THE SCOPE OF A HISTORICAL DICTIONARY
The OED tag ‘On Historical Principles’ is understandably emulated
without realizing its full implications.
A historian of human culture or of language has no control
over the selection of his data. The historical processes themselves which he
presumably wants to understand have sieved the data for him – from
his point of view this sievage may be felicitous or disastrous or
indifferent; in any case the result is always going to be fragmentary.
I am of course speaking of his principal data – the so-called
primary sources. In relation
to the history of language these are texts (or specimens of the use
of language). Besides being fragmentary, the data is also
going to be ‘dead’ in the sense of being a residue of transitory events. For the older stages of the language, for example,
we may have texts available for a postmortem possibly with the help
of an interpretative tradition, but no sample user of the language
with his Sprachgefűhl. The
historian, in order to make up for this double handicap, resorts to
the principle of uniformity.
Hutton, a practitioner of geology which is also a historical
inquiry unlike physics or chemistry, formulates it in the following
manner (as summarized in Labov 1971: 482): processes which operated
to produce the geological record are essentially the same as the ones
now taking place around us – weathering, sedimentation, volcanic activity,
earthquakes, and so on.
A student of human history unlike that of natural history has
a second string to his bow – which can be as treacherous as it can
be helpful. The human beings whose history is being investigated
are disposed to stand back from their own activities even as they
are immersed in them and maintain a running commentary. Indeed some of these comments may stem from
contemporary or subsequent historians’ activity itself. In other words, the so-called secondary sources
of a historian may have been left behind by his forerunners.
Focusing more narrowly on the history of language we must at
the outset make a distinction between a synchronic analysis of a historical
stage of a language to which we have access only through texts and
not through live contact on the one hand and linguistic history proper
which diachronically connects various stages (including of course
the contemporary stage if any). By a historical dictionary we shall mean a
dictionary that is historical in this second sense. Whether we have in mind the first or the second sense of the historical
study of a language, we may say that the historian of language has
before him as data the following:
All available texts – written texts or, in very recent times, mechanically
recorded texts-duly authenticated, edited, dated where possible, and
above all assigned to the correct corpus or subcorpus (in Medieval
North India, for example, it is not always easy to say whether the
text in Old Braj or Old Khari Boli or Old Braj influenced by Rajasthani
or Old Rajasthani influenced by Braj and so on; in Eighteenth-Century
western Europe, it is important to say whether the text is a true
narrative or a simulated narrative, colloquial or literary, aristocratic
or simulated aristocratic or bourgeois and so on; the linguistic features
in a manuscript text or letter may be assignable to the official author,
the ghost-writer, the redactor, or the scribe).
All available secondary responses to language (thus, a parenthetical
comment like ‘if I may be forgiven for using a vulgar expression’,
helps us in locating the expression that follows along the stylistic
scale; definitions of technical terms offered by writers in the field
or annotations of earlier texts by later commentators are best regarded
as belonging to his category).
All available previous linguistic descriptions (thus, the attestation
of a vocable together with its gloss in an old lexicon or any etymological
proposals are clearly on a different footing from the primary attestations).
The dividing line between the last two is not a hard and fast
one. For example, the etymological proposals made
by Sanskrit authors may not always be valid, i.e., may remain unusable
under [c]; but they may still throw light on the contemporary concepts
behind the words so etymologized, i.e., may remain usable under (b). We will call citations under (b) and (c) meta-citations
to distinguish them from the citations proper under (a). Meta-citations can be treacherous in that they
are a fertile source of ghost senses and even ghost words.
The use of the qualification ‘all available’ used earlier is
of course to be taken with several grains of salt.
Only in the really unfortunate case, like that of Gothic or
Avesta, is this idea easily attainable and therefore painfully desirable.
(For such languages, the historical dictionary may well turn
out to be a concordance for the whole corpus).
In the case of Sanskrit the problem is the opposite one – of
principled selection from the embarrassing riches.
The selection will be at three points – selection of texts
for extraction, selection of excerpts for filing, and selection from
filed excerpts for citation in the published dictionary.
The discussion of the principles that will help us in grading
the texts in order of relevance or in fixing the density of extraction
for each selected text or in determining the density of extraction
for each selected text or in determining the size and composition
of a fair sample is beyond the confines of this paper. I may simply note in passing that the single
rubric like ‘Sanskrit’ may conceal a variety in spite of the relative
stabilization, even fossilization of certain of its varieties over
long stretches of time. The
coinage of rubrics like ‘Vedic’, ‘Classical’, ‘Neo’ is only the first
approximation – we have to make finer chronological, geographical,
social, and stylistic slices.
