Theoretical Overview of Columbia School Linguistics

Columbia School Theory

by Wallis Reid

 Columbia School linguistics takes as its ultimate object of explanation the perceptible sounds of speech (or sequence of symbols in a written text). Instances of human communicative behavior, then, are its primary data. It accounts for these observables by postulating abstract cognitive systems upon which speakers appear to be operating. Columbia School is thus an avowedly explanatory enterprise and neither a throwback to behaviorism nor an instance of an Externalized (E-) language approach.

The basic structural unit in the cognitive systems is a sign—a signal paired with a meaning.  (This is reminiscent of Saussure's signe linquistique composed of a signifiant and a signifié). Both the signals and their meanings are language-particular rather than universal; each language offers its own semantic categories. The analytical problem for the Columbia School linguist is to determine the actual identity of these signal-meaning units in a particular language. This is done by testing proposed signs against actual usage; a hypothesized meaning must make some form of contribution to every message for which its signal is used. This demonstration involves both case-by-case analysis of authentic examples in context, and quantitative testing for predicted skewings of signals throughout a text.  

 The explanation for the appearance of a particular signal in a text is that its associated meaning—hypothesized and tested by the analyst—contributes semantically to the message being conveyed. A typical Columbia School question would be, What motivates speakers of English to say sometimes broken hearts and other times breaking hearts? What is the difference in meaning between the signals -ing and -en that is guiding their choice? Or, What motivates speakers of Spanish to say sometimes le escribí and others lo escribí, sometimes le llamo, others lo llamo? The sequential order of signals is addressed as well; for example, What motivates the choice between the order of signals in he left the house windowless and he left the windowless house? Here we find two kinds of explanations: either a particular feature of word order is due to natural iconic principles, or it is itself a signal of a grammatical meaning.

 This mode of explanation is goal-directed rather than formal because the choice of a sign, either grammatical or lexical, is explained in terms of what its 'chooser’—the language user—is attempting to accomplish, rather than in terms of syntactic rules. While the linguistic system has a well-defined structure, its deployment is affected by an open-ended number of factors and is thus not amenable to algorithmic formalization. Language use is creative in the everyday sense of that word.

 A second aspect of linguistic creativity is the conceptual leap that the hearer must make between individual linguistic meanings and the intended message. The message is more than the sum of its semantic parts because the meanings of many common signs are imprecise, functioning more as hints to the message than as conceptual fractions. Hearers must rely on context, social setting, life experience and common sense to jump to a message that is under-determined by the semantic input. This gap between semantic  input and output means that Columbia School theory espouses an inferential model of communication rather than the familiar compositional model.

 Columbia School is a highly restrictive theory that operates only with constructs motivated by its two orientations (that language is a communicative system and that it is used by intelligent humans). It does not assume the a-priori validity of the categories of the Western grammatical tradition. And its only data are  the observed asymmetry of sounds in speech or of graphic marks in writing. In Columbia School work one does not ask why reflexive pronouns in English are also sometimes emphatics (He saw himself in the mirror, He saw it himself); one seeks to understand the distribution of the form -self in its entirety, including these but also other utterance types. One does not inquire about English overt and null complementizers (He said he was coming, He said that he was coming); one seeks to understand the distribution of that in its entirety, including these but also other utterance types.

Typical Analytical Problems Studied by Columbia School Linguists

by Wallis Reid

All Columbia School grammatical analyses begin with a question about the distribution of a linguistic form in a particular language. Listed below are examples of such questions for which analyses have been done (full references in Bibliography).

1. Why do English speakers sometimes select a  –self form, while on other occasions select forms like he or she? (Stern 2001, 2006)                       

She bought flowers for herself.

She carried an umbrella with her.




2. Why do English speakers sometimes opt for -ing forms and on other occasions for to forms? (Wherrity 2001)

After a year she will like living in France.

Some day, she would like to live in France.

 

3. Why do English speakers sometimes put the ‘recipient’ in a to phrase and elsewhere do not? (Huffman 1996)

 I sent a package to Mary

I sent Mary a package

 

4. Why do singular event words in English sometimes occur with plural subjects (and vice versa)? (Reid 2001)            

The sex lives of Roman Catholic nuns does not, at first blush, seem like promising material for a book [Taken from a review in Newsweek]                            

 

5. Why is inversion in English used both for questions and non-questions?

Should he leave before supper?  He can eat in town.

