Chronology of Generative Grammar... complicated theory in a simplified way..

The theory of language structures first proposed in Chomsky's Syntactic Structures (1957). Just as physics studies the forms of physically possible processes, so linguistics should study the form of possible human languages. This would define the limits of language by delimiting the kinds of processes that can occur in language from those that cannot. The result would be a universal grammar, from which individual languages would derive as, in effect, different ways of doing the same thing. Variations such as vocabulary and principles governing word order would be revealed as different applications of the same underlying rules. Tacit knowledge of this universal grammar is pre-programmed, an innate biological endowment of normal human infants. The argument with which Chomsky supported the claim for such an endowment is known as the argument from the ‘poverty of stimulus’: it is argued that language-learning proceeds so fast in response to such a relatively slender body of ‘data’ that the infant must be credited with an innate propensity to follow the grammar of everybody else. The extent to which this argument treats the infant as a theorist or ‘little linguist’ has been much debated (see also language of thought hypothesis). In Chomsky's original model, language consists of phrase structure rules and transformations. 

Phrase structure rules represent the grammatically basic constituent parts of the sentence (e.g. a sentence might be a noun phrase + a verb phrase). Transformation rules change relations (as in the active/passive transformation) and determine how complex sentences may be formed from more simple ones. This latter function became taken over by the phrase structure rules in the later work (Aspects of the Theory of Syntax, 1965) that introduced what became known as the Standard Theory. In this, phrase structure rules perform the task of defining the deep structure of a sentence, from which its surface structure is thought of as derived by means of possible repeated transformations. Deep structure bears some affinity to the idea of the logical structure of a sentence, thought of as the representation of the sentence that reveals its inferential properties. The notion did not survive in Generative Semantics, one of the successors to Chomsky's Standard Theory, in which transformation rules map semantic representations onto surface structures. The introduction of semantics is often thought to be a necessary amendment to the purely syntactic and grammatical approach of Chomsky's early theory.

After the standard theory came the Extended Standard Theory, and eventually Government-binding Theory, both of which maintain an abstract and mathematical approach to the discovery of linguistic principles of the highest generality. Philosophically most interest has centered on the claim that complex grammatical principles might be innate, and on the relationship between syntax and semantics that is presupposed in the idea of a generative grammar. In general, philosophical formalists have been more interested in the possibility of unraveling concealed semantic structure, rather than in the more purely grammatical problems of linguistics.

Historical development of models of transformational grammar:

• Standard Theory (1957-1965)
• Extended Standard Theory (1965-1973)
• Revised Extended Standard Theory (1973-1976)
• Relational grammar ( 1975-1990)
• Government and binding/Principles and parameters theory (1981-1990)
• Minimalist Program (1990-present)

Frameworks

There are a number of different approaches to generative grammar. Common to all is the effort to come up with a set of rules or principles that will account for the well-formed expressions of a natural language. The term generative grammar has been associated with at least the following schools of linguistics:

Transformational grammar (TG):

• Standard Theory (ST)
• Extended Standard Theory (EST)
• Revised Extended Standard Theory (REST)
• Principles and Parameters Theory (P&P)
• Government and Binding Theory (GB)
• Minimalist Program (MP)

Monostratal (or non-transformational) grammars:

• Relational Grammar (RG)
• Lexical-Functional Grammar (LFG)
• Generalized Phrase Structure Grammar (GPSG)
• Head-Driven Phrase Structure Grammar (HPSG)
• Categorial Grammar
• Tree-Adjoining Grammar

Useful Semantic Network..

ConceptNet: A Practical Commonsense Reasoning Tool-kit

Reasons why I do not care grammar formalism...

http://www.aclweb.org/anthology/C/C88/C88-2095.pdf

Today's Passing Thought.....

Programming today is a race between software engineers striving to build bigger and better idiot-proof programs, and the Universe trying to produce bigger and better idiots. So far, the Universe is winning.

Concluding remarks on non evidence of syntactic rules in non-human communications...

