Introduction
Sentence parsing, the process by which individuals interpret and assign structure to linguistic input, remains a central topic in psycholinguistics. Understanding how ambiguous or complex sentences are processed provides insight into the cognitive mechanisms underlying language comprehension. The sentence “When Fred eats food gets thrown” presents a particular challenge due to its potential garden-path structure and syntactic ambiguity, which can lead to multiple interpretations. This essay explores how different theoretical models of sentence parsing—namely the serial processing model, the parallel processing model, and constraint-based approaches—account for the interpretation of this sentence. Drawing on empirical evidence associated with each framework, the discussion highlights their strengths and limitations in explaining human parsing behaviour. The purpose of this essay is to evaluate these models, considering their applicability to real-world language processing and their implications for understanding cognitive mechanisms in psychology.
Serial Processing Model and Garden-Path Effects
The serial processing model, often exemplified by the Garden-Path Theory, posits that language comprehension occurs sequentially, with the parser initially adopting a single syntactic structure based on heuristic principles such as minimal attachment or late closure (Frazier and Rayner, 1982). In the case of the sentence “When Fred eats food gets thrown,” a serial parser might initially interpret “food” as the direct object of “eats,” following the simplest structural analysis. However, upon encountering “gets thrown,” the parser must reanalyse the structure, recognising that “food” is likely the subject of a new clause rather than the object of the first. This reanalysis aligns with the garden-path effect, where initial misinterpretation leads to processing difficulty.
Empirical support for this model comes from eye-tracking studies, which demonstrate increased fixation times and regressions when readers encounter disambiguating information in garden-path sentences (Frazier and Rayner, 1982). For instance, Rayner and Frazier (1987) found that readers exhibit longer reading times at points of syntactic ambiguity, suggesting that reanalysis imposes a cognitive load. Applied to the sentence in question, this model predicts processing difficulty at “gets thrown,” as the parser must backtrack to correct the initial misparse. However, a limitation of the serial model is its inability to account for individual differences or contextual influences, as it assumes a uniform parsing strategy. Thus, while it offers a clear explanation for initial errors in interpreting ambiguous sentences, it may oversimplify the complexity of human language processing.
Parallel Processing Model and Multiple Interpretations
In contrast to the serial model, the parallel processing approach suggests that the parser simultaneously considers multiple syntactic structures during comprehension, weighting them based on probability or frequency (MacDonald et al., 1994). For the sentence “When Fred eats food gets thrown,” a parallel parser would entertain both the interpretation of “food” as the object of “eats” and as the subject of a new clause from the outset. The parser eventually resolves ambiguity by selecting the structure with the highest likelihood or consistency with the input as more information becomes available.
Empirical evidence for parallel processing is derived from studies using ambiguity resolution tasks. For example, MacDonald et al. (1994) demonstrated that readers are sensitive to frequency-based cues when resolving ambiguous structures, maintaining multiple parses until disambiguation. In the given sentence, a parallel model might predict less pronounced processing difficulty compared to the serial model, as alternative structures are already under consideration. However, this approach can be criticised for its computational implausibility, as maintaining multiple parses simultaneously may be cognitively resource-intensive. Furthermore, it struggles to explain why certain ambiguities, such as garden-path sentences, consistently result in processing disruptions if all possibilities are equally available. Despite these limitations, the parallel model offers a more flexible account of how ambiguous inputs like the sentence in question are interpreted.
Constraint-Based Approaches and Contextual Influences
Constraint-based models propose that sentence parsing is guided by a combination of syntactic, semantic, and contextual constraints, which interact dynamically to influence interpretation (Trueswell and Tanenhaus, 1994). Unlike serial or parallel models, which prioritise syntactic structure, constraint-based approaches suggest that multiple sources of information—such as lexical frequency, thematic roles, and discourse context—shape parsing decisions. For the sentence “When Fred eats food gets thrown,” a constraint-based parser might initially favour “food” as the object of “eats” due to lexical frequency (as “eat” typically takes a direct object). However, semantic plausibility and the subsequent input “gets thrown” would prompt a rapid re-evaluation, potentially interpreting “food” as the subject of a passive construction.
