Free download lexical approach

WordPress Shortcode. Share Email. Top clipped slide. Download Now Download Download to read offline. The lexical approach Nov. Technology Education.

JolexM Follow. The lexical approach. Lexical approach c-map. Lexical approach. Grammatical meaning. English vowel , monothongs, diphthong, triphthongs. Triphthongs phonetich practice. Competency based language teaching. Related Books Free with a 30 day trial from Scribd. Related Audiobooks Free with a 30 day trial from Scribd. The lexical approach 1. This information facilitates the acquisition process by grouping items of vocabulary in semantic fields.

Wray is in favour of the automatic repetition of these units for their acquisition and argues that the storage of these sequences is produced holistically, as if they were single units. This process facilitates acquisition because, as was detailed above, learners do not need to analyse morphologically or syntactically such units for comprehension.

We agree with Cortina on her recommendation of three graded steps in the process of acquiring formulaic sequences: automatisation, contextualisation and production. The first is related to guided activities, while the second and third are related to semi-guided and free production activities. This pedagogical approach allows learners to work independently during the first two steps, whilst the teacher can indulge the practice of oral skills.

Among consciousness raising activities, meaningful tasks, such as dictogloss, seem to be the best option for lexis acquisition. We recommend dictogloss of texts containing collocations, as it is the bridge to afford automatisation. These tasks encourage students to analyse and process language more deeply, which facilitates both the learning and retention of information in long-term memory.

We suggest that there is also an inherent creative and motivational element involved. The tasks should be selected to allow the input to become intake and thus, to produce automatic retrieval. Some recommendations to develop collocations in the classroom are given below: Handling with general and specific dictionaries and other reference tools.

Making vocabulary learning enjoyable and stimulating. Guessing the meaning of lexical items from context. Using real situations in simulations. Working in groups: helps learning independence and exchange knowledge. Practising rhetorical functions. Noticing collocations and language patterns. Recycling and repetition of tasks. Working with language corpuses. A variety of word association games and exercises using the diverse techniques of cloze procedure. Further hints through underlining.

We have found it appropriate to maintain their original headings: Matching and definition a. Now, try to define each of these terms in the same way: storage device flatbed plotter laser printer colour monitor impact printer daisywheel printer disc drive line printer voice recognition device b. Choose the correct definition for the multi-word noun compound: 1. Optical character recognition input a. Basic telecommunication access method a. Cathode ray tube display unit a. Building compounds Preparation: the teacher must collect enough pieces of text scientific papers, instruction manuals, specialised newspapers and magazines… to be able to give at least half of one to each student.

A page or two is not enough for this exercise. Find six such combinations and be prepared to teach their meanings to the rest of the class. This may well mean bringing in the context in which you found them.

Teach your collocations clearly and briefly. In the next class: 1. Then the group is asked to produce other nouns that combine well with the first noun. The result may be: Qualifying Head Head Head Head Head Head Head noun noun noun noun noun noun noun noun System chart communi- design analysis failure generation control cation language Address bus Format mapping calculation modifica- range control sorting tion register field 2. Decoding compounds: rewriting: The aim of this activity is to decompose noun compounds or to construct them from a decomposed text, understanding the relationships between their elements.

Conclusion This paper has offered a general insight into the methodological foundations of the lexical Approach and its implications in formal teaching situations. As recent l2 research indicates, learning collocations, apart from increasing the mental lexicon, leads to an increase in written and spoken fluency.

Thus, an understanding of collocation is essential for all learners, especially for those on upper-intermediate and advanced level courses for the active improvement of oral output within and outside classroom practice. For that purpose, we have made some recommendations for classroom practice, as well as illustrations of tasks oriented to one of the most important lexical problems in EST discourse: the so-called Noun Compounds.

References Aitchison, J. Words in the mind- an introduction to the mental lexicon. Oxford an New York: Blackswell Publishers ltd. Chambers, F. System, 25 4 , Cortina, B. Deignan, A. Centre for English Studies. Birmingham: Uni- versity of Birmingham.

