How Do Readers Know Where to Look Next? Local Information Distributions Influence Eye Fixations

Abstract

Skilled readers' eye movements were recorded as they inspected sentences in preparation for comprehension questions. The sentences were written around target words that had uneven distributions of information, in that the words were predictable, given either the first few letters or the last few letters. Some parts of these words can be described as being more important than others for successful word recognition, and the experiments asked whether inspection patterns reflect the uneven distributions. Four types of ten-letter word were used: words with highly redundant endings (e.g. engagement), with moderately redundant endings (e.g. repatriate), with moderately redundant beginnings (e.g. superstore), and words with informative beginnings and endings (e.g. amalgamate). Redundancy was defined operationally in terms of the number of possible completions of a word given the first few or last few letters. A highly redundant ending is therefore one that occurs frequently and cannot be used to identify the word. Words gaining just one fixation received this fixation nearer to the word's centre than in the case of the first fixation upon words gaining two fixations. The single fixation also had a longer duration, giving support to the notion of a convenient viewing location, which, when achieved, can lead to a net saving in inspection time. In the case of words with redundant endings gaining just one fixation, the fixation was nearer to the beginning of the word than in the case of words with informative endings, and this influence of word type upon the location of the first fixation was interpreted in favour of the parafoveal processing hypothesis of eye guidance during reading.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 122 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.