The first part of the disambiguation process takes place in the morphology. When ROBIE identifies a word that has a morphological ending, the morphology must adjust the features of the word. For example, when blocked is identified, the feature "ed" must be added to the list of features for block. At the same time, a portion of the disambiguation takes place. If block is defined as both a noun and a verb, then blocked is not a noun. The morphology causes some features to be added, such as "ed, past" and some features to be removed such as "tenseless." As features that are no longer applicable are removed, so also are parts of speech and their associated features that are no longer applicable. For blocked, the features "noun, ns, n3p" will be removed and the features "adjective, ed, past" will be added.

The morphology will identify words such as adverbs, adjectives, and verbs in a similar way. The morphology used is very similar to that of Winograd (1972) and of Dewar, Bratley, and Thorne (1969); the part-of-speech additions and deletions are taken from Marcus (1980). Although this technique may seem obvious, it is included to point out that a majority of the occurrences of part-of-speech ambiguity can be resolved or reduced on the basis of the morphology alone.

Now that we have allowed words to have multiple parts of speech and the morphology can be used to trim some of the ambiguity, we need a simple technique for disam­biguating words to a single part of speech. Again, refer­ring to Occam's Razor, what is preferable is a simple and general technique for all types of disambiguation.

Given the above mechanisms - multiple definition and disambiguation by the pattern matching, let us see how a few simple examples are handled. Consider:

(1) The falling block needs painting.

We will look only at the words falling and block in this example. The word falling is defined as a verb and an adjective in the dictionary and block is defined as a noun and a verb.

While parsing this example, after the word the has initiated an NP and been attached to it as a determiner, the rules to parse adjectives are activated. The rule ADJECTIVE has the pattern [adj], and matches the word falling. Falling is then attached and disambiguated as an adjective. Recognition of falling as a verb does not occur. As there are no more adjectives, ROBIE will acti­vate the rules to parse the headnoun. (ROBIE's grammar assumes that all words between the first noun and the head noun of an NP are nouns; see section 2.6.) The rule NOUN with the pattern [noun] will match on the word block, and it will be attached as a noun. Hence block will also be disambiguated without the verb use being consid­ered by ROBIE.

Other ambiguities inside the noun phrase will be handled in a similar way. This approach will usually cover the situation of singular head nouns, verb/adjective ambiguity and many other pre-nominal ambiguities. This works because the noun phrase has a very strict word order. When an ambiguous word is found, only one of its meanings will be appropriate to the word order of the noun phrase at that point. This approach can be thought of as an extension of the basic approach of the Harvard Predictive Analyzer (Kuno 1965).

This strategy will also often disambiguate main verbs. For example, consider the following sentences:

(2) Tom hit Mary.

(3) Tom will hit Mary.

(4) The will gave the money to Mary.

In (2), hit is the main verb. In the dictionary, hit is also defined as a noun, (as in card playing). The parser will attach Tom as the subject of the sentence and then activate the rules for the main verb. Since hit has the feature "verb", it will match that rule and be attached and disambiguated as a verb. Again, other possible parts of speech are not considered.

The word will could be a noun or a modal as sentences (3) and (4) demonstrate. In (3), will cannot be part of the headnoun with Tom, so the NP will be finished as above. The rules for the auxiliary will then be activated and the word will then matches the pattern [modal] and is attached to the AUX.

In (4), the word will is used as a noun. Since it follows the determiner, the rules for nouns will be acti­vated. The word will then matches the pattern [noun] and attaches to the NP as a noun.

The same approach will also disambiguate stop and run in the following sentence. Since stop is sentence initial and can be a tenseless verb, the rule IMPERATIVE will match, and it will be disambiguated as a verb. The word run, which can be a noun or a verb, will be handled as will in (4).

(5) Stop the run.

Now let us consider a more difficult example, the word to. To is defined as an auxiliary verb and a preposition in ROBIE, as illustrated by these sentences:

(6) I want to kiss you.

(7) I will go to the show with you.

In (6), to is the infinitive auxiliary, while in (7) to is a preposition. This analysis is based on that of Marcus (1980:118). Our two buffer cell lookahead is sufficient to disambiguate these examples.

The buffer patterns for the above sentences are:

By looking at the following word, to can be disambigu­ated. In (7), the word the cannot be a tenseless verb, so the first pattern does not match. In (6), the second buffer does not have the feature "ngstart", so the rule doesn't match.

However, the above patterns will accept ungrammat­ical sentences. To reject ungrammatical sentences, we can use verb subcategorisation as a supplement to the above rules. One cannot say:

(8) *I want to the school with you.

