Skip to content

THE UNIVERSALS ARCHIVE:

THE UNIVERSALS ARCHIVE:
A Brief Introduction for Prospective Users Frans Plank and Elena Filimonova
(Universität Konstanz)

[Originally published in Sprachtypologie und Universalienforschung 53 (2000) 109-123; updated by Frans Plank and Thomas Mayer, Aug. 2006]

Abstract

THE UNIVERSALS ARCHIVE aims to be a comprehensive documentation of the linguistic universals on record. Currently including over 2,000 entries, it has for some time now been available on the internet as a searchable database. For its updating and overhaul the cooperation of the typological community is solicited. Here we outline what kind of information is held in the archive and how it can be accessed and used.

1. The Remit

Typology is about how languages, or rather their grammars and lexicons, can and cannot differ from each other, and universals are the laws regulating crosslinguistic variation. Although the typological research programme has been under way for several centuries now, there is regrettably little awareness of just how many such putative laws have been accumulating. Often what comes to mind, even that of many a practicing typologist, is little more than (a subset of) the 40 or so universals suggested in GREENBERG (1963) and conveniently collected in the appendix of this modern classic. Perhaps, if universals had been named after their discoverer as laws are in other sciences (even when they are so unlawlike as Saussure?s Law and other such diachronic regularities, limited in scope to particular families or indeed languages), their impact would have been greater. In actual fact, there are laws galore on record, though few have been officially passed by the typological legislature and no systematic effort has ever been made at proper codification. In view of their significance for the field, these scattered laws, whatever their substance and quality, deserve to be collected and recorded and to be made available for ready consultation. Occasionally, more or less narrowly circumscribed selections of universals have indeed been catalogued, notably in BURQUEST ET AL. (1982), KIRBY (1995), and PLANK (1998)1. Within the national typology programme of the Deutsche Forschungsgemeinschaft, initiated in 1996, a more comprehensive effort in this direction has been made by the project “Sprachbaupläne”, based at the Fachbereich Sprachwissenschaft of the Universität Konstanz2. The remit of this project was to document existing universals, no matter how hypothetical their validity, and not to suggest new ones: a putative universal needed to be in print, or at least be forthcoming, in order to get into THE UNIVERSALS ARCHIVE. All that was added when a universal was being archived were comments of various sorts, contributed by the archivists or again drawn from the literature. Although this collection of published universals was far from complete and the documentation of those already collected sometimes lacked polish, THE UNIVERSALS ARCHIVE was made available as a searchable database on the internet in 1999, as a tool for typologists and as an invitation for its further extension and refinement. What follows is a brief introduction of this archive for the special benefit of those who are accessing it at this address: http://ling.uni-konstanz.de/pages/proj/sprachbau.htm

1. We hesitate also to mention DÉCSY (1987).
2. The project was directed by Frans Plank, with Elena Filimonova as a research associate. We gratefully acknowledge the archiving assistance of Shin-Sook Kim, Arista Da Silva, and Bernhard Metz, as well as technical help from Thilo Dannenmann and Emilia Nagy.

2. The Coverage: Universals

What is being documented in THE UNIVERSALS ARCHIVE are linguistic universals. This notion is neither trivial nor self-explanatory; still, a few remarks will have to suffice here concerning what is and is not covered by the archive. First, although the dividing line is sometimes difficult to draw, to qualify for inclusion in the archive universals ought to be substantive rather than being “design features”, whose universality is definitional (such as vocal-auditory channel, evanescence, duality of patterning; see HOCKETT 1963): lacking them, a communication system, instead of being considered a counterexample, would not be called a natural “language”3. Second, the substantive universals that are entered in the archive are statements preceded by a universal quantifier ranging over all natural languages; and in their basic form, which is not always easy to discern, they are either unconditional (“For all languages, there is or is not p”) or conditional, and in particular implicational, prescribing co-variation of two or more logically independent variables (“For all languages, if there is or is not p, then there will or will not be q”). Thus, to be universal, properties must be universally shared or, if they are not shared by all languages, they must not vary independently across the domain of all languages. This is a somewhat restrictive notion of universality, favouring a conception of universal grammar as the intersection of particular grammars rather than as their union set, this latter set being the grand repository of anything that any language is free to select. Third, universals are often conceded even by their proponents to be mere tendencies rather than categorical laws; such statistical universals are entered in the archive, unlike those regularities explicitly claimed to be limited to areas or families (Euroversals, Bantuversals, and the like). Fourth, laws are timeless; still, it is possible for universals to have a time dimension – as in claims that whenever a language has a particular property, it must have acquired it in a particular way and from a particular source. Fifth, going again for archiving liberality, the properties figuring in the universals documented include all conceivable kinds of units, categories, constructions, rules, constraints, principles, or relationships between them, regardless of how concrete or abstract, theory-neutral or theory-bound. 3. We cannot here broach the issue of whether universals which do not hold for sign languages are to be considered substantive (and invalid) or definitional.

