You are on page 1of 64

1

An Introduction to Applicative Universal Grammar


Bernard Paul Sypniewski
Rowan University
Copyright Bernard Paul Sypniewski, Woodbine, NJ 08270 1999
Freely distributable for educational purposes
2
Introduction
Sebastian Shaumyan's Applicable Universal Grammar (AUG) has generated considerable
interest but the technical nature of his novel description of language has discouraged many linguists
from investigating AUG in any thorough way. This monograph was written to make it easier for those
curious about AUG to understand this interesting linguistic theory.
The primary source for AUG is A Semiotic Theory of Language Shaumyan (1987); this
monograph also relies on papers written by or co-authored by Professor Shaumyan since the
publication of A Semiotic Theory of Language. Despite the complexities of the text, the book is only
an outline of AUG and its basic notions. Shaumyan currently plans a series of books to expand
on his 1987 work. This monograph does not include the same detail as Shaumyan (1987).
Certain portions of the original text, which, in some cases, are substantial in size, will be touched
upon only lightly. For example, AUG depends on combinatory logic. Combinatory logic is more
advanced than the logic taught in schools. While an understanding of combinatory logic will satisfy
the reader that Shaumyan rests AUG on a sound logical basis, an understanding of combinatory
logic is not necessary for an appreciation of AUG. Shaumyan illustrates his book with many linguistic
examples. I will omit most of the examples in Shaumyan (1987) in this monograph; I urge the
interested reader to refer to the original text for examples beyond those provided here.
I will describe AUG roughly in the order in Shaumyan (1987). Briefly, that order is:
A description of the semiotic theory of language;
Phonology;
Genotype Grammar;
Phenotype Grammar; and
Methodology.
A number of technical expressions
1
, e.g., genotype and phenotype grammar, may be unfamiliar to
the reader, especially in the sense that they are used in AUG. I will define these expressions
when necessary for the understanding of the underlying theory.
Part 1 - An Overview of the Semiotic Theory of Language
AUG is not a theory about the workings of any particular language. AUG explains
what is common to the way that all languages work; hence, the word universal in the name
Applicative Universal Grammar. AUG is a new way of looking at language that is frequently at
odds with established linguistic theories. AUG analyzes the semiotic properties of language
while other systems, e.g., Generative Transformational Grammar, ignore the semiotics of
language. We will see that the semiotic properties of language are not additions to linguistics but
are fundamental to the understanding of the operations inherent in all languages. AUG has two
major components: applicative grammar and a two level phonology. Applicative grammar,
itself, has two major components:
a formal calculus which contains the notions of inherent syntactic functions and the
superposition of functions, and
a series of semiotic laws which constrain the calculus.

1
I will use the word expression for technical words and
phrases. AUG uses the word term in a technical sense. I wish
to avoid the obvious possibility of confusion
3
Taken together, these components explain language and linguistic phenomena that are
difficult to explain or are inexplicable using other linguistic theories. While a semiotic analysis
of language is capable of describing artificial languages and languages which non-humans
(bees, ants, apes, etc.) might be capable of using, AUG is a theory limited to human language.
For AUG, natural language is a culturally determined, mental phenomenon. Shaumyan
(1987) contains a six part semiotic definition of language. Shaumyan and Sypniewski (1995)
contains a somewhat different definition of natural language. Despite the differences in detail,
these two definitions are complementary. The semiotic definition of language in Shaumyan
(1987:2) posits a sign system with the following properties:
two semiotic strata;
sequencing;
use of rules;
structure;
hierarchical stratification; and
semiotic relevance.
Shaumyan and Sypniewski (1995:1) gives the following definition for natural language:
Natural languages are instruments of communication that transpose - each in its own way - an
unanalyzed continuum of human experience into a system of discrete bilateral linguistic units.
Each linguistic unit is a concept linked to a phonic segment - the sign of the concept. The
sign, in turn, is represented by phonemes - discrete acoustic distinctive elements.
In this monograph, I will rely on the original six-part semiotic definition of language, but I will also
describe the notion of linguistic unit mentioned in the latter definition, since it is slightly better
developed in the paper than in the original text.
Two Semiotic Strata
Since AUG is a semiotic theory, the most basic notion in AUG is the notion of the sign. It
is crucial to understand that signs and meanings are not things. Rather, a sign and its
meaning populate a reciprocal relation, referred to as the sign of: meaning of relation, if we
examine the relation from the point of view of the sign, or the meaning of: sign of relation if we
examine it from the viewpoint of the meaning. AUG emphasizes that the relation is more
important to linguistics than the sign entity; AUG specifically denies that there is such a class of
entities that have certain inherent properties that can be called "signs". Shaumyan and
Sypniewski (1995:1) AUG differs from other linguistic theories by emphasizing relations
rather than entities. Confusing sign with sign of will make AUG more difficult to understand. I
cannot overstress how important it is to understand that AUG describes linguistic relations rather
than "things". In many linguistic theories, things like words or phrases have inherent properties.
In AUG, linguistic objects or expressions only exist in relation to other linguistic objects or
expressions. If there is no relation, there is nothing for AUG to concern itself about.
For AUG, the description of sign is different than descriptions of the sign in other semiotic
theories or philosophies. For example, Peirce talks about a sign as an entity and insists that it
be connected with an object which he often defines in tangible terms, see, e.g. Peirce
(1991:240-241) Peirce's notions of the sign (his notions change over the course of his
writings) are general semiotic notions and are rarely intended to have specifically linguistic
uses. Peirce has layers upon layers of signs in his philosophy. In AUG, the notion of the sign is
considerably simpler. The notion of the sign of: meaning of relation in AUG is much closer to
Saussure's notions of signifier and signifie, but is differently developed. Both Shaumyan and
Saussure reject the notion that language is a stock of labels stuck onto different aspects of
reality. The notions of sign in AUG and in Saussure are not identical. For AUG, sign and meaning
are not primitive notions; sign of and meaning of are the primitives. In other words, there are no
4
signs or meanings outside of a bilateral relation. Saussure sees signs as "real objects", while
saying that signs and their relations are what linguists study Saussure (1966:102). The sign of:
meaning of and meaning of: sign of relations are not symmetrical. A sign communicates a
meaning but a meaning does not communicate a sign. Saussure has two descriptions of the
sign: one as a unilateral notion and another as a bilateral unit. AUG eliminates Saussures
bilateral sign as unnecessary and redescribes the notion of the unilateral sign. Shaumyan
(1987:16-17) AUG's notion of the sign must be understood broadly. While many linguistic signs
are sound sequences, AUG does not limit signs to sound sequences. Indeed, the absence of a
sound sequence can be a sign. AUG permits what it calls zero signs. Signs can have any level of
complexity. Due to the central importance of the sign for AUG and the innate complexity of
linguistic signs, the delimitation of linguistic signs is a fundamental problem that presents great
difficulties... Shaumyan (1987:4)
The implications of the sign of: meaning of relation is substantial. Several of these implications
have been developed into principles. One important principle is the Principle of the
Differentiation of Signs:
If two signs are different, they must be differentiated by different sequences of phonemes.
Shaumyan (1987:5)
Phonemes do not necessarily have meanings but all phonemes have a differentiating (diacritic)
function. The Principle of the Differentiation of Signs says that if two signs are different, they
must be distinguished from each other by different phonemic patterns
2
. While signs can be
sequences of phonemes, a phoneme is not a sign. A sign is an entity (understood broadly) that
has meaning, i.e., a sign is an entity that can be part of a sign of: meaning of relation. Strictly
speaking, AUG is not concerned with either signs or meanings in an abstract sense. AUG is
concerned only with sign of: meaning of relations Shaumyan and Sypniewski (1995:1). An
important implication of AUGs concern with the sign of: meaning of relation will be discussed
later in the section on Semiotic Relevance. The reader must understand the importance of the
diacritic nature of the sign. A sign is not just a marker or a flag for a meaning. A sign
distinguishes the meaning to which it is related from all other meanings. AUG does not place any
significance on the degree to which one meaning is different from other meanings. What is
important is that one meaning can be differentiated from other meanings through signs.
Language has two distinct semiotic strata, the system of signs and the system of
diacritics. Phonology studies the diacritic system and grammar studies the sign system. The
notion of two semiotic strata requires AUG to describe two levels of grammar: the universal, or
genotype level, and the language specific, or phenotype level. Genotype grammar is unique to
AUG. The greatest portion of Shaumyan (1987) is taken up with a description of genotype
grammar.
Sequencing
Sequencing, i.e., the order in which signs are produced, is an obvious and fundamentally
important part of language. Sequencing is a powerful method for the construction and
development of new signs, words, and sentences. Sequencing pervades AUG but is not a
separately developed notion.

2
For two phonemes to be different, i.e., for them to be
considered to be two different phonemes, there must be
something about each phoneme that distinguishes it from the
other.
5
Rule Use
AUG sees language as a rule-based sign system. Rules are useful for eliminating or limiting
errors, for example, in the selection of usable sequences of sound. Language is selective and
redundant. No language uses all possible sequences of sounds, letters, or words. The total
number of possible sequences is enormous and beyond our capacity to memorize. We use
rules to help construct acceptable sequences and reject unacceptable ones. Since there are
fewer rules than sequences, it is easier to remember the rules than to remember all of the
acceptable sequences. A grammar is a description of the set of linguistic rules used by a
particular language community.
Different rules may produce different acceptable sequences with identical meanings.
For example, sentences can be either active or passive and still express the same meaning.
The ability to create equivalent but different constructions, called the polytectonic property of
language, is an important, inherent quality of natural language. Artificial languages are
monotechtonic and produce only unique constructions.
Structure
The fact that sentences have structure accounts for language's ability to develop a
potentially infinite number of sentences. The structure of a sentence is a description of the
interrelationship of the parts of the sentence. Sentence structure is the network of syntagmatic
relations between its parts and paradigmatic relations between each part and all other
expressions that can be substituted for it Shaumyan (1987:8). Syntagmatic relations are the
relations between parts of the sentence; paradigmatic relations are relations between words
that can be substituted for each other in a sentence. A complete statement of the structure of a
sentence requires the mention of all its syntagmatic and paradigmatic relations.
Words and sequences of phonemes have a structure similar in many ways to that of
sentences. For example, any two roots that can acceptably share a common suffix or any two
suffixes that can be admissibly attached to a single root are examples of syntagmatic and
paradigmatic relations (respectively) of words and phonemic sequences. Contrast is a word
used to designate syntagmatic relations and opposition is used for the designation of
paradigmatic relations.
Hierarchical Stratification
Language is functionally stratified. AUG identifies a large number of functional strata. Six
functions stand out:
Sounds, which have no linguistic function, embody phonemes and linguistic units;
Phonemes have a differentiational (or diacritical) function;
Morphemes signify root or non-root concepts;
Parts of speech have symbolic functions;
Sentence parts have a syntactic function; and
Sentences are units of communication.
AUG also notes functions like deixis and several communication functions: a representational,
a vocative, and an expressive function for sentences.
Hierarchical stratification is central to AUG. A main concern of AUG is to discover the laws
that describe and control the functional levels of language.
6
Semiotic Relevance
The Principle of Semiotic Relevance is implied by the sign of: meaning of relation in
language. The Principle states that the only semiotically relevant distinctions between meanings
are those which cause a difference in signs, and vice versa. When a language community sees
a difference between two aspects of reality, it will express that difference by using different
signs; we can also say that a language community sees an significant difference between two
aspects of reality if that community refers to those aspects of reality with different signs. The
sign of: meaning of relation is conventional, i.e., it is the product of a society Shaumyan
(1987:11-12); see also Shaumyan and Sypniewski (1995). Two similar signs that do not reflect
different meanings are variants of the same sign; two meanings, though conceptually similar, that
produce different signs are different meanings.
A relevant distinction in one language may not be relevant in another. What may be
important semiotically may not be important extrasemiotically, and vice versa. There is a
tension between the semiotic and extrasemiotic aspects of language that is resolved in a
speech act. A speech act transforms a semiotic aspect of meaning into the extrasemiotic, so
that the transmitted message can be properly understood by the hearer. Shaumyan (1987:12)
Semiotic Relevance is empirically testable. The existence of ambiguity supports the
validity of the Principle of Semiotic Relevance. An ambiguous sentence not only has two
possible meanings but, because of the structure and sequencing of the sentence, the hearer
or reader cannot determine with assurance which meaning is intended. If the Principle of
Semiotic Relevance were invalid, ambiguity would not be recognized because there would be no
way to differentiate meanings
3
. For AUG, ambiguity is caused by the improper correlation of the
distinctions between signs and their meanings in a particular sentence.
Homonymy may be seen as providing a counter-example to the Principle of Semiotic
Relevance. A homonym may be seen as a single sign that represents two or more distinct
meanings. Under the Principle of Semiotic Equivalence, we expect two signs rather than one;
yet, we see what appears to be one sign. AUG sees homonyms as two different signs which
share the same form but which belong to two different sign of: meaning of relations. In
Shaumyan (1987), such a sign is called a polyvalent sign; in Shaumyan and Sypniewski
(1995), it is referred to as a polymorphic sign. AUG rejects the notion of homonymy.
Linguistics as part of Semiotics
AUG claims that linguistics is part of semiotics because natural language is a subset of the
class of all sign systems but the class of all sign systems is broader than the class of all
languages. Linguistics studies the natural sign system that is human language. To function as a
human language, a sign system must be capable of expressing all of the concepts that a
language community needs expressed. This notion, referred to as cognitive flexibility, means that
message transmitted by a sign system can be translated into a human language but, because of
the richness of human language, the converse is not always possible.
As a consequence of cognitive flexibility, language must have a diacritic system to enable
the production of a potentially infinite number of signs. A sign is produced by combining

3
A sign and a meaning form a sign of: meaning of relation.
If two relations are similar, we can be confused. We may not
understand which relation is intended. The fact that we can
distinguish two meanings means that there are two sign of:
meaning of relations. If we could not distinguish two
meanings then there would only be one relation.
7
diacritics; a sign can be seen as nothing more than a combination of diacritics. Cognitive flexibility
makes natural language the type of system known as a "very large system" in cybernetics.
Because of the richness of "very large systems", such systems cannot be completely observed or
controlled
4
. The richness of very large systems is due to the very large number of distinctions
that such a system is capable of making.
If a natural language is seen as a very large system in the cybernetic sense, we
immediately perceive a problem: the explanation of language acquisition by children. AUG
hypothesizes the existence of a simpler sign system that underlies language. AUG names this
simple sign system the linguistic genotype and assumes that linguistic genotypes are common to
all human languages
5
. This is a fundamental notion without which AUG cannot be understood.
Natural languages are embodiments of linguistic genotypes. The functioning of a natural
language simulates the functioning of a linguistic genotype. Linguistic genotypes do not exist
separately from natural languages but are "built into" all natural languages. AUG studies
linguistic genotypes, i.e., the common semiotic properties of natural languages. To put it
another way, AUG sees all human languages as functioning the same way on what AUG refers to
as the genotype level. As we will see, AUG proposes that any human language operates,
simultaneously, on two different levels: the phenotype level, roughly corresponding to the level of
operation described by traditional linguistics, and the universal, genotype level.
Linguistics is frequently seen as a branch of psychology. Psychological principles are not
needed for the understanding of sign systems. While a psychological investigation of language
can provide us with important insights, such an investigation is not necessary for a linguistic
understanding of language. Indeed, any psychological investigation of language presumes
linguistic principles. AUG rejects the claim that linguistics is a branch of psychology.
Some Goals for Linguistic Theory
Despite the substantial differences between languages, all languages have significant
similarities. For example, every language differentiates between the functions of predicates and
terms in a sentence. Language differences are not unlimited. Take word order; because of the
nature of human speech, word order can occur in one of a limited number of patterns. There
are constraints on the forms that natural language can take; an example of a constraint on
language is the aforementioned Principle of Semiotic Relevance.
A linguistic theory must have several goals:
1. the essential semiotic properties of natural language must be defined and their consequences
must be explored;
2. it must account for linguistic facts of languages and equate them to the basic semiotic
properties of all natural language;
3. it must create typologies and grammars for languages; and
4. a linguistic theory must be able to justify its abstractions.

4
There is a similar idea in the growing field of non-linear
dynamics.
5
Shaumyan (1987) does not discuss language acquisition very
much but it is possible to argue that what children (or other
language learners) acquire is the phenotype rather than the
genotype of language.
8
Much of the book consists of tracing the abstractions proposed by AUG which are
based on an examination of the consequences of the semiotic description of language.
Synchronic and Diachronic Linguistics
Synchronic facts are the syntagmatic and paradigmatic oppositions between linguistic
units existing at certain times in the history of a language; diachronic facts are correspondences
between two successive states of a language. AUG recognizes a paradoxical relation between
the synchronic and diachronic facts of a language. Diachronic facts both account for synchronic
facts and are irrelevant to them. Synchronic facts are logically independent of diachronic
facts. Synchrony is both logically prior to diachrony and more fundamental to the study of
language. Synchrony also shows that language has the two semiotic strata that we have
previously seen: phonology and grammar, with the term grammar taken in its broadest sense.
Languages often have dialects which can be divided into four types, two of which are
synchronic and two diachronic. Shaumyan (1987) only considers one type : a monochronic,
monotopic linguistic system in which time and space are excluded.
AUG and Generative Grammar
There are several obvious differences between generative grammar, as proposed by Noam
Chomsky, and AUG. Generative grammar is not a sign system; AUG is. While there are rules in
Chomsky's system, rules are external to language; in AUG, they are an integral part of
language. In AUG, grammar cannot be understood separately from meaning. The sign of:
meaning of relation - the basic linguistic unit - prohibits grammar from being separated from
meaning. Chomsky uncouples grammar and meaning, a basic element of his linguistic ideas.
According to Shaumyan, Chomsky has confused the notions of lexical and grammatical
meaning. Chomsky's famous Colorless green ideas sleep furiously is, for Chomsky,
grammatically correct but meaningless. For AUG, the sentence may be lexically meaningless
but it is grammatically meaningful. Grammatical meanings are categorical and functional
meanings, the most general meanings that a linguistic unit can have. Unless we understand the
grammatical meanings of the linguistic units in a sentence, we cannot even tell whether an
expression is meaningless or not.
Grammatical meanings are morphological and syntactic categories Shaumyan (1987:25)
In Shaumyan and Sypniewski (1995:2-8) the notion of lexical and grammatical meanings
expands beyond that in the book. The distinction between lexical and grammatical meanings
is theory neutral. The Principle of Obligatoriness must be understood before the
notion of grammatical meaning can be understood. There are a closed set of notions in every
language which are needed to express the language users' view of reality; an example is a
grammatical meaning which expresses time, e.g., tense. In some languages, like Russian, time
is expressed grammatically; in Indonesian, the time perspective is expressed by the lexical
meanings of language. The grammatical and lexical meanings of some words are coordinated,
e.g., cat has a lexical meaning which refers to an object and a grammatical meaning - term -
which also refers to an object. The word rotation, on the other hand, has uncoordinated
lexical and grammatical meanings. The grammatical meaning of rotation - term - refers to an
object but its lexical meaning refers to a process or continuing event.
Grammatical meanings can be either inflectional or derivational. Derivational meanings
are structural meanings, e.g. the suffix -er. Inflectional meanings divide into two classes:
functional meanings and specifying meanings. The functional meaning of a morpheme
describes the syntactic function of a word, e.g., predicate, modifier. A specifying meaning is a
meaning belonging to particular types of words, e.g., predicates have specifying meanings for
tense. The notions of grammatical and lexical meanings are not synonymous with syntaxeme
9
and semanteme as used in other linguistic theories. There is another opposition related to
lexical and grammatical meanings: syntactic and semantic meanings. Using this terminology,
semantic meaning covers lexical meanings and all grammatical meanings except strictly
syntactic ones. Many linguists consider syntactic meanings to be merely functions. AUG
disputes this. Syntactic meanings are highly abstract; they denote connections between words.
Generative grammar sees syntax as a form of lexical meaning. AUG considers this to be
a confusion between grammatical and lexical meanings. There is one sense in which the
distinction between semantic and syntactic meanings provides a significant insight. Semantic
meanings are directly connected to the world; syntactic meanings are only indirectly related to
the world. Seeing syntax as the relation between words in different positions in a sentence
and semantics as the relations between words which may fit into a particular place in a
sentence, we can consider syntax to be a set of "horizontal" relations and semantics to be a
set of "vertical" relations.