(9) *I will hit to wash you.

In English, only certain verbs can take infinitive complements. To can only be used as an auxiliary verb starting a VP when the verb can take an infinitive complement. Hence, by activating the rules to handle the VP usage only when the infinitive is allowed, the problem is partly reduced. Also by classifying the verb for PPs with the preposition to, the problem is simplified. This is merely taking advantage of subcategorisation in verb phrases. Taking advantage of this subcategorisation greatly reduces, but does not eliminate, the possible conflict.

We have seen what to do if the verb will only accept a toPP or a VP. The final difficult situation arises whenev­er the following three conditions are true:

• the verb will accept a toPP and a to VP,

• the item in the second buffer has the features "tenseless" and "ngstart" and,

• the toPP is a required modifier of the verb. Although this situation rarely arises, the above rule will make the wrong decision if the ambiguous word is being used as a noun. In this situation, ROBIE will make the wrong decision, and has no capability to better decide. By default, the principles of Right Association and Mini­mal Attachment apply as discussed in Frazier and Fodor (1978).

A free text analysis done on a cover story in TIME magazine (1978) resulted in 55 occurrences of the word to. The two rules mentioned above in conjunction with verb subcategorisation gave the correct interpretation of all of these. These rules were also checked on the MECHO corpus (Milne 1983) and the ASHOK corpus (Martin, Church, and Patil 1981). There were no violations of these rules in either of these.

Adjective/noun ambiguity is beyond the present scope of this research and is handled in a simple-minded way. If the word following the ambiguous adjective/noun word can be a noun, then the ambiguous word is used as an adjective. In other words, all conflicts are resolved in favour of the adjective usage. This problem arises in these examples:

(10) The plane is inclined at an angle of 30 degrees above the horizontal.

(11) A block rests on a smooth horizontal table.

In (10), horizontal is a noun, while in (11), it is an adjective. The above algorithm handles these cases.

This approach takes advantage of the lookahead of the deterministic parser. A word should be used as an adjec­tive if the following word can be an adjective or a noun. However, this approach would fail on examples such as:

(12) The old can get in for half price.

(13) The large student residence blocks my view.

In this section we have seen many examples of the reso­lution of ambiguity. To handle these examples, we mere­ly constructed a compound lexical entry for each word, composed of the features of each part of speech the word could be and allowed the pattern matching to perform the disambiguation. This technique has been used by Winograd (1972). Why does this work so well?

English has a fairly strict structural order for all the examples presented here. Because of this, in each exam­ple we have seen, the use of the word as a different part of speech would be ungrammatical. Although these tech­niques have been used for non-deterministic parsers, their effectiveness has not been investigated for a deter­ministic parser.

Most ambiguities are not recognised by people because only one of the alternatives is grammatical. In many situations, when fixed constituent structure is taken into account, other uses of an ambiguous word are not possible and probably not even recognised. Since fixed constituent structure rules out most alternatives, we have been able to handle the examples in this paper without any special mechanisms. In the introduction to this paper, it was stated that a clean and simple method of handling ambiguity was desired. I feel that this goal has been met for these examples.

Before we proceed, let us look at an assumption Marcus made in his parser, that it would be given only grammat­ical sentences. This assumption makes life easy for someone writing a grammar, since there is no need to worry about grammatical checking. Hence no provision was made for ungrammatical sentences and the original parser accepted such examples as:

(20) *A blocks are red.

(21) *The boy hit the girl the boy the girl.

This simplification causes no problems in most sentences, but can lead to trouble in more difficult exam­ples. If the parser's grammar is loosely formulated because it assumes it will be given grammatical examples only, then ungrammatical sentences may be accepted. If the syntactic analysis accepts ungrammatical sentences as grammatical, then it is making an error. Using grammat­ical constraints actually helps parsing efficiency and disambiguation. In the next sections we look at the consequences of this assumption as well as those of rejecting ungrammatical sentences.

We know that the verb group has a complicated but rela­tively fixed constituent structure. Although verbals have many forms, they must be mixed in a certain rigid order. We also know that the first finite verbal element must agree with the subject in person and number. That is, one cannot say:

(23) The boy are run.

(24) The boy will had been run.

(25) The boys had are red. etc.

While Marcus's parser enforced these observations to some extent, he did not follow them throughout his parser. We want to enforce this agreement throughout ROBIE. Checking the finite or main verb, to be sure that it agrees in number with the subject, will lead to the rejection of the above examples. This was done by adding the agreement requirement into the pattern for each relevant rule as will be explained later.

