Table of Contents
The Nature of Categorisation and its relevance to word meaning
The grammatical constraint, syntactic structure, and the compositionality of meaning representation
Lexicalisation, the TRH, and the question of a “universal” syntactic representation
List of works used:
Universität zu Köln
Prof. Dr. J. Erickson
Ray Jackendoff’s semantic theory presented in “Semantics and Cognition” proposes that semantics is more than a field of linguistic studies; it studies the very nature of thought and experience as well. In this essay, I will look at certain aspects of this theory and compare them to more traditional approaches to semantics. It will become apparent that Jackendoff’s theory, in the end of the day, is not only concerned with semantics but with the whole area of cognition as well
The Nature of Categorisation and its relevance to word meaning
In this part of the essay, I will first describe more traditional approaches of semantic theories to categorisation and try to illustrate their inadequacies. Then I will go on to explain a theory provided by Ray Jackendoff which ought to address those problems adequately. Categorisation, about which Jackendoff says that
“…an essential aspect of cognition is the ability to categorise; to judge that a particular thing is or is not an instance of a particular category”
(Jackendoff, 1983: 69),
has to be an integral part of every semantic theory, as categorisation not only comprises encoding / decoding of linguistic information, but ultimately the core of meaning analysis, the way we store and process information.
A very popular and often – used approach to categorisation was the theory of necessary / sufficient conditions. Jackendoff explains the theory as
“The meaning of a word can be exhaustively decomposed into a finite set of conditions that are collectively necessary and sufficient to determine the reference of the word.” (Jackendoff, 1983: 112)
Therefore, if the number of characteristics is too low, thus if necessary conditions are not met, or if the quality of the characteristics is not adequate to make it a member of one distinct category, thus sufficient conditions are not met, the #entity# will not be identifiable. I will try to augment the explanation using examples:
Assuming an #entity# has the following characteristics: 4 legs, alive. These characteristics are necessary and sufficient conditions to categorise this #entity# as ANIMAL. If I then identify barking to be another characteristic, my #entity# could also belong to the categories DOG, WOLF etc. In this case, sufficient conditions are not met to categorise the #entity# as any special animal. I will need to identify domesticated, for example, to meet sufficient conditions to identify it as DOG. If I identified white with black dots instead of domesticated, I could even identify it as ANIMAL, DOG, DALMATION. Thus the higher the number of characteristics, i.e. the more conditions I can obtain about an #entity#, the better I can categorise this #entity# as I have more necessary and sufficient conditions for more categories.
At this point, I would like to introduce two more notions about information, dictionary and encyclopaedic information. Dictionary information will be defined as information about, or characteristics of, an #entity# supplying necessary / sufficient information. Encyclopaedic information I will call additional information about that #entity#. Using the above example, my dictionary information about the DALMATION is alive, 4 legs, barking, white with black dots. But there is a lot more I can say about that #entity#. I also know that it will be carnivorous, a mammal, domesticated etc. This information, which is not needed for categorisation, is encyclopaedic.
There are, however, serious problems with the theory of necessary / sufficient conditions. If categorisation were as described above, word meaning would be definite and absolute, and there could be no ambiguous categorisation. Alas, that is not the case. I will use an example presented by Jackendoff to make this fact clear, the vase / cup / bowl paradigm. If people were presented with objects as shown in (1.1), most would call (1.1.a) a vase, (1.1.c) a cup and (1.1.e) a bowl. But judgement about (1.1.b) and (1.1.d) would be unclear.
illustration not visible in this excerpt
(1.1.b) could not quite be called a vase, but neither be judged to be a cup either. The same problem would apply to (1.1.d). It can certainly be agreed upon that everybody knows what a vase, a cup and a bowl are, but still judgements about (1.1.b) and (1.1.d) would not be unanimous. Thus word meaning is not definite and absolute.
A theory attempting to solve this problem is the fuzzy set theory. This theory proposes that categorisation does not have to be absolute, but rather “relatively true”. Thus (1.1.b) can be a member of the category CUP by 60%, or a penguin could be categorised as a BIRD up to 71%, for example. But if a penguin would be a bird by 71%, what would the other 29% be? Although this theory addresses the problem of absolute vs. non – absolute word meaning, it can hardly provide with a solution.
