween parts
of a whole in the Third Investigation.6
As is well known, independent parts (or Stcke, pieces) are such
that theycan be considered in isolation, whereas dependent parts
(or moments) mustnecessarily occur with other parts of a certain
sort in a whole of a determinatesort. Thus, a horses head, its
tail, its left front leg are such detachable pieces,whereas its
color and shape are non-detachable moments of the whole (nocolor
without extension and vice versa). In this respect moments of a
wholeplay a fundamental role since they are necessarily related to
other parts of adeterminate sort in a whole of a determinate sort
they are founded on themin relations of either unilateral or mutual
dependence: they are, as it were,crystallizations of the system of
relations that bind the parts into one consist-ent whole.7 This is
what Husserl applies to language and meaning: grantedthat
linguistic objects, word-complexes, are made out of different
significantparts, we can now ask whether all parts have the same
kind of signification(in this case sentences are summative wholes
of significant parts) or whetherit is possible to distinguish
between different types of signification, just as itis possible to
distinguish between dependent and independent contents, mo-ments
and pieces (in which case, sentences or complex linguistic objects
areinternally articulated in virtue of determinate types of
relations that it wouldbe possible to characterize).
Husserl emphatically adopts the latter alternative and proceeds
to a three-step determination of what is meant by dependence in the
realm of lin-guistic significations. First he draws an exclusively
grammatical distinctionbetween two general types of linguistic
parts in discourse; then he leads thisdifference back to a
genuinely semantic distinction; and ultimately he foundsthe
linguistic semantic distinction on a cognitive distinction, i.e., a
generaldistinction between types of meaningful contents given in
intentional acts.
Here is how the argument goes: To begin with, Husserl refers to
the oldmedieval parsing of the elements of discourse into two
general subsystems:categorematic and syncategorematic expressions.
The distinction is originallypurely grammatical: sentences are
traditionally said to be compounded bywords that have a proper
signification (generally, nouns, verbs, and adjectiveswere
considered as such), called categorematica, and words (like
prepositions,conjunctions, particles, pronouns, as well as
grammatical units such as inflex-ions, prefixes, suffixes, etc.)
that only take on a signification together with
-
56 PEER F. BUNDGAARD
other words; the latter are called syncategorematica, or
co-signifying words.Husserl now presents the following alternative:
(1) The distinction is indeedmerely grammatical, i.e., only
relative to the particular make-up of each em-pirical language. In
that case, only categorematic words are linked to repre-sentations
proper (they express the latter), while syncategorematic
wordsmerely play a role at a grammatical level, i.e., they only
have a function asconnectors within the linguistic system of
expression, but they do not expressany meaning whatsoever. If so,
the syncategorematic words which help tobuild up [the] expression
are, properly speaking, quite meaningless, only thewhole expression
really has a meaning (54/304). (2) The completeness
orincompleteness of a linguistic expression is a reflection of a
completeness orincompleteness at an underlying semantic level. In
this case, syncategorematicado not simply play a
functional-connecting role at a grammatical level ofexpression,
they have a signification in their own right, yet their
significationis not a complete or independent one, but an
incomplete and dependent one.
Husserl adopts the latter solution: categorematica are semantic
pieces of alinguistic whole, syncategorematica are semantic moments
of a linguisticwhole. The grammatical distinction is an expression
of an essential semanticdistinction, strictly analogous to the
distinction between dependent and inde-pendent contents of
intentional acts. The answer is adopted on the grounds ofthe
following claims, which show not only the comprehensive character
ofHusserl's approach (it addresses the essence of its object,
language, at threedistinct and interrelated levels: a grammatical,
a semantic, and a cognitive one),but also the kind of mereological
principle governing the unity of complexlinguistic objects that he
aims at laying bare. Let us consider the two funda-mental
assumptions.
The first claim follows of course directly from the attribution
of depend-ent significations to syncategorematic expressions:
linguistic complex objectsare simply tokens of complex objects in
general, thus the laws governing theirunity are the same as those
established within the general theory of wholesand parts. Parts are
combined with each other by virtue of their essentialdependent or
independent contents, i.e., by virtue of the type of connectionto
other parts they require. The a priori governing the realm of
linguisticsignifications is therefore a mereological a priori:
Having called syncategorematic meanings dependent, we have
already said where wethink the essence of such meanings lies. In
our enquiries into dependent contents in gen-eral, we have given a
general determination of the concept of dependence: it is this
samedependence that we have to recognize in the field of meaning;
Dependent contents, westated above (Third Investigation, 57), are
contents not able to exist alone, but only asparts of more
comprehensive wholes. This inability [of dependent significations
to exist
-
57THE IDEAL SCAFFOLDING OF LANGUAGE
alone] has its a priori governing ground in the specific
essences of the contents in ques-tion. Each dependence points to a
law to the effect that a content of the sort in question,e.g. sort
can exist only in the context of a whole G(AB...M), where B...M
stand for de-terminate sorts of content. Determinate, we said,
since no law merely asserts connec-tion between the sort A and any
[arbitrary] sort of [context] [. . .] Law involves
specificdeterminateness of context: dependent and independent
variables have spheres limited byfixed generic or specific
characters. We have mainly employed as examples the concretethings
of sensuous intuition. We could, however, have brought in other
fields, those ofact-experiences and their abstract contents
(5859/311312).
The argument is conducted from an entirely a priori point of
view and, at thislevel, it is quite clear. To the extent that we
can identify dependent parts withina linguistic whole, and to the
extent that such parts call for completion, andnot just any
completion, but of a determinate kind, there must exist laws
gov-erning the way in which such parts should be completed, with
what kind ofparts, and in what types of connection. That is to say,
just like dependent con-tents in general, dependent significations
are essentially prior or more funda-mental than independent
significations, since they point at or imply the lawsgoverning
their completion.8
However, if we remember that here Husserl is engaged in a
linguistic in-quiry, we may be entitled to ask for specifications
of this semantic scaffold-ing of language: what does it mean for a
linguistic content, by virtue of itsspecific essence, to require a
context of a determinate sort? In fact, it shouldbe easy to
illustrate this point within the linguistic domain in view of the
factthat in the Third Investigation Husserl provided enlightening
examples of whathe meant by dependent content, relations of
foundation, and relations of uni-lateral or mutual dependence
within the domain of intuition. Yet, in the FourthInvestigation,
there is no linguistic counterpart to, say, the relation
betweencolor and extension such as it is minutely analyzed in the
preceding Investi-gation. In other words, it is quite difficult to
tell how exactly the laws invokedapply to the domain of linguistic
significations, and, certainly, how laws gov-erning the relation
between contents of intuition should rule the unity of lin-guistic
parts.