AUG does not claim that the grammatical and lexical meanings are distinct. Blending
grammatical and lexical meanings forms a heterogeneous whole. Shaumyan and Sypniewski
(1995:6). Grammatical and lexical meanings are part of an opposition grammatical meaning:
lexical meaning, which is frequently ignored in other linguistic theories. Shaumyan and
Sypniewski (1995:7) Grammatical meanings are not directly accessible. Lexical meanings
are the meanings of word stems; grammatical meanings are the meanings of inflectional
morphemes, prepositions, conjunctions, and other devices such as word order. Shaumyan and
Sypniewski (1995:7) Grammatical meanings need not be "words" nor do all words need to
have lexical meanings, as can be seen from words like prepositions. Words like articles and
prepositions must have meanings (at least grammatical meanings) because they are signs, but
they need not have lexical meanings. Shaumyan (1987:25)
To return to the differences between generative grammar and AUG, we note that generative
grammar is a mathematical formalism that promises a set of rules for generating "correct"
sentences and only "correct" sentences for a particular grammar. Even granting the claim that the
rules so generated actually work as the theory requires, the mere fact that a mathematical
formalism creates working rules does not imply that the language which it claims to describe works
like the mathematics it uses or the rules that it generates. AUG aims at an explanation of
linguistic phenomena and not merely a simulation of them. Generative grammar is not intended
to be explanatory. AUG does not reject the use of mathematical or logical tools in the
description of language; an examination of Chapter 3 of Shaumyan (1987) will dispel any such
notion. AUG claims that it is not enough to merely create "correct" sentences. We need to know
why sentences are "correct". In order to be satisfactory, any linguistic theory must not distort or
ignore linguistic facts. AUG claims that Generative Transformational Grammar cannot meet these
goals. Shaumyan (1987:23-31)
Part 2 - Phonology
In AUG, the main goals of phonology are :
the strict separation of functional and physical levels of sounds used in language; and
the study of the interrelationship between the two layers of phonology.
Shaumyan sees a dangerous gap between phonology and experimental phonology;
phonologists often ignore experimental data and experimental phonologists are often uncertain of
the validity of phonologists' theoretical explanations of their data. Shaumyan believes that a study
of the interplay of the functional and physical levels of language sounds is capable of restoring
communication between these two camps. AUG's two phonological levels are at odds with the
phonology of generative grammar. After its origin around 1959, generative phonology replaced a
10
defective multilevel phonology. AUG does not propose a return to the older defective phonology
but rather to rehabilitate some of its more useful ideas and an to advance new ideas.
Phonemes are diacritics that are part of the diacritic of: sign of relation. The initial description of
the phoneme in Shaumyan (1987:32) does not equate a phoneme with a sound but, rather, with
a linguistic entity which distinguishes signs. Although speech can be seen as strings of
segmented phonemes, speech is a actually a continuous flow. The central problem of phonology
is the solution of the problem of speech sound identity. Speech continuity creates an intrinsic
variability in speech sounds. AUG has a principle, called the Principle of Speech Sound
Variability, which states that speech sounds are environmentally modified during speech acts.
Different language groups classify similar sets of sounds in different ways (phonic relativity).
Phonic relativity creates the problem of the identity of speech sounds.
In order to address the problem of speech sound identity, Shaumyan uses the Principle of
Differentiation of Signs which requires that linguistic signs be differentiated by different diacritic
sequences. The position of sounds helps us in our examination. When comparing groups of
sounds, it is important to note the position of phonemes with regard to other phonemes in each
set. Each set of phonemes has two distinctive features: concrete distinctive oppositions and
concrete phonemes. When ordering each set by some positional relation, the relations that order
the sets of sounds in each position are known as the concrete distinctive relations. The terms
(phonemes) in these relations are known as concrete phonemes. Concrete phonemes are not
merely physical sounds; concrete distinctive oppositions semiotically function to distinguish one sign
from other signs. Concrete distinctive oppositions have important properties called concrete
distinctive features. A set of concrete phonemes ordered by a concrete distinctive opposition
is called a paradigmatic class of concrete phonemes. Shaumyan (1987:32-35)
With these new concepts in mind, the problem of speech sound identity can be rephrased as:
What determines the identity of concrete phonemes that occur in different positions? Shaumyan
(1987:35) Because AUG sees phonemes as relational entities, the question can be rephrased as:
What determines the identity of the structure (isomorphism) of paradigmatic classes of concrete
phonemes? Shaumyan (1987:35) AUG postulates the Law of Phonemic Identity:
Two paradigmatic classes of concrete phonemes K
i
and K
j
are identical if their relational
structure can be put into a one-one correspondence, so that to each concrete phoneme x of K
i
there corresponds a concrete phoneme y of K
j
, and to each concrete distinctive opposition r of K
i
there corresponds a distinctive opposition s of K
j
, and vice versa. There is a one-one
correspondence between concrete phonemes x and y, and between concrete distinctive oppositions
r and s, if the difference between x and y and between r and s is reducible solely to the effect of
positional variation. Shaumyan (1987:35)
AUG resolves the problem of speech sound identity by splitting the concept of speech sounds
into two distinct parts: speech sound proper and the concrete phoneme. Speech sounds can be
distinguished by physical analysis; concrete phonemes are distinguished by the Law of
Phonemic Identity. Although concrete phonemes are physical entities, the distinction between
concrete phonemes is not based on their physical characteristics. Shaumyan (1987:36) The Law of
Phonemic Identity is capable of predicting the behavior of phonemes in situations where different
concrete phonemes are identical with respect to their physical properties, and, ... where two
identical concrete phonemes are completely different with respect to their physical properties.
Shaumyan (1987:36-39) The notion of phonetic features becomes split into two segments:
functionally identical phonetic features (concrete distinctive features) and physically identical
phonetic features (phonetic features proper). Classes of functionally identical concrete phonemes
and classes of physically identical sounds are logically independent of each other. Shaumyan
(1987:39)
The following terminology will be used in this section:
11
An abstract phoneme is a class of concrete phonemes.
An abstract distinctive opposition is a class of concrete distinctive oppositions.
An abstract speech sound or sound type is a class of concrete speech sounds.
An abstract phonetic feature is a class of concrete phonetic features.
Phonology has two levels: physical and functional. For convenience, Shaumyan follows the
mathematician Markov in using identifying idealizations. He creates abstract phonemes. Simply put,
an abstract phoneme is the result of a convention which says that if an identical phoneme occurs
more than once in a word, it will be convenient to treat the phonemes as multiple occurrences of
one phoneme rather than as separate phonemes. In titillate, the phoneme t occurs three times
rather than there being three separate phonemes t, for example.
When compared, the definition of phoneme as a diacritic and the definition of phoneme as a
concrete speech sound that functions as a term in distinctive opposition are complementary
definitions. Speech sounds have a dual nature. The functional and physical identities of a speech
sound are complementary but mutually exclusive ideas. AUG refers to the sound/diacritic
combination rather than just the sound as a phoneme. The concepts of sound and phoneme are two
different levels of abstraction and are not related by class inclusion or membership. Shaumyan
draws an analogy to waves and particles in physics to explain the relationship between sounds and
diacritics. He calls this type of relation a "centaur" relation. AUG requires that the functional and
physical levels be strictly distinguished at all times.
Speech Flow Segmentation
Speech sounds segment both physically and functionally; the segments are independent of
each other. Shaumyan cites different opinions about whether linguistic criteria make any
difference in speech sound segmentation. Speakers segment speech into discrete units, a
phenomenon Shaumyan calls the physical segmentation of the speech flow. The aim of AUG's
phonology is to determine the basis of functional segmentation.
The single concrete phoneme is the minimal part of a linguistic sign Shaumyan (1987:43)
but a minimal part of a linguistic sign may consist of several sounds. On the other hand, two sounds
may consist of one or two concrete phonemes. Shaumyan (1987:44) Two considerations enter
into the functional interpretation of two signs. First, a sequence of two sounds can be either
cohesive or noncohesive. If cohesive, the sequence is a single concrete phoneme; if noncohesive,
the sequence is two concrete phonemes. Second, in a sequence of two sounds, if either sound can
be replaced by another sound or by a "zero" sound, then the sequence represents two phonemes;
otherwise, the sequence is a single phoneme. Sometimes a sequence of two phonemes may
appear in a position like a single phoneme that is an element of the sequence. In such cases, the
single phoneme can be seen as the "realization" of the two-phoneme sequence. Shaumyan
(1987:45-46)
A Return to the Problems of Functional Identity
Some sounds cannot acceptably follow certain other sounds while other related sounds can.
Shaumyan uses the English language example of b, d, and g following s. None of these voiced
stops can acceptably follow s in an English word, but the unvoiced stops p, t, and k can. The
question arises: when a sound x cannot appear in position p
1
but a related sound y can appear in
position p
1
, can x be said to be functionally equivalent to y in positions where both can occur? If
we remember that AUG concerns itself with relations rather than "objects", we can anticipate
the answer. Functional equivalence depends on whether x and y can be members of a binary
relation x:y. In the case of the sounds mentioned above, the phonological oppositions p:b. t:d, or g:k
12
do not exist because p, t, and k can appear after s but b, d, and g cannot. P, t, and k after s merge
with s into one phoneme; b, d, and g cannot. Such mergers are called neutralizations; the product
of a neutralization is called an archiphoneme.
Phonological Antinomies
AUG recognized that there are certain antinomies in phonology at least as early as
Shaumyan (1968). Antinomies of the paradigmatic and syntagmatic identification of phonemes
are created by the consequences of two assumptions:
Speech sounds are discrete physical elements.
Speech sounds function as diacritics; they distinguish signs according to the Principle of
Semiotic Relevance.
The first assumption predicts that speech sounds will not be identical but the second
assumption permits identical sounds. The first assumption suggests that two speech sounds
cannot constitute one phoneme, while the second assumption permits just such a merger.
Furthermore, a third theoretical antimony emerges. If we assume that the role of phonemes is to
distinguish signs and we further assume that phonemes are acoustic elements, we get what
Shaumyan calls the antinomy of transposition, which concerns the transposition of types of signs.
The first assumption permits us to substitute one type of sign for another, e.g., a graphic sign for a
acoustic sign. We say that the "medium" of the sign makes no difference to a sign of: meaning of
relation. However, if the second assumption is valid, that phonemes are acoustic signs, once we
transpose a phoneme into, say, written form, we destroy the phoneme and, hence, the sign. Noting
the consequence of accepting the second assumption, we might be tempted to reject the first
assumption as invalid. AUG's response is that we are able to distinguish between the phoneme and
its acoustic substance. A phoneme is, so to speak, the value of an acoustic entity, not the entity
itself. Shaumyan (1987:50-54)
AUG asserts that linguistic ideas can, and do operate according to a notion akin to Niels
Bohr's Complementarity Principle in physics. Shaumyan (1987:54-55) Complementarity is a
notion that several descriptions of a single phenomenon can be simultaneously maintained and
can prove useful, even if they appear to be contradictory; rather than being contradictory, in
reality, they are complementary, so that if they are taken together, we get a better and, perhaps,
more complete description of some aspect of reality than any single description provides
individually. As physics has shown, some descriptions of an aspect of reality, when seen from
different points of view, can appear to contradict other descriptions of the same reality, despite the
fact that both descriptions are accurate. The key to understanding a complementarity principle is to
note that many phenomena, like language and quantum level physics, are so complex that an
observer can only observe and describe a portion of the phenomenon at any given time. The
observer has a "point of view" which can and often does change with the observation.
Although neither traditional Western logic nor linguistic theory takes "point of view" into account,
quantum mechanics, which has been called the most successful scientific theory in human
history, could not function without due consideration being give to the point of view of the
observer and the effects of the observation on the observed. Shaumyan proposes a similar
consideration in linguistic theory. Incorporating the notion of complementarity to linguistics is one
of AUG's major advances over traditional linguistic theory. The notion of complementarity allows us
to study language without understating its inherent complexity. While it is beyond the scope of
this monograph to proceed further, it should be noted that AUG is strongly influenced by quantum
mechanics. See, e.g., Shaumyan's references to Bohr in a discussion of phonological
investigation Shaumyan (1987:58-60) and his development of the linguistic notion of
"superposition", which we will see later, originally a quantum mechanical notion developed by
Dirac.
13
Shaumyan acknowledges that his three antinomies are not without their critics. The reader
is referred to Shaumyan (1987:55-58) for a review of some criticisms and Shaumyan's refutation
of them. None of the criticisms, even if accepted, does damage to AUG. Shaumyan also reviews
some alternate theories of the phoneme, especially that of Trubetzkoy, and shows where, in
several places, the theories are inconsistent. Shaumyan (1987:61-66)
Phonological Syntagmatics
Every language strictly constrains sequences of phonemes; sequences that would be
acceptable in one language are not acceptable in another. Sequences have structure due to the
composition and arrangement of their individual phonemes. The minimal sequential phonological
element is the syllable; other, larger phonological elements also exist. Shaumyan uses
Trubetzkoy's term culminative function to describe the role of accent in words. The culminative
function uses accent to distinguish words from each other. Each word has one accent; in speech,
we can count the number of words in the speech flow by counting accents. The accent is placed on
a particular syllable, the accented syllable becoming the "center" of the word, with all other syllables
becoming "satellites" of it. Shaumyan (1987:66)
While Trubetzkoy noticed the culminative function with regards to accent, AUG sees a similar
function within syllables. A syllable is centered around its vowel feature. One segment of the
syllable becomes central, while the other segments are satellites of it. AUG sees an analogy
between the phonological functions of words and syllables. Shaumyan (1987:66-67) These
observations lead to the conclusion, at odds with some phonological theory, that vowels and
consonants cannot appear in the same position and that only vowels can be opposed to vowels, and
consonants to consonants. Shaumyan offers several proofs of this conjecture Shaumyan
(1987:67-69)
Stress has another function besides the culminative function. Stress can distinguish signs.
Usually stress does not have this secondary function, but we should be alert for those occasions
when stress does. Shaumyan (1987:68-69)
A syllable and a phonological word are not the same. A phonological word can be a sign; a
syllable cannot. Syllables do not have meaning (syllables cannot enter a sign of: meaning of
relation) by themselves. Monosyllabic words are not exceptions or counter-examples to this rule.
Monosyllables only function as words because they have meaning; they do not have meaning
because they are words. As an example, language speakers of Vietnamese (a highly monosyllabic
language) take monosyllables and make them into sign of: meaning of relations. Unless and until
that is done, a monosyllable is not a sign.
AUG observes that syllables are universal phonological units, i.e., every known language
uses syllables. However, the phonological word, which requires a stress as a component, is not
universal. In some languages, the number of stresses in a particular speech flow does not
signal the number of words in the flow. Shaumyan gives examples from French and German. In
the latter case, in which compounding words is a major method of vocabulary expansion, a
compound word may have several stresses. The largest phonological unit is a phonological
sentence with intonation. Intonation has both a culminative function and a distinctive function.
Shaumyan (1987:69-70)
From the analysis briefly described above, Shaumyan derives four laws applicable to all
languages:
The Law of Phonological Opposition
No language can have an opposition of vowels and consonants. Vowels oppose vowels;
consonants oppose consonants.
14
The Law of Phonological Contrast
Vowels and consonants contrast (have syntagmatic relations).
The Law of Phonological Functions
The culminative function is obligatory only for vowels; the distinctive function is obligatory for
consonants and optional for vowels.
The Law of Minimal Vocalism
The minimal vocalism is either two vowels with a distinctive function or one vowel with no distinctive
function.
These laws are not inductive; they are the products of an analysis of the semiotic properties
of language. Shaumyan (1987:70-71)
The Law of Minimal Vocalism implies that there can be languages with a monovocalic
phonological system. The Law of Phonological Opposition says that only vowels can oppose
vowels, and consonants . This law presupposes at least two phonemes. Any vocalism that has a
distinctive function must, therefore, have at least two phonemes. The culminative function is based
on contrast and, according to the Law of Phonological Contrast, vowels and consonants can
contrast. Therefore, a minimal vocalism, limited to contrast, must be confined to one vowel.
Shaumyan (1987:70-71) Shaumyan cites some controversial examples of languages, e.g., from
the Northern Caucasus, which fit this description. A fuller exploration is promised in another book.
Shaumyan (1987:71)
The Structure of Syllables
A syllable has three parts: an onset, a nucleus, and a coda. The combined nucleus and coda
are called the core. Vowels are the phonemes that function as the nucleus, the irreducible portion,
of the syllable. A syllable without a coda is "open"; a syllable with a coda is "closed". In some
languages, closed syllables occur either rarely or not at all, but no language has syllables that do
not have onsets. There is a basic opposition - onset:core. Every language must have this opposition
but not every language must have the opposition nucleus:coda. These two oppositions form a
hierarchy. If the analysis of a syllable is seen as a constituency tree, the opposition onset:core is the
most basic branch point on a tree, with the opposition nucleus:coda branching from the core.
There is an intimate relationship between the quantity or intonation of a syllable and its core.
The quantity of a syllable is fixed by the core. We see that long nucleus = short nucleus + coda
Shaumyan (1987:72-73) In some languages which have cores consisting of only short vowels,
stresses must pass to other vowels since short vowels cannot be stressed. These are the so-called
"light" syllables. A "heavy" syllable consists of either a long vowel, two short vowels, a short vowel
plus a consonant, or some combination of these three.
A basic problem for phonology is the definition of the syllable boundary. Shaumyan
introduces the notion of the interlude to address this problem. Shaumyan (1987:73) The
interlude is a string of one or more consonants between two vowels. An interlude may be either
unary or binary. A binary interlude has two components: the left part (the coda of the preceding
syllable) and the right part (the onset of the succeeding syllable). A unary interlude is a binary
interlude without the right component. Of the two possible components of an interlude, the right part
is the most important since the left part presumes that there is an onset (all languages have
onsets) but an onset does not presume a coda (not all languages have codas).
AUG assumes that there is a relationship between syllable boundaries and word structure.
Ideally, the same constraints on the sequencing at the beginning of a word should also operate on
the beginning of a syllable and the same constraints that operate on the sequencing at the end of
15
the word should operate at the end of the syllable. AUG defines the following basic principles for
determining syllabic boundary:
The interlude constitutes the onset of the next syllable, unless the preceding syllable cannot
be open because of vowel sequencing. In this case, the necessary number of consonants must
be detached from the onset and added to the coda of the preceding syllable.
If the interlude cannot occur at the beginning of a word, as many consonants as necessary are
detached in order to make the interlude acceptably appear in a word-initial position. The
detached consonants are attached to the coda of the preceding word. Shaumyan
(1987:73-74)
Shaumyan acknowledges that his proposed definition of vowels and consonants is not
generally accepted. Some linguists dispute the universality of Shaumyan's phonological definitions,
specifically the observation that sometimes consonants act as syllabic nuclei, with vowels as their
satellites. While agreeing that one and the same phoneme may function sometimes as a
syllable nucleus and sometimes as a nonsyllabic phoneme in the same language Shaumyan
(1987:75), Shaumyan points to the primary and secondary functions of a phoneme to support his
position. The primary function of vowels is to serve as syllabic nuclei and, secondarily, they
function as consonants; the primary function of consonants is to serve as satellites of vowels and
their secondary function is to serve as syllabic nuclei.
The notion of primary and secondary functions of phonemes is based on observations of the
"range" (the distribution of phonemes within a syllable) of phonemes. The range of phonemes is
not related to the statistical notion of frequency but is solely defined by its distributional possibilities.
Shaumyan (1987:75) If the range of a phoneme is greater when it serves as a syllabic nucleus, the
primary function of that phoneme is to serve as a syllabic nucleus; when its range is greater when it
serves as a satellite, then that is its primary function.
Clements and Keyser introduced the notion of the "extrasyllabic consonant" to describe
what they saw as a consonant which is not part of any syllable, e.g., the initial consonant in the
English pronunciation of Gdansk. Shaumyan sympathizes with the motivation for the creation
of the "extrasyllabic consonant" but views it as unnecessary. The g in Gdansk exercises its
secondary function and becomes a syllabic nucleus; to ease pronunciation an ultra-short schwa is
added. Shaumyan (1987:76-77) The notion of "extrasyllabic consonant" is an ad hoc solution to a
particular problem; the notions of primary and secondary functions of a phoneme are general
and arise from an analysis of the structure of a syllable. Furthermore, extrasyllabic consonants
make otherwise regular linguistic units irregular while the notion of primary and secondary
functions of the phoneme do not. The latter notions, which are more general than the notion of
extrasyllabic consonants, explain the phenomena in a more precise way and arise from a semiotic
analysis of language, are, therefore, to be preferred to the notions developed by Clements and
Keyser. Shaumyan (1987:77)
Prosodic Features
Prosodic features are elements in a speech flow whose duration differs from the duration of
phonemes. Usually, prosodic features have longer duration than phonemes, as can be seen in
syllables or phonological words that are longer than a single phoneme. However, prosodic
features are sometimes shorter than phonemes, as in the case of a syllabic nucleus splitting into two
consecutive units (morae). Prosodic features can be divided into two groups: accentual and
nonaccentual prosodic features.
Stress can be either bound or free. Stress is bound in a given language if it always appears
in one and the same place or if its place is strictly determined by the phonemic structure of a
word. Stress is said to be free if it can occupy various places in a word, independent of the
16
phonemic structure of the word. Free stress has both a culminative function and a word
differentiation function.
Depending on the manner of setting off a stressed syllable, AUG recognizes several types
of stress:
(a) strong, or dynamic, stress (the stressed syllable is set off by greater tenseness of its articulation);
(b) quantitative stress (the stressed syllable is set off by increased lengthening of the articulation
of the vowel); and
(c) tonal, or musical, stress (the stressed syllable is set off by a change in the tone pitch).
Shaumyan (1987:78)
Nonaccentual prosodic features are relevant to more than one syllable. Their only purpose is
word differentiation; they have no culminative function.
A mora is the smallest speech segment that can carry a prosodic feature.