Buffers 1 and 2 must agree before a rule relating the subject and verb can match. This check looks at the number code of the NP and the person/number code of the verb and checks whether they agree. The routine for subject/verb agreement is very general and is used by all the subject/verb rules. The routine can only check the grammatical features of the buffers.

Marcus (1980) did handle some part-of-speech ambigui­ties. The words to, what, which, that, and have could all be used as several parts of speech. For each of these words he also used a Diagnostic rule. These Diagnostic rules matched when the word they were to diagnose arrived in the first buffer position and the appropriate packets were active. Each diagnostic would examine the features of the three buffers cells and the contents of the Active Node Stack. Once the diagnostic decided which part of speech the word was being used as, it either added the appropriate features, or explicitly ran a gram­mar rule. Marcus did not give each word a compound lexical entry as we have done here.

Most of the grammar rules in his parser were simple and elegant, but the diagnostics tended to be very complex and contained many conditionals. In some cases they also seemed rather ad hoc and did not meet the goal of a simple, elegant method of handling ambiguity.

For example, consider the THAT-DIAGNOSTIC:

[that][np] - in the Packet CPOOL (Clause pool of rules) "If there is no determiner of second and there is not a qp of second and the nbar of 2nd is none of massn.npl

and 2nd is not-modifiable then attach as det else if c is nbar then label 1st pronoun, relative pronoun else label 1st complementiser."

Notice that if the word that were to be used as a deter­miner, then it would be attached after the NP was built! This is his primary rule for disambiguating the word that. Marcus's parser also had three other rules to handle different cases.

It seems that these rules did not "elegantly capture generalisations" as did the rest of his parser. I consider these rules undesirable and feel that they should be corrected to comply with my criteria for simple and elegant techniques in resolving ambiguity. I wanted a method that used no special mechanism, or routine, other than that needed to parse grammatical sentences. These diagnostics are certainly special mechanisms and do not meet this goal. Can we cover the same examples in a more simple and principled way?

In this section, we look at each of these diagnostics in turn and show how they have been replaced in the newer model. We also look at a few other examples of ambigui­ty which Marcus did not handle, but are related to our discussion here.

The handling of to by Marcus's diagnostic can be replaced by the method outlined in Section 2.5. This method was motivated to handle grammatical sentences and meets our criterion for a simple approach.

For both what and which, the ambiguity lies between a relative pronoun and a determiner. The following exam­ples show various uses of both words:

In ROBIE, that is defined as a singular determiner, a pronoun, a relative pronoun, and a complementiser. Marcus had four diagnostics to handle the word that. We have seen one of these at the start of this section. In this sub-section we see how these four diagnostics can be replaced in a simple way. Let us consider how to handle the uses of that one at a time.

Firstly, as a determiner. The following sentences illus­trate the problem in identifying this usage.

(40) I know that boy should do it.

(41) 1 know that boys should do it.

Marcus assumed that PARSIFAL would be given only grammatical sentences to parse. If determiner/number agreement is not given to a parser, then it will, incorrect­ly, make that a determiner in (41), producing the wrong analysis. The way to prevent this is to enforce number agreement in the rule DETERMINER by insisting that the determiner agree with the noun in number. The deter­miner usage will be grammatical only when the headnoun has the same number. If we make this a condition for the rule to match, then that will not be made a determiner in (41) and ROBIE will get the correct parse.

For this case, the agreement check would make sure that one of the following patterns match:

The above two cases are handled properly because number agreement blocks the interpretation of the (41) as a determiner. This approach leads to the correct pref­erence, when there is an ambiguity and accounts for the difficulty in (42) versus (43):

(42) That deer ate everything in my garden surprised me.

(43) That deer ate everything in my garden last night.

The second experiment in Milne (1983), showed that (42) is a garden path sentence, while (43) is not. In both sentences, it is believed the subject uses the word that as a determiner. Deer is both singular and plural, so it fits the above rule. In (42), that must be used as a complem­entiser to make the sentence grammatical. The approach outlined above will use that as a determiner in an ambig­uous case such as this.

These two simple techniques, word order and agree­ment, are sufficient to handle all the examples we have just presented. In addition, free text analysis has shown no violations to this approach (Milne 1983). These tech­niques provide the same coverage as Marcus's diagnostic, with the added bonus that the determiner is attached before the NP is built.

