Page 363 -

12.5 Hybrid data collection methods 353

Comparison of the data from children and adults generated several
important insights. Children were slower and less accurate than adults. Children
also tended to hover over targets, reentering much more frequently than adults.
Hourcade and colleagues found that while Fitts' law does apply to children, the
model is more accurate for adults.
Mouse motion paths (or “trails”) provide striking illustrations of the
differences between children and adults. While adults move accurately and
quickly between targets, 5-year-old children went far afield, often overshooting
targets. Four-year-old children were even worse, moving all over the screen
(Figure 12.9). Children were also much more likely to repeatedly enter, leave,
and reenter targets (Figure 12.10).

Based on these differences in performance profiles, Hourcade et al. suggested
several possible approaches to designing interfaces for young children. Possible so-
lutions include larger icons, smaller mice, expanding targets, slower or accelerated
mouse movement, and constrained motion between selections (using directional ar-
rows and a selection button) (Hourcade et al., 2004).

12.5 HYBRID DATA COLLECTION METHODS
If one source of HCI data is good, then two must be better. Multiple channels of
data collection can be combined to overcome the shortcomings of any one approach.
Logging user interactions with a word processor may be a good start toward under-
standing how various controls are used, but log files provide little, if any, contextual
information that might prove invaluable for interpretation and analysis. Video re-
cordings or direct observation of users at work can help researchers understand users'
goals, frustrations, state of mind, and satisfaction (or lack thereof) with the system
being used. Taken together, these data sources provide a much more detailed and
complete picture of user activity than either would on its own.
Combining outputs from log files and video sessions may require specialized soft-
ware support. Ideally, we might want to identify an event in a log file and jump right
to the video recording that displays what the user was doing at that moment. HCI
researchers have developed tools that provide this synchronized access (Hammontree
et al., 1992; Crabtree et al., 2006; Fouse et al., 2011). These features may also be
available in professional tools for usability studies, such as the interaction recording
tools discussed in Section 12.3.3.
Similar approaches can be used for evaluation of web-based systems. Web log
files and web proxies are limited in their ability to capture details of user interac-
tions—and possibly in their ability to identify distinct users. However, they are rela-
tively easy to configure and deploy. An instrumented web browser can provide vastly
greater detail regarding specific user events, but it requires installation of unfamiliar
software on end-user computers, potentially limiting the number of participants. One