REWRITE
FORMULAS IN GLOSSING
The
aim of the present paper is a modest one.*
We
have all been waiting lately for the promised extension of descriptive
linguistic techniques to the field of semantics.
Any such extension is going to face the following groups of
problems :
1.0 Investigation or discovery, to include –
1.1 Collection
of data – primarily, a corpus of messages and secondarily, previous
attempts at scientific or pre-scientific description; and
1.2 Collation and analysis – primarily, breaking
the code underlying the corpus and secondarily, “structural restatement”
of previous descriptions.
2.0 Presentation in a form that can be tested
by others. Any of three styles
may be adopted –
2.1 A set of instructions for encoding a message,
that is, generating a well-formed utterance (GENERATIVE PROTOCOL).
2.2 A set of instructions for decoding a message,
that is, deciding whether a given speech event is well-formed and,
if so, exhibiting its structure (DIAGNOSTIC PROTOCOL, comparable to some extent to the decision
procedures of logic).
2.3 A direct presentation of the code (underlying
well-formed utterances and adumbrated in 2.1 and 2.2 above) in classificatory
terms (DESCRIPTION OF THE SYSTEM).
3.0 Criticism or evaluation, especially with
a view to choosing between alternative analyses.
What
I propose to do here is to adopt the so-called rewrite formula, a
trick successfully employed in presenting grammar (and lately, phonology)
as a generative protocol as defined above, and try it out in one area
of the presentation of semantics – namely, translation glossing. To
use Jespersen’s triad, the three modes of presentation described above
may be said to proceed in the following manner :
Generative : Notion Function Form
Diagnostic : Form Function Notion
Descriptive : From Function Notion1
The general schema for
the formula will be (read x → y as “rewrite x as y”) :
{From with Function} (with
“Gloss”) →
{From with Function} with
“Gloss”
where “Gloss” will be
either absent on the left hand side or different on the two sides.
While {Form with Function} will be specified by grammatical spelling
with or without limiting conditions, “Gloss” will be a translation
in the metalanguage. For the purpose of this paper, we shall not attempt
to show the usefulness and relevance of translation glossing beyond
saying that it can find a place eventually in the diagnostic protocol.
In all my examples, I shall use Marathi as the language whose
forms are to be supplied with gloss and English as the metalanguage. A few simple examples may be set forth :
(1)
bhau →
bhau “brother”
(2)
lek masculine → lek masculine “son”
(3)
culta →
culta “father’s brother”
(4)
dhak¶- →
“who is born later than ego”
(From now on the carryover
of {Form} from the left hand side will be left understood.)
(5) adj. + subst. → “adj. + subst.”
Applying these instructions, we get :
(6)
dhak¶a bhau → “brother who
is born later than ego”
*(7)
dhak¶a lek
masculine →
“son who is born later than ego”
Obviously we are in trouble, because we have failed to account
for the conditioned modification of the gloss in (4) in this context. A suitable restriction will be :
(8) dhak¶
– when adj. To lek masculine, etc. → “who came to have
that relationship with ego as junior to others who have the
same relationship with ego”
This will ‘take care’ not only of
(9)
dhak¶a lek masculine
“son who came to have…ego” (i.e., “younger son”)
but also of :
(10) dhak¶i
bayko → “wife who came to have … ego” (i.e. younger wife”)
(11) dhak¶i
sun → “son’s wife who came…ego”
(A moment’s thought will show that the new gloss of dhak¶- cannot do duty in (6).)
so far, we have set up three kinds of rewrite formulas for
translation glossing : (a) the one giving the basic notional range
of a particular simple form or formal feature; (b) the one deriving
the gloss of a composite form by combining the glosses of its components
– for example, in (6) we have pooled together the glosses of
dhak¶-, bhau, and adj.+subst.; and (c) the one requiring a modification
in the gloss of a simple form when it appears as a component of some
larger composite form of a specified type – for example, (7) calls
upon us to interpret dhak¶
- not as “who is born later than ego” but as “who came to have that
relationship with ego as junior to others who have the same relationship
with ego” when it appears in the environment, for instance, to lek——
-a ahe.
One may wonder at this point, why all this fuss, why not replace
both (4) and (8) by a single formula?
*(12) dhak¶- →
“younger”
This will fit (6) as well
as (9), (10), (11). We shall
return to this point a little later after considering a few more examples
:
(13) culti → “fathers’s brother’s
wife”
*(14) dhak¶a
culta → “father’s brother who is born later than
ego”
This does not work, so we try putting culta along with
lek masculine, bayko, etc. in (8) and deriving therefrom.