If we now turn to the second claim underpinning Husserls
argument, wesee that it in fact clearly specifies what is meant by
linguistic content andits specific essence. The main tenet of the
argument is functional. Languageserves the purpose of expressing
representations or experiences (whether ac-tually on-going,
recalled or imagined). Now, experiences are themselves madeout of a
great many partial experiences, combined into a whole or intended
asconnected in a specific way. Husserls point here is simply that
if languagehad not disposed of the necessary means to express both
partial experiences
-
58 PEER F. BUNDGAARD
or representations and the specific way in which these are
intended to com-bine, it would be incapable of expression, and thus
not be language:
Language has not been lead by chance or caprice to express
presentations9 with namesinvolving many words, but by the need to
express suitably a plurality of mutually coher-ing
part-presentations, and dependent presentational forms, within the
enclosed self-suf-ficiency of a presentational unity. Even a
dependent moment, an intentional form ofcombination through which,
e.g., two presentations unite in a third, can find
semanticexpression, it can determine the peculiar meaning-intention
of a word or complex of words.Clearly, we may say that if
presentations, expressible thoughts of any sort whatever, areto
have their faithful reflections in the sphere of
meaning-intentions, then there must be asemantic form which
corresponds to each presentational form. This is in fact an a
prioritruth. And if the verbal resources of language are to be a
faithful mirror of all meaningspossible a priori, then language
must have grammatical forms at its disposal which givedistinct
expression, i.e., sensibly distinct symbolization, to all
distinguishable meaning-forms (5455/304305)
It is worthwhile to stress once again that the approach is
semantic through andthrough. The unity of complex linguistic wholes
is defined in terms of semanticrelations of foundations between
dependent significations and those specificcontexts of meaning in
which they must occur.10 What is more, in the veinof the First
Investigation, the approach is semantic twice rather than
once:syncategorematic expressions have genuine significations
(which must im-ply that these are amenable to semantic
description); the significations arethemselves correlates to
certain pre-linguistic11 contents of meaning acts thatin turn
constitute a complex, configured whole of partial meaning acts, so
thatthe dependence or independence of each signification is a
result of its capac-ity or incapacity to constitute the full,
entire meaning of a concrete act ofmeaning. In short, as Husserl
concludes, a linguistic (dependent or independ-ent) content and its
specific content is an articulation of the content of a
pre-linguistic meaning act. Syncategorematic expressions are thus
linguistic formsthat specify grammatically the content inherent to
a specific mode of inten-tion (or, as quoted above: they are
grammatical forms that express intentionalforms of combination of
partial representations):
A meaning, accordingly, may be called independent when it can
constitute the full, en-tire meaning of a concrete act of meaning,
dependent, when this is not the case. It canthen only be realized
in a dependent part-act in a concrete act of meaning, it can
onlyachieve concreteness in relation to certain other complementary
meanings, it can only existin a meaningful whole. The dependence of
meaning qua meaning thus defined determines,in our view, the
essence of the syncategorematica (59/312).
Once again this claim is not sustained by any example or
concrete analysis ofhow precisely dependent expressions and
significations specify and articu-
-
59THE IDEAL SCAFFOLDING OF LANGUAGE
late pre-linguistic dependent contents of expressible thoughts
or intentionalmodes of combining partial representations. Husserl
does indeed claim thatany example proves the point that
syncategorematica express such inten-tional forms of combination,
but no linguistic example in particular is recruitedto corroborate
this assertion.
We may now sum up Husserls claims concerning the unity of
meaning quameaning:
Husserl sets out from the purely grammatical distinction between
categor-ematic and syncategorematic expressions.
He founds the grammatical distinction on a semantic distinction
betweenindependent and dependent significations. So doing, he
asserts the exist-ence of two general subsystems in language that
each contribute its type ofmeaning. The categorematic (or lexical)
subsystem contributes independ-ent significations that can be
apprehended per se; the syncategorematic (orgrammatical) one
contributes significations that are unbounded, vague,and call for
completion. Due to the laws that govern the configurationsof parts
into wholes, independent parts (or moments) require not just
anywhole whatsoever, for they are not amenable to all sorts of
completion; ratherthey require completion of a specific sort.
Though Husserl does not con-clude this himself, it follows from the
above that the meaning contributedby syncategorematic expressions
is the general semantic frame or seman-tic structure within which
they are to appear.
Moreover, linguistic significations are defined as correlates of
meaningsintended in specific meaning acts. Dependent and
independent significationsare counterparts of dependent and
independent contents of meaning acts,so that the structure
governing the combination of significations at a lin-guistic level
is a structure already ruling the configuration of
pre-linguisticintentional acts. Thus, syncategorematic expressions
are not merely gram-matical or syntactic connectors, they are
faithful reflections of inten-tional forms of combination through
which partial representations uniteinto one complex
representation.
Finally, it should be stressed that the object language as such,
which isassessed in this functional approach, is not accessed as a
self-contained,autonomous object, i.e., by virtue of its specific
essence qua that kind ofobject, but rather by virtue of its being
an object whose essential functionis to be a symbolic vehicle, a
means of expressing, faithfully reflecting, andrearticulating
already formed, structured, or configured pre-linguistic con-tents
of meaning acts.
-
60 PEER F. BUNDGAARD
As already mentioned, the Fourth Investigation inflects from its
course around10, i.e., exactly after having reached this stage of
the argument. Before ex-posing what direction this inflection
takes, it may be useful to spell out whatproblems follow from this
attempt at a semantic-conceptual definition of theunity of
linguistic meaning. It may also serve as an explanation of why so
fewscholars have noticed the heterogeneous character of the Fourth
Investigation,and, a fortiori, why so few (if any?) scholars have
appreciated Husserls at-tempt to develop a semantic combinatorial
system on mereological grounds.The problems are basically the
following:
(1) Husserl provides absolutely no systematic examples for the
fact thatsyncategorematic expressions are understood, even when
they occur in iso-lation; they are felt to carry definite moments
of meaning-content, momentsthat look forward to a certain
completion which, though it may be indetermi-nate materially, is
formally determined together with the content in question,and is
circumscribed and governed by it (56/301). The absence of
exampleshas obviously had serious drawbacks for the argument, since
no one seems tohave shared Husserls confidence in the self-evidence
of his point. None ofthe linguistic theories that have either
explicitly acknowledged their debts toHusserl or been indirectly
influenced by him have elaborated these semanticaspects of his
considerations. On the contrary, at least until the beginning ofthe
1980s, they have remained a sort of terra obscura for linguistic
theory.We will see in a while that the situation has radically
changed.
(2) The other critical point concerns the nature of the
grammatical a priorisuch as Husserl defines it in the first nine
paragraphs. The aim was to laybare the unity of meaning qua
meaning, independently of it being logicallyconsistent or not; the
issue was thus the internal formal unity of a specificsymbolic
system, language. Now, although Husserl does establish
essentialdistinctions between the components of this system, and
even if he does as-sign essentially different semantic functions to
these components that canbe defined independently of psychology or
other empirical sciences, and,finally, even though he does address
and answer the question about whatprinciples govern the combination
of these element into coherent wholes these principles clearly
exceed and are by no means reducible to the linguisticsphere qua
linguistic. In short: if the Fourth Investigation as it has
oftenbeen said, and rightly so is an application of the insights
achieved in theThird Investigation, then the grammatical a priori
allegedly achievedthrough this application is in fact a
sub-mereological a priori which is ulti-mately governed by the very
same synthetic or intuitive a priori ruling thecombination of
perceptual parts in intuition. The tools with which we elu-cidate
the internal unity of linguistic meaning complexes are identical to
the
-
61THE IDEAL SCAFFOLDING OF LANGUAGE
ones with which we elucidate all other sorts of complex objects,
such asobjects in visual and acoustic perception.12
(3) Finally: in the first nine paragraphs, Husserl has proposed
a theory ofhow contents of linguistic expressions are configured
into consistentwholes. He has assumed that certain types of
grammatical expressions(syncategorematica) specify certain
intentional forms of combination (depend-ent contents of meaning
acts), so that founding relations between partialrepresentations
can be faithfully reflected in language. Yet, he has notconsidered
nor a fortiori elucidated the apparently crucial property of a
sym-bolic system like language, namely, that expressions must
follow each otherin a determinate way, according to determinate
forms of linear distribution,in order to make sense: what is given
as configured, n-dimensional contentsin experience must be
redistributed linearly in determinate, one-dimensionalcombinations
of expressions; and this linear constraint seems to constitute
thespecific difference and thus essence of language, the genuine
grammatical apriori.
