In brief

NSM Semantics in brief

The basic idea of the NSM approach is that we should try to describe complex meanings in terms of simpler ones. For example, to state the meaning of a semantically complex word we should try to give a paraphrase composed of words which are simpler and easier to understand than the original. This method of semantic analysis is called "reductive paraphrase" and the products are called "explications". Reductive paraphrase prevents us from getting tangled up in circular and obscure definitions, problems which greatly hamper conventional dictionaries and other approaches to linguistic semantics. No technical terms, neologisms, logical symbols, or abbreviations are allowed in reductive paraphrase explications– only plain words from ordinary natural language.

Assuming that reductive paraphrase can be made to work as a method of analysing meanings, it follows that every language has an irreducible "semantic core" which would be left after all the complex expressions had been dealt with. This semantic core must have a language-like structure, with a lexicon of indefinable expressions ("semantic primes") and a grammar, i.e. some principles governing how the lexical elements can be combined. The semantic primes and their principles of combination constitute a kind of "mini-language" with the same expressive power as a full natural language.

If languages all have irreducible semantic cores, how do we find them? By semantic experimentation; i.e. by attempting to actually explicate meanings of many different kinds from many different languages, aiming always to reduce the terms of the explications to the smallest possible set. This is exactly what Anna Wierzbicka and colleagues have been doing over a period of more than thirty years. The set of 64 semantic primes (Goddard and Wierzbicka Eds 2002, Goddard Ed 2008) is the fruit of that program of research, which, it should be stressed, is not yet regarded as complete.

When Wierzbicka and colleagues claim that DO, BECAUSE, and GOOD, for example, are semantic primes, they are claiming (i) that these words are essential for explicating the meanings of numerous other words and grammatical constructions, and (ii) that they cannot themselves be explicated in a non-circular fashion, i.e. that their meanings are conceptually simple. The same applies to other examples of semantic primes such as: I, YOU, SOMEONE, SOMETHING, THIS, HAPPEN, MOVE, KNOW, THINK, WANT, SAY, WHERE, WHEN, NOT, MAYBE, LIKE, KIND OF, PART OF. Notice that all these terms identify intuitively intelligible meanings which are grounded in ordinary linguistic experience.

Although the primes just mentioned have just been expressed using English words, NSM researchers believe that they are semantic fundamentals in all languages and therefore that there will be equivalent expressions ("exponents") in all languages. A growing body of research, begun in Goddard and Wierzbicka (eds 1994) and continued in Goddard and Wierzbicka (eds 2002), Goddard (ed. 2008), and range of other publications, suggests that this is indeed the case. NSM studies have been carried out in a wide range of a languages, including English, Russian, French, Spanish, Polish, Danish, Italian, Ewe, Amharic, Malay, Japanese, Chinese, Korean, Mbula (PNG), East Cree, Yankunytjatjara, Arrernte, and Maori, among others.

Finally, a few more words about the conceptual status of semantic primes. In Anna Wierzbicka's writings the terms 'semantic' and 'conceptual' are used more or less interchangeably, the idea being that semantic primes represent elements of linguistic conceptualisation, i.e. elements out of which complex linguistic concepts are built. Some critics find this usage (i.e. semantic = conceptual) objectionable and insist that independent psycholinguistic evidence is required before one can make any conceptual deductions from purely semantic analysis. Wierzbicka's view (like that of Leibniz before her) is that semantic analysis is by its nature a conceptual inquiry, because meanings are not external or objective entities, but, so to speak, creatures of the mind. This is no mere terminological issue, but it would take too long to go into this issue in any depth here.

Semantic Primes

The NSM model has changed a lot since it was first advanced in the early 1970s. In Anna Wierzbicka's 1972 book Semantic Primitives, only 14 semantic primitives were proposed and in her 1980 book Lingua Mentalis, the inventory was not much bigger. Over the 1980s and 1990s, however, the number of proposed primes was expanded greatly, and has now reached a total of 64. The same period also saw the development of some important new ideas about the syntax of the semantic metalanguage. The current proposed primes can be presented, using their English exponents, in the Table below. Perhaps not surprisingly, the inventory of primes looks like a natural language in miniature. As with any language, the "mini-language" of semantic primes has a grammar, as well as a lexicon. This is described in brief below.