3. Guide Through an Archive Under Construction

To get started on our guided tour, the following universal, one of Moravcsik?s Many Laws, serves to illustrate what kind of information is maximally provided in an entry in the archive. (picture of entry no. 31)

For identification, entries are serially numbered, reflecting the order in which universals have been entered. Universals which are thematically related can be found and collected together by following up cross-references in the Comments section. Of course, when searching for particular domains or keywords, this will define thematic subsets of universals in the first place.

The documentary part of an entry begins with the field Original, where the universal is reproduced approximately as stated in the original source. For various reasons, including readability and limitations of space, it is not always advisable to reproduce the source literally. Also, we translate into English even when source languages are among the more widely read ones (such as Russian). Subsequently the original statement is, if necessary, standardized in style and terminology. Such transformations are always meaning-preserving, and sometimes meaning-enhancing. It is these Standardized versions that are recommended to archive users for all practical purposes. Ideally they ought to be Authorized Versions, but obtaining authorization has so far not always proved viable. (So far as possible, we continue trying.)

We had intended to standardize and indeed formalize universals yet further; but this has largely remained a task for the future. Presently the field Formula is of little utility, and is best ignored by users.

The fields Keywords and Domain classify universals by content at different levels of generality. At the highest level we currently distinguish these domains and domain-combinations:

syntax, morphology (sometimes differentiated as inflection and word formation), phonology (sometimes specified as prosodic phonology), phonetics, semantics, and lexicon. Lower-level keywords are devised as we go along.

A separate inventory of keywords, categorizing universals at lower levels is available for the benefit of on-line searchers of the archive. Further keywords are devised as we go along archiving.

The following three fields Type, Status, and Quality classify universals by kind.

The basic “types” of universals are unconditional (like #1181, a lexical one, for a change) or conditional. The most important conditional relationship is that of material implication (one-way or mutual); but from the way relationships are sometimes named and characterized in the sources (e.g., “p correlates with q”, “p and q are mutually conducive”, “p favours q”, “p authorizes q”) it is not always easy to determine whether this is the intended reading. Here (and perhaps further in the field Comments) attention is also drawn to possibly misleading formulations of universals as implications in the original source. Not infrequently, what is given in the form of an if-then statement is not really a material implication but a kind of conditional relationship which is paraphrasable as “q, provided/given that p”. Implications can be transposed under negation of both implicans and implicatum: “if p then q” is equivalent with “if not q then not p” (e.g., “if there is a trial, then there is a dual” = “if there is no dual, then there is no trial”); if relationships fail the test of transposition, they are not genuine implications, or need some tinkering to reveal their implicational essence. Universals ##12 and 170 are examples. (pictures of entries no. 1181, 12, 170)

Status is about the historical nature of universals. When they are purportedly of timeless validity, proscribing or prescribing structures independent of previous or later states of a language, they are referred to as “achronic”. “Diachronic” universals are laws of change, linking structures of an earlier with those of a later stage of a language and perhaps specifying a mechanism of how to get from one to the other. The problem is that these labels are frequently a matter of interpretation, and ultimately raise profound questions about the relationship between typology and diachrony (see e.g. PLANK 1999, PLANK & SCHELLINGER 2000). Without going into this issue here, we refer back to #31 as an example of a universal that is clearly intended as (and presumably indeed is) achronic. So is #17, while #38 is clearly intended as diachronic; and this pair goes to show how intricately achronic and diachronic universals may be related to each other, with the latter potentially rendering the former redundant. Universal #1287 posits hierarchies which are equally meaningful as timeless crosslinguistic constraints on segment inventories and as steps in the historical spreading within a language of the process in question, with the author of the universal himself wavering between an achronic and a diachronic reading. (pictures of entries no. 17, 38, 1287)

