Spektator. Jaargang 2

[p. 470]

Hans den Besten, Henk van Riemsdijk, Catherine Snow.
Ambiguous sentences: perceptual strategies?

Recent research in psycholinguistics has centered around attempts to identify sentence-processing strategies (Bever, 1970). An interesting beginning along these lines has been Ruwet's treatment of two superficially identical constituents which behave differently in movement transformations. For example, of the two ‘de’ NP complements in ‘le portrait d'Aristote de Rembrandt,’ only the last one can be moved as described in the following strategy:

All else being equal, in the absence of particular morphological marks or lexical-semantic information, each time a sentence has absolute structural amibuity in surface structure, because one and the same category occurs twice, this sentence is only acceptable in the [...] reading for which at least one of these two occurrences keeps the position it had in deep structure. (Ruwet, 1971)

The NP PP PP sequence which Ruwet treats here has the further characteristic that its components can have four different possible structural relationships with each other dependent on the preceding verb. These structures and sentences containing them are:

Type 1: (V (NP (PP PP)))
De rechter (las (de drukproeven (van het boek (van Koster))))
(The judge read the proofs of the book of Koster)
Type 2: (V (NP PP) PP)
De verzamelaar (kocht (een kopie (van het boek)) (van de antiquair)
The collector bought a copy of the book from the antique dealer)
Type 3: (V (NP PP PP))
De kunstliefhebber (bekeek ((het schilderij) (van de soldaat) (van de amateur)))
(The art lover looked at the painting of the soldier by the amateur)
Type 4: (V NP (PP PP) )
De minister ontving (een rapport) (van de leden (van de commissie)))
(The minister received a report from the members of the committee)

The question arises which of these sentence types is the most difficult to process. This is however impossible to determine directly since it is in

[p. 471]

practice impossible to find enough unambiguous sentences of any type. Therefore we decided to assess the relative difficulty of the types by testing sentences which were systematically ambiguous between two readings (triply ambiguous sentences are not difficult to find) to see which of the interpretations was dominant.

A further point of interest is the comparison between picture nouns and non-picture nouns as head nouns in these structures. Ruwet formulated his strategy on the basis of examples of both. We wished to compare sentences containing picture nouns with sentences containing non-picture nouns because the linguistic description of picture nouns remains an interesting problem. (Ross, 1967)

Method

Materials: Sentences which were ambiguous between types 1 and 2, 1 and 3, 1 and 4, and 2 and 3 were produced. We wished also to test sentences ambiguous between types 2 and 4 and between types 3 and 4, but could not find sentences like this. In two sentences of each type the direct object was a picture noun, and in two sentences of each type the direct object was a non-picture noun. A complete list of the 16 sentences we used is given in the appendix. Twenty non-ambiguous but rather difficult sentences of approximately the same length as the test sentences were used as filler sentences.

Subjects: 43 adult native speakers of Dutch were tested.

Procedure: A formalized instruction was read to the subjects. This explained that the subject would be asked to paraphrase each of a group of sentences after reading the sentence out loud. The sentences were typed on separate sheets of paper and arranged in a booklet. After reading each sentence the subject turned the page and saw a word from the sentence. He was to begin his paraphrase with this word. For the filler sentences this word was any noun or pronoun from the sentence. For the test sentences this word was always one of the last two nouns in the sentenee, which one being balanced across sentence types and across subjects. This begin-word procedure was used to avoid ambiguous paraphrases. Five randomly determined orders of presentation of the test sentences were used. The filler sentences, with which the test sentences were alternated, were always presented in the same order. The subject's paraphrases were recorded for later scoring.

Results

The complete results for eaeh sentence are presented in the Appendix.

[p. 472]

Differences between sentences were relatively minor and so will be ignored here. A clear dominance hierarchy for the sentence types was established, i.e.:

4 > 1 > 2 = 3

This hierarchy was identical for picture nouns and for non-picture nouns except for one case, that of the 1/3 sentences. For these sentences the type 1 interpretation was strongly dominant for the picture nouns, but the type 3 interpretation was slightly dominant for the non-picture nouns. Two sentences (De verzamelaar...and Vrouwen verwerpen...) were found to be unusually difficult, but results for these sentences were not atypical.

Discussion

Although we had no clear hypothesis concerning the order of difficulty at the beginning of this experiment, the results are surprising and no straightforward interpretation immediatly presents itself. The most promising way to explain the 4 > 1 > 2 = 3 hierarchy we found appears to be to postulate more than one perceptual strategy. Three strategies interaeting in specific ways can predict the hierarchy. These three strategies are:

1.

Verbs which take source phrases (receive, buy) establish (a) an expectation that a source will be named and (b) an expectation that a from-phrase which occurs immediately after the direct object will identify that source. This suggests that sentences containing source verbs will be more difficult than sentences containing simple transitive verbs (verbs which take a direct object but no prepositional complement), and that sentences with source verbs followed by a from-phrase which does not identify the source will be more difficult than similar sentences in which the from-phrase does identify the source. This strategy is clearly reminiscent of suggestions made by Fodor, Garrett, and Bever (1968). It may also be the case that if the first occurring from-phrase does not identify the source then the expectation of a source is nullified. Thus type 4, in which the source is identified immediately after the direct object, would be easier than type 2, in which the source identification is delayed, and than type 1, in which the source is never identified. This strategy in its strongest version seems very similar to the type of strategy suggested by Ruwet (1971).