In AUG, prosodic features are relational entities independent of definite phonic substance.
When differentiating prosodic and nonprosodic elements of speech, We do not differentiate
between fundamentally different substances. Prosodic and nonprosodic elements are simply
speech elements that are susceptible to different types of analysis. At this point, Shaumyan
reintroduces the Principle of the Transposition of Phonological Structure:
Any phonological structure can be transposed from one phonic substance to another as long as
the phonological relations characterizing this structure remain intact.
Shaumyan (1987:79)
The Principle of the Transposition of Phonological Structure predicts that a phonic substance
that may be a prosodic feature in one language may not be a prosodic feature in another
language. Shaumyan gives the example of the glottal stop which is a phoneme in Arabic, is a
distinctive feature characterizing glottalized phonemes in a few American Indian languages, and is
neither a phoneme nor a distinctive feature in Lettish or Danish, but is an accent, that is to say, a
prosodic feature.
Comparison with Generative Phonology
The phonological component of generative-transformational grammar consists of two levels:
systematic phonemics and phonetics. The systematic phonemic level converts representations of
underlying phonemic forms into phonetics. Systematic phonemics roughly corresponds to the
morphophonemic level of non-transformational grammar, and systematic phonetics roughly
corresponds to phonetics in non-transformational grammar. Generative transformational
grammar rejects the notion of the existence of a phonemic level and replaces it with generative
phonology. Shaumyan examines some of the claims of Morris Halle using the Principle of Semiotic
Relevance. Halle claims that the morphophonemic (phonemic) level is unnecessary and can be
dispensed with. Halle says that the existence of the phonemic level is based on the following:
A phonological description must include instructions for inferring (deriving) the proper
phonological representation of any speech event, without recourse to information not contained in
the physical signal.
Shaumyan (1987:81)
17
Halle notes that the procedures referred to in what he calls Condition 3a (the quote above) are
"essentially analytical". Halle says that Condition 3a creates an unscientific classification
method; ...it sets up a distinction between phonemes and morphophonemes for the sole reason
that the former can be identified on the basis of acoustic information alone, whereas the latter
require other information as well. Since it has never been shown that phonology differs from other
sciences in such a way as to warrant such a deviation in classification procedure (the opposite - that
phonology is a science just like all others - is usually asserted), Condition 3a is an uncalled-for
complication and should be dispensed with.
Shaumyan generally agrees with this criticism. However, after examining the example of
Russian phonology cited by Halle, Shaumyan shows that it was incorrectly interpreted. Reanalyzing
Halle's example raises the question of the correct basis for phonological theory. A comparison of
the goals of phonology in AUG and generative transformational grammar is instructive.
Phonology, in AUG, studies a system of distinctive oppositions. The starting point for this study is
the phonemic level as an ideal system of distinctive oppositions whose properties are
determined by the Principle of Semiotic Relevance. Shaumyan (1987:83) This ideal system of
distinctive oppositions is not based on analytical procedures; it is the theoretical basis from which
possible systems are deduced, i.e., it is a theoretical systems used for the purpose of
understanding real world systems. Since generative phonology studies a system of alternations,
it must base its theory on the morphophonemic level of language.
Shaumyan sees this important difference between AUG and generative phonology: a system
of distinctive oppositions is inherent in all languages but a system of alternations is not. A system of
alternations is important to languages that have such a system but not all languages do. It is
important to keep one feature of AUG always in mind: it is intended as a universal theory. AUG is
a theory of linguistic features which underlie all languages and which is intended to explain the
workings of all languages by constructing a set of basic, universal, theoretical models. Universality
is a goal of all aspects of AUG. While other theories purport to be applicable to all languages,
their emphasis on universality is not as central as it is with AUG. A consequence of the quest for
universality is the level of generality of AUG's description of linguistic features. Shaumyan
notes that generative phonology would have a difficult time defining the notion of "distinctive
oppositions" but that the notion of "alternations" is not significantly difficult to define in AUG's
terms (Alternations are nothing other than a subset of distinctive oppositions that hold between
allomorphs of morphemes. Shaumyan (1987:84)).
Alternations are founded on the synchronic structure of the phonetic level of language.
Shaumyan concludes that this poses a constraint on the type of rules that can be developed. The
Condition of Synchronic Motivations reads:
No rules should be formulated in phonetic terms if they do not correspond to synchronic
processes in the structure of the phonemic level.
Shaumyan (1987:84)
This condition prevents the substitution of diachrony for synchrony in phonological studies.
Shaumyan creates an integrated phonological theory with two basic goals. The first is the
study of systems of distinctive oppositions as "autonomous objects" and the study of systems of
alternations as objects partially subordinated to systems of distinctive oppositions. Distinctive
oppositions are autonomous because they are independent of systems of alternations. An
integrated phonological theory also has a methodological goal of studying the constraints on
phonological model making. We have already seen an important constraint: the Condition of
Synchronic Motivation. Shaumyan notes three others:
18
[t]he strict separation of functional and physical equivalence of sounds and acoustic
features;
functional and physical segmentation of the speech flow;
phonetically and nonphonetically conditioned alternations
Shaumyan (1987:86)
Shaumyan's brief suggestions about phonological model building will be omitted, except to note
that none of the suggestions are dogmatic. Perhaps because of its universal outlook, AUG does not
insist that its goals be met by following the "one true path".
By ignoring language's nature as a sign system, generative phonology has confused
synchrony and diachrony as well as confusing the phonemic level with the morphophonemic and
phonetic levels of language. Shaumyan examines the claim that the Ordering Hypothesis (that rules
are ordered) is "one of the best supported assumptions of linguistic theory". He disagrees with this
assertion. While agreeing that language is hierarchically structured (indeed, hierarchical
stratification is one of the bases of AUG), Shaumyan points out that the Ordering Hypothesis refers
to an extrinsic and, therefore, arbitrary ordering rather than an intrinsic and natural ordering. The
erroneous claim for the Ordering Hypothesis is a result of the formal design of generative
transformational grammar. Generative transformational grammar is capable of producing
phonological objects; generative transformational grammar assumes that, since it can produce
phonological objects with mathematical procedures, the mathematics used accurately
describes the way that language works. Shaumyan correctly notes that The generative model is
based on logical necessity, but logical necessity does not necessarily conform to empirical
necessity. Shaumyan (1987:87) and again that the mathematical consistency of a theory does not,
ipso facto, mean that the theory is a correct description of reality. Shaumyan (1987:92) The
generative notion of linguistic competence is the unfortunate conflation of synchrony and
diachrony. Shaumyan reviews a number of examples from Chomsky's works and shows the
confusion between synchrony and diachrony that he claims for generative phonology. Shaumyan
(1987:88-92)
Part Three - Genotype Grammar
Genotype grammar is a notion unique to AUG; the uniqueness of genotype grammar is a
consequence of AUG's special concern for universality. Genotype grammar is a grammar of a
"universal" language. By this, I mean that genotype grammar is not intended to be the
grammatical theory of any specific language but, instead, is intended to be a grammatical
description of the common features of all languages.
A universal description of language imposes certain requirements on its underlying
formalism. Obviously, the formalism of a genotype grammar cannot be based on the formal
description of any particular language. To do so would create two significant problems:
we would have to justify the selection of the language on which we decide to base our
formalism.
the act of selection suggests that the language chosen as the basis for a universal formalism is
somehow "better" for the job than other languages; i.e., that some language might be
more "universal" than others (somehow, the chosen language has more "universal" features
than others or that a "universal" language description should be somehow like the chosen
language) or that some languages are somehow "deviant" or not as "universal" as others.
AUG avoids these problems by creating a formalism that is not based on the grammar of any
specific language or on the formalism of any other linguistic theory.
19
Genotype grammar is one of the two types of grammar discussed in AUG. As mentioned,
genotype grammar is a universal grammar. AUG also posits a phenotype grammar, specific to each
language. The genotype level of a language can be understood as the functional level of a
language while the phenotype level can be understood as the syntagmatic level of the language. A
complete description of any language is a description of both its genotype and phenotype levels.
Basic Terms
The minimal sign is the morpheme.
A combination of morphemes is called a syntagm.
Morphemes, words, and groups of words are called syntagmatic units.
A language has functional units. Functional units are observed when we view language
as an instrument of communication.
The basic functional unit is the sentence. The purpose of the sentence is to transmit
messages from sender to receiver. An importance consequence of viewing language as an
instrument of communication is that unless a word or group of words functions as a unit of
communication, i.e., a sentence, the word or group of words does not transmit a message.
After the sentence, the most basic functional element is the predicate. The predicate functions
as the "representative" of the sentence; the predicate is the irreducible minimum component
of a sentence.
It may be objected that English does not permit a declarative sentence to consist of only a
verb. An English declarative sentence must have both a predicate and a subject. Shaumyan
responds that this correct observation is an observation of the phenotype level of English. There
are rules that require English to have sentences that meet this description, but these rules do
not apply to all languages. In no language can a sentence lack a predicate. Recall that the
genotype level of language is the universal level; the fact that the predicate is indispensable to
sentences in all languages is the basis of AUG's claim for the centrality of the predicate. The reader
should not think that AUG says that the well-known fact that English requires properly formed
declarative sentences to have subject terms is an incorrect observation. AUG simply says that
such an rule does not apply to all languages, another well-known fact. Rather than declaring that
languages which are not consistent with a syntax similar to Englishs to be somehow deviant or
nonnormative, AUG places this rule on the phenotype (language specific) level rather than on the
genotype (universal) level.
The term predicate and other terms we will see (primary term, secondary term, tertiary
term, term modifier, and predicate modifier) are intentionally chosen to be neutral terms of art.
Shaumyan (1992:8) Each of these terms must be understood as referring to functions on the
genotype level of language and not on the phenotype level. When we say something like "the
predicate is indispensable to all languages", the reader should understand this to mean "the
predicate is indispensable to the description of the genotype grammar of all languages".
Another important functional element is the term. The term is a complement to the predicate.
The predicate denotes a situation; the term designates a participant in the situation denoted by
the predicate. Predicates and terms can be modified by predicate modifiers and term modifiers.
These four, predicates, terms, predicate modifiers, and term modifiers, are not the only
functional units that inhabit a genotype grammar, but they are the only four described in any detail
by Shaumyan (1987). The four types of functional elements are not simply "preferred terms" for
verb, noun, adverb, and adjective. The words verb, noun, adjective, and adverb are the names of
syntagmatic units. Syntagmatic and functional units belong to different levels of the description of
20
language. Syntagmatic and functional elements do not have a one-to-one correspondence. As we
will see, functional units can be composed of groups of words as well as single words. A functional
unit can also be as small as a single morpheme. The latter observation leads to the distinction
between two types of syntactic connections:
syntactic connections between functional units and
syntactic connections between units that realize functional units, that is, syntactic connections
between syntagmatic units.
Shaumyan (1987:97)
Shaumyan cites considerable confusion between these two types of syntactic connections in
current linguistic literature, agreement and government being examples. Shaumyan (1987:97) A
basic observation of AUG is that the syntactic structure of a sentence is independent of its
syntagmatic representation. This observation leads directly to the observation that grammar has two
levels (or that language has two levels of grammar).
The genotype and phenotype levels of language can be described as follows:
Genotype grammar comprises functional units - predicates, terms, modifiers - and abstract
operator-operand relations between these units. Phenotype grammar comprises syntagmatic units -
morphemes and words - and connections between them in terms of linear order and their
morphological properties.
Shaumyan (1987:97)
The terms genotype and phenotype are consciously borrowed from biology to emphasize the
relation between the two linguistic levels. Just as genes are invariant with respect to different
phenotypes, the rules of the genotype level are invariant with respect to different phenotypes
(individual languages). Acceptance of the two level description of language puts us on the road to
the solution of the basic question of linguistic theory... What factors contribute to the similarities
and differences between natural languages? Shaumyan (1987:97)
Shaumyan uses passivization as a demonstration of the importance of seeking language
invariants, i.e., genotype rules and descriptions, for linguistic phenomena. Generative
transformational grammar sees the change from an active form of a sentence to a passive form as a
transformation. While there are many proposals for exactly how passivization should be viewed in
generative transformational grammar, all proposals basically come down to transforming a sentence
in the form NP1 V NP2 to NP2 V
(pass)
NP1. We can see that this is so for English sentences but
can we maintain such a rule universally? Shaumyan says no. The generative transformational
rule just mentioned relies on the linear order of English as its basis. In Malagasy, an active
sentence is in the form V NP2 NP1 (with NP1 being the subject and NP2 being the object); in its
passive form, it becomes V
(pass)
NP1 NP2. In many languages, such as Russian, word order is
irrelevant for passive transformations. Russian active and passive sentences can have the
same word order. Russian active and passive forms depend on the case markings of its terms.
A universal rule for passivization cannot be stated in terms of word order, as generative
transformational grammar suggests. Word order and the system of case markings are language
dependent; in the terminology of AUG, they exist on the phenotype level. To create a universal
passivization rule, we must examine the functional units on the genotype level. For our purposes
here, we can see the predicate as a binary relation involving an actor and a patient. Passivization is
the conversion of this binary relation. Conversion is an operation that creates the converse of a
relation R, which we shall call R'. R' holds between X and Y iff R holds between Y and X. A passive
predicate is the converse of an active predicate. In an active sentence, the primary term is the
agent and the secondary term is the patient; in a passive sentence, the primary term is the patient
21
and the secondary term is the agent. Primary and secondary terms will be described later, but, for
now, the reader should keep in mind that the names primary and secondary do not imply linear
position or a fixed word order.
The rule that the passive is the converse of the active is universal, i.e., the rule is not language
dependent. One of the hallmarks of a universal rule is that it is predictive. In three example
languages, English, Malagasy, and Russian, the passivization rule correctly predicts the outcome
for each language. The predictiveness of universal rules makes AUG testable in a way that
generative transformational grammar is not. Another result predicted by our rule is that, since our
rule is an operation on the predicate, we expect the predicate to be modified in some way. We note
that the predicates in our examples are morphologically marked to indicate their passive form.
Shaumyan (1987:97-101)
Some Basic Notions of Genotype Grammar
Communication, i.e., a language act, has three functions, expressive, vocative, and
representational functions, depending whether the language act is focused on the sender,
receiver, or some subject matter external to the language act, respectively.
Shaumyan concentrates on the representational function of language acts.
The Principle of Representational Relevance is a postulate that describes some of the basic
elements of communication that are considered by AUG. The Principle states that, after we
remove everything in the language that is irrelevant to the representational function of the
communication, we find that there are only three types of linguistic expressions remaining: the
"names" of objects, the "names" of situations, and methods for constructing either or both types
of "names". Shaumyan (1987:101) I have drawn attention to the word "name" because it is both
of central importance and easily misunderstood.
AUG calls the "name" for an "object" a term and the "name" for a "situation" a sentence.
AUG calls any means for creating terms or sentences operators.
A term should not be misunderstood to be a noun. While a term can be a noun, a term can
also be what is traditionally referred to as a noun phrase.
For example, car, the car, the big car, and the big, yellow car are all examples of a term. Each is
the "name" of an object, here only coincidentally a tangible object in the real world. Each example
term can refer to the exact same object or to different objects; nonetheless, each example is both
a "name" and a term. It should be noted that "term" is the "name" of the linguistic function of each
of the language examples just given.
Like a sentence, a term can be thought of as being constructed of parts. Unlike generative
transformational grammar, AUG does not recognize a term, once constructed, as still being
composed of parts, except for convenience. Generative transformational grammar, which is
based on first order logic, sees a phrase like the big, yellow car as actually being the + big + yellow
+ car, in short, a unit composed of arbitrary parts. AUG, which is based on second order logic, sees
the big, yellow car, once constructed, as a single functional unit. This distinction, which at first
seems to be subtle and, perhaps, not worth noticing, has significant implications. Some are
explored in Shaumyan (1987).
Operators permeate genotype grammar.
An operator is any kind of linguistic device that acts on one or more expressions called its
operands to form an expression called its resultant. Shaumyan (1987:101)
22
Operators must operate on something; an operator cannot function without operands. An operator
may have one or more operands.
An operator with n operands is referred to as an n-place operator.
Operators should not be confused with verbs. While a verb is a type of operator, it is far from being
the only type of operator in genotype grammar.
Any kind of linguistic device can be an operator; to take a few examples, morphemes,
intonation, and even zero signs can be operators.
Shaumyan does not simply graft second order logic on top of linguistics and declare the result
to be a new linguistic theory. He uses the techniques of second order logic to explore and explain
linguistic phenomena and, from the results of his explorations, develops a new linguistic theory
(AUG). For instance, Shaumyan notes that logic treats all operands as being equally connected to
its operator. In short, no operand receives any preference in treatment by the operator. Linguistics
shows that language does not quite work this way, that language shows degrees of connectedness
between operands and their operators. Here are some examples:
A transitive verb (an two-place operator) is more closely connected to its direct object (secondary
term) than with its subject (primary term).
The combination of a transitive verb and its object is equivalent to an intransitive verb. One can
replace the other without loss of meaning, e.g., to dine can replace to eat dinner and vice
versa.
This last observation explains certain types of ambiguity. Nouns derived from intransitive verbs are
"oriented" toward the subjects of the actions described and nouns derived from transitive verbs are
oriented towards the objects of their actions. Some verbs, e.g. shoot, can be used both
intransitively and transitively. When such a verb becomes a noun (shooting), it creates the
ambiguous situation of whether the action it describes is oriented towards its subject or object (the
shooting of the hunters). Shaumyan notes that the orientation of nouns derived from transitive
verbs towards the object of the actions described is a universal tendency.
It has been shown, first by Moses Schnfinkel and later by Haskell Curry, that an act of
combination involving an n-place operator is actually a series of one place combinations repeated n
times with the operator using different operands each time. Shaumyan sees the implication of this
logical observation for linguistics. Take the sentence The hunter killed the bear Shaumyan
(1987:103). In this sentence, killed is a two place operator, the hunter is the primary term, and the
bear is the secondary term. Rather than applying the operator killed to all its terms at once, AUG
applies it first to the secondary term, as the previous observations suggest it should. The resultant at
the first step of the application gives us two units: a one-place predicate combined with its
secondary term ((killed the bear)), a unit now equivalent to an intransitive verb, and the primary term
(the hunter). The new, one place predicate is now applied to the primary term. The resultant is
the sentence (also called an s-type or, simply, s): ((killed the bear) the hunter). The process which
we have just run through is called an application, hence the name of Shaumyan's theory,
Applicative Universal Grammar.
The Applicative Principle formally defines the process which we have just seen:
An n-place operator can always be represented as a one-place operator that yields an (n-1)-place
operator as its resultant.
Shaumyan (1987:103)
23
Keep in mind that the results of applying an operator to an operand is another operator; second-
order logic discusses operations on functions which yield other functions. Even a sentence is an
operator; it is a zero-place operator, i.e., an operator which cannot operate any further (there is
nothing left to operate on). The sentence is the basic unit of communication in all languages.
Here are the definitions of a one, two, and three place predicate, and primary, secondary, and
tertiary terms.
Definition 1. If X is an operator that acts on a term Y to form a sentence Z, then X is a one-place
predicate and Y is a primary term.
Definition 2. If X is an operator that acts on a term Y to form a one-place predicate Z, then X is a
two-place predicate and Y is a secondary term.
Definition 3. If X is an operator that acts on a term Y to form a two-place predicate Z, then X is a
three-place predicate and Y is a tertiary term.
Shaumyan (1987:103-104)
Notice that the definitions above are not the definitions of operator but of predicate. The notion of
predicate roughly corresponds to the traditional notion of verb, but is not be restricted to the
traditional notion. The phrase that acts on in the above definitions can also be read as combines
with. When an n-place predicate combines with a term, it produces an (n-1)-place predicate that is
the linguistic structure of the predicate combined with the linguistic structure of the n-ary term. It is
this combination which is referred to as the (n-1)-place predicate and which operates on the n-ary
term.
Application can be described in applicative tree diagrams. The interested reader should
refer to Shaumyan (1987:104) for examples.
Constituency
Operators and operands are interconnected by constituency and dependency relations. An
operator and its operand are related as parts to a whole (constituency relation) with its resultant. An
operator and operand exist in a dependency relation; either the operator or the operand can be the
head of such a relation with the other being the dependent. Both constituency and dependency
networks exist simultaneously.
If A is an operator and B is its operand, C is the resultant of applying A to B. A and B are called
the immediate constituents of C. In general, C can have an arbitrary number of constituents.
Constituents are not defined in terms of word order. Constituents need not be linearly adjacent
to each other, as in generative transformational grammar.
If C is the resultant of applying the operator A to the operand B, then either A is the head of a
dependency relation with B as its dependent, or vice versa.
If C belongs to the same linguistic category as B, then B is the head and A is the
dependent, and vice versa.
If B is the head and A the dependent, then A is called the modifier of the head; otherwise, A is
called the complement of the head. Shaumyan (1987:106-107).
Shaumyan gives this example. There are several operator : operand combinations in the sentence
Bill bought new books. Applying new (an operator) to books (its operand), we get (new books). New
24
is the modifier of books because new books is in the same category as books, both being terms.