Ellis, N. Vocabulary acquisition, word structure, collocation, word-class, and mean- ing. Cambridge: Cambridge University Press. Firth, J. Jakobson, R.

Lewis, M. The lexical approach: the state of ELT and the way forward. Hove, England: language Teaching Publicactions. Hove: language teaching publications. McIntosh, A. Patterns and ranges. Language Harper and Row, New York. Miller, J. Weinert Spontaneous spoken language: syntax and discourse. Oxford: Oxford Clarendom Press. It assumes that anybody who limps is likely to be hurt, which seems to be a reasonable assumption.

A hearer will always try to accommodate the information that will help admit an utterance into his context, as long as he believes that the speaker is knowledgable about the information. But this is the wrong reading for 47a. But if we ignore this small technicality, then this is so far so good. Here van der Sandt gets it right. Presumably, Mary will not see him unless he visits, and only if she sees him will she perceive his physical injuries and come to learn, as the speaker has, that John is hurt.

The visual contact is apparently not an issue in 47a. Here is an example where world knowledge plays a certain role: 48 a. While 48b should project global accommodation , 48a should not local satisfaction. In the story, the villain named Darth Vader revealed to the protagonist Luke Skywalker that he was his father. The other world knowledge the readers will need in order to understand 48a is that Emperor Palpatine is another villain in the story, notorious for his cunning and deception.

Perhaps this has something to do with him being misled by the grand trickster of the epic, Palpatine, who lied to Luke about his father being Vader for the purpose of making him feel bad and regret. Standing in stark contrast with 48a is the rather comical statement 48b. There is very little doubt in this case that the factive presupposition is projected to the global scope.

The stipulation we have been making for 47a and 48a , is a kind of bridging reminiscent of the ones below Beaver, : 49 a. Jane sat in the car.

She adjusted the rear-view mirror. Mirror linked to car b. If I go to a wedding then the rabbi will get drunk. Rabbi linked to wedding c. An old woman hit me. The brass knuckles cut deep. Brass knuckles linked to old woman d. The links are stipulated in order to make the most sense of the individual sentences. Each of the bridging links in 49 , as in 47a and 48a , are between partial matches.

The ways in which we bridge these partial matches are either through some form of entailment between the meaning of words like 47a , or via some elaborate set of reasoning and heuristics cars have rear mirrors, the weddings could be Jewish, Palpatine intended to make Skywalker sad because he is evil, etc. The information we gain through bridging can be accommodated, this accommodation need not always take place in the global scope, but it often does, and either way it will have far reaching consequences for the presupposition resolutions that follow, as we saw in 47a and 48a.

It is unclear how we derive bridging information, however, and exactly how or where we should accommodate it. I fully admit that these are very rudimentary speculations, but again nobody has ever looked into these types of sentence con- structions closely enough. The puzzles presented here certainly cannot be answered by linguistics alone. Why does the change of a few words makes a presupposition disappear or emerge?

I expect the answer to be ontological and epistemological in nature. Here is a well known example: 50 France has no King, and the King of France is not bald. Presupposition failure is one of those targets. And indeed, sentence combinations like 50 are hard to make sense of unless they are reactions to someone else who has made an utterance carrying the faulted presupposition. The Consistency Constraints will prevent global accommodation of the unique French King.

The next and only option is to locally accommodate in the scope where the presupposition is originally triggered. But there is nothing wrong with a sentence like 50 , so this should not happen.

Beaver sketches some proposals to ameliorate this situation: Since not all speech acts are informative, perhaps we should make informativity a constraint only on assertive sentences, but not on denials. Furthermore, we may claim that informativity is not an absolute constraint, but only a preference- when two readings are available, then either the more informative one or the one with the least redundant sub-DRSs is the most preferred.

These proposals have not been spelled out in any detail, so it remains to be seen whether they indeed solve our problem instead of getting us into new ones. Karttunen gives the correct readings for the following sentences. Van der Sandt, however, runs into some problems Beaver, : 51 a. Maybe John has a daughter. If she is over 18, she has probably left home.

