Concepts, Meaning, and The Lexicon: Philosophy of Language Meets The Syntax-Lexical Semantics Interface

Concepts, Meaning, and the Lexicon: Philosophy of Language Meets the Syntax-Lexical Semantics Interface
Michael Glanzberg
University of California, Davis
November 2010
Michael Glanzberg (UC Davis)
Concepts, Meaning, and the Lexicon
November 2010
1 / 67
Words and Concepts I

We have (I shall assume) concepts.
Assume they are something like mental representations. Assume they are not part of the language faculty proper.
Enter into wider cognitive processes (e.g. metaphor comprehension). Some animals (non-verbal creatures!) appear to have something like concepts (though perhaps not just like ours) (e.g. Gallistel, 1990).
A relatively weak thesis: not (yet!) trying to take a stand on issues of how much language-like structure concepts have, or on the priority of thought and language.
The language faculty includes a lexicon.

Lexical entries include semantic, syntactic, and phonological information. Include meanings of words. Determine a wide range of syntactic properties, including what used to be known as D-structure.
November 2010
2 / 67
Words and Concepts II

The two relate.
Our grasp of concepts and our grasp of word meanings tend to go together. Our words are naturally assumed to express our concepts. Thus, natural to suppose that concepts are in effect meanings. Hence, expect concepts to play an important role in the semantic portion of the lexicon.
Themes for today.

The grammatical aspects of the lexicon put a surprising amount of space between our ordinary concepts and the semantics encoded in the lexicon. Though concepts do play an important role in the lexicon, there is an equally important role for the linguistic packaging of concepts. Shows how word meaning can relate to concepts, but still be distinctively linguistics. Raises some old and hard questions about linguistic relativism for concepts.
Michael Glanzberg (UC Davis) Concepts, Meaning, and the Lexicon November 2010 3 / 67
The Plan
The Lexicon: Grammar and Packaging of Meaning
Roots and Concepts
Word Meaning, Understanding, and Grammar
Language and Concepts Revisited
November 2010
4 / 67
Some Big Facts about Word Meaning I

Variety and idiosyncrasy.
Not really important, but Websters Third contains over 45,000 entries. More important: lexical categories are open classes. Lexical categories host idiosyncrasy in what we can express. Presumably, derived from some idiosyncrasy in what concepts we form, or how we think about the world. Fits well with the idea that the meanings of genuinely lexical items are closely tied to our concepts.
Structure.
Some reasonably coherent classes, e.g. aspectual classes or semantic classes (mostly for verbs) like sound emission (buzz), contact by impact (hit), psych verbs (frighten, fear), etc. (cf. Levin, 1993). Limited number of thematic roles. Restrictions on organization, e.g. thematic hierarchies.
Some Big Facts about Word Meaning II

Some aspects of structure could apply to concepts too, e.g. no surprise that concepts can be grouped in terms of categories like sound emission, contact by impact, etc. Some aspects of structure seem to indicate distinctively linguistic constraints, e.g. the limited number of thematic roles contrasts with our extensive ability to conceptualize participation in events in many different ways.
Interactions with grammar.

Relatively uncontroversial that there are some, though very controversial just what, and how meaning and grammar interact. Example: telicity correlates with temporal modier selection: (1) a. John was happy for an hour/*in an hour. b. Max found Mary in an hour/*for an hour.
November 2010
6 / 67
Some Big Facts about Word Meaning III

Example: properties of argument realization. Much-discussed case from Fillmore (1970).
Break class: bend, fold shatter, crack, . . . Hit class: slap, strike, bump, stroke, . . . Break verbs but not hit verbs enter into the causative alternation: (2) a. i. The boy broke the window. ii. The window broke. b. i. The boy hit the window. ii. * The window hit.
Similar surface syntactic properties in transitive occurrences, but different semantic properties, e.g. Rappaport Hovav & Levin (1998) take hit to be a manner verb, while break is a result verb.
Some grammatical interactions might derive from interactions with associated concepts, but hard to see how e.g. the hit/break contrast would (though NB categories of contact by impact and change of state).
Some Big Facts about Word Meaning IV

Variation across languages.
Blush (English) versus arrossire (Italian) (Levin & Rappaport Hovav, 1995; McClure, 1990). Near synonyms, but:
Blush an activity. Atelic. Arrossire is an achievement. Telic. Something like become red in cheeks.
Counterpart of hit in Lhasa Tibetan obligatorily takes a locative marker on the argument for the contacted object (Levin & Rappaport Hovav (2005) citing DeLancey MS): (3) shing*(-la) sta=re-s gzhus-pa tree-LOC axe-ERG hit Hit the tree with an axe.
Insofar as concepts like BLUSH and HIT are presumably similar for speakers of Italian, English and Tibetan (pace Whorf and Sapir), this is hard to explain simply in terms of associated concepts.
November 2010
8 / 67
The Packaging Approach

