The Role of Context in Defining the Functional Part

The Role of Context in Defining the Functional Part One potential limitation to the Carlson-Radvansky et al. (1999) work is that the functional bias was defined with respect to a functional part that was predetermined by the experimenters as functionally important, particularly in the context of a given located object (e.g., the tube of toothpaste interacts with the bristles, rendering the bristles the functional part). It is likely that these parts are not always considered the most functional parts of the objects. Indeed, although all reference objects showed a positive functional bias, the degree of bias varied widely across pairs of objects and across diVerent types of functional interactions (see Carlson and Covell, in press). Moreover, many objects have many functional parts, and these parts may become diVerentially important depending upon the goal for which one is using the object. Kenny and Carlson (2003) began an investigation into the role that context plays in defining the functional importance of the parts of an object. At the beginning of the experiment, participants watched a video that depicted objects involved in one of three types of interactions; this constituted the critical context manipulation. Two of the interactions emphasized diVerent functional parts of the reference object. For example, in a scene with a toaster serving as a reference object, one interaction 140 Laura A. Carlsoninvolved putting a piece of bread into the slots of toaster. The bread was the located object, and the functional part was the slots. In contrast, a diVerent interaction involved pressing down the lever activating the toaster. The hand was the located object, and the functional part was the lever. In the third interaction, the toaster was simply picked up and then put back down. There was no located object, and no functional part was emphasized. This was the neutral context condition and served as a baseline from which to assess the relative importance of the functional parts in the absence of context. A given participant saw only one type of interaction for a given object, but across participants, each type of interaction was viewed by at least seven participants for each of 18 reference objects. Following the video, all participants performed an identical placement task in which they were asked to place pictures of a beanbag (a functionally unrelated object) near pictures of the objects that they viewed in the video; these objects thus served as reference objects. A functionally unrelated located object (beanbag) was used so that placements would not be biased by virtue of the identity of the object and its possible interaction with the reference object. This is important because we were interested in assessing whether a previous interaction with a reference object would bias one’s later interpretation of space surrounding that object in a seemingly unrelated task. The prediction was that if the context provided by the video served to highlight one functional part of the object over another, and if this functional salience played a role in defining space around the reference object, then one should see a bias in participant’s placement of the located object toward that functional part. This would contrast with Logan and Sadler’s (1996) findings from a placement task in which participants were instructed to draw a dot ‘‘near’’ a square; placements were distributed around all sides of the reference object. Data for two reference objects (bottle opener and toaster) are shown in Fig. 4. The symbols correspond to the placements of the beanbags by diVerent participants and are coded as a function of the type of interaction that participants viewed prior to the placement task. Four features in the data can be noted. First, for participants who saw the neutral context, placements were at both functional parts and in other locations around the reference object. Second, for participants who viewed a functional context, there was a systematic bias to place the beanbag near the functional part that was emphasized by the context. For example, participants viewing the bread context for the toaster were more likely to put the beanbag near the slots than near the lever or anywhere else, whereas participants viewing the lever context for the toaster were more likely to put the beanbag near the lever than the slots or anywhere else. To quantify this eVect, for each object, we defined a midpoint between the two functional parts and measured Using Spatial Language 141placements in the two context conditions relative to this midpoint, coding them as positive if they were biased toward a contextually emphasized part and negative if they were biased away from the contextually emphasized part. Given that the distances between the parts varied across objects, the deviations were converted into proportions of the distance between the midpoint and the center of the functional part, with 0% indicating placements over the midpoint and 100% indicating placements over the center of the emphasized functional part. On average, there was a positive bias (M ¼ .31, SEM ¼ .11) that diVered significantly from 0 [t(17) ¼ 2.74, p < .014], indicating a preference for placing the located object with respect to the emphasized functional part. Third, this bias was not complete; not all Fig. 4. Placements of the beanbag around a toaster (A) and bottle opener (B) coded as a function of the observed interaction with the object. 142 Laura A. Carlsonparticipants within a given context placed the beanbag with respect to the emphasized functional part. This is similar to the incomplete bias (mean deviation <100%) obtained by Carlson-Radvansky et al. (1999) for placements of a functionally related object relative to a preselected functional part. Presumably the center of mass of the object also played a role in biasing the placements; this is currently being investigated. Finally, the functional parts seem to have diVerential salience, as reflected in diVerent degrees of bias. To demonstrate this, for each object, placements were recoded as deviations from the midpoint of the distance between the centers of the two functional parts, with negative values associated with placements biased to the leftmost functional part (arbitrarily defined as part 1) and positive values associated with placements biased to the rightmost functional part (defined arbitrarily as part 2). Mean deviations and statistical comparisons are shown in Table I. The fact that the functional bias toward part 2 was significantly diVerent from the neutral context indicates an influence of the video context on performance in the spatial language task. The fact that the neutral context had a negative value indicates a preexisting bias to place the object toward functional part 1. Moreover, emphasizing part 1 did not enhance this bias. Current work is examining more precisely how the type of functional part impacts its salience. Table I also shows the mean number of placements that were biased positively toward part 2. These were computed by calculating the number of positive placements (out of eight possible) for each object and then averaging across objects. The relatively large number of objects that were biased positively for the context emphasizing part 2 indicates that the deviation percentages were not due to large deviations associated with one or two particular objects. TABLE I Mean Deviations and Number of Objects Exhibiting a Positive Deviation as a Function of Context Conditiona Context Mean deviation Mean No. biased toward part 2 Emphasize part 1 .109 (.138) * 3.6 (.422) * Emphasize part 2 .505 (.148) y 5.7 (.341) y Neutral .106 (.158) * 3.9 (.491) * a Mean deviations are expressed as the proportion of distance between midpoint and the functional part, with negative deviations reflecting a bias toward part 1 and positive deviations reflecting a bias toward part 2. The maximum number of placements per object was 8. Conditions with different superscripts are significantly different ( p < .01). Standard errors are in parentheses. Using Spatial Language 143