Do individuals have an Occam’s razor for word meanings?

Question

Background and Example

On the German Stack Exchange, we had a lengthy discussion regarding the meaning of the word Gefäß. It was undisputed that a Gefäß is:

1. an item which can contain liquids
2. movable
3. not flexible

The argument was about whether

    4. artificial (in the sense of being modified by humans in any way)

should be added to the definition. My opponent would not call something Gefäß which complied to points 1 to 3 but not to point 4, e.g., a peculiarly shaped stone; I however would. The problem was there are very few objects for which point 4 makes a difference, as the vast majority of objects complying with points 1 to 3 also complies with point 4 anyway. Thus I considered the discussion rather moot, as few people have ever made up their mind as to whether point 4 is essential to the term Gefäß and it would hardly affect the actual usage of the term.

In particular, this made me think of whether people are actually likely to include aspects such as point 4 into their personal definitions of words, as it seems innefficient to me.

Question

Is there any study or general linguistic argument regarding whether we tend to include less relevant details in our personal definitions of words, i.e., details which do not strongly affect the range of meanings a word covers. Or do we have some sort of Occam’s razor instead, which keeps our personal definitions simple?

Be aware that I do not ask how to solve the dispute in the example or analyse the example issue. I am really interested in general tendencies towards or against detailed word definitions.

Answer 1

Semantics (and to a certain extent cognitive or even social psychology) studies the sort of thing you're after. But nobody (I know of) would formulate the problem in this way. An ancient school of Buddhist semantics (the name of which eludes me) postulated that the meaning of a word was everything that it wasn't.

However, the whole idea of meaning being the kind of definition that you would find in the dictionary has been completely rejected by many in semantics. It's still a useful shortcut to say things like 'a word means something' or 'a word has a meaning' but nobody would think of meaning as an unproblematic thing, any more.

Meaning is best described as a process which has a linguistic, cognitive (psychological) and social dimension. Words and phrases trigger certain ideas, images, frames, etc. which blend with others as well as the context. The way this happens depends on the language, the speakers knowledge and also the shared knowledge of a community.

So in effect, we don't have a personal definition of a word in our head. We have a complex network of uses, images, propositions, etc. associated with the word. A definition such as the one you listed is simply another way of using the word in context. It is a product of social interaction and an expression of our metacognitive ability (ability to reflect on our speech and cognition), not the actual meaning or a description of that meaning.

From a linguistic perspective, it is important to think of dictionaries as just another text (albeit one with metalinguistic aims) and not a template for some sort of a mental lexicon. That does not mean a dictionaries or word definitions are not useful (or even indispensable tools). They are. But they must not be confused for anything more fundamental than that.

Answer 2

Preface: As far as I know, this issue is usually not discussed in linguistics (however, see e.g. concept learning and conceptual inference). The usual view in semantics would be that words (lexical items) are not stored as 'definitions', i.e., sets of necessary and sufficient conditions. Instead, there is a mash-up of prototypicality, semantic similarity and frequency effects, as well as some properties on top. Whether and how these sources of information are combined to give you 'the' meaning of a word is not clear at all. A trend on the rise is to analyze a lot of these inferences, e.g. about the properties of entities denoted by a word, as non-monotonic or probabilistic inferences.

A succinct and general answer to your question: No (at least not that I know of).

A more useful answer: There are two ways to go about this. (1) Assume that speakers store world-knowledge and relate it to the words they use, thereby enriching their lexical content. If a Gefaess is usually an artifact, then speakers may make a (!defeasible!) inference that a Gefaess is an artifact. This inference lives as long as no evidence to the contrary is encountered. In this sense, artifact is stricly not a part of the lexical meaning of Gefaess but an enrichment. (2) Test whether Gefaess behaves like other artifacts, linguistically. If they do across probes, then being an artifact is a part of the lexical meaning of Gefaess.

Addendum, since Gricean reasoning was mentioned already: Pragmatics is usually taken to deal with language use and not the information stored/related to a word (which is what I take your question to be about). A Gricean theory could tell help you decide whether speakers infer that Gefaess is an artifact (see (1) above) but not whether your own definition of Gefaess includes information about it being an artifact.