The big facts reveal a combination of idiosyncrasy and structure associated with word meaning, both within and across languages. An approach to the big facts: Packaging meaning.
Meanings are a combination of some idiosyncratic content and structural elements. The structural elements package the idiosyncratic content. Can do so differently, e.g. blush and arrossire package the same idiosyncratic content of reddening of cheeks in different ways. Limited range of structural elements should provide some explanation of structure within a lexicon. Structural elements have substantial linguistic properties, including grammatical ones, providing a basis for explaining interactions with grammar.
Ordinary concepts connect most closely to idiosyncratic content, not linguistic structural elements.
November 2010
9 / 67
Big Facts, Big Questions

The packaging view invites some big questions.
What is the nature of the idiosyncratic elements? What is the nature of the structural elements? How do they combine?
These are all a mix of foundational and empirical issues. Goal today is to shed some light on their foundational aspects.
Ask how the idea of packaged meanings relates to our ordinary concepts. Argue it shows surprising space between such ordinary concepts and the meanings our words actually have. Argue it shows that substantial structural articulation, even syntax-like structure, must be part of word meaning even if it is not part of the intuitive concepts related to those words. Thus, afrm a somewhat Fregean idea that only in the context of a sentence does a word have a meaning. Ask if the relation between ordinary concepts and word meanings indicates a form of linguistic relativism.
Two Views of Packaging

To do this, consider two examples of approaches to the lexicon and its role in grammar. Structure in lexical entries.
Structural and idiosyncratic elements combine in the lexicon. Word meanings are rich articulated structures.
Structure in syntax.
Simple lexical entries: word meaning has minimal linguistic structure. Structure is determined by syntactic processes, xing the environment in which an expression appears.
Will not try to decide between these (it is substantially an empirical matter). Will use them to help isolate enough features of idiosyncratic and structural elements to address our foundational questions.
November 2010
11 / 67
Structure in Lexical Entries

Following Levin & Rappaport Hovav (1995); Rappaport Hovav & Levin (1998); Levin & Rappaport Hovav (2005). Widespread idea, cf. Bierwisch & Schreuder (1992), Pinker (1989), Wunderlich (1997), etc. An event decomposition approach. Predicate decomposition within the lexical entry describes decomposition of an event into structural components. Example, open: externally caused change of state. (4) a. open b. [[x ACT] CAUSE [BECOME [y OPEN ]]]
Features of the analysis: packaging within the lexicon.

A root element OPEN . An event-structural frame, built from elements including CAUSE and BECOME. Decomposes the meaning into a two-part event structure.
Not going to worry about whether the details are correct.

A Little about Argument Realization I

Semantic structure in the lexicon determines grammatical behavior of expression. Main example: determines how arguments are projected in syntax. Does so via some interface linking rules. For example: lexical entry for open predicts usual transitive argument structure of agent subject and theme object.
Argument structure present in lexical entry via variables and the subevents they are in. Argument XPs in syntax correlate with these. Linking rule: immediate cause variable projected as external argument. Linking rule: variable for object undergoing change projected as direct internal argument (Levin & Rappaport Hovav, 1995).
Semantic structure does most of the work of xing argument realization, supplemented by linking rules.
A Little about Argument Realization II

