The NSM specification An overview of Natural Semantic Metalanguage

16 difference between the two can be found. Establishing semantic equivalence involves a deliberate attempt to find differences. One repeats the process of proposing differences and testing whether the differences actually exist. After exhausting all reasonable possibilities, two forms can be provisionally accepted as equivalent. Probably the most helpful tools for isolating the relevant primitive senses of a word are canonical sentences. A canonical sentence contains a primitive in an elementary syntactic configuration. Often there is sufficient disambiguating context to eliminate incompatible non-primitive senses in polysemous words. Consider the following canonical sentences: I saw someone there. If I do this, people will say bad things about me. I think this is bad, but I don’t know. When that happened, I felt something good. I want you to do it. There are many kinds of things. These two people lived at the same time. Some people can do this – some people can’t. Don’t say anything more about it Good people do not do things like this. I live near here. This thing is above this other thing. John is now on this side of me. There is something inside this. I don’t know, maybe John knows. This thing can move. A very short time after that I heard something. This thing has two parts. He thought that someone like me could do it. These sentences are composed entirely of primitives or near primitives and exemplify the elementary syntax of their respective primitive. They should have exact counterparts in all languages. This use of canonical sentences reflects a new view of primitives in NSM. A primitive is no longer just an isolated lexeme, but a lexeme and its primitive combinatorial syntactic properties. Thus, it is just as legitimate to view canonical sentences as the true primitives of the system, since they exemplify the valid combinations of primitives. At the primitive level, the distinction between lexicon and syntax is blurred and one cannot be separated from the other.