When we apply bought to (new books), we get (bought (new books)). Since (bought (new
books)) and (new books) are in different linguistic categories, (bought (new books)) being a
predicate, bought is the head of a dependency relation with (new books) being its complement.
Once again, applying (bought (new books)) to Bill, we get another dependency relation with (bought
(new books)) as its head and Bill as its complement because the resultant of applying (bought (new
books)) to Bill is ((bought (new books)) Bill), which is a sentence or s-type.
The notions of head and dependent in AUG are more general than in current dependency
grammar. Dependency grammar cannot describe dependency between functional units while
AUG can. AUG does not view dependency as a primitive concept. Dependency grammar cannot
define dependency; AUG, using the notion of application, can. AUG sees no conflict between
constituency and dependency but, rather, sees these notions as complementary. Shaumyan does
not see constituency as being superior to dependency or vice versa; neither can be dispensed
with. Both constituency and dependency, in AUG, are relations between operators, operands,
and resultants.
Sentence Structure
Before discussing AUG's conception of sentence structure, we need to mention certain basic
notions. In order to understand sentence structure from a genotype point of view, we must realize
that discussions about the details of the genotype stratum are not discussions of syntax. The
genotype stratum is composed of functional units. These functional units are universal and
irreducible. A comment such as Neither words nor combinations of words belong to syntax proper
Shaumyan (1987:109) may, at first, seem startling, but it must be understood as a comment based
on the two-strata view of language. The genotype stratum does not have syntax; the phenotype
level of language has syntax. The genotype stratum is composed of abstract types that all
languages have. The phenotype level of a language is a syntagmatic realization of the
universal genotype level peculiar to that language. In order to emphasize this distinction, Shaumyan
coined the word syntaxeme for functional units.
Predicate Frames
A predicate frame is a combination of a predicate and a number of operands (terms) which will
result in a sentence. The smallest predicate frame consists of a predicate and a primary term.
Shaumyan does not use the word subject but prefers primary term. The primary term is
privileged; the name refers to a participant whose role, whether active or passive, is emphasized
by its choice as the primary term Shaumyan (1987:110) Shaumyan refers to this role as a "pivot".
The importance of the primary term is reflected in the linguistic fact that it alone among terms is
indispensable to a sentence.
Traditionally, sentences are seen as an expansion of the basic sentence that consists of only a
predicate. In AUG, the sentence is a reduction of structures based on the predicate :
primary term relation. We will see the reason for this claim when we review the Principle of
Maximal Differentiation. AUG recognizes that a predicate, by itself, can form a complete
sentence. In this case, the predicate : primary term relation is dispensed with. In some languages,
e.g., Latin, sentences without explicit primary terms are quite common.
A two-place predicate plus a secondary term and a one-place predicate without a term are
syntactically equivalent. A three-place predicate with a term is syntactically equivalent to a two-
place predicate without a term. Tertiary and secondary terms are more closely connected to their
predicates than primary terms. Sentences which contain more than one term must be constructed
by incorporating terms into predicates. Note that the notion of predicate and term combination
is not a theoretical construct but results from observation of linguistic facts. Shaumyan gives
25
examples of predicate/verb combinations collapsing into simple predicates, e.g., form a circle
round collapses to encircle Shaumyan (1987:112)
Predicates and terms can be operated on by modifiers. Taking the case of a term, we
observe that the elementary form of a term is a noun and the elementary form of a term modifier is
an adjective. Term modifiers are not limited to adjectives; the role of term modifier can be taken by
other words or groups of words. AUG recognizes that words have primary and secondary
functions. The primary function of an adjective is to be a term modifier; the primary function of a
verb is to be a predicate but one of its secondary functions is to be a term modifier. Even a
sentence has primary and secondary functions. Its primary function is to be a unit of communication
but its secondary function is to be a term modifier , e.g., in a clause beginning with which (which
John ate - John ate can function as a well formed sentence outside of the clause). An adverb has a
primary function of being a predicate modifier but other words that are not adverbs have a
secondary function of predicate modification. An important syntaxeme is the sentence modifier,
such as unfortunately in Unfortunately, it is too late.
A sentence can be transposed into a term. Once the sentence is transposed into a term, it
can be used like any other term. For example, John has left is a sentence By transposing this
sentence into a term, we can use it as a term modifier: I know that John has left . The conjunction
that serves to transpose the sentence John has left into a term. When sentences are used as
terms, they can be either coordinated or subordinated. When two sentences X and Y, are used as
terms in another sentence Z, X and Y are said to be coordinated when they can interchange their
positions within Z while Z maintains its original sense. X and Y are said to be subordinated when
Z changes its meaning when X and Y change positions within Z.
Functional Transposition and Superposition
A transposer is an operator which changes the linguistic category of its operand to
another linguistic category; the transposed operand is called a transponend, and the process is
called transposition.
Transposition is different from transformation as the term is used in generative transformational
grammar. Transformation changes a specific set of structures through a specific set of functions.
Transposition has no such limitations; transpositions are relational operations that change the
linguistic categories of the operands of transpositional operators. AUG does not require
transformations at all. What does transposition get us? Transpositions enable AUG to explain
linguistic phenomena that are difficult for other forms of grammar to explain. Transposition
highlights the role of linguistic categories (what Shaumyan calls word categories). Word categories
have a dual nature. Words have some very general lexical meaning and some definite syntactic
function. We may think of a word as having a denoting function and a genotype function. For
example, a noun denotes some object and functions as a term. These functions are inherent in
words and coordinate with each other, i.e., certain syntactic functions and denoting functions
naturally are found together. When a word is transposed, both functions are effected. Words have
primary and secondary functions. Primary syntactic functions are inherent in words; secondary
syntactic functions are created through transposition. We have already seen that even sentences
can be transposed from their primary function as an independent unit of communication.
AUG abandons concepts like verb and noun because not every language distinguishes
between lexical classes but every language has predicative and nonpredicative functions
Shaumyan (1987:115). Recall that such assertions are a direct result of AUG's attempt to be a
grammar that is universally applicable to all languages.
The interplay of primary and secondary functions have very important implications for AUG.
For diachrony, when a word X receives a secondary function F, there is a universal tendency to
replace X with another word or group of words Y with F as its primary function. Concerning the
26
typological classification of languages, we note that inflectional languages have different types of
transposers from analytical languages. A central problem for AUG's language typology is an
accurate assessment of the use of transposers within a particular language.
Superposition is an important, novel notion that is related to transposition. Superposition is a
notion borrowed from quantum physics that states (roughly) that a particle can exist in several
energy states at the same time.
Superposition in AUG states that a transposed syntactic unit does not change from one
category to another. Instead, it acquires a new function superposed on top of its old
function. A superposed word can exercise each of its functions if necessary.
Take the verb instruct. The suffix -ing added to instruct superposes the verb instruct into the
gerund (a verbal noun) instructing. The superposed function can be thought of as a syncretic
function, i.e., a new function that combines its original and superposed functions.
Valence and Voice
Valence is the number of operands that an operator has.
In AUG, the notion of valence is generalized; all operators, not just predicates, have valence. The
valence of an operator can be changed. For example, one English valence-changing rule says that
when root vowels of one-place predicates, we can derive two-place predicates: to fall, a one-place
predicate, becomes a two-place predicate to fell. The tree has fallen changes to Someone has felled
a tree. The meaning of the sentences are the same but the valence of their predicates has changed.
A true decrease in valence is caused by some formal device. Some languages, like English, do
not have formal devices for decreasing valence.
Conversion is a process that neither increases nor decreases valence. Instead, it permutes the
roles of terms.
For example, X killed Y can be converted to Y was killed by X. In the first case, X is a primary term,
denoting an agent and Y is a secondary term, denoting a patient; in the second X is a secondary
term, denoting an agent and Y is a primary term, denoting a patient. Conversion can be applied to
any many-place operator. Conversion shows us that there are actually two types of valence:
quantitative valence and relational valence. We have already seen quantitative valence.
Relational valence is an orientation of an operator with respect to its operands determined
by the rules of conversion. Shaumyan (1987:119)
The grammatical category into which predicates are sorted by their quantitative and relational
valence is called voice. A language can have many different voices, but no language will have
all possible voices.
Sentence Construction Typology
There are two broad types of language: those that have a transitive:intransitive opposition and
those which do not distinguish between the transitive and intransitive but distinguish between
the active and inactive. Shaumyan (1987:120) The first group is further subdivided into
accusative and ergative languages.
A transitive construction is a sentence with a two-place predicate in which the primary term
can denote an agent and the secondary term a patient (accusative languages) or, vice versa
(ergative languages).
27
An intransitive construction is a sentence with a one-place predicate in which the primary term
can denote either an agent or a patient.
Agent and patient are considered primitive notions by AUG. AUG uses the terms agent and
patient in preference to subject and direct object because the traditional definitions and uses of
these terms causes difficulties for linguistics. If the term subject denotes an agent, and the term
direct object denotes a patient, as they do in transitive constructions, these terms can cause
obvious confusion when applied to ergative languages in which the denotations of subjects and
objects are reversed. Agent and patient are denotations, not syntactic functions. We will not explore
other problems created by the traditional nomenclature; the reader is referred to the text. Since they
are not universal notions, they are not of use to AUG.
The primary term : secondary term opposition is a central concept in the syntactic
organization of language. Since intransitive constructions have primary terms only, the range of
primary terms is wider than that of secondary terms. The notion of range is defined by the
Markedness Law. The Markedness Law applies to any syntactic opposition, not just the
primary term: secondary term opposition.
If A is an unmarked term and B is a marked term, the opposition A : B is called the
markedness relation between A and B.
The range of a semiotic unit is the sum of its syntactic positions. According to the Markedness
Law, if the range of A is wider than the range of B, the set of relevant features of A is
narrower than the set of relevant features of B. B has a plus relevant feature. A relevant
feature is an essential feature that is part of the definition of a semiotic unit, i.e., a relevant
feature is an element of the definiendum of a semiotic unit.
The Markedness Law leads to the Dominance Law.
The Dominance Law states that in a given sentence, the marked term cannot appear without
the unmarked term, but the unmarked term can appear without the marked term Shaumyan
(1987:123).
A corollary to the Dominance Law states that the unmarked term is central and obligatory to a
sentence.
The marked term is both marginal and dependent. As a consequence of this law, languages
manifest the following general tendency: ergatives can, but absolutives cannot, be eliminated in
transitive constructions in ergative languages; accusatives can, nut nominatives cannot, be
eliminated in transitive constructions in accusative languages. Shaumyan (1987:123)
Shaumyan uses a small chart to describe the Applicative Hierarchy in which square
brackets surround members of a predicate frame:
[Primary term > Secondary term > (Tertiary term)] > Oblique term
The parentheses around the tertiary term indicate that it is marginal with respect to the primary
and secondary terms; it occurs only in ditransitive constructions. The range of predicate frames is
broader than the range of oblique terms. Oblique terms do not appear in every construction;
predicates must.
The oblique term is a term transposed into an adverbial that serves as an operator modifying
the predicate frame. Shaumyan (1987:123-124)
28
Types of Meaning
Shaumyan cites some supposed counterexamples to his Applicative Hierarchy. In order to
dispense with these counterexamples, Shaumyan discusses his notion of meaning. In AUG, there
are two kinds of meaning: lexical and grammatical. Grammatical meanings can be thought of as
descriptions of the roles in a sentence that different genotypes play. For instance, the grammatical
meaning of a primary term is agent; the grammatical meaning of a secondary term is patient.
Shaumyan speaks of these as mappings. in Shaumyan (1987); mapping does not appear in
Shaumyan and Sypniewski (1995). Agent and patient are universal to all languages. These two
grammatical meanings are present in every language but other grammatical meanings, such as the
grammatical meaning of the oblique (a spatial reference) may appear differently in different
languages. In other words, the precise grammatical meaning of the oblique or the tertiary term in a
language is determined by the linguistic needs of its language community Shaumyan (1987:125);
Shaumyan and Sypniewski (1995:3) The grammatical meanings agent and patient are invariant
for all languages.
AUG asserts the Principle of Obligatoriness:
Any language contains a closed set of meanings, called grammatical meanings, which are
obligatory for presenting various aspects of experience.
Shaumyan and Sypniewski (1995:3)
Grammatical meanings and lexical meanings are distinct and should not be confused, even
though grammatical meanings are strictly formal devices and are not directly accessible. Lexical and
grammatical meanings are blended together to form a single heterogeneous object. Even though
grammatical meanings are not directly accessible, their existence can easily be shown by a
thought experiment; see Shaumyan (1987:126-127); Shaumyan and Sypniewski (1995:6-7)
Grammatical meanings and lexical meanings are not always coordinated.
The Principle of Free Coordination of Grammatical and Lexical Meanings states:
The grammatical and lexical meanings of a word or a construction may refer to the same
or different types of experience.
Shaumyan and Sypniewski (1995:3)
The English word cat has coordinated grammatical and lexical meanings. Lexically, cat
refers to an object, and its grammatical meaning is either an agent or a patient, i.e., its grammatical
meaning also refers to an object (understood broadly). The word rotation has uncoordinated
lexical and grammatical meanings. While rotation is a term and, therefore, is either an agent or
patient, its lexical meaning refers to a process (the process of rotating); its lexical meaning has
something of the aspect of a predicate to it, unlike cat which strictly refers to an object. If cat were to
be used as a predicate, as in to cat about, it would have uncoordinated lexical and grammatical
meanings, since, despite its being used as a predicate, to cat about has something of an object
about its lexical meaning (in the sense that to cat about can be understood as to act like a cat in
some way perhaps by moving silently about). Shaumyan and Sypniewski (1995:3-4)
There are two classes of grammatical meanings: inflectional and derivational.
Derivational grammatical meanings are structural, e.g., the suffix -er determines the grammatical
meaning of the root to which it is appended.
Inflectional grammatical meanings can be further subdivided into two groups: functional and
specifying meanings.
29
Morphemes with functional meanings are morphemes capable of giving words syntactic
functions, e.g., morphemes capable of making words predicates.
Morphemes with specifying meanings are morphemes that serve to distinguish words within a
common functional group, e.g., morphemes that serve to distinguish predicate tenses.
Shaumyan and Sypniewski (1995:4)
Lexical meanings are the meanings of morphemes that are word stems; grammatical
meanings, on the other hand, may come from inflectional morphemes, prepositions,
conjunctions, and other devices, such as word order. Shaumyan (1987:127)
Many linguists reject the opposition of grammatical meanings and lexical meanings or
confuse grammatical meanings with functions. Many linguists do not consider syntactic meanings to
be meanings at all. Rather, they separate syntax and semantics and claim that syntax operates
separately from semantics. AUG rejects the syntax semantics separation and proposes instead a
distinction between grammar and lexicon. While grammar and lexicon interact to some degree,
the two are relatively independent. Even so, Shaumyan sees that the distinction between
syntax and semantics is useful in some respects. What Chomsky and Montague refer to as
semantic meanings are directly connected with the real world whereas syntactic meanings, being
functional, that is, connective meanings, are connected with the real world only indirectly.
Syntactic relations have their own meaning, they do not need an interpretive component. Syntax
is a mold for semantics. Syntax and semantics relate to each other not as a formal and
interpretive component of grammar but as the distributional form and the content of the
distributional form. Shaumyan and Sypniewski (1995:5) By keeping the distinction between
grammatical and lexical meanings before us, our descriptions of language become more
precise. Because of the confusion of grammar and lexicon in some linguistic theories, the
proposed semantical notion of agency is vague; as a grammatical meaning in AUG, agency
becomes quite precise and correlates with linguistic fact.
In order to demonstrate the benefits of AUG's grammar : lexicon dichotomy, with its notions
of grammatical and lexical meanings, Shaumyan examines ergativity in considerable detail. While
I will refer the reader to the text for the details of Shaumyan's arguments about ergativity
Shaumyan(1987:129-145), there are a number of comments made during the course of that
argument to which we should refer as being pertinent to a general outline of genotype grammar.
The Criterion of the Nonomissibility of the Subject bears some discussion. Essentially, the
primary term in a sentence must be present; a proper sentence will be formed even if all other
terms can be omitted. Ellipsis is not the omission of a term. If we do not display a term by using a
form of ellipsis, we can always recover the term but we cannot recover an omitted term. Omitting
the secondary term from John ate clams gives us the sentence John ate. We cannot determine
that the omitted secondary term was clams when we see the sentence John ate. Shaumyan points
out that different languages have different rules of ellipsis. In Latin, for example, a pronominal
primary term is used only for emphasis; it is normally subsumed into the predicate ending. In
Russian, the opposite takes place; the pronoun is dropped for emphasis.
A markedness relation is a binary relation between a marked and an unmarked semiotic
unit. In a relation between two semiotic units A and B, if the range of A is wider than B, then A's set
of relevant features is narrower than B's set. A is the unmarked unit, B is the marked unit.
According to the Dominance Law, The marked term in a sentence cannot occur without the
unmarked term, while the unmarked term can occur without the marked term. Shaumyan
(1987:139) The unmarked term is central and independent; the marked term is marginal and
dependent. The unmarked term is the primary term; the marked term is the secondary term.
The marked term is omissible because the unmarked term does not presuppose the
occurrence of the marked term. And the unmarked term is nonomissible because the marked
term presupposes the occurrence of the unmarked term; that is, it cannot occur without the
30
unmarked term. Shaumyan (1987:139) Shaumyan notes that there can be deviations from this
law, but that these deviations are not counterexamples. For instance, if the meaning of a
transitive predicate is not complete without the marked term (a direct object), the marked term
cannot be omitted. Note that this is not a syntactically caused deviation; it is a deviation caused by
the semantics of the sentence.
Shaumyan posits the Law of Duality:
The marked term of an ergative construction corresponds to the unmarked term of an accusative
construction, and the unmarked term of an ergative construction corresponds to the marked term of
an accusative construction; and vice versa, the marked term of an accusative construction
corresponds to the unmarked term of an ergative construction, and the unmarked term of an
accusative construction corresponds to the marked term of an ergative construction.
Shaumyan (1987:140)
The Law of Duality describes ergative and accusative constructions as mirror images of each other.
Shaumyan notes that this law also applies to phonology. For phonology, he restates the Law of
Duality as follows:
The marked term of the opposition lax:tense corresponds to the unmarked term of the opposition
voiced:voiceless, and the unmarked term of the opposition lax:tense corresponds to the marked term
of the opposition voiced:voiceless; and, vice versa, the marked term of the opposition
voiced:voiceless corresponds to the unmarked term of the opposition lax:tense, and the unmarked
term of the opposition voiced:voiceless corresponds to the marked term of the opposition lax:tense.
Shaumyan (1987:141)
The main problem with ergativity, when seen in light of the Markedness Law, can be
restated:
How can we resolve the contradiction that ergative, which is the secondary (marked) term of a
clause, is treated under the above rules as if it were the primary (unmarked) term of a clause?
Shaumyan (1987:141)
Shaumyan relies on functional superposition to resolve the apparent contradiction posed by the
restatement. First, he notes that only a primary term can occur in intransitive clauses. If a primary
term of an intransitive clause were to be identified in a transitive clause, it would have to be
superposed (i.e., given a secondary function) to one of the terms of the transitive clause. There are a
total of three possibilities:
only the primary term of a transitive clause can be identified with the primary term of an
intransitive clause (no superposition);
only the secondary term of a transitive clause can be identified with the primary term of an
intransitive clause (a superposition of the function of the primary term); or
both the primary and the secondary terms of a transitive sentence can be identified with the
primary term of an intransitive sentence (no superposition or the superposition of the function of
the primary term of an intransitive clause onto the secondary term of the transitive clause).
Shaumyan (1987:141)
Accusative and ergative languages do not use the above mentioned possibilities equally.
Accusative languages use only the first of the three, while ergative languages use all three. In short,
31
functional superposition occurs in ergative languages and not in accusative languages. To explain
why functional superposition occurs in ergative languages only, Shaumyan cites the semiotic notion
of iconicity. The agent-patient relation is considered to be more "natural", i.e., with the agent as the
starting point of an action described by the sentence and the patient as the end point, the sentence is
seen as more accurately describing the "real world" than other types of constructions. An ergative
sentence has an absolutive-ergative hierarchy in which the primary term of the sentence represents
the patient and the secondary term represents the agent. Hence, ergative languages "need"
functional superposition to help them accurately represent reality. As a consequence of his
discussion of ergativity, Shaumyan defines accusative construction and ergative construction and
makes the following observations:
The accusative construction and ergative construction are two representations of the abstract
transitive/intransitive clause pattern: primary term + transitive predicate + secondary
term/primary term + intransitive predicate.
Primary term is represented by subject in the accusative construction and by absolutive in the
ergative construction. Secondary term is represented by direct object in the accusative
construction and by ergative in the ergative construction.
The abstract clause pattern and its representations are characterized by the Markedness Law
and the Dominance Law.
There is a correlation between the accusative and ergative constructions characterized by the
Law of Duality.
