Selecting a Reference Frame: Evidence for Inhibition

Selecting a Reference Frame: Evidence for Inhibition a. An Initial Demonstration Given that these parameters operate at diVerent levels of representation, it is important to specify the locus at which a given process operates. Carlson-Radvansky and Jiang (1998) examined this question in the context of selecting a reference frame when diVerent sources of information do not agree on the orientation to assign a given axis (for an overview, see Carlson, 1999); an example would be rotating the reference object into a noncanonical orientation, thereby dissociating the absolute and intrinsic reference frames (see Section IIC2 and Fig. 2, bottom). Competition is assumed to result due to the simultaneous activation of these reference frames (Carlson-Radvansky & Irwin, 1994), and resolving this competition requires the selection of one frame over another. Carlson-Radvansky and Jiang (1998) demonstrated that the selection of a reference frame involved inhibiting the nonselected frame actively, such that use of that frame on subsequent trials was diYcult. This is consistent with the manner in which selection operates within other domains, such as visual attention, in which selection is more eYcient when the nonintended items are made less accessible (for reviews, see Fox, 1995; May, Kane, & Hasher, 1995). With respect to spatial language, Carlson (1999) speculated that such inhibition could serve as a useful mechanism that would bias speakers to consistently use the same reference frame across utterances within a conversation, thereby enabling coordination between speaker and listener (Garrod & Anderson, 1987; Schober, 1993). The general paradigm is illustrated in Fig. 5 using stimuli from Klatt and Carlson (2003). Trials were grouped in pairs, with each pair consisting of a prime trial followed by a probe trial. There were diVerent types of trial pairs, defined by the prime display. For all types, the probe trials were identical, using the same objects and the same spatial term; in the case of the probe trial in Fig. 5, the reference object was a sport utility vehicle, the located object was a ball, and the spatial term was ‘‘above.’’ The critical dependent variable was the response time on these probe trials for verifying that the spatial term corresponded to the spatial relation between the objects in Using Spatial Language 145the display. Participants were instructed to define ‘‘above’’ both with respect to the absolute reference frame and with respect to the intrinsic reference frame, such that they should make a ‘‘yes’’ response if the located object was placed with respect to either frame. Fig. 5. Prime and probe displays illustrating the negative priming paradigm used to assess the level at which inhibition was applied within the reference frame. 146 Laura A. CarlsonOf interest was how the probe response times varied as a function of the diVerent types of prime trials. In the control prime trial, participants would respond ‘‘yes’’ on the basis of defining above with respect to the absolute reference frame. Because the reference object (soccer ball) does not have predefined intrinsic sides, the intrinsic reference frame is not active. Consequently, there is no competition, and hence no need for inhibition. In contrast, in the end point prime trial (referred to as the matched relation condition by Carlson-Radvansky & Jiang, 1998), the reference object (shoes) has predefined sides, thereby supporting the use of an intrinsic reference frame. The reference object has been rotated, dissociating the intrinsic frame from the absolute frame. This results in competition among the frames and a consequent need for selection. Note that the placement of the located object is above with respect to the absolute reference frame. Accordingly, participants would select absolute above, as on the control prime trial. The diVerence is that the intrinsic frame is available and presumably active on the experimental prime trial, but not on the control prime trial. On all probe trials, the located object is placed above with respect to the intrinsic reference frame; accordingly, participants need to use intrinsic above. If selection involves inhibition of active competing frames, then intrinsic above would have been inhibited on the end point prime trial but not the control prime trial. As a consequence, use of intrinsic above on the end point probe trial should be more diYcult than on the control probe trial, as reflected in longer response times. This diVerence between experimental and control probe trials is referred to as negative priming. Carlson-Radvansky and Jiang (1998) found significant negative priming on the end point trials, suggesting that inhibition was applied to intrinsic above on the prime trial. Importantly, they also found significant negative priming on same axis trials (mismatched relations, Carlson-Radvansky & Jiang, 1998), illustrated in Fig. 5. On these trials, the spatial term on the prime trial (i.e., below) is at the opposite end point as the spatial term (i.e., above) on the probe trial. Because an equal amount of negative priming was observed for these trials as for end point trials, this suggests that inhibition was applied to the intrinsic vertical axis, encompassing both end points. This could suggest that selection of the absolute frame over the intrinsic frame on the prime trial occurred prior to the assignment of particular end points, consistent with the idea that orientation can be assigned separately from direction (Logan, 1995, 1996). b. How and Why Is Inhibition Applied? Central to understanding how and why inhibition is applied during reference frame selection is determining why a given frame is activated on the prime trial. Klatt and Carlson (2003) set about to address this question. One obvious possibility is that it is Using Spatial Language 147activated because it corresponds to the spatial term that is being verified. For example, assignment of the spatial term ‘‘above’’ on the prime trial would involve picking out the vertical axis on the absolute frame, picking out the vertical axis on the intrinsic frame, determining whether the placement of the located object fell along the vertical axis of either frame, and if so, choosing the appropriate frame and inhibiting the other. Klatt and Carlson (2003) reasoned that if this were the case, then negative priming should only be observed when spatial terms referring to the same axis are used on prime and probe trials (as in end point and same axis trials), but not when spatial terms corresponding to diVerent axes are used on prime and probe trials, referred to as across axes trials, illustrated in Fig. 5. For example, use of the term ‘‘right’’ on the prime trial should active the left/ right axis on both absolute and intrinsic frames, with a corresponding inhibition of the intrinsic left/right axis. As such, negative priming should only be observed when left/right are probe terms and not when above/below are probe terms. To assess this issue, in separate experiments, negative priming was assessed using above/below or left/right as probe terms and above/below, front/back, and left/right as prime terms. The resulting negative priming eVects and standard errors of the mean are presented in Table II as a function of prime and probe term. Counter to the hypothesis that the spatial term on the prime picks out the appropriate intrinsic axis that is subsequently inhibited, the pattern of negative priming eVects indicates that the intrinsic above/below axis is always inhibited, regardless of the prime, and that the intrinsic left/right axis is never inhibited. This interesting result suggests a primacy associated with the intrinsic above/below axis, TABLE II Mean Negative Priming Effect (in ms) and Standard Error of the Mean (in Parentheses) as a Function of Prime Trial Type a Prime term Probe term Level Negative priming Above/below Above/below End point 48 (13.6)* Below/above Above/below Within axis 32 (18.1) y Left/right Above/below Across axes 38 (9.6)* Front/back Above/below Across axes 29 (5.9)* Left/right Left/right End point 3 (20.9) Right/left Left/right Within axis 14.3 (19.5) Above/below Left/right Across axes 15.9 (14.2) Front/back Left/right Across axes 3 (14.2) a Symbols after negative priming effect indicate outcome of tests of significance against 0 ms, with *p < .001; y p ¼ .08. 148 Laura A. Carlsonperhaps due to its involvement in object recognition (i.e., in assigning the top and bottom sides to the reference object). This would be consistent with the idea that intrinsic axes underlying object recognition are recruited for spatial language use within the intrinsic reference frame (Logan & Sadler, 1996). Moreover, these data indicate that inhibition of a nonselected frame occurs prior to assigning the spatial term to a specific axis or end point and is applied to axes that are active as a consequence of earlier processes, such as identifying the objects during the constituent step of finding the relevant objects