A distinction between estimator and system variables is made in the eyewitness research literature between two categories or types of variables that influence the accuracy of eyewitness accounts. System variables are those that are (or can be) under the control of the justice system, whereas estimator variables cannot be controlled by the justice system. Examples of system variables include factors such as the instructions given to eyewitnesses prior to their viewing a lineup or the number of people who are used in a lineup. Examples of estimator variables include factors such as how good a view the eyewitness had of the perpetrator during the crime or whether the witness and perpetrator were of the same or different race. The estimator versus system variable distinction tends to be tied to a temporal unfolding of events, in the sense that events that occur before or during the witnessing experience are necessarily relegated to estimator variable status whereas system variables begin to come into play later, once the investigation is under way. There is no presumption in the estimator variable versus system variable distinction that one category of variables has more impact on eyewitness accuracy than the other. Nevertheless, this distinction, first articulated in 1978 by Gary L. Wells, has tended to result in a higher premium being placed on system variables because these can be used to help minimize eyewitness errors in actual cases, whereas estimator variables can only be used to postdict how the variables might have influenced the eyewitness.
The study of system variables has generally been tied to policy-related recommendations on ways to improve how crime investigators interview eyewitnesses and on ways to improve how lineups are constructed and conducted. The study of estimator variables, in contrast, has more often been tied to the development of expert testimony that can assist triers of fact (e.g., judges, juries) in deciding whether to accept the testimony of an eyewitness as having been accurate or mistaken. In fact, however, system variables are as relevant to expert testimony as are estimator variables, and in recent years, it has become more apparent that estimator variables and system variables are not independent. In general, the impact of system variables is likely to depend somewhat on the levels of the estimator variables. An obvious example of this dependence is when the estimator variables are highly favorable to the existence of an extremely deep, solid memory. If memory is strong enough, system variables would not likely have much impact. For instance, system variable research shows that it is critical for eyewitnesses to be warned prior to viewing a lineup that the actual perpetrator might not be present, because the absence of such a warning leads eyewitnesses to select someone from a lineup even if the actual perpetrator is not present. However, if the eyewitness’s memory is strong enough (e.g., attempting to pick one’s own mother from a lineup), the presence or absence of this warning is of little consequence. Hence, a complete understanding of eyewitness performance clearly requires research on both system and estimator variables.
Generally, system variables can also serve the function of being estimator variables, but estimator variables cannot be system variables. In some cases, however, variables that traditionally have been considered estimator variables have taken on system-variable-like properties. The confidence of an eyewitness, for instance, has traditionally been considered an estimator variable because it was presumed to be beyond the control of the justice system, and the emphasis of the estimator variable research on eyewitness confidence was to find out how well or poorly it postdicted the accuracy of the eyewitness. Now, however, there is a great deal of research showing that procedures that are under the control of the justice system affect the confidence of the eyewitness and the magnitude of the confidence-accuracy relation. In this sense, eyewitness confidence, traditionally an estimator variable, has taken on some of the properties of a system variable.
In eyewitness identification research, system and estimator variables have been further subdivided in recent years into two types—namely, suspect-bias variables and general-impairment variables. Suspect-bias variables are those that influence the eyewitness specifically toward identifying the suspect from a lineup, whereas general-impairment variables simply reduce the overall performance of the eyewitness. An example of a general-impairment system variable is when the lineup administrator fails to instruct the eyewitness that the perpetrator might not be in the lineup. In this case, the instruction failure impairs the eyewitness’s performance (by making the eyewitness insensitive to the possibility that the correct answer might be “not there”) but does not specifically bias the eyewitness toward the suspect any more than it biases the eyewitness toward the nonsuspects in the lineup. An example of a suspect-bias system variable is when a lineup is structured in such a way that the suspect stands out as the obvious choice (e.g., as the only one who fits the description of the culprit). An example of a general-impairment estimator variable is poor viewing conditions at the time of witnessing. Poor viewing conditions might impair the eyewitness’s performance on the lineup, but poor viewing conditions do not specifically bias the eyewitness toward the suspect any more than they bias the eyewitness toward the nonsuspects in the lineup. An example of a suspect-bias estimator variable is when the eyewitness has source confusion, such as when an innocent suspect is picked out of a lineup because he was familiar; but he was familiar because he had been a customer, not because he was the person who robbed the clerk.
- Wells, G. L., Memon, A., & Penrod, S. (2006). Eyewitness evidence: Improving its probative value. Psychological Science in the Public Interest, 7, 45-75.
- Wells, G. L., & Olson, E. (2003). Eyewitness identification. Annual Review of Psychology, 54, 277-295.