39 |
known as *blending* - creating surfaces that |
known as *blending* - creating surfaces that |
40 |
meet several existing surfaces smoothly. |
meet several existing surfaces smoothly. |
41 |
|
|
42 |
|
Displaying relationships between nodes by lines or arcs is efficient as |
43 |
|
physical connection makes the nodes to be perceived as parts of a single |
44 |
|
object, according to the law of connectivity (one of the laws of |
45 |
|
perceptual organization). Therefore, the viewer needs no conscious effort |
46 |
|
to perceive the semantic relationship between the nodes. Fillets aim to |
47 |
|
enhance the perception of connectivity by implementing the connecting |
48 |
|
lines with continuity. The node and the connecting line form one object, |
49 |
|
without an abrupt disruption in perception, thus making it easy to follow |
50 |
|
the connection between nodes. |
51 |
|
|
52 |
Our use of fillets is entertainingly analogous to the use in mechanical |
Our use of fillets is entertainingly analogous to the use in mechanical |
53 |
engineering: fillets ensure that the human perception system |
engineering: fillets ensure that the human perception system |
54 |
doesn't break an object and a connection starting |
doesn't break an object and a connection starting |
60 |
grouping of visual elements, in this case, |
grouping of visual elements, in this case, |
61 |
grouping of the node and the connection. |
grouping of the node and the connection. |
62 |
|
|
63 |
|
The only disruptions in a fillet graph will be when the connections cross |
64 |
|
each other. Line crossing is one serious factor making it hard to read |
65 |
|
complex node-link graphs. With fillets, tracing a connection is |
66 |
|
perceptually easy even if it crosses with other connections. |
67 |
|
|
68 |
|
We tested the easiness of perceiving fillets in a controlled laboratory |
69 |
|
experiment with ten naïve participants. Eight different graphs were |
70 |
|
tested, of which one was implemented with "perfect" fillets. The other |
71 |
|
seven graphs were "incomplete" fillets or different common node-link |
72 |
|
graphs lacking visual continuity. All types of the graphs were implemented |
73 |
|
in three different sizes. The task of the participants was to recognize as |
74 |
|
fast as possible a connection going behind a node. There was only one |
75 |
|
target connection in a graph. A participant performed this task 24 times |
76 |
|
per graph type. |
77 |
|
|
78 |
|
We found that among these eight graph types, fillets did indeed enable the |
79 |
|
fastest perception of connection going behind a node. This indicates that |
80 |
|
fillets have such inherent properties that can help perceiving the |
81 |
|
structure of node-link graphs and thus understanding their contents, as |
82 |
|
these two are closely intertwined. We plan to carry more experiments to |
83 |
|
prove the efficiency of fillets also in more realistic use situations. |
84 |
|
Results from earlier research (Irani & Ware, 2003) indicate that more |
85 |
|
natural-looking nodes and links improve users' recall of the structure of |
86 |
|
the graph. We expect that fillets will prove their usefulness in tracing |
87 |
|
connections in graphs but also comperehending the structure and possibly |
88 |
|
the contents of the graphs. |
89 |
|
|
90 |
|
|
91 |
|
Irani, P. & Ware, C. (2003). Diagramming Information Structures Using 3D |
92 |
|
Perceptual Primitives. ACM Transactions on Computer-Human Interaction, |
93 |
|
Vol. 10, No. 1, March 2003, Pages 1-19. |
94 |
|
|
95 |
|
|
96 |
Figure: ambiguity, 1/2 page |
Figure: ambiguity, 1/2 page |