Need to explain causative alternation: (5) a. Mary opened the door. b. The door opened. One option: binding of external cause within the lexicon (Levin & Rappaport Hovav, 1995). Maybe via existential-quantier like operation in the lexicon, operates on lexical entry to provide modied structure for linking rules. Something like [[x(x ACT)] CAUSE [BECOME [y OPEN ]]]. Unaccusative structure results: only internal argument is projected.
November 2010
14 / 67
A Little about Argument Realization III
Another possibility: really an argument projected, typically realized by some morpheme, but not overt in English. Reexive morphemes in Romance (cf. Chierchia, 2004): (6) La porta si aperta. The door REFL is opened.
November 2010
15 / 67
A Syntactic Analysis I
The tradition from Larson (1988), Baker (1988), Chomsky (1995), Hale & Keyser (1993, 2002), Kratzer (1996), and Marantz (1997). I am following a version presented by Harley (2007). Little v analysis, causative variant: (7) vP
DP Mary v0 0 / CAUSE
v v0 A open DP the door SC A tA
November 2010
16 / 67
A Syntactic Analysis II
Little v analysis, inchoative variant: (8) v0 v0 0 / BECOME A open DP the door vP SC A tA
November 2010
17 / 67
A Syntactic Analysis III
Important features.
Roots are elements of syntax, frequently non-verbal (Hale and Keyser emphasize the role of adjectival roots in creating causatives. Marantz thinks in terms of acategorial roots.) Structural elements like CAUSE are added by distinct syntactic positions. The intuitive verb is created by a syntactic process involving the root and v, such as incorporation. Packaging is thus articulated in syntax, not in the lexicon itself. Argument structure is determined by syntax, e.g. external argument is Spec of v.
November 2010
18 / 67
A Syntactic Analysis IV
Some issues to put aside.
The version sketched (Marantz, 1997; Harley, 2007) assumes difference between causative v projecting a specier and inchoative v not. Then not clear how to explain why some verbs fail to alternate, require agentive external arguments, etc. Original Hale & Keyser (1993, 2002) version is even more syntactically based. (Verbal heads are typically empty. Assume A requires its verbal host to project a specier. Allows embedding under a higher verb, and then the EPP requires adjoining an external argument.) Syntax for Hale and Keyser is l-syntax, within the lexicon. Meaning is located in a syntactic construction. Some relations construction grammar (Goldberg, 1995), but also important differences.
November 2010
19 / 67
Similarities and Differences
On both views, roots are relatively simple in linguistic structure. Both package roots. One does so by articulating structure in the lexicon. The other via syntactic structure. Genuinely distinct positions on nature of the lexicon and on mechanism of argument realization. But enough similarity across the views that deciding between them can be a complex matter. I shall not try to argue for one over the other.
November 2010
20 / 67
Where Are We?

So far, we have:
Seen the idea that lexicon involves roots and the packaging of roots with distinctively linguistic items. Seen that it addresses some of the big facts about meaning. Seen two examples of what packaging might be like.
Now, on to:
Argue roots are concepts. Argue that they are not the ordinary concepts that correspond to a verbs meaning. Thus, show there is more space between our ordinary concepts and our lexicon that a view embracing concepts might have lead us to expect. Wonder what we should conclude from this about grasp of meaning. Wonder what we should conclude from this about our concepts.
November 2010
21 / 67
Where Are We?
Roots and Concepts
November 2010
22 / 67
What Are Roots? I

We have seen two uses of a notion of root. Both suppose: Roots are sources of idiosyncratic meaning. Roots are source of openness of lexical categories. Roots (tokens!) are linguistically more or less atomic (Grimshaw, 2005): few if any linguistically signicant properties of lexical entry determined by the root. Rather, linguistic properties like argument structure are determined by the congurations (syntactic or semantic) in which roots appear. Roots are typed.
Rappaport Hovav & Levin (1998): types include STATE, MANNER, LOCATION/THING, etc. Hale & Keyser (2002): types are associated with syntactic categories Noun, Adjective, Preposition, and sometimes Verb . Some persistent match-ups: e.g. STATE corresponds to Adjective.
What Are Roots? II

Some cross-cutting of frameworks. Many manner verbs are analyzed as denominal by Hale and Keyser. Run is the incorporation of a nominal root with a verbal head, but has a frame [x ACT RUN ] for Levin and Rappaport Hovav.
The root type determines what structures it can appear in.

Frames select types for Levin and Rappaport Hovav, e.g. [y STATE ]. Syntax imposes restrictions on congurations for Hale and Keyser. Categorial properties of root types (selectional properties) determine which syntactic congurations they appear in.
Care about linguistic atomicity:

Root type is linguistically signicant for determining which structures a root can gure in, but it does not generate these structures. Linguistic explanations e.g. of argument structure, run off the structural congurations, not the root types.
November 2010
24 / 67
What Are Roots? III

Root types have core semantic properties.
STATE/Adjectives are fundamentally predicates of individuals. THING/Nouns (not DPs) pick out kinds of things, e.g. shelf, oil. (NB for Hale and Keyser get only external arguments, to account for unergative structure.) MANNER/Verbs/Nouns characterizes kinds of events.
Core semantic property of roots is predicational/categorizing nature. So roots are:

Elements with content. Contents are predicative. Contents are groupable according to type. Types determiner linguistic congurations in which roots can occur.
November 2010
25 / 67
Roots Are Concepts?

Invites the idea that roots are concepts. Concepts are typically predicative, in that they classify elements that can fall under them. Concepts can be grouped into types according to their kinds of contents. Explained if lexicon contains elements built from concepts:
Open-ended and idiosyncratic aspects of the lexicon. Typing of roots. Linguistic atomism of roots: they are outside the language faculty, and so no substantial generalizations about language apply.
Controversial, but some empirical support from a role for concepts in acquisition (e.g. Bloom, 2000). The interface picture:
Roots are points of interface between the language faculty and the wider cognitive makeup of the person. Function like pointers to concepts outside of the language faculty proper.
Roots Are (Usually) Monadic!

