He sat looking at the silent telephone.

A)The splitting procedure

The first division should be placed after He. So we receive two constituents of the first level: He and sat looking at the silent telephone. The test of deletion shows that none of the constituents can be dropped. We should conclude that the structure (in this case the sentence) has no governing element. Such structures are called exocentric that presupposes that the governing element if any is outside the structure. The first constituent consists only of one unit and cannot be divided any further. The second constituent can. It is to be split also into two in the following way: sat and looking at the silent telephone. This time the deletion test proves positive for the second constituent and we have to assume that the leading element in the structure is sat. Such structures, those that have the governing element are called endocentric because the governing element, the centre of the structure is within. Since the governing element can substitute the whole structure in the larger unit it is represented though the distributional class of the leader, in our case it is a verb in the Past Indefinite, symbolised by V_ed.Here again the first constituent, the leader, cannot be split any more but the second can. After the third division we have looking and at the silent telephone. this structure also has its leader: looking, and can be represented by the symbol of its distributional class — V_ing. The governing constituent being again simple we should try and divide the second one. This constituent should be divide like that: at and the silent telephone. The deletion test yields negative results for both the constituents, thus the structure is exocentric. The structure then should be represented in the larger unit by the symbols of the distributional class of the first element, here it is a Preposition and the symbol is prep, and the symbol of the leading element of the second constituent. The next structure presents some difficulties. It is not clear where the division should be placed. It might be placed either before silent or after it. Cutting the structure after silent produces a combination of elements that have no connection with each other. Placing the cut before it ruins the intrinsic connection between the Article and the Noun. But if take the Article as a modifier of the whole group, then the cut should be made before silent. In such case we receive two constituent none of which reacts positively to the deletion test: the and silent telephone. Again we received an exocentric structure and again it should be represented in the larger unit by two symbols — Det for the article and the symbol of the governor of the second constituent. This structure easily falls into two constituents — silent and telephone, of which the second is the leader as the deletion test proves; so the whole structure should be represented by it symbol N. Now we can draw a diagramme showing the relation of the words in the sentence we analyse putting in the point of divisions the symbols of the distributional classes to which the leading constituents belong To these symbol it is necessary to add the letter P indicating that the symbol represents not a single word but a Phrase. Since the initial structure is an exocentric one (and sentences are always exocentric) we include into the system of representation a special symbol S for the sentence.

NP VP

VP V_ingP

rprDetNP

DetNP

Pron V_ed V_ingprp Det A N

He sat looking at the silent telephone.

B. The assembling procedure

The opposite way, connecting the smaller units into larger strings, follows the same principles as the splitting procedure: string at every step should consist of only two IC's, the IC's should undergo the deletion test to define the leader of the structure,

If we take the same sentence (He sat looking at the silent telephone.) and try to apply to it the assembling procedure we may start either with left hand unit (he), or with the right hand one (telephone). In any case we are to group together only TWO units that is we can add to the initial one (he or telephone, respectively) only one unit. So if we start with He we can add to it only sat. The deletion Both constituent react negatively to the test and we should treat the structure as exocentric and can compose the first group: ((He)—(sat)). The next step makes the structure: (((He)—(sat))—>(looking)) in which deletion test proves positive for the constituent (looking) making the first constituent ((He)—(sat)) the host of the group. But the next step shows that we cannot add to the string we have made only one unit. It means that at this point of the structure of the sentence two composite constituent are connected. This is supported by the fact that we cannot unite the following two elements (at and the). The next pair (the and silent) yields the same result. And only when we come to the pair (silent and telephone) we can do it. But exactly as during the splitting procedure we find some inconvenience in joining only these two elements leaving the third (the) out. Yet we managed to create a string ((silent)<—(telephone)). After that we can make the string ((the)—((silent)<—(telephone))), then, (((He)—(sat))—>(looking)), and, finally, (((He)—(sat))—>(looking))—>((at)—((the)—(( silent)<—(telephone)))

If we now compare the two modes of the IC analysis we cannot but see that they are different only in the order of operations — the splitting starts with the whole sentence and breaks it up into smaller and smaller units, establishing the relations between each pair of constituents until the smallest further unbreakable elements known as ultimate constituents. The assembling procedure starts with the ultimate constituents and gradually assemble the constituents till we receive the largest structure a sentence. All other features including difficulties are exactly the same. Both modes yield exactly the same information: they relations of dependence of ultimate units if such relations actually exist in the structure analysed. The tree of dependence and the brackets can automatically be converted one into another.

The description of the distributional method of extracting information about Language and its Grammar presented in the §§5-7 show both strong and weak features of the method. The strongest feature of it is so called objectivity: exclusion of meaning as one of the criteria for linguistic units makes the procedure seem mechanical, automatically producing the same results whoever uses the procedure. But unfortunately it only seems, because it is impossible to omit the meaning even from the primary description. In §6 when forming the distributional class of "Verb Inflections" we omitted the element |er| without any expansion. The reason is simple — it is practically impossible to prove using only the data of the distribution of |er| that it is not a element of the "Verb Inflections" distributional class. Still we know that the unit |er| makes new words but not forms of words as other inflections do. If we did not know it we could have included |er| into "Verb Inflections" and further interpretation of the verbal system would have met insoluble problems. In §7 the deletion test and the very possibility of correct splitting and uniting — all these features are controlled by the Meaning. Th Immediate Constituents Analysis has its own specific drawback — each constituent can include two and only two constituent parts. The problems resulting of this rule are illustrated in §7.

The limitations of the distributive analysis are not accidental. They are due to the image of the object under analyses, the idea of what is Language. Language was modelled as a purely formal system in which contrast of form indicated contrast of meanings. Difficulties and limitations of the distributional analysis are results of this image of Language. But Language is not only a purely formal system in which contrast of form indicated contrast of meanings, it is also a functional, meaningful, cognitive and operational system. Other ideas of Language may cause other types and methods of analysis.