Though this topic is quite interesting, can not devote more time on it. I have been given another challenge to propose some implementable algorithm on Ontology which can be integrated to the existing retrieval engine and also it should show marked improvement in results obtained. Have to start reading about it. So here goes my observations on the above mentioned topic:

Though artificial chimp signaling systems have some analogies to human language (e.g., use in communication, combinations of more basic signals), it seems unlikely that they are homologous. Chimpanzees require massive regimented teaching sequences contrived by humans to acquire quite rudimentary abilities, mostly limited to a small number of signs, strung together in repetitive, quasi-random sequences, used with the intent of requesting food or tickling. This contrasts sharply with human children, who pick up thousands of words spontaneously, combine them in structured sequences where every word has a determinate role, respect the word order of the adult language, and use sentences for a variety of purposes such as commenting on interesting objects. Though we can find some rough parallel in animals, as for instance, in perceptual abilities, it is obvious that human not only segment speech, but are also capable of continuously map the segments onto an enormous lexicon of minimally contrasting yet semantically distinct words. In all the cases of language-trained apes, they might act as if they have imbued some syntactic properties or recursive abilities, but there is no test to this hypothesis. Same is the case with CHF’s (2002) hypothesis that FLN (Faculty of Language in Narrow sense) may be unique to human. Although they have argued that most if not all of FLB (Faculty of Language in Broad sense) is shared with other species, what might be crucial is the degree of capacity. As for example, memory is essential to recursion. A computer program with a badly defined recursive procedure comes to a standstill, because it runs out of memory to store a record for the infinite number of loops. Likewise, the role of working memory aided by two subsidiary systems of phonological loop and visuospatial sketchpad is important in language. Future investigations of the linguistic capacities of language-trained animals should search for such other crucial attributes of human syntax, including the use memory, attributing mental status to others, reflecting the knowledge the signaler (primates) intends its audience to acquire rather than the knowledge it has and so on.

Why do non-human primates lack Syntactic (Recursive) Rules?

Finally back to some serious work!! Since last week, was trying my writing skills to build 'tareefon ke pul' for the boss & dignitaries and giving welcome speeches.... being just linguist is not enough in this place, I guess!!! But I was right when I said--- "Dr. D  presents a challenge that calls forth the best in people and brings them together around a shared sense of purpose."

So..... some sounds produced by monkeys are functionally semantic. There is also some evidence of primitive form of syntax in some nonhuman primates. But there appears to be no syntactic structure to the calls or vocalizations of these primates. Instead, sequences of calls tend to consist of either the same call repeated a number of times or of the pairing of two calls typically associated with different emotional states to express an intermediate state. There is also no evidence that an alarm call can be modified to elaborate upon the characteristics of the predator currently in question. Through repetition and changes in amplitude (both of which seem to carry prosodic information), alarm calls can serve to inform others of the immediacy of danger. They cannot, however, specify whether the predator is big or small, sleeping or stalking, in a tree or on the ground. Although the alarm calls can be equated with the first word stage of very young children, by the time the child attains age one, the linguistic capacities of the child surpasses that of the vervets. Children at this stage clearly use grammatical structures to infer the meaning of words (Pinker, 1984), and they will respond differently to a grammatically well-formed command than to an ill-formed one.

Cheney & Seyeafarth (in Why animals don’t have language) has given a possible explanation in the form of non-human primates’ inability to recognize argument structure. They fail to see that an event can be described as a linear sequence in which an agent performs an action. In any case, even if monkeys and apes do mentally tag events with syntactic properties (who does what to whom); they certainly fail to map these tags onto a communicative system in any stable or predictable way. However, some primates such as captive apes, ‘kanzi’ and some dolphins, after much training were able to produce phrases that differ according to agent, action or modifier. Although language-trained animals may be relatively proficient in the comprehension of phrases, it is doubtful whether even language-trained apes can produce phrases with any consistent syntactic structure.


In a recent paper Fitch & Hauser (2004) argued that non human primates can master the finite state grammar (FSG) which is at the lowest level of complexity. However, according to Chomsky, human language use requires the mastery of the next level in the complexity hierarchy, termed the “phrase structure grammar” (PSG). “In addition to concatenating items like a FSG, a PSG can embed strings within other strings, thus creating complex hierarchical structures and long-distance dependencies”. The aim of the F&H paper was to show that although the abilities to master a PSG are available to all normal humans, they are not available to monkeys.

It is a well-established fact human language exhibit hierarchic, generative and recursive nature.

Hierarchy of language:

Sentence   ←        the umpires talked to the players

Phrase      ←        the umpires                    talked to the players

Word       ←        the       umpires              talked         to     the    players

Morpheme ←      the    umpire     -s           talk   -ed    to     the    play-er-s

Phoneme    ←      D@    Vmpayr   z       tOk     t    to   D@    pleI   @r  z

The term recursion needs to be distinguished from iteration. An iterative procedure makes something to repeat over and over again. A recursive procedure, on the other hand, is a procedure, which is defined in terms of itself. Language, in principle can be characterized by a set of recursive rules (embedding sentences within sentences). A rule system which is recursive can not be accounted for by a finite state automation, because the latter can only produce regular grammar. Finite state grammar, although is able to capture left and right embedded recursive structures, can not represent center embedded recursive structures. Examples of center embedded recursion are:

1. The boy the girl saw fell.
2. The boy the girl the cat bit saw fell.
3. The boy the girl the cat the dog chased bit saw fell.

