Journal of Experimental Psychology: Human Perception and Performance 2015, Vol. 41, No. 3, 807-815
© 2015 American Psychological Association 0096-1523/15/$ 12.00 http://dx.doi.org/10.1037/xhp0000050
The Selection and Usage of Information for Perceiving and Remembering Intended and Unintended Object Properties Brandon J. Thomas and Michael A. Riley University of Cincinnati The current study was designed to investigate the linkage between intention, attention, and information selection and detection. Two experiments determined whether the intention to perceive maximum overhead reaching height with the use of handheld rods with different rotational inertia resulted in the ability to remember the rods’ heaviness after they were removed from view. Participants were partially successful at distinguishing the heaviness of rods but only when visual information was restricted and reaching height was perceived using the rods by dynamic touch. The results support an ecological approach to attention, and suggest that information picked up for perception can support the memory of other properties of the environment if their informational bases are related. Keywords: direct perception, affordance, remembered affordance, attention, nonspecifying information
In the ecological approach to perception-action (E. J. Gibson, 1969; Gibson & Pick, 2000; J. J. Gibson, 1966, 1979) attention is defined as an active process of selecting and obtaining goal relevant information. Many informational variables are potentially available in the ambient energy arrays to which animals’ percep tual systems are sensitive. When performing a particular percep tual task, a perceiver attends to and selects one or some of those variables for detection, but ignores others. Michaels and Carello (1981) described attention, thus construed, as the control o f infor mation detection. This formulation suggests that an organism’s intention is a driving force of attention, and implicates purposive, exploratory (i.e., information-generating or -seeking) behaviors as major parts of the selection process involved in attention. An organism that intends to perceive a certain property, affordance, or event will behave to select information relevant to that perceptual goal (Turvey, Carello, & Kim, 1990). The current study was designed to further test how intention, attention, and the selection and detection of information affect perception. Specifically, the study was conducted to address the question of whether an intention-attention-exploration linkage for detecting one object property could support remembering object properties other than the intended property if the informational variables that specify the properties in question correlated with each other. The preceding perspective on intention, attention, and explora tion has been supported by the results of several empirical studies. For example, in Arzamarski, Isenhower, Kay, Turvey, and Mi chaels (2010) participants were instructed to perceive by dynamic touch (Carello & Turvey, 2004) either the width or length of unseen handheld objects. Participants accomplished those two
distinct perceptual goals by selectively attending to the informa tion that uniquely specified the to-be-perceived object dimensions, and different intentions were accompanied by attention toward and detection of different informational variables (see also Michaels & Isenhower, 2011; Riley, Wagman, Santana, Carello, & Turvey, 2002; Stephen, Arzamarski, & Michaels, 2010). Similarly, partic ipants can discriminate differences between two unseen rods held by the same hand (Turvey, Carello, Fitzpatrick, Pagano, & Kadar, 1996) and the full and partial length of a single unseen rod (Carello, Santana, & Burton, 1996) by selectively exploiting dif ferent informational variables. Thus, the intention of perceivers constrains the information attended and detected. Michaels, Weier, and Harrison (2007) found similar results when participants perceived the affordances (action capabilities) of multiple tools in a number of different behavioral tasks. Perception of tool suitability was accomplished by attending to task-specific properties of the tools and information about each tool was gen erated by differential exploratory movement patterns. Further more, participants tended to rely on visual information when it was available, even when concurrently available dynamic touch infor mation could have supported perception of the intended affor dance. In other words, the perception of affordances was achieved by selectively attending to information in the optic array, even though the affordances were also specified by information avail able for pickup in another energetic medium by a different per ceptual system (i.e., modality). This is consistent with other studies that have found that vision is consistently preferred over other modalities for the perception of affordances and object properties when information in other energetic media is available (Bongers, Michaels, & Smitsman, 2004). A basic principle of J. J. Gibson’s (1979) ecological approach is that information (patterns within ambient energy arrays) lawfully specifies properties of the animal-environment system. A number of studies suggest that individuals sometimes use variables that merely correlate with specifying variables— so-called nonspecify ing variables (Gilden & Proffitt, 1994; Jacobs & Michaels, 2001; Jacobs, Runeson, & Michaels, 2001; Runeson & Vedeler, 1993;
This article was published Online First April 13, 2015. Brandon J. Thomas and Michael A. Riley, Center for Cognition, Action, and Perception, Department of Psychology, University of Cincinnati. Correspondence concerning this article should be addressed to Brandon J. Thomas, Mail Location 0376, Department of Psychology, University of Cincinnati, Cincinnati, OH 45221-0376. E-mail: [email protected]
807
THOMAS AND RILEY
808
