Three tasks are used to evaluate how well the visualizations convey the shape of intersecting surfaces. This section will describe the most general aspects of each task (i.e. the wording of the question, the nature of the possible responses). Descriptions of the visualizations or the trial data are discussed in later sections.
The first task (distance) asks the participant to estimate and compare distances between two sur- faces. The second task (local shape) asks the participant to estimate and compare local shape between two surfaces. The third task (global shape) asks the participant to recognize individual shapes in a pair of intersecting shapes.
These user study tasks were chosen in consultation with scientists, as described in Chapter 2. The tasks needed to satisfy the following requirements:
• use shape tasks commonly employed in the literature (adapted as necessary for two intersecting surfaces), and
• facilitate fast, simple judgments from the participants.
Because the scientists are interested in both the distance between intersecting surfaces and the shape of intersecting surfaces, the user studies include tasks evaluating performance for estimating and comparing both distances between surfaces and local surface shape. Typically, either a distance task or a local shape task alone would be considered sufficient to determine the effectiveness of a visualization technique at conveying surface shape1. I include both of these tasks to directly address scientists’ questions about layered-surface data. As will be shown, these two tasks (as implemented here) produced different rankings of the evaluated visualization techniques.
The scientists’ questions are predominantly about comparing the two surfaces, so I decided that the tasks should involve comparisons between two surfaces. To facilitate quick responses from the partic- ipants, the tasks are of a forced-choice design with two possible responses to each trial. Forced choice also removes the need to train participants to manipulate direct-measurement widgets to respond to tasks (i.e. the orientation probe developed by Koenderink and van Doorn [KVD95]). Because one of the visualization techniques included in the studies renders only one surface (color mapping), forced choice also avoids the problem of indicating where on the invisible surface the participant should es- timate local shape. With forced-choice tasks, participants could perform many trials in a relatively short amount of time (compared to manipulating widgets). However, these tasks yield only binary responses instead of metric errors.
5.1.1 Distance
The distancetask evaluates how well a visualization technique conveys surface shape by ask- ing participants to compare distance between intersecting surface at two points. Participants make judgements about the relative distances, reporting which distance appears shorter. The distance task
1Langer and B¨ulthoff list the most commonly used shape perception tasks and weigh their advantages and disadvantages
evaluates how effectively a visualization technique can be utilized to explore questions likeWhere are the two surfaces separated by more or less than some threshold?(see Section 2.3 for a discussion of scientists’ questions about distances between intersecting surfaces).
The distance task presents the participant with a forced choice between two regions in a display of intersecting surfaces. Participants are shown a display of two intersecting surfaces with two dis- tinct regions indicated by perceptually distinguishable markers. The participant is presented with the following question:
In which circled region do the two surfaces appear to be closer together?
The participant compares the relative distances and responds with the region containing the smaller distance.
5.1.2 Local shape
Thelocal shapetask evaluates how well a visualization technique conveys surface shape by asking participants to estimate local surface orientation. An estimate of local surface orientation is equivalent to an estimate of the local surface geometry (specifically, the surface normal at that point). Accurately estimating the local surface normal is an indication that the participant understands the shape of the surface in that region. Participants make judgements about relative differences in local surface orien- tation, reporting which differences appear smaller.
The local shape task evaluates how effectively a visualization technique can be utilized to explore questions likeAre the apparent differences between surfaces consistent with some missing features? (see Section 2.3 for a discussion of scientists’ questions about distances between intersecting sur- faces). Additionally, the local shape task evaluates how effectively a visualization technique enables the participant to estimate local shape on one of the pair of intersecting surfaces.
This task presents the participant with a forced choice between two regions in a display of in- tersecting surfaces. Participants are shown a display of two intersecting surfaces with two distinct regions indicated by perceptually distinguishable markers. The participant is presented with the fol-
lowing question:
In which circled region do the two surfaces appear to be more similarly oriented or parallel? The participant compares the local surface orientations for both surfaces in each region, then compares the orientation differences between regions, then responds with the region containing the smaller difference.
5.1.3 Global shape
Theglobal shapetask evaluates how well a visualization techniques conveys surface shape by ask- ing participants to recognize objects. Participants determine if a pair of intersecting objects includes a target object, reporting the target object’s presence or absence from the pair. The global shape task evaluates how effectively a visualization technique enables the participant to understand each of the intersecting surfaces as an individual object.
This task presents the participant with a forced choice between the presence or absence of a target object in a display of intersecting objects. Participants are shown a display of two intersecting objects that when displayed alone would be easily recognizable (e.g. fruit, vegetables, domestic animals). The participant is presented with the following question:
Is thetargetpresent in this pair of objects?
The participant reconstructs the shape of each object as if displayed individually, determines if either displayed object is the target object, then responds if the target object is absent or present in the pair.