Response Latency

Response latency is the speed or ease with which a response to a survey question is given after a respondent is presented with the question. It is used as an indicator of attitude accessibility, which is the strength of the link between an attitude object and a respondent's evaluation of that object. While response latency has been used for some time in cognitive psychology lab experiments, its use in surveys came about more recently. In telephone surveys, response latency is measured in milliseconds as the elapsed time from when an interviewer finishes reading a question until a respondent begins to answer.

There are four stages that survey respondents use when answering questions: (1) question comprehension, (2) retrieval of information from memory, (3) integration of the information to form a judgment, and (4) selection of an appropriate response option. Response latency measures the time it takes to retrieve, form, and report an answer to a survey question. Response latency data can provide much useful information about attitude accessibility that can be incorporated into data analysis. For example, when attitudes are modeled to predict subsequent behavior, respondents with more accessible attitudes (indicated by shorter response times) often exhibit a stronger relationship between the attitude and the subsequent behavior. Attitude accessibility as measured by response latency is just one way of measuring the strength of an attitude, but it can be consequential for attitude stability, a respondent's resistance to persuasion, as well as the influence of the attitude on behavior. Response latency has also been used as a method of pretesting survey questionnaires in order to identify problematic questions.

Response latency was first used in cognitive psychology lab experiments where the timer measuring the response latency is a function of the participants' own reaction time to a self-administered instrument. When adapted for use in telephone surveys, it is generally measured via a voice-activated or “automatic” timer (which requires special equipment) that senses when a response is given or through an interviewer-activated or “active” timer embedded into the programming of computer-assisted telephone interviewing (CATI) software. The active timer requires the interviewer to start and stop the timer at the appropriate time using the computer keyboard and then verify that the time measurement is valid. Response latencies are coded as invalid if the interviewer fails to apply the timer correctly or if the respondent asks for the question to be repeated. Response latency can also be measured using a “latent” or unobtrusive timer that is programmed into CATI software and is invisible to both interviewers and respondents. Latent timers simply measure the total duration of each question from the time the question appears on the interviewer's screen until the moment the respondent's answer is recorded. Such timers also can be used in computer-assisted personal interviewing (CAPI) and computerized self-administered questionnaires (CSAQ).

Regardless of how the response latency data are collected, the distribution of responses is frequently skewed, and the data require careful examination and cleaning before analysis. Invalid data and extreme outliers should be removed and the data transformed to eliminate the skew. Depending on how the data are collected, researchers using response latency data may also want to control for baseline differences among respondents in answering questions and interviewers in recording survey responses because some respondents are naturally faster in answering questions and some interviewers are naturally faster in recording responses.

...