While the shifts observed in the citations and metacitations
in the given language will constitute the primary residue of historical
processes, one must not lose eight of the evidential value of data
from (a) ancestral languages (e.g. data from Pali for a historical
lexicographer of Sinhalese); (b) cognate languages; (c) descendent
languages; (d) donor and recipient languages; and (e) substratum and
superstratum languages. In the case of Sanskrit, evidence from Avestan,
Old Persian, Middle Iranian, MIA (especially Pali in its triple capacity
as descendent, recipient, and substratum), NIA (especially early NIA),
Tibetan, Dravidian, Munda, Arabic, Old Javanese, etc. may be helpful
in various ways – including the reconstruction of obscure senses. There is of course no documentary evidence for the pre-Sanskrit
In the long and arduous process of collation and interpretation
that will follow and that will in turn feedback to the activity of
collecting and selecting the sources should be envisaged as falling
into two somewhat distinct passes corresponding to the vital distinction
between chronicle and history proper.
A chronicle records the more accessible historical facts and
documents them. A history reconstructs or extrapolates the less accessible facts
and interprets and theorizes about the processes that account for
all the facts – whether accessible or extrapolated.
The existing historical dictionaries are (when they are historical
at all and not merely sketchy guides for the contemporary reader to
the interpretation of the older texts of the language) are either
chronicles or chronicles struggling to be history but not quite making
it. It is far from my intention to suggest that
a chronicle is merely a poor cousin to history. A good chronicle is an indispensable foundation for a good history;
it is good precisely because it is conscious of its importance. A
bad chronicle is bad either because it is sham history or because
it is not inspired by a sense of relevance but merely compiled.
Now, it is perhaps true that given the present state of Sanskrit
scholarship, given the vastness of the total corpus, and given the
chronic uncertainties of the pre-Muslim chronology of India, we can
only attempt a lexical chronicle for the family of languages known
as Sanskrit or Old Indo-Aryan postponing a lexical history to a rather
remote future. At the same
time, in order to become a good chronicle this lexical chronicle should
be inspired by a clear notion of what a good lexical history of Sanskrit
would be like. One may even
go a step further at the practical level, and say that at least some
of the entries in it should be preliminary sketches of properly historical
entries. More modestly one may require of this lexical
chronicle a historical sophistication.
(1) A historian’s
sense of evidence, scholarly conscience, and sense of fairplay to
his reader will lead him not only to the careful weighing of evidence
and sifting of the well-established fact from the strong conjecture
and the open question but also to put the reader in possession of
the evidence and the theories rejected. A historical dictionary – whether of the chronicle
variety or of the history one – will do well to be liberally sprinkled
with question marks. It will
separate the editor’s conclusions and guesses from the evidence and
give a “fair” sample of the evidence and the contending theories –
so as to permit the reader to draw his own conclusions.
Where the editor is doubtful about the relevance of a piece
of information he will include it with the label “potentially relevant”.
A clear distinction should be established between primary citations
– our level (a) – and secondary meta-citations – those at level (b)
or (c) or bordering on either.
(3) Where absolute chronology cannot be maintained,
attempts should be made to evolve techniques for establishing and
presenting to the reader a relative chronology – at least a relative
slabwise chronology comparable to the geological stratum-labels like
Upper and Lower Jurassic.
(4) One must not
mistake the merely marginal or ad hoc (and therefore attractively
curious or exotic) for the central core feature (and therefore often
dully quotidian or run-of-the-mill).
A felicitous literary departure is grist to the linguistic
historian’s mill chiefly because it may set a new norm for later writers
to follow. A sampling of mediocre
writers is as much needed as taking the unavoidable classics.
A historical dictionary is not a chrestomathy of literary jewels,
though of course it certainly need not deliberately avoid the jewels. A locus classicus is important because it is very often a pace-setting
event. The other pitfall is
that the commendable desire to be exhaustive may rob the editor (and
therefore eventually the user of the dictionary) of the proper perspective.
It is more useful to give a larger number of citations exhausting
all the occasional or short-lived uses of that vocable. Marginal uses
include instances of words coined ad hoc only to be discarded later
and also insertions from another language not as a borrowing but only
as a citation or a temporary switch from the main languages. While offering explicit statistical information
or even intelligent guesses about the shifting ‘popularity’ of a word
or a word-sense from period to period may not be feasible, a historical
dictionary may find it possible to imply such information “in a rough
and ready way, - by the proportions of references given under the
different heads” (Aitken 1971).