That can only be a complementiser when a that S- is expected. Hence the rules using that to start an embed­ded sentence are only activated when the verb has the feature THAT-COMP. The rules in THAT-COMP will fire when that is followed by something that can start an NP. This ensures that the S- will have a subject and means that that will be taken as a pronoun in the following sentences:

(44) I know that hit Mary.

(45) I know that will be true.

but it will be taken as a complementiser in these sentences:

(46) I know that boys are mean.

(47) I know that Tom will hit Mary.

It seems that, unless the S- has a subject, the pronoun use of that is preferred. Otherwise one would have a complementiser followed by a trace, rather than a unmarked complementiser, followed by a pronoun. This rule provides more complete coverage than Marcus's diagnostic since it examines the second buffer.

The rule to handle pronouns in general is of low prior­ity and will only fire after all other uses have failed to match. That is treated in the same way.

That will be identified as a relative pronoun only if it occurs after a headnoun and the packet NP-COMPLETE is active. This situation will be handled in the same manner as the usual relative clause rules and will then cover:

(48) I know the boy that you saw.

(49) I know the boy that hit you.

The most difficult case for that is when the verb is subcategorised:

That is, it can take an NP subject, followed by a that S-. For these examples, ROBIE may have to decide if the series of words following that is a relative clause or an embedded sentence.

In the following sentences, the lookahead would have to be more than three buffers. (Brackets indicate words in the buffers. The last word is the disambiguating word.)

It can be seen that in these sentences the disambiguat­ing word is outside our three buffers. How do people handle these, and what should our parser do? In Milne (1983) it was shown that when the syntax could not resolve the ambiguity with its two-buffer lookahead, the decision of which interpretation to use might be made using non-syntactic information. It was also stated that if context can affect the interpretation of the sentence, then non-syntactic information is being used to select the

Robert MUne

interpretation. The reader can experiment for himself and see that context does affect the interpretation of these sentences. Therefore it is predicted that non-syn­tactic information is being used to interpret these sentences, and that this problem should be resolved not on a semantic basis but on a non-syntactic one.

This explains why some of these examples cause diffi­culty and others do not. The psychological evidence from cases using that is scant, and I feel no conclusions can be reached here. My theory predicts that context will strongly affect these examples and, if they are strongly biased to the incorrect reading, a garden path should result.

One well-known example in this area is (52):

(52) I told the girl that I liked the story.

(53) I told the girl whom I liked the story.

(54) I told the girl the story that I liked.

These examples were tested in Milne (1983). The results suggested that (52) was read faster than the other two examples. Many of the subjects were questioned informally after the experiment about their interpretation of the sentence. All reported only one meaning; the S-reading. None of the subjects said that they noticed the relative clause reading, hence the result. The experiment however, was not designed formally to distinguish these.

To handle the examples we have seen in this section, Marcus had four diagnostics, one of which was very complicated. I have just shown how to handle all four cases of that without any special rules, merely substitut­ing enforced agreement and rejecting ungrammatical sentences.

Let us now look at the elimination of Marcus's HAVE-DIAGNOSTIC in relation to the use of agreement we have been discussing in this section. The problem with have is illustrated by the following sentences:

(55) Have the students take the exam.

(56) Have the students taken the exam?

In these, we must decide if have is an auxiliary verb or a main verb and whether the sentence is a yes-no question or an imperative. The sentences have the same initial string until the final morpheme on take. To handle this case, Marcus (1980:211) used this rule:

"RULE HAVE-DIAG PRIORITY: 5 IN SS-START

If 2nd is ns,n3p or 3rd is tenseless

then run imperative next else If 3rd is not verb

then run yes-no-question next else if not sure, assume it's a y/n-q and run yes-no-question next"."

This rule seems to be necessary in order to distinguish between the question and the imperative. If one tries to ascertain exactly what occurs, the apparent complexity is revealed. Note also that Marcus's rule defaults to a yes-no question twice in this diagnostic. The following sentences illustrate the distinction this rule makes.

(57) Have the boy take the exam.

(58) Have the boy taken the exam.

(59) Have the boys take the exam.

(60) Have the boys taken the exam?

It can be seen that YES-NO QUESTION should run only when the NP following is plural and the verb has "en" (i.e., taken). [Only (60) has a plural noun, the boys, and the verb taken.] This can also be understood as: the sentence is an imperative if the item in the 2nd buffer is not plural and the verb is tenseless. Thus, the first three examples above are Imperatives because either the noun (boy) is singular (57 and 58) or the verb is tenseless (59). The second part of the rule takes care of the fact that the third buffer must contain a verb for the imperative, as this would be the main verb of the embedded sentential object.

