Parallel Functioning Hypothesis to Explain Pronoun Resolution and Processing Load: Evidence from Eye-tracking

Abstract

Although the pronoun resolution mechanism has not been clearly defined, research in language acquisition, computation, and production has proposed various explanations for it in various languages. Among these models, the Parallel Function Hypothesis (PF) suggested that in a complex sentence, if the co-referential noun phrases are in the same grammatical functional category, the Parallel Function Pronoun Resolution (PFR), the statement is easier to process than the Non-parallel Function Pronoun (non-PFR) Resolution. In the current paper, we studied the processing times of these pronouns through eye-tracking methodology, and compared the first fixation, total fixation and difficulty level of PFR and non-PFR pronouns, as explained in the PF for pronoun referencing mechanism. In the Eye-mind Assumption (EMA), Just and Carpenter suggested that a person’s attention could be followed through the visual display in eye movement recordings, in which the duration of fixation on a word was interpreted as the processing of information. In other words, the difficulty in processing could be manifested on eye tracking data, especially on fixations. It was also clearly put forward that the eye tracking data revealed information on information processing regarding syntactic processing difficulty. The results of the current study suggested that pronouns which were resolved to a noun phrase which had a similar grammatical function were fixated less than pronouns which were resolved to a noun which had a different grammatical function. Unlike first fixation durations, total fixation durations of PFR pronouns were significantly shorter than those of non-PFR pronouns. Therefore, in terms of syntactic processing load, it could be stated that our findings clearly illustrated that non-parallel pronoun resolution involves more processing load than parallel resolution.

Acknowledgements

Successful completion of this study was only made possible by the support and cooperation of Dr. Ursula Stickler and Graham Healing of the Open University.

Notes

¹ For details of the explanations, please refer to Heron (1957), McKeever and Huling (1971), Diamond (1977), Pollatsek et al. (1981), Orbach (1967) and Battista and Kalloniatis (2002).

² In the original study, Wolf, Gibson, and Desmet (2004) manipulated sentence sets with coherence relations, namely similarity and cause effect. For the purposes of the current study, the coherence relations are not included in the analysis.