4. Some Advice for Users at a Distance

THE UNIVERSALS ARCHIVE is a FileMaker Pro 4.1 database. This software is easy to deal with and serves the intended purposes reasonably well. In particular, without much ado it enables users of the archive to do and to see at a distance what can be done and seen locally. One crucial difference, however, is that those visiting the archive on the internet cannot alter the database directly; if this is what they desire, they have to prompt the archivists to take local action. And, annoying locals no less than visitors, formatting and special characters do not transmit well on the web. We tend to use Symbol and PalPhon to supplement the standard font (Times) whenever needed, especially in phonological/phonetic and semantic universals, but visitors may not be able to see it (yet, pending Unicode), even if they choose the same fonts on their screens. Profuse apologies. Visitors can access THE UNIVERSALS ARCHIVE in two different ways. They can browse through it or they can search it for anything they hope to be in it. To familiarize themselves with what they can reasonably expect to find, it is recommended that they first browse a bit, scrolling through the myriad universals in chunks of ten. To improve user-friendliness, there will in future be pop-up windows in particular for keywords, the likeliest guides for systematic searches. (To the same end, it is planned to add links from references given in individual universals files to the separate bibliographies for universals sources and for counterexamples and comments.) When browsers turn searchers, they can retrieve universals in terms of any string of symbols or of any combinations of symbol strings which occur in any field of a universals file, from Number to Comments. If they simultaneously want to search for items from more than one field (e.g., Original and Domain), they need to determine whether they should be joined with “AND”or “OR”. As is likewise explained on the web page, searchers then only need to enter the item(s) to be searched for in the relevant right-hand field(s) and click on “Search” below. There are two options for searching, to be determined in the pull-down menus of the left-hand fields: “begins with” looks for any words beginning with the character strings entered (e.g., “case” in the fields Original or Standardized will find you files where “case”, “cases”, and “case system” occur in these fields; “aggl” will find you “agglutination”, “agglutinative”, and “agglutinating”; “num” will find you “number”, “numeral”, and, who knows, “numerology”, “numinous”, “numbskull”), while “is” only looks for words exactly identical with the term you enter. Within one field, any number of words can be entered, separated by blanks. This is a possibility that will be particularly appreciated by typologists, who are after all out to discover co-variation, and for co-variation you need (at least) two variables. For example, if you are curious whether switch-reference has been claimed to co-vary with anything about the order of any kind of elements, simply enter these two words (switch-reference order) in any of the fields “Keywords”, “Original”, or “Standardized” and click on “Search” ? and the archive will return you three files, ##613, 614, and 615. Those who suspect that tone somehow correlates with case (e.g., with the number of tones inversely proportional to that of cases), and type in these two words together, will be pleased (or disappointed) to discover that no one before them has gone on record with such a suspicion, to the best of the archivists? current knowledge. (picture of entries no. 613, 614, 615, 45)

5. Linking Implications

Sometimes a single conditional universal itself involves more than two variables. The best-known instances are claims that a pattern – such as agreement (cf. universal #45), relativization, number distinction, or palatalization (#1287 above) – follows a hierarchy with several positions. Otherwise, however, information about how individual universals link up to form whole chains and networks, thus defining “the great underlying ground-plans of language” (that “tout se tient” is unlikely), is far less convenient to extract from THE UNIVERSALS ARCHIVE. Lacking a suitable programme, it is so far only by hand that local as well as distant users can piece together implications as in the following examples.
Examples
(1) #671 IF basic order is SOV, THEN question particles are almost always sentence-final; IF basic order is not SOV, THEN question particles tend to be sentence-initial. (Ultan 1978) #495 Provided that question particles or affixes are positioned relative to the sentence as a whole, IF they are initial, THEN there are prepositions, and IF they are final, THEN there are postpositions. (Greenberg 1963) thus: SOV => Sent Q => Postp; not SOV (=>) Q Sent => Prep [less secure]; from which follows, since implication is transitive: SOV => Postp; not SOV => Prep [less secure]; which encouragingly fits in with: #491 IF basic order is SOV, THEN with overwhelmingly more than chance frequency there are postpositions. (Greenberg 1963) (2) #615 IF there is switch-reference marking, THEN basic order is mostly verb-final. (Haiman & Munro 1983) #438 IF basic order is verb-initial or free, THEN head-marking is favored, and vice versa. IF basic order is verb-medial or verb-final, THEN dependent- marking is favored, and vice versa. (Nichols 1992) #437 IF there is ergative alignment, THEN there will be dependent-marking, and vice versa. IF there is stative-active or hierarchical alignment, THEN there will be head-marking, and vice versa. (Nichols 1992) thus: switch-reference => verb-final => dependent-marking <=> ergative alignment. (3) #627 IF there is Suffixaufnahme, THEN basic order is verb-initial or verb-final. (Plank 1995) #438 IF basic order is verb-initial or free, THEN head-marking is favored, and vice versa. IF basic order is verb-medial or verb-final, THEN dependent- marking is favored, and vice versa. (Nichols 1992) #896 IF there is head-marking, THEN adjectives tend to be grammatically similar to verbs; IF there is dependent-marking, THEN adjectives tend to be grammatically similar to nouns. (Dixon 1997) #801 IF adjectives are nouny, THEN verbs are tensed; IF adjectives are verby, THEN verbs are not tensed. (Stassen 1997) #1031 Tensedness correlates with OV, Casedness of nouns, AND construction, absolute converbs; Non-Tensedness correlates with VO, Non-Casedness, WITH construction, no absolute converbs. (Stassen 1997) thus: Suffixaufnahme & verb-initial => head-marking => verby adjectives => verbs non-tensed =>nouns non-cased => WITH construction => no absolute converbs OR Suffixaufnahme & verb-final => dependent-marking => nouny adjectives => verbs tensed =>nouns cased => AND construction => absolute converbs; which fits in with: #664 IF converbs are prominent, THEN basic order is OV, and vice versa. (Haspelmath 1995) (4) #635 IF there is Suffixaufnahme, THEN nominal inflection is agglutinative. (Plank 1995) #589 IF morphology is agglutinative, THEN there is no grammatical gender. By transposition: IF there is grammatical gender, THEN morphology is not agglutinative. (Meinhof 1936, Renault 1987, Moravcsik 1994) #247 IF there is an accusative case distinct from a nominative, THEN there is grammatical gender. By transposition: IF there is no grammatical gender, THEN there is no accusative distinct from a nominative. (Palmaitis 1987) #440 IF alignment is non-accusative, THEN verb-initial basic order is favoured. (Nichols 1992) thus: Suffixaufnahme => agglutinative (nominal) morphology => no gender => no nominative-accusative case marking => verb-initial favoured; which fits in with: #627 IF there is Suffixaufnahme, THEN basic order is verb-initial or verb-final. (Plank 1995) #638 IF there is Suffixaufnahme, THEN alignment is ergative or otherwise non-accusative. (Plank 1995)
Often enough, when implications which are entered individually in the archive are connected, and transitivity is employed to link directly variables which are only linked indirectly in such chains, there is no encouragement but cause for concern. However, discovering inconsistencies through such uses of the archive is just as well, since at this stage of the typological enterprise only few universals can rightfully shed the epithet “putative”.