(5) The true historian
is not the one who is lost to the present, but rather the one who
has a lively sense of (a) the past that lives in the present and (b)
contemporary history that is being made around him.
It should not sound incongruous if I urge the prospective lexicographer
of Sanskrit to look at contemporary chronicling of semantic change,
the record of the lexical stock market (vocable A displacing vocable
B in sense X after a tough competition), and the birth and death register
of words – as seen, for example, in American Speech or Vie et langage.
(A convenient sampling from British English is available in
Bhide 1948, 1970.) OIA vocables must have been subject to the same
forces of weathering, sedimentation, and volcanic displacement to
which NIA vocables are subject. In
talking about volcanic displacement I have alluded to the possibility
of major discontinuities of usage.
Only an unawareness of this fact and of the earlier-mentioned
distinction between central and marginal can explain but not justify
the blitheness with which post-Dravidic or neo-Sanskrit innovations
are often “supported” by Sanskrit scholars by authentic but irrelevant
citations and, worse still, metacitations. (The justification, if indeed it can be called such, is very often
the Brahmanical reluctance to admit that the well of Sanskrit has
been defiled by non-Indo-Aryan borrowings or substratum interferences.)
The insights gained by an awareness of the living past and
contemporary history and their application through historical imagination
to the traditional scholarly exposure (what is known as “being steeped
in Sanskrit”) should be made available to the team of editors and
eventually to the readers by being codified in a series of notes on
recurring features, problems, cultural domains, and processes.
Thus there could be a series of alphabetized entries entitled,
say, astronomy and astrology, ayurveda, compounds, easternisms, euphemisms,
gatis and other idioms, kosa literature, literary expressions, nyayas,
plant names, Prakritisms, proverbs, verb paradigm, and so on. The entry on compounds will, among other things, spell out the editorial
decisions as to what compounds to include as part of the working capital
of the language (Sprachgut) and what to exclude as ad hoc nonce-formations
that, like most phrase and clause level formations, constitute merely
the linguistic turnover. The
entry on euphemisms will tell us, for example, what the euphemizable
items are, what the characteristic modes are, whether there have been
any major changes, and so forth.
(One may thus report that English ladies stopped sweating in
the 19th century or that Marathi speakers hide behind English
borrowings in talking about unpleasant things like widowhood, impotence,
or even death.) The drafting
of such Guide Entries at an early stage in the editing is a desideratum.
They will set up a comprehensive code of mutually consistent
basic editorial decisions so that the specific decisions in a given
entry will not be ad hoc decisions but applications of certain principles.
Of course these principles will also spell out how much ground
is left for the exercise of editorial discretion. It is only fair
that the reader should be given a glimpse of the editorial kitchen.
(A selection of these has a claim for being included even a
Finally, one may turn to the vexed question of the arrangement
of the various uses of a vocable within the entry in a dictionary.
On the face of it, the answer seems to be fairly straightforward:
in a descriptive, synchronic dictionary the arrangement should be
in an order of descending frequency as seen in the attestations of
the period under consideration; in a historical, diachronic dictionary
the arrangement should be in an order of increasing recency of the
first attestation of that sense. If this apparently straightforward
answer does not work, it is not merely on account of practical exigencies
like the reader’s convenience and the printer’s incompetence (this
last factor cannot certainly be ignored in an underdeveloped country).
The reasons are deeper reasons.
(1) The synchronic order is certainly not liner – sense 1, sense
2, etc., in a mechanical order of frequency will be a travesty of
how the word behaves actually. The
uses at any given stage in the history are best thought of as a multiply
spreading and branching network.
Ideally, a descriptive lexical entry will open with a diagram
consisting of a set of variously interconnected nodes.
Since the order of printing is linear, the subdivisions will
be arranged and numbered in a quasi-linear fashion : to take a sample
1, la, lal, 2a, 2b, 2bl, 2b2,
This is an alphanumeric notation which alternates figures and
letters and which selects one node (numbered 1) as the “unmarked”
or neutral meaning – often called Grundbedeutung.
As a rule but by no means invariably this basic meaning is
the most frequent, the least specialized, the most general, the lest
figurative, the historically earliest, etc.
The critical test is whether the meaning is the one which will
most readily suggest itself to the user if the vocable is mentioned
out of context rather than used within a context.
In relation to a synchronic account of a language that is no
longer current, one may want to recast this test slightly: the basic
meaning is the one that an editor interpreting and annotating a text
is most disposed to pick up so long as there are no clear contextual
pointers in a different direction.