It is exactly this grammatical misre that Husserl sets out to
remedy in theremaining paragraphs of his Investigation. The issue
now is not how contentsconfigure into consistent wholes, but how
specific categories of expressionscombine into formally well-formed
sentences. What was formulated in termsof the ordered co-existence
of semantic parts according to essential laws ofunity is now
formulated in terms of the linear combination of syntactic
cat-egories.
The syntactic a priori from semantic configuration to
syntacticconnectivity
The attempt to satisfy the demand for a pure grammatical a
priori implies aprogressive, yet radical, redetermination of the
notion form of combinationor connective form. Here is how the
re-construal is fulfilled in 10, A pri-ori laws governing
combinations of meanings.
(1) If there exist dependent significations, there exist a
priori laws regu-lating their combination into new meanings
(61/317). Insofar as dependencemeans need of completion, the laws
in question regulate the meanings needof completion by further
meanings; i.e., they indicate the forms and kindsof context into
which such dependent significations must be fitted (ibid.).The
argument is, at this stage, undistinguishable from the
mereological-se-mantic approach in the preceding paragraphs.
(2) No parts can be assembled without connective forms (or
momentsimplying foundational relations to other parts) governed by
laws of combina-
-
62 PEER F. BUNDGAARD
tion that restrict or prescribe the kinds of combinations that
are possible forthe combining members in question. In other words,
each sphere has its ownmaterial a priori governing the combination
of its members: In no sphere isit possible to combine items of any
and every kind by way of any and everyform: the sphere of items
sets a priori limits to the number of combinatorialforms, and
prescribes the general laws for filling them in (62/317). The
pointis simply that in each material sphere (or generic family) of
objects, we willfind rules proper to that sphere, which govern the
connection of the kind ofparts we find in that sphere. In the
sphere of, say, intuition, we will for exam-ple find determinate
dependence relations between color and extension; thus,in the
sphere of meaning, we must expect to find a priori rules that
specifi-cally govern the combination of linguistic elements.
It is at exactly this point (62 ff./317ff.) that the change of
scope intervenes.From now on, and abruptly so, the approach is
exclusively linguistic or gram-matical: the combination of
linguistic meanings has been cut off from its foun-dation on the
configuration of intentional acts and their contents (for
thesebelong to another material sphere with other characteristic
material a priorilaws). In other words, the form of combination
that Husserl earlier referredto as intentional, and thus pre- or
extra-linguistic, is now construed as apurely linguistic connector,
independent of any act-intentions or relation toany
extra-linguistic object. Here is how Husserl now develops the
mereologyof linguistic units:
(a) What characterizes the combination of linguistic meanings is
that theymust be connected in specific, antecedently determined
ways: only certainsequences of words make sense, others, even
including the very same words,yield only a heap of meanings, never
a single meaning (IVth LI: 62/318).Now, just as the relation of
mutual dependence between color and extension,is not one that
concerns one specific quality and one specific surface qua,
say,exactly this red nuance and exactly this apple surface, but one
that concernsthe genera of which they are the lowest differences
(color, extension), thelaws governing the combination of linguistic
elements do not apply to the lin-guistic units qua specific
meanings, but to the essential genera (wesentlichenGattungen),
i.e., the semantic categories (Bedeutungskategorien) to whichthey
belong.13 So, if a sentence like without cherished I alphabet an
vowelsdoes not make sense, it is not due to the incompatibility of
the particular se-mantics of these specific words, nor to the words
themselves, but to the factthat these words as instances of
specific semantic categories do not combineaccording to the rules
that govern the combination of semantic categories.
(b) Husserls sketch of a pure logical grammar rides rather
heavily on thisnotion of semantic category. It indicates the
following fundamental princi-
-
63THE IDEAL SCAFFOLDING OF LANGUAGE
ple: whenever we have a well-formed expression, say, This tree
is green, wecan, by means of formal abstraction, obtain the
corresponding pure syntac-tic form14 this S is p. Now, this ideal
form can be instantiated in indefinitelymany, but still evidently
restricted ways. The variability is indefinitely richinsofar as any
nominal matter say, this carved square root, this hungrybikini,
etc. can instantiate S (or substitute this tree in the former
example)without affecting the grammatical integrity of the
sentence; consequently, anyadjectival matter say agnostic or
skilled can instantiate p (or replacegreen in the example). On the
other hand, the variability is strongly con-strained: a token of
one categorical class can only be replaced by a token ofthe same
categorical class. In this blue raven is green, the integrity of
thecomplex linguistic form remains invariant, whereas in this
careless is green,it has disintegrated due to an illegal exchange
of syntactic categories and thusan illegal combination of
categories which is not supported by any correspond-ing pure form
of meaning.
(3) The conclusion is now quite simple: a construction of any
sort makessense linguistically (notwithstanding its logical
consistence) if and only if itis sustained by (or is an
instantiation of) a pure syntactic form combining se-mantic
categories in a predetermined way. The result of this is a research
pro-gram for the pure logical grammar whose essential and
foundational principleHusserl by now considers as evidently
established:
The task of an accomplished science of meanings would be to
investigate the law-gov-erned, essence-bound structure of meanings
and the laws of combination and modifica-tion of meaning which
depend upon these, also to reduce such laws to the least number
ofindependent elementary laws. We should obviously also need to
track down the primitivemeaning-patterns and their inner
structures, and, in connection with these, to fix the
purecategories of meaning which circumscribe the sense and range of
the indeterminates thevariables in a sense exactly analogous to
that of mathematics that occur in such laws(68/328)
Yet, from this passage, as well as from others, it is not
difficult to identify twoin fact quite different dimensions of the
exposed research program. On theone hand, Husserl posits the
necessity of establishing the laws of combina-tion; in the light of
the first paragraphs of his investigation, this demand canhardly be
a surprise: given that syncategorematic expressions, and thus
de-pendent significations, are unsaturated, they must be combined
with otherexpressions, and since combinations, by essence, are not
random, the task mustconsist in establishing the lawful way in
which meanings compound. On theother hand, however, this demand has
obviously become obsolete by the re-cently introduced notion of
ideal Satzform, the pure form of meaning, which
-
64 PEER F. BUNDGAARD
constitutes an a priori or pre-formatted syntactic template
(like S is p); ob-solete, because if such a syntactic template or
primitive meaning-pattern exists, it has already solved the problem
of combining partial meanings:whatever appears in such a form
appears as ipso facto combined by virtue ofthat form. Consequently,
the other, and prevailing, aspect of the research pro-gram
announced by Husserl, consists in tracking down and
circumscribingthe number of essential primitive syntactic forms
that predetermine the essen-tial modes of combination of semantic
(read: syntactic) categories. Consider:
Our first task, therefore, in a purely logical form-theory of
meanings, is to lay down theprimitive forms of meaning with the
requisite purity just described. We must fix the primi-tive forms
of independent meanings, of complete propositions with their
internal articu-lations, and the structures contained in such
articulations. We must fix, too, the primitiveforms of compounding
and modifications permitted by the essence of different
categoriesof possible elements (. . .). After this, we must
systematically survey a boundless multi-tude of further forms, all
derivable by way of repeated compounding or modification
(69/329330).