Five important points about identifying semantic primes are:

1. A mere list is not sufficient, in itself, to identify the intended meanings, if only because many of these English exponents are polysemous (i.e. have several meanings), but only one sense of each is proposed as primitive. While it is claimed that the simplest sense of the exponent words can be matched across languages (i.e. that they are "lexical universals"), it is recognised that their secondary, polysemic meanings may differ widely from language to language. A fuller characterisation indicates for each proposed prime a set of "canonical contexts" in which it can occur; that is, a set of sentences or sentence fragments exemplifying its allowable grammatical contexts.
2. When we say that a semantic prime ought to be a lexical universal, the term "lexical" is being used in a broad sense. An exponent of a semantic prime may be a phraseme or a bound morpheme, just so long as it expresses the requisite meaning. For example, in English the prime A LONG TIME is expressed by a phraseme, though in many languages the same meaning is conveyed by a single word. In many Australian languages the prime BECAUSE is expressed by a suffix.
3. Even when semantic primes take the form of single words, there is no need for them to be morphologically simple. For example, in English the words SOMEONE and INSIDE are morphologically complex, but their meanings are not composed from the meanings of the morphological "bits" in question. That is, the meaning SOMEONE does not equal "some + one"; the meaning INSIDE does not equal "in + side". In meaning terms, SOMEONE and INSIDE are indivisible.
4. Exponents of semantic primes can have language-specific variant forms (allolexes or allomorphs, indicated by ~ in the table above). For example, in English the word 'else' is an alloex of OTHER; likewise, the word 'thing' functions as an allolex of SOMETHING when it is combined with a determiner or quantifier (i.e. this something = this thing, one something = one thing).
5. Exponents of semantic primes may have different morphosyntactic characteristics, and hence belong to different "parts of speech", in different languages, without this necessarily disturbing their essential combinatorial properties.

All these factors mean that testing the cross-linguistic viability of the semantic primes is no straightforward matter. It requires rich and reliable data, and careful language-internal analysis of polysemy, allolexy, etc. Though cross-linguistic testing of this kind is still in progress, the prospectus seems promising. The existence and lexicalised status of semantic primes has been confirmed for more than 30 languages of widely different linguistic types and from widely different cultural settings. There are a couple of difficult and/or contested cases, but the great balance of evidence favours the hypothesis that semantic primes are both universal and have lexical exponents in all languages.

An NSM Sketch Grammar

Universal semantic primes have an inherent grammar—a "conceptual grammar"—which is the same in all languages; that is, each semantic prime has certain combinatorial properties virtue of the particular concept it represents. The formal realisations (marking patterns, word order, constituent structure, etc.) may differ from language to language without these underlying combinatorial properties being disturbed. Although groups of primes share particular properties and can be regarded as falling into natural classes, virtually every prime has some idiosyncratic properties, giving each one a distinctive syntactic signature. The cross-linguistic viability of the current model of NSM grammar has been checked in considerable depth and detail across a range of typologically divergent languages. Key references are Goddard and Wierzbicka (eds., 2002) and Goddard (ed., 2008). Some of the main aspects of the NSM grammar can be sketched as follows.

Substantives and substantive phrases

The elements which combine directly with substantives include: (a) determiners, e.g., 'this someone', 'the same thing', 'somewhere else', (b) quantifiers, e.g., 'two things', 'many people', 'one part', 'many kinds', (c) evaluators, e.g., 'something good', 'something bad', and (d) descriptors, e.g., 'something small'. They can also be modified by like-phases, e.g., 'people like this', 'someone like me'. There is much more to the grammar of substantive phrases than can be covered here; for example, the relational substantives KIND and PART each can head phrases sui generis, which can be termed "classifier phrases" and "part-hood phrases", respectively, such as 'one thing of this kind' and 'one part of this thing'. Some of the quantifier primes have a "subset" valency optio, e.g. 'one of these people/things', 'two of these people/things', 'some of these people/things'.

Predicates, valency options and complements

We can think of a simple NSM clause as consisting of a predicate, such as HAPPEN, DO, SAY or WANT, together with one or more substantive phrases whose nature is constrained by the identity of the predicate. In addition to a minimal frame, predicates typically allow extended frames in which additional arguments identify or fill out the aspects of the situation implied by the nature of the predicate. These optional extras are termed "valency options". For example, HAPPEN allows us to speak not only of 'something happening', but also of 'something happening to someone' or 'something happening to something'. Likewise, with DO it is possible to add an additional argument and speak of 'doing something to someone' or 'doing something to something'. This frame can be further extended to speak of 'doing something to something with something' (an "instrument" option). Another option for DO is the "comitative" option, as when we speak of 'doing something with someone'. This means that all the following arrays are licensed by NSM sytnax. Furthermore, all predicates can combine with the "meta-predicates" or "predicate operators" NOT and CAN.