The primary and secondary terms may exchange their functions, so that the function of the
primary term is superposed onto the secondary term, and the function of the secondary term is
superposed onto the primary term. Shaumyan (1987:143)
With these observations, Shaumyan sees a way to make Equi and Subject Raising
applicable to both accusative and ergative languages. To do so, Equi and Subject Raising must be
generalized to Equi-Primary Term Deletion and Primary Term Raising. These new abstractions are
based on the premise that accusative languages and ergative languages are variants of the primary
term: secondary term construction. Shaumyan does not deny that there is a difference between
ergative and accusative languages, in some cases a very sharp difference. By examining the
genotype strata of both ergative and accusative languages, Shaumyan can formulate laws that
respect and model both the similarities and differences between language types. Doing so allows
him to merge the notions of syntactic ergativity and morphological ergativity into one notion, which he
calls the Integrated Theory of Ergativity.
Implications of the Integrated Theory of Ergativity for Linguistic Typology
The examination of significant anomalies (what Shaumyan calls linguistic phenomena)
advances linguistic theory. Linguistics should be a science whose theories, laws, and principles are
subject to empirical verification. Shaumyan does not simply create a theory, such as the Integrated
Theory of Ergativity and rest on its theoretical constructions. A substantial portion of chapter three
consists of various tests of his theory. The consequences of the Integrated Theory of Ergativity will
be summarized below.
The Integrated Theory of Ergativity suggests that the only processes that can be considered
to be formal ergative processes are those which correspond to ergative morphology.
When Shaumyan uses the term morphology, he means the term in the broad sense of any coding
device of a language. Shaumyan (1987:146) Ergative constructions can be found in languages that
do not have ergative morphology. Appealing to the Principle of Semiotic Relevance, we can see
32
that ergative constructions can only form a distinct grammatical category in a language if those
constructions are distinguished from other classes of constructions by at least one coding rule.
Shaumyan (1987:147) Shaumyan gives a Russian nominal example to show that languages can
have ergative patterns but not a true formal ergative construction category and then contrasts this
example with one from Dyirbal, which Shaumyan claims to have a true formal ergative category.
Ergativity is identical with agentivity if we define the meaning 'agent' as a class of meanings
characterized by the same coding devices as the syntactic function 'ergative'. Shaumyan
(1987:149)
This assertion is opposed to current views of the 'agent', which claim that the notion 'agent' is a
nonformal, purely semantic, notion. Shaumyan says that the current view is based on a
misunderstanding of idea of meaning. As we have seen before, Shaumyan has already noted two
types of meanings : lexical and grammatical. Lexical meanings are closer to reality than grammatical
meanings, which are based on methods of presentation. The ergative case is a formal mode of
presentation (a grammatical meaning) and not a lexical meaning. A noun in the ergative is,
grammatically, an agent. While the ergative case has a formal status in ergative languages, it does
not have a formal status in accusative languages. The current view of agentivity (citing Comrie as a
typical example) rests on a confusion of the lexical and grammatical, i.e., strictly formal, meanings of
terms.
Shaumyan looks at the Keenan-Comrie Accessibility Hierarchy, and, especially, Comrie's
definitions of syntactic ergativity and accusativity, to determine whether the Accessibility Hierarchy
can be considered universal. By examining the Hierarchy's predictions for Mayan, Shaumyan says
that, as stated, the Accessibility Hierarchy is not universal because it is based on nonuniversal
concepts. Shaumyan (1987:153) By using the Markedness Law, Shaumyan restates the Hierarchy
as an Applicative Hierarchy for both accusative and ergative languages. The restatement not only
accounts for those examples correctly accounted for by the original Keenan-Comrie Accessibility
Hierarchy but also correctly accounts for examples for which the Keenan-Comrie Accessibility
Hierarchy proves deficient. Shaumyan (1987:154)
Shaumyan once again issues a caution that he does not advocate replacing, e.g., the terms
subject and direct object with the terms primary term and secondary term but, rather, the concepts
subject and direct object with the concepts primary term and secondary term. It is replacing the old
concepts with new concepts that permits the accurate restatement of the Accessibility Hierarchy and
not merely the replacement of terms. Replacing the old terms with new terms accomplishes nothing
by itself.
Since the Law of Duality tells us that accusative and ergative languages are mirror images of
each other, ergative languages cannot have a true passive voice like accusative languages.
However, they must have a converse of the accusative's passive voice according to the Law of
Duality. In a pure ergative language, we can call the voice the antipassive. A split ergative
language can have the converse of an accusative's passive voice as part of its own accusative
subsystem.
What some linguists call the passive voice in ergative languages is a construction resulting
from demotion of ergative. In a true passive construction, the patient is topicalized when the
predicate is converted. In the demotion of the ergative, the patient is topicalized without the
predicate's being converted. Shaumyan refers to ergative demotion as quasi-passive constructions.
Shaumyan (1987:155) Whether we use the term passive to cover all these types of constructions or
not is a terminological matter of little concern; what is of importance is the proper grouping of
constructions. Accusative languages can have passive constructions with converted predicates, the
demotion of primary terms, and the promotion of the secondary term to the position of a primary
term, while an ergative language cannot. The Law of Duality requires that an ergative language
have an antipassive construction. On the other hand, ergative languages can have passive
constructions in which only the secondary term (agent) is demoted, while an accusative language
33
cannot. In accusative languages, the secondary term denotes a patient; therefore, demoting a
secondary term that is an agent is not possible.
Ergative languages view actions from the perspective of the patient (primary term); accusative
languages view actions from the perspective of the agent. Ergative sentences tend to present
actions as already accomplished (past tense) while accusative sentences tend to present actions
in progress (present tense). Ergative languages display various splits in case marking.
Shaumyan explains these splits as a special instance of the Law of Duality:
The marked term of the syntactic construction with the ergative tense/aspect corresponds to the
unmarked term of the syntactic construction with the nonergative tense/aspect, and the unmarked
term of the syntactic construction with the ergative tense/aspect corresponds to the marked term of
the syntactic construction with the nonergative tense/aspect; and, vice versa...
Shaumyan (1987:157-158)
In light of the proposed definitions of ergative construction and ergative language, the currently
accepted class of ergative languages might have to be revised. Georgian is an example.
Based on the Integrated Theory of Ergativity, Shaumyan suggests a regimen of fieldwork
to further confirm his findings and advance our understanding of ergativity. The interested reader is
referred to Shaumyan (1987:158-160)
An Informal Theory of Passivization
Supplementing the work begun in Descles, Guentcheva, and Shaumyan, 1985,1986,
Shaumyan next describes his theory of the passive.
Preliminary Terminology
A passive sentence with three parts: an nonagent, a passive predicate, and an agent, is called a
long passive sentence.
A passive sentence with two parts: a nonagent and a passive predicate, is called a short passive
sentence.
The active and the long passive form a correlation, which leads to the opposition active:passive.
Using John closed the door and The door was closed by John as examples of active and long
passive sentences, respectively, Shaumyan argues that John corresponds to by John, as the door
(secondary term in the active sentence) corresponds to the door (primary term in the long passive
sentence); was closed is the converse of closed. In English, there is also another opposition: long
passive: short passive.
According to Shaumyan, the definition of the short passive is ambiguous. It may either imply
an unspecified agent or may not imply an agent at all. The former implication is not a formal feature
of the language but depends exclusively on context. This is the case for languages other than
English. While some languages have markers for the designation of unspecified agents, the majority
of the world's languages do not; in AUG's terminology, the unspecified agent is a zero term. Even so,
the unspecified agent must be considered an "integral part of short passive". Shaumyan (1987:162)
The notion of a zero term (the unspecified agent) is needed to explain the mechanics of the
passive voice. In short passives, the unspecified agent is an inherent part of the meaning of the
predicate [b]ut certain contexts may superpose special functions onto predicates in short passives...
Shaumyan (1987:162) The passive predicate in a short passive sentence is a one-place predicate
34
that is derived by applying the converse of a two-place predicate to a zero term. This analysis gives
rise to a concern for the definition of the predicate in a long passive sentence. Shaumyan says that
we cannot treat the secondary term in a long passive sentence as a regular secondary term
because by-phrases are usually used as oblique complements. The function of a by-phrase in a long
passive sentence sharply differs from their function as oblique complements. Shaumyan (1987:163).
To demonstrate the sharp difference, Shaumyan uses the following example:
Mary was killed by John by the seashore.
The phrase by John is opposed to Mary and is the nucleus for by the seashore. The phrase by the
seashore can be replaced with other oblique term, e.g., in the room, but phrases such as this cannot
replace by John. So, we are forced to conclude that not every by-phrase functions the same way. By
John is a secondary term, but not an oblique complement, in the sentence above.
AUG's basic notion of functional superposition requires that we distinguish between the
inherent syntactic function of by-phrases and syntactic functions that may be superposed upon them.
The range of an inherent function is wider than the range of a superposed function. Since the role of
being an oblique complement is inherent in by-phrases, and the role of acting as secondary terms in
long passive constructions is a superposed function, the role of oblique complements is wider than
the role of being a secondary term in a long passive construction. When a function is superposed,
like that of a secondary term in long passive constructions, the linguistic unit does not change from
one function to another. Functional superposition is a notion that says that a linguistic unit may have
several functions at the same time. Therefore, when a by-phrase is superposed to a secondary term
in a long passive construction, it has both its inherent function (oblique complement) and its
superposed function (secondary term). Note that one functional superposition may "cause" another.
The predicate in a long passive construction is, inherently, a one-place predicate; when a by-phrase
is superposed to a secondary term, the predicate is also superposed to a two-place predicate.
Prepositional phrases in active constructions can be seen to have similar shifts. For
example, to sit on the beach has a one-place predicate followed by a prepositional phrase. However,
in to agree on the terms, the preposition has lost its concrete meaning and becomes a marker, that
is, a word introducing, in this case, a secondary term. The prepositional phrase has become a
secondary term and the one-place predicate to agree has become superposed to a two-place
predicate. Shaumyan also cites examples from languages other than English. Shaumyan (1987:163-
165)
In order to discuss the short passive, Shaumyan first introduces the notion of
derelativization, which is a notion brought into linguistics from logic and may turn out to be useful in
linguistics, as well. Shaumyan (1987:165) He quotes Quine's discussion from Word and Object:
Commonly the key word of a relative term is used also derelativized, as an absolute term to
this effect: it is true of anything x if and only if the relative term is true of x with respect of at least one
thing. Thus, anyone is a brother if and only if there is someone of whom he is a brother. Where the
relative term is a transitive verb, the corresponding absolute term is the same verb used
intransitively.
Relative terms also combine with singular terms by application, to give absolute general
terms of a composite kind. Thus, the relative term 'brother of' gives not only the absolute general
term 'brother' but also the absolute general term 'brother of Abel'. Similarly, the relative term 'loves'
gives not only the absolute general term 'loves' (intransitive) but also the absolute general term 'loves
Mabel'. Again the relative term 'at' gives the absolute general term 'at Macy's'.
Shaumyan (1987:165)
Using Quine's concept, Shaumyan characterizes the meaning of passive predicate in short passive
constructions as the derelativized converse of an active predicate. Derelativization explains why
35
short passives are often ambiguous: since passive predicates are one-place predicates, their
converse meaning must be supported either by the context or, as in long passive constructions, by
agent modifiers; without this support, the converse meaning is lost. Shaumyan (1987:165)
Shaumyan proposes a hypothesis about the inherent function of the passive:
The inherent function of passive is the conversion of the active predicate and the
derelativization of the converse predicate, which involves the superposition of the function of the
secondary term onto the predicate modifier.
Shaumyan (1987:166)
The predicate modifier, in this case, is the linguistic unit, such as a by-phrase, which has predicate
modification as its inherent function. The function of being a secondary term is superposed on top of
the inherent function of predicate modification, which, it should be stressed, is not destroyed, but only
"masked", as it were.
Shaumyan supports his hypothesis with a number of observations. Long passive
constructions, in those languages which permit them (not all do), are much less common (therefore,
much less important) than short passive constructions. Short passive constructions topicalize non-
agent terms (making them primary terms), remove agents, and change active predicates into their
derelativized converses. A long passive the agent is mentioned, but the nonagent is topicalized.
Even though topicalization of the nonagent term is the result of converting the active predicate,
nonagent topicalization is not necessarily connected with the passive voice; it may occur in active
constructions as well. This means that there are two types of nonagent topicalization, one that results
from the conversion of the active predicate, and one that occurs in active constructions. Some
languages, like English, did not always have long passive constructions. Long passives only became
a feature in English in the sixteenth century, when the marker by became dominant and replaced
many older agent markers.
Turning to a further description of the relation between the active and the passive,
Shaumyan notes that passive predicates are derived from active predicates, meaning that the active
voice is the basic voice. In most languages, the passive voice is morphologically marked; even in
languages where this is not the case (Mandarin), passive constructions are marked by the use of
particles and a change in word order. In most languages, the subject of a declarative construction
cannot be referential-indefinite. Speakers must resort to existential-presentative constructions, like
there is a...
In transitive active constructions, the primary term (the agent) is often determined while the
secondary term (the patient) may or may not be. In passive constructions, the primary term (the
patient) is usually determined while the secondary term (the agent) may or may not be determined.
When the secondary term in an active construction is determined, it is simple to construct the
corresponding passive construction. When the secondary term of an active construction is
undetermined, it is not always possible to create a satisfactory passive construction. Passive
constructions have a narrower range than active constructions. Shaumyan notes that there is an
equivalence but not an identity between active and long passive constructions. By equivalence,
Shaumyan means that there are some essential properties of the meaning of long passive
constructions and the meaning of corresponding active constructions that are invariant under the
operation of the conversion of predicates. Shaumyan (1987:169) For example, the topic of an active
sentence and a long passive sentence derived from it are not the same.
A change in topicalization is always accompanied by a change in definiteness or
indefiniteness. The sentences An automobile was stolen by the man, The automobile was stolen by
a man, and The man stole an automobile have different meanings which are reflected by the articles
the and an, even though there is an underlying invariant relation between man (the agent) and
automobile (the nonagent) in each of the sentences.
36
Passive predicates tend to denote a state resulting from a past action. This leads us to
notice that in some languages, such as contemporary English, some passive phrases are
ambiguous (not clearly referring to a process or a state); in other languages, e.g., contemporary
Russian, passive constructions which refer to processes have fallen into disuse. Passive
constructions are presented from the point of view of the nonagent. Its focus is on the effect of the
process on the nonagent rather than on the process itself. The active construction expresses an
action from the point of view of the agent, so it is not necessary for the passive to do so.
Many languages passivize the tertiary term as well as the secondary term. In some
languages, like English, tertiary term passivization is relatively rare, being restricted to a small
number of predicates. In other languages (Shaumyan cites Malagasy), tertiary term passivization is
common. However, regardless of the frequency of the occurrence of tertiary term passivization in a
language, it is always subordinate to the passivization of secondary terms, leading to the Condition
on the Passivization of Tertiary Terms:
Any language that passivizes tertiary terms must passivize secondary terms, but the reverse
is not true; the passivization of secondary terms does not presuppose the passivization of tertiary
terms.
Shaumyan (1987:171)
At this point in the discussion of passive constructions, Shaumyan defines both the long and
short passive constructions. These definitions will be set forth at length, but it is worth noting that
Shaumyan mentions that there are apparent contradictions in the definitions. He resolves the
apparent contradictions after stating the definitions.
The short passive construction is defined by the following conditions:
1) A one-place passive predicate is derived from a two- or three-place predicate in two steps:
i) formation of the converse of the active predicate;
ii) derelativization of the converse predicate by applying it to a zero term
denoting an unspecified agent.
2) The passive predicate is applied to a noun phrase that serves as the primary term of the short
passive construction.
3) The primary term of the short passive construction denotes a nonagent and is a functional
counterpart of the secondary or tertiary term of the corresponding active construction.
The long passive construction is defined by the following conditions:
1) The passive predicate is derived in two steps, as in the short passive construction.
2) The passive predicate first is modified by a noun phrase that serves as an oblique term and then
is applied to a noun phrase that serves as the primary term.
3) The primary term of the long passive construction denotes a nonagent and is a functional
counterpart of the secondary or tertiary term of the corresponding active construction.
4) The function of the secondary term is superposed onto the oblique term and the function of the
two- or three-place converse predicate is superposed onto the passive predicate. As a result of
the superposition, the oblique term of the long passive construction turns into a functional
counterpart of the primary term of the corresponding active construction.
37
5) The passive predicate takes on the function of a two- or three-place converse predicate.
Shaumyan (1987:171)
Passive predicates, according to the definitions above, have the same form in both short and long
passive constructions, but they are described as functioning in different ways. The short passive
predicate functions as a one place predicate which results from the application of a converse
predicate to a zero term and the long passive predicate is said to be a one-place predicate which
functions as a two- or three-place predicate. Furthermore, the agent in a long passive construction
was said to be an oblique term that functions as a secondary term.
What Shaumyan refers to as the paradoxical structure of the passive is a direct result of the
attempt of a language to create symmetrical linguistic structures by functional superposition. The
functions of two and three place predicates are superposed on a one-place predicate and the
function of a secondary term is superposed on the oblique term. Every language has the potential of
creating symmetrical linguistic structures, even though it is not necessary for a language to realize
them. Symmetrical structures must meet two conditions:
they must have converse predicates, and
they must have the same number of terms.
Short passive constructions are the symmetrical counterpart of active constructions. In Old and
Middle English, for example, the agent could be marked by the dative case or by certain
prepositions. In the course of linguistic change, by replaced all of these structures, turning short
passives into long passives. The passive voice, and especially the long passive construction, are a
linguistic luxury. Many languages do not have passive constructions at all, even though some have
verbal adjectives with a passive meaning.
Impersonal Passive Constructions
The fact that passive predicates in some languages, like German, can be derived from
intransitive active predicates may, at first, be seen as a counterexample to Shaumyans informal
theory of passivization. To avoid any possible contradiction, he posits the existence of a zero dummy
term. Shaumyan considers the postulation of an additional abstract entity to be justified in this case
because it is well motivated by directly observable empirical properties of sentences and the
operation of conversion Shaumyan (1987:173). The introduction of a zero dummy secondary term
into active intransitive sentences creates a well-motivated analogy between the derivation of passive
predicates from active two-place predicates Shaumyan (1987:173). To counter the argument
that impersonal passive constructions do not satisfy the conditions implied by the definition of the
passive, Shaumyan relies on the Principle of Polymorphism. The principle is simple and can be
stated:
There are different ways of constructing the same sentence, and not all ways are possible for
every sentence.
Shaumyan (1987:174)
Shaumyan sees that the key to understanding the passive is that passive constructions can always
be created by applying a passive predicate to the terms which are their operands. Shaumyan
(1987:174) For him, the Principle of Polymorphism solves some of generative transformational
grammars difficulty with accounting for passive constructions; generative transformational grammar
only permits one method of creating passive constructions. AUG does not so limit passivization.
38
Conversion can be objected to by pointing to some languages, like Polish, which have
passive constructions that do not promote a patient to a primary term. Shaumyan reminds us that the
application of a two-place predicate to a term gives us a one-place predicate. Languages have one
of two options: either use a transitive sentence with a transitive predicate and a patient or use an
intransitive sentence with an intransitive predicate, i.e., a one-place predicate and term. Polish
decided to use the latter construction. Shaumyan notes:
The possibility of this choice can be explained by the fact that the crucial function of
passivization is demotion or suppression of the agent, rather than promotion of the patient to the
position of the primary term; passive constructions are used when the agent is either unknown or
unimportant.
Shaumyan (1987:175)
Passive and Antipassive
Ergative languages cannot have a passive voice. Accusative language have the opposition
active : passive; pure ergative languages have the opposition ergative : antipassive. Even though an
antipassive construction may look like an active construction, there is a fundamental difference
between them. The active construction is fundamental and normal, while the antipassive
construction is derived and stylistically marked. Shaumyan (1987:176). Most ergative languages do
not have antipassive constructions; they only have ergative constructions. It should be noted that not
all accusative languages have passive constructions. It is, however, much more common for ergative
languages to lack antipassive constructions than for accusative languages to lack passive
constructions.
Antipassive constructions topicalize the agent. Antipassive constructions are used in those
languages with ergative constructions in which the topic is the nonagent rather than the agent. In
most ergative languages, however, superposing the functions of the primary term on the secondary
term results in topicalizing the secondary term and, since the secondary term denotes the agent, the
agent becomes topicalized. Therefore, most ergative languages do not need the antipassive. These
languages also do not need a passive construction because they already have an analogue to the
passive construction by denoting the agent without conversion of the ergative predicate Shaumyan
(1987:176).
There are two types of antipassive construction, the short and long antipassive. Shaumyan
defines antipassives as mirror images of passives. Short antipassive constructions meet the
following conditions:
1) a one-place antipassive predicate is derived from a two- or three-place ergative predicate in two
steps:
i) formation of the converse of the ergative predicate;
ii) derelativization of the converse of the ergative predicate.
2) The antipassive predicate is applied to a noun phrase that serves as the primary term of the short
passive construction.
3) The primary term of the short passive construction denotes an agent and is a functional
counterpart of the secondary term of the corresponding ergative construction.
The long antipassive must meet these conditions:
1) The antipassive predicate is derived in two steps, as in the short antipassive construction.
2) The antipassive predicate is first modified by a noun phrase that serves as an oblique term and is
then applied to a noun phrase that serves as a primary term.
3) The primary term of the long antipassive construction denotes an agent and is a functional
counterpart of the secondary term of the corresponding ergative construction.