For Levin and Rappaport Hovav, they are typically monadic elements of types like STATE, THING, MANNER. For syntactic approach, they are typically monadic elements like Nouns or Adjectives, which take at most one argument. Not an accident: the systems we have looked at build complex congurations which provide argument positions for verbs. They are built around simple roots, which are typically monadic. For instance, in the form of the syntactic approach we looked at, arguments are invariably speciers. Some complications:
For Hale and Keysers own approach, P roots are relational. For Levin and Rappaport Hovav, can have e.g. manner roots like SWEEP . Tokens determine additional argument position, but not one corresponding to an event structure position. Might get an extra event argument, but not counting that.
Even so, monadic roots are fundamental for and pervasive in lexical entries on both approaches we have considered.
Many Ordinary Concepts are Polyadic! I

We appear to posses some genuinely polyadic concepts (e.g. Pietroski, forthcoming).
Not always easy to x the adicity of a concept: e.g. TRIANGLE can be monadic (x is a triangle) or triadic (lines x, y , z form a triangle) (Pietroski, forthcoming). Even so, natural to assume that some familiar concepts are at least dyadic, e.g. HIT, BREAK, etc. Many of our concepts appear to involve agency or causality, as well as an affected object.
Empirical support?
Studies of reorientation in rats suggest they rely on geometric relations (e.g. Gallistel, 1990). Likewise for children at 1.52 years (e.g. Spelke, 2002). Long tradition (from Michotte, e.g. 1963) of studying the perception of causality indicates perception of causality in adults and in infants as young as four months (e.g. Scholl & Tremoulet, 2000; Saxe & Carey, 2006).
Many Ordinary Concepts are Polyadic! II
These suggest we nd relational concepts, even in non- or pre-linguistic creatures. But, the empirical situation remains murky, and many of the results can support multiple interpretations.
November 2010
29 / 67
Roots Are Not the Concepts We Thought They Were

Roots (often) cannot be the concepts naturally seen to correspond to verbs.
Concepts like BREAK are polyadic. But the root of break is monadic, on both approaches we have considered.
Moreover: Roots never encode agency or causality, while many concepts do. Roots can still be pointers to concepts, but not the ones we expected. Lexical items are thus not the linguistic packaging of intuitively corresponding concepts. The lexical entry for break does not simply package BREAK. A feature of the interface.
Interface between lexicon and wider conceptual abilities (substantially) constrained to pointers to monadic concepts. Typing of roots and other aspects of packaging rely on this.
No Easy Way Out I

An easy way to get monadic concepts: suppress additional arguments.
Might seem like a device available in lexicalization, for tting polyadic concepts into frames requiring monadic roots. Allows the mapping of words to concepts to remain fairly tight. Existential quantication as the natural mechanism: BREAK(x, y ) xBREAK(x, y ) = BROKEN ?
Intuitively seems wrong: xBREAK(x, y ) is not a state, whereas the lexical entry seems to call for something stative. Overgeneration? Predicts the existence of verbs that seem somewhat dubious.
If existentially binding arguments in concepts yields state-like roots for the lexicon, expect it could apply to either argument of a dyadic concept. Hence, we should also expect to have things like y BREAK(x, y ), yielding a root BREAKER .
No Easy Way Out II

Are there verbs like:
[[x ACT] CAUSE [BECOME [y BREAKER ]]].
Requires agentive object? Not an experiencer. Seems unlikely?
[BECOME [x BREAKER ]].

Something like an internally caused change of state verb, but change of state is itself a causal one?
Not really sure, but these seem rather dubious.
xBREAK(x, y ) predicts wrong meaning.

Assume agentive root. Then frame describes x doing an act of causing there to be someone who breaks y . Seems to count, e.g. contributing to the delinquency of a minor as breaking a window.
Seems we cannot just force ordinary concepts into the lexicon in any such simple way.
November 2010
32 / 67
The Causal Case I