Neither of these readings are sound. Again, the same solution could be applied to help us get around this. This is the same strategy that helped solve 28a in the last chapter: 28a Everyone who serves his king will be rewarded. Or if we move the trigger to the consequent position: 5 Beaver and Zeevat , p. I do not think this reading should be available, despite how reasonable it sounds.

Judging from these examples alone, intermediate accommodation may come across as a shot in the arm, but much to our dismay, this does not work for The same problem arises in conditional statements Beaver, : 53 a. If Mary buys a car, she will sell her Cadillac. If a woman buys a car, she will sell her Cadillac. But this is not what 53b means. The meaning of 53b has the odd implication that every woman has a Cadillac.

Intermediate accommodation of 53b turns out to be a rather contrived rationalization of 53b , and like 52 , it looks as if a presupposition trigger in the consequent of a conditional should not be allowed to accommodate in the antecedent. Beaver provides the following contrasts to illustrate this point: 54 a. Every German woman drives her car to work.

Between and yesterday, every German woman drove her car to work. But as soon as the genericity is removed, as in 54b , it appears that global accommodation is preferred - The meaning of 54b appears to have an implication that every German woman has a car, despite how counterintuitive that may sound. Each trigger only has one accommodation site. Both are partial models with discourse entities, properties and rela- tions.

They both are arranged hierarchically with subordinate relations between the internal structures6. Zeevat , p. Beaver illustrates this point with the following example: 55 a. Bill called Mary a Republican. The speaker of 55a intends to use the trigger too to imply that Mary was insulted by somebody other than John in the context of utterance. A hearer would be inclined to deduce that calling somebody a Republican would constitute an insult, therefore Bill insulted Mary- this information is globally accommodated.

This however, does not licence 55b even on the basis of 55a , because for all we know, Mary may not have said anything about Bill in her diary. These conditions. The presupposition of the lexical trigger therefore must hold in any context where the trigger appears if the concept is to be meaningful at all. Lexical triggers should not only be accommodated in the high- est suitable site, but also locally along its projection path. Single Accommodation 56 a. Both 56b and 56c follow from 56a.

A very delicate observation made by Zeevat and Beaver is that it is only when lexical presuppositions are embedded in intensional contexts that multiple accommodation is required: 57 a. Mary thinks it unlikely that it will stop raining.

Mary hopes that it will stop raining. Mary doubts that it will stop raining. Things are a bit more complicated than just straight out accommodation everywhere however- Neither 57c says Mary hopes it is raining nor does 57d say she doubts it is raining. Instead, 57c means the speaker and Mary believe that it is raining, but Mary hopes that it will stop, and 57d implies that the speaker and Mary believe that it is raining, and that Mary doubts it will stop.

As we can see, presupposition projection in attitudinal contexts can be very nuanced, and there is clearly a lot of work that needs to be done in this area.

Semanticists are not exempt from this. The ma- jority of presupposition studies tend to focus on the projection problem, and they do so by concentrating on single presuppositions, studying how they behave in various contexts.

In this pursuit, we have grown accustomed to setting up simple exam- ples which overlook one crucial aspect- presupposition interaction. To my knowledge, there has only been one attempt in the contemporary literature that turned to this issue Kamp, a , even though presupposition interaction is a very commonplace phenomenon.

These are two separate but interdependent issues. The former will be explained in the next sub-section. The latter is a topic that has been overlooked at least since Karttunen , and for all we can tell, was not discussed before that time either. In the literature, the context in which a presupposition is admitted is as close as we come to visualizing its semantic content.

But what precisely individual presup- positions consist of, what sort of referential functions they serve and what types of conditions are involved are questions that are completely brushed aside. If we want to know how presuppositions interact with one another, and about the outcome of such interactions, then it becomes imperative that we inquire into the meaning of individual triggers using the logical framework we choose to work with. To illustrate how presuppositions interact, consider these examples: 58 a.

John did not pass the exam. I know this because I know he did not show up at the exam. The narrow scope reading is not so complicated, Walter has a rabbit, and that rabbit is on the run now, as it was before7. A single rabbit owned by Walter was on the run and is now on the run again.

