SOME LINGUISTIC ARGUMENTS IN FAVOUR OF Mr BAR-HILLEL'S 
"NON UNIQUENESS" HYPOTHESIS 
By Prof. REVEIN 
Institut de Philologie Slave de l'Acad~mie des Scimnces 
30 a avenue Troubnikovski 
MOSCOU 
RESUME 
I. In his "Four Lectures" (1963) Mr Bar-Hillel advanced a hypothesis 
of "different grammars for the same language lying peacefully side by side 
somewhere in our brain". In recent years the writer -quite independantly from 
Bar-Hillel (whose "lectures" were unknown to him)- came across many arguments 
in favour of this "non-uniqueness" hypothesis. Stylistic considerations involved 
were discussed in a paper read at the Semiotic Conference in Kazimlerz (Poland) 
in September 1966. In the following some grammatical questions are briefly 
sketched (the whole being presented in the writers book "Methods of Modelling 
and Typology of Slavic Languages" - to appear in summer 1967);; 
2. There are no linguistic or loglc obstacles for producing words - 
out of morphs or simple syntactic groups (phrases) - out of words or morphs by a 
FS (finite state) grammar. Mr Vauquois and his colleagues have already demonstra- 
ted this fact for words in their brilliant study on the use of models in mechani~ 
cal translation. 
3. As for simple syntactic groups, one can get some insights into 
their structure by interpreting states on the diagramm of the FS grammar as 
linguistic categories (following in this a proposal by Bar-Hillel and Shamir 
from a work of 1960). For this the writer considers a restricted FS grammar 
(the diagramm of such grammar does not contain two arrows labelled by the same 
non-empty word). Under this restriction the distribution of non-elimlnable 
empty words provides for an adequate segmenting of the group. It is 
-I- 
/ 
worth mentioning that there is only on~ restricted FS graummr for a given 
language (if abstraction is made from arrows labelled with the empty word), 
4. The described restriction must be abandoned, if homonymous 
constructions are investigated. Nevertheless it will be shown that even here the 
~ 
distribution of the empty word labelling some arrows provides for an explanation 
of homonymity (e.f. there can be two different paths So they S I are $2~ S 4 flying 
S 5 planes So " and So they S 1 are $3 flying S6~S 5 planes So, giving the sentence 
they are flying planes. 
5. It can be assumed that the units of the level of words and that 
of syntactic groups (in some cases also of simple sentences) are generated not' 
only by a device equivalent to a phrase Structure grammar (e.f. a dependency gram- 
mar), but may be generated by a simplifier device (a restricted or non-restrlcted 
FS grammar), the only linguistically relevant difference being the fact that 
the former provides for a hierarchization of the parts (which can be semantically 
interpreted for all endocentric construction as a property-attributing relation), 
and the latter does not. 
6. Almost the same applies to the relation of phrase structure 
grammars to transformational grammars, the latter providing for a still higher 
degree of hlerarchization (and semantic depth). 
7. Following Bar-Hillel we assume that the speaker (or the hearer) 
constantly swltches, over from one to another way of production according to 
the needed (or possible) degree of hierarchization. In one particular case - 
that of £he seml-idlomatlc expressions- such a solution seems obvious (one 
can store them in the vocabulary as whole or construct them from parts). But 
the most of human speech seems to be seml-idiomatic in a broader sense. 
8. The non-uniqueness of syntactic description can be compared with 
the non-uniqueness of phonemic solutions as described first by Yuen Ren Chao 
and shown in its full importance in ~ remarkable paper by Vjach. 
-2- 
9. On contemplating the picture of Grammatic Activity as a 
constant switching over from one device to another, one is Struck by the 
analogy to the picture of many computing devices working in parallel, which 
according to yon Neumann ("Computers and Brain") explanesthe m~racle 
of human intuition in genral. 
/ 
-3- 