Roots for causative verbs tend to describe result-states. Often adjectival (in the Hale-Keyser system).
(Break is less clear: interpreted as a result-state causative, but zero-related nominal break. Hale & Keyser (2002) gloss it as a denominal verb.)
Result states in Levin and Rappaport Hovavs lexical entries (Rappaport Hovav & Levin, 1998):
Open: [[x ACT] CAUSE [BECOME [y OPEN ]]] Break: [[x ACT] CAUSE [BECOME [y BROKEN ]]] Contrast with run: [x ACT RUN ] Contrast with hit: [x ACT HIT y ].
Rather than select a concept corresponding to the act in question, selects one capturing the result state of the act. Rebuilds causal and agentive aspects in purely linguistic terms.
November 2010
33 / 67
The Causal Case II
Indicates a lexicalization process much more complex than merely mapping to concepts.
Select a specic type of monadic concept, e.g. result state. Related to the ordinary concept associated with a verb, but not always the most directly connected one. Use that as a root. Package the root in a way that reconstructs key aspects of the ordinary concept, within a linguistically determined frame.
November 2010
34 / 67
Why Do That? I
A few speculations. Much discussed points from the acquisition literature: learning verbs is hard (cf. Poulin-Dubois & Graham, 2007).
Many possible verbs meanings apply to events children observe (cf. Fisher et al., 1994; Gleitman, 1990). Delay between event and description. Bias towards nouns in early vocabulary (some controversy).
Perhaps(!) selecting certain sorts of roots, e.g. result states, helps simplify this process. Might be a form of linguistic bootstrapping (using linguistic sources of information to help in the task of learning words or other aspects of language).
The Hale and Keyser approach might be closer to syntactic bootstrapping (Fisher et al., 1994; Gleitman, 1990). The Levin and Rappaport Hovav approach also builds in some elements closer to semantic bootstrapping (Pinker, 1989, 1994).
Why Do That? II
Might reduce choices if e.g. constrained to nd result states rather than any number of causal concepts associated with an event? Uniform monadicity is generally a simplication in complexity (Hurford, 2007; Pietroski, forthcoming). Might simplify other aspects of lexicon to have monadic roots and a few elements to package them into polyadic constructions, and so simplify the wider task of language acquisition?
November 2010
36 / 67
Where Are We?

So far, we have:
Seen that roots are concepts. But they are not the concepts we should expect. They are typically monadic concepts, like result states, etc. The lexicon packages these in ways that recover features of our ordinary concepts, like causality.
Now, on to:
Argue that our grasp of meaning reects the complex packaging in the lexicon. Hence, grasp of meaning, even if derived from our concepts, is grasp of complex linguistic structure. Note this indicates a sort of context principle. Then ask what it might tell us about our concepts.
November 2010
37 / 67
Where Are We?
Roots and Concepts
November 2010
38 / 67
Fine, but I Want to Know about MEANING

Much of the motivation for the sorts of highly structured lexical entries we have been exploring is grammatical (often syntactic). How does this relate to the philosophical idea of meaning: something like what I understand when I understand a word, what I am trying to convey to my hearer, etc.? It appears the meaning of a verb in this sense cannot simply be the root.
Clear that we recognize causal or agentive aspects of many causative verb meanings: open and break do not have meanings simply providing result states, but indicate that an act of opening or breaking was carried out. Thus, the only place in the lexicon we will nd these aspects of meaning is in the contribution of the frame, especially the CAUSE element.
But do we actually grasp the frame elements, or import the idea of causation from e.g. the ordinary concept, rather than the lexicon?
Meanings from Structural Elements I
Claim: the abstract element CAUSE is part of our grasp of the meanings of many causative verbs. Argument (from best explanation).
Common feature of a wide range of verbs entering into the causative/inchoative alternation, including, among many: abate, decrease, expand, grow, soak, topple, as well as drop, break, bend, etc (cf. Levin, 1993) Common feature is a feature of our grasp of these words. Common element can be glossed as a very abstract notion akin to causation: all of these involve something like an agent doing something like bringing about a change into a result state.
November 2010
40 / 67
Meanings from Structural Elements II

In many cases, CAUSE is not strictly causation: (9) a. Dissension toppled the government. b. Time abated the damage. c. John broke the window.
Some of these (e.g. break might involve canonical instances of causation. Many, like topple and abate involve a much more abstract and inclusive notion of bringing about a change of state. (Dissension does not enter into genuine causal relations!) Likewise a very extended sense of agency. Figurative language? Not obviously, though if so, the gure would have to stem from some cause-like notion anyway, so not a worry. (Possibility of gurative extension of causal aspects is further evidence that the abstract elements make their way into our understanding.)
November 2010
41 / 67
Meanings from Structural Elements III
Frequently noted point that in certain ways, CAUSE is more restricted than causation (e.g. Dowty, 1979; Parsons, 1990; Pietroski, 2005): (10) a. I caused the window to become broken, by hiring a kid to throw a brick through it. b. # I broke the window, by hiring a kid to throw a brick through it.
Common suggestion that CAUSE is restricted to direct causation (cf. Fodor, 1998).
November 2010
42 / 67
Meanings from Structural Elements IV
Common aspect of our grasp of meaning explained if it is the abstract notion of CAUSE contributed by the structural frame in the lexicon. Conclude that we indeed grasp that highly abstract notion as part of our understanding of a wide range of causative verbs. We presumably infer real causation in cases like break from our worldly knowledge of how things become broken.
November 2010
43 / 67
Meanings from Structural Elements V