Let us look more closely at the reason why only (60) is a question. Firstly, if the sentence is a yes-no question, then aux-inversion must occur. When this happens, Have will be adjacent to the verb that was in the third buffer. In order for ROBIE to continue, the verb must have an "en" ending, or have and the next verb will not agree in aspect. This is the basis for discrimination in the earlier examples (57-60).

Secondly, in (57) and (58), the noun phrases are singular and both sentences are imperatives. Had the sentence been a yes-no question, have would need to agree with the subject, which must then be plural.

Hence, in effect, Marcus's rule checks for number agreement between the subject and verb, and checks that the fixed order of the verb group is obeyed. Let us now look at other situations where this is necessary.

PARSIFAL would accept the following ungrammatical strings:

For a yes-no question, the inverted auxiliary must agree with the verb after it has been inverted. To stop these ungrammatical constructions, we must enforce verb agreement. The pattern for the rule YES-NO QUESTION should be:

[auxverb][np][verb], agree(auxverb,verb),agree(verb,np).

This constraint enforces agreement of the verb and auxiliary verb and the subject and verb. Again this check is based only on the linguistic features of the buffers.

Such a constraint effectively blocks the ungrammatical constructions. (The parser will fail if the auxiliary has been inverted, since the auxiliary will not be parsed.) Also the subject NP must agree with the auxiliary verb, so we can also add "agree(auxverb,np)" to the rule, as we did with the HAVE-DIAGNOSTIC! So, by correcting the yes-no question rule, the HAVE-DIAGNOSTIC is redundant.

In this section we have seen that Marcus's HAVE-DIAGNOSTIC can be replaced by merely exploit­ing agreement. It should be pointed out that although this approach has the same coverage as Marcus's diag­nostic, it is wrong in some cases. Milne (1983) has a full discussion.

There is a class of ambiguities that can be resolved mere­ly by enforcing subject/verb agreement. In this section, we see an example from the class of words with the features noun, verb, final-s (plural). If we have two words that can be a plural noun or a singular verb, we can enumerate four cases. Let us look at these possibil­ities and see that these cases can be disambiguated by simple rules using subject/verb agreement. The follow­ing examples illustrate all the possibilities:

(65) The soup pot cover handle screw is red.

(66) The soup pot cover handles screw tightly.

(67) The soup pot cover handles screws tightly. {68} The soup pot cover handle screws tightly. {69} The soup pot cover handle screws are red.

Each of the words pot, cover, handle, and screw can be either a noun or a verb. The "end of constituent" prob­lem is to find out which word is used as the verb and which words make up the complex headnoun. The possi­ble distributions of the morpheme "s" among two words gives us four cases. We deal with each of these in turn.

Case 1: In (65) each noun is singular. For this case all ambiguous words must be nouns and part of the head­noun. Due to subject/verb agreement, a singular noun must match a 3rd person singular (v3s) verb, i.e, one without the letter "s". This case excludes that possibility since none of the words have an "s" at the end. Hence they must all be nouns.

Case 2: In (66) handles is a plural noun and each word before it must be a noun. When a singular noun/verb word follows handles, the word (screw) must be a verb and handles is the last of the headnouns. It is not possi­ble to use handles in this situation as a verb, and screw as a noun because of subject/verb agreement.

Case 3: The examples in this case have two consec­utive plural nouns as in (67), where both words have noun/verb ambiguity. (Do not confuse plural "s" with possessive "'s").

When the first plural is a noun, then the second one can be a verb only if it is part of a different constituent. Examples of this are the following. (Sentences beginning with "?" are considered grammatical but unacceptable to most readers.) (70) ?The soup pot machine handles screws easily.

(71) The soup pot machine handles screw easily.

(72) Which years do you have costs figures for?

(73) Do you have a count of the number of sales requests and the number of requests filled?

Because there is a non-plural headnoun followed by a plural headnoun, this case is really a subset of Case 4. In general, the problems and issues for Case 4 dominate the resolution of this ambiguity.

Case 4: Sentences (68) and (69) both have the same word initial string until after screws, but in (68) screws is a verb while in (69) screws is part of the headnoun. In this situation, where the final word in a series is plural, each word before it must be a noun. The word itself can be either a noun or a verb, depending on what follows. These can be recognised as a pair of potential garden path sentences, as discussed in Milne (1982). Therefore, this is the case to which the Semantic Checking Hypoth­esis applies and the predictions of Milne (1982) apply.

