An essential step in the development of methodology for the use of eye movements to yield on-line information about cognitive processing is the establishment of reliable means of differentiating scans elicited under different cognitive processing conditions. Numerous research reports include findings that the observer's task influences the way that he or she scans a visual display. However, the means of quantifying differences between scans evoked under different experimental conditions vary widely across a vast range of studies within and between disciplines.

Authors (e.g., Yarbus) of some of the seminal studies employing eye movement measures to study cognitive processes support their conclusions that there are differences between scans by relying on their (and their readers') subjective impressions of overall differences when comparing scan patterns superimposed upon pictures viewed. In some cases, one needs only to look at the mapping of fixation points onto a visual display to see clear differences in where fixations are concentrated for one versus another condition. Very often, though, such differences are not so obvious. Further, even when one may subjectively judge that two scan patterns look different, it is often difficult to characterise quantitatively how the scans differ. The same dependent measures that may differentiate one scan from another (e.g., the proportion of total looking time spent focusing on various features of a picture) may not apply to all other pairs of contrastive scans.

A framework for the selection of appropriate dependent measures and analysis procedures is offered here in an effort to foster improved validity of eye movement analysis and interpretation and to achieve greater consistency in the reporting of data in the research literature concerning eye movements and cognitive processes.

An outline of the paper follows.

1. Issues in selecting and reporting eye movement measures
2. Consideration of potential dependent measures for scan differentiation (advantages and disadvantages of specific measures highlighted)
   1. Measures pertaining to fixation on "target(s)"
      1. Most fixations on target(s)? (binomial)
      2. Greatest duration on target(s)? (binomial)
      3. Total number of fixations on target(s)
      4. Total duration on target(s)
      5. Proportion of total number of fixations in the scan attributed to target(s)
      6. Proportion of total duration in the scan attributed to the target(s)
   2. Measures pertaining to time to arrive at target(s)
      1. Number of fixations to arrive at target(s)
      2. Duration to arrive at target(s)
      3. Proportion of total fixations in scan to arrive at target(s)
      4. Proportion of total scan duration to arrive at target(s)
   3. Measures of constraint or sequence
      1. Number of items fixated in the scan
      2. Number of fixations per scan
      3. Total duration per scan
      4. First-order asymmetric lambda
      5. Other measures of sequential predictability
3. Statistical relationships among dependent measures: Correlations among durational measures, numbers of fixations, and proportional measures
4. Conclusions
   1. Improving rationale for and validity of eye movement analysis and interpretation
   2. Improving consistency in reporting of data concerning eye movements and cognitive processes