LECTURETTE #3 MOVEMENT-TRANSFORMATIONS VS. CONSTRAINTS How does REST differ from ST? During the post-Aspects development of the Standard Theory, many changes took place in the basic assumptions of the framework. In this and the next few lecturettes, i'm going to try to summarize the most important of these. First of all, there was a major shift in attention from movement-trans- formations to constraints. The Aspects model focussed on identifying and defining transformations. In generative syntactic scholarship from the 60's (and later) you find a lot of talk about specific transformations -- the Passive Transformation, Equi-Deletion, Subject-to-Object Raising, Dative Shift, 'For-To' Deletion, 'There'-Insertion, that sort of thing. Attempts were made to generalize across various transformations, but it was commonly assumed that a syntactician's job was primarily to define exactly what a specific transformation did and under what circumstances it was appropriate. What generalizations were proposed were regarded as actual transformations themselves, superseding other transformations which were then regarded as special instances thereof. Furthermore, transformations were typically regarded as language-speci- fic; it is obvious, after all, that German's 'Passive Transformation' is different from that of English, and 'Dative Shift' doesn't seem to ope- rate in French at all. This means, of course, that every child learning a language has to learn a bevy of discrete transformations. In his seminal 1968 Ph. D. dissertation Constraints on Variables in Syntax (published in 1986 under the title Infinite Syntax! -- God knows why!), John R. 'Haj' Ross argued that there were certain things transfor- mations just can't do, and furthermore that at least some such con- straints were cross-linguistic -- they seem to hold for all languages, or at least an impressive number of them. During the 70's, while much of the formal language of the Aspects model ('Structural Descriptions', 'Structural Changes', etc.) continued to be used, the 'cutting edge' of syntactic research, at least in the Chomskyan school (which began to lose whatever hegemony it had had in the field of generative syntax at this time, with the development of Relational Gram- mar), followed up on Ross' work in trying not so much to define specific transformations as to identify constraints on transformations and on the implicit power of the transformational formalism. After all, it was rea- lized, the formalism for defining transformations in the Aspects model could do an awful lot -- there was a huge range of conceivable transfor- mations that could be formally described but which didn't seem to be attested in any known human language. If the goal of grammatical theory is to understand precisely what it takes to be a human language, then obviously the Aspects formalism was missing something. Some of the more generally important constraints that began to be defined during this period will come up for discussion later. (Of course, it has to be understood that some constraints have been proposed as linguistic universals only to lose that status due to subsequent research. For in- stance, i am currently working on a project addressing the validity of Ross' 'Coordinate Structure Constraint' as a linguistic universal. The important thing is not so much the validity of any particular proposed constraint as the general notion that constraints make up a significant part of Universal Grammar.) For now, what is important is that by the end of the 70's it was becoming increasingly obvious, at least to some people (including Chomsky), that the way the theory was progressing there was really very little point in trying to maintain any kind of ontologi- cal distinction between different transformations. It was becoming in- creasingly clear that the theory could operate with one single, generic transformation, which came to be known as 'Move Alpha' and was understood to mean 'Move any constituent Alpha anywhere (as long as no relevant con- straints are violated in the process)'. In saying this, i must grant that at certain stages of work in the REST framework it has proved unfeasible to reduce every single reasona- bly-well-motivated transformation to an instance of 'Move Alpha'. For this reason, some have proposed as an alternative a slightly broader generic transformation, commonly called 'Affect Alpha'. (Where 'Move Alpha' is typically interpreted as 'Move anything anywhere', 'Affect Alpha' is interpreted as 'Do anything to anything'.) In addition to moving a constituent from one part of a constituent structure to another, Affect Alpha would presumably be able to do such things as rearrange the constituent structure in the vicinity of a certain constituent without actually moving anything. It might also be able to delete certain nodes; the question of node-deletion will come up in Lecturette #6. However, it must be noted that Affect Alpha is of dubious status. There are many proponents of REST who favour a 'strict' view of the transforma- tional component, according to which only transformations describable as instantiations of Move Alpha are admissible. These people would regard any evidence for a broader/looser generic transformation Affect Alpha as merely evidence that the proper analysis hasn't yet been discovered. While i have to confess that in my heart i find the notion of Affect Alpha attractive, my head points out that a theory that can dispense with it is ipso facto a stronger theory. So i will in the future continue to treat Move Alpha as *the* generic transformation in REST. This shift of focus is symptomatic of a major difference between REST and most of what had preceded it in the field of syntactic theory: Up until sometime in the 70's at least, the main focus in grammatical theory was on 'rules' -- statements about precisely what 'happened' in the genera- tion of a given string. But in REST the focus is on general *principles* of grammar. What *happens* is Move Alpha; what grammatical theory is about is the general principles that govern and constrain it. We will come back to this push toward generality again and again in the coming weeks, and you will see that 'rules' in the traditional sense are of decreasing moment in the theory -- in fact, it is considered desirable to dispense with them entirely. (I believe, though i am not confident, that a similar trend has been taking place in phological theory. Some of you who know more about that area may be able to clarify this.) At this point the question may be raised, If the transformational compo- nent of the grammar can be reduced to a single, generic transformation, Move Alpha, why can't it be eliminated altogether? This is a very good question, and indeed there are several quite respectable frameworks of syntactic theory that dispense with movement-transformations entirely. But it should be noted that these frameworks make use of completely dif- ferent means of stating the generalizations that Standard Theory des- cribes by means of movement-transformations. For instance, in an agentless passive clause like The door was opened. 