Should he leave before supper, he can eat in town. (Diver, ms)

                  

6. Why does the word gun appear after fired in the first sentence and before fired in the second one when both sentences could describe the same scene?

The soldier fired a gun.

The gun fired a 40-mm shell.                                

 

7. In French, why is the form le used with some event words and the form lui used for others? (Huffman 1983, 1997)

Je le suis.              'I follow him.'

Je lui obeis.           'I obey him.'                             

 

8. In English, why does the characterizing word sometimes appear before the characterized and other times after it? (Davis and Hesseltine, 2020)

The boys painted the red barn.

The boys painted the barn red.                  

 

9. Why do speakers of English typically say ‘she came in a taxi’ but ‘she came on a bus’? (Reid, 2002)

 

10. In Spanish, why does the characterizing word sometimes appear after the characterized and other times before  it? (Klein-Andreu 1968, 1983)

 un viejo amigo        'an old (longtime) friend'

un amigo viejo        'an old (elderly) friend'           

 

11. In German, why is the least active participant appear sometimes with den and other times with dem with different event words? (Zubin 1972, 1980)

Sitta sieht den Jungen an              'Sitta looks (ansehen) at the boy.'

Sitta sieht dem Jungen zu             'Sitta looks (zusehen) at the boy.'

                                                                              

12. In Italian, why is a singular in-focus participant sometimes egli and sometimes lui? (Davis 1992, 1995a).

 

Egli non rispose       'He did not answer.'

Lui  non rispose        'He did not answer.'             




13. Can a connection be found between Hebrew triconsonantal roots  (CCC) and their meanings?  For example: /C-r-C/  roots appear to reflect a  general semantic field of       'Change in Structure' (Tobin, 2001).

 

14. In Swahili and other Bantu languages, a different and smaller set of event words is used in the negative than in the affirmative. Why? (Contini-Morava, 1989).

 

15. In Swahili and other Bantu languages, noun class prefixes signal information about noun class membership as well as number.  But the classes do not all pair into binary singular-plural sets.  What number information is actually signaled by the noun class prefixes and how does it relate to the noun classification system?  (Contini-Morava, 2000). 

 

16. Is there a semantic difference between what are traditionally called the synthetic and the periphrastic comparative and superlative constructions in English? (Tobin 1990)

I couldn't have made a more distinct comment if I tried.

... In fact, a distincter comment than that is difficult to imagine.

 

17. Why does one usually find that the response to someone saying I love you is

 I love you too rather than I also love you or I love you also? (Tobin, 1990)

 

What kinds of data do Columbia School analyses use?

by Wallis Reid

 Typically, Columbia School analyses rely heavily on data drawn from natural language use, often written but occasionally spoken. Fabricated data are occasionally used, but only to illustrate familiar and conventional usage. This is for two reasons. First, natural discourse provides examples of striking expressive creativity that would be suspect if made up by the analyst himself. Second, naturally-occurring examples offer the possibility of appealing to redundant features in the context in support of the interpret-ation claimed by the analyst. If the analyst were to create both the example (with its interpretation) and the context, the former could not be tested against the latter.

 Diver 1969 uses a single text, The Iliad.  Huffman 1997 uses a wide variety of French novels, adding to them all three volumes of  DeGaulle’s war memoirs. Davis 1992 uses texts supplemented by spoken data.  Klein-Andreu uses taped, spoken conversation in her work (see papers in Contini-Morava & Goldberg 1995 and Contini-Morava & Tobin 1999).  Reid 1991 draws heavily from contemporary written and spoken journalism. Questionnaire data play a significant role in Garcia and Otheguy 1972, Zubin 1976, Reid 1991 and Wherrity 2001. Wherrity 2001 also provides the first example of a corpus-based analysis.

 

Contrasts between Columbia School and Chomskyan frameworks

by Wallis Reid

The early outline of Columbia School theory emerged just as Chomsky’s Aspects of the Theory of Syntax appeared in the mid 1960's. While defining a linguistic position such as that of Columbia School in terms of its relation to the Chomskyan paradigm is no longer necessary, it is still a short-hand way of locating it on the linguistic map.

