The Garden Path Model Of Sentence Processing

A central aim of psycholinguistic enquiry is to provide a logical account of how individuals construe structurally ambiguous sentences (Ni, Grain & Shankweiler, 1996). Human sentence processing is an intricate process directed by rules of language and syntax (Ferreira, Christianson & Hollingworth, 2001). Of late, researchers have begun to investigate and explain the human mechanics of how individuals build meaningful sentences from words (Whitney, 1998). A key issue of debate, on sentence processing, is whether language processes are integrated or not. Indeed, Fodor (1983) and Frazier and Fodor (1978) argued that an internal central processor contained a number of autonomous modules that matched to phonological, lexical and syntactic components of language. According to Harley (2010) the prerequisite of research is to identify whether sentence parsing is an innate syntactic module that only uses structure to parse sentences. However, researchers such as Trueswell, Tanenhaus and Garnsey (1994) have argued that semantics can guide individuals into parsing sentences.

The debate whether sentence processing is interactive or modular can be related to the models of parsing. According to Frazier and Fodor (1978) models of parsing go some way in explaining how individuals construe the syntactic structure of a sentence. In one-stage models, syntactic and semantic rules are used simultaneously to construct a syntactic representation. However, in two-stage models semantic data is drawn upon in the second stage (Whitney, 1998). Investigators have used ambiguous sentences to study how individuals parse. This appears practical as ambiguous sentences distort processing ability. Indeed, Harley (2010) noted that “it is difficult to discern the operations of the processor when all is working well. For this reason, most research on parsing has involved syntactic ambiguity” (p. 289).

Best services for writing your paper according to Trustpilot

Premium Partner
From $18.00 per page
4,8 / 5
Writers Experience
Recommended Service
From $13.90 per page
4,6 / 5
Writers Experience
From $20.00 per page
4,5 / 5
Writers Experience
* All Partners were chosen among 50+ writing services by our Customer Satisfaction Team

A noteworthy model of sentence processing was developed by Frazier and Fodor (1978), called the ‘Garden Path Model’ (GPM). According to Altmann, Garnham and Dennis (1992) and Ferreira, Bailey and Ferraro (2002) the model proposes that the language processor initially calculates the syntactic structure of a sentence without analysing or considering the context. Consider the sentence “Mary saw the man with the binoculars”. Most individuals would interpret the sentence to mean that Mary used the binoculars as a tool to see the man. However, it could also mean that the man had the binoculars. According to Ferreira and colleagues when an individual encounters an ambiguous sentence only one meaning is primarily processed. If the meaning attributed does not make sense, the sentence is reparsed using the available contextual knowledge available. This essay will critically evaluate the Garden Path Model of sentence processing. In particular, it will look at the Constraints Based Model of parsing and provide evidence in support of each model.

The GPM is a two-stage parser that fundamentally relies on two main principles; ‘minimal attachment’ and ‘late closure’ (Altmann, Nice, Garnham & Henstra, 1998). According to Ferreira and Clifton (1986) minimal attachment suggests information should be attached to the phrase marker using the fewest syntactic nodes possible. While late closure takes arriving information and attempts to incorporate it into the clause or phrase being processed at the time. However, if conflict arises between minimal attachment and late closure, minimal attachment will take priority (Fodor, 1983). For example, consider the sentence “the horse raced past the barn fell” (Bever, 1970; Ferreira et al., 2002). When this sentence is read, it can be interpreted as grammatically correct until the final word ‘fell’. According to Harley (2010) when an individual reaches the word ‘fell’ they realise that they have been led up the garden path. Thus, realising that the first interpretation was incorrect and the sentence has to be re-analysed. After back tracking and re-analysing the sentence, the correct understanding of the sentence is construed (“the horse that was raced past the barn was the one that fell”). Frazier and Fodor (1978) pointed out that the GPM processor ascribes the noun phrase to the first verb, thus when ‘fell’ appears it appears evident that the sentence is incorrect.

In support of minimal attachment and late closure Rayner and Frazier (1987) conducted a study in which eye movement was compared during parsing of two sentences. The first sentence “The criminal confessed that his sins harmed many people” was a control sentence that had no ambiguity. The second sentence “The criminal confessed his sins harmed many people” was phrased in a way that led the participant up the garden path. According to Harley (2010) and Rayner, Carlson, & Frazier, 1983) the latter sentence leads the parser to treat the noun phrase ‘his sins’ as a direct object of confessed instead of the subject of ‘harmed’. Minimal attachment postulates when the participant reaches ‘harmed’ they will realise they have been led up the garden path and need to re-analyse the sentence to make sense of it. Indeed, Rayner and Frazier (1987) found that the sentence that contained ambiguity, led to participants experiencing more difficulty, than the sentence which was unambiguous. The evidence shown by Rayner and Frazier suggests that the parser is conforming to minimal attachment and late closure. That is, individuals minimise the cognitive processing load by attaching a new clause in a way that is syntactically permissible but uses the fewest nodes possible. Likewise, Ferreira and Clifton (1986), Frazier and Rayner, (1982) and Mitchell (1987) support the notion that syntactic analysis is processed autonomously without any need of guidance from semantic factors.