'door' behaves in some respects like the subject (it precedes the verb and controls verb-agreement) and in some respects like the direct object (most obviously in the semantic sense, in that it is the patient or theme of the action). In transformational frameworks like the Standard Theory and Relational Grammar (RG), this fact is stated by claiming that it *is* *both* subject and direct object, but at distinct levels. In non-trans- formational frameworks like Generalized Phrase-Structure Grammar (GPSG), Lexical-Functional Grammar (LFG), Head-Driven Phrase-Structure Grammar (HPSG), and Role & Reference Grammar (RRG) superficial features such as grammatical relations (e.g., the subjecthood of 'door') are represented in one way (e.g., by a constituent-structure tree diagram), while such things as 'deep', semantic relations are represented in a completely dif- ferent way (e.g., in the verb's 'logical' or 'argument structure'), and the connection between the two, which in the Standard Theory is represen- ted by a movement-transformation, is described in terms of the verb's lexical entry. But given the Standard Theory's insistence on defining all grammatical information in terms of constituent structure, both the 'deep' semantic relation of patient-hood or theme-hood and the superfi- cial grammatical relation of subjecthood must be described in such terms, which necessitates a movement-transformation to relate the two. It should be understood that, in spite of the focus of interest in REST on constraints on movement-transformations, movement itself is regarded as the major factor in generation. As one of my professors once put it, the transformational component may have been maximally generalized and the number of transformational rules reduced to one, but that one is quite hearty. It is movement that drives all of the processes of deriva- tion, in the view of this framework. Think of it as being somewhat like breathing; we linguists don't talk much about breathing, but we can't talk without it. Likewise, to a proponent of REST movement is a given; what is interesting is why it occurs under such-and-such circumstances and why it doesn't under such-and-such other circumstances. In principle, movement can take place whenever and wherever it does not result in violations of any constraints. For a while during the 80's it was commonly supposed that in fact it took place freely; i remember Chomsky in 1988 claiming that a person's internalized grammar routinely overgenerates, creating gobs and gobs of garbage for every decent sen- tence that person actually uses. But increasingly during the late 80's proponents of REST, including Chomsky, pushed for a principle according to which movement only takes place when it is necessary, i.e. when fai- lure to move would result in ungrammaticality. This principle is some- times referred to as the 'Least Effort' or 'Laziness Principle'. More ominously, it is sometimes called the 'Orwellian Principle', on the ana- logy of a political principle in Orwells' _1984_: 'If not forbidden, then obligatory; if not obligatory, then forbidden.' Call it what you will, the Laziness Principle has recently been enshrined as a fundamental prin- ciple of the Minimalist Program. The 'Orwellian Principle' is often invoked in the physical sciences, but not in the same way it is in REST. In syntactic theory it typically means 'Move if you have to in order to avoid ungrammaticality; otherwise stay put'. In physics it means, if i understand it correctly, 'This event will take place *sometime*, unless there is some explicit natural law (which we may or may not know about) preventing it'. For example, one theory says that there is, in fact, no fundamental law preventing protons from decaying; therefore, according to this theory, a typical proton must decay -- sooner or later. But 'sooner or later' does not mean that protons must in fact decay every time you look at them; obvi- ously, they don't. The theory predicts that in this case 'sooner or later' means roughly squijillions of years, so the likelihood of any particular proton actually decaying in anyone's lifetime is very small. (Last i heard, there was no actual experimental evidence that the proton could, in fact, decay, and the theorists are getting antsy about it.) If the physicist's notion of the Orwellian Principle were applied to syntax, it would mean something like, 'Unless there is some good reason why con- stituent A can't move to position P, expect it to move there *on occa- sion*', not 'If A is allowed to move to P, then it *must* do so *under all circumstances*'. In my own dissertation, i argued that the standard REST version of the Orwellian Principle, or Laziness Principle, was too strong: the corpora of early Indo-European languages like Latin and Sanskrit are full of re- orderings that apparently occur not because the grammar requires them but solely because it does not forbid them. In other words, the physicist's understanding of the Orwellian Principle seems to apply better to the grammar of such languages than the standard syntactician's understan- ding. Partly for this reason, i am sceptical of the Laziness Principle as it is currently enshrined in the Minimalist Program. Best, Steven --------------------- Dr. Steven Schaufele 712 West Washington Urbana, IL 61801 217-344-8240 fcosws@prairienet.org **** O syntagmata linguarum liberemini humanarum! *** *** Nihil vestris privari nisi obicibus potestis! ***