Withagen & van Wermeskerken, 2009).1 Perceptual accuracy is typically lower when a perceiver is using a nonspecifying variable than it could be if the perceiver was using a specifying variable. The degree of accuracy depends on the magnitude of the correla tion of the nonspecifying variable with the property in question. With feedback, perceivers can improve perceptual accuracy by shifting to a reliance on specifying variables (i.e., those specific to the intended perceptual task: Jacobs et al., 2001; Michaels, Arzamarski, Isenhower, & Jacobs, 2008; Michaels & de Vries, 1998; Smith, Flach, Dittman, & Stanard, 2001). This process is called attunement (Michaels et al., 2008) or the education o f attention (J. J. Gibson, 1966). The relevance of that work for the present study is that a single informational variable can correlate with multiple properties in addition to uniquely specifying some other property. If a perceiver intends to perceive a certain property, such as the length of a handheld rod, then the perceiver’s intention-attention-exploration linkage might incidentally support the perception of properties that merely correlate with the information detected about length. The main hypothesis that motivated this project was that an intention to perceive a certain property constrains attention and that the infor mation picked up for that intention can be used to report other properties of the environment if the two properties’ informational bases are related, but not if they are uncorrelated. To investigate these issues we leveraged the extensive empirical literature on perception of hand-held objects using the haptic subsystem of dynamic touch. Research on dynamic touch has provided an understanding of the different informational variables used to perceive, for example, both the length and heaviness of hand-held rods. Perceived length is a power function of the rods’ first moment of inertia raised to a positive value and the third moment of inertia raised to a negative value,
lp = KW*,
(i)
where Lp is perceived length, 7, is the first moment of inertia, and 73 is the third moment of inertia (Fitzpatrick, Carello, & Turvey, 1994)2. Perceived heaviness is a function of the object’s mass, inertia ellipsoid volume (V), and inertia ellipsoid symmetry (5), with the latter two quantities given by V=4Tt/ 3(I ] X /2 X /3r 1/2
(2)
S = 2 • /3/ ( / | + /2).
(3)
The important point is that the variables f and / 3 influence both the perceived length and perceived heaviness of a wielded object (Shockley, Grocki, Carello, & Turvey, 2001). The intention to perceive the length of an unseen rod via dynamic touch might therefore partially support perception of the rod’s heaviness, even when perceivers lack that explicit intention from the outset, if perceivers are detecting information that expresses the influence of those mechanical quantities. In support of that possibility, it has been shown that training on a number of perception-action tasks transfers to performance of other similar tasks (Abeele & Bock, 2003; Hamilton, 1964; Rieser, Pick, Ashmead, & Garing, 1995; Stephen & Hajnal, 2011). For instance, Stoffregen, Yang, Giveans, Flanagan, and Bardy (2009) found that practice locomoting with a wheelchair resulted in more
accurate judgments of the minimum lintel that could be passed under, even though pass-under-ability was not explicitly practiced. Similarly, Wagman (2012) found that practice in overhead reach ing while standing on toes improved the perception of reaching with an object without standing on toes. Also, perceptual training on perceived length of rods by dynamic touch transfers to per ceived location of the center of percussion of the rods (Withagen & Michaels, 2007); despite that, Cooper, Carello, and Turvey (1999) found that reports of length and center of percussion were perceptually independent. The capacity for transfer across tasks suggests that individuals pick up informational variables that in cidentally share common constraints. Though the perception of different properties may be separable, their respective informa tional bases might correlate and thus support the perception of different properties. Again, the primary hypothesis of this study is that, by virtue of constraining information selection and detection, the intention to perceive a property will partially support the memory of properties that share an informational basis with the intended property, but not of other properties that do not. We tested this hypothesis using the remembered affordance paradigm (Wagman, Thomas, McBride, & Day, 2013) because that paradigm permits an evalu ation of what it is that perceivers fundamentally tune in to when they encounter some environmental circumstance (Thomas & Ri ley, 2014). The straightforward premise of this approach is that people will remember most accurately what it is they attend to in the first place. It is possible that individuals can accurately remem ber object properties that differ from the properties they initially intended to perceive if the informational variable detected corre lates with the “unintended” object property (i.e., if the detected information acts as a nonspecifying variable). This should, how ever, depend on the salience and availability of the type of infor mation to the perceiver, which we manipulated by making differ ent perceptual modalities available to the perceiver. For example, in accordance with the results of Michaels et al. (2007), perceivers might attend to the optic array and utilize simple geometric infor mation to perceive object length when the object is visible, but wield the object to utilize dynamic touch when vision is unavail able. The latter might incidentally support perception of object heaviness, but the former might not. Participants were tasked with reporting their maximum over head reaching height while wielding rods that were either visible (Experiment 1) or occluded (Experiment 2). Two different rods were used in each experiment. They were the same length, but one had a weight attached to the end, increasing its mass and rotational inertia relative to the other (Table 1); the attached weight was not visible (both rods appeared to have a uniform tubular shape). After participants estimated how high they could reach with the rods, the rods were removed from view. Participants then reported their
1 The values of nonspecifying variables are not arbitrary or random— they are lawfully determined and thus specific to some aspect of the animal-environment system— but the variables do not specify the intended property. 2 The current study assumed that information picked up to perceive maximum overhead reaching height using a handheld rod is the same as that picked up by participants in studies on the perception of length. In those studies, participants are typically instructed to report perceived object length as the distance they could reach with the object (Riley et al., 2002; Withagen & Michaels, 2007; Withagen & Michaels, 2004).
