An Overview Of The Garden Path Model English Language Essay

Compare and contrast the Garden Path Model and the Constraint Based Model of sentence processing. What are their relative strengths and weaknesses?

Human sentence processing is an intricate process governed by syntactic and language rules. The goal of all research in this field is to discover how people use language. (Ferreira, Christianson and Hollingworth 2001). For humans language is an essential part of what entails being a human and the main feature that differentiates us from other beings. Life without language is difficult to comprehend since it is what dominates our social and cognitive activity. The Psychology of language concerns itself with the psychological processing that takes place when communicating with other people.

There are two main levels of analysis in the comprehension of sentences. Firstly there is the analysis of the syntactical structure of each sentence, which is known technically as ‘parsing’. Secondly there is the study of the intended meaning of each sentence, which is known as ‘pragmatics’. Sentence processing research investigates how people process the arrangement of words in a sentence focusing particularly on syntactic ambiguity. (REF)

This essay details and evaluates two theories that dominate research on syntactic ambiguity resolution i.e. parsing: the garden path model and the constraint-based model.

According to the two-stage theory of the garden path model, the sentence processor initially adopts a single analysis using only a restricted range of information. (REF)

In contrast, constraint based theories claim that multiple analyses of a syntactic ambiguity are activated in parallel and the processor immediately uses all sources of information. (REF)

An Overview of the Garden Path Model:

The most prominent and influential psychological theory of sentence processing is the Garden Path Model (Frazier, 1987a; Frazier & Rayner, 1982).

The garden path model is an autonomous 2-stage parser in which initial attachments are made purely on the basis of syntactic information.

According to the garden path model, the sentence processor when dealing with temporarily ambiguous sentences initially uses only information about the syntactic structure of the sentence to adopt a single analysis. Other non-structural sources of information such as the semantics, context and frequency of structures are in use during later stages of processing. When the initial analysis is contradictory with the information that becomes available later, the processor reanalyzes and thus processing difficulty arises.

The garden path model is governed by two main parsing strategies that determine people’s initial attachment in temporary ambiguous sentences: ‘late closure’ and ‘minimal attachment’. Late closure refers to the process of trying to attach new parts of a sentence to the phrase or clause that is currently being processed. (SYK OF LANG REF) . For example in the sentence ‘John went to dinner with the daughter of the professor who was named in the newspaper last week’, who was named in the newspaper the daughter or the professor? If it is decided that it was the professor who was named in the newspaper, the principle of late closure has been followed. The relative clause of ‘who was named’ was attached to the ‘professor’ as it is the most recently processed noun phrase.

The minimal attachment principle states that new incoming material should be attached in such a way that is syntactically legal and results in the least number of syntactic nodes in the phrase structure tree. ( SYK OF LANG REF)

If there is a conflict between the two strategies, minimal attachment takes precedence over late closure.

An overview of the constraint based model

The constraint-based model in contrast is an interactive one-stage model. According to the constraint-based model the processor uses multiple sources of information including syntactic, semantic, discourse and frequency based, called constraints. One of the assumptions of the constraint model is that sentence interpretation is the product of interaction between multiple constraints/sources of information, thus semantic, pragmatic and syntactic information are taken into consideration from the start (MacDonald et al, 1994; Trueswell, et al 1994). .

Although processing is generally quite easy, when two analyses have an equal activation, this results in competition. This would therefore result in processing difficulty as it takes a long time before the correct analysis is recognized and the incorrect one inhibited. There is no form of reanalysis in this explanation as all analyses are activated from the beginning of the ambiguity therefore disambiguation would not require the construction of analyses that were not initially considered.

NEEDS 2 BE FINISHED!!!!!!

There has been a large amount of research testing the two opposing models. These studies investigate weather the use of non-syntactic information is delayed relative to syntactic information.

The evidence for the garden path model comes from a substantial number of experiments. Rayner et al (1982) conducted an eye-movement study from which they concluded that reading time was longer when a sentence was inconsistent with the late closure principle. They presented participants with the sentences:

Since Jay always jogs a mile and a half this seems a short distance to him.

Since Jay always jogs a mile and a half seems a very short distance to him.

The first sentence is consistent with late closure and thus does not cause the processor any problems. The second sentence is not entirely consistent with late closure; the processor tries to attach the noun phrase ‘a mile and a half’ to the first verb ‘jogs’. When we reach ‘seems’ it is clear that this structure is wrong and the reader has been led up the garden path. The word ‘seems’ is a disambiguating region and Frazier and Rayner found that first fixation in disambiguating region is longer for the second sentence compared to the first.