2.6 The NSM specification

In its early days, NSM’s sparse treatment of the primitive syntax W 1980 was the reason for legitimate concern among a few linguists, especially McCawley 1983. While W was well aware of the problem, she believed it premature to speculate about the primitive syntax until the set of primitives itself had matured. The first and long-awaited treatment of the syntax appeared in W 1991c. The syntax is simply a list specifying which primitives may combine with which primitives. At that time, the set had expanded from the original 13 to a set of 37 primitives. Goddard and W’s 1994 ground-breaking study tested this set against 17 languages and paved the way for the current set of 54 primitives W 1996. For The point is that since each word occupies a unique slot within a language’s system of paradigmatic relationships, we should not also expect its counterpart i.e., its semantic equivalent in another language to occupy an analogous slot. In NSM, one says that these words have different ‘resonances’, although they are exact equivalents semantically. 17 the purposes of this study, however, we will be focusing on the 37 primitives of the 1994 volume and their syntax. 1994 PRIMITIVES Substantives I YOU SOMEONE SOMETHING PEOPLE Mental Predicates THINK KNOW WANT FEEL SAY Determiners THIS THE SAME OTHER Quantifiers ONE TWO MANY ALL Actions, Events DO HAPPEN Evaluators GOOD BAD Descriptors BIG SMALL Time WHEN AFTER BEFORE Space WHERE UNDER ABOVE Taxonomy, Partonomy KIND OF PART OF Metapredicates NOT CAN VERY Interclausal Linkers IF BECAUSE LIKE For the convenience of the reader, I have compiled the following phrase structure grammar of NSM based on W’s 1996 treatment. Her treatment is generally explicit, but is not always completely explicit. In several cases, it was necessary to make reasonable educated guesses about W’s underlying intent. In the rules which follow, square brackets disambiguate constituent grouping, parentheses mark optional constituents, italicized words are the SE markers identifying a given valence, and forward slashes separate possible fillers. The categories themselves have been chosen purely for the sake of representing the NSM syntax in an economical fashion. I adopt some category names from the NSM literature and I invent a few. In any case, the theoretical significance of these names is poorly understood at present and I have chosen them primarily for convenience. The numbers in category labels are used simply to differentiate one phrasal level from another. Our first phrase structure rule defines the phrasal category of “modifier”: m1  VERY BIGSMALLGOODBAD There are three “general determiners”: d0  [THISANOTHERTHE SAME] These “general determiners” may combine with other NSM determiners in the following fashion: d1  [[THIS OTHERONETWOALL]THE SAME[OTHER than d0]] m1 d1  ONETWOMANYALL of d0 m1 LIKE also seems to function like a determiner in certain contexts: d1  LIKE IYOU[d0 m1 s1] For LIKE, there is an additional requirement that the head substantive “agree” with the argument to LIKE: 44 someone like you 45 someone like this thing 18 More precisely, the head substantive must be the indefinite “counterpart” of the argument of LIKE. I have divided the substantives into several groups. There is the lexical category of “pure” substantives: s1  SOMEONEPEOPLESOMETHING There is a phrasal category based on the “pure” substantives: s2  [IYOU[d1 PERSON][d1 PEOPLE][d1 THING]THIS] There is a phrasal category based on the “personal” substantives: sp  [IYOU[d1 PERSON][d1 PEOPLE]] There is a stipulation which applies to the two above phrasal categories: the combination ONE PEOPLE must be excluded from their output. Finally, there is a phrasal category based on the “complemental” substantives: sc  VERY GOODBAD THINGTHISTHE SAME The following list shows how “pure” predicates and these substantive phrases may combine to form simple clauses: sp THINK [scabout s2THE SAMEsc about s2THE SAME] sp WANT [sc] sp KNOW [scabout s2THE SAMEsc?MUCH?ALL about s2THE SAME] sp FEEL [scSOMETHING LIKE THIS] sp SAY [sc?MUCH] about s2THE SAME to sp sp DO [sc] to s2 with s2 in d1 PLACE sc HAPPEN to s2 in d1 PLACE Question marks indicate tentatively proposed valences. In addition, the following has been proposed as an alternative frame for the primitive FEEL: sp FEEL [GOODBADTHISLIKE THIS] Substantive phrases may also combine with “attributive” predicates to form a clause: s2 is m1 s2 is VERY GOODBAD [for sp] s2 is a VERY GOODBAD s1 Again, there is a requirement the object must be the indefinite “counterpart” of the subject: 46 you are a good person 47 I am a bad thing 19 Substantive phrases may also combine with “substantive” predicates to form a clause: s2 is d1 PARTKIND of s2 s2 is in d1 PLACE Finally, substantive phrases may also combine with “relator” predicates to form a clause: s2 is THE SAME KIND OF s1 as s2 s2 is UNDERABOVELIKE s2 For these frames, there is also a requirement of non-identity between subject and object. 48 you are above this thing 49 I am like me Temporal adjuncts are required only for the following kinds of clauses: at BEFOREAFTER d0 TIME, SAY-clauseDO-clauseHAPPEN-clause They are optional in all other clauses. There are clausal operators which combine with clauses to yield other clauses: CAN THINK-clauseSAY-clauseDO-clauseHAPPENGOOD-for-clauseBAD-for-clause NOT clause The following predicates take clausal complements, forming complex clauses: sp THINKKNOWWANT clause sp SAY clause to sp Finally, the following primitives conjoin two clauses together into a complex clause: clause LIKE clause IF clause, clause BECAUSE clause, clause Our grammar generates a finite number of possible clauses. Unfortunately, as it stands, it also generates only a finite number of clause combinations. The few primitives which can combine clauses together operate on a finite number of clauses. Fortunately, there are non-lexical means of combining clauses. It is these non-lexical combinatorial operations which gives NSM an infinite generative capacity. Clauses may be linked together by indexical or iconic means, in the Peircean sense. By ‘indexical’, I refer to the use of the primitive THIS in explications. 50 John said something bad. 51 this is not good. THIS maybe used either anaphorically or cataphorically; it may refer to either a substantive or a group of clauses. 20 The spatial arrangement of clauses on the printed page represents iconically the relationships between clauses. Indented lines represent iconically relations of subordination and vertically-aligned contiguous lines represent iconically linear order. Consider the explication of frustration: 52 X feels FRUSTRATION  a. X feels something b. sometimes a person thinks something like this: c. I want to do something d. I can’t do this e. because of this, this person feels something bad f. X feels like this The linear order of the clauses is significant. If, for example, we reversed the order of the two inner components: 53 a. X feels something b. sometimes a person thinks something like this: c. I can’t do this or something d. I want to do something or this e. because of this, this person feels something bad f. X feels like this The resulting explication does not sound quite the same as the original although the meaning components are exactly the same. Iconic relationships between clauses are not so much an empirical statement as they are an observation. In natural language some clauses appear to be related to one another as PEERS whereas some clauses appear to be SUBORDINATED with respect to other clauses. The question is how does one mark this relationship? W p.c. does not believe that this is the sort of relationship which can be marked by lexical means, so she represents these relationships iconically through the spatial arrangement of clauses. What is probably more important is the claim by NSM, that relationships such as this one whatever one calls the relationship can be marked by some linguistic means whether by pauses, intonation, stress, or something else in any language. To rephrase this somewhat, every language should have some way of marking this kind of relationship. Indentation of lines RESEMBLES the relationship of subordination, in that, the indented lines appear to be UNDE R the unindented line and the unindented line appears to COVE R the indented lines. There should be, in every language, some linguistic correlate to indentation and vertically-aligned contiguous lines on the printed page.

2.7 Concerns about NSM theory