**5.3 Discussion**

The results show that the haptic and visual representations of the same 3D objects were remarkably similar, suggesting that information critical to visual concepts were generally maintained following haptic inspection. As was the case in our previous studies that explored multidimensional scaling following the learning of categorical structure, the degree of structure was generally enhanced following learning (Homa et al., 1979; Zaki & Homa, 1999). As predicted, the conceptual spaces were more tightly structured following haptic examination. What seems likely is that there exists a dominant set of features critical to similarity that are comparable to the visual and haptic modality but that additional information, perhaps idiosyncratic, is more available in the visual modality. This would explain why the spaces were highly correlated and yet why the haptic space was more tightly structured.

Fig. 9. Mean Structural Ratio for the six MDS solutions

correlations were computed among the six conditions, using as input the individual structural ratios for each object. These 15 correlations were positive and high, ranging from r = + .817 to r = + .981; the average correlation was r = +.924. A sample space – in this case, the MDS space following systematic learning in the haptic modality - is shown in Figure 10. What is clear is that the three haptic categories are clearly defined. Comparison with the original space (Figure 1) clearly reveals that the category prototype (P1, P2, P3) has become centered within each category rather than occupying the location at the extreme points of

The results show that the haptic and visual representations of the same 3D objects were remarkably similar, suggesting that information critical to visual concepts were generally maintained following haptic inspection. As was the case in our previous studies that explored multidimensional scaling following the learning of categorical structure, the degree of structure was generally enhanced following learning (Homa et al., 1979; Zaki & Homa, 1999). As predicted, the conceptual spaces were more tightly structured following haptic examination. What seems likely is that there exists a dominant set of features critical to similarity that are comparable to the visual and haptic modality but that additional information, perhaps idiosyncratic, is more available in the visual modality. This would explain why the spaces were highly correlated and yet why the haptic space was more

**Categorical Structural Ratio**

the two transformational paths.

**5.3 Discussion** 

tightly structured.

Fig. 9. Mean Structural Ratio for the six MDS solutions

**No Lrn**

**Random**

**System**

**No Lrn**

**Visual Haptic**

**Random**

**System**

**0**

**0.1**

**0.2**

**0.3**

**Mean Structural Ratio**

**0.4**

**0.5**

Fig. 10. Three dimensional MDS space following systematic learning in the haptic modality