39
4) The function of the secondary term is superposed onto the oblique term; the function of the two- or
three-place converse predicate is superposed onto the antipassive predicate. As a result of the
superposition, the oblique term of the long antipassive construction turns into a functional counterpart
of the primary term of the corresponding ergative construction.
Shaumyan (1987:176-177)
Alternative Theories of Passivization
Shaumyan briefly compares his theory of passivization with three others: generative
transformational grammar, relational grammar, and the demotion theory of passivization. The reader
is referred to the text Shaumyan (1987:177-193) for the details of Shaumyans criticisms. I will only
sketch them here.
Generative Transformational Grammar
Shaumyan criticizes generative transformational grammar for defining passivization solely in
terms of word order. He sees that this definition raises three problems:
Some languages, such as Russian, use the same word order for both active and passive
sentences.
Generative transformational grammar must treat what is really the same in different languages
as different rules, causing generative transformational grammar to miss essential linguistic
principles.
Generative transformational grammar claims that passive constructions can only be derived from
active ones. Shaumyan sees this as a consequence of the reliance on the Autonomous Syntax
Hypothesis, which claims that grammatical morphemes are for the most part meaningless and
are inserted for purely formal purposes.
To correct the last problem, Shaumyan posits the following unnamed law:
If a language has passive constructions with agentive phrases, it must have passive
constructions without agentive phrases, but the reverse is not true: if a language has passive
constructions without agentive phrases, it may or may not have passive constructions with agentive
phrases.
Shaumyan (1987:178)
Relational Grammar
Relational grammar and generative transformational grammar both share the abstract notion
of an underlying syntactic structure, called deep structure in generative transformational grammar
and the initial stratum in relational grammar. There are some technical differences, but these notions
are similar in the sense that they are the fictitious entities from which empirical linguistic structures
(surface structures in generative transformational grammar, and the final stratum in relational
grammar) are derived. Generative transformational grammar uses a dynamic meta-language for the
derivation (transformations), while relational grammar uses a static meta-language (grammatical
relations between strata).
It is obvious that Shaumyan thinks more highly of relational grammar than generative
transformational grammar. Relational grammar explains passivization in functional terms;
40
furthermore, relational grammar has predictive power. Shaumyans criticism of relational grammars
explanation of the passive is that relational grammars view of passivization is not a universal one. It
is based on the syntax of accusative languages, rather than derived from an examination of all
languages. Shaumyan points to the fact that even for accusative languages, relational grammars
notion of passivization leads to problems.
The Demotion Theory of Passivization
Shaumyans comments on this theory are very brief. Essentially, he sees the demotion
theory as an attempt to explain passivization without recourse to any abstract notions. Shaumyan
feels a need to use abstractions like the zero dummy term to make the explanation of passivization
smoother. The demotion theory may have gone too far in its otherwise laudable quest to eliminate
unnecessary abstractions.
The Formalism of Applicative Grammar
AUGs formalism derives from combinatory logic, a second order logic, and categorical
grammar. [T]he linguistic theory based on the resulting formalism [Shaumyan] call[s] applicative
grammar Shaumyan (1987:193). We will see later that Shaumyans theory does not derive from
an application of combinatory logic to linguistic reality. Rather, AUG gives enormous weight to the
notion of combination. Combinatory logic is a much better formal fit for a linguistic theory which is
based on combination than is first order logic which forms the basis for the formalism of
Generative Transformational Grammar (GTG). When thinking about AUG, one fact must always
be kept in mind: Applicative Universal Grammar is a theory about behavior. An essential
difference between AUG and GTG is that AUG considers word behavior while GTG considers
only linguistic states, which are mathematical (logical) constructs. This difference makes a
substantial difference in the formalism of the two theories. While not specifically mentioned in the
book, it is likely that AUG cannot properly be represented using GTGs formalism.
Basic Rules, Terminology, and Notions
There are three basic elements (categories or types) of expressions which all languages
must have:
terms;
sentences;
operators
Terms, as we have seen, are the names of objects (broadly conceived), sentences are the names
of situations, and operators are expressions that combine expressions to form other expressions
Shaumyan (1987:194)
What makes AUG an applicative grammar, or, rather, what makes any grammar an
applicative grammar? Shaumyan notes:
In applicative grammar, operators are connected with one another by a network of formal
definitions, which eventually reach the ultimate definientia - term and sentence.
Shaumyan (1987:194)
The formal definitions of which Shaumyan speaks are actually definitions of relations. The quote
above hides the activity that goes on in the applicative parsing of a sentence. To understand what
we will see in upcoming pages, we must understand the meaning of the phrase eventually reach
the ultimate definientia - term and sentence. A genotype of a sentence is a description of the
relations that make a particular group of words into a single sentence. To parse a sentence, for
41
AUG, means to display those relationships. So, what I mean by genotype parse of a sentence is
the process by which a sentence in natural language is reduced to a description of the relations
between the words in that sentence to each other and to the sentence as a whole using the
notions and notation of AUG. A genotype parse is completed when all relations within the
sentence are accounted for. The result of a genotype parse of a sentence is what is referred to as
an s-type or, simply, s. The s-type is the ultimate definiendum of the sentence because it
describes the interrelation of all the parts of the sentence.
What is the ultimate definiendum of a term? A term, for AUG, is not necessarily a single
word, although, of course, it may be. As I mentioned before, term and noun are not synonymous.
A term may be composed of one or more nouns, one or more modifiers, such as adjectives, and,
perhaps, other words such as conjunctions. For AUG, the noun is not the constituent of the s-
type, the term is. Of course, when a term and a noun are co-extensive, i.e., when the noun is the
only element of the term, the noun may appear to be a constituent of the s-type, but it is only an
appearance and not an underlying reality. When a noun is not the only element of a term, the
noun will function as a nucleus for the term, i.e., the word (assuming there is only one noun in the
term) to which all other elements of the term relate. When a term consists of more than one word,
each word combines with the nucleus of the term. The result of the combination is still a term. It is
the entire term, i.e., the relation of all elements of the term to the nucleus of the term that is a
constituent of the s-type. AUG manipulates the entire term as a single unit. A term which is
composed of ten words is considered to be a single term; the term is no longer treated as ten
words for the purposes of a genotype parse but, rather, as one block. Adding or subtracting
words from a term changes the semantic value of the term but does not change the fact that AUG
still deals with the term as a single unit. For example, each of the following phrases analyzes as a
single term:
car
a car
a yellow car
a big, yellow car
a big, shiny, yellow car
a big, new, shiny, yellow car
a big, new, shiny, expensive, yellow car
Example 3.1
Each term has a different set of relations between its elements and must be analyzed in such a
way as to make those relations clear. However, at the end of the analysis, each phrase will have
the same analysis: t or term. Each different phrase in Example 3.1 is a term because each
phrase functions as a term, i.e., each phrase functions as a single unit which is a term.
In order to combine expressions, there must be some device to combine expressions.
These formal devices are called operators. Operators are, themselves, expressions. Operators
combine expressions to produce new expressions. That is their sole role in AUG. In order to
understand the notion of operation, a bit of logic is required. In 1926, Moses Schnfinkel showed
that all functions can be reduced to one-place functions, an idea that was rediscovered by Haskell
Curry, who developed combinatory logic.
In AUG, all operators are one-place operators. At first, it may seem contradictory
because many operators, predicates especially, are referred to as two- or three-place operators.
While this is the case, an n-case operator is only applied once. Recall that the role of an operator
is to take an expression and make a new expression from it. That new expression may be
another operator, a term, or a sentence. If the result of applying an n-place operator to an
expression is another expression that is an operator, the new operator will be an n-1-place
operator. While Shaumyan does not use the word, I will call the number of places that an operator
has its arity; so, for example, a 3-place operator will be said to have an arity of three.
42
An example may help clarify this discussion. We will use the sentence
Jim showed Mary pictures
Example 3.2
Shaumyan(1987:194) and use the notation o
n
to refer to an n-place predicate (p) or term (t).
Before explaining the example, I must say a word about the notions of primary, secondary, and
tertiary terms. Primary, secondary, and tertiary terms are simply terms and are no different than
any other types of terms. Primary, secondary and tertiary refer to roles in a sentence. No term is
inherently a primary, secondary or tertiary term; a term takes on one of these functions when it is
constituent of a particular sentence. So, in the sentence Jim showed Mary pictures, Jim is a
primary term, Mary is a tertiary term, and pictures is a secondary term, but in the sentence Mary
showed Jim pictures, Mary is the primary term, Jim is the tertiary term, and pictures is a
secondary term. A primary term can be thought of as a subject, a secondary term as a direct
object, and a tertiary term as an indirect object as long as it is recalled that term refers to function
and not to syntactic type.
Showed is a 3-place predicate because showed is an operator that takes three terms
and, ultimately, reduces them to a sentence. It does so in this way. First, showed is applied to the
tertiary term, Mary. The new expression produced by an operator is called the resultant. The
resultant of applying the operator showed with type p
3
to Mary with type t
3
is a two-place operator
(p
2
) showed Mary. The operation of applying the predicate to the tertiary term produced a new
predicate with a reduced arity. It is important to note that the 2-place predicate is not showed but
showed Mary. The operation has combined the 3-place predicate and the tertiary term into a new
predicate. Showed Mary is then applied to pictures (t
2
) to produce showed Mary pictures with
type p
1
. The 1-place predicate, showed Mary pictures, is then applied to the primary term, Jim (t
1
)
to produce an s-type, which is the sentence, Jim showed Mary pictures.
Several things should be noted about our example. First, the notation we used is not
AUGs formalism. We will see that formalism shortly. Second, a multi-place operator is actually a
series of 1-place operators, repeatedly applied. Third, operation is an act of reduction; we can
think of it in terms of energy reduction. Not all operators operate on terms. An operator operates
on an operand. An operand is an expression. Operators can, in theory, operate on any
expression to produce any other expression. Part of AUGs formalism is a shorthand for
describing the operator, the operand and the resultant. Let us now turn to the formalism for
describing operators.
If XY is an expression in which X is an operator and Y is an operand and if the
expression XY belongs to a type v and the expression Y belongs to type u, then X is an operator
which changes an expression of type u into an expression of type v. The formal notation for this is
Ouv. Think of operation, or, as Shaumyan refers to it, operationality Shaumyan (1987:195), as a
black box. U goes in the left side; v comes out the right side.
Figure 3.1
43
Shaumyan refers to Ouv as the operationality primitive, which reads the type of operators from u
into v Shaumyan (1987:195). The operationality primitive is the most general description of
operationality in AUG. The rule for classifying XY is:
If X is in Ouv and Y is in u, then XY is in v.
Shaumyan creates a calculus from this by assuming that there are primitive operators (O-types):
c
1
, c
2
, ... Rule T is his formal definition of an O-type.
RULE T: a. Primitive types c
1
, c
2
, ... are O-types.
b. if u and v are O-types, then Ouv is an O-type.
Shaumyan (1987:195)
Rule Tb is the notion of combination that we briefly discussed. Rule T says that O-types
(operators) are either primitive types of operator (as yet undefined) or combinations of primitive
types. In Rule T and in the rest of this introduction to AUG, the word type will be used in a special
way. Type means a class of operators. It should be noted that Shaumyan considers sentences
and terms to be 0-place operators for the sake of generality Shaumyan (1987:194). Rule T can
also be described by a tree diagram:
u v
---------------------
Ouv
Figure 3.2
As we have seen, in AUG, every operator has an operand and a resultant. In the operationality
primitive (Ouv), u is the operand and v is the resultant. Briefly, a formula like Ott is an instance of
the operationality primitive. u is the first t, meaning that this operator takes a term as an operand.
v is the second t, meaning that this operator produces a term. Ott, then, is an operator that takes
a term and produces another term. The two ts in Ott are to be understood as two semiotically
different ts. u and v each may be simple, in the sense that each may consist of one word, or each
may be complex, in that they consist of many words.
The notation yX is the notation for the relation belongs to. yX reads expression X belongs
to type y. Using this notation, we can take the rough tree diagram in Figure 3.2 and make it a
rule:
RULE E: OuvX uY
v(XY)
v(XY) is the resultant of applying X to Y as indicated by the horizontal line. Rule E reads if
expression X belongs to O-type Ouv, and expression Y belongs to O-type u, then expression XY
belongs to O-type v Shaumyan (1987:196). Shaumyan notes two things. First, I want to stress
the relativity of the concepts operand and resultant: the three expressions X, Y, and XY are all
operators. But besides, Y is an operand and XY is a resultant of X. Shaumyan (1987:196).
Second, the notation that has just been developed is arbitrary. It is adopted because it is
unambiguous and convenient. It is possible and acceptable to describe Rule E and any other rule
in AUG by any other notation that is both unambiguous and convenient.
Rule E has an obvious consequence known as Rule E1. It can be stated:
RULE E1: v(XY), uY
OuvX
44
It reads if the resultant of the application of expression X to expression Y belongs to O-type v,
and expression Y belongs to O-type u, then expression X belongs to type Ouv Shaumyan
(1987:196). Rule E1 defines an operator in terms of its operand and the resultant. Shaumyan
derives another related rule, Rule E2, which defines an operand and a resultant in terms of their
operator:
RULE E2: XY, OuvX
uY, v(XY)
If we construct an expression XY by Rule E and if X is in type Ouv, then Y must be in O-type u
and XY belongs to the O-type v.
Rule E provides us with a definition of well-formedness with respect to type. Here is
Shaumyans definition:
I call an expression X well formed with respect to type if it is constructed from
expressions Y
1
, ... Y
n
by Rule E.
Shaumyan (1987:197)
We are now in a position to reduce the sentence in Example 3.2 to its ultimate
definiendum using AUGs formalism.
1. OtOtOts showed t Mary
2. OtOts (showed Mary) t pictures
3. Ots ((showed Mary) pictures) t Jim
4. s(((showed Mary) pictures) Jim)
Example 3.3
Example 3.3 is the reduction of the sentence in Example 3.2 in Shaumyans formal
calculus. The lines in Example 3.3 are not part of AUGs formalism and are only numbered for
convenience. Indenting the lines for readability is optional but common. Each line consists of an
operator and operand, even line 4. Recall that AUG considers a s-type and a term to be zero-
place operators, so the line contains no operand. Starting with line 2, each line is both an
application of an operator to an operand and the resultant of the application of the operator to its
operand on the line above it. The operator always appears on the left; the operand always
appears on the right.
O-type notation takes a little while to get used to, but, after a while, it becomes easy to
read and, more to the point, unambiguous. Example 3.3 uses five different symbols: OtOtOts,
OtOts, Ots, t, and s. The simplest are t and s, which stand for term and sentence (or s-type)
respectively. Ots, OtOts, and OtOtOts stand for, respectively 1-, 2-, and 3-place predicates. All O-
type symbols start with the capital letter O. There are two fields following the O, which may be
thought of as the operand field and the resultant field. We need to know to what the operator
applies and, once the operation has occurred, what will be produced. Schematically, we have:
O operand field resultant field
Figure 3.3
Figure 3.3 can be read Whenever the operator is applied to whatever is in the operand field,
whatever is in the resultant field will be produced. OtOtOts, therefore, can be read as:
whenever this operator is applied to t, OtOts will result. OtOts says: whenever this operator is
applied to t, Ots results. Ots says: whenever this operator is applied to t, s results, i.e., an s-type
45
or sentence is produced (the ultimate definiendum of the sentence is reached). Informally,
OtOtOts is called a 3-place predicate because it takes three terms (t) to reach a sentence and,
analogously, OtOts is a 2-place predicate and Ots is a 1-place predicate. A 1-place predicate is
also called an intransitive predicate; a 2-place predicate is also called a transitive predicate; a 3-
place predicate is also called a ditransitive predicate.
What a symbol like OtOtOts does not tell us is which word has this type. The symbol is
the result of applying rules T, E, E1, and E2 to words in a sentence to determine their types.
Words may have different types in different sentences; see, e.g., the discussion on functional
superposition above. Using the sentence reduction in Example 3.3, let us take a line and
examine what it says. Line 1 is:
OtOtOts showed t Mary
Example 3.4
Example 3.4 shows that it is not enough to merely display the types in the reduction; we must
also display the words which we claim are in those types. Here, the word showed is type OtOtOts
and the word Mary is type t. The horizontal line in Example 3.4 indicates that showed is being
applied to Mary. Example 3.4 does not show the resultant of the application but only indicates
that there should be a resultant.
OtOtOts showed t Mary
OtOts (showed Mary) t pictures
Example 3.5
Example 3.5 shows the resultant of the application of showed to Mary that we saw in Example
3.4. Notice two things. The resultant of the operation is, itself, an operation as indicated, once
again, by the horizontal line. The operator on the second line is not showed; it is (showed Mary)
in type OtOts. (showed Mary) is the 2-place predicate (OtOts) produced by the application of the
3-place predicate showed to the term Mary in the first line. Example 3.5 demonstrates the
difference between AUGs notion of the predicate and the traditional notion of the verb. A
predicate is an operator that may be a combination of a verb and other types of words that are
treated as a single unit. This may take a bit of getting used to but it is a concept that is absolutely
vital to understanding AUG. All operators, when applied to an operand, result in a combination of
expressions, which combination is treated as a new, single expression.
Example 3.6 continues in the same vein as Examples 3.4 and 3.5:
OtOtOts showed t Mary
OtOts (showed Mary) t pictures
Ots ((showed Mary) pictures) t Jim
Example 3.6
The 1-place predicate resulting from the application of (showed Mary) to pictures is ((showed
Mary) pictures) in type Ots. The parentheses show which expression has been applied to which
other expression. In a sentence as simple as the one we are examining, one might be tempted to
eliminate the parentheses as a nuisance, but in more complex sentences, the parentheses add
structure to what, without them, might be a difficult notation to read. Example 3.6 emphasizes the
fact that it is the entire resultant of an operation (assuming that the resultant is something other
than a 0-place operator) which is applied to an operand.
OtOtOts showed t Mary
OtOts (showed Mary) t pictures
46
Ots ((showed Mary) pictures) t Jim
s(((showed Mary) pictures) Jim)
Example 3.7
Example 3.7 shows the entire reduction. The reduction stops at the fourth line when an
s-type is produced. Being a 0-place operator, an s-type has no operand and cannot, therefore, be
applied to anything. Therefore, there is nothing further to do.
The Syntactic System of Applicative Universal Grammar
The syntactic system of applicative grammar consists of the following notions:
1. primitive O-types: t, s;
2. rules for constructing composite types from the primitives:
a. primitive O-types t, s are O-types,
b. if x and y are O-types, then Oxy is an O-type;
3. expressions belonging to O-types;
4. rules for constructing expressions: Rule E, Rule E1, Rule E2;
5. nine combinators (or combinatory operators): I, C, C
*
, W, B, S, K, , ;
6. rules for applying combinators: reduction rules and expansion rules;
7. replacement rules;
8. deductive processes: expansion and reduction.
Shaumyan (1987:198)
We have already discussed one through four on this list. We will see the rest, in more or less
detail, in the rest of this section. We will begin with Shaumyans definitions of 1-, 2-, and 3-place
predicates and primary, secondary, and tertiary terms.
DEFINITION 1. If expression XY belongs to type s, expression Y belongs to
type t, and expression X belongs to type Ots, I call X a one-
place predicate and Y a primary term.
DEFINITION 2. If expression XY belongs to type Ots, expression Y belongs to
type t, and expression X belongs to type OtOts, I call X a two-
place predicate and Y a secondary term.
DEFINITION 3 If expression XY belongs to type OtOts, expression Y belongs
to type t, and expression X belongs to type OtOtOts, I call X a
three-place predicate and Y a tertiary term.
Shaumyan (1987:198)
Shaumyan next mentions the primary term: secondary term opposition. This opposition forms the
nucleus of a sentence, if there is a 2-place predicate. The primary term exists within this
opposition if there is a 2-place predicate in the sentence. If the predicate is a 1-place predicate,
the primary term exists outside this opposition.
AUG has the notions of head, modifier, and compliment, which are related to each other
according to Definition 4:
DEFINITION 4. Let AB be a well-formed expression. It follows from Rule E that A belongs to O-
type Oxy, B belongs to O-type x, and AB belongs to O-type y. Either A or B can be considered
the main constituent of expression AB, called its head: if x y, the head is B, and if x [~] y, the
47
head is A. If B is the head, A is called a modifier of the head. If A is the head, B is called a
complement of the head. The term dependent denotes modifiers and complements together.
Shaumyan (1987:199)
Since modification is such an important notion in linguistics, let us look at the idea a little deeper.
The notion of modification in AUG is the notion that expressions can be composed of one or more
parts. If there is more than one part to an expression, each part plays a particular role. The head
of an expression is the central part of the expression, the nucleus, if you will. All other words in
the expression are related to the head in one of two ways: by modification or by complement.
Modification combines with the head of an expression and makes a new expression. A
complement is, essentially, an operand for the head of an expression.
Shaumyan uses an example to clarify the notions of head, modifier, and complement.
