Cognition and Eye Movements

edited October 30, 2002

... a remarkable thing... when he was reading, his eye glided over the pages and his heart sensed out the sense, but his voice and tongue were at rest. -- St. Augustine (5th century A.D., describing St. Ambrose)

To be short, they be wholly given to follow the motions of the minde, they doe change themselves in a moment, they doe alter and conforme themselves unto it in such maner, as that Blemor the Arabian, and Syreneus the Phisition of Cypres, thought it no absurditie to affirme that the soule dwelt in the eyes... --A. Du Laurens (1599) A Discourse of the Preservation of Sight: of Melancholike Diseases; of Rheumes, and of Old Age

The perceptions of naked outlines, as in the diagrams of Euclid and the alphabetical characters, are to say the least of it, three parts muscular and one part optical. -- - A. Bain (1855) The Senses and the Intellect


  1. Reading
    1. Saccades made during reading
      1. Characters are most useful measure of saccade amplitude
      2. Pattern of eye movements during reading
    2. Relation between eye movements and cognition is not straightforward. To conclude that observer is getting the information at the location and for the duration of each fixation involves a number of assumptions that are not supported by the data.
    3. Observers usually fixate objects they want to identify...
      1. e.g., when comparing two line drawings, each object is fixated
    4. ... because limited information can be extracted peripherally...
      1. Factors
        1. Acuity
        2. Attention
        3. Visual masking
    5. ... but some information is available from peripheral retina.
      1. e.g., predictable words are fixated less frequently than expected
      2. peripheral preprocessing
        1. amount of preprocessing inversely proportional to discriminability of stimuli
        2. Parafoveal preview benefit
          1. During fixation a string of letters is presented parafoveally. After making a saccade to the word, it is named. Response time and errors are measures of performance. The experimental manipulation involves changing the characters while the saccade is in flight.
            1. Benefit when only certain characters match, e.g., benefit for naming TRASH when initially seeing TRAIN
            2. Benefit even is letter case is changed, e.g., benefit for naming TRASH when initially seeing trash
            3. Beginnings of words matter more than ends of words, e.g., no benefit for naming TRASH when initially seeing CRASH.
            4. Note that Ss rarely aware of the change in the letters!
    6. Moving window technique
      1. Different kinds of manipulations
      2. Perceptual span
        1. Determine effects on performance of limiting view of different extent. Original notion that extent of saccade during reading is just long enough to make some perceptual span overlap through successive fixations
        2. Asymmetric - 4 characters to left + 14 characters to the right for English readers
          1. Opposite asymmetry for Hebrew
          2. Different for Chinese, Japanese
        3. Different functional perceptual spans
          1. information about inter-word spaces up to 15 characters to the right
          2. word shape information up to 10 characters to right
          3. letter information up to 6 characters to right
    7. Optimum viewing position phenomenon
    8. Optimum viewing with free movements
      1. variation with kind of word
    9. One theory of reading is that two processes are happening concurrently. One process guides gaze according to word size, spacing, etc. - lower level visual attributes. This one works more rapidly. The 2nd process involves linguistic analysis. It is slower and guides eye movements more precisely - shifting gaze to optimum viewing position, prolonging fixation duration.
  2. Picture viewing
    1. Scan path - idiosyncratic pattern of fixations
    2. Analyses of free scanning eye movements
      1. Spatial distribution of fixations remains stable
      2. Fixation duration depends on degree of contextual liklihood and serial order
        1. context affects fixation duration. the more unlikely the object, the longer the fixation
    3. Viewing cartoons with captions