Jackendoff proposes a different approach to categorisation. While he rejects the standard decompositional theories, formulated in terms of necessary and sufficient conditions, he rather redefines a
“…nonstandard notion of decomposition that meets the usual objections to necessary and sufficient conditions and that squares with the character of other perceptual and cognitive phenomena” (Jackendoff, 1983: 109)
than supporting other theories that are formulated in terms of prototypes, meaning postulates or associative networks. His theory, preference rule system, takes the approach that categorisation is not done by YES / NO, but by preferential judgement. This leads to categorisation not of YES / NO, as both is necessary / sufficient conditions and in fuzzy set theory, but in graded categorisation YES / NO / NOT SURE.
The first difference is Jackendoff’s distinction not by #entity# / category, but [TOKEN] [TYPE] distinction. A [TOKEN] can be defined as the thing to be categorised, and the [TYPE] as the category. Thus #the pen in my hand# would be a [TOKEN] and a dog would be a [TYPE]. According to Jackendoff, [TYPES] are creative and infinite. A [TYPE] can be created for every [TOKEN], thus for everything one can ever experience or imagine. Moreover, all these things can be grouped, leading to the creation of new [TYPES]. For example, I can create the [TYPE] [THINGS LIKE THE CAR THAT JUST DROVE OUTSIDE MY FLAT], and the group [TYPE] [CARS]. Furthermore, both [TOKEN] and [TYPE] are conceptual structures. According to Jackendoff, the level of conceptual structures is the interface level at which information gathered via all senses (auditory, visual, tactile etc.) can be processed and interchanged. This is important in two aspects. Firstly, it means that categorisation is not limited to linguistic information, but central to all cognitive thinking and learning. Secondly the process of categorisation can be defined as follows:
“In short, a categorization judgement is the outcome of the juxtaposition of two conceptual structures” (Jackendoff, 1983: 78)
Jackendoff identifies five symptoms about the preference rule system (Jackendoff, 1983: 152)
1. Judgements of graded acceptability and of family resemblance:
Judgement of graded acceptability proposes a process similar to fuzzy set theory, implying that a [TOKEN] does not have to have all the characteristics a [TYPE] includes. This way a penguin, although unable to fly, can be acceptable as [TYPE] bird. Even though judged to be a bird, however, there can still be some doubt about the judgement, hence graded acceptability. The term family resemblance explains this a bit further. Family resemblance suggests that all [TOKENS] connected with one [TYPE] do not all have to share all characteristics, but that all [TOKENS] must share some of them. The [TYPE] car for example implies 4 WHEELS, although few would not call the BMW Isetta a car. Moreover, some cars do not use petrol as fuel, cabriolets do not have a roof and formula one cars do have neither a windshield nor a license plate. Still, they are all categorised as [TYPE] car.
2. Two or more rules, neither of which is necessary, but each under certain conditions can be sufficient for judgement:
This symptom is closely connected to the first, postulating that a [TOKEN] can be judged to be a certain [TYPE] even if some requirements are not met. Nevertheless, there must be more than one rule to make a judgement. Using the rule 4 WHEELS alone the Isetta could not be judged to be [TYPE] car, but the rules MOTORISED, CABIN, 3 WHEELS are sufficient to make it a car, not a lorry or a trike.
3. Balancing effects among the rules apply in conflict:
This symptom points to the relation among the rules themselves. Even though 4 WHEELS certainly is an important rule for [TYPE] car, it can be balanced by CABIN, MOTORIZED, 3 WHEELS. Especially 3 WHEELS appears to be contradictory, it can still balance out a very crucial rule. There are also rules that do not appear to be very important, like WINDSHIELD, LICENSE PLATE, LIGHTS, 2 SEATS OR MORE, etc, which all do not apply for formula one cars, but all of these rules can be balanced by 4 WHEELS, MOTORIZED, for example.
4. A measure of stability based on rule application:
Jackendoff defines stability as the certainty with which one connects [TOKEN] with [TYPE]. To use the vase / cup / bowl paradigm (1.1) again, the examples (a), (c) and (e) will be judged to be a vase, a cup and a bowl respectively with more certainty that (b) or (d). The reason for that lies with the graded acceptability. These two [TOKENS] do not fulfil the characteristics of their [TYPE] as well as (a), (c) or (e). Thus the judgement of these three would be more stable than the judgement of (b) and (d). Still, as people judge them to be either [TYPE] vase, cup or bowl, there is a minimal amount of stability in their judgement. I chose the word minimal, because if you told them “No, (b) is certainly not a vase, but a cup”, they might revert their judgement and agree.
- Quote paper
- Stefan Prahl (Author), 2000, Semantics. Structure and function, Munich, GRIN Verlag, https://www.grin.com/document/6813