*(15) dhak¶a
culta → “father’s brother who came to have that relationship
with ego as junior to others who have the same relationship with ego”
This does not work either.
A similar situation obtains with culti. The results
that we do desire as consistent with the facts of Marathi speech are
respectively.
(16) dhak¶a
culta → “father’s dhak¶- brother”
(17) dhak¶i
culti → “father’s
dhak¶- brother’s
wife”
(The incongruous dhak¶-
in gloss can of course be suitably gotten rid of later by applying
(6).)
It is obvious that the rewrite formulas of type (b) and (c)
(combinatory derivation and conditioned modification) cannot cope
wit this. To set up special formulas of type (a) individually
for these two (and similar other) cases is a costly and possibly wasteful
procedure, though I suppose we shall be forced to adopt this course
in the case of recalcitrant idioms like
(18) (a)
buddhi →
“intellect”
(b) b«½ → “power”
(c) buddhib«½ ~
budb«½ →
“chess”
(19)
culta “…brother / sister…” with
culti
etc.
dhak¶-
→ “…dhak¶- brother
/ sister…” (with no further
gloss for dhak¶-)
(Here ‘etc.’ covers other items like dir “husband’s brother”,
n«¸«nd “husband’s
sister”.)
This strikes me as in some way analogous to the grammatical
rewrite formulas for obligatory transformation : further analysis
is necessary at this point.
Coming back to the point raised by *(12), we realize that dhak¶- of (19) cannot be translated by “younger”
in English :
*(20) dhak¶a culta →
“younger uncle”
“Younger uncle” in English is rather to be paraphrased “uncle
who is younger than other uncles” (cf. *(15) above).
This brings home to us an important point about the glossing
metalanguage we may use in the presentation of semantics – namely,
that one should not depend on detailed congruences between the object-language
and the glossing metalanguage (in our case, Marathi and English respectively). Ultimately, our primary aim here is not a ‘transfer grammar’ between
Marathi and English (with both as object languages) but a description
of Marathi in its own terms. The
glossing metalanguage that is to serve as our tool, therefore, is
not the natural English spoken in some speech community or other but
an artificial version of it.
We have so far indicated five possible types of glossing formulas
: (a) the simple, (b) the simple additive, (c) the conditional, (d)
the simple suppletive, and (e) the replacive – exemplified respectively
by (1), (6), (8), (18c), and (19). A few more examples may now be
given to exhibit one more possibility :
(21)
bhau “…brother…” with cul«t
culta
dir
etc.
preceding → “…father’s brother’s son…” (with no further gloss for cul«t)
Deriving from this :
(22) cul«t bhau → “father’s brother’s son”
(23) cul«t culta
→ “father’s father’s brother’s son”
(24) cul«t dir → “husband’s father’s brother’s son”
But there is a further complication :
(25)
cul«t bhau “….brother….” with
culta
dir
etc.
cul«t preceding “…father’s brother’s son…”
Applying this,
(26) cul«t cul«t bhau →
“father’s father’s brother’s son’s son”
More simply, we can say that (21) is a replacive formula capable
of recursive application. Another example of the recursive sub-type
of type (e) is
(27)
bhau “…brother…” with mav«s
culta
dir
etc.
preceding → “mother’s sister’s son….”
Applying (21) and (27) jointly and, recursively, we are in
a position to supply glosses to possible forms like
mav«s bhau
cul«t mav«s bhau
mav«s cul«t bhau
mav«s maves cul«t bhau
It should perhaps be added
that different but analogous formulas will be needed to explain cul«t
and maves used with other kin terms than bhau, culta, dir, etc.
I might as well pause here, since my chief aim was raising
problems rather than indicating solutions.
COLOPHON
* An earlier version of this paper was presented
at the Autumn Seminar of Linguistics held at M.S. University of Baroda
in October 1962, to whose co-participants I am indebted for comments
and suggestions.
This version was published in Indian Linguistics 24:25-30,1963.
- Popularly, the three may be dubbed as the speaker’s, the listener’s,
and the bystander’s grammar respectively. It must be clearly understood though that
the first two are at best “official” reports of what happens in
the actual traffic of language – bearing to it as much resemblance
as cadaver dissection does to what the surgeon sees when he opens
up the living body. Incidentally Chomsky and Halle’s rejection of
the demand for explicit discovery procedures applies legitimately
only to code-breaking. (1.2) (an activity undertaken by the analyst)
and not to decoding (2.2) (an activity on the language-user’s part
observed by the analyst). I have a suspicion that Halle at least
is guilty of confusing the two.