Nevertheless, caution is needed when drawing conclusions from studies proposing that sentence processing is modular. Studies appear to rely on the arguments of null effects. A failure to find a significant effect of semantic sentence processing does not mean that semantics are not used. It is possible that the manipulations used in such studies are relatively weak and are not sensitive to such an effect.

However, there is evidence from electroencephalogram (EEG) and event-related potential (ERP) studies that support the GPM. Both EEG and ERP’s measure semantic and syntactic anomalies (electrical activity) in specific brain areas (Datta et al., 2004). An occurrence of a semantic anomaly creates a negative ERP of 400ms (N400), while a positive syntactic irregularity creates an ERP of 600ms (P600) (Osterhout & Nicol, 1999). For example, in the sentence “the horse raced past the barn fell” a syntactic anomaly will arise because the word ‘fell’ does not fit the sentence. According to Harley (2010) the P600 is produced in response to a grammatical error. It is thought that P600 is an electrophysiological marker of the GPM (Osterhout, Holcomb & Swiney, 1994). This suggests that syntactic handling is autonomous and that sentence structure is independent of semantics when processing sentences. However, there are two significant flaws to this study; firstly, there is little clarification on exactly why or how the abnormality occurs. Secondly, EPR’s are imprecise in determining what area of the brain is producing the activity. Thus it is possible that syntactic and semantic processing may run simultaneously, indeed, this is the premise put forward by the ‘Constraint Based Model’ (CBM) of Parsing (Whitney, 1998).

The CBM suggests that semantic factors are used in sentence processing. According to Tanenhaus, Spivey-Knowlton, Eberhard and Sedivy (1995) the CBM proposes that garden path sentences are analysed using syntactic and semantic information in a one-stage process. Indeed, Altmann and Steedman (1988) and Trueswell, Tanenhaus and Kello (1993) argued that the ‘Incremental Interactive Theory’ (IIC) proposes, contrary to the garden path theory, that context does influence the way individuals parse sentences.

To support this argument, Altmann et al. (1992) conducted an experiment that measured eye movement in response to how context affects garden path sentences. Consider the sentence “The fireman told the man that he had risked his life for to install a smoke detector”. The GPM postulates that the sentence should lead to a garden path effect. However, the IIC claims that ambiguous sentences can be resolved if a unique referent can be found. Accordingly, the resolution of ambiguities is context dependent, i.e. the context can bias the processor towards or away from a garden-path effect. According to Harley (2008) one-way of approaching this ambiguity is to provide more than one possible referent for ‘the man’ in the sentence. For instance, “A fireman braved a dangerous fire in a hotel” or “He rescued one of the guests at a great danger to himself”. Indeed, measurements of eye movements verified that garden path ambiguities lead the processor to construct representation based on the best semantic fit.

Similarly, Altmann and Steedman (1998) measured reading times on the following ambiguous sentences:

“The burglar blew open the safe with dynamite and made off with the loot”

“The burglar blew open the safe with the new lock and made off with the loot”.

Participants were exposed to context specific sentences of either one or two safes. According to Harley (2010) if one safe was mentioned, (S2) would be more difficult to read because ‘the safe with the new lock’ would be made redundant, causing additional processing difficulty. Likewise, if the context sentence mentioned two safes (S1) would be more difficult to read because the noun phrase ‘the safe’ fails to identify either safe. Altmann and Steedman propose that the human processor syntactically parses and interprets the sentence on a word by word basis, then simultaneously makes a decision using the context.

The GPM argues that semantics take no part in the first stage of sentence processing. However, the studies cited claim that the CBM is plausible and that semantics do play an important role in processing sentences. Nevertheless, the CBM has been criticised. Mitchell (1994) argued that it is difficult to rule out modularity, it is conceivable that syntactic information may be available foremost, then within a couple of hundred milliseconds semantics follow, rather than being activated simultaneously. In short it is difficult to separate the modular and interactive views. One conceivable reason is that context is not always immediately available and so has to be retrieved from memory, thus takes longer to compute.

To investigate this hypothesis Tanenhaus et al. (1995) developed a method for studying spoken language processing. The paradigm monitored eye movements as participants listened to sentences that instructed them to move or touch objects, to see if visual context influenced initial parsing. Participants listened to the sentence “Put the apple on the towel in the box”. GPM predicts that the participant will interpret the sentence as ‘place the apple on the empty towel’ until they hit the words ‘in the box’, then the participant will realise they have been led up the garden path. Tanenhaus and colleagues manipulated the context in that there was only one apple on the towel, or two apples with one on the towel. Results confirmed that participants were incorporating the visual information and auditory words. Therefore they were not led by minimal attachment. Data from both written and spoken studies provided evidence that context can influence sentence processing.