Further evidence: we recognize other event-structural aspects of verb meaning.
Consider again an inchoative (and apparently unaccusative) like open: (11) The door opened.
Our grasp of the meaning of the verb here includes a sense of change of state, witnessed by our willingness to infer that the door was previously not open. Grasp of content not exhausted by the root OPEN , which is purely the state of being open. Again, explained if we attribute grasp of the structural elements to speakers. Again, captures a pattern across a wide range of verbs.
November 2010
44 / 67
Grasp of Structures I
If this is right, grasp of the meaning of a verb is grasp of not simply of a concept, but of a highly structured linguistic object. Might be a highly structured lexical entry, like: (12) [[x ACT] CAUSE [BECOME [y OPEN ]]] Grasp includes that of root concepts like OPEN and of structure of frame. Might be a syntactic structure, like: (13) v0 v0 0 / CAUSE
v SC A open DP the door A tA

November 2010 45 / 67
Grasp of Structures II
Grasp includes terminal nodes, and structure of the syntactic conguration. Is it strange to have grasp of meaning include so much structure, maybe even so much syntactic structure? Not really, if you remember the role of morphology. Grasp of inchoative open becomes like our grasp of -en verbs like redden: (14) v0 v0 -en BECOME A red DP the sky vP SC A tA
November 2010
46 / 67
A Context Principle? I
Frege instructs us that It is only in the context of a sentence that words have any meaning. Many reject this: makes no sense of how we build meaningful sentences out of words compositionally, how words encode concepts, how words are acquired, etc. A modied principle: Only in the context of a grammar (language, faculty, etc.) does a word have a meaning. Grasp of word meaning is grasp of something grammatically complex (but not necessarily a sentence).
Grasp of word meaning is grasp of complex lexical items. Only exist within a grammar that provides the packaging elements, and combines them in linguistically acceptable ways. Can be syntactically structured congurations, or purely lexical ones. Regardless, cannot have the needed grasp without having the grammar.
A Context Principle? II
Still a role for concepts.
Still provide idiosyncratic content. Provide roots in the lexicon. Form cores of meanings of words.
Still have a substantial role for compositional construction of meaning.

Sentence meaning is built up from lexical entries plus syntax. Speaker can have independent grasp of those meanings, and use them to compose a sentence whose meaning is thereby grasped.
But, see word meaning as built up from selected concepts by grammatical processes.
November 2010
48 / 67
Where Are We?

So far, we have:
Argued that word meanings are not simply the corresponding ordinary concepts.
Roots are concepts, but they are packaged by distinctively linguistic structure in the lexicon. The roots packaged are monadic, while ordinary concepts are frequently polyadic. Packaging provides linguistic reconstructions of features like cause or agency.
Argued that our grasp of our words includes grasp of the the structural components of lexical entries.
Hence, grasp of meaning is in part grasp of complex linguistic structure. Implies a modied context principle.
Now, on to:
Reconsider the status of ordinary concepts in light of our conclusions about grasp of word meaning. Reconsider the relation of concepts to language. Fret about linguistic relativism.
Where Are We?
Roots and Concepts
November 2010
50 / 67
A Puzzling Consequence
Conclusion reached so far suggests we have parallel elements:

Ordinary concepts like BREAK. Complex lexical entries like [[x ACT] CAUSE [BECOME [y BROKEN ]]] (or syntactic version).
Have argued we do grasp complex lexical structures. If so, we seem to have in our cognitive repertoire two distinct break elements. This might seem intuitively odd, or dubious from our introspective access to how many concepts we have.
November 2010
51 / 67
Options I
1
Conclude that structural aspects of the lexicon really reect our wider cognitive abilities (various forms of cognitive grammar, Jackendoffs (e.g. 1990) conceptual semantics?).
Conclude that the lexicon really provides an articulation of our ordinary concepts. Have been implicitly arguing against this:
Argument that CAUSE is distinctively linguistics. Role of argument projection and other aspects of grammar in determining structural elements. Restricted range and distribution of structural elements. Documented variation across languages of some features of packaging.
Reject the supposition that our ordinary concepts like BREAK lack distinctively linguistic structure.
Never really had such concepts prior to language development? Had prototypical versions that are replaced by lexical ones? Implies some degree of linguistic relativism.
November 2010
52 / 67
Options II
Accept the consequence and minimize the oddness.