Here is not the place to comment on the notion of formal
analyticity whichHusserl draws on in his sketch of a pure logical
grammar (in contradistinc-tion to the synthetic a priori on which
he grounds his analysis of dependentand independent
significations).15 Neither shall I expose the fundamental prob-lem
of circumscribing the inventory of genuinely primitive syntactic
formsand of establishing the fundamental generative principles that
enable the pro-duction of new, empirically given sentences from
these primitive forms.16 In-stead I shall simply stress the
essential and both unrelated and unarticulated differences between
the semantic approach prevailing in the first 9 para-graphs of the
Investigation and the purely syntactic approach dominating thelast
paragraphs. These differences can be boiled down to the following
three:
1. As already commented on, the laws of combination required by
the first,semantic, approach are made obsolete by the second,
syntactic, approachto the extent that whatever linguistic parts
appear in a meaningful sentencethese are always already combined by
virtue of the ideal syntactic formsustaining the sentence in
question. In other words, the distinction betweendependent and
independent meanings is no longer operative. In this ap-proach,
there is no point in parsing linguistic parts into dependent
momentsand independent pieces. All parts are indeed dependent with
respect to thewhole. One may of course proceed to classify parts
according to the semanticcategories they fall under, but this
distinction is not mereological in the firstsense; that is to say,
it does not concern the nature and function of the parts
-
65THE IDEAL SCAFFOLDING OF LANGUAGE
qua parts in a given whole. The primary primitive object is not
the parts,nor their types of mutual relations, but the very
syntactic gestalt itself. Thismeans that the parts are only
accessible insofar as they occur in a pre-de-termined whole, a
primitive syntactic form, which qua primitive and pureis naturally
indecomposable and irreducible to either its parts or the
rela-tions between them: the form is in and of itself the very
determination ofthis relation. In the first, semantic account,
however, the unity of the wholeis certainly claimed to rely on
specific dependent meanings, which couldbe strictly defined as
figural or unifying moments: i.e., parts that do not onlyrequire
being combined with other parts, but also specify the
(semantic)nature of the combination required. The difference
between these two ap-proaches seems thus characterizable in terms
of gestalt theory: in the firstapproach, linguistic wholes are
composed of genuine parts (dependent andindependent), and their
internal order or structure can, as it were, be tracedback to the
dependent parts; in the second account, no parts are genuine,
allare abstractly distinguishable parts, variables in the already
constitutedsyntactic form itself.17
2. In both approaches, laws of combination are invoked. In the
first approach,these are specified as laws governing the
configuration of parts, by virtueof their meaning, in what could be
called an n-dimensional quality space;in the second, however, they
are specified as laws of connectivity govern-ing the linear
distribution of parts by virtue of the syntactic category to
whichthey belong.
3. Finally, the object of the first approach is assessed through
its being essen-tially a vehicle, serving the purpose of expressing
and re-articulating alreadyformed Vorstellungen or expressible
thoughts, whereas in the secondapproach it is assessed as an
autonomous, self-contained symbolic system,by virtue of its own
symbolic congruity, independently of what it expresses.
Having arrived at this stage, i.e., having characterized the two
approaches andspelled out their differences, the time has now come
to provide some evidencethat would justify the relevance of the
first, semantic approach. In other words,I will try to show what a
genuinely semantic combinatorial system underpin-ning the
composition of linguistic significations consists in, and more
spe-cifically with respect to Husserl I will supply a selection of
examplesthat, I believe, convincingly illustrate what it means for
a dependent mean-ing to require a semantic whole of a determinate
sort. Questions like thesehave already been addressed for a great
many years within linguistics, andI will therefore simply compare
these theoretical results with Husserls keynotions:18
-
66 PEER F. BUNDGAARD
Examples of grammatical expressions that specify intentional
forms ofcombination
Consider the following case:19 a complex representation
(Vorstellung) com-posed by two partial representations, say, (1)
They are married, (2) They donot live together, and an intentional
form of combination of (1) and (2). If arepresentation exists that
combines (1) and (2) into one complex representa-tion, then,
Husserl claimed, there must be a semantic correlate to that
globalrepresentation, and to the semantic form there must be a
specific grammati-cal correlate, i.e., a way of faithfully
expressing the intended meaning. Thisentails that not only the
partial representations, but also the intentional formof
combination should be expressible. Consider now the following
possiblecombinations of (1) and (2):(a) They are married, BUT they
do not live together.(b) They are married, AND they do not live
together.(c) They are married, OR they do not live together.In each
case, the partial representations remain the same, yet the global
mean-ing varies considerably according to the way in which the
relation between(1) and (2) is intended and expressed. Conjunctions
like the above clearly donot merely laterally combine one partial
representation (or proposition) withthe following, they also
express the very way such representations are intendedto relate.
(They are the exact linguistic correlate to the intentional form
ofcombination Husserl referred to above.) This justifies the
characterizationof the meaning of a syncategorematic expression in
mereological terms as thefigural moment that combines partial
significations into a specific semanticwhole. In other words as we
shall see in more detail below the meaningcontributed by dependent
contents, like but, is not merely a block of signi-fication added
to already given blocks of signification in a summative wholeof
significations; rather, the meaning of a dependent content is the
very glo-bal relation it expresses between the partial
representations it combines. Thus,the meaning of but could be
characterized as follows: in a complex con-struction compounded
with but, whatever is to the left of but and what-ever is to the
right of it are intended as contrasting or conflicting contentsin
some respect; they take on this additional, and crucial, semantic
value byvirtue of the dependent content that combines them. Thus,
dependent contentsdo not simply require determinate contexts; in
fact, their meaning is the kindof semantic whole into which the
partial significations are combined. Thissemantic whole could in
the case of but be diagrammed as a schematic func-tion (in Freges
sense) like this: X > [but] < Y.
-
67THE IDEAL SCAFFOLDING OF LANGUAGE
To get a firmer hold on how this semantic organization differs
from syn-tactic combination, consider the following: granted that
independent meaningscontribute a schematic whole of a determinate
sort (yet, vaguely, incompletely,i.e., without any specific
arguments), we can claim that whatever specificmeaning but assumes
in a given sentence (depending on the independent parts(or
arguments) it combines), it necessarily implies a relation of
contrast,conflict, or tension between its relata. It thus demands
completion of aspecific sort, namely, through arguments that must
in one way or another beconstrued as conflicting. Consider at
present what consequences this has forpure logical grammar. If it
is the case that, for example, but requires com-pletion that
satisfies its general schematic meaning of contrast, then we cana
priori state: if XYZ in XYZ but XYZ means exactly the same thing in
eachof its occurrences, then the expression is agrammatical for
essential reasons. Itis therefore an analytical fact that he is a
man but he is a man is agrammatical(or semantically ill-formed) if
by he is a man I mean the same thing twice,because the law
governing the completion of the dependent signification hasnot been
satisfied.
This is of course not the case from a purely syntactic
perspective. Here wecan apply our method of formal abstraction on a
sentence like he is a priest,but he is a sinner and thus obtain an
ideal syntactic form combining certainsyntactic categories (or
classes of variables), such as He is [Nominal matter],CONJ. He is
[Nominal Matter] and from there, through substitution, we
canconstruct a perfectly congruous sentence like He is a man, but
he is a man.In other words, the difference between the two
combinatorial systems is sodeep that what in one system epitomizes
grammaticality is in the other a pa-rade example of
agrammaticality.20
In the next section, I will provide more examples, all borrowed
from Talmy(2000), in order to give a more substantial
characterization of the configurationalsemantic system underpinning
the organization of linguistic significations.Needless to say that
it is far beyond the scope of this paper to propose a sys-tematic
presentation of Talmys work in particular and Cognitive
Linguisticsin general. I will therefore only include material that
is directly relevant forthe present purpose: i.e., evidence that
shows how the general mereologicalprinciples of semantic
organization, purported by Husserl, can be found to holdin
language.
