United States Department of Transportation
i
Prediction of destination entry and retrieval times using keystroke-level models
-
1998-04-01
[PDF-981.09 KB]
Details:
-
Alternative Title:Prediction of destination entry and retrieval times using keystroke-level models : FAST-TRAC Phase IIB deliverable
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Corporate Publisher:
-
Abstract:Thirty-six drivers entered and retrieved destinations using an Ali-Scout navigation computer. Retrieval involved keying in part of the destination name, scrolling through a list of names, or a combination of those methods. Entry required keying in the destination’s name, the longitude, and the latitude. For young subjects, mean interkeystroke intervals were 1.7 s for initial cursor keystrokes, 1.5 s for initial enter, letters, and shift keystrokes, 1.2 s for numbers, and 0.6 s for space (different from the “standard” value in keystroke-level models of 1.2 s for a “worst” typist). Second keystrokes were about 1 s for letters, 0.7 s for cursor actions, and 0.5 s for numbers, similar to the standard time for complex codes (0.75 s/keystroke). For more than 2 cursor keystrokes, times were about 0.5 s. Age differences were large, with middle-aged drivers taking 40 percent longer and older drivers 120 longer than young drivers. Mental (thinking) times averaged 2.2 s, much greater the standard time (1.35 s). Equations were developed linking keystroke-level predictions to the actual times. Linear equations based on pure keystroke-level models accounted from 41 (all subjects) to 78 percent (young subjects only) of the variance of retrieval times and 12 to 39 percent of entry times. For tailored keystroke-level models (with experimentally based values for keystroke times, and adjustments for age and lighting) the variance accounted for was 58 and 83 percent (all subjects and young subjects, respectively) for retrieval, and 47 and 49 percent for entry. Use of linear equations and tailored models significantly reduced the size of prediction errors, making the predictions useful for engineering evaluations of alternative interfaces.
-
Format:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