In these sentences one relative is nested within another. In their paper H & F (2004) aimed to show that cotton top tamarin monkeys are unable to process such kind of long distance dependencies.

In their experimental demonstration, F&H used a particular PSG, termed An Bn. This grammar generates center-embedded constructions. The A and B elements were drawn within separate sets of eight CV syllables, and were further distinguished by their acoustic characteristics. The A syllables were spoken by a female and the B syllables by a male, so that the two classes of syllables differed by voice pitch, quality, and other particularities of the voice sources. Participants (undergraduate students on the one hand, and cotton-top tamarins on the other) first heard a set of sentences following the patterns AABB or AAABBB. In the subsequent test phase, they heard novel sentences, half following the same grammar (A_n B_n) and half following a finite-state grammar (AB_n), which generated either ABAB or ABABAB sentences. Students were asked to state whether the pattern of each novel sound was the same as or different from the pattern heard during the familiarization phase. They scored 85% correct on this discrimination task. The performance of the Tamarins was assessed through their visual orientation towards the loudspeaker, an increase in looking rate being taken as indicative that the sounds were perceived as different. Interestingly, Tamarins displayed an equivalent rate of looking to strings that violated the rules of the grammar and to strings that were consistent with those rules. Obviously, their failure to selectively look at the non-consistent strings could be due to multiple causes, such as a perceptual inability to discriminate the acoustic properties of A and B syllables. In order to eliminate alternative interpretations, F&H inverted the two grammars for other groups of subjects. It turned out that Tamarins trained with ABAB or ABABAB sequences showed a significant increase in looking to the AABB or AAABBB strings when they were displayed during the test phase. Because this inverted task involved the same perceptual abilities as the first one, F&H inferred that the failure of tamarins trained with the PSG provided a demonstration of their inability to master this class of grammar.

Perruchet & Rey (Does the Mastery of Center-embedded Structures distinguish humans from nonhuman primates?) however, argues that F & H’s conclusions are not compelling. Their first argument is that since it is not possible to demonstrate that the achievement in a specific task t implies the mastery of a grammar g, then how could it be possible to conclude that the failure in t attests that g cannot be mastered?  The implication is that “the failure of tamarins in the F&H task is relevant with regard to their alleged inability to master a PSG only if it can be asserted with a reasonable confidence that the achievement of humans in the same task attests for their mastery of the PSG” (Learning Hierarchical Structure 1).

Secondly, the center-embedded structures, which F&H designated as their main target throughout their paper, it is doubtful that such structures exist in natural language (even if they exist, they are not manageable whenever the number of embeddings exceeds one), and in fact the instance provided by Chomsky (if…then) is better described as a genuine center-embedded construction (Christiansen & Chater, 1999).

Some pics from 4th Anniversary of Xobdo....

feels ecstatic to be a part of something to promote my language & my state... wish I could be present there, but it gives me an immense pleasure that my father received the appreciation certificate given to me on my behalf...thanks, Deuta..

Papa, the back bencher... :)

http://www.enajori.com : the first monthly bi-lingual (Assamese & English) online magazine..

Feeling very happy to post this news that we (some Axomiya students & professionals across the globe) are going to introduce our own golden era of music, cinema and literature to our current generation through the world of internet along with our rich history, culture, wildlife and tourism. Endeavouring on a greater attempt to showcasing our own Assam, we are going to launch the first monthly bi-lingual (Assamese & English) online magazine, http://www.enajori.com on the 13th of March, 2010. The vision of this website is to create a single platform wherein the entire gamut of Assam, in its entire cultural and artistic splendour, is encapsulated; a humble attempt to put Assam’s rich culture and heritage on the world map so as to enlighten the people of the world about the state of Assam and her people.  This monthly bi-lingual e-magazine will serve to strike a chord with the non-resident Assamese people who are scattered around the world as also those living within. And hopefully, it will also enlighten about our culture and our very existence as 'INDIAN' to the rest of the Indians who believe that West Bengal is the eastern limit of India and beyond that there exists a huge jungle inhabited by some uncivilized, uncultured people with no sense of so called traditional valuation like them.