Talmys closed-class semantics
Talmys work shows several striking analogies to Husserls Fourth
Investi-gation. In the very first lines of the first chapter of his
monumental Toward a
-
68 PEER F. BUNDGAARD
Cognitive Semantics (2000), he too calls attention to a
universal, purely for-mal design feature of language, namely, that
it is composed by two subsys-tems: a grammatical and a lexical one.
The lexical one, which is co-extensivewith Husserls categorematica,
consists of open word classes that are richin members and readily
augmentable (i.e., roots of nouns, verbs, and adjec-tives). The
grammatical one, co-extensive with Husserls syncategorematica,is
made out of closed word classes that have few members and are not
sub-ject to or likely to undergo any variation (all other word
classes preposi-tions, conjunctions, verb particles, etc. but also
all sorts of tense and casemarkers, suffixes, prefixes, specific
syntactic constructions, and grammati-cal categories as such).
Next, Talmy, just like Husserl, attributes specific semantic
functions to eachsubsystem. A complex linguistic expression (e.g.,
a sentence) serves the pur-pose of articulating and thus evoking in
the listener a specific experientialcomplex or a cognitive
representation (Talmy 2000, p. 22). The point isnow that elements
of both subsystems represent conceptual material i.e.,have a
meaning but not the same type of material. Keeping in mind that
lan-guage according to Husserl must dispose of semantic forms to
express not onlybounded contents of meaning acts, but also the way
these are intended to re-late (their intentional form of
combination), it is easy to appreciate the af-finities to Talmys
conception:
The grammatical and lexical subsystems in a sentence seem
generally to specify differentportions of a cognitive
representation (CR). Together, the grammatical elements of a
sen-tence determine the majority of the structure of the CR, while
the lexical elements togethercontribute the majority of its
contents. The grammatical specifications in a sentence,
thus,provide a conceptual framework or, imagistically, a skeletal
structure or scaffolding forthe conceptual material that is
lexically specified (Talmy 2000, p. 21).
Such assumptions, as well as Husserls in his first approach,
clearly call intoquestion the task assignment dogma within
linguistics, according to whichperception provides referential
access to objects, whereas language combinesthe symbols referring
to these representations exclusively in terms of its ownlaws of
linear composition.21 Rather, it claims that linguistic
representation isnot a two-relata affair between linguistic
structure proper and reference ob-jects, but a three-relata affair,
so that language, thanks to certain expressiveand semantic
properties, not only refers to a scene, but also articulates
struc-tured modes of experiencing or conceptualizing this
referential scene. Thecrucial claim here is then strictly like
Husserls: closed-class elements playa privileged role (as dependent
meanings) since they make out the inventoryof semantic forms
through which the structures of a complex representation
-
69THE IDEAL SCAFFOLDING OF LANGUAGE
are specified, or, to put it differently, through which a
referent scene is ex-pressed linguistically such as it is intended
in experience. Talmy thus concludesthat a study of semantics as a
genuine combinatorial system in language mustbe a systematic study
of the kind of structure specified by closed-class ele-ments, since
they are in charge of the conceptual organization within
lan-guage:
The present chapter advances the position that this set of
grammatically specified notionscollectively constitutes the
fundamental conceptual structuring system of language. Thatis, this
crosslinguistically select set of grammatically specified concepts
provides the basicschematic framework for conceptual organization
within the cognitive system of language(Talmy 2000, p. 21).
I will now turn directly to the examples that may serve to
answer our initialquestions:
1. If to require a determinate sort of whole implies that a
dependent mean-ing (or a closed-class element) in fact contributes
the kind of structural wholein terms of which the linguistic parts
are combined, how is this structuralwhole or schema to be
identified and characterized?
2. How can it be shown that dependent meanings express and
specify inten-tional forms of combination, as Husserl had it, or
conceptual structure,as Talmy calls it?
Consider the following examples:
1.(a) The cat is on the car(b) The cat is in the car(c) The cat
is two feet from the car
2.(a) The boat is on the water(b) The boat is in the water
Traditionally, prepositions, syncategorematic expressions, are
considered asserving a binding function within a clause for
instance, by specifying therelation in which one referential
element stands to another. Yet in 1 (a,b,c),the prepositions do not
simply specify a spatial relation between two referen-tial objects,
they also, and crucially so, express a specific
conceptualizationof, in this case, the car, which in (a) is
conceptualized as a surface (with all
-
70 PEER F. BUNDGAARD
other properties abstracted away); in 1 (b) as a volume or a
container (withall other properties abstracted away); and in 1 (c)
as a point (with all otherproperties abstracted away). What the
examples reveal is that a specific modeof perceptually intending
the car is specified by the prepositions, namely tothe effect that
only certain of its spatial properties are referred to, while
allothers are neglected.22
The second set of examples illustrate the same phenomenon, but
with anadditional and quite essential specifying feature. 2 (a,b)
belong indeed to afamily of expressions that refer to exactly the
same state of affairs, but alter-nate in the way they schematize
it. So 2 (a) and 2 (b) of course differ by virtueof the schematic
content of the prepositions and the way they idealize thewater as a
surface and a volume, respectively. Yet these
alternatingschematizations are best construed as yet another
specification, namely, ofpoint of view, on specifies a distal point
of view from which the water is givenin experience as a homogeneous
plane (no significant or perceivable move-ment of waves, etc.),
whereas in specifies a proximal point of view fromwhich the water
is given in experience with its mass and voluminous charac-ter
(waves licking the hull, etc.).
Examples of alternations in schematization are indeed
particularly inter-esting. They display obvious semantic
differences in expressions referring toexactly the same objective
reference scene. Such differences are therefore notexplainable
solely with respect to the reference object (it remains
invariant),nor with respect to the full-fleshed semantics of the
opened class words em-ployed (these are basically the same),23
rather they should be explained withrespect to the describable
properties of the experiential act in which the refer-ential object
is intentionally given. In short, the whole combinatorial systemin
charge of the semantic organization of language clearly exceeds the
strictdomain of language and grammar. Basic semantic features
displayed in lan-guage are simply not assessable in purely
linguistic terms. They are essen-tially grounded on characteristics
and structures of perception and intentionalexperience as such.24
Thus, in cases of alternations in schematization, the dif-ferences
are readily and sometimes quite subtly reducible to gestalt
dif-ferences between figure/ground structures in the experienced
referent scene,intentional distribution of attention to a reference
scene, perspective, andmodes of perceptual apprehension. In each of
these cases as we shall verybriefly see below the pre-linguistic
organizing structures are linguisticallyspecified by closed-class
elements and their dependent meaning (be they sin-gle words or
specific syntactic constructions).
Semantic differences in linguistic expressions due to
figure/ground inver-sions are obvious in cases like a is parallel
to b vs. b is parallel to a a re-
-
71THE IDEAL SCAFFOLDING OF LANGUAGE
sembles b vs. b resembles a, a beats b vs. b is beaten by a (cf.