In the rare cases where a vocable may have more than one equally
viable neutral senses (for example, “rank, class”, “sequence, arrangement”,
and “mandate” in the case of the word order in English), one
can start with a summary paragraph of main senses 1, 2, 3 the order
of the subsequent paragraphs being based on some non-synchronic extrinsic
criterion (Consider the Concise Oxford Dictionary entry for
the word order). Note that a given sense under one vocable may
bear a relation of perfect synonymy with a given sense under another
vocable. The lack of perfect
synonymy normally refers to the vocables in their total ranges. An incidental advantage of this notation is
that it permits us to make use of the anuv¤tti principle – everything
said under 1 is to be carried forward to 1a, 1a1; that under 1a to
1a1; and so on – unless the contrary is stated.
The diachronic order is also
more complicated. Suppose
there are five main historical stages (to be denoted by Roman numerals)
and six senses (to be denoted by capital letters). The situation may
well look like this in a chart that may be placed at the beginning
of the entry.
Note the eclipse of Sense B in stage IV and its revival in
stage V. It is not enough to hunt for the earliest citations
far a given sense; one must hunt for other ‘critical’ citations- the
latest, the earliest revival, the latest before a gap, the transitional
citation linking two senses, etc. Again, while the order between C,
D or between C, E can be rationalized under the chronology principle,
that between D, E cannot. The actual dates or first available attestations
are seldom the dates of first usage except in the special case of
the innovative locus classicus. The
same goes for the dates of the last attestations or of gaps in attestations,
Resort will have to be taken, therefore, in such a tie (as the one
between D, E) to some non-diachronic, extrinsic criterion.
Even assuming that the difficulties mentioned under (2) could be resolved
and assuming that the dated record is ample enough to permit such
clear periodwise assignment of the various senses, the fact remains
that in a fundamental sense the sample chart may be a travesty of
history. A true picture will rather look, say, like
1, 1a, 2, 2a
1, 2, 2a, 3
2, 2a, 3, 3a, 3b
2, 2a, 2b, 3, 3a, 3b, 3b1
The loss of sense 1 after stage III may thus render senses
2, 3 disjoint. Also, there may be reorderings. Sense 2b may change its “loyalty” and get attached
to 3; what is merely a minor ‘shade’ or ‘slant’ under Sense 1aat one
stage may become a distinct Sense 1a1 at the next stage; Sense 2 may
get “promoted” to the status of being Sense 1, while the erstwhile
Sense 1 may become marginal; what are Senses 1 and 2 at one stage
may have been etymologically quite different vocables or at least
etymological doublets at an earlier stage; and so on.
In essence, the diachronic picture has to incorporate the successive
synchronic pictures like the stills in a film.
The distinction between chronicle and history is based, to
some extent, on the distinction between the documentation of the past
and its reconstruction. Since documentation and reconstruction feed
on each other and since a good chronicle is always struggling towards
being a good history, the two perspectives, synchronic and diachronic,
cannot be kept apart. A good
diachronic account calls for a prior sound synchronic analysis
of each of the stages. Also, the diachronic order and the synchronic
order of the latest stage may often resemble each other. Thus, the sample alphanumeric set given under
(1) above may reflect the diachronic order.
1, 2, 3
1, 1A, 2, 2a, 2b, 3
1, 1a1, 2a, 2b, 2b1, 2b2, 3
This suggests that an arrangement of this entry in the following
manner will not be too difficult to read or too distorting. Thus, the earliest sense is also the unmarked, basic sense.
It is true that this cannot accommodate reorderings of the
kind described earlier. But
this need not seriously disturb us if we realize that such reorderings
are not very common; that, where they do occur, supplementary visual
aid can be provided in the entry; and that, for a dictionary that
is a lexical chronicle only struggling to prefigure a future lexical
history, all this is unrealistically ambitious in any case.
We must not of course let our preoccupation with citations
and their arrangement make us lose sight of the simple fact that a
historical dictionary is also a dictionary and as such faces all the
problems that a synchronic descriptive dictionary faces.
If it doesn’t face them, it means it is not doing its whole
job. (I have elsewhere (Kelkar
1970-1) discussed some of the general problems faced by and dictionary
whether historical or not and whether bilingual or not and whether
bilingual or not.) The tendency
to look upon the explanations of meaning in a historical dictionary
merely as convenient expendable tags for each of the citation subsets
in an entry is dangerous. (Cf. Aitken 1971: In getting out the defining
characteristics of the genre of which OED is a paradigm example. Aitken frankly states: “The definitions and
descriptive notes, which are also a normal feature of such dictionaries,
may be regarded as fulfilling a somewhat secondary purpose, that of
sign posts or labels to the particular subset of quotations which
follows.”) The same goes for the neglect of idioms and
collocations. After all the
reader has as much right to the earliest citation for the idiom as
such as to that for the individual words in the idiom. The reader
of a historical dictionary of Bangla would want to know since when
Bangla speakers started “eating” water and cigarettes.