The rabbit on the run before belonged to Walter when it was on the run, but no longer does at the time when 58a is uttered. The standard view found in projection theories generally contends that if one presupposition fails, then the entire sentence should fail to obtain a truth value, or as Strawson puts it, the sentence is meaningless- end of the story.

I am not so sure if this is the end of the story though. Take 58b , clearly, even if 58b should have no truth value for lack of King of France, its other presuppositions, such as Walter, his rabbit, and the occurrences of such a rabbit being on the run- they should have their place in the context nevertheless.

The way I see it, this is not a problem about truth value interpretation, but rather, a problem about failure to bind and failure to accommodate- what should we do with the other presuppositions in the same sentence?

A theory of presupposition interaction would obviously have to account for occasions when certain presuppositions fail while others succeed, and what to do under each of the circumstances.

Another interaction can be found in 58c. There are many examples of presuppositions interaction in day to day life, and their varieties go far beyond those mentioned in If we are indeed serious about how presuppositions of complex sentences are projected from their parts, then we ought to start from the most fundamental parts of these sentences. Assuming a word is the smallest building block to compositionally makes up a sentence, a prudent investigation of the projection problem ought to begin with a study of the words- whether they are presupposition triggers or not.

In fact, there is a general lack of systematic proposals for a syntax-semantic interface that includes the presuppositional phenomena8. A syntax-semantic interface must accomplish two vital and interrelated objectives: On the one hand, we must have a systematic method to construct semantic representations in the form of a DRS from any given syntactic structure of any commonly accepted syntactic framework.

On the other hand, the semantics of the sentence as a whole should be based on the contributions of the words which appear in the sentence. This point is perhaps most potently demonstrated in examples 38 - 40 , which involve partial match. The relation between two partial match candidates only becomes available once we have the means to compute the meaning of the words in question.

Suitability conditions can only be articulated in a theory that takes into account the idiosyncratic meanings of individual words. World knowledge is indispensable as a consequence of the attention that is paid to individual word meanings; this is why it is not a coincidence that bridging almost always occurs in the form of partial match.

The contrast between single versus multiple accommodation 55 - 57 was shown to be closely connected to the class of triggers involved: anaphoric, and lexical. In response to this, I intend to adopt a reductionist9 , and perhaps even radical perspective that emphasizes the importance of individual word meanings and the role they play in constructing semantic representations. The most unconventional aspect of such a lexical perspective is perhaps how we incorporate those presupposition reso- lution strategies that are needed for the presuppositions generated by a word as part of the meaning of the word itself.

It is a philosophical position that complex systems are nothing but the sum of their parts, and that an account of a complex system can be reduced to accounts of its individual constituents. The lexical perspective to pre- supposition is reductionist in the sense that I begin with the assumption that there are no general, universal principles which govern the behavior of presuppositions.

Instead, by looking at the smallest components of an expression- either words or morphemes- the overall presupposition behavior can be calculated based on the interactions of these small denominators. The assumption draws a close connection between the meaning of a word and the way in which that word interacts with the context, as well as other words which carry their own presuppositions. Essentially, resolution becomes an intrinsic component of word meaning, instead of some external set of guidelines put into action by some hypothetical overseer.

These include: The information necessary to determine whether two individuals are identical for binding , or if they have bridging links; the constraints and preferences necessary to resolve the presuppositions that a word triggers. These aspects however, barely touch the core of our subject matter. What I think is the most important element of the meaning of a word, and what constitutes the semantic representation of a lexical entry, is a set of referential functions in the form of discourse referents, and a list of all relevant discourse conditions attached to those referents, such that they ensure that the context update of the utterance must be both consistent and informative with respect to the meaning of the word as we understand it.

Consistency and informativity seem to be a recurring theme here, but these are merely the minimum criterion of any context update, and the only attributes that are common to all words. The contribution of individual words goes far beyond consistency and informativity. This point is perhaps made most succinctly by raising three intimately related questions: 1.

What exactly is accommodated in the process of presupposition accommodation? And 3. How do we build up a context so that the correct inferences about the discourse may be drawn from it? To even begin answering these questions however, we are left with nothing but the lexical approach I have proposed. Consider the following Kamp, b : 59 I gave the workers a generous tip.

