Prepositional Phrase Attachment Ambiguities in Declarative and Interrogative Contexts: Oral Reading Data
Date of Degree
Janet Dean Fodor
Phonetics and Phonology | Psycholinguistics and Neurolinguistics | Semantics and Pragmatics | Syntax
parsing, prosody, human sentence processing, reanalysis, ambiguity
Certain English sentences containing multiple prepositional phrases (e.g., She had planned to cram the paperwork in the drawer into her briefcase) have been reported to be prone to mis-parsing of a kind that is standardly called a “garden path.” The mis-parse stems from the temporary ambiguity of the first prepositional phrase (PP1: in the drawer), which tends to be interpreted initially as the goal argument of the verb cram. If the sentence ended there, that would be correct. But that analysis is overridden when the second prepositional phrase (PP2: into her briefcase) is encountered, since the into phrase can only be interpreted as the goal argument of the verb. Thus, PP2 necessarily supplants PP1’s initially assigned position as goal, and PP1 must be reanalyzed as a modifier of the object NP (the paperwork).
Interrogative versions of the same sentence structure (Had she planned to cram the paperwork in the drawer into her briefcase?) may have a different profile. They have been informally judged to be easier to process than their declarative counterparts, because they are less susceptible to the initial garden path analysis. The study presented here represents an attempt to find a behavioral correlate of this intuitive difference in processing difficulty.
The experiment employs the Double Reading Paradigm (Fodor, Macaulay, Ronkos, Callahan, and Peckenpaugh, 2019). Participants were asked to read aloud a visually presented sentence twice, first without taking any time at all to preview the sentence content (Reading 1), and then again after unlimited preview (Reading 2). The experimental items were created in a 2 x 2 design with one factor being Speech Act (declarative vs. interrogative) and the other being PP2 Status, i.e., PP2 could only be an argument of the verb iv (Arg), as above, or else PP2 could be interpreted as a modifier (Mod) of the NP within the preceding PP, as in She had / Had she planned to cram the paperwork in the drawer of her filing cabinet(?).
Participants’ recordings of Reading 1 and Reading 2 were subjected to prosodic coding by a linguist who was naive to the research question. Distributions of prosodic boundaries were statistically analyzed to extract any significant differences in prosodic boundary patterns as a function of Speech Act, Reading, or PP2 Status. Logistic mixed effect regression models indicated, as anticipated, a significant effect of PP2 Status across all analyses of prosodic phrasing, and a significant effect of Reading for both analyses of prosodic phrasing that included boundary strength. Speech Act was a significant predictor in one of prosodic phrasing, but the hypothesized interaction (between Speech Act and PP2 Status) was not significant in any model.
Another analysis concerned the amount of time a participant spent silently studying a sentence after Reading 1 to be confident they had understood it before reading it aloud again (Reading 2). The time between readings is referred to as the inter-reading time (IRT). It was assumed that a longer IRT signifies greater processing difficulty of the sentence. Thus, IRT was hypothesized to provide a behavioral correlate of the intuitive judgement that the interrogative garden paths are easier to process than the declarative ones. If a correlate had been found, it would have taken the form of an interaction between the two factors (Speech Act and PP2 Status) such that the IRT difference between Arg and Mod sentence versions was smaller for interrogatives than for declaratives. Ultimately, however, no statistically significant interaction between Speech Act and PP2 Status was found.
Further studies seeking behavioral evidence of the informal intuition motivating this research are proposed. Also offered are possible explanations for why the intuition is apparently so strong for some English speakers, and why, if so, it is not reflected in IRT. Significant ancillary findings are that interrogatives are in general more difficult to process than corresponding declaratives. Also, inter-reading time (IRT) in the Double Reading paradigm is confirmed as a useful measure of sentence processing difficulty given that within the declarative sentences, the garden-path (Arg) versions showed significantly longer IRTs than the non-garden-path (Mod) versions.
Peckenpaugh, Tyler J., "Prepositional Phrase Attachment Ambiguities in Declarative and Interrogative Contexts: Oral Reading Data" (2019). CUNY Academic Works.
Phonetics and Phonology Commons, Psycholinguistics and Neurolinguistics Commons, Semantics and Pragmatics Commons, Syntax Commons