hide
Free keywords:
-
Abstract:
The encoding and recognition of a large-scale environment was studied using virtual reality technology. A computer model based on the Cafe’ Liechtenstein, a house situated in
the centre of Tübingen, Germany, was developed using a 3D modeling program. Simulated walkthroughs were implemented using SGI Performer programming libraries. To facilitate natural learning of the environment Ss had to find and acknowledge 3D spatially located markers (consisting of two-digit codes) which appeared only when Ss were close
enough. The test consisted of images of all locations corresponding to the position of the markers acknowledged during training. Also included was an equal number of images of ‘surprise’ locations not corresponding to marker locations. This determined whether recognition
performance generalized to all views, not just those indicated by markers. As a test of recognition from novel perspectives, the Ss movements during training were
restricted. During testing, images from both familiar and unfamiliar directions were shown to Ss (all were images of actual and surprise marker locations). Ss were assigned
to one of two groups: The ‘active-explorers’ controlled their movement through the scene using a Spaceball 3D input device, while the passive-movers only observed these movements.
Twelve naive subjects were assigned randomly to each group. We used a 3-way mixed design to analyze the d’ sensitivities in an old/new recognition task. Mode of learning (active/passive) was a between groups factor while marker (present/absent) and view direction (familiar/novel) were within-subject factors. The results revealed no overall significant effect of active/passive observers. There was however a highly significant effect of viewing direction with familiar direction views being recognized much easier than novel views. There was no overall significant effect of marker presence/absence although
there was a marginal interaction between view direction and marker present/absent. In this case the familiar views of locations where markers were present were recognized
easier than those familiar direction views that were not indicated by markers. There was no similar difference in performance corresponding to the novel direction views. Not surprisingly, this indicates that the attentional factor (of having acknowledged the presence of a marker) makes a particular location more memorable. What is interesting however is that the novel views did not show such a difference. The d’ for novel direction views was
always above 1.0 and this indicates that the novel directions were recognized as a natural process of encoding the scene and not as an artifact of drawing Ss attention to marker locations.
In conclusion, the superior performance for familiar directions views after natural yet restricted (active or passive, simulated) movement through a scene shows that encoding of structural information is view-based and egocentric. However, novel direction views were also recognized and the properties of the mental encoding that facilitate this ability must be further investigated.