Rayner and Frazier (1987) also monitored the eye movements of participants, to test the minimal attachment principle, when reading sentences such as:

The criminal confessed his sins harmed many people

The criminal confessed that his sins harmed many people.

When the participant begins to read the first sentence, minimal attachment leads to the use of the structure that contains the fewest number of nodes. So in this example when we get to the words ‘his sins’ is the object of ‘confessed’, i.e. we are led to believe that the criminal confessed his sins. Hence the reader is led up the garden-path in this first sentence and must then reanalyze when they read ‘harmed’. Sentence two on the other hand does not lead to a garden path because of the word ‘that’ which blocks the object analysis of the sentence. When the reader gets to ‘that’ he treats the noun phrase ‘his sins’ as a subject of complement ‘harmed many people’. Rayner and Frazier found that readers had a problem when they reached ‘harmed’ in sentence one but not in sentence 2.

Evaluation etc

In serial 2 stage models such as the garden path model, only syntactic information controls the initial analysis, and there is a preference for a second stage that uses the semantic information.

In parallel constraint-based models use multiple analyses from the start and both syntactic and non-syntactic information is used together to activate alternative representations. From the vast research conducted on both models it is difficult to claim any one model being the better account for sentence processing.

The garden path model can be classed a strong model as it provides a simple and coherent account of key processes involved in sentence processing. The constraint-based model also has key strengths; in particular it carries the notion that there can be varying degrees of support for different syntactic interpretations of a sentence. As someone reads a sentence, the accumulating syntactic and semantic evidence gradually leads the reader to produce a definite syntactic interpretation. It seems efficient that readers should use all the relevant information from the start when trying to work out the syntactic structure of a sentence.

Those in favour of the Garden Path Model argue that the effects that are claimed to support constraint based models arise because the second stage of parsing begins very quickly and that many experiments that are supposed to be looking at the first stage are in fact looking at the second stage of parsing. Any interaction observed is occurring at this second stage that begins very early in processing. They argue that experiments supporting constraint based models are methodologically flawed and that constraint based models fail to account for the full range of data.

On the other hand those in favour of the constraint-based model argue that researchers favouring the garden path model use techniques that are not sensitive enough to detect the interactions involved and that the non-syntactic constraints used are too weak. Supporters of the constraint-based model argue that the theory of minimal attachment does not work for all sentences. (ALTMANN ET AL STUDY).

Carreiras and Clifton (1993) found evidence that readers do not always follow the principle of ‘late closure’. They presented sentences such as ‘The spy shot the daughter of the colonel who was standing on the balcony’ If the reader was to follow the principle of ‘late closure’, this sentence would be interpreted as the colonel rather than the daughter that is stood on the balcony. However they did not strongly prefer either interpretation, which is contrary to the principles of the garden path model. When the same sentence was presented in Spanish, there was a clear preference for assuming that it was the daughter who was stood at the balcony. This again is contrary to theoretical predictions. (cog syk a student handbook ref)

According to the garden path model, prior context should not influence the initial parsing of an ambiguous sentence. However several studies have been conducted since that concludes that context does affect the initial parsing process. For example Tannenhaus et al (1995) made participants listen to the ambiguous sentence ‘Put the apple on the towel in the box’ and then recorded participants eye movements to observe how the sentence was interpreted. If the principles of the garden path model were followed the sentence would be interpreted so that it was understood that the apple should be placed on the towel because this is the simplest syntactic structure. When the context did not remove the ambiguity, participants did indeed interpret the sentence as: the apple should be placed on the towel. However when the visual context was two apples, one on a towel and the other on a napkin, participants rapidly interpreted ‘on the towel’ as a way of identifying which apple was to be moved and so did not make the mistake of focusing on the towel presented on its own. Spivey, Tanenhaus, Eberhard and Sedivy (2002) also conducted a similar experiment to the one described above and came to the same conclusion; context has a large effect on eye movement The pattern of eye movements was very similar for unambiguous sentences and for ambiguous sentences with a less confusing context.

There are various limitations with the constraint-based model. Firstly the model is not correct in its view that all relevant constraints or sources of information are used immediately. (Boland and Blodgett 2001 EXPAND). Secondly within McDonald et al’s (1994) theory, little is said about the detailed processes involved in generating syntactic structure for complex sentences. Also it is assumed within the constraint-based theory that various representations are formed in parallel with most of them subsequently being rejected. However there is little evidence to prove this assumption.