Talmy 2000,volume 1, chapters 5 and 6, for a comprehensive
exposition of figure/groundstructure in language). Cases where
semantic differences amount to differencesin modes of
experientially intending the situation or in perceptually
appre-hending the referent scene are particularly interesting (they
have been thor-oughly analyzed by Langacker [1991, chapter 3]).
Consider the following:
3 (a) The troops went across the desert3 (b) The troops crossed
the desert
Here the basic semantic contents of the preposition across and
the verb crossare the same. Langacker (1991, pp. 78 ff.) suggests,
however, that there is adifference. According to him the verbal and
the prepositional expressions re-fer to distinct ways of
intentionally accessing the components of a complexreferent scene.
This difference is not specified by any contents (schematic ornot)
of the words used, but rather by virtue of the notional contents of
the gram-matical categories used. The claim is, then, that a
preposition implies a sum-mary scanning in which the various facets
of a situation examined incumulative fashion, so that progressively
a more and more complex con-ceptualization is built up; once the
entire scene has been scanned, all facetsof it are simultaneously
available and cohere as a single gestalt (Langacker1991, p. 78). On
the other hand, a verb (here: cross) implies a sequentialscanning,
in which the various phases of an evolving situation are exam-ined
serially, in non-cumulative fashion; hence the conceptualization is
dy-namic, in the sense that its contents change from one instant to
the next(Langacker 1991, pp. 7879). In other words, subtle but
linguistically perva-sive differences between parallel expressions
like the above are explainablein terms of differences in (1) mode
of conceptualization (scanning), and interms of (2) the internal
structure of the latters intentional correlate. Thus in3 (a) the
process and its components are intended as a complex atemporal
ge-stalt, with no specific focus on the distinct phases of the
figures (the troops)movement; whereas in 3 (b) the scene is
intended with specific focus on thetransient movement, from phase
to phase, of the troops.25
The above presentation does not of course pretend to be either
systematicor, a fortiori, exhaustive. It is intended as a mere
general survey of certainconceptual contents specified by language,
ones that linguistic structure istuned in, as it were, to express.
It does, however, help us better understandwhat exactly Husserl
anticipated in his sketch of a mereological semantic sys-tem of
language. It supplies us with concrete examples of how
syncategorematic
-
72 PEER F. BUNDGAARD
expressions and their dependent meanings are linked to or
contribute to de-terminate sorts of semantic wholes, i.e., general
schematic frames in terms ofwhich the component elements of complex
linguistic expression are eitherconceptualized or combined. It
demonstrates how extra-linguistic cognitiveor intentional features
such as perspective, point of view, and attention are notonly
capable of being expressed linguistically, but also play a
fundamentalrole in organizing linguistic contents that are not
explainable in traditionalintra-linguistic terms. Finally, it
provides convincing evidence for the exist-ence of what Jackendoff
(1987) called the intermediary level where languagestructure is
compatible with perception (exactly the level whose
existenceHusserl claimed on a priori grounds) the level of
conceptual structure. Inthis respect Cognitive Linguistics in
general and Talmys or Langackers worksin particular are not
strictly linguistic theories; they are comprehensive cog-nitive
theories that study and lay bare (1) the relative dependence of
linguis-tic structure on pre-linguistic structure; (2) the
essential tenets of pre-linguisticstructure; and, finally, (3) the
design features of the linguistic system that makeit capable of
systematically expressing and re-articulating such a
conceptualstructure.26
I will now conclude this study with a couple of epistemological
remarksand some minor conjectures as regards the essence of
language.
Concluding remarks
Research in Cognitive Linguistics is one among several examples
of the factthat progression in the empirical sciences does,
occasionally, elucidate un-clear or vague elements of philosophical
thought, or that it does provide theevidence that completes only
partially or incompletely developed philosophi-cal hypotheses. That
Cognitive Linguistics serves this function with regardsto Husserls
Fourth Investigation is in a way quite natural if we consider
thefact that Husserlian phenomenology as I briefly commented on in
my Pre-amble indirectly or through sedimentation constitutes a
genuine source forseveral of the fundamental hypotheses purported
by this recent research pro-gram in linguistics, the most crucial
among these being the idea that the scaf-folding of language
construed as a means of combining linguistically whathas already
been composed intentionally is determinable in terms of a se-mantic
combinatorial system.
Now, this semantic approach to language has not been popular in
linguis-tics (at least not in the eighty years that followed the
publication of the Logi-cal Investigations). The reason is most
likely that linguists (and philosophers
-
73THE IDEAL SCAFFOLDING OF LANGUAGE
as well) had much difficulty in seeing what such a semantic
system shouldconsist in (whereas many theories about formal
syntactic systems of combi-nation were available, at least since
Bolzano). Even granted that dependentsignifications require
semantic completion of a determinate sort and thus provethe
existence of principles governing this completion, it was for many
dec-ades difficult to see how these principles could be amenable to
systematiccharacterization, and consequently how linguistic
inquiries should be con-ducted in these terms.
As we have seen, Husserl himself was not eloquent on this issue.
Perhapsbecause he himself did not know how the idea of grammar was
to be estab-lished within such a comprehensive theoretical
framework, i.e., how it couldbe possible to characterize the
systematic correlation of the level of grammati-cal expressions,
the semantic level of significations, and the level of
pre-lin-guistic presentations (Vorstellungen). This may be the
reason why he notonly changed, but certainly also substantially
reduced his scope, and redefinedboth his object and the a priori on
which he would ground it.
This change and restriction of scope is also likely to respond
to another,already mentioned, difficulty: the fact that in his
first semantic approach,Husserl lays bare essential structures of
language that indeed govern the con-figuration of parts into wholes
in general, that is to say, far beyond the realmof linguistic
significations. This could be and has certainly been considered
aquite annoying circumstance in at least one respect: If the issue
is language,and if language is a phenomenon endowed with certain
properties and char-acteristics that make it evidently
distinguishable from other objects, then theessential distinctions
proper to language should, allegedly, be drawn withinthe sphere of
grammar itself. The essential properties assigned to languageshould
be extracted from language itself and only from there. Otherwise
therisk of a metabasis looms large, i.e., the danger of
characterizing an object ofone type (linguistic meanings) in terms
pertaining to an object of another type(such as contents of
perception).27
All the above difficulties have, I believe, been overcome. We do
indeedknow what a semantic combinatorial system is supposed to look
like and howit is likely to be characterized; we have substantial
(though diverging) ideasof what functions it serves, and we can
conceive why and how it is connectedto thought in general and
perception in particular. As concerns the latter,Talmys works
illustrate rather well how it is possible both to found linguis-tic
structure on pre-linguistic conceptual structure and to provide
systematicdescriptions and characterizations of the linguistic
system per se (for in-stance, in terms of what categories and types
of structure closed-class ele-ments specify).28
-
74 PEER F. BUNDGAARD
Finally, in the present paper I have of course not wished to
disregard theimportance of syntactic constituency. It is certainly
a remarkable fact thatsentences such as this medieval prime number
is scared are meaningful andperfectly congruous (at least following
Husserls first rough sketch). Obvi-ously, no language would be
capable of expressing anything if it did not fol-low some
principles of syntactic combination. (It is also true for
unpluggedcomputers that they cannot calculate.) What I have called
into question is ratherthe (wrongly) entrenched assumption,
inaugurated by Husserl in the FourthInvestigation, that the essence
of language is reducible to principles govern-ing the linear
combination of linguistic parts in virtue of their syntactic
formand regardless of their semantic import. There is no a priori
reason why theobject language should be reduced to only this
aspect. At least, it seemsequally justified to endorse a
holistic-functional view and consider languagea symbolic system
that qua symbolic essentially serves the purpose of express-ing
thoughts in a very large sense. In this case, of course, an urgent
task con-sists in accounting for the means of which language
disposes to articulate thesecontents faithfully, that is to say,
according to the way they have been expe-rienced or intended. What
is of particular interest here is that strong and de-scribable
elements of constituency can be laid bare in this domain. Of
coursenot in the linear sense of constituency, but, as it were,
transversally to it, byvirtue of the very contents expressed.