In conclusion, there is empirical evidence to suggest that syntactic preferences such as minimal attachment and late closure are independent of semantic and contextual factors. The GPM is a two-stage parsing model that assumes that minimal attachment operates during the parsers first pass of the sentence. Only later the parser consults contextual or semantic information (Frazier and Fodor, 1978; Altmann et al., 1992; Ferreira et al., 2002). Further support from EEG and ERP studies also support this view.

However, investigators argue a modular view relies too much on the null effect and that manipulations are too weak to detect semantic processing. Consequently, it would be irrational to suggest that semantics and contextual information are not involved in parsing. Indeed, investigators who favour the CBM (Altmann and Steedman, 1988; Trueswell et al., 1993; Tanenhaus et al., 1995), suggest that the human processor syntactically parse and interprets the sentence on a word by word basis, then simultaneously makes a decision using the context. Similar results have been found in auditory parsing studies.

Largely, the evidence points to the fact that the GPM is wrong. The CBM identifies an important phenomenon in that individuals use semantics to parse sentences. However, the CBM only works well as a critique of the GPM that is; it only gives a framework for explaining sentence processing not a theory. Future research in sentence processing needs to be aimed at developing a theory that helps investigators to make specific predictions on when interaction effects will occur and when they will not.


Altmann, G., & Steedman, M. (1988). Interaction with context during human sentence processing. Cognition, 30, 191-23.

Altmann, G. M., Garnham, A. A., & Dennis, Y. Y. (1992). Avoiding the garden path: eye movements in context. Journal of Memory and Language, 31(5), 685-712.

Altmann, G. M., Van Nice, K. Y., Garnham, A. A., & Henstra, J. A. (1998). Late closure in context. Journal of Memory and Language, 38 (4), 459-484.

Bever, T. G. (1970). The cognitive basis for linguistic structures. In J.R Hayes (ed), Cognition and the development of language (pp.279-362). New York: Wiley.

Datta, S., Lyon, I., Mackintosh, B., McLannahan, H., Murphy, K., Naish, P., Nettle, D., Romero, I., Toates, F., & Whatson T. (2004). Learning Language. Milton Keynes.: Open University.

Ferreira, F., Bailey, K. D., & Ferraro, V. (2002). Good-enough representations in language comprehension. Current Directions in Psychological Science, 1, 11-15.

Ferreira, F., Christianson, K., & Hollingworth, A. (2001). Misinterpretations of garden-path sentences: implications for models of sentence processing and reanalysis. Journal of psycholinguistic research, 30(1), 3-20.

Ferreira, F., & Clifton, C. R. (1986). The independence of synaptic processing. Journal of Memory and Language, 25(3), 348-368.

Fodor, J. A. (1983). Modularity of mind. Cambridge: MIT Press.

Frazier, L., & Fodor, J. D., (1978). The sausage machine: a new two stage parsing model. Cognition, 6, 291-326.

Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14, 178 – 210.

Harley, T. (2008). The Psychology of Language: from data to theory. New York: Psychology Press.

Mitchell, D. C. (1987). Lexical guidance in human parsing: locus and processing characteristics. In M. Coltheart (Ed.), Attention and performance XII: The psychology of reading (pp. 601-618). Hillsdale, NJ: Erlbaum.

Mitchell, D. C. (1994). Syntactic processing. In M. A. Gernsbacher (ed), Handbook of Psycholinguistics. New York: Academic Press.

Ni, W., Grain, S., & Shankweiler, D. (1996). Sidestepping garden paths: assessing the contributions of syntax, semantics and plausibility in resolving ambiguities. Language And Cognitive Processes, 11(3), 283-334.

Osterhout, L., Holcomb, P.J. & Swinney, D.A. (1994). Brain potentials elicited by garden-path sentences: evidence of the application of verb information during parsing. Journal of experimental psychology Learning, memory, and cognition, 20(4), 786-803.

Osterhout, L., & Nicol, J. (1999). On the distinctiveness, independence, and time cause of the brain responses to syntactic and semantic anomalies. Language and Cognitive Process , 14, 283 – 317.

Rayner, K., Carlson, M., & Frazier, L. (1983). The interaction of syntax and semantics during sentence processing: eye movements in the analysis of semantically biased sentences. Journal of verbal Learning and Verbal Behaviour , 22, 258 – 374.

Rayner, K., & Frazier, L. (1987). Parsing temporarily ambiguous complements. Quarterly Journal of Experimental Psychology, 39(A), 657-673.

Tanenhausm, M. K., Spivey-Knowlton, M.J., Eberhard, K. M., & Sedivy, J. C. (1995). Integration of visual and linguistic information in spoken language comprehension. Science, 268, 1632-1634.

Trueswell, J. C., Tanenhaus, M. K., & Kello, C. (1993). Verb-specific constraints in the sentence processing: separating effects of lexical preferences from garden paths. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 528-553.

Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing: use of thematic role information in syntactic ambiguity resolution. Journal of Memory and Language, 33(3), 285-318.

Whitney, P. (1998). The Psychology of Language. New York: Houghton Mifflin Company.

You Might Also Like