Bill bought new books
Example 3.8
Here is the tree diagram for Example 3.8, with line numbers as before:
1. Ott new t books
2. OtOts bought t (new books)
3. Ots (bought (new books)) t Bill
4. s((bought (new books)) Bill)
Example 3.9
Example 3.9 uses a new symbol: Ott. Ott, also known as a term modifier, is an operator that
takes a term (t) and makes it into a new term (t). In Example 3.9, the term modifier in line 1 is the
word new and the term is the word books. Applying new to books results in the new term (new
books). Recall the discussion of Example 3.5: (new books) is the new term, not books.
books is the head of the expression (new books); new, by Definition 4, is the modifier of
the head, books. In line 3, the expression (new books) does not modify bought. bought is the
head of the expression (bought (new books)). Notice once again that it is not books which is the
complement of bought; rather it is the expression (new books), the term that resulted from the
application of new to books. On line 2, bought is applied to the term (new books). The application
results in the formation of a 1-place predicate (line 3) which is then applied to another term, Bill.
The result is an s-type but notice that the head of the s-type is ((bought(new books)) and not
simply bought. It is the result of line two found on line 3. Expressions are being combined
throughout the tree diagram. A combined expression becomes a new expression; the entire
combined expression is one unit. It is that unit which is either the head of an expression, the
modifier of the head of an expression, or the complement of the head of an expression.
The need to use the combined expression as a head, modifier, or complement explains
why line 1 consists of the application of a term modifier to a term and not a predicate to a term. In
order to apply a predicate to a term, the term must be correctly formed, i.e., the ultimate
definiendum of the term must be reached. For complicated terms (and for complicated predicates
and other expressions as well), the entire term must be resolved before an O-type can be applied
to it.
Definition 5 introduces a bit of shorthand, the adjoined type symbol.
48
DEFINITION 5. A type symbol is called adjoined if it is introduced into the syntactic system by a
definition of the form z = Oxy where z denotes and adjoined type symbol and Oxy denotes an O-
type where x and y are either other adjoined type symbols or t or s.
Shaumyan (1987:199)
There are four basic adjoined type symbols, a, m, p
1
, and c. These type symbols are shorthand
for the following:
a = Ott (a term modifier)
m = Oss (an operator which makes a sentence into a new sentence, i.e., a sentence modifier, like
suddenly, in Suddenly, the lights went out.)
p
1
= Ots (a 1-place predicate)
c = Ost (an operator which makes a sentence into a term, e.g., that in That the Phillies would lose
was a foregone conclusion)
We can use this shorthand to construct symbols for combined types, e.g., instead of the notation
OtOts for a 2-place predicate, we can equivalently say Otp
1
. We can introduce as many adjoined
type symbols as we wish but we must be able to define what any introduced type symbol means.
The step by step definition of an adjoined type symbol is called a definitional reduction Shaumyan
(1987:200).
Shaumyan describes an alternate method of constructing tree diagrams which is
essentially the reverse of the one shown here. I will leave it to the interested reader Shaumyan
(1987:200-201)
Superposition
The notion of superposition is unique to AUG. Being unique, it may at first be difficult to
understand so we will begin slowly. Superposition is a notion borrowed from quantum physics. In
quantum physics, electrons can be in several energy states at the same time. The notion of
superposition is a notion about types, i.e., linguistic categories. It contrasts with the common
grammatical notion of parts of speech, which is a surprisingly underdeveloped notion, largely
inherited from the late Classical Greek grammarians. Let us examine that idea so that we can see
just what superposition contrasts with.
A part of speech is the name for the syntactic role played by a word, e.g., cat is a noun
and play is a verb. Traditional grammar considers that a word has an inherent type; in other
words, cat is always a noun and play is always a verb. Of course, a moments reflection will show
cracks in this scheme. Play certainly functions as a verb, as in
Dennis Rodman plays power forward
Example 3.10
but it can also function as a noun, as in
That was an impressive play that Dennis Rodman just made.
Example 3.11
Plays in Example 3.10 and play in Example 3.11 play different roles in their sentences. Plays is
a verb and play is a noun. They are different parts of speech and, therefore, different words, or so
the traditional theory goes. Traditional grammatical theory emphasizes the differences between
words like plays and play but spends less time discussing their similarities. Plays and play
49
obviously come from the same root and have related meanings. They each have forms that are
identical with the other. It is impossible to unambiguously say what play and plays mean if these
words are uttered outside of a sentence because the noun form singular and the verb form
present plural indicative are both play and the noun form plural and the verb form present
singular indicative are both plays. The similarities between the two words are strong and not
accidental.
Traditional grammar has a problem in explaining these similarities. If play the verb and
play the noun are different words with fixed, discrete parts of speech, why are they alike? The
theory of superposition solves this problem by eliminating the idea of a fixed part of speech.
Superposition means that, in a specific context, a single word may play more than one role, i.e.,
have more than one type, at the same time. In order to explain superposition further, we need
some additional terminology. The superponend is a term used to describe the word being
superposed. Superposer is the name of the operator that superposes a superponend. A
superponend, therefore, is the operand of the operator known as a superposer. Once operated
on by its superposer, a superponend has what is called a stratified Shaumyan (1987:202) or
bistratal Sypniewski (1996:1) function, which Shaumyan has also called a dual Shaumyan
(1996a:).
Superposition is an operation just like any other operation in AUG. It has an operator (the
superposer), an operand (the superponend) and a resultant (the bistratal function. Here are
Shaumyans formal definitions of superposition and superposer:
DEFINITION 6. Let E be an expression of type x, and let E take on type y on type x. Then E
shall be said to belong to a type z, such that z is stratified into y superposed onto x. Type z is
represented by the formula <y/x>, where the slash (/) indicates the stratification of type z into type
y superposed onto x, which are enclosed in angle brackets.
DEFINITION 7. An operator R of type Ox<y/x> shall be called a superposer.
Shaumyan (1987:202)
Superposition produces words that can take on multiple roles at the same time.
Shaumyan uses the following example to demonstrate superposition:
John proposed our immediately leaving the office
Example 3.12
The word leaving has been superposed. As the result of the superposition, leaving is modified like
a term by the term modifier (Ott) our. It also acts like a predicate, in this case, a 2-place predicate
(OtOts), in that it takes a secondary term the office as well as being modified by a predicate
modifier (OOtOtsOtOts) immediately. The word leaving performs these roles (or a mixed role, if
you will) simultaneously. The sentence in Example 3.12 is well-formed under any theory of
grammar or linguistics. The theory of superposition explains the behavior of leaving, i.e., how one
word can act, simultaneously, as though it were a member of two distinct functional types.
Superposition is more than just a theory that says that words may have more than one
type. Superposition is context sensitive. By this, I mean that the context in which a word finds
itself must be taken into account to determine whether superposition has occurred. Another way
of saying this is to say that superposition is not permanent. If we think of what part of speech
means, we will see that the notion of part of speech assumes that a word is always identifiable as
being a member of a particular type. It is as though a noun carries an identification card which,
when demanded, indicates that the word is a noun. Superposition makes no such claim. A word
may be superposed, i.e., have a bistratal function, in one context but not in another. The word
leaving is superposed in the sentence in Example 3.12 but not in the sentence in Example 3.13:
50
We are leaving the office.
Example 3.13
Superposition requires that there be an operator (a superposer) present in the context. If there is
no superposer, there is no superposition. It is important to realize that a superposer need not be
another word. A superposer can be anything that a sign can be, including the context itself. If we
believe that superposition has taken place, we must be able to identify the superposer. As has
already been mentioned, superposition is, at its foundation, simply another operation. All
components of the operator must be present for there to be an operation.
Context is not defined in Shaumyan 1987. Context is an abstraction that means, roughly,
the semiotic surroundings of a word. For example, in Example 3.12, leaving exists in a context in
which it is modified by a term modifier and a predicate modifier. Context is not simply the
sentence because not all parts of a sentence will directly or even indirectly effect every word in
the sentence. Context depends on the linguistic and semiotic level that we are examining. For
instance, the context of a phoneme may be different than that of a word. Some elements, e.g., a
suffix, may be a direct part of several contexts at the same time, e.g., the morpheme -ing can be
part of a phonetic context and the context of a word (because it could be a superposer) at the
same time.
Shaumyan lists six rules for superposition. I will list them and then comment on them.
Rule S1: Ox<y/x>A xB
<y/x>AB
Rule S2: OxyA <x/z>B
yAB
Rule S3: OxyA <z/x>B
<z/x>AB
Rule S4: <Oxy/z>A xB
yAB
Rule S5: <z/Oxy>A xB
<z/y>AB
Rule S6: <Oxy/z>A <x/u>B
yAB
Shaumyan (1987:202-203)
In Rules S1-S6, A is an operator and B is an operand. Expressions in angle brackets are, of
course, stratified types.
S1 is a version of Rule E. It simply says that if A is a superposer and it is applied to B, the
resultant is AB of stratified type <y/x>. Recall the membership notation, zX, meaning that X
belongs to type z. Rule S1s first line applies A which is a member of type Ox<y/x> to B which is a
member of type x. The notation may seem a little awkward at first but it isnt. The left-hand side of
the first line of S1 says that A is a superposer because it uses membership notation zX (X
belongs to type z). A becomes X and Ox<y/x> becomes z; A belongs to type Ox<y/x>. I will leave
it to the reader to see how Rule S1 is a variant of Rule E.
51
S2 says that if an operator applies to a superposed operand, the resultant is of the same
type as the resultant of the operation if it were applied to a operand with a non-stratified type.
Shaumyan gives an example from our Example 3.12.
Ott our <t/Ots> immediately leaving the office
t our immediately leaving the office
Example 3.14
Here the phrase immediately leaving the office has a stratified type of a term stratified onto a 1-
place predicate (<t/Ots>). A term modifier (Ott) our is applied to the phrase. The resultant (our
immediately leaving the office) is type t because that is what results from the application of Ott to
a term. The stratified type has a type t stratified onto a predicate. It is the term function that is
modified by Ott. You might ask: where did the predicate go? The answer is: nowhere, it is still
there. It is, if you wish to think of this way, inside the term. The predicate function is modified by a
predicate modifier (immediately) just as though it (the predicate modifier) modified a non-
stratified predicate, e.g., He immediately sent the letter. The resultant, our immediately leaving
the office, acts like a term with respect to the rest of the sentence while it has a predicate function
within it. Recall that a term is considered one unit. An expression may have internal structure.
That internal structure must be accurately described. Superposition is one means toward that
accurate description.
S3 is similar to S2. Shaumyan uses this illustration to explain Rule S3:
OOtOtsOtOts immediately <t/OtOts> leaving
<t/OtOts> immediately leaving
Example 3.15
Notice however that the resultant type is stratified, even though the operator was a simple
predicate modifier. The difference is apparent in the notation of the two rules:
Rule S2: OxyA <x/z>B
yAB
Rule S3: OxyA <z/x>B
<z/x>AB
In Rule S2, the superposer is applied to an operand with a stratified type in which type x is
stratified onto type z and vice versa for Rule S3. In the notation <y/z>, type y is stratified
(superposed) onto type z. Type y is the primary type of the resultant and type z is the secondary
type of the resultant. The difference between Rules S2 and S3 is the difference between
operating on the primary (S2) or secondary (S3) type of the operand. If the primary type of the
operand is being operated upon, the resultant takes this primary type; if the secondary type is
operated on, the resultant takes the stratified type.
Rules S4 and S5 also form a pair. In these rules, a superposer is applied to a simple
operand. Notice the difference in the superposers in Rules S4 and S5:
Rule S4: <Oxy/z>A xB
yAB
Rule S5: <z/Oxy>A xB
<z/y>AB
52
In Rule S4, the primary type of the superposer is an operator; in Rule S5, the secondary type of
the superposer is an operator. In Rule S4, the operator type of the superposer is applied to the
operand and the resultant has a type that the operator, which is part of the superposer, would
normally produce. In Rule S5, the operator is applied to the operand but, since the operator is
the secondary type of the superposer, the resultant of the operator becomes the secondary type
of a new, bistratal function which, with z as its primary type, becomes the type of the resultant.
Shaumyan uses a passive example to illustrate Rule S6. His example sentence is:
This hypothesis was advanced by Einstein
Example 3.16
Rule S6 is a rule which describes what occurs when a superposer who is, itself, a stratified type is
applied to an operand which is a stratified type. In the long passive construction , as in Example
3.16, the intransitive (1-place) predicate takes on the function of a transitive (2-place) predicate.
Therefore, the stratified type of was advanced is <OtOts/Ots> (or, if you will: <p
2
/p
1
>). The agent
term (in this case by Einstein) in a long passive also has a stratified type: term is stratified onto
predicate modifier (<t/OtOts>). Note that the stratified type for agent is term stratified onto the
primary function of the predicates stratified type. When the predicate in Example 3.16 is applied
to its agent term, the primary function of the predicate gets applied to the primary function of the
agent term. The entire sentence reduction appears below.
<OtOts/Ots> was advanced <t/OtOts> by Einstein
Ots was advanced by Einstein t this hypothesis
s was advanced by Einstein this hypothesis
Shaumyan (1987:204)
When the primary function of the predicate was applied to the agent term, the predicate became
Ots. Now the 1-place predicate is applied to the remaining term, thereby reducing the entire
expression to an s-type. The agent performs both as a term and as a predicate modifier.
A term are either a primary term (t
1
), a secondary term (t
2
), or a tertiary term (t
3
) in a
sentence. Every term can take on the function of another term Shaumyan (1987:204) by
superposition. Shaumyan gives a Russian example Otcu xolodno Shaumyan (1987:204) which
translates as Father is cold. In Russian, otcu is in the dative case, a tertiary term, but functions as
the primary term of the predicate xolodno. Otcus type is <t
1
/t
3
>.
With respect to terms, every predicate is either a regular predicate or a predicate-
superposer. If the predicate is a regular predicate, e.g., the English predicate is cold, the
predicate will not superpose the terms to which it is applied. Predicate-superposers, e.g., the
Russian xolodno, superpose terms. We can think of regular predicates as having these type
descriptions:
1-place regular predicate: Ot
1
s
2-place regular predicate: Ot
2
Ot
1
s
3-place regular predicate: Ot
3
Ot
2
Ot
1
s
Figure 3.2
These formulae remind us of the order of the application of the predicates. The chart in Figure
3.2 permits the calculation of all possible predicate-superposers once we take a constraint into
account. A predicate-superposer cannot operate on two terms of the same type. Figure 3.3
shows all possible predicate-superposers.
53
1-place predicates; Ot
2
s,
Ot
3
s
2-place predicates: Ot
3
Ot
2
s,
Ot
2
Ot
3
s
3-place predicates: Ot
1
Ot
2
Ot
3
s,
Ot
1
Ot
3
Ot
2
s,
Ot
2
Ot
3
Ot
1
s,
Ot
2
Ot
1
Ot
3
s,
Ot
3
Ot
1
Ot
2
s
Figure 3.3
Combinators
Since combinators will be new to most readers, I will briefly describe what
combinators, what they do, and why AUG uses them. The notion of the combinator used in AUG
comes from the combinatory logic of Curry and Feys {1958) and Curry, Hindley, and Seldin
(1972). Combinators are abstract logical devices that perform some constructive or
deconstructive function. This is, frankly, a poor but, for our purposes, useful definition of
combinator. As used in AUG, the combinators of combinatory logic are used to assemble,
rearrange, and disassemble linguistic expressions. In other words, AUG is an application of
combinatory logic. In his book, Shaumyan mentions the basic combinators which are used in
AUG Shaumyan, (1987:207). These combinators, with names such as C, I, and , are taken
directly from combinatory logic but are developed into linguistic operators as part of AUG.
If combinators are thought of as reduction rules, every combinator takes the form: X
reduces to Y. Reduction means that the application of a combinator to X results in Y, which may
or may not be some altered form of X
6
. The reduction relation is symbolized by a right-pointing
delta (>). To look at a simple example, take the combinator, C, known as a permutator
7
. When C
is applied to a 2-place operator, R, it produces a new relation between the operands of R, which
can be symbolized thusly:
CRXY > RYX
Shaumyan (1987:207)
Because of the result that C produces, C can be said to produce a converse relation or a
conversion. See also: Curry (1977:118); Hindley and Seldin (1990:191)
8
The reducibility relation
is reflexive, transitive, and symmetric. Shaumyan (1987:219)

6
The identity operator, I, does not result in a change. It
exists, primarily, for formal purposes. I is called an
identificator.
7
For a formal description of C, see Hindley and Seldin
(1990:24).
8
It should be noted that AUG also has expansion rules,
which perform the opposite function of reduction rules.
Expansion is symbolized by a leftwards-pointing delta (<).
Unless otherwise noted, expansion and reduction rules are
similar except for the direction of their operation.
54
How does this admittedly abstract notion help our understanding of language? Shaumyan
has noted that C can be used to help explain the passive construction. Here is his example
Shaumyan (1987:207):
(1) This book was bought by Peter yesterday.
and its converse:
(2) Peter bought this book yesterday.
How has C been applied? C has not been applied to the words of these sentences, but to their
genotype functions. C has been applied to the primary and secondary terms and has produced
the converse of the relation of the primary and secondary terms. In (1), this book is the primary
term and Peter is the secondary term; in (2), the converse is the case. C can be reversed; we can
go from (2) to (1) as easily as from (1) to (2). Notice that the words in (1) and (2) are not the
same. The actual construction of the sentences, with the phrase by Peter in (1) and Peter in (2),
as well as changes in the expression of the predicate (was bought in (1) and bought in (2)), are
consequences of the application of C. In other words, the actual words used are an expression of
the underlying functional restructuring of the sentences.
Understanding this point is crucial to the understanding of AUG. AUG claims that there
are two levels to language. Here they are. The combinators act on the genotype level; the
expression of their actions are the phenotype level of language, i.e., the level that we are directly
in contact with.
The Normal Form and Other Abstractions
There are several other abstract notions that are important to understanding AUG. One
such abstraction is the notion of a normal form. The content of a sentence, called a semantic
field, is different from the form of the expression of that content. Several different sentences can
express the same content. Shaumyan gives this example:
a. John sent a letter to Mary.
b. John sent Mary a letter.
c. A letter was sent to Mary by John.
d. John, he sent a letter to Mary.
e. John, he sent Mary a letter.
f. Mary was sent a letter by John.
g. Mary, she was sent a letter by John.
Shaumyan (1987:220)
The normal form of a sentence is the simplest expression of a semantic field. AUG claims that the
meanings of a-g are the same, i.e., they are invariant.
Normal form explains what happens in reduction and expansion rules. A reduction rule is
a rule that takes a sentence in non-normal form and maps it onto a sentence in normal form. An
expansion rule reverses this process. It takes a normal form sentence and maps it onto a non-
normal form sentence. Reduction rules are rules that would take sentences b-g above and map
them onto (convert them to) sentence a. Expansion rules are rules that would take sentence a
and map it onto any of sentences b-g.
The notion of normal form has an interesting implication that every sentence has two
structures: an R-structure and an S-structure. The S-structure is the structure of the sentence as
the sentence normally appears. The R-structure is the combination of all the reductions of a
sentence which are necessary to reduce the S-structure to its normal form.
55
The reader should consult Shaumyan (1987:220-228) for worked out examples of the
reduction of sentences to normal forms.
The Dummy Term
For purely formal reasons, AUG postulates the existence of a dummy term. A dummy
term has a null meaning and is used just in the way any other term is used. Dummy terms are
used to explain impersonal sentences, for example, where there is no overt subject (primary
term). Dummy terms serve to resolve the conflict between syntactic and semantic requirements:
since dummy terms have a null meaning, they do not change the semantic context of an
impersonal sentence, and at the same time allow satisfaction of the structural requirement that
every sentence must have a subject. Shaumyan (1987:228) It should be noted that in a
translation, a sentence in one language may need a dummy term, while the other sentence may
not.
I will not give any further explanation of combinatory logic or of Shaumyans description
of the combinators used by AUG. Anyone interested in the theoretical aspects of the use of
combinators in AUG should carefully read Shaumyan (1987:207-251). Shaumyan develops AUG
formalizations of reflexive constructions and sentence nests as examples of how AUG treats
complex linguistic problems. Shaumyan (1987:238 et seq.) For anyone wishing some additional
background on combinatory logic without surrendering to the tender mercies of Curry and Feys,
ibid., and Curry, Hindley, and Seldin, ibid., Curry (1977) and Hindley and Seldin (1990) may prove
helpful. Curry (1977) only treats combinatory logic in passing but it explains the motivation for
combinatory logic quite well. Hindley and Seldin (1990) is an introductory textbook on
combinatory logic and lambda calculus, neither of which are subjects one thinks of when the word
elementary is used.
Shaumyan ends his chapter on the formalism of AUG with a brief comparison of AUG
and other linguistic systems as well as commenting on the place of AUG among semiotic
systems. I will postpone a review of these comments until later in this paper.
Part 4 - Phenotype Grammar
AUG claims that all languages exist on two levels, the genotype and the phenotype level.
We have examined Shaumyans explanation of the genotype level and now turn our attention to
his description of the phenotype level of language. The genotype level, the universal semiotic
basis Shaumyan (1987:284) of all language, can be understood as the functional level of
language. The phenotype level of language is the syntagmatic level of language. The genotype
level is the same for all languages; the phenotype level differs from language to language. Being
a theory of both levels of language, AUG has both a genotype grammar and a phenotype
grammar. Part 4 of this monograph explores Shaumyans phenotype grammar.
