THE THEORETICAL IMPLICATIONS
OF
RELATIONAL MARKERS AND STRUCTURAL PREPOSITIONS
FOR
JUNCTION THEORY

 Brigham Young University
Translation Sciences Institute
Circa 1975

 Editor’s Note

This article is an amalgamation research notes and class handouts for Linguistics 501, a graduate course in advanced Junction Grammar taught by the author during the 1970’s. A non-trivial challenge presented by ongoing research was to develop a discovery procedure for modeling invisible mental processes. The content of the article is reflective of that objective. Principles (1) and (2) are not dated – they were important then and still are.

Overview 

Orientation

The phrase structure grammars utilized by TGG in its various forms use what amounts to a single joining operation, namely, concatenation. In the case of conjunction, however, transformationalists have found it necessary to introduce a conjunction marker between nodes to indicate the existence of a special relationship. Might such notation be, in effect, a first step towards the introduction of junction operations generally? Time will tell. The various relational markers for conjunction (and, or, but, etc.) are positive evidence that this is not a misguided move. But what additional evidence is there to justify the many junction operations used in Junction Grammar (JG)?

It is not the purpose of this brief exposition to restate what has previously been written about the semantic contrasts between adjunction, conjunction, and subjunction. Rather, we will direct our attention to specific phenomena which have more recently come to light that tend to support the use of an extended inventory of specialized junction operations, even more extensive than those formerly postulated.

Prosodic Relational Markers

First, positive evidence that the specializations of subjunction recently introduced into junction notation are in fact valid is available in the form of voice inflections which cue them to the listener. Specifically, the contrast between restriction and non-restriction (‘.*’ and ‘.*.’ respectively) is coded as the presence or the lack of a pause in the articulation stream when the modifier follows its antecedent.

1)  John, whom we know …(.*.)

2)  The John we know …(.*)

Further, the contrast between right and left subjunction is realized as emphasis on the more specific operand of the subjunction. Consider, for example, the inflection occurring on interrogative pronouns, or other forms joined to such modifiers asonly (outside the article), even,also, etc. From the perspective of junction theory, a system of representation that does not motivate prosodic relational markers is hurting in the same way that a grammar does which fails to provide for conjunctive markers in a principled way.

Structural Prepositions

It is obvious that junction relations in basic data are in some cases recoded as nodes during the process of extracting information from the information net and preparing it for use in conversation.  For example, consider the prepositions ‘of’ and ‘by’ in such expressions as ‘the teaching of English by the professor.’  The existence of such prepositions is positive evidence for postulating at least two kinds of adjunction, rather than a single kind. Moreover, the fact that relations may in fact be recoded as nodes suggests a technique for detecting junction contrasts not yet recognized in the notation of JG.

Specializing Adjunction

For example, consider the agentive and obgentive readings of the structural preposition of or the corresponding adjective suffix ’s. The ambiguity of these forms (the two distinct readings) clearly demonstrates that the relation between subject and predicate is not identical to that between verb and object:

1)      The reading of John was sub par. (‘John reads” - agentive)

2)      The reading of opposition literature was prohibited. (‘… read literature’ - obgentive) 

One is therefore justified in postulating distinct adjunction operations for the environments in question.  In this case, since the adjunctive generalization previously made for these relations seems basically sound (the J-rule schemata thereby obtained have been explanatorily powerful), one would introduce specialized adjunctions for the relations in question.  

Similarly,‘John seeks a house’ and ‘John builds a house’  correspond to ‘John’s seeking/search for (*of) a house’ and ‘John’s building of (for*) a house,’ where the node counterparts of the respective adjunctions again contrast. Notice also that ‘by’ does not fit with equal ease in both environments: ‘The building of a house by John’ versus ‘*the seeking for a house by John.’ While this may be simply a preference for ‘of’ in the presence of ‘by’ (notice ‘the seeking of John for a house’), intuition identifies a different role for John as an agent in each case (a creator for the predicate with ‘build’, but not for the predicate with ‘seek’).  We note also in this regard that the nominal form of ‘build’ (‘building’) can refer to the result of John’s act, while the corresponding form of ‘seek’ (‘seeking’) cannot. Thus, in accordance with our hypothesis, if adjunctive contrasts are presumed to exist for ‘build’ and ‘seek’, the lexicalizations of their respective nominals are governed in terms of that same contrast.

