Contents:
As soon as the more complex word or word sequence has been assembled and entered in memory, it is available for access. Hay proposes that each instance of direct access of the complex unit strengthens that path of access and weakens the access through the component parts, at the same time weakening the relations with these parts and bringing on gradual loss of analysability. Of course, all this happens outside the controlled context of the experiment, where other factors become very important.
Analysability is maintained in the contexts in which the base word is also primed and it is lost in contexts where the base is not primed. The same would apply to 48 Chunking and degrees of autonomy compositionality. Semantic and pragmatic shifts that reduce compositionality are aided by frequency or repetition, but their source is in the contexts in which the complex unit is used. The loss of associations with component parts leads to increasing autonomy Bybee and Brewer , which is the topic of the next section. In the network model of Bybee as well as the approach of Hay , , these relations, which embody analysability and compositionality, can be of varying strengths and they may even be lost entirely.
The fact that we can document degrees of autonomy arising independently from each of these mechanisms as well as cases where two or three work together means that autonomy is an emergent property of linguistic units. The following examples are meant to distinguish among the three mechanisms to demonstrate their independence.
These contractions are all transparent semantically and some have clear enough internal structure, being equivalent in syntax and semantics to their uncontracted counterparts. It is thus very likely that that they are accessed directly rather than composed from two parts. Semantic change without phonetic reduction seems to be characteristic of grammaticalization in South-east Asian languages, for example Bisang In fact, semantic change with little or no phonetic change is quite common and occurs in the creation of new constructions, such as the WXDY?
Some analysability is retained in this expression, as we can perhaps recognize the two words that comprise the expression.
Even though the erstwhile prepositional phrase is now written as one word, very little, if any, phonetic reduction is apparent. Our question now is whether autonomy due to repeated direct access can occur without the other types of changes. As an expression reaches high frequency it becomes more difficult to find examples without phonetic reduction or semantic shift, but the examples used in an early proposal of the concept of autonomy in Bybee and Brewer appear to show the effect of direct access without other effects.
Thus 1st sg. There are quite a variety of changes that occur in the various documented dialects of Spain and Provence, but in none of them is the 3rd sg. As the 3rd sg. In support of the gradience of autonomy, it can be noted that the 1st sg. Thus relative autonomy due to repeated direct access and lexical strength is the factor behind the Conserving Effect of frequency.
This does not mean that frequency causes phonetic reduction, or meaning changes, only that repetition is an important factor in the implementation of these changes. For phonetic reduction, repetition of the bias towards reduction leads to changed exemplar clusters, as explained above; for semantic or pragmatic shifts, repetition within certain contexts leads to new associations of the expression with a meaning, as will be discussed more below.
Even autonomy in the sense of direct access is created by frequency of use only because the human brain adjusts to repeated access by creating shortcuts. Thus derived words such as disease, business and breakfast have become disassociated from their etymological bases through semantic and phonetic changes. In grammaticalization we also find cases of complete autonomy, as we will see in the next section. It should be noted, however, that a certain degree of analysability remains well into the process of grammaticalization.
The origins of the lexicon: how a word store evolved. Continuing with our sample data in Table 3 for the sake of consistency, we choose from a class of clustering methods called "hierarchical agglomerative clustering" which combines or "agglomerates" the most similar cases rows in Table 3 into groups and then those groups into larger groups to form a tree-like structure called a "dendogram". When langue and parole are defined in this way, there is a gap between both: what is the mediating factor that bridges the distance between the social and the psychological, between the community and the individual, between the system and the application of the system, between the code and the actual use of the code? They found that a higher conditional probability of the function word given the preceding word predicted vowel shortening in the function word, even in bigrams which are not usually thought of as any kind of structural units such as them and , sometime in , where the , and fine and. Perridon eds. Annual Review of Cognitive Linguistics 4: —
This sentence contrasts with the Present Perfect sentence He has written the letter. Thus we have he has just arrived in the Present Perfect Smith The meaning of the construction is not compositional in the sense that one cannot work strictly from possessive have and the participle meaning to arrive at the anterior sense of a past event with current relevance. See Chapter 6 for a fuller discussion of grammaticalization. As is well-known, this construction reduces to something spelled as gonna, in which the erstwhile allative or infinitive marker is fused with the preceding participle.
Another sort of evidence for the loss of compositionality and analysability in grammaticalization can be found in the sequences of auxiliaries would have, could have, should have and might have Boyland However, the current meaning of these sequences is counterfactual, a meaning that is not compositional. Even low degrees of grammaticalization can provide evidence for loss of compositionality and analysability. See examples 8 and 9. As for analysability, speakers surely recognize all the words in the expression, but there is evidence that the relationships among 52 Chunking and degrees of autonomy the words are not always grasped.
Note that in the nineteen examples of this construction that occur in the BNC, the object of from is always first person; there are sixteen examples of me and three of us. It is thus a discourse device the speaker uses to disclaim a certain stance. While for some, the relationship between far and from might be quite transparent, it is interesting that seven of the examples in the BNC used the preposition for instead of the historically correct from. Thus examples 10 and 11 : 8 Far be it from me to stand in the path of true love.
