RECURSIVE FUNCTIONS P0R COMPUTATION OF NATURAL SECRET 
LANGUAGES 
Jouko J. Sepp~nen 
TeKoLa Computing Centre, Helsinki University of Technology, . 
02150 ESPO0 15, Pinlaz~ 
The paper describes a system called RENTTU, a collection 
of functions for computation of transformations from natural 
languages into several types of natural secret languages such 
as Pig Latin etc. By a natural secret language, sometimes 
called cant in English or children's play language we mean s 
natural language transformation, which preserves important 
morpho- and phonological qualities of the underlying source 
language. Such languages seem to make up a rich culture, which 
is a mixture of lore, linguistics, cryptology, and - if you 
wish - of computing. 
Today the phenomenon lies probably nearest to folklore 
and children's culture, but in earlier times they have, and 
perhaps still do constitute important skills of the lower and 
underground world. In the Middle Ages they were widely used 
among ~ beggars, tramps, pedlars, vagabonds, thieves etc. To 
Europe they seem to have come through the Bysantitun from the 
East, where they still today continue to live an active life. 
As a linguistic phenomenon they seem to be universal 
and appear in some form and extent practically in all languag- 
es and cultures. Yet it seems that they have attracted little 
attention from the part of linguistics or cryptology. Most of 
the treatments seem to fall within SOCtOlln~Aistics or folk- 
lore. As cryptological methods the natural secret l~age 
- 251 - 
transformations are fairly simple and their value barely 
extends beyond the children's playground. Though interesting- 
ly enough they seem to well meet five out of the six so call- 
ed Kerckhoff's criteria often used to assess usability of 
cryptographio methods and systems. 
A~Vhow, they work well and fulfill their purpose in 
situations for which they are aimed for: exchange of oral 
messages in real time in a situation, where the unintended 
are present, Since the message does not leave permanent trace, 
it is not possible to examine it afterwards. 
For linguistic research they can be valuable in that 
they exhibit some phenomena and processes on the morpho- and 
phonologlcal levels that can serve as a framework for extra- 
linguistic experimentation, 
From computing and programming point of view the sub- 
Ject is of definite 'interest, since it involves nontrIvial 
morpho- and phonological recognition, context conditions and 
transformations, which offer an instructive excercise in 
linguistic progrsmming. 
Several types of families of secret languages, or rath- 
er classes of language transformations, are identified. The 
typology can be categorized as follows: 
I. Syllabic transformations, which may involve deletions, 
replications, insertions or substitutions of syllables or 
phonemic elements under certain contextual conditions. 
2. Transformations that involve syllable permutations, possib- 
ly intermixed with transformations of the above type. 
3e Types of transformations, which involve nonsyllabic splitt- 
ing of words into segments that ere interchanged with 
segments of a given key word or with those of the next 
word. 
Finally some hybrid types occur, which combine features 
of various types in one language. 
- 252 - 
Q 
As an example a detailed analysis is given of the Kontti 
language, which is widely known in Finland and belongs to the 
third category. In Kontti the split point is determined by a 
number of rules, ~hich are conditioned by some metric and 
prosodic features together with the vowel consonant structure 
of the two words. The transformations involve both durational 
and harmonic accommodations to preserve the metric qualities 
and to meet the vowel harmony laws in Pinnish. 
Most delicate rules and conditions are found in splitt- 
ing criteria concerning various kinds of diphtong, vowel pair 
and long vowel combinations, in which the Finnish language is 
fairly rich. Some cases take real ear for the l~age to be 
correctly resolved, but all rules and conditions can, as soon 
as they have become clearly established, be programned. Thus 
a full competence of the Konnti language can be implemented 
on the computer. 
Prom the linguistic point of view the transformations 
are interesting in the sense that they fully preserve the 
phono- and morphological weli-formedness of the underlying 
language, i.e. the resulting words and word forms could well 
be real substance of the source language. 
Another notable feature is that the reverse transform- 
ation based on the same set and order of rules restores the 
original words and word forms except for certain degenerate 
cases, which involve neutral vowels that are incapable of 
• receiving information about harmonic quality when it is 
char~ed. 
Presently the RENTTU system accounts for some 20 lan- 
guages known in Finland including some based on Swedish. It 
consists of a collection of some 50 neat LISP functions 
written in a purely functional style. The functional style 
of programming is not only well suited or applicable to the 
problem, but also clearly reflects the nature of both the 
- 253 - 
lingu£stie and the computational aspects of the processes 
involved. It also results in elegant and concise function 
definitions. The author would be grateful of any material or 
observations concerning such languages in other countries. 
The system has been implemented in INTERLIEP on the TeKoLa 
DECSystem-20 and is available as a listing from the author 
on request. 

References

8epp~nen J. (1981), Sananmuodostus leikki - Ja salakieliss~, 
Publications of the Linguistic Association of Yinland, 
8. Sananmuodostus, Symposio Seiliss~, 3.-4.9.1981, 
Turku 

0~ansuu H. (1916), Leikki - Ja se~ekielista, Suomenkielen 
tutkimuksen ty~me~ta, Gummerus, Jyv~skyl~ 
• - 254 - 
