**1.3 Stimuli for Experiments 1-3**

The initial studies used complex 3D shapes, shown in Figure 1, that were composed of three abstract prototypical shapes and systematic distortions. Objects were originally modeled in the Maya 3D modeling software produced by Autodesk. Initially, 30-40 3-dimensional virtual forms were generated using a shape growth tool within the Maya suite, and 20 were chosen for multidimensional scaling. Three forms were then selected from the multidimensional space (MDS) that were moderately separated from each other and which appeared to be equi-distant from each other in three dimensions. These 3 forms become the prototypical forms for three categories. The surface of each prototype was then subdivided into a very small polygon mesh which gives objects a more organic appearance.

The Maya's shape blend tool was used to generate forms that were incremental blends between all pairs of the 3 prototype forms. This resulted in a final category population of 24 3-dimensional objects, where each prototype was transformed, along two paths into the other two prototypes. The distortion setting used in the shape blend tool was set to .14, which allowed for 7 forms to be generated between each prototype pair. The forms were then converted from Maya's file format which could then be steriolithographically printed using a ZCorperation Zprinter. Each of the objects was smooth to the touch and of the same approximate weight and overall size.

#### **1.4 Theoretical issues**

This structure was selected to address a number of additional issues. First, each prototype occupied the endpoints of two transformational paths and was the only form capable of readily generating its distortions. However, unlike the vast majority of studies in categorization, each prototype was not otherwise central to its learning (or transfer) patterns but was positioned at the endpoints of two transformational paths. We were interested in whether these prototypical objects would, nonetheless, exhibit characteristics typically found in recognition and classification. For example, the prototype is often falsely

Haptic Concepts 7

ambiguity of category membership also plays a role, as well as similarity to old training objects, then (false) recognition should be reduced, compared to the new objects. This was because the midpoint objects could not be unambiguously classified into a single prototype,

Experiment 1 examined visual (V) or haptic (H) category learning followed by a transfer test in the same or alternate modality (VV, VH, HV, HH). Half of the subjects received random or systematic training. Particular contrasts were of special interest: (a) Transfer differences between the VV and HH conditions should reveal whether visual concepts are learned better than concepts learned haptically; (b) VV vs. VH and HH vs. HV should indicate how much information is lost when tested in an alternate modality; and (c) VH vs. HV would indicate whether information is transferred more readily from one modality

Objects were placed on a small table next to the participant. An opaque dark blue curtain was hung between the stimuli and participant and could be slid back and forth along a rod situated 10 feet above, allowing the participant to view or handle the object. This allowed the experimenter to select a designated stimulus to present to the subject, while hiding the remaining 23 stimuli. The stimuli were shown one at a time. Four types of objects can be identified: (a) 12 old objects, 4 from each category prototype, that were presented during learning; (b) 6 new patterns, 2 from each category; (c) 3 prototypes; and (d) 3 midpoint objects. The latter objects were midway between either of two prototypes and, therefore, could not be unambiguously assigned to a single prototype category. A schematic representation of the 24 objects, separated by the three categories and transformational

The learning phase was composed of four study-test trial blocks. On each study block, the 12 learning objects were shown randomly or systematically blocked by category, labeled as A, B, or C for the subject. Following this, the objects were presented in a random order and required verbal classification of the object (A, B, or C). Following their judgment, corrective verbal feedback was provided. For subjects in the systematic condition, the three categories were presented in a counterbalanced order, although patterns belonging to a given category

On the transfer test, all 24 objects were presented in a random order, which included the four training patterns in each category (old), the three category prototypes, and nine new objects. As indicated in Figure 2, three of the new patterns were located midway between the two prototypes and were, as a consequence, analyzed separately from the remaining new objects. On the transfer test, the subject was required to make a double judgment to each object. The first judgment was a recognition judgment – is this object old or new? The

second judgment was a classification judgment (is it an A, B, or C pattern?).

since either of two prototype classes would be correct.

**2. Cross-modal category transfer** 

to the other.

**2.1 Method** 

paths is shown in Figure 2.

were shown in a random order.

**2.2 Procedure** 

recognized as an old pattern and classified better than other new exemplars (Metcalfe & Fisher, 1986; Nosofsky, 1991; Shin & Nosofsky, 1992). However, exceptions in recognition to this outcome have been obtained (Homa, Goldhardt, Burruel-Homa, & Smith, 1993; Homa, Smith, Macak, Johovich, & Osorio, 2001), apparently when the prototype is a unique pattern rather than composed of features identically contained in its exemplars.

Fig. 1. The categorical space composed of 24 shapes; each category prototype is located at the vertex

In the present experiments, all objects including the prototypes were unique patterns, composed of novel and not identically repeated features or components. Second, two types of new patterns were used in transfer, those that were positioned between old training forms and those that were located at the midpoint of the transformational paths generated from different prototypes. In effect, each midpoint stimulus was a form that was positioned within a 'gap' that was positioned in the middle between two prototypes. We were interested in whether an object that fills a gap and flanked by two training patterns from different prototypes would be less likely to be falsely recognized as old than other new patterns that were similarly flanked by two training patterns but which was closer to the category prototype. If similarity to close training neighbors in learning dictates (false) recognition, regardless of the category membership of the neighbors, then recognition of the midpoint objects should be similar to recognition of the new objects. Alternatively, if

recognized as an old pattern and classified better than other new exemplars (Metcalfe & Fisher, 1986; Nosofsky, 1991; Shin & Nosofsky, 1992). However, exceptions in recognition to this outcome have been obtained (Homa, Goldhardt, Burruel-Homa, & Smith, 1993; Homa, Smith, Macak, Johovich, & Osorio, 2001), apparently when the prototype is a unique pattern

**P3**

**<sup>14</sup> <sup>19</sup>**

**20 21**

**12**

**13**

**11 10**

A Greco-Latin square was used to assign stimuli to the factors of category label, prototype representing each category (P1,P2,P3) and name assigned to the category (A,B,C).

In the present experiments, all objects including the prototypes were unique patterns, composed of novel and not identically repeated features or components. Second, two types of new patterns were used in transfer, those that were positioned between old training forms and those that were located at the midpoint of the transformational paths generated from different prototypes. In effect, each midpoint stimulus was a form that was positioned within a 'gap' that was positioned in the middle between two prototypes. We were interested in whether an object that fills a gap and flanked by two training patterns from different prototypes would be less likely to be falsely recognized as old than other new patterns that were similarly flanked by two training patterns but which was closer to the category prototype. If similarity to close training neighbors in learning dictates (false) recognition, regardless of the category membership of the neighbors, then recognition of the midpoint objects should be similar to recognition of the new objects. Alternatively, if

Fig. 1. The categorical space composed of 24 shapes; each category prototype is located at

**<sup>4</sup> <sup>2</sup> <sup>3</sup> <sup>6</sup> <sup>5</sup>**

rather than composed of features identically contained in its exemplars.

**18**

**17**

**<sup>16</sup> <sup>15</sup>**

**P1**

**1**

the vertex

ambiguity of category membership also plays a role, as well as similarity to old training objects, then (false) recognition should be reduced, compared to the new objects. This was because the midpoint objects could not be unambiguously classified into a single prototype, since either of two prototype classes would be correct.