For me occurred six times and for us once.
Phonetic change further obscures the individual parts of the expression. Use in context can affect meanings and inferences, and meaning changes lead to a loss of compositionality. As we will see in the next section, frequently made inferences from the context can become part of the meaning of an expression or construction. This suggests no clear divide between aspects of the meaning that are derivable from context and those that are inherent to the lexical item or construction.
This is due in part to direct access, but also to the similarity matching Meaning and inference 53 that goes on in categorization. As mentioned, both are more likely to occur in high-frequency items. That drives me crazy is more likely to be used than that makes me insane for this reason. Finally, we should consider whether or not decreases in autonomy are also possible.
As Hay has pointed out a constraining factor in developing autonomy is the frequency of the base word, or in the case of phrases, the component words. There may be some cases where a lower or decreasing frequency of a complex form leads to its becoming more analysable and compositional. Rather it seems more realistic to assume that the meaning of forms and constructions is involved in an interesting interaction of the specific with the general.
For instance, in grammaticalization we see meanings generalize and become more abstract, while some specific meanings or functions are retained. Indeed, considering the way that categories evolve, it would not be unexpected to find that one category had split into two, given certain contexts of use.
Cambridge Core - Cognition - Language, Usage and Cognition - by Joan Bybee. 'It used to be a cliché that humans understand new utterances by constructing analogies with previous utterances. A fully-fledged articulation of this idea was.
Consider first the fact that many times the identification of an invariant meaning is not feasible. Some English examples, all of which can be replicated cross-linguistically, include the fact that the future marker will, which has a pure prediction meaning in many contexts 12 , also has intention uses, as in 13 , and willingness uses, as in Examples are from Coates Since an invariant meaning has to be the most abstract meaning available, the more specific meanings have to be derived from the more abstract, in a reversal of the usual diachronic relation.
As for the problems such a hypothesis poses, consider how one might try to derive the willingness meaning of 14 from the prediction meaning shown in What in the context could be relied upon to provide that meaning? The root possibility use 17 , which I hypothesize once united them, has mostly disappeared in spoken language Bybee b.
I will wander along to your loo, if I may. I tell you this so that you may make arrangements elsewhere if you are able. Consider also the use of Past Tense in hypothetical protases of conditional sentences, as in 18 and 19 examples from the BNC. There are various diachronic scenarios that lead to the disjoint meanings of the same morpheme and to the more specific or more abstract uses, but none of these developments could occur if speakers were limited to a single abstract meaning for each morpheme.
The diachronic development that leads to the polysemy in these morphemes is discussed in Chapters 6 and To the extent that invariant meanings are proposed, they still have to be modulated in context.
Most proposals for invariant meaning assume that such calculations occur online Michaelis Thus a rich representation of meaning that includes the inferences that have been made during use is necessary to explain the common meaning changes that occur in context. In Middle English will was still used with a sense of volition Bybee and Pagliuca , but it occurred often in contexts indicating intention, as it still does see example The intention examples are particularly common with a first person subject, but they also occur with third person, as in In such a case, however, the expression of an intention of a third person implies a prediction that the predicate will be carried out.
Through such implications, the will construction takes on prediction meaning. It then is extended to use in unambiguous predictions such as 21 and You better go otherwise she will be annoyed! The polysemy illustrated here for will and may has developed diachronically from a richer lexical meaning that has been modulated in context; the frequent inferences made in context have been registered in memory and have become conventionalized as part of the meaning of the auxiliary.
This process would not be possible if speakers and hearers always assigned one invariant meaning to a grammatical form. The meanings developed for use in language come about because meaning is always situated in context. It is important to realize that our experience of the physical world and our social relations is neither uniform nor flat; it is not just a one or two 56 Chunking and degrees of autonomy dimensional conceptual space.
It is vastly more varied in its topology. There are some situations that are more important and more frequently arising and referred to than others. Time is not just two dimensional in our experience either; nor is it experienced or talked about independently of modality, both epistemic and agent-oriented. In fact, I argue in Chapter 10 that only an analysis that takes into account the very concrete uses to which language is put will be able to explain both the similarities and differences among languages.
We have also mentioned the gradience in the sequential associations which depends upon how often a particular transition within the sequence occurs. Considerable evidence indicates that this process refers to specific sets of items that have been previously experienced and stored in memory. Given the specificity of constructions and the way they are built up through experience with language, the probability and acceptability of a novel item is gradient and based on the extent of similarity to prior uses of the construction.
On the phonological level we could cite an example such as the pronunciation nucular for nuclear, which is very likely based on the strength of the sub-word string —ular as in popular, regular, or even more specifically, binocular. While we are most interested here in analogy as a processing mechanism, we also gain important insights by exploring its role in language change and child language acquisition. Despite the apparent success of 1 in describing how a new form such as leaped could come into existence to replace earlier leapt , it is very doubtful that linguistic analogies are very often of this type.