Finally, the dictionary tends to take upon itself the duties
of a word finder (or thesaurus) and a cultural encyclopaedia and the
historical dictionary need be no exception. A historical dictionary
need be no exception. A historical
dictionary of English should not only trace the trajectories of the
two words hound and dog independently but also their
intersection – dog displacing hound as the least specialized
word for Canis familiaris. This
could be achieved by appending a word-finder-type synonymy section
to the entry for dog. Similarly
the entries for teacher or nurse will not only record
the first attestation and increasing use of she in relation
to these nouns but also connect this with women’s emancipation.
The danger, to my mind, is not the non-realization of the ambition,
but the pedestrian shirking of an ambition.
In other words our modesty may land us with a bad chronicle.
I am aware that this paper is short on live examples and long
on simulated schematic examples illustrating an ambitious theory. But a non-Sanskritist can justify his presence in a gathering of
distinguished-Sanskritists discussing a Sanskrit historical dictionary
only by venturing to tread on a ground that the angles may want to
keep away from.
I. The Anatomy of a Historical Dictionary
Guide to the scanning of an entry
Guide consisting of alphabetized entries
Skeleton linguistic analysis-rules for proceeding from phonological
spelling to phonetic spelling(s); from grammatical analysis into morphemes
to phonological spelling; from grammatical label to privileges of
occurrence; from orthographic spelling to phonologic and phonetic
spelling; from transliteration to original orthography; from base
to productive derivatives
Body of the dictionary
Appendix of alphabetized entries of borderline cases like proper names,
bound affixes and bases, abbreviations.
The Anatomy of a Historical Dictionary Entry
Lemma: transliteration, original script spelling, phonological spelling,
phonetic spelling (or any subset of these in terms of which the rest
can be predicted)
Alternant spelling (of various sorts)
Grammatical function-class label
Etymology: ancestral, co-successor, successor forms
(taking you beyond the confines of the language concerned
Grammatical structure: simple or complex?, if complex the constituents
and the structure – type label operating entirely within the language
Accompaniments: inflectional paradigm, syntactic selections, idiomatic
collocations, concord and government label
Explanations: descriptions, glosses, cultural notes
Citations (critical and other) and metacitations
Word-finder section: comparables (synonyms, antonyms, hyponyms, hypernyms,
words confused) and derivatives (affixal, reduplicative, compositional)
To be inserted at appropriate points in the above arrangement are
distinctive alphanumeric tags ( 1, 1, 1a, etc. ), style
labels, social labels, region labels, period labels
In Sanskrit the traditional grammatical analyses (in terms of U¸ādi
sūtras, the man grammatical sūtras, etc.) will
be cited under (e), whether valid or not (if not valid, to be introduced
by “pace”). The traditional
etymologies will be cited under (d) if valid and as metacitations
under (h) if invalid. The annotations in commentaries will be cited
as metacitations under (h). The
loose term derivation spans (d) and (e) above.
A.J. 1971. Historical dictionaries
and the computer. In : The Computer in literary and linguistic
research … a Cambridge symposium. London: Cambdrige University,
Oven, 1971, p.3-17.
H.S. 1948. A Study in the
development of the English vocabulary.
U. of Bombay, Ph.D. diss. Unpublished.
-- 1970. Lexicographical
notes on English. Indian Linguistics
31, 162-73. Based on Bhide 1948.
Ashok R. 1970-71. The Anatomy
of a dictionary entry with samples proposed for a Marathi-English
dictionary, Indian linguistics 29 (1968). 143-9; 30(1969). 50-64.
William. 1971. Methodology.
In: DINGWALL, William Orr (ed.). A Survey of linguistics science.
College Park, Maryland: Linguistic Program. U. of Maryland. Pp. 412-91.
and slightly enlarging this paper from the version read at the Seminar,
I benefited from the discussions. This was published in: A.M. Ghatge
et al. ed. Studies in historical Sanskrit lexicography, Deccan
College, 1973, p.57-69.
The paper was presented at a Seminar on Historical Sanskrit
Dictionary at Deccan College, Pune, December, 1972. subsequently I
came across Aitken 1971 which confirmed some of my hunches.