The verb ‘love’ affords yet another example of adjunctive contrast between verb and object: ‘John loves Mary/ John loves money’ versus ‘John’s love for Mary/ John’s love of money.’ A study of verbs in general in the context of compatible structural prepositions may prove useful in refining the inventory of adjunctive relations used by the languages of the world.

It has also been proposed that an adjunctive specialization is responsible for the ambiguity of sentences such as (3), where one reading is that what happened was accidental and the other is that what happened was done on purpose:

(3) He bumped an old lady.

 While this is a plausible proposal, difficulties arise for it in the case of conjoined PV’s. If the adjunctive operations were responsible for the ambiguity in question, one would expect one reading or the other to hold for an entire predicate. In the case of conjoined predicates, however, each conjunct is independent of the others in this respect. Thus, in a sentence such as (4), the first predicate is perceived to have occurred accidentally, but not the second:

           (4) Jim hit his thumb with the hammer and then proceeded to curse.

To treat the contrast in question in terms of an adjunctive specialization would require that a different adjunctive relation be given for each PV conjunct. This, of, course, would be a radical departure from the standard formalism of JG.

Others have proposed that the contrast is in the subject, i.e. that a feature specifies the subject noun as being either an agent or a non-agent with respect to the predicate. This solution, however, runs into the same difficulty – the subject will have to be specified differently for each conjunct of the predicate.

Another solution is suggested by the fact that the ambiguity of (3) can be removed by attaching a modifier to the predicate (He bumped an old lady on purpose/accidentally). Similarly, for (4) each predicate conjunct can be modified differently (Jim accidentally hit his thumb and proceeded to swear on purpose.) This suggests that the contrast in question resides in predicates themselves rather than in subjects or in the relation between subjects and predicates.

Case

Insofar as a comparison with other methods of linguistic analysis and description is concerned, we have here entered into the domain of what is generally referred to as case. We defer further comment, however, for another day. For the moment, let us concentrate on the theoretical implications of relational markers and structural prepositions.

Discovery Procedures

Basic versus Pragmatic Data

We return now to the observation that junction relations in basic data are in some cases recoded as nodes in pragmatic data during the process of extracting information from the information net and preparing it for use in conversation. This scenario gives rise to the following conditional that may be applied to the task of positioning components of a language model in relation to one another:

 1. If a medium of expression has the property of being symbolically mixed, i.e. some forms decode as operands but others decode as operations (relations if the data is viewed statically), then one may safely assume that the medium in question draws upon and recodes data from other sources.

When (1) is applied to JG Level II (Type II if you wish) data, i.e. to syntacto-semantic junction trees, our attention is again directed toward the Ad affixes (of, by, ‘s) discussed above that decode as (refer to) adjunctive relations holding between subjects and predicates, or between verbs and objects. To cite another example, given an expression such as Fred’s purchase of the books, ‘s refers to a subject-predicate relation and of refers to a verb-object relation. Now, inasmuch as some nodes in the corresponding J-tree have reference to relations, we know that language of this ilk (the vernacular) in its observable, overt form represents a mixed medium. Where did the mix come from? 

We have postulated in this regard that basic data, i.e. non-contextualized data, is imported from sensory sources and serves in its unadulterated form as the code of preference for constructing one’s information net. Pragmatic data, on the other hand, is presumed to reside on a ‘scratch pad’ (work area) which accommodates data being organized for output. The agent in charge of these mechanisms (that’s you) would accomplish the conversion of basic data to pragmatic data, recoding some operations as operands and inserting new operations and/or operands as required to meet the constraints for well-formed junction trees and the needs of discourse.

Analysis

Returning now to the matter of specialized adjunctions and discovery procedures for detecting junction contrasts not yet recognized in the notation of JG, let us formalize the principle of analysis that has guided the foregoing discussion :

2.  If semantic operations (junctors) A and B are, through a coding process, realized as operands (nodes) which contrast in their reference, then A and B are distinct operations.

Duality

There is a notable parallel between the operand~operator alternation observed when linguistic structures are ‘disturbed’ and wave~particle duality in physics. With that as a closing note, we defer discussion of its implications to another place and time.

[Home] [Origins] [Article Archive] [Foundations] [Formalizations] [Analyses] [Pedadogy] [Forum] [Guest Book] [BYU]

Copyright© 2004 Linguistic Technologies, Inc.