6. Archive Yourselves

THE UNIVERSALS ARCHIVE currently includes over 2,000 entries, and is constantly being overhauled and updated4. Since much relevant literature has so far remained unread by the local archivists, and some has no doubt been misread, the coverage of structural areas is patchy, and what is covered may fall short of perfection. To bring the archive up to scratch we solicit the cooperation of the typological community. We welcome constructive comments of all kinds. And in particular: Do send in, or correct, your own universals and those of your friends and enemies – or at any rate counterexamples to those of the latter! FRANS PLANK & ELENA FILIMONOVA (& THOMAS MAYER) Fachbereich Sprachwissenschaft Universität Konstanz 78457 Konstanz Germany {frans.plank, thomas mayer}@uni-konstanz.de, ef263@cam.ac.uk 4. In the interest of documenting not only the constraints on, but also the possibilities of crosslinguistic diversity, there are also plans to have a separarte department of the archive devoted to the opposite of universals ? namely, rara, rarissima, and nonesuches. See now DAS GRAMMATISCHE RARITĂ„TENKABINETT.

References

  • BURQUEST, DONALD A. et al. (1982). Language universals. Research Papers of the Texas SIL 11.
  • DÉCSY, GYULA (1987). A Select Catalog of Language Universals. Bloomington: Eurolingua. FILIMONOVA, ELENA (1999a). Toward universals of pronominal plural inflection, or Do pronouns always take after nouns?, in: Proceedings of the 34th Colloquium of Linguistics. Bern: Lang.
  • FILIMONOVA, ELENA (1999b). The noun-phrase hierarchy and relational marking: The counter-evidence. Unpublished, Universität Konstanz. [Now in Linguistic Typology 9 (2005) 77-113.]
  • GREENBERG, JOSEPH H. (1963). Some universals of grammar, with particular reference to the order of meaningful elements, in: GREENBERG (ed.) (1963), 73-113. GREENBERG, JOSEPH H. (ed.) (1963). Universals of Language. Cambridge, Mass.: MIT Press. HOCKETT, CHARLES F. (1963). The problem of universals in language, in: GREENBERG (ed.) (1963), 1-29.
  • KIRBY, SIMON (1995). EUROTYP Noun Phrase Universals Archive. http://www.ling.ed.ac.uk/~eurotyp/
  • PLANK, FRANS (1998). The co-variation of phonology with morphology and syntax: A hopeful history. Linguistic Typology 2, 195-230. PLANK, FRANS (1999). Split morphology: How agglutination and flexion mix. Linguistic Typology 3, 279-340.
  • PLANK, FRANS, SCHELLINGER, WOLFGANG (1997). The uneven distribution of genders over numbers: Greenberg Nos. 37 and 45. Linguistic Typology 1, 53-101.
  • PLANK, FRANS, SCHELLINGER, WOLFGANG (2000). Dual laws in (no) time. Sprachtypologie und Universalienforschung 53, 46-52.