One thanked me. The other one left without saying a word. As the impersonal pronoun much like the German Man; 3.

There are a number of interactions in 59 , and it would take an entire chapter for me to do full justice to them. The only point I intend to make here with this example remains the same as before: The study of presuppositions must eventually be rooted in a study of the individual word triggers.

Until we become fully knowledgable of the conditions they impose on context and how they interact, we will never be able to fully predict their presuppositional behavior.

Only case studies can provide comprehensive coverage. As an ending disclaimer: I strongly caution that although one may be tempted by such a lexical perspective to assume that every presupposition triggering word will generate its own presuppositions, and that it does so in a way that is in principle inde- pendent from the ways in which other lexical presupposition triggers generate theirs- this is obviously false.

This will provide the ground for observing presupposition and meaning interactions, while opening up a greater potential for world knowledge to play a part in DRT. From the studies of various lexical entries we may then induce a generalized theory that expands the current framework- on the one hand, we will be able to distinguish those properties of presuppositions which are unique to individual words or expressions, but also, on the other hand, those properties that are shared amongst sets of triggers, including those which are common to all presuppositions.

It is those latter properties that can be seen as the essential part of a general theory of presuppositions. The semantic representations of both words and complex expressions are written with the notations outlined in a recent version of DRT that incorporates the handling of presuppositions Kamp et al. Originally proposed by van der Sandt , this version of DRT takes the view that presuppositions should generally be treated as anaphoric expressions.

I assume a similar syntax here, thus the trees I will be using should be thought of as generated by GPSG rules. The semantic representations are written in the notation of an extended version of DRT that incorporates presuppositions Kamp et al. This framework was successfully implemented by Blackburn and Bos in Verbmobil Bos, Given the following lexical entries: 1 Lexical entries operate under a syntax-semantics interface.

In this dissertation, we will not be focusing on any particular syntactic theory, but rather try to include some of the necessary elements that belong to any plausible syntax-semantic interface.

The syntactic analysis I assume will simply incorporate sets of syntactic properties of analyzed sentences which any viable theory of syntax should make available in some form. John climbed a mountain. The leaf nodes are lexical entries, each represented by a lambda DRS, and all nodes except the root are intermediate stages of the seman- tic construction.

The result of this construction is a Preliminary DRS. The reason for this is because there are no unresolved presuppositions in 60b , as there are no apparent presupposition triggers involved. Adherents of a descriptivist theory of meaning such as Russell hold the view that proper names are concealed descriptions- that the meaning of a proper name is essentially a set of properties that belong to the individual that bears the name.

This set of properties can be expressed as a complex description. It is obvious but important to note that in general, uniqueness must be understood as uniqueness within some contextually restricted and salient domain, this contex- tual domain is itself anaphoric to a set of entities which the interlocutors share in the common ground.

If necessary, this process is repeated all the way up to the global DRS. This process is called binding. This second process is called presupposition accommodation. Like binding, accommodation can, according to van der Sandt, take place at the global, the local, and some intermediate level.

Global accommodation is when the presuppositional DRS is added to the global DRS, and this is the most pre- ferred option. If global accommodation would produce an inconsistency, redundancy, or an unbound variable, then intermediate accommodation is resorted to, and when that also fails, local accommodation remains as the last option.

Local accommodation of these presuppositions amounts to adding to the consequent or the nuclear scope itself. What is needed instead, are algorithms which specify for each type of presupposition how presuppositions of that type can be resolved. The distinction between them manifests itself in that with a few exceptions I will ignore , the referents of singular descriptions are single individuals and those of plural descriptions sets of two or more individuals or, in logical terminology, the referents of singular descrip- tions are atomic and those of plural descriptions, non-atomic individuals.

The grammatical number of the description, i. In this notation, representations of presuppositions are always left-adjoined to the represen- tation of the assertive content of the sentence part in which those presuppositions are triggered. Moreover, presuppositions which are generated by other presuppositions are left-adjoined to the representations of those. This way, we keep track of the local origins of presuppositions, and so it is possible to recognize from the preliminary sen- tence DRS what is the local context of each represented presupposition and what, if any, are its intermediate contexts.