The venue and the time are mentioned below:

Venue: Guwahati Press Club, Guwahati

Date: 13th of March, 2010 (Saturday)

Time: 3pm

Inauguration by: Nirupama Borgohain

(Non) evidence of Syntactic (Recursive) Rules in non-human Primates...

Though this is not my usual topic of interest, but few readings have really captured my linguistic bent of mind to read about it and post my views. So for the time being, deviating from regular Minimalist Sytax.....

Continuing in the line of discussion proposed by Hauser, Chomsky and Fitch that Syntactic Recursion is the only unique aspect of human language differentiating them from non-human primates, lets look at some evidences and experiments from literature related to the syntactic rule learning ability of the non human primates.  Zuberbuhlar’s (2004) study shows that vocalizations and differences in alarm calls by monkeys can be treated as semantic signals in the sense that monkeys compares signals according to their meanings and not just their acoustic properties. To date, there has been no evidence that they go beyond this simple semanticity. Whatever syntactic evidence found in animal communication, is phonological syntax, referring to rules that specify the assembly of smaller vocal units into larger ones. In phonological syntax, the units, like the letters in a written word, have no independent meaning. In lexical syntax, the units, such as the words in an English sentence, have meanings which contribute to the overall meaning of the whole signal. Many bird species can learn songs with phonological syntax. Oscine birds, which learn complex songs, are very distant relatives of humans. Many other birds, and more closely related species, including most mammals, do not produce calls composed of independent subunits. Our closest relatives, the apes, do produce long calls composed of subunits.


Efficient and wide spread adoption of new words would seem to require that both speaker and hearer attribute intention and belief to one another. It seems possible, then, that some sort of rudimentary theory of mind might be necessary for the learning of words and language. If true, this hypothesis might explain the lack of vocal learning by monkeys, because all evidence to date suggests that monkeys cannot attribute mental states to others. In contrast, word learning in even very young children seems to be accompanied by primitive mental state attribution. Clearly, young children of one and two years of age do not have a fully developed theory of mind, in the sense that they attribute false beliefs to others. By the age of one year, however, they already seem to understand that words can be mapped onto objects and actions in the world. Crucially, this understanding seems to be accompanied by a form of “social referencing,” in which the child uses other people’s direction of gaze, gestures, and emotions to appraise a situation. Infants by the age of 18 months also actively attend to the speaker’s gaze and focus of attention when inferring the referent of the speaker’s utterance, as if they have developed some tacit understanding that gaze and attention are a reflection of underlying knowledge.


No doubt, we need to look at more literature & studies to dig further into this topic, still even on the basis of this limited reading, I feel that at least some species of nonhuman primates possess in their natural communicative repertoire a small number of calls that serve as semantic labels for objects. Nonetheless, these same animals never seem to create new calls or labels for objects. Why should an animal that already possesses a small number of semantic signals in its vocal repertoire be unable to create new labels for other objects and events in its environment? Why is there so little evidence for learning and modification in the natural calls of nonhuman primates and other mammals? Finally, if animals can be taught to obey sentence like commands by humans, why do they not also spontaneously produce sentences, and why is there no evidence for syntax in the natural communication of animals? Studies have reported that syntactic abilities are conducted under human designed artificial communicative system, but so little is known about their natural communicative abilities unaffected by human influence. Moreover, though it is fair to call such abilities in apes ‘syntactic’, they are still far removed from the human ability to organize sequences of words into complex hierarchically organized sentences. Little is known about the ability of apes to learn hierarchically structured behaviors, although all researchers seem to expect apes to be less proficient at it than humans.

What defines language and Language Faculty? based on Hauser, Chomsky & W Tecumseh, 2002

In continuation of my earlier post (dated 11 Nov, 2009) Is possessing a language a quintessentially human trait?


The boundary line between Human and non-human primates is characterized by language. Janes R Hurford in his article ‘Human Uniqueness, Learned Symbols and Recursive Thoughts’ says that equating language with communication is counterproductive. There are two features of human language that no other animal communication system involves: learned arbitrary symbols and recursive, semantically composed syntax. All words of any human language are symbolic. There is an arbitrary relationship between the signified concept and the signifier form. Human language also has semantically composed syntax. Words are put in different sequences to convey different messages and the messages are composed from the meanings of the words. However, some degree of syntactic patterning is also found in other primates in their communicative calls. But how they manipulate extremely long and complex signals is questionable.

