Transformations

The still-dominant view among generative linguists seems to be that surface forms of sentences are transforms of deeper forms which are themselves sentential in nature. Originally these deeper structures were thought to be ‘kernel’ sentences of active, declarative form; during the past decade they have gradually become more abstract, but still forms of sentences. The implication of the very recent work on presuppositions, as well as of my little demonstrations, would seem to be that what is ‘transformed’ into a surface sentence is not another ‘sentence’ (hyper or otherwise) but rather a momentary cognitive state which is not linguistic at all yet has its own complex semantic structure.

Entering the human information-processing system via purely perceptual signs and events eliminates any obvious dependence, at least, upon prior sentences. Furthermore, ideally, such perceptual inputs serve to ‘anchor’ the states of the cognitive systems of different speakers so that transformational equivalents can be observed directly when they Simply Describe. I have already noted in passing several instances of surface transforms - for example, the combination of a present or even past-tense verb with toward to express the same Future feature as will )rolled toward substituting for will hit) or the interchangeability of and it )hits, etc.) and which )hits, etc.).

The reason I said above that ‘ideally’ speaker cognitive states are anchored by common perceptual input is that in these demonstrations - as compared with color film strips, for example - the observers may actually be perceiving somewhat different events and they certainly may be attending to different aspects of events selectively. Careful inspection of the 26 sentences produced by demonstration #10 (Table 2) clearly indicates this: for example, subject # 1 says the small dark hall hit the large blue ball (reporting only the contact event), whereas #2 says the man rolled one black ball along the table, hitting the blue plastic ball (reporting the presupposed Agent, ROLLING as well as HITTING events, and even the TABLE locale). But we can also note genuine transformations of the ‘ same ’ semantic information across different describers: # 3 implies with a passive construction )a small black ball was rolled against the larger blue ball) the same presupposed Agent that # 2 made explicit; # 14’s sentence is the passive version of # 1 ’s simple active )a blue ball was struck by a smaller black ball) #9 expresses MAN ROLLING BALL and BALLS COLLIDING in two sentences conjoined by and: the compound noun squash-ball expresses the same information for speaker #18 as does small black rubber ball for other speakers. Any distinction between syntactic and lexical transformations would be hard to draw - what, for example, about the obvious equivalence of #7’s collides with, #10’s rolls into, and #11’s rolls and hits?

The two demonstrations requiring repeated sentencings were designed deliberately to get at transformations, both across repetitions and within single speakers. As an illustration of transformations across repetitions, I checked the relative frequency of passive constructions across THE BLACK BALL HITS THE BLUE BALL AND THEN THE BLUE BALL HITS THE ORANGE BALL (#12-14); there were only 19% passives for #12, which rose to 50 % for #13, and then fell again to 34% for # 14 - clearly suggesting that this construction is a non-preferred but readily available transformation.

A few examples of within-speaker (across-repetitions) transformation must suffice: for #12-14, speaker # 1 says the small dark ball hit the blue ball which hit the orange ball, then provides us with one of the rare center-embeddings the blue ball the dark ball hit hit the orange ball, and finally gives us a reverse entity ordering the orange ball was hit by the blue ball which had been hit by the dark ball, all necessarily cast into the passive form; for speaker #6 sentence 1 is the forward active and sentence 3 is the backward passive, but sentence 2 is an interesting combination of directions, the blue ball was hit by the black ball and hit the orange ball. For #18-20, speaker #15 tells us first that a little black ball is perched on the end of a long tube which in turn is standing inside a can cover, then that three objects are piled atop each other - from bottom up, a can cover, a tube and a ball, and finally, with extraordinary efficiency, that a ball on a tube is in a can; speaker #22 does a kind of phrase-shuffling act, moving from a ball sits on top of the cylinder which stands in the lid, into a cylinder holding at its top a ball stands on the lid, and finally to a lid has standing in it a cylinder which is topped by a ball.

If such sentences are all transformations of the same cognitions established by the perceptual input - and they certainly do seem to be informationally equivalent - then they would seem to pose some interesting problems for the generative grammarian. For example, if they are semantically equivalent transformations, then they should, at some depth level, trace back to the ‘same’ hyper-sentence. But what would this be? Or another question: what, in a generative grammar, ‘triggers’ one surface structure as opposed to some other? To say, for example, that some Manner T-passive is attached to the entire, deep-structure tree of speaker #1’s third effort above is entirely post hoc and begs the real question, at least from the point of view of a performance model.

One thing suggested by these examples and the total data is that particular transformations are ‘triggered’ by the entity which is the focus of the speaker’s momentary interest or attention and is therefore likely to be represented in his first encoded noun phrase. Once an observer of perceptual event #12 has begun encoding the orange ball. . ., he is, so to speak, stuck syntactically by this decision in conjunction with the cognitive ‘ facts of the world ’ - that THE ORANGE BALL WAS HIT BY THE BLUE BALL (not the reverse), and that THE BLUE BALL WAS PREVIOUSLY HIT BY THE BLACK BALL (not the reverse), and so forth. This would suggest that the job of syntax is not central but rather peripheral in ordinary language - merely accommodating lexical decisions made on the basis of the fleeting interests and motivation entertained by speakers - a notion that will surely raise the hackles of many linguists.