We start with cases where one or more antecedents are found, to which the descriptions might be bound. But in fact, in general that is not the right answer. The resulting set of possible readings is ranked by a preference order, which is determined by full versus partial matching, relative distance along its projection line, discourse principles, and non-linguistic knowledge.

A full discussion of the discourse factors that co-determine the choice of the preferred interpretation if the resolver leaves open a number of logical possibilities is beyond the scope of this paper van der Sandt, , p.

As one can see, van der Sandt stopped short of spelling out in terms of what makes one reading more preferred over the other when partial match is involved. To put this in more formal terms, I borrow the notion of Value Set from Krahmer and van Deemter , p. I will use lowercase letters when I use these terms more generally, or in the loose sense in which they are employed in much of the literature on the same topic.

Under the value-based perspective, there are exactly three possible types of Partial Match. Type 1 is when the Value Set of the potential antecedent is subsumed by the Value Set of the anaphor. In other words, the antecedent is entailed by the anaphor or that the antecedent is more informative e. Type 2 is exactly the other way around. The Value Set of the anaphor is subsumed by that of the antecedent.

In other words the anaphor is more informative, and in the event that it is bound to the antecedent, it actually adds information to it e. What this entails is that some form of accommodation has taken place implicitly, because the Value Set determined by the descriptive content of the antecedent is now restricted to a smaller Value Set than it was before resolution took place. We will see a full spectrum of examples in the next subsection.

Type 3 is a rather special kind of partial match in the sense that it does not occur very often in day to day usage, and when it does, it usually involves plurals.

Type 3 is when neither the anaphor nor the antecedent entails the other, but do not exclude each other either. Thus the highest degree of compatibility is equivalent to Full Match in our terminology. One possibility for this is perhaps explained by Grice , according to which one must be as informative as required, but not more informative than necessary. Speech production goes beyond the scope of this chapter and I will leave it at that.

The lowest degree of compatibility viz. Partial match falls anywhere within the area intermediate between those two extremes. This is very common: In particular, it is the standard situation for anaphoric pronouns, e. Before we move on to some examples of Partial Match, there is another component of our taxonomy that needs to be made explicit. So far we have only discussed the value-based approach.

This is the more prevalent perspective in the formal semantics tradition. But the other consideration, which I call the form-based approach, is important too. In this approach, matching of the anaphor with an an- tecedent is based on the form of their respective NPs. There are two 7 A thing to be said about intensionality here is that when the NPs of an anaphor and a potential antecedent are identical, they necessarily refer to the same set of entities in all worlds.

In this case, the comparison involves not only the head noun of N, but also its other parts. The strongest and most common kind of form-based match is a form-based Full Match in which the expression that are being compared are identical. When N is a simple noun, form-based matching may also take a weaker form than value-based Full Match i.

When N is a complex phrase, value-based matching procedures others than formal identity may occasionally play a role as well; but such procedure will not be considered here. A referential expression is lexically given if there is either an identical expression, a synonym, or a hyponym within the preceding discourse that acts as lexical head of the phrase that may serve as its antecedent.

A form-based Full Match is obtained when this lexical head is identical with the lexical head of the anaphoric description. What can the form-based approach contribute to antecedent selection and how can it interact with the value-based approach?

I assume that from a logical point of view the form-based approach plays a role that is secondary to the value-based method. It is only in cases where several competing antecedents are present in the context DRS, and every one of them has the same type of Value Set relation with the anaphor. We will see some more examples of this in the following sections.

There will also be a data related discussion about the roles of form-based and value-based selection towards the end of this chapter. The results of which suggest that the form-based approach plays a far bigger role in resolution than was previously believed, and that certain measures such as recency and segmentation could be built into a form-based approach to improve the accuracy of a resolution algorithm. The role we have determined for the form-based approach is a secondary one.

But is that really the case? While it remains necessary to look at more examples, a partial answer could be provided here towards that end: Unlike the value-based approach, a form-based analysis cannot be used to classify the full spectrum of possibilities.