Empirical support for this model is evident in studies demonstrating the influence of contextual and probabilistic cues on parsing. Trueswell et al. (1993) found that semantic fit and discourse context significantly affect ambiguity resolution in sentences with temporary syntactic ambiguity. Applied to the target sentence, a constraint-based approach might suggest that real-world knowledge (e.g., food cannot “get thrown” by itself in an active sense) guides the parser towards a passive interpretation. While this model excels in incorporating diverse linguistic and non-linguistic factors, it can be critiqued for lacking specificity in predicting when and how certain constraints take precedence. Nevertheless, its emphasis on interaction between multiple information sources makes it particularly relevant for understanding complex or ambiguous sentences.
Comparative Analysis and Implications
Comparing the three models reveals distinct strengths and limitations in accounting for the interpretation of “When Fred eats food gets thrown.” The serial processing model effectively predicts initial misparsing and reanalysis, supported by evidence of processing difficulty in garden-path structures (Frazier and Rayner, 1982). However, it overlooks contextual and individual variability. The parallel processing model offers a more nuanced perspective by considering multiple interpretations simultaneously, though it struggles to explain consistent processing disruptions (MacDonald et al., 1994). Finally, the constraint-based model provides the most comprehensive framework by integrating diverse linguistic cues, though it lacks precision in prioritising constraints (Trueswell and Tanenhaus, 1994).
Indeed, the interpretation of the target sentence underscores broader challenges in psycholinguistic research. Each model captures different aspects of human parsing but fails to fully account for the interplay of cognitive and linguistic factors. For instance, while serial models highlight structural preferences, they may not explain why some readers resolve ambiguity more efficiently, possibly due to experience or context. Arguably, a hybrid approach combining elements of all three frameworks may offer a more accurate representation of sentence processing, though this requires further empirical validation.
Conclusion
In summary, the serial, parallel, and constraint-based models of sentence parsing provide distinct explanations for interpreting the ambiguous sentence “When Fred eats food gets thrown.” The serial model accounts for initial misparsing through garden-path effects, supported by eye-tracking evidence (Frazier and Rayner, 1982). The parallel model offers flexibility in maintaining multiple structures, while constraint-based approaches highlight the role of contextual and semantic cues (MacDonald et al., 1994; Trueswell and Tanenhaus, 1994). Each framework contributes valuable insights, yet none fully captures the complexity of human language comprehension. The implications of this discussion extend beyond theoretical debate, influencing fields such as language acquisition and cognitive neuroscience by informing how linguistic input is processed in real time. Generally, future research should aim to integrate these models, testing their predictions against diverse linguistic inputs to better understand the dynamic nature of sentence parsing.
References
- Frazier, L. and Rayner, K. (1982) Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14(2), pp. 178-210.
- MacDonald, M. C., Pearlmutter, N. J. and Seidenberg, M. S. (1994) The lexical nature of syntactic ambiguity resolution. Psychological Review, 101(4), pp. 676-703.
- Rayner, K. and Frazier, L. (1987) Parsing temporarily ambiguous complements. Quarterly Journal of Experimental Psychology, 39(4), pp. 657-673.
- Trueswell, J. C. and Tanenhaus, M. K. (1994) Toward a lexicalist framework for constraint-based syntactic ambiguity resolution. In: Clifton, C., Frazier, L. and Rayner, K. (eds.) Perspectives on Sentence Processing. Hillsdale, NJ: Lawrence Erlbaum Associates, pp. 155-179.
- Trueswell, J. C., Tanenhaus, M. K. and Kello, C. (1993) Verb-specific constraints in sentence processing: Separating effects of lexical preference from garden-paths. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19(3), pp. 528-553.