Motivate parallel lexical and ordinary conceptual elements. Minimize the differences: both have core conceptual elements in common. The position that has been implicitly adopted up till now.
I shall briey explore the second two options.
November 2010
53 / 67
Option 2 and Linguistic Relativism I
Suppose our seemingly ordinary concepts are really articulated lexical entries with distinctively linguistic structure. That structure can vary across languages. Thus suppose we only have the concepts we have in virtue of having language, and which language determines which concepts we have. The specter of the Whorf-Sapir hypothesis (Carroll, 1956) raises its head!
November 2010
54 / 67
Option 2 and Linguistic Relativism II
But, an extremely mild form of linguistic relativism (not really Whoran).

Concepts could share root elements. Variation only across a limited range of packaging options. Does not indicate radical differences in ways of conceptualizing the world or kinds of thoughts. (No claim e.g. that one group has a fundamentally different concept of time.) Only indicates small differences in just how we conceptualize events (e.g. for blush, change to state versus enduring state).
Not obviously unacceptable, or vulnerable to the many objections raised to Whorf and Sapir.
November 2010
55 / 67
Option 2 and Linguistic Relativism III

A somewhat more Vygotskian (1962) model: language involved in shaping concepts as they develop.
Might start with some coarsely individuated concepts (vague conglomerations of individuals). Rene via development of more structured lexical entries. In effect replace primitive concepts.
Still implies modest linguistic relativism, but not in the basic sources of thought. An updated version.
Start with coarsely individuated concepts, e.g. conceptual roles, causal covariation, etc. Not ne-grained enough to determine particular predicates of events. Rely on linguistic structure to provide additional grain and x event predicates. Fully articulated predicates of events replace coarse-grained concepts when available.
Option 2 and Linguistic Relativism IV

The updated version allows:
Us to grant that at a certain level of abstraction we have the same thoughts. Us to hold that fully articulated concepts can depend on language. For mild linguistic variation in fully articulated concepts.
Might render option 2 acceptable?

Updated Vygotskian view allows us to better capture a sense in which thought is not language-relative. But avoids the parallel conceptual repertoires. Evidence on way or another? Perhaps from pre-linguistic children or non-linguistic animals?
November 2010
57 / 67
Option 3 and Lexicalization I
Option 3 proposes it is not so odd to have ordinary concepts and word meanings come apart.
Only in certain respects: linguistic packaging and root selection. Makes word meanings not simply the concepts closely linked to those words.
Much like the Vygotskian alternative of option 2, but without the replacement thesis.
Have ordinary concepts characterized by e.g. functional role or causal covariation (or whatever else you like). Have lexical entries that are more nely individuated, e.g. predicates of events (for verbs). The latter need not supplant the former.
November 2010
58 / 67
Option 3 and Lexicalization II

Probably a fair bit of our most sophisticated thinking makes direct use of lexical items.
When you avoid breaking something, you can rely on BREAK. If you ask if breaking involves two events or one, or if the inference in the causative alternation is valid, you may be making use of the lexical entry for break.
Makes lexicalization semantically as well as syntactically substantial.

Produces new highly articulated elements of cognition whose contents we grasp. These differ in content from ordinary concepts (though in small and subtle ways). They encode linguistic as well as root conceptual structure.
November 2010
59 / 67
Which Option to Choose?

Already noted I have in passing provided reasons not to take option 1. Options 2 and 3 both seem viable. I prefer option 3:
Do not see any direct support for the replacement thesis. Do see a role in cognition for ordinary concepts. See a parallel role for lexical items in sophisticated thought.
But not clear to me what tests would conclusively favor one over the other.
Can observe the role of a concept in an agents non-linguistic life. But, not clear if we can probe for the sorts of ne-grained differences in concepts involved without relying on word meanings? So, not clear what would tell us whether an ordinary concept was replaced, rather than merely that a word meaning was under scrutiny.
So, conclude by asking which of these options we should really choose.

