The Direction of Interaction

It is important to note that the size of this functional bias is much smaller (<20%) than the influence obtained by Carlson-Radvansky et al. (1999) (72% for functionally related objects; 45% for functionally unrelated objects). One possible reason for this difference is the nature of the interaction between the objects. Recall that in the Carlson-Radvansky et al. (1999) study, the reference objects were selected with the constraint that the located object interacted with it from above or below. Thus, placing the located object above or below the reference object resulted in a placement that served to facilitate the interaction between the objects (e.g., toothpaste "above" the bristles of the toothbrush). In contrast, in the present set of stimuli, the located object typically interacted with the relevant part of the reference object from the side. For example, the mute and mouthpiece are applied to their respective parts on the trumpet by moving in from the left or right direction; the head must be tilted and move from the left to fit within the headphones; and the curling iron clasps the hair from the sides. This horizontal direction of interaction is typical for the set of items that we used. Thus, an association between the objects is not sufficient for obtaining a strong functional bias. Rather, the placement of the objects with respect to the spatial term must enable a fulfillment of the simulated interaction. We assessed this idea by conducting the placement task with the spatial term "near." We selected "near" because Prasada and Ferenz (2001) used a rating task to show that "near" is influenced by the functional interaction among the objects. Similarly, "near" is related semantically to "approach," and Morrow and Clark (1988) observed effects of object characteristics on judgements of distance associated with the term "approach." This makes it likely that the term "near" will be susceptible to the functional bias in our placement task. Second, and most importantly, the definition of "near" is not restricted to a particular part of the object. Indeed, Logan and Sadler (1996) had participants place an "X" near an "O," and found that the placements were scattered close to but all around the O, consistent with a geometric definition based on distance from the edges of the reference object, without further specification of a particular edge. Thus "near" offers an important contrast to the terms "above" and "below" that restrict placements to a particular side of the object. As discussed previously, this type of restriction may alter the degree of functional influence, particularly if the direction of interaction between the located objects and the reference object is not from "above" and "below."

The "near" condition was embedded in the "above" and "below" trials in the preceding experiments, with all participants receiving all terms, and each term paired with each type of located object across participants. The predictions for "near" were similar to those for "above" and "below,"

although a stronger functional bias was anticipated, given that the spatial term no longer restricted placements to a particular side that may or may not have coincided with the direction of interaction between the objects. More specifically, with "near" it was expected that participants would be able to place the located object at any side, and could therefore position their placements so as to facilitate the interaction between the objects, should they wish to do so.

Figure 3.7 shows the near placements around the three sample reference objects from Figure 3.5, broken down by type of located object. The differences between Figures 3.6 and 3.7 are readily apparent. First, placements of the neutral object in the center columns are generally clustered around the two functional parts rather than around the midpoint of the object. Second, placements are functionally biased toward the appropriate functional part, with occasional exceptions. Third, the gradation in the functional placements that was observed for above and below that seemed to reflect a combination of functional and geometric factors is much less evident for near. Rather, the functional placements tend to illustrate a complete bias. Note also that placements of the located objects were generally consistent with the direction of interaction between the objects, enabling fulfillment of their interaction.

We quantified these effects, following the procedures used for the "above" and "below" data. Specifically, for the located object functionally related to Fi (the leftmost part in Figure 3.7), the mean percentage deviation of the placements was -56%; for the located object related to F2 (the rightmost part), the mean percentage deviation was +36%; and the mean percent deviation for the beanbag was -30%, indicating a bias toward the leftmost functional part. These deviations were all greater than 0, with a significant difference between the deviation associated with Fi and the deviation associated with F2. In addition, all of these deviations were significantly greater than the size of the deviations observed in the "above" and "below" data. These data indicate a substantial functional influence due to the identity of the located object, with placements biased toward the appropriate functional part, and an influence due to the functional parts of the reference object, as indicated by the bias to place the neutral object near Fi. Thus, as in the "above" and "below" data, in addition to the contribution of the identity of the located object, there was also an effect due to the salience of a particular part of the reference object. Whether this part was highlighted by virtue of its functional importance, perceptual prominence or some other dimension is unclear.

As with the "above" and "below" data, we also classified the placements as geometric, with 18% of placements occurring in this category; functional, with 60% of the placements occurring in this category, and consistent with the opposite function, with 22% falling into this category. For the beanbag placements, 72% were in the functional category and 28% were geometric.

figure 3.7. Placements of the located objects near the three sample reference objects. See Figure Caption 3.6 for description of the layout.

These classifications for the functionally related objects are largely a reversal of the pattern of data observed for "above" and "below," and indicates that "near" placements are more susceptible to functional influences than geometric influences. We attribute this to the fact that "near" placements were not restricted to any particular edge of the reference object, and thus could be placed so as to facilitate the interaction between the objects, consistent with a simulation of the interaction between the objects.

0 0

Post a comment