The basic sign
i
on the phenotype level is the word. A combination of words constitutes a
word syntagm. The sentence is a functional unit of the phenotype level
ii
. Usually, a sentence is
composed of a word syntagm, i.e., more than one word, but, occasionally and in some languages
more frequently than others, consists of only one word, e.g., the Latin amo. Phenotype grammar
has two broad tasks. The first is to study the structure of the word as the basic sign on the
phenotype level. The second is to study the syntagmatic structure of sentences. Because
sentences are combinations of words, phenotype grammar pays particular attention to the study
of word groups.
The Word
Bloomfield described the word by using the concepts of free and bound forms. A bound
form is a linguistic entity, e.g., a suffix, which cannot appear on its own but only as part of a larger
56
form or collection of morphemes. A free form is one that can appear by itself. A free form that
cannot be subdivided into smaller free forms, i.e., a minimal free form, is a word; a free form that
can be so subdivided is a phrase. A bound form cannot be a word.
Shaumyan does not find this description to be satisfying. He cites three essential
problems:
The conflict between Bloomfields description of the word and word use
Bloomfield confuses the phonological representation of the word with grammatical notions of
the word
Bloomfields description of the word means that every variant of the same word (plurals, etc.)
is a separate word
Bloomfields notion of the minimal free form suggests that any word can stand on its own
but, in our every day speech, this is not the case. Possessive pronouns and articles, while
recognizable as words, cannot stand on their own. For example, the word the uttered by itself is
essentially meaningless; it needs to be associated with other words, i.e., it needs to be part of a
word syntagm, to be meaningful and to be used properly. Nonetheless, we consider possessive
pronouns and articles to be words.
iii
While many phonological and orthographic words represent unique grammatical words
but such is not always the case. Shaumyan cites the English word shut that can represent three
different tenses (present, past, and past participle).
iv
Bloomfields description leads to the conclusion that do and does are separate words, but
we do not treat them as such, but as variants of the same word. In Bloomfields description of the
word, the relation between word forms is not developed.
Shaumyan says that the definition of word in applicative grammar must be independent
of the notion of the morpheme. He proposed, instead, that the definition of word be based on the
notion of syntactic function. Shaumyan (1987:285) This proposal naturally flows from the claim
that language has two levels, a functional and a syntagmatic level. Genotype functions map onto
phenotype functions. The phenotype grammar of a language is the expression of the genotype
functions of all languages by a specific linguistic community. With this background, here is
Shaumyans definition of word:
A word is a minimal linguistic sign that is capable of having various syntactic and paradigmatic
functions either (1) by itself or (2) together with a word of type (1) meeting in the latter case the
condition of separability. Minimal means that a word contains no other word.
Shaumyan (1987:285-286)
A word of type (1) is called an autonomous word and a word of type (2) is called a non-
autonomous word.
In the following discussion, I will use the convention that any capitalized word refers to
the abstract word rather than to the non-capitalized word by itself. Thus, the word RUN refers to
the class of words (the abstract word) which are forms of the autonomous word RUN. So, run,
ran, runs, are different forms of RUN and are, collectively, referred to by RUN. RUN, the
autonomous word, is invariant in its symbolic function, meaning that it always represents the
collection of its forms.
Why are run, ran, runs, etc. forms of RUN rather than separate words? They are
separate forms because they are syntactic and paradigmatic variants of RUN. They are all related
to RUN rather than being separate, unrelated words. RUN represents a single paradigmatic
function to which all of its forms relate, although each represents a slightly different variation of its
autonomous word. Bloomfields definition of word implies that each word is independent of all
57
others. Shaumyans definition of word correctly points to the relation between forms of
autonomous words.
Autonomous word does not refer to a word that is orthographically separated from other
words. Words that are orthographically distinct from one another are said to be separable. A
suffix, for example, is said to be inseparable because it cannot function without being attached to
a word. A suffix, such as -ing, by itself, is meaningless. The key to the distinction between
autonomous and non-autonomous words is the capability of the word in question to perform
various syntactic and paradigmatic functions by itself. Shaumyan compares the Russian and
English words for book (kniga, book). BOOK has singular and plural forms but KNIGA has many
more forms due to Russians having a more highly developed case system than English. Many
Russian forms, e.g., the genitive singular knigi, are equivalent to an English prepositional phrase,
in this case, of the book. The Russian suffix -I is obviously non-autonomous and inseparable. Of,
on the other hand, is separable but this separability is not enough to make it autonomous. Of
effects the syntactic functions of other words rather than itself. Therefore, it is non-autonomous.
Autonomous words are words that have either a representative or vocative function:
nouns
adjectives
adverbs
verbs
pronouns
interjections
Non-autonomous words are:
prepositions
articles
conjunctions
Autonomous words can be further categorized. Autonomous words which have a representative
function (as opposed to a vocative or expressive function) can be either symbolic (verbs, nouns,
adjectives, and adverbs) or deictic (pronouns).
Shaumyan cites three advantages for his definition of word, some of which we have
already seen.
1. This definition of word groups words into abstract categories, such as RUN or WALK, which
have forms which express the syntactic and paradigmatic functions which the abstract word
performs, instead of seeing each variant of an abstract word as a word in its own right.
2. The correlations between autonomous words and syntaxemes makes the description of
primary syntactic functions and the description of the processes which transpose primary to
secondary syntactic functions easier.
3. The word replaces the morpheme as the central unit of morphology. Morphemes become
properties of the word instead of words being sequential strings of morphemes.
Shaumyan sees that post-Bloomfieldian linguistics has taken a wrong term in its view of
the word. Morphology was generally ignored and the word lost its theoretical status. He says:
If we accept the notion of syntactic functional units, we must accept the word as the basic
morphological unit that can serve as a syntactic functional unit. Morphemes in themselves cannot
58
serve as syntactic functional units. A morpheme can have syntactic functions solely as part of a
word. Only a word or combination of words can serve as a syntactic functional unit.
Shaumyan (1987:287-288)
In this quotation, we can see AUGs emphasis on function as the basis for determining the nature
of linguistic units. Morphemes cannot function independently; therefore, they cannot be basic
linguistic units. The word is the minimal linguistic unit because it is the smallest linguistic unit that
can take on a syntactic function. This last remark should not be taken to suggest that AUG see
the word as a monolithic entity. Words can be analyzed on the level of morphs. It is worth noting
some of the definitions and descriptions of different aspects of morphology that Shaumyan
mentions.
Shaumyan relates the notions of morph, morpheme, and allomorph as follows:
The morph is a minimal unit that comprises a minimal sign represented by a phoneme sequence
or one phoneme and its meaning. A class of morphs with the same meaning but different
phonemic is called a morpheme, and the morphs representing the same morpheme are called the
allomorphs, or variants, of that morpheme.
Shaumyan (1987:288)
It should be noted that the meaning referred to above is likely to be highly abstract. For example,
the word renewable has three morphs: re, new, and able. Both re and able are abstract and are
unable to stand on their own, i.e., they are not autonomous, and, therefore, are not words.
Meaning, by itself, is not enough to turn an otherwise non-autonomous morph into a word.
The variants of a morpheme can be conditioned phonetically or phonemically. A
phonetically conditioned variant of a morpheme can result from the replacement of a voiceless
consonant by a voiced one or vice versa. Shaumyan (1987:288)
Shaumyan distinguishes between lexical and grammatical morphemes:
A lexical morpheme indicates a fundamental concept, a concept of subject matter (lexical
meaning). A grammatical morpheme indicates a subsidiary, usually a more abstract concept
(grammatical meaning). Grammatical morphemes comprise closed-set items. The sets of items
are closed in the sense that, as a rule, no new items can be created to extend these sets. lexical
morphemes comprise open-set items. The sets of items are open in the sense that new items can
be created to extend these sets.
Shaumyan (1987:288)
This distinction is similar to the autonomous word - non-autonomous word distinction. There are
many linguists who would dispute that grammatical meanings are meanings at all. There are
several ways of addressing this objection, some of which are noted in Shaumyan and Sypniewski
1995. One way of addressing this objection is to examine the purpose of language. Language
exists for communication (and other purposes) but communication depends on the accurate
transmission of information. There are many ways of transmitting information, as the diversity of
the worlds languages shows. The objection that grammatical meanings are not meanings stems
from the tacit or overt belief that meaning (however defined) comes from the real world rather
than from language itself. Words are assumed to be entities with meanings of their own.
Grammatical meanings are abstract meanings which often color the meanings of some other
word or words rather than being able to stand as separate meanings by themselves. To suggest
otherwise is to suggest that there are words with no meaning at all. A simple test will show that
this is not so.
59
Examine the sentence
(4.1) John sent the letter to Mary.
The tertiary term is to Mary. Remove the words which have no meaning: the and to. We get:
(4.2) John sent letter Mary.
Sentence 4.2 is barely comprehensible and only so because we know that it is extremely unlikely
for John to mail Mary to a letter. The words the and to make the sentence easier to understand by
more clearly defining the roles (secondary term - the letter and tertiary term - to Mary) but they
also add something to the meaning of the words letter and Mary. The word the makes letter
specific; the word to makes Mary into the patient. The last point is interesting. Sentence 4.2 could
easily be a random collection of English words. Except for our expectations of word function
based on linear position, sentence 4.2 could just as easily said that Mary sent the letter to John
as vice versa.
Grammatical morphemes come in two types: functional and derivational morphemes.
Functional or inflectional morphemes indicate word functions, e.g., -am in portam indicates the
accusative singular. Derivational morphemes are used to derive new words, e.g., -er in worker.
Derivational morphemes modify the meaning of lexical morphemes. An unmodified lexical
morpheme is a root. A modified root is called a stem. Grammatical morphemes, also called
affixes, distinguish between word forms.
Shaumyan introduces the notion of the suprafix to help explain word formation. A
morphological formative is an elementary linguistic element that is not a root. Assume that we
have a root X and a meaning Y not expressed by a root, e.g., X is dog and Y is plural. There are
two separate methods. We can simply add X to another linguistic unit, Z, with the meaning Y to
get the new word. Z is an affix. The second method involves some alteration to X in order to
derive the new word. This alteration is referred to as a suprafix. The notion of the suprafix helps
to understand how we can combine X and Y into a new word W. There are five types of
suprafixes:
1. A suprafix whose meaning is correlated with a shift of a prosodic feature. For example, in
deriving a noun from a two-syllable verb, the stress often shifts to the first syllable.
2. A suprafix whose meaning is correlated with subtraction. Shaumyan gives the example of the
masculine adjective in French (e.g. blanc) being derived from the feminine (blanche) by
subtraction.
3. A suprafix whose meaning is correlated with an alternation of phonemes. Simple English
examples are take:took and man:men.
4. A suprafix whose meaning is correlated with reduplication. A Malaysian example is orang
(man) : orang-orang (men, people).
5. A suprafix whose meaning is correlated with transference (meaning converting change of the
context of the word). Play, the noun, is derived from play, the verb by being combined with
articles: a play, the play.
Shaumyan (1987:289-290)
Shaumyan notes that under his theory a word is a collection of a stem and formatives.
Derivational analysis cannot simply rest with the analysis of the immediate constituents of the
stem. AUG differs from GTG in this respect and solves a problem that GTG cannot. Affixes may
be linearly ordered but suprafixes are not.
60
There are two types of affixes: agglutination and fusion. A language can use both types.
Fusion involves a change to the root, e.g., the Russian nominative singular konec, which has an e
in its root, looses the e in the root for its other cases, such as the dative singular koncu.
Agglutination involves simple addition, e.g., the Turkish nominative singular son becomes sona in
the dative singular. Russian uses a combination of affixes and fusion while Turkish uses only
agglutination. In Russian, an affix can have more than one meaning, e.g., -am signifies both the
dative and the plural. In Turkish, affixes have only one meaning. Russian affixes are non-
standard in the sense that a given affix cannot be used in all situations. The dative singular is
expressed by -u but the dative plural is -am. In Turkish, the affixes are standard, e.g., a dative
suffix is the same in both singular and plural. A separate suffix -lar is used to make a noun plural
(sona - dative singular, sonlara - dative plural). The root that results from fusion (in Russian)
cannot be used as an autonomous word; an affix needs to be added to it in order to make it into
one: knoc- must become, e.g., koncu. In Turkish, affixes are added to a root that can be an
autonomous word. Son is the nominative singular. The predominance of the type of affixation
divides the worlds languages into agglutinating languages and fusional languages. Shaumyan
(1987:290-292)
Besides the morphological formatives we have just discussed, there is another type of
grammatical formative called a syntagmatic formative. The essential difference between
morphological formatives and syntagmatic formatives is where they express grammatical
meanings. Morphological formatives are internal and syntagmatic formatives are external in
regard to words. Shaumyan (1987:292) As a consequence, morphological formatives combine
grammatical and lexical meanings inside a word and syntagmatic formatives separate
grammatical from lexical meanings outside of a word. Grammatical meanings are expressed by
three types of formative:
prepositions and other non-autonomous words (segmental syntagmatic formatives)
word order (segmental syntagmatic formative)
sentence intonation (suprasegmental syntagmatic formative)
If a language chiefly employs non-autonomous words and word order (segmental syntagmatic
formatives) to express grammatical meanings, that language is said to be an analytic language. A
language which chiefly uses morphological formatives to express grammatical meanings is called
a synthetic language. In a synthetic language, if a word W is removed from a sentence, W retains
its grammatical characteristics. Latin is a synthetic language; any Latin noun and adjective, in
isolation, can be identified with respect to case, gender, and number. In synthetic languages,
morphological structure determines a words syntactic properties. As Shaumyan says, [t]he word
in synthetic languages is independent of environment Shaumyan (1987:292). On the other hand,
a word in an analytic language, in isolation, has only a lexical meaning. It derives its grammatical
characteristics from its context. Word order is not generally important for synthetic languages;
when word order is an important consideration, it is important for stylistic reasons rather than
grammatical ones.
Concord or agreement is a characteristic of synthetic languages; analytic languages have
considerably less concord than synthetic languages. Concord is the obligatory repetition of some
grammatical meaning, e.g., in the Latin in tribus partibus, both the grammatical meanings dative
and plural repeat in tribus and partibus. The adjective is said to take or agree with the case of
the noun that it modifies. Concord is less important for analytic languages because analytic
languages rely on proximity to show grammatical relations between words. In the English in three
parts, there is no repetition of case markings in three and parts.
Government is a relation between two words in different grammatical classes in which the
grammatical class of one word determines the grammatical class of the other word. Prepositions
in Latin, for example, govern the case of accompanying nouns: ad urbem, ab urbe.
61
Shaumyan takes issue with current linguistic literature in which concord and government
are treated as phenomena that correspond to syntactic dependencies between functional units.
Shaumyan (1987:295) He says that concord and government are morphologically conditioned
connections between words that do not necessarily reflect dependencies between syntactic
functional units. Shaumyan (1987:295) In fact, morphologically conditioned dependencies can
conflict with syntactic dependencies between different functional units. In Latin, the predicate is
morphologically marked to agree with the primary term: Caesar vincit. From a morphological point
of view, this suggests that the predicate depends on the primary term, but the reverse is the case
from a syntactic point of view. Let A be the head of a syntagm and B be its dependent. There are
four possibilities for morphological marking:
1. only A is marked;
2. only B is marked;
3. both A and B are marked;
4. neither A nor B are marked.
Shaumyan notes:
Syntactic connections between functional units, on the one hand, and concord and government,
on the other hand, are completely independent of each other.
Shaumyan (1987:296)
Neither can be explained in terms of the other. Shaumyan points to the problems which a failure
to note this last point can lead to.
The notion of linguistic category is discussed next. A linguistic category is a class of
expressions that share a common meaning or function denoted by common morphological or
syntagmatic formatives Shaumyan (1987:297). A word can belong to different categories.
Phonological units, e.g., vowels, also belong to linguistic categories. Note that the definition of
linguistic category says that a member of a category is an expression, not a word. Shaumyan
says:
In accordance with the distinction of derivational and inflectional morphemes, we must
distinguish derivational and inflectional grammatical categories. Since the range of occurrence of
inflectional categories is wider than that of derivational categories, the latter are subordinated to
the former.
The essential changes of the grammatical system of a language result from the shifts
between related inflectional and derivational categories. The change of an inflectional category
into a derivative category is called lexicalization. The opposite change of a derivational category
into an inflectional category is called grammaticalization.
Shaumyan (1987:298)
A linguistic category has two distinctive features: a common grammatical meaning (as opposed to
a common lexical meaning) and common morphological and syntagmatic formatives. The latter
feature is actually a relation. Books and feet are both members of the linguistic category plural
even though, as plurals, they are not formed in the same way. What is important is that they are
both members of singular:plural relations (books:books, foot:feet) and there is some
morphological change that indicates which member of the relation they are. Both words are
morphologically marked as plurals. The fact that they are both marked as plurals places them into
the linguistic category plural rather than the marking that is used. Of course, there may be a
category that admits only plural nouns ending in -s or some variant thereof, but such a category
may or may not be linguistically important.
62
Shaumyans discussion of phenotype grammar concludes with a brief discussion of the
category of case. Shaumyan understands case as:
...systematic differences in the forms of nouns, adjectives, and participles corresponding to the
differences between their roles as functional units in a sentence.
Shaumyan (1987:298)
There are three types of case systems:
active
ergative
accusative
In an active case system, the contrasting syntactic functions agent:nonagent are marked. The
agent is expressed in the active case; the nonagent is expressed in the inactive case. Verbs are
either active or stative rather than transitive or intransitive. In an ergative system, case depends
on whether the verb is transitive or intransitive. The primary term of a transitive or intransitive verb
is expressed in the accusative case; a secondary term of the transitive verb is expressed in the
ergative case. Like the ergative system, the choice of cases in an accusative system is also
dependent on the contrast between transitive and intransitive verbs. The primary term for either
type of verb is expressed in the nominative case and the secondary term of a transitive verb is
expressed in the accusative case. In terms of their case systems, the ergative and accusative
systems are mirror images. Shaumyan (1987:299)
In an accusative system, there are two classes of cases: grammatical and concrete
classes. The grammatical class consists of:
the nominative (primary term)
the accusative (secondary term)
the genitive (transposes a sentence into a noun group)
The concrete cases are:
the instrumental (movement across)
the ablative (movement from)
the allative (movement to)
the locative (lack of movement)
Shaumyan posits a hierarchy of positions, with grammatical cases being central. His comments
include the following diagram:
Verb > Grammatical Case > Concrete Case > Adverb
This hierarchy is purely relational and is independent of the linear order of words. Concrete cases
are intermediary categories between grammatical cases and adverbs.
Shaumyan (1987:300)
The secondary function of grammatical cases is to serve as concrete cases and vice versa,
suggesting that grammatical and concrete cases are subject to superposition.
63
Comments on Linguistic Methodology
Conclusions and Observations
Bibliography
1
de Saussure, Ferdinand. 1966 A Course in General Linguistics. NY: McGraw-Hill
Descles, Jean-Pierre, Guentcheva, Zlatka, and Shaumyan, Sebastian. 1985. Theoretical Aspects of
Passivization in the Framework of Applicative Grammar. Amsterdam/Philadelphia: John Benjamin Publishing
Co.
__________________________________. 1986. "Reflexive Constructions: Towards a Universal Definition in
the Framework of Applicative Grammar", Linguistical Investigations, 2
Shaumyan, Sebastian. 1968 Problems of Theoretical Phonology. The Hague: Mouton
__________________ 1987 A Semiotic Theory of Language. Bloomington and Indianapolis: Indiana
University Press
__________________ 1992 "The Stratification of Language". Proceedings of the Nineteenth Conference of
the Linguistic Association of Canada and the United States
Shaumyan, Sebastian and Segond, Frederique. 1994 "Long-Distance Dependencies and Applicative
Universal Grammar", unpublished manuscript
Shaumyan, Sebastian and Sypniewski, Bernard Paul. 1995 "Semiotic Analysis of Meaning", Proceedings of
the Twenty-Second Conference of the Linguistic Association of Canada and the United States

i
Even though we are discussing a different level of language, we are still discussing a semiotic system. The
phenotype level of language is part of a larger semiotic system (AUG) just like the genotype level. The sign is
fundamental to all semiotic systems.
ii
Of course, the sentence exists on the genotype level as well.
iii
Possessive pronouns and articles are not the only word types which pose problems for Bloomfields
description of the word. These two categories are merely illustrative. Some languages, like Chinese, have
words in Bloomfields sense of minimal free unit, which are treated as words by language in that they are
represented and pronounced like any minimal free form, which have what appear to be purely functional
meanings, e.g., ma which, when added to the end of a declarative sentence, converts the sentence into a
question.
iv
Another type of word, not mentioned by Shaumyan, which gives Bloomfields description of the word
difficulty is the homophone. Some words are homophones but not homographs. For example, the Far East
Chinese-English Dictionary (1993:1865), by my count, lists 119 entries for different inflectional variants of zhi.
While I have not worked out the implications of superposition and transposition with regard to this
phenomenon, it is quite apparent that something other than a simple one to one, phonological to orthographic
relation occurs here.

You might also like