References I
Baker, M. C. (1988). Incorporation. Chicago: University of Chicago Press. Bierwisch, M. & Schreuder, R. (1992). From concepts to lexical items. Cognition, 42, 2360. Bloom, P. (2000). How Children Learn the Meanings of Words. Cambridge: MIT Press. Carroll, J. B. (Ed.) (1956). Language, Thought, and Reality: Selected Writings of Benjamin Lee Whorf. Cambridge: MIT Press. Chierchia, G. (2004). A semantics for unaccusatives and its syntactic consequences. In A. Alexiadou, E. Anagnostopoulou, & M. Everaert (Eds.), The Unaccusativity Puzzle, pp. 2259. Oxford: Oxford University Press. Chomsky, N. (1995). The Minimalist Program. Cambridge: MIT Press. Dowty, D. R. (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel.
References II
Fillmore, C. J. (1970). The grammar of Hitting and Breaking. In R. A. Jacobs & P. S. Rosenbaum (Eds.), Readings in English Transformational Grammar, pp. 120133. Watham: Ginn. Fisher, C., Hall, D. G., Rakowitz, S., & Gleitman, L. (1994). When it is better to receive than to give: Syntactic and conceptual constraints on vocabulary growth. Lingua, 92, 333375. Fodor, J. A. (1998). Concepts: Where Cognitive Science Went Wrong. Oxford: Oxford University Press. Gallistel, C. R. (1990). The Organization of Learning. Cambridge: MIT Press. Gleitman, L. (1990). The structural sources of verb meanings. Language Acquisition, 1, 355. Goldberg, A. E. (1995). Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press.
References III
Grimshaw, J. (2005). Semantic structure and semantic content in lexical representation. In Words and Structure, pp. 7589. Stanford: CSLI Publications. Hale, K. & Keyser, S. J. (1993). On argument structure and the lexical expression of syntactic relations. In K. Hale & S. J. Keyser (Eds.), The View from Building 20, pp. 53109. Cambridge: MIT Press. (2002). Prolegomenon to a Theory of Argument Structure. Cambridge: MIT Press. Harley, H. (2007). The bipartite structure of verbs cross-linguistically. Ms. Hurford, J. R. (2007). The Origins of Meaning. Oxford: Oxford University Press. Jackendoff, R. (1990). Semantic Structures. Cambridge: MIT Press.
November 2010
63 / 67
References IV
Kratzer, A. (1996). Severing the external argument from its verb. In J. Rooryck & L. Zaring (Eds.), Phrase Structure and the Lexicon, pp. 109137. Dordrecht: Kluwer. Larson, R. K. (1988). On the double object construction. Linguistic Inquiry, 19, 335391. Levin, B. (1993). English Verb Classes and Alternations. Chicago: University of Chicago Press. Levin, B. & Rappaport Hovav, M. (1995). Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge: MIT Press. (2005). Argument Realization. Cambridge: Cambridge University Press. Marantz, A. (1997). No escape from syntax: Dont try morphological analysis in the privacy of your own lexicon. University of Pennsylvania Working Papers in Linguistics, 4, 201225.
References V
McClure, W. (1990). A lexical semantic explanation for unaccusative mismatches. In K. Dziwirek, P. Farrell, & E. Mejas-Bikandi (Eds.), Grammatical Relations, pp. 305318. Stanford: CSLI Publications. Michotte, A. (1963). The Perception of Causality. London: Methuen. Translation by T. R. Miles and E. Miles from the French edition of 1954. Parsons, T. (1990). Events in the Semantics of English. Cambridge: MIT Press. Pietroski, P. (2005). Events and Semantic Architecture. Oxford: Oxford University Press. (forthcoming). Semantic monadicity with conceptual polyadicity. Ms. Pinker, S. (1989). Learnability and Cognition. Cambridge: MIT Press. (1994). How could a child use verb syntax to learn verb semantics? Lingua, 92, 377410.
References VI
Poulin-Dubois, D. & Graham, S. A. (2007). Cognitive processes in early word learning. In E. Hoff & M. Shatz (Eds.), Blackwell Handbook of Language Development, pp. 191211. Oxford: Blackwell. Rappaport Hovav, M. & Levin, B. (1998). Building verb meanings. In M. Butt & W. Geuder (Eds.), The Projection of Arguments, pp. 97134. Stanford: CSLI Publications. Saxe, R. & Carey, S. (2006). The perception of causality in infancy. Acta Psychologia, 123, 144165. Scholl, B. J. & Tremoulet, P. D. (2000). Perceptual causality and animacy. Trends in Cognitive Science, 4, 299309. Spelke, E. S. (2002). Developing knowledge of space: Core systems and new combinations. In A. M. Galaburda, S. M. Kosslyn, & Y. Christen (Eds.), The Languages of the Brain, pp. 239258. Cambridge: Harvard University Press.
November 2010
66 / 67
References VII
Vygotsky, L. S. (1962). Thought and Language. Cambridge: MIT Press. Edited and translated by E. Hanfmann and G. Vaker, from the Russian edition of 1934. Wunderlich, D. (1997). Cause and the structure of verbs. Linguistic Inquiry, 28, 2768.
November 2010
67 / 67

Concepts, Meaning, and The Lexicon: Philosophy of Language Meets The Syntax-Lexical Semantics Interface

Uploaded by

Copyright:

Available Formats

Concepts, Meaning, and The Lexicon: Philosophy of Language Meets The Syntax-Lexical Semantics Interface

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Concepts, Meaning, and The Lexicon: Philosophy of Language Meets The Syntax-Lexical Semantics Interface

Uploaded by

Copyright:

Available Formats

Concepts, Meaning, and the Lexicon: Philosophy of Language Meets the Syntax-Lexical Semantics Interface

Michael Glanzberg (UC Davis)