Distinguishing the Shared and Unique Components of the Language Faculty (FLB & FLN)

In a recent research program, Hauser, Chomsky and Fitch are of the opinion that Language faculty can be divided into a broad (FLB) and narrow sense (FLN). FLB includes two components- sensory motor and conceptual intentional. FLN is the abstract linguistic computational system, independent of other systems with which it interacts and interfaces. The internal architecture of FLN is a computational system that generates infinite number of new sentences from a finite set of rules. A core property of FLN is recursion, often termed as discrete infinity. “This recursive mechanism entails a procedure that calls itself, specially combining the discrete store of words and sentence constituents into hierarchical phrase structures that can be further embedded with other hierarchically arranged phrase structures” (Tincoff & Hauser, to appear). Each expression of this mechanism is a pairing of sound and meaning and free from the restrictions, for instance, working memory, lung capacities etc. that often limits the FLB aspects. 

Comparative Methods

To discuss how these various components of language faculty have evolved, evolutionary biologists have suggested three methods:

1. Homologies: Traits which are not uniquely human evolved for an earlier function directly inherited from a common ancestor.
2. Homoplasies: A trait not uniquely human, shared by other species, but not present in their common ancestors, evolved for other purposes and was redesigned for language. FLB is derived, uniquely human adaptation for language.
3. A trait which is uniquely human, is a phylogenetically novel trait evolved only in human.

Evidence from homologous mechanism:

FLB is strictly homologous to animal communication. FLB is composed of the same functional elements that underlie communication in other species. Sensory motor system includes abilities used in vocal production, categorical perception and imitation which are also shared by other primates. On the side of vocal production, birds and non human primates naturally produce and perceive formants in their own species typical vocalization. Rich multimodal imitative capacities are seen in dolphins and birds, with most song birds exhibiting a well developed vocal imitative capacity. In a recent study, cotton tamarins have shown a perceptual mechanism for discriminating the rhythmic classes of spoken language. Conceptual intentional system includes conceptual representations (number, color spatial referents etc.). These abilities also to some extent shared by non human primates, only that they can not express it. What is strikingly different is the ability to form a lexicon. Even if they have it, the lexicon is small and no evidence is there for the capacity to generate novel strings. This is where, FLN comes, which is the only unique trait of human language.

FLN as defined by Hauser is especially the mechanism of recursion. Fitch & Hauser (2004) provide the most direct test of the FLN on the perceptual studies of some cotton top tamarins. The result shows that monkeys share with human the basic computational mechanism for finite state grammar. However, they fail to extract phrase structure grammar. In the experiment, tamarins were familiarized with a set of syllables following An Bn grammar. When they were tested with new strings, half violating the rule, half consistent with it, the tamarins failed to recognize the violations from the grammar consistent ones. This result shows that non human primates can not master hierarchically embedded structures.

Diachronic Dimension in Explaining Language Universals

Diachronic considerations to the study of language universals would be to look for universals in language change. This dimension consists of answering the question: how do languages evolve into the variation states to which Implicational universals refer? More recently, researchers have focused on the diachronic aspect of language universals instead of being satisfied with the formulation of the general principles offering a summary of cross-linguistic patterns, without suggesting how they might come into existence. In fact, explaining language universals in this line of research was suggested by Greenberg long back. Greenberg, as cited in Bybee (1989) made this point in outlining a new program for general linguistics, one that could deal with all statements pertaining to all languages, by saying that whether a particular typological pattern was frequent or rare was ‘the resultant of two factors, one of origin and the other of survival’ (p. 89). His statements emphasizes that we must look to the diachronic dimension to learn how the conventions of grammar arise if we are to know why they take the particular form that they do.

Bybee explains this phenomenon in the light of word order universals (correlation in the order of constituents in SOV and VSO languages) to which Hawkins (1979) offers an explanation in the form of his Cross Category Harmony Principles and Venneman (1972) as an innate disposition in speakers to grasp the operator-operand relation. Nevertheless, this mode of explanation is not complete unless we answer the ‘how’ question. What are the causal mechanisms that bring generalizations in language? The kind of Implicational universals of the form ‘if P, then Q’ follows logically but it presents a prediction, not necessarily an explanation. For instance, a survey was conducted by the author of this post to figure out the typological properties of classifiers in South Asian languages and arrived at an implicational generalization that-

If a language has large class of numeral system, then it definitely has numeral classifiers.