• something HAPPENS - [minimal frame]
• something HAPPENS to someone/something - [undergoer frame]
• someone DOES something - [minimal frame]
• someone DOES something to someone/something - [patient frames]
• someone DOES something with something - [instrument frame]
• someone DOES something with someone - [comitative frame]

In some cases, NSM researchers propose valency options which are seldom recognised in mainstream grammars. For example:

• someone THINKS about someone/something - [topic frame]
• someone THINKS something (good/bad) about someone/something - [topic + complement frame]
• someone THINKS like this: " – – " - [quasi-quotational frame]
• (at this time) someone THINKS that [ —— ]S - [propositional complement frame]
• someone SAID something - [minimal frame]
• someone SAID: " – – – " - [direct speech frame]
• someone SAID something to someone - [addressee frame]
• someone SAID something about something/someone else - [locutionary topic frame]
• someone SAID (some) words to someone else - [directed words frame]
• someone SAID something (not) with words - [verbal means frame]
• I SAY: – – – - [performative frame]

Complex sentences

Most obviously, there is an inherently bi-clausal construction connected with the semantic prime IF, exemplified in sentences such as: 'If you do this, people can think something bad about you'. The primes KNOW and THINK have the capacity to take propositional complements. For example, 'I know that something bad happened to this someone'. WANT is also complement-taking, but it has special characteristics that set it apart from the other complement-taking predicates.

Several other complex sentence structures are allowable in the natural semantic metalanguage. These include (i) "quasi-relative clauses" headed by PLACE or TIME; for example, 'I am far from the place where I live', (ii) adverbial clauses of time and reason, e.g., 'when something like this happens, ...', 'it happened like this because someone wanted it', (iii) "analogy clauses" with LIKE~AS as a linking element, e.g., 'I want to do something bad to this someone, like this someone did something bad to me'.

All these combinatorial possibilities mean that the phrasing of NSM text can be quite complex.

Chart of NSM Semantic Primes

This chart, first released in a preliminary version 2010, displays the English exponents of the primes in a layout that groups syntactically similar elements together. The entry for each prime also includes some of its basic combinatorial properties.

• Chart of NSM semantic primes (The chart is designed to be printed on A3 size paper.)

Semantic Molecules

Semantic primes can be thought of as simple and irreducible "atoms" of meaning. But though the whole enterprise of NSM semantics rests ultimately on semantic primes, NSM researchers recognise that some explications necessarily incorporate certain complex semantic units, termed "semantic molecules". These are non-primitive meanings (hence, ultimately decomposable into semantic primes) that function as units in the semantic structure of other, yet more complex words. The notion is similar to that of "intermediate-level concepts" in the Moscow School of Semantics, but with the additional constraint that semantic molecules must be meanings of lexical units in the language concerned.

From a conceptual point of view, the NSM claim is that some complex concepts are semantically dependent on other less complex, but still non-primitive, concepts. For example, semantic explications for words like 'sparrow' and 'eagle' include 'bird' as a semantic molecule. The cognitive claim is that the concept of 'sparrow' includes and depends on the concept of 'bird'. In this case, the relationship is taxonomic: 'sparrows' and 'eagles' are both 'birds [m] of one kind' (molecules are marked in explications with the notation [m]). Many semantic molecules are taxonomic superordinates, but semantic molecules can enter into semantic structures in many other ways. For example, explications for 'walk' and 'run' include 'feet [m]' and 'ground [m]', where the relationships are, roughly, body-part instrumental and location, respectively. Explications for 'cut' and 'chop' include 'sharp [m]', where the molecule characterises the nature of a physical instrument. Explications for 'fork', 'spoon' and 'plate' include 'eat [m]', where the relationship relates to intended purpose or function.

It has to be stressed that whether or not a given word-meaning is a semantic molecule is an empirical issue, it is not arbitrary or a matter of convenience. The sole criterion is semantic necessity, and this can only be determined by detailed semantic analysis.

Many semantic molecules are language-specific, but it appears that a limited number, perhaps 20 or so, may be universal or near-universal. These include some body-part words, such as 'hands', some environmental terms, such as 'sky' and 'ground', and some basic social categories, such as 'children', 'women', and 'men'. Semantic molecules be "nested", enabling a great compression of semantic complexity. Research into semantic molecules is still at a relatively early stage, with many interesting research questions as yet unanswered.

it is worth adding that sometimes NSM explications make use of complex terms of a different kind; namely, derivational bases. For example, the explication for a derived word like 'illness' includes the base element 'ill [d]', where the notation [d] marks its status as a derivational base. The secondary or extended senses of a lexeme sometimes incorporate its primary sense, and, when this occurs, the base meaning is also marked with [d]. To clarify the difference between semantic molecules and derivational bases, one can say that derivational bases have a highly localised and specialised provenance, i.e. each one is confined to a handful of semantically related words. Semantic molecules, on the other hand, range widely across the lexicon. It can also be noted that the concept of semantic molecules is an NSM innovation, whereas the idea that derived forms are semantically built on the meanings of their base forms is commonly accepted in linguistics.