Listed below are a series of pithy and provocative contrasts between the Columbia School framework(CS) and the Chomskyan paradigm (CP), many of which distinguish Columbia School from various formal approaches as well.

 

Kinds of Problems

CP:  big questions about language and mind.

CS:  small questions: why do linguistic forms appear where they do?

 

Kind of Data

CP:  decontextualized sentences fabricated by the analyst

CS:  data drawn from extended discourse. Context crucial to testing meanings.

 

Nature of Language

CP: Essentially a representational system and only incidentally a communicative system.

CS : Essentially a communicative system (and not a representational system at all).

 

Basic Theoretical Unit

CP:  the sentence

CS:  the linguistic sign

 

Relation between Form and Meaning

CP: sentences have a formal structure independent of its meaning structure

CS: linguistic form and linguistic meaning united in the linguistic sign

 

Corollary 1: Syntax

CP: formal rules for what word sequences the language allows.

CS: no syntax in that conventional sense

 

Corollary 2: Grammaticality

CP: a bedrock irreducible fact about sentences

CS: an epiphenomenon that seems real chiefly to linguists.

 

Nature of Linguistically Encoded Meaning

CP: precise fractions of the message.

CS: imprecise hints and clues to the message

 

Relation between word and sentence ‘meaning’

CP: compositional relation

CS: instrumental  relation, i.e. meanings help to communicate a message without necessarily being conceptual fractions of the message

 

Model of Communication

CP: a telementational  model of communication

CS: an inferential model of communication

 

Psychological Process of Speaker

CP: rule-governed behavior

CS: goal-directed behavior

 

Psychological Process of Hearer

CP: decoding the message

CS: inferring the message from the linguistic meanings signaled and the larger context. 

 

Status of literal, a-contextual sentence ‘meaning’

CP: a linguistic object that can be formally represented.

CS: personal psychological experience; not a linguistic object.

 

Grammatical Categories

CP: a universal inventory of grammatical categories

CS: no universal inventory; language-particular grammatical categories

 

Object of Explanation

CP: grammaticality, structure and meaning judgments; language  “acquisition”

CS: the phonetic and graphic output of people using language to communicate.

 

Mode of Explanation

CP:  aspires to purely formal explanation

CS:  non-formal, functional explanation

 

Genetic endowment for language

CP: a language faculty consisting of formal and substantive linguistic universals.     

CS: a symbolic ability; problem-solving; jumping to conclusions; abduction

 

Faculty specificity

CP:   the language faculty is independent of other psychological faculties

CS:   Both linguistic structure and language use are shaped by recognizable human    psychological characteristics evident in other human behavior.

 

Linguistic meaning in Columbia School theory

by Alan Huffman

 The exact role of meaning in linguistic analysis is viewed in a seemingly limitless variety of ways by the various schools of linguistic thought. Yet nearly all, from formalism to functionalism, from traditional grammar to logical semantics, in practice agree on one thing: that linguistic meaning is compositional. The assumption is that sentence or propositional meaning is the sum total of the discrete meanings of the lexical and grammatical morphemes and the syntactic structures that compose it.  This requires a mapping between each fraction of sentence meaning and some lexical, morphological or syntactic feature of a sentence.

Columbia School theory embraces an alternative view of linguistic meaning, one aptly called an instrumental view. In this view, linguistically encoded speaker input is very sparse compared to communicative output.  The term ‘meaning’ is reserved for the constant input of a signal, such a grammatical or lexical morpheme.  Meanings are versatile tools that nudge output in one direction or another, but this much richer output, the message, is inferred by the hearer.  Messages are context dependent, but not derivable algorithmically.  Their inference is helped by the hint-like meanings plus contextual and extra-linguistic knowledge.  In this view, the role of  human intelligence, creativity, and inference is given explicit recognition in the communicative process, but not built into the communicative instrument.

The English word with provides a very simple illustration of this conception of linguistic meaning. The form with occurs in messages of ‘means’ or ‘instrument’ as in

Henry cut the cake with a knife.

 But a slight change to:

Henry cut the cake with the bride

changes the message radically. Now instrumentality is absent, and the message is one of cooperation. In retrospect, however, it is clear that the source of these messages can be traced to elements of context other than with itself. ‘Cutting’ requires an instrument; a ‘knife’ is a cutting instrument. A ‘bride’ is not a cutting instrument; rather, ‘bride’ and ‘cake’ conjure up images of harmony and cooperation.  With another change, we get yet a third message:

 Henry cut the cake with a smile

now a message of ‘manner’ or ‘accompaniment’.  Yet ‘accompaniment’ is literally true also of the first two messages: if I cut the cake with a knife, or with a bride, I am in fact accompanied by the knife or bride.

I suggest that the meaning of with is something like Attendant Circumstance. To be sure, this is a very sparse contribution, much less precise than the numerous messages for which it is used.  Yet a sparse contribution is all that is required if speakers in fact rely largely on inference to derive messages. Moreover, this contribution seems to apply to all uses of with.  If, for instance:

 Henry had a fight with his bride,


we don’t need to set up an “adversarial” with, distinct from the others, since that element can be attributed to the word ‘fight’.  It takes two to fight; and if the bride is the ‘attendant circumstance’ of Henry’s fighting, it can easily be inferred that she is his adversary.  Thus, if we recognize that people put information together and jump to appropriate conclusions, we don’t need multiple with’s, neither homonyms nor polysemes nor prototype and network.

                                                      

The Absence of Autonomous Syntax from Columbia School Linguistics

by Ricardo Otheguy & Betsy Rodríguez-Bachiller 

One of the most intriguing ideas of the late William Diver was his deep skepticism regarding the existence of an autonomous syntactic component. Diver's position, stated often in his lectures and sketched out in some of his publications, has been echoed in the published writings of many of his followers (Diver 1977, 1982, Huffman 1997:188, 341, Reid 1991, and, for a summary, Contini-Morava 1995). 

The Diverian position holds that while it is certainly true that one can observe syntactic patterns, such as words being grouped in phrases or constituents, or constituents appearing in certain orders or displaying certain co-occurrence restrictions, these observations do not justify the postulation of autonomous syntactic constructs in the underlying grammar of the language. The Diverian grammar consists only of signs, that is, constructs made up of lexical or grammatical meanings and the signals that express them. It is true that analytical experience has shown that the signal side of a sign is in some cases a syntactic positioning of words or constituents.

But aside from grammatical meanings signaled by positional signals, the grammar envisioned by Columbia School analysts does not have any syntactic content; there are no parameters and no settings, no phrase markers, no rules of order, no phrase-structure trees delineating initial structures, no movement rules, no constructions, and no blocking or filtering mechanisms that interact to specify the allowable order of constituents.

In Diver’s view, the incorrect belief that such constructs are part of the grammar of language derives from the a-priori nomenclaturist assumption that (a) the basic concepts proposed by traditional grammar to study the Sentence and its parts are essentially correct and, more importantly, that (b) the data for the study of language consist of complementary sets of grammatical and ungrammatical Sentences, and of relations within and between Sentences, all already conceived of, prior to analysis, through the theoretical categories of traditional, sentence-based grammar.[i] In the absence of these unwarranted assumptions, so the argument goes, no justification remains for the postulation of a syntactic component, as most of the 'facts' that would motivate it become epiphenomenal. Diver wanted to admit as data for linguistic science only those facts that would unavoidably force themselves on observers unencumbered by the a-priori, nomenclaturist categories of the tradition.

But how to account for those compelling syntactic facts that do meet Diver's requirement? For surely, even under Diver's uniquely austere definition of the data of linguistics, there remain syntactic patterns of order and co-occurrence that need to be explained.  In simplest terms, Columbia School linguistics posits no syntactic component because it believes that other, non-syntactic constructs—linguistic and non-linguistic—provide a better account of the syntactic facts. In general terms, Columbia School scholars explain syntactic order by appealing to a linguistic factor, namely the language-specific meanings of positional signals, and to a psycholinguistic factor, namely language-independent principles of discourse organization and processing.

Some facts of order can be explained through the meanings of positional signals alone, others solely through psycholinguistic principles, while still others, perhaps the majority, through the interaction of positional signals and processing principles[ii]. But in any case, once a successful analysis has produced robust hypotheses on the meanings of positional signals and the substance of processing principles, and provided that what Diver regarded as pseudo-facts are set aside, the need for an autonomous syntactic component disappears.