While making this kind of generalizations we must explain the relation between P and Q and how the development of one influences the other. Since it was also observed in the data that languages which have few numerals, restricted to one or two (or perhaps three), borrow higher numbers from neighboring languages (also relies on some form of deictic indication for higher number or else borrow from neighboring languages) and thereby borrow the classificatory system too.

Bybee (1989) has identified certain causal factors, to qualify as explanations and as contributors to the creation of the particular grammatical phenomenon. They are explanations based on processing ease, on iconicity, on cognitive or semantic factors or on typical discourse structure. “In order to attain valid explanation for language universals it is necessary to attend the causal mechanisms operational in the establishment of grammatical conventions and find the general dynamic principles behind the causal mechanisms. This view requires that explanation have both diachronic and synchronic dimension. The diachronic dimension plots the paths of development of grammatical conventions across time, and synchronic dimension fills in the small steps along these paths by referring to the way in which language users manipulate the linguistic conventions that they have inherited for their conceptual and communicative purposes” (p. 375).

Few Concluding Remarks:

From the above discussion, we can say that linguists who search for universals in language will generally find what they are looking for. The goal of language universal studies is to define what a possible language is. Both the approaches discussed so far (Generative vs. Typologists) try to answer this question and both believe that there are universal constraints that define the answer to this question. But after reviewing both the approaches and the ongoing debates between the two, one genuine question that arises, is- what conclusions can be drawn from this? Does universality imply innateness? Or, implicational universals limiting the variations offer a description of logically possible languages. Here is a list of some conceivable ways of explaining language universals without innate grammar:

• Many similarities between languages are adequately explained by their having a common origin. It appears quite likely that all human languages have a common origin, if one goes far enough back in time. One would have to assume that several distinct groups of proto-humans independently developed language. Now, this is certainly possible, but the evolution of a single innate universal grammar, common for all mankind, actually requires that all languages have a common origin, in order to be compatible with standard Darwinism.
• In order to be a useful instrument for communication, a language has to meet certain basic criteria. Is it possible that some principles of Universal Grammar can be explained by their being, logically or pragmatically, necessary features of a language?
• Can the rules of Universal Grammar be derived from our general (non-linguistic) cognitive structures, along the lines of the Cognitive models?

While coming to Typological studies, in general, their claim to study a wide range of languages remains valid and constant, as we can see that recent work in Generative grammar has indeed broadened the range of languages considered (although extremely restrictive). Comrie (1984) in his reply to Coopman’s review of his Language Universals and Linguistic Typology has mentioned- “Chomsky (1981) makes clear that generative grammarians have come to realize that an adequate study of syntax within universal grammar requires the study of languages of different types (in the current terminology: with different parameter value)” (p. 155). However, keeping apart the theoretical and methodological distinctions between the two approaches, and emphasizing more on Typological studies, we find a number of problems with investigating all languages:

• Many languages have become extinct.
• Given language change, many new languages will arise. (however, the diachronic dimension in explanation for language universal have thrown much light on this area)
• There is further problem of choosing a representative sample language which should be free from some biases:

- Languages chosen must be from a range of genetic language family.

- In establishing a sample language, one should ensure that each group of language should be given equal    representation.

- Areal distribution of language spoken in a same geographical are tend to influence one another through borrowing or shared innovation, which are not necessarily language universals.

Based on the available evidence, it seems very hard to draw any firm conclusions in this field. Several theories exist which are compatible with what is known about language acquisition and universals, none of which is clearly preferable. It appears likely that no single theory contains the whole truth, but that many theories have some grains can be of it. However, one new line of inquiry can be conjectured based on the recent arguments by Hauser, Chomsky and Fitch (2002) that the uniquely human and uniquely linguistic aspect of human language is syntactic recursion. Recursive property is the only crucial aspect of the computational mechanism of human language that accounts for the rich, expressive and open-ended power of human language. Of course, this view should not be taken as the only possible explanation for universals rather as a new entrant to this program of research, always subjected to testability and falsifiability.