SELECTION AND USAGE OF INFORMATION
Table 1 Metric Properties o f the Three Rods Used in Experiments 1 and 2 Property
Light rod
Heavy rod
Standard rod
Length Mass Handle (diameter) Distal end (diameter) /, (g X cm2) h (g X cm2) 7, (g X cm2)
46 cm 11L4 g 1.3 cm 5.2 cm 36,766 35,739 1,780
46 cm 163.1 g 1.3 cm 5.2 cm 126,657 125,183 2,945
46 cm 50.5 g 1.3 cm 1.3 cm 28,971 28,401 604
Note. Moments of inertia are calculated with respect to the wrist as the origin.
remembered reach-ability with each rod, and then reported the remembered heaviness of each rod. They were not instructed to pay attention to the rods’ heaviness while they were wielding them, so participants had no explicit need to attend to heaviness until they were later asked to report it. We hypothesized that reports of both perceived and remembered reach-ability would be scaled to the visible rod length when the rods were visible (and thus in Experiment 1 there would be no significant effect of rod) but would be based on the rods’ inertial properties when the rods were occluded in Experiment 2 (in which case there would be a significant effect of the rod— when partic ipants had to rely on dynamic touch, the rod with the extra attached weight and consequently greater /, would be perceived to increase reach-ability more than the other rod). We additionally hypothe sized that remembered heaviness would be discerned when the rods were occluded in Experiment 2. Both length and heaviness are related to I, and I3, so the intention to detect information about how the rods’ length enhanced reach-ability should incidentally support perception of heaviness. We did not expect participants to easily discriminate the heaviness of the rods when the rods were visible in Experiment 1, since the intention to perceive reach ability visually would not encourage participants to detect infor mation related to the rods’ heaviness (cf., Michaels et al., 2007).
809
Method Participants. Thirty-seven undergraduates from the Univer sity of Cincinnati participated in this experiment for course credit. Written informed consent was obtained prior to data collection. Given the overhead reaching task and the height constraints of the laboratory and apparatus, participants were required to be no taller than 173 cm. The average height of participants was 164.4 cm (SD = 5.1 cm). The sample consisted of 32 females and five males, and their average age was 20.0 years (SD = 3.9 years). Materials and apparatus. When providing reaching height reports, participants instructed an experimenter to use a pulley and string to raise or lower a marker consisting of 11 stacked washers (11 g mass, 3 cm diameter, 2 cm tall) until they felt it was at their maximum reaching height. Maximum reaching height was defined as the maximum height at which the marker could be touched with the fingertips of the right hand or the distal tip of the rod held in the right hand with the arm or rod fully extended above the head, without lifting either foot off of the floor or standing on tiptoes. The marker was suspended via a pulley system attached to the top of a planar surface (264 cm tall X 165 cm wide). To create a uniform background behind the marker, gray sheets were draped over the surface. A tape measure affixed to the back of the vertical surface was used to measure the height of the suspended marker once participants finalized their perceptual reports. The tape mea sure was not visible to participants. The apparatus is depicted in Figure 1. Two 46 cm long aluminum rods were used to extend partici pants’ reach. Both rods had a 50 g circular weight (5.2 cm diameter and 1.9 cm tall) attached 13.9 cm from the bottom of the rod.
Experiment 1 In Experiment 1, visual and dynamic touch information were both available to participants. We predicted that participants would perceive reach-ability using geometric, visual information. There fore, perceptual reports of reach-ability would not differ between the two wielded rods, since they were the same length despite one having the extra attached weight. We also expected that partici pants would initially be unable to differentiate the heaviness of the two rods, since their intention would not support the pickup of information about heