Back to basics

This chapter looks at the type of grammar Chomsky finally decided was needed: a principles and parameters model. It outlines the basic principles of such a grammar.

‘There are three things in life you must never run after: a bus, a woman and a theory of transformational grammar. There will be another one along in a moment,’ commented one well-known linguist. Chomsky’s 1980s grammar – his so-called Government and Binding (GB) approach – was radically different from the Standard Theory, and his 1990s proposals – his Minimalist Program – were further away still. He even abandoned transformations!

This chapter presents a bird’s-eye view of his later ideas. It explains what the grammar was trying to do in general, and glosses over the intricate technical details. More information on these are presented in the works suggested for further reading at the end of the book.

[]() Universal Grammar (UG)

Chomsky became particularly concerned with the learnability problem. How did children manage to learn language so efficiently? They must, he assumed, be born equipped with Universal Grammar (UG), a basic outline knowledge of key language properties.

But if UG was inbuilt in the brain, why []()were languages so different from one another? UG, he argued, was only partially wired up. Children were born with an inbuilt knowledge of basic linguistic principles, but these needed supplementing. The inherited framework must be backed up with ‘parameters that have to be fixed by experience’. A parameter is an essential property with inbuilt variation. For example, temperature is a parameter of the atmosphere: temperature must always exist, but is set each day at different levels.

Perhaps youngsters were faced with an array of linguistic switches, he suggested, which had to be switched one way or another. They instinctively knew the basic options, but needed to find out which had been selected by the language they were learning.

Once they had discovered this, multiple repercussions would follow – just as if, say, animals had to opt for air or water as their basic environment, which would in turn bring about a number of inevitable consequences. A simple decision to choose one option rather than another at a particular point would have repercussions throughout the grammar. Relatively few decisions may need to be made, but they would have far-reaching effects.

Chomsky labelled this framework a Principles and Parameters (P and P) model (see ).

[]()

Figure 18.1.

The possible option points remained []()speculative, though some suggestions were made. For example, within phrases, one possible choice might be between having the head (main word) at the beginning of a phrase (as in English PP _up_ the tree) or at the end (as in Turkish, which says, as it were, the tree _up_). This would have far-reaching effects on the form of sentences. For example, an English sentence such as:

[]()The man who fell downstairs broke his leg.

would, in a language such as Turkish, turn out to have a literal translation, something like:

[]()The downstairs-fell man his leg broke.

Pro-drop might be another crucial option. There might be an important division between languages which allow their speakers to drop the pronouns at the beginning of a sentence, and those which do not. For example, Italian is a pro-drop language. It is possible to say either:


Sono Inglese.	or	Io sono Inglese.
*Am English		I am English

In English, only the second option, the one with the pronoun, is possible. The path taken would have repercussions throughout the grammar, some of them unexpected ones. For example, consider the English sentence:

[]()Angela thought that Hezekiah was stupid.

This has a corresponding question:

[]()Who did Angela think was stupid?

For some strange reason, the word that has to be omitted in the question. You cannot say:

[]()\*Who did Angela think that was stupid?

This odd fact seems to be characteristic of []()non-pro-drop languages. Pro-drop languages such as Italian seem able to leave that in.

In the long run, Chomskyan linguists hoped to specify all of the crucial option points, and their repercussions. These then would be built into the overall framework. If this was ever achieved, then linguists would have gone a long way towards achieving their ultimate goal of specifying UG.

[]() From deep structure to D-structure

Superficially, the most obvious difference between the 1980s TG and the 1960s Standard TG was the renaming of some of the essential ingredients, so as not to confuse them with the old ones. Deep structure in its altered form was relabelled D-structure, and the revised surface structure was relabelled S-structure. The old semantic representation was superseded by LF ‘logical form’, and the phonetic representation was labelled PF ‘phonetic form’ ().

[]()

Figure 18.2.

These levels were linked by processes which had certain superficial resemblances to the older-style TG: PS rules specified the D-structure. D-structure and S-structure were linked by transformational operations, though (as explained in the previous chapter) the transformational component was a mere shadow of its former self. PF rules converted the S-structure into PF, and LF rules converted the S-structure into LF ().

[]()

Figure 18.3.

However, not one of these levels was truly []()similar to the comparable level in an old-style TG, nor were the rules which operated on them the same. For example, the LF rules and LF contained a considerable amount of material which belonged strictly to the syntax in a Standard TG.

[]() Government and binding

The new-look 1980s grammar was presented by Chomsky in a series of lectures published under the title: Lectures on government and binding (1981). The name has stuck. It is usually referred to as government-binding theory, or GB. This somewhat strange label was because GB paid particular attention to two aspects of grammar which were virtually ignored in the Standard Theory, government on the one hand, and binding on the other. Government involved the notion of a constituent having power over others, and binding dealt with the linking or binding of items to one another. Let us briefly outline what each involved.

The general notion of government has been around in linguistics for a long time, in that it has long been recognized that some words have influence over, or govern other words. In Latin grammars, for example, grammarians spoke of prepositions ‘governing’ nouns, since a preposition such []()as contra ‘against’, as in ‘fight against the Gauls’, caused the following word, Gauls, to have a particular ending. In the more recent (somewhat altered) usage, the word ‘government’ was usually used in connection with heads of phrases which influenced others in their immediate locality, in the sense of requiring them to exist. For example, in English a verb such as hit governs a following NP, as in hit the donkey, and a preposition such as up also governs an NP, as in up the tree. The nodes involved are normally sisters, daughters of the same mother (Chapter 7).

[]()

Figure 18.4.

However, quite often there is an important relationship between words which are on different branches and on different levels, as in the sentence:

[]()Drusilla had a dream about herself ().

Drusilla and herself are on different branches, and different levels, yet clearly they have an intimate relationship, which needs to be carefully specified, since one could not randomly alter the sentence around. It is impossible to say, for example:

[]()\*Herself had a dream about Drusilla.
[]()\*Drusilla had a dream about Peter kissing herself.

Furthermore, it is important to understand the relationship in order to interpret sentences properly. Consider the sentences:

[]()Henry read the report about Toby stabbing himself.
[]()Henry read the report about Toby stabbing him.

It is essential to realize that Toby was stabbed in the first sentence, but Henry (or someone else) in the second.

[]()

Figure 18.5.

A major part of Chomsky’s GB theory was to try to specify exactly which parts of trees []()influenced one another, and which could be linked in their interpretation. He and his followers tried to draw up a wider notion of government, known as command. A principle known as c-command (from ‘constituent command’) specified which constituents had power over others in a total tree structure. There was some dispute about exactly how to phrase c-command, but in general (and somewhat simplified), it said that when a node branched, items on the first branch had some influence over those on the second branch, irrespective of how high or low on the tree they came. So, in , the first NP Drusilla c-commanded the VP, and every node under it. The main verb had c-commanded the NP following, and every node under it, and so on.

The notion of c-command enabled one to specify relationships and restrictions between different parts of the sentence. For example, one could say that the word herself in the sentence about Drusilla must be c-commanded by the person referred to. This precluded impossible sentences such as:

[]()\*Herself had a dream about Drusilla.

Furthermore, the notion could be used with other constructions, such as:

[]()The politicians argued with one another.

One could specify that a phrase such as one another must be c-commanded by the phrase it referred back to, so precluding:

[]()\*Each other argued with the politicians.

More generally, one could say that words []()which refer back to others, traditionally known as anaphors, have to be c-commanded by the words they refer back to, their antecedents. Therefore, a general structural relationship, that of c-command, enabled one to specify simply a large number of apparently separate restrictions, which had to be stated one by one in an old-style transformational grammar.

Binding was strongly interlinked with the notion of c-command. Briefly, a binding principle stated that when two NPs are coindexed – refer to the same thing or person, as with Marigold and herself in Marigold cut herself – then the antecedent (Marigold) must c-command the anaphor (herself). If so, it has been properly bound: in other words, there is a proper link between the two NPs. In contrast, a sequence such as:

[]()\*Herself cut Marigold.

would be impossible, because herself is not properly bound. Binding related above all to the interpretation of sentences, since it showed which items were to be interpreted as linked together.

All this may seem somewhat like common sense, given these simple examples, and it looks at first sight as if one could have just rephrased all this by saying antecedents come before anaphors. But the necessity of specifying a structural relationship between NPs becomes apparent as sentences get more complex, as in:

[]()Who did Marigold claim cut herself?

The surface structure of this sentence would be something like:

[]()Who did Marigold claim t cut herself.

We need fairly detailed mechanisms []()to specify the linking of herself and t, and t and who, so allowing one to interpret who and herself together, rather than Marigold and herself (t for ‘trace’ was explained on pp. 259–60).

To summarize so far, the government and binding approach was particularly concerned about relationships between constituents. It specified which constituents had power over others, the overall purpose being to express clearly which nodes on a tree were interlinked. Only certain links were possible, and discovering and describing these clarified the principles underlying sentence interpretation.

[]() Broadening the range

The Government and Binding (GB) model showed a shift of perspective. In the early days of transformational grammar, a grammar was essentially a device which specified what was, and what was not, a well-formed sentence. In the GB version, Chomsky had moved the emphasis to the general principles and relationships which exist within language.

The GB model was the first within the Principles and Parameters (P and P) framework (p. 268). It contained a number of different components, or modules, each of which played a role in the whole. For example, theta-theory or θ-theory – short for ‘thematic relations theory’ – dealt with who did what to whom by specifying the roles played by NPs, such as agent or theme (Chapter 7). The Empty Category Principle (ECP) specified how to deal with apparent gaps in the structures, as in:

[]()Aloysius wants – to go.

The person who Aloysius wants to go is not overtly specified, yet native speakers know that it is Aloysius.

But considerable arguments arose as to which []()modules did what, and conflicting proposals were made by different researchers. Chomsky himself proposed some amendments to the theory in his book Barriers (1986). But in his later work, The minimalist program (1995), he largely gave up on specifying the tasks allotted to each module.

[]() The bare bones

Chomsky tried to pare down his linguistic theory to the bare bones of language, which is why he referred to his 1995 work as a Minimalist Program. He was trying to find basic laws of nature, such as a linguistic equivalent of the law of gravity (Chapter 3).

The main feature retained was the switch-setting of the Principles and Parameters (P and P) approach. Two levels of structure were abolished: D-structure (the descendant of deep structure) and S-structure (the descendant of surface structure) no longer appeared.

In the bare-bones model, the lexicon fed into a ‘computational system’. This checked that the word combinations were in accord with basic linguistic principles. The lexicon fed into a ‘spell-out’ which specified the pronunciation. The end-point was meaning on the one hand, and pronunciation on the other ().

The linguistic principles which guided the system remained sketchy, but they were essentially principles of ‘economy’ or simplicity. The most straightforward was Shortest Move.

Consider the sentence:

[]()Angela has asked Henry to find her hat.

Suppose you wanted to query who Angela had asked and what she wanted found:

[]()Angela has asked who to find what?

[]()

Figure 18.6.

Normally, any word beginning []()with wh is brought to the front of a sentence. But in this case, only the wh-word which moves the shortest distance can come forward: it is possible to say:

[]()Who has Angela asked to find what?

But is impossible to say:

[]()\*What has Angela asked who to find?

This was the type of broad-ranging linguistic principle which Chomsky was hoping to identify, though, as he himself admitted, much remains to be done. ‘Current formulation of such ideas still leaves substantial gaps,’ he commented.

Chomsky’s model of language is not the only model being worked on, as was pointed out in Chapter 3. However, owing to the huge influence of his work, this model has been given prominence in this book. But where do we go from here? This will briefly be considered in the next section.

[]()[]()THINGS TO REMEMBER

[]()Chomsky became concerned about the learnability problem: how children manage to learn language so efficiently.
[]()He assumed children must be born equipped with Universal Grammar (UG).
[]()UG contained basic information about key language principles.
[]()The inherited framework needed to be supplemented by parameters that had to be fixed by experience.
[]()This was referred to as a Principles and Parameters (P and P) model.
[]()A parameter is an essential property with inbuilt variation.
[]()The new grammar became known as government-binding theory (GB).
[]()Government was concerned with elements which were required to exist by the heads of phrases.
[]()Binding was concerned with the interpretation of sentences containing phrases which are linked together.
[]()Chomsky’s ideas were labelled the Minimalist program, because (he assumed) it contained the bare bones of language structure.