Typological Generalizations in Language Universals

(With reference & continuation to my post Language Universals and Chomskyan Account dated 5th Feb, 2010)

Traditionally, typology was used as an alternative method for pursuing the same goals as Generative grammar: to determine the limits of possible human languages. The works of Joseph Greenberg has been particularly instructive and influential in this regard. Greenberg (1963) represented universals in logical form, namely as Implicational universals and biconditional universals (a language has X if and only if it also has Y). He constructed an areally and genetically diverse sample of 30 languages in order to infer empirically valid universals arguing that grammatical categories must be compared across languages as on an ultimately external semantic basis. He then applied this method to word order and morphological categories, constructing a total number of 45 universals. According to him, search for universals must focus on the distribution of types found in languages, such as the preference of suffixing over prefixing, and the correlation between different typological features across languages, such as the correlation between preposition-noun order and genitive-noun order. That means significant universals are to be found in the constraints on cross-linguistic variation, not on cross-linguistic uniformity. His 1963 paper established the basic methodology of what is known now as the Typological approach to grammar deriving major empirical results and offers an explanation used widely today in Typological analyses. At the same time, Chomsky was also arguing that linguists should seek for language universals, but via deductive reasoning through the study of individual language. Chomsky and Greenberg at the same time proposed opposing theories about universals of grammar (particularly syntax), how they are to be discovered, defined and how they are to be explained. These later came known as Chomskyan and Greenbergian approach to language universals and later characterized more broadly as the Formalist and the Functionalist approaches to language.

Ontologically, Greenberg’s universals are very different from the principles on which Chomskyan universals focuses.

• Under Greenbergian framework, speakers need not know about the universals, either implicitly or explicitly. Greenberg’s universals are not part of I-language.
• Typologists tend to assume a functionalist framework for language description. Consequently, the universals they state are normally couched in much more concrete terms than the abstract UG principles which generativists seek to discover.
• Just as Greenbergian, universals are ontologically very different from UG principles, so the research methods of typologists are quite different to those of generative linguists. We saw earlier that generative linguists deduce the existence of particular UG principles from the facts of language acquisition via a poverty-of-the-stimulus argument (or, at least, that is what they claim to do). Greenbergian universals, in contrast, are arrived at inductively, by comparing data from large numbers of languages. Inductive generalizations, however, are only as good as the data on which they are based. For this reason, typologists do not study random groups of languages, but use special techniques to construct representative language samples. In particular, the Typologist's sampling techniques are designed to avoid genetic and aerial bias: if a sample contains too many languages from the same family or the same geographical area, then it may not provide a reliable picture of cross-linguistic variation.

More about Greenbergian Universals
Greenberg’s universals are known as Implicational Universals (universals relate the presence of one property to the presence of some other property). Logically, this universal can be symbolized as- if P, then Q, which results in four possibilities:

a)      both P and Q

b)      P and not-Q

c)      Not-P and Q

d)     Not-P and not-Q

E.g., if a language has VSO as basic word order, then it has prepositions. Welsh has both VSO and prepositions (a). English has prepositions but not VSO (c). Japanese- no VSO, no prepositions (d). However, possibility number (b) no prepositions, but VSO is not attested. In formulating implicational universals, it is always important that the rigid interpretations of material implication be followed, and in particular to note that a given implicational universal always allows three of the logical possibilities while disallowing one, that counts as a counter example to an implicational universal. For certain properties of language, it seems that we can state whether they found in natural language without reference to any other properties of the given language. For instance, ‘all languages have vowels’ makes absolutely no reference to any other item that must or must not be present. These are called Non-implicational universals.

Greenberg’s list also produced another parameter along which universals may be classified, namely distinction between absolute universals and universal tendencies, to which there are exceptions. An example of a tendency is the following (from Greenberg 1963):

"With overwhelmingly more than chance frequency, languages with dominant order VSO have the adjective after the noun. The existence of such tendencies makes perfect sense within the functionalist framework. It is asserted that the limits to cross-linguistic variation emerge from the requirements of language use, but functional pressures may take a while to make their effects felt through language change".

Linguists working in the Greenbergian tradition allow for variation in the explanation of the existence of the universals, which they isolate. It is considered possible that some universals may require one type of explanation, while others may require explanation of another kind. For instance, some universals, such as the fact that all languages have at least three persons and two numbers may be explained from the point of view of discourse pragmatics. This Functional and Pragmatic approach to explanation of language universals argue that certain universals serve to make language more functional, either as a communication system or as more, particularly relative to the communicative needs of humans. Comrie (1989, p.28) proposes a similar explanation of the fact the existence of first or second person reflexive forms in a language implies the existence of third-person reflexive.

Some language universals may be explained by some reference to the Processing demands placed on language users. This explanation appeal to two major considerations: human beings are limited capacity machines- there memory constraints and the relative ease or difficulty involved in processing certain structures in comprehension and production.