Accordingly, verbs, for example, do notonly specify a specific mode
of action, but correlatively to this mode of ac-tion they also
unfold a schematically configured scenario with a determinatenumber
of participants and determinate relations between these (thus
giveimplies a giver, an object, and some receiver, and a
specifically oriented trans-fer of the object; it shares this
schematic content with another verb like send).This configurational
content does warrant a genuine semantic combination ofmeanings, and
it does so through structural properties both different and priorto
the combinatorial principle governing the well-formed linear
combinationof a subject, a verb, an object, and an indirect object
in a sentence.29
This is, I believe, what the first part of Husserls Fourth
Investigation aimedat exposing. The crucial question is not how
language combines words lin-early, but how language, under this
one-dimensional constraint, is capable ofexpressing everyday
thoughts and experiences composed in pluri-dimensionalcomplex
wholes.
Notes
1. In the following, references to the Logical Investigations
are given directly in the textwith no further indications, like
this: (XX/YY). The first page indication refers to
-
75THE IDEAL SCAFFOLDING OF LANGUAGE
Findlays translation of the Logical Investigations, the next to
the original second edi-tion in German.
2. On Jakobson and Husserl, see first of all Elmar Holensteins
indispensable RomanJakobsons Approach to Language Phenomenological
Structuralism (1976). E. CassirersStructuralism in Modern
Linguistics (1945) also emphasizes this connection. Mulligan(1988)
characterizes the mereological foundation of K. Bhlers theory of
language, and,finally, Stjernfelt (2000) is a very instructive
general survey of the role of mereology,the theory of parts and
wholes, in linguistics and semiotics.
3. The title of Lakoff and Johnsons Philosophy in the Flesh
(2000) is of course indebtedto Merleau-Pontys concept of la chair
and to Husserls concept of Leib. Johnson (1987)characterizes his
method as one of descriptive phenomenology. The claim that
linguisticsignifications, expressed by signs, are sustained by
meaning-conferring and meaning-fulfilling acts constitutes the
initial core argument of the First Investigation. The foun-dation
of predicative judgments on ante-predicative judgments is at the
heart of HusserlsExperience and Judgment (1939).
4. One of the first philosophers to consider the Fourth
Investigation as a precursor of mod-ern logic and incidentally of
modern linguistics was Bar-Hillel, who in his paper from1957,
Husserls Conception of a purely Logical Grammar, concludes: [. .
.], we maysay that Husserls conception of a purely logical grammar
has to be regarded, in a veryessential and pregnant sense, as a
forerunner of Carnaps conception of a general logi-cal syntax. One
has only to omit the detour through the realm of meaning, and the
re-liance upon an apodictic evidence and to add a mastery in modern
symbolic logic andits philosophy in order to perform the transition
from Husserl to Carnap (Bar-Hillel 1957,p. 369). J.-L. Gardis
(1975) very instructively develops the theoretical affinities
betweenChomsky and Husserl, and much closer to us, also the French
phenomenologist J. Benoistconcludes one of his numerous works on
the Fourth Investigation with the followingremark: One of the
logical consequences of the considerations developed in the
presentwork, particularly as regards the syntactic a priori, would
certainly consist in re-actual-izing the discussion about the
connection between Chomsky and phenomenology andabout the
possibility of developing a Chomskyan interpretation (minus
mentalism) ofphenomenology (Benoist 1999, p. 68; my translation).
And finally, D. Mnch criticallyobserves: If we look at the fourth
Logical Investigation we can see that Husserl is aforerunner of
Chomsky. In this investigation Husserl applies the basic concepts
of for-mal ontology which he had developed in the third
investigation to grammar. Languageis conceived as a field which is
guided by purely formal rules, which has to be studiedby a
discipline he calls pure grammar [. . .] This grammar has an
algorithmic charac-ter, too. But Husserl does not only support a
computational approach to language. Moreo-ver, it is one of his
central claims that our intentions are restricted by the laws of
purelylogical grammar. Thoughts which are not in accordance with
these laws are nonsense,i.e., they cannot be directed towards an
intentional object (Mnch 2002, p. 203).
5. To avoid all possible misunderstandings I stress as such: the
concepts used by Husserlin the first nine paragraphs (dependent and
independent meanings, relations of founda-tion, etc.) neither have
been ignored nor are they to be considered as essentially
incom-patible with syntactic analysis as such. As we have already
seen, structural linguisticsexploits exactly these conceptual tools
in its characterization of language as an autono-mous, functional
system. And as Mulligan (1988) has shown there is nothing
intrinsi-cally inconsistent in a mereological approach to syntax:
in his Theory of Language, Bhler
-
76 PEER F. BUNDGAARD
indeed exposes the basic tenets of such a syntactic theory (to a
large extent developed inthe very same terms by the founding father
of modern theory of syntax, Lucien Tesniere(1959). My point is
simply that no one seems to have further elaborated, even
appreci-ated, Husserls sketch of a comprehensive theory of semantic
unity in mereological terms.Even Benoist (1999, 2001), who clearly
identifies Husserls ambition, rejects it; alas,on a priori grounds,
without reference to any linguistic theory, let alone any
linguisticfact.
6. I slightly modify Findlays translation of the German
Selbstndigkeit and Unselbstndig-keit. In order to follow, at least
a bit, the German constructions, he chooses independ-ence for
Selbstndigkeit and non-independence for Unselbstndigkeit; to avoid
themeanderings of a double negation, I prefer the straightforward:
independence forSelbstndigkeit and dependence for
Unselbstndigkeit.
7. This is why moments, though dependent, are said to be prior
to pieces: Strictly speak-ing our approach is positive in the case
of what is dependent, negative in the case of whatis independent
(Third Investigation, 13/241).
8. Cf. We have recognized that the seemingly indifferent
distinction between categorematicand syncategorematic expressions
corresponds to a fundamental division in the realmof meanings. We
took the former as our starting point, but the latter revealed
itself asbasic, as the prime foundation of the grammatical
distinction (58/310).
9. Presentation translates Vorstellung (idea,
representation).10. In view of the subsequent, entirely syntactic,
analyses, this statement is indeed aston-
ishing: it is easily seen that Husserls mereological foundation
of linguistic constituencyis not at all applicable to syntax in his
sense. Let us consider the syncategorematic ex-pression A as a
syntactic category regardless of any signification we now apply
thelaw according to which it should appear in a (syntactic) whole
of a specific sort G(AB...M), where B...M stand for determinate
sorts of syntactic categories. Obviously theextension of this law
(say, the number of syntactic constructions in which (A, e.g.,
theconjunction but, can occur) is indefinitely big, and thus quite
indeterminate. Now,compare this to its semantic counterpart, where
but expresses an intentional form ofcombination; in this case, but
requires only one sort of context: one in which a rela-tion of
contrast, tension, or conflict is intended to hold between two
partial representa-tions combined by but. I will discuss this in
further detail below.
11. Pre-linguistic in the sense that the act or the meaning
intention precedes the forma-tion of linguistic signification.
There is intention of meaning before there is
specificsignification. Consider the expression tiger safe nature
reserve; it is impossible to saywhether it means that the reserve
is a safe place for the tigers or that it yields protectionagainst
the intrusion of tigers. As Husserl put it in the First
Investigation, the differencedepends on the meaning-conferring
intentional acts and the way they animate the word-ing.
12. Benoist (1999, chapter 5) also stresses this. So does
Mulligan, yet without consideringit a problem: Bhler, however, like
Husserl and Marty, was aware, as most contempo-rary writers are
not, that these notions [of dependence and constituency] can be
system-atically applied above and below the level of syntactic
structure as well as out-sidelanguage (Mulligan 1988, p. 208).
13. Cf., The impossibility attaches, to be more precise, not to
what is singular in the mean-ings to be combined, but to the
essential kinds, the semantic categories, that they fallunder
(62/318). Semantic category translates Bedeutungskategorie, but
what is
-
77THE IDEAL SCAFFOLDING OF LANGUAGE
meant here is nothing but syntactic category. The fact that the
same term, Bedeutungis used indistinctly throughout the whole
Investigation (yet in two different senses) mayhave served to blur
the essential differences between the two approaches.
14. Findlays translation does not faithfully render Husserls
expressions. In the Germanversion, Husserl says Gehen wir
formalisierend von der gegebenden Bedeutung [. . .]zur
entsprechenden Bedeutungsgestalt, zur Satzform ber, so erhalten wir
dies S ist p,eine Formidee, die in ihrem Umfang lauter selbstndige
Bedeutungen befat (pp. 318319). Findlay conflates Bedeutungsgestalt
and Satzform into pure form of mean-ing (p. 62), which contributes
to obliterating the fact that whenever expressions suchas meaning
or semantic (category) are used in this and the following
paragraphs,what is meant is syntactic function or syntactic
category.
15. On this issue, cf. Jocelyn Benoists analyses in Benoist
(1999, chapter V; 2001, chapterIV). The present paper is in a way
indebted to Benoists minute scrutiny of the FourthInvestigation in
a whole series of works. My conclusions are nevertheless exactly
op-posite his, most clearly as regards the question whether or not
semantics on mereologicalgrounds is possible: I do not, contrary to
Benoist, believe that this question is decidableon purely a priori
grounds.
16. This defines in fact the very core and purpose of Chomskys
transformational and gen-erative grammar in its orthodox version.
In this respect, Husserl is evidently a forerun-ner of this
linguistic theory.
17. Abstractly distinguishable parts is B. Smiths expression
(1994, p. 270) and is used tocharacterize the nature of wholes and
the status of their parts according to the BerlinSchool of Gestalt.
The gestalt theoretical difference evoked here is indeed the one
be-tween the Austrian theory of Gestalt (von Ehrenfels, Husserl,
Stumpf) and the BerlinSchool of Gestalt (Wertheimer, Koffka, Khler)
on this see Smith (1994, chapter 8).Mulligan (1988) provides
analyses of Karl Bhlers theory of language based on the
samedistinction.
18. Simply to show how familiar this type of questioning is
within linguistics, I shall quotefrom R. Jackendoffs recent book.
Foundations of Language (2002): Assuming that thefunction of
language is the expression and communication of thought, I will
identifysemantics as the organization of those thoughts that
language can express [. . .] Thethoughts expressed by language are
structured in terms of a cognitive organization calledconceptual
structure. Conceptual structure is not part of language per se it
is part ofthought (Jackendoff 2002, p. 123). Incidentally, it is
worthwhile noticing that the firstof Jackendoffs assumptions is
almost literally identical with Husserls view accordingto which
language serves the purpose of articulating expressible thought,
and the ar-gument arrives at the same conclusion: if language is to
express, faithfully, contents ofthought, it must dispose of means
to organize semantic contents, i.e., it must be sustainedby a
system capable of combining contents which is not derived from or
likely to beassimilated to syntax.
19. The following examples are not mine. Charles Fillmore uses
them, yet to a different end,in his article Frame Semantics, which
is, by the way, a genuine piece of mereologicalsemantics that
extends the dependence of significations right into the heart of
thecategorematic expressions: By the term frame I have in mind any
system of conceptsrelated in such a way that to understand any of
them you have to understand the wholestructure in which it fits;
when one of the things in such a structure is introduced into
atext, or into a conversation, all of the others are automatically
made available (Fillmore
-
78 PEER F. BUNDGAARD
1982, p. 111). The meaning of sell or seller is in other words
founded on a wholescenario implying a buyer, an object to be
bought, etc.
20. I stress: it is agrammatical not for logical reasons, but
for reasons inherent to the lawsgoverning the combination of
meanings at that level.
21. This dogma finds one of its first and clearest expressions
in Husserls second, syntactic,approach in the Fourth Investigation.
According to this view, there is no other structur-ing principle in
the domain of language than the one sustained by the ideal
syntactic forms.
22. This is indeed, according to Talmy, the fundamental feature
that justifies the characteri-zation of closed-class contents as
schematic. The schematic contents of, for instance,prepositions are
thus semantic correlates to a previous conceptual process of
abstrac-tion where single elements or whole reference scenes in all
their bulk and physicalityare differentially boiled down to match
ascribed schemas (Talmy 2000, p. 220). Im-plied in such use of
schemata is thus a cognitive process of idealization, i.e., a
modeof intending a referent scene only with respect to those
properties that instantiate theschema, and correlatively a process
of abstraction of all those properties that are notrelevant for the
schema. A schema underlying the preposition in could, once again
inFregean terms, be called a two-argument function (or a vague
predicate) that specifies arelation of containment between one
element conceptualized as a figure and a groundconceptualized as a
volume. Hence, as Talmy often has remarked, the neutrality of
theschema as to the material differences of the scenes to which it
is applied, cf: the waterin the teaspoon, the baby in her arms, the
fish in the ocean.
23. Husserl makes the very same point in his First Investigation
( 12).24. As Talmy remarks: Clearly, the language-related faculty
of the brain evolved to its
present character in the presence of already existing cognitive
domains, including thatof vision, and no doubt developed in
interaction with their mechanisms of functioning,perhaps
incorporating some of these (Talmy 2000, p. 96).
25. Examples are in fact legion here. Consider, for example, the
difference between a ver-bal and a prepositional construction in
French: (a) Ils longent la rivire vs. Ils marchentle long de la
rivire. Talmy, too, operates with the same distinction between a
synoptic(summary) mode as in There are some houses in the valley
and a sequential mode as in There is a house every now and then
through the valley (Talmy 2000, p. 71). Thesystematic exposition of
the notional contents of grammatical categories and their rela-tion
to modes of intentions can by the way be traced back to R. Ingarden
(1931), who al-most literally anticipates certain fundamental
assumptions of Langackers; he too addressesthe issue of whether or
not there is an essential difference between nominal and
verbalsignifications and claims that the difference is due to the
quite different types of inten-tionality, i.e., to the way the
intentional correlates have been intentionally construed bythe
corresponding [nominal or verbal] significations (Ingarden 1931, p.
7778).
26. According to Talmy, the wide variety of grammatical elements
endowed with semanticfunctions group in a rather restricted set of
schematic categories, which in turn clusterin only four different
general s