2.

Constituents tend to be closed off as soon as possible. Bever suggested a version of this strategy for processing noun phrases (1970) and some evidence that it is used has been offered by Chapin, Smith, and Abrahamson (1972). On the basis of this strategy type 4, in which the

[p. 473]

direct object ends after the first noun phrase, would be easier than types 1 and 3, in which the direct object is modified by two prepositional phrases, and than type 2, in which the direct object is modified by one prepositional phrase.

3.

Complex or long constituents in the middle of a sentence are more difficult to process than complex or long constituents at the end of a sentence. Although this has never been tested psycholinguistically (as far as we know), there exists a fair amount of linguistic evidence that complex constituents tend to shift into sentence-final position (e.g. Ross, 1967). On the basis of this strategy, type 4 would clearly be dominant over type 2.

If these three strategies all function together but with different strengths or probabilities, then they can explain our results. We are therefore testing for the existence of each strategy independently in three experiments in which short-term memory for sentences and numbers is the dependent variable.

Institute for General Linguistics; University of Amsterdam. November 1972.

References

Bever, T. The cognitive basis for linguistic structures. In: J.R. Hayes (ed.), Cognition and the Development of Language. New York: Wiley, 1970.

Chapin, P.G., Smith, T.S., & Abrahamson, A.A. ‘Two factors in perceptual segmentation of speech’. Journal of Verbal Learning and Verbal Behavior, 11 (1972), 164-173.

Fodor, J.A., Garrett, M., & Bever, T. ‘Some syntactic determinants of sentential complexity, II: Verb structure’. Perception and Psychophysics 3 (1968), 453-461.

Ross, J., Constraints on variables in syntax. Unpub. Ph.D. thesis, MIT, 1967.

Ruwet, N., How to deal with syntactic irregularities: Conditions on transformations or perceptual strategies? Unpub., 1971.

Appendix

Picture-noun sentences.	Type	Times Chosen
De rechter ontving de drukproeven van het boek van Koster.	1/2
The judge received the proofs of Koster's book.	1	32
The judge received the proofs of the book from Koster.	2	7

[p. 474]

De minister ontvangt de eerste druk van het rapport van de commissie.	1/2
The minister receives the first printing of the committee's report.	1	30
The minister receives the first printing of the report from the committee.	2	9
De redacteur bekeek de foto van de wagen van Jansen.	1/3
The editor looked at the photo of Jansen's car.	1	39
The editor looked at Jansen's photo of the car.	3	1
De toeschouwers bekijken de affiches van de massale demonstratie van de Dollc Mina's.	1/3
The spectators look at the posters of the mass demonstration of Women's Lib.	1	26
The spectators look at the Women's Lib posters for the mass demonstration.	3	9
De antiquair kocht het boek van de vriend van Smit.	1/4
The antique dealer bought Smith's friend's book.	1	13
The antique dealer bought the book from Smith's friend.	4	24
100 Kamerleden krijgen handtekeningenlijsten van tegenstanders van de abortuswetgeving.	1/4
100 MP's get lists with signatures of the opponents of the abortion law.	1	11
100 MP's get lists with signatures from opponents of the abortion law.	4	24
De verzamelaar kocht het schilderij van de soldaat van de amateur.	2/3
The collector bought the painting of the soldier from the amateur.	2	14
The collector bought the amateur's painting of the soldier.	3	11
De marechaussee vordert spandoeken van wit linnen van de demonstranten.	2/3
The police confiscate white linnen banners from the demonstraters.	2	15
The police confiscate the demonstraters'white linnen banners.	3	22
Non-picture noun sentences
De stalknecht vernam de aankomst van 10 schimmels van Boltini.	1/2
The stableboy heard about the arrival of Boltini's 10 horses.	1	36
The stableboy heard about the arrival of 10 horses from Boltini.	2	3
De politie verneemt de verdwijning van de woonwagens van de Zigeuners.	1/2
The police are informed of the disappearance of the Gypsies'caravans.	1	32
The police are informed of the caravan's disappearance by the Gypsies.	2	6
De columnist voorzag de verwijdering van de stafleden van de faculteit.	1/3
The columnist foresaw the removal of the staffmembers of the faculty.	1	14
The columnist foresaw the removal of the staffmembers from the faculty.	3	24
Vrouwen verwerpen afhankelijkheid van abortusteams van Volksgezondheid.	1/3
Women reject dependence on abortion-teams from Public-Health.	1	18
Women reject the dependence of abortion-teams on Public Health.	3	13

[p. 475]

De commissaris verwijderde het insigne van het uniform van de agent.	1/4
The commissioner removed the policeman's uniform-insignia.	1	13
The commissioner removed the insignia from the policeman's uniform.	4	22
De arts leent de wagen van de organisatoren van de actie.	1/4
The doctor borrowed the movement's organiser's car.	1	13
The doctor borrowed the car from the movement's organisers.	4	22
De agent vernam de verwijdering van de demonstranten van het hoofdburo.	2/3
The policeman was informed of the removal of the demonstraters by the headoffice.	2	21
The policeman was informed of the removal of the demonstraters from the headoffice.	3	18
De krant verneemt de verwijdering van de Zigeuners van Publieke Werken.	2/3
The newspaper was informed of the removal of the Gypsies by Public Works.	2	20
The newspaper was informed of the removal of the Gypsies from Public Works.	3	19