They can only be derived when the form-based approach functions in conjunction with the value-based approach. The reason for this is that even if some of the descriptive contents of the anaphoric description overlap with those of a potential antecedent, it is still perfectly possible that they have No Match in the value-based analysis, and therefore constitute a contradiction. But in reality, these two are No Match according to the value-based analysis, and they should not be allowed to bind because their descriptive contents contradict one another.

For this reason, we will in discussing the examples to which much of the remainder of the chapter is devoted, continue to structure the discussions as we have already begun to do.

There are two notions of Full Match, the value-based, and the form-based Full Match. This, however, is not the case the other way round, a value-based Full Match does not necessarily entail a form-based Full Match think of synonyms. Yesterday, an uncle of mine bumped into a man.

The man fell to the ground. Yesterday, a man bumped into an uncle of mine. Yesterday, a guy bumped into a man. Yesterday, a man bumped into a guy. Yesterday, a man hit an uncle of mine on the back. Yesterday, John bumped into a guy. He was really hurt and the man took him to the hospital. Yesterday, Jean bumped into a guy. All of the examples in this chapter, unless otherwise indicated, should be read on a purely textual basis, that is, pragmatic factors are to be excluded.

This preference is so strong that neither of the sentences appears to be ambiguous. The penchant to bind to the antecedent with the same head noun is so strong that it even overwrites inferences-based preferences, as seen in 61e. Common-sensically, a man is far more likely to fall over as a result of being punched than fall over as a result of punching somebody. The consequence of this is that the hearer is stuck with the burden to rationalize why he fell in order to make full sense of this reading e.

From these examples we learn that the resolution algorithm should not only take the accessibility or the value of the relevant discourse referents into consideration, but binding with the antecedent that has a form-based Full Match should always be prioritized above all other alternatives. There is more to form-based matching than simply seeking an antecedent with the same head noun. To avoid digression, let us take an example where pronouns are already taken care of through means other than a lexical based resolution algorithm.

Consider 61f. Or is it introduced through inferences based on world knowledge, in order to facilitate the resolution of other anaphora during the semantic-pragmatic phase as in 61f and 61g?

Under the form-based approach, only those nominal conditions which are intro- duced simultaneously with a potential antecedent are considered. This conclusion entails that the interpreter must remember how a referent is introduced into the discourse universe. Several empirical studies reinforce this intuition. Jarvella and Herman have shown that listeners are good at recalling antecedents within the last sentence they heard, and less so as the antecedent becomes less recent.

In their studies, head nouns are matched before any further constraints are added. Take the following examples: 62 a. A jolly good fellow walked into a bar and sat down next to a grumpy old man. The man spoke. The man who was in a real good mood spoke. There are three conventional forms of compound nouns typically found in English dictionaries- closed e.

We will consider the one that is most prone to confusion. But this stops as soon as we look at 62c , head noun matching would yield the unintuitive reading which entails that the grumpy old man is in a real good mood.

The form-based approach does not, and therefore should be applied after the value-based analysis to deal with ambiguities where two or more potential antecedents are present with the same type of Partial Match as the anaphor and same distance on the accessibility path. In any case, when there is a Full Match by either standards, accommodation cannot occur.

But these are exceptions rather than the rule, and their occurrences vary from language to language. Examples from each type and an explanation of their properties are described in this and the following subsections. Value-Based Partial Match, Type 1 : 63 a. Krahmer and van Deemter set up their examples in the form of conditionals to highlight the dilemma between choosing binding and accommodation, as well as the choice of where accommodation should take place since binding always takes place in the scope of the antecedent.

We begin with a familiar example, 63a , which was used by van der Sandt Nevertheless, the partial match set up here is such that the conditions associated with the descriptive content of the antecedent entail that of the anaphor- if x is an oriental girlfriend then x is a girlfriend. There appears to be no particular correlation between the resolution strategy and the type of partial match we are dealing with. Through such relations, the noun e.

A cooperative hearer will be forced to to try to establish this link without guidance from the explicit context, by choosing what appears to her for whatever reason the most plausible candidate presumably John, in this case , and this is the reason why 63b has a preference for binding. It has nothing to do with partial match.