In that paper, the idea of potential garden path sentences is presented. These are sentences that may or may not lead to a garden path. Each garden path sentence has a partner, which is similar but not a garden path. It is proposed that the decision as to how to resolve the ambiguity that may lead to a garden path should be made by semantics and not by syntax. This theory is called the Semantic Checking Hypothesis. For full details see Milne (1983).

In this section, we have looked at resolving a simple case of noun/verb ambiguity. In order to resolve this ambiguity, it was necessary merely to exploit agreement between the subject and verb in number and person.

Due to number and subject verb agreement, these facts have a linguistic base. They rely on the fact that a final "s" marks a plural noun but a singular verb. If the verb is v3s (verb agrees with a 3rd person, singular noun, as with the "s"), then the subject of the verb must be singular, or else the sentence is ungrammatical. This is why all the words before the v3s word must be nouns. If any of these words were used as a verb, then subject-verb agreement would be violated. This is why (67) is ungrammatical. If the verb is v-3s (agrees with any noun phrase except 3rd person, singular i.e., no "s"), then the subject cannot be singular. (65) has no plural subject and so cannot have a v-3s verb. In (66) handles provides a plural subject, so screw, which is v-3s, can agree.

We now consider noun/modal ambiguity as demon­strated by can and will. Both can be either a noun or a modal (i.e., could, should, would, can, will, might, etc.):

(74) The trash can was taken out.

(75) The trash can be taken out.

(76) The paper will was destroyed.

(77) The paper will be destroyed.

Another problem is the word her, which can be used as a pronoun or as a possessive pronoun. Note that we can say:

(82) Tom kissed her.

(83) Tom kissed her sister.

Clearly in (82) her is a pronoun and in (83) her is a possessive determiner. When multiple part-of-speech definitions were added to ROBIE and the simple disam­biguation method used, ROBIE always made her a posses­sive determiner.

This difficulty arose in Marcus's parser because the rule to start a NP was ordered before the rule to parse a pronoun. These rules were copied directly into ROBIE's grammar. Since the word her has both the features "ngstart" and "pronoun", it could match both rules. Unfortunately, as Marcus's rules were stated, it always matched the NP starting rule, and hence was used as a determiner by the parser. This indicates one problem that can arise in the writing of a parser grammar.

To handle possessive determiners, PARSIFAL and ROBIE have a rule with the pattern:

This rule will match a possessive pronoun after it has been made into an NP. It will also match any possessive NP, such as: the boy's or the boy's mother's. The rule then adds the feature "determiner" to the NP, making it eligi­ble for the NP starting rule. By degrading the possessive NPs to determiners, both parsers easily handle examples of left branching such as:

(84) The boy's mother's brother is his uncle.

Another problem arose in (82) because the possessive NP rule was not sufficiently constrained. It is possible to use her as a determiner only where the next word can be part of a noun phrase with that determiner. To enforce this, the second buffer is checked to be certain that its contents will take the determiner. Using this approach, her in (82) would not be converted to a possessive deter­miner. The rule DETERMINER can run only if the next item will "take a determiner".

This check is made by the syntactic category of the following word, rather than by a specially marked feature. This check could be done by having a list of all the possible categories as the pattern of the second buff­er. As an implementation detail, this is in the form of an agreement check, merely to simplify this rule and to show its generality.

The only remaining problem occurs when the verb can take one or more objects and the item after the word her can be either the second object, or an NP with her as a determiner. For example:

(85) I took her grapes.

(86) He saw her duck.

(87) I gave her food for the dog.

The examples presented above are all examples of global ambiguity, which is discussed in more detail in Milne (1983). In these cases the check of "Will the next word take a determiner?", may or may not lead to the wrong analysis. This problem also interacts with the top-down component of verb phrase parsing and the semantic restrictions presented by it.

The conflict between the determiner and possessive usage can be modelled as a conflict of rule priorities. If the possessive use is preferred, then this rule should match first. Conversely, if the object use is preferred, then the object rule should match first. Any error in reading these examples would be due to one rule having priority over the other, when the reverse should be the case. Finally, notice that with no help from either into­nation or context, either analysis is possible. That is, there is not enough information in the sentence to deter­mine a unique interpretation.

We have now shown how to replace all the diagnostics Marcus used. In doing this, we enforced number and verb agreement on the rules before they could run. This was motivated to reject ungrammatical items, rather than for the handling of ambiguity. While there are still a few problems due to global ambiguity, the approach reported here has the same coverage as Marcus's diagnostics, and provides a better explanation of why people have trouble on certain sentences.