To some extent, for certain properties language universals are explainable in terms of human perceptual and cognitive apparatus (Hawkins, 1988 p.17). If a language has Color system, it will distinguish at least black and white. If it has three colors, the third will be red, if it has four, then the fourth will be either green or yellow. Hawkins (1988) cites Kay and McDaniel to point out that this universal can be explained by reference to the neural anatomy of the color vision of the humans.

Core Concept: Morpho-syntactic Features

It is not the actual word that is relevant here. What seem to be relevant are rather properties of words.

The pigs grunt

*The pig grunt

Agreement relation among words gets reflected in morphological forms of words. A morphosyntactic feature is a property of words that the syntax is sensitive to and which may determine the particular shape that a word has. Features seem to be the core elements of languages that relate sound and meaning.

The sheep bleat

The sheep bleats

The plural feature clearly has an effect not just on the morphology of the word, but also on its meaning: in this case it affects whether we are talking about one child or more than one; one man or more than one, and so on. Features that have an effect on semantic interpretation in this way are called Interpretable features. We shall see that the notion of Interpretable features, and its opposite, Uninterpretable features, will play a significant role when we come to build up our theory of syntax.

In an intuitive sense, the features responsible for the morphological difference are also responsible for a semantic difference. Usually a plural noun is associated semantically with a group of entities in the world, rather than with a single entity. However, once again the link between syntax and semantics is not strict: the word scissors is plural in form, but refers to an entity which is usually conceived of as a single thing.

The most important set of features that is relevant to syntax is the set of Category features.These are the features that are responsible for separating words into the traditional word classes of noun, verb, adjective, preposition, plus a few others. These features certainly seem to have some semantic basis: nouns, by and large, tend to pick out objects in the world, while verbs tend to pick out events, or actions. Of course, as is well known, this semantic criterion is by no means foolproof. A noun like race, for example, seems to pick out something that is clearly an event, and it’s difficult to think of a verb like exist as denoting an event or action.

If we want to make the generalisation that whole classes of words (nouns vs. verbs) are restricted as to the features that may occur on them ([past] cannot occur on nouns). Moreover, at least some of the words in these classes show distinctive morphological endings that reflect the class they are in. Finally, the two classes broadly seem to have different interpretations. These considerations would be enough evidence to allow our language learner to posit that there are features in English that distinguish these classes. We therefore posit the feature [N] which is part of the featural specification of nouns and [V] which serves in the same capacity for verbs. A plural noun like men will then have the partial feature specification in:

men [N, plural]

In summary, then, we have four major word classes, which we usually abbreviate as just N, V, A and P and which we could distinguish using the four features [N], [V], [A] and [P]. Another term for these features is major category features. We can refer to [N] as the Nfeature of a word and [V] as the V-feature and so on. Note that since we have four categories, it is once again possible to define these by using two features. Here is one possibility:

a.  noun [N]

b.  verb [V]

c.  adjective [N, V]

d.  preposition [ ]

This feature system also has the curious result that prepositions do not have category features. This is almost certainly not correct. An alternative, here, would be to adopt the binary system, giving the following feature specifications:

a.  noun [+N, –V]

b.  verb [–N, +V]

c.  adjective [+N, +V]

d.  preposition [–N, –V]

So, we have met morphological features and major category features. There are also semanticfeatures. In addition, words have a pronunciation, which, if we are to be consistent, must be specified featurally as well. The features responsible for a word's pronunciation are termed phonological features, so a lexical item turns out to be a collection of phonological features, semantic features and morphosyntactic features. It also appears to be the case that phonological features are universal in the sense of their being a universal set, some portion of which is selected in particular languages. Each language, then, specifies what its lexicon (i.e. its set of lexical items) is, constructing its lexical items from putatively universal featural domains. The child learning a language is faced with the task of internalising this lexicon on the basis of the evidence that he/she hears and sees.

In general, then, we assume that syntactic relations like agreement access syntactic features, and not phonological or purely semantic features. Some syntactic features have a transparent effect on interpretation (so, by and large, nouns with a [plural] feature in English are interpreted as referring to more than one entity), but some do not. This idea, that syntactic relations hold between purely syntactic features is known as the Autonomy of Syntax Hypothesis.

The morphosyntactic feature does not always have a predictable semantic interpretation, something which reinforces the idea that the semantic interpretation of a feature is not accessible to syntactic processes. These types of features, person, number and gender, go under the general name of Phifeatures (often written f-features). F-features appear to be interpretable, and are motivated by both semantic and morphological facts. The agreement relation we saw above ensures that some subset of the f-features on subjects, agrees with those on verbs.