To avoid those problems caused by 63a and 63b , Krahmer and van Deemter use the sentence in 63c to illustrate the type of partial match where the antecedent is more informative than the anaphor. In spite of that, there are several issues involved in this example which I think should not be confused. First, there is the complication introduced by the plural. The second consideration is form-based- what kind of role does the head noun play in resolution of this particular type of Partial Match?

How does this type of Partial Match correlate with any of the resolution strategies? We will set plurals aside for the moment and consider 63d , a proximate clone of 63c , except that now we are dealing with singular NPs. This sentence exhibits a very strong preference for binding over global accommodation.

Thus, the fog of ambiguity Krahmer and van Deemter insisted on is suddenly dispelled, all while we have succeeded in preserving the kind of Partial Match that 63c was originally intended to illustrate9.

If we allow this in- ference, then binding seems to me very strongly preferred Take 63e for example, 9 The type of Partial Match we are dealing with and having had preserved in 63d can be spelled out. It is absolutely critical to set up the examples correctly if we want to discover anything meaningful about the relationship between resolution behavior and the way in which Partial Matches are categorized. In particular, we must guard against unwanted interference from inadver- tent bridging inferences.

Several other measures must be taken to produce the most revealing examples. The lack of competing antecedents in the discourse context makes that they can be used to illustrate only the alternative between binding and accommodation.

An isolated antecedent in context carries a strong preference for binding the only alter- native being accommodation , since pronominal satisfaction is the preferred strategy for a speaker in such situations.

Because of this complication, it is not easy to have from such examples what we want to know about binding. What we really need is a variety of Partial Matches, both in value and in form, in order to see which one of them the anaphor picks up as the most suitable binding candidate This is a very basic premise and there is nothing wrong with taking it for granted.

Yesterday, he saw a young man, a thief, a robber, a war criminal, and a drug dealer. The criminal is to go on trial at 5pm. After work, he1 had another brief exchange with the criminal. The2 dealer was complaining to his2 lawyer while the criminal paced up and down impatiently.

The same example gets complicated quickly as soon as we make a few changes: 65 John1 works at the courthouse as a clerk. Yesterday, he saw a thief, a robber, a war criminal, and a drug dealer. The criminal who is accused of selling drugs is to on trial. The lettered sentences are unrelated to one another. After work, he1 had another brief exchange with the man who is said to have violated the Geneva Conventions. The2 dealer was complaining to his2 lawyer.

The criminal paced up and down impatiently. What is interesting about this example is that we are seeing a competition between the two criteria we have been using to classify our partial matches. So the form-based approach takes the front seat. What is going on here? The answer lies in that 65a is not really an example of Type 1 Partial Match, according to the way we have been classifying them. A drug dealer one sees in a courthouse is essentially equivalent to the criminal who is accused of selling drugs, they denote the same set of individuals in every Model that is consistent with the world we know When such relation holds between the two, form-based considerations do not come into play.

A drug dealer one sees in a courthouse is probably not dealing drugs, but instead, standing trial. Since one is technically innocent until proven guilty, the alleged criminal can only be accused of selling drugs, unless we have evidence from context that he is indeed convicted and sentenced.

Some experiments have been done in this area, and what has been generally acknowledged is that interpreters tend to remember the descriptive content of a discourse referent better when its last occurrence falls within a 1 to 4 sentence window.

Interpreters also tend to prefer an antecedent when it is introduced by the same head noun as the anaphor, but the memory used to keep track of this deteriorates over time as the antecedent becomes less recent. For me, 65c is genuinely ambiguous between the two readings. This is by all means plausible, but salience is not a lexical property, so I will leave the matter at this.

Furthermore, if there is more than one Type 1 potential antecedent available in the discourse context, then we must use the form-based approach to look for the one with the closest matching NP, starting with the head noun. Falling short of that, a textually more recent Type 1 antecedent is preferred over one that is less recent.

If John owns a donkey, he will be worried about the purple farmer-eating donkey on the loose.