Some Remarks on J.L.Mey's Paper for the International 
Conference on Computational Linguistics, Sweden 1969 
P.Sgall, E.Haji~ovi 
Mey's criticism of the functional approac~ to gene- 
rative description concerns (1) the formal properties 
of the system proposed by Sgall et el. (its weak gene- 
rative power, recursivit$~), and (2) some itl~or!~al que- 
stions connected with the mentioned approach. 
(1) From the formal point of view, Mey's paper con- 
tains many quite unclear points and errors, which make 
his cl~Jms unfounded. Some of those may be due to a me- 
re unpreciseness and carelessness Jn formulations (cf. 
for instance P.7, where he speaks ~,bout "e language 
that is not CF, or may be not even regular", w!ich is 
as if one says "This mineral is not found in Europe, 
not even in whole Switzerland") but others hsve a more 
consequential bearing on his further argumentation. He 
confuses (p.3) the trsnsl~tion by the means of s push- 
down store transducer in Evey's sense (henceforth pdt) 
with the question of CF-preservation in the sense of 
Ginsburg (and Rose); he dose not seem to realize fully 
(pp.3-4) that Sgall et al use only the notion of pdt 
in the sense of Evey and not in that of Ginsburg. 
Ginsburg and Rose's theorem is not identical with Evey's 
theorem 2.6.6, which is based on a different definition 
of pdt, connected with notions of input and output lan- 
guages defined by the means of the notion of computation. 
Although the theorem of Ginsburg and Rose, partly inspi- 
- 2 - 
red by that of Evey, needed a correction, it does not 
follow that Evey's theorem is wrong. Of course, it 
would be of interest to analyze the relationships be- 
tween Evey's system of notions and that of Ginsburg, 
es well as to give an explicit account of the eventual 
bearing of Ginsburg and Rose's result with respect to 
the theorems and proofs contained in Sgall et el. But 
Mey does not undertake an~ such analysis in his paper; 
without giving any proofs he simply assumes that one of 
these results is c0ntradicted by the others. 
Thus we can state that Mey has not shown that a 
system of the discussed type generates a language that 
is not context-free, to say nothing about his clearly 
exaggerated claim (P.7) of having 'shown" ths~ the 
lengusge generated by such a system "simply never" is 
context-free (cf. the bottom of p.4, where Ginsburg and 
Rose's formuletions are rendered in a rather c6nfused 
w~y). 
Further, Mey is not right in claiming that a device 
of the discussed type is "practically equivalent" to 
(universel) Turing machine (P.5), or that its output 
l~nzu~ge is not necessarily a recursive set 
(pp.4, 5, 9). As shown in ~lha (1966), quoted in Sgali 
(1967), and as ststed again in Chapter 5, of Sgell et 
el., s recognition procedure for the system under con- 
sideration does exist (this is ensured by the preserva- 
tion of the 1,n~th of the strings). 
b 
-- 3 I 
(2~) The informal parts of Mey~ criticism contain 
first / of all hls question "what about the remaining 
input, where does it all come from?" (P.7); in case 
that the output language of a device in our system (or 
the terminal language of the grammar)i8 a proper sub- 
set of the input language of its successor in the sequen- 
ce of devices, the difference between these two languages 
should be taken as a formal counterpart pecific syntax 
of a given level ( cf. the distinction between ftik and 
blik, or that between or cqmed at home and golf pla2ed 
John;cf. Sections 2.1.4 and 2.2.5 in Sgall et el.). 
We cannot discuss here st length questions of this 
kind, es well as other items, which are,under the given 
conditions, rather questions of taste. Certainly, the 
significance of the eventual possibility (not claimed 
by us) to use the existing CF-recognition routines (May, 
P.5) or the significance of the results obtained by 
European traditional and structural linguistics may be 
appreciated differently. We do not take the "time-hono- 
red" European linguistic tradition as a linguistic argu- 
ment, but we do not want to give a mere preference to 
American traditional high-school gra~,ar over the Euro- 
pean one. We would like only to recall that in the de- 
velopment of the transformational theory there are vari- 
ous points showing that a more careful attitude to the 
"classical" linguistics could have saved some detours; 
so one would have been able e.g. to see earlier the ne- 
- 4 - 
cessity to distinguish between a deep (or semantic) 
3tructure and a surface one (and not to provide the 
transfor~:~ational component with recursive propertles)~ 
Ys it not clear that the transformational description 
does lose, successively, at least some of the properties 
distingulsling it from a description of the stratifica- 
~ional or functional type? 

References

R.J.Evey (1963), The Theory and Application of Pushdown 
Store Machines, Mathematical Linguistics and 
Automatic Translation, Rep. No. NSF-10,Harvard 
Comput.Lsb., Cambridge, Mass. 

S.Ginsburg (1966), The Mathematical Theory of Context- 
Free Grammars, New York 

S.Ginsburg & G.Rose (1966), "Preservation of Languages 
by Transducers", Inf.Contr.9, 153~176. 

S.Ginsburg & G.Rose (1968), " A Note on Preservation Or 
Languages by Transducers", Inf.contr.12,549-552. 

J.~\[.Mey (1969), On the Preservation of Context-Free 
Languages in a Level-Based System, International Conference
on Computational Linguistics, Sweden 

A.~ha (1966), On the Recognition Procedure for Pushdown 
Store Transduoers, The Prague Bulletin of 
Mathematical Linguistics 5, Pp.3-15 ~ ' 

P.S~ell (1967), Generativn~ popis jazyk~ a ~esk~ deklina- 
ce ( A Generative Description of Language and 
the Czech Declension), Praha 

P.Sgall et al. (1969), A Functional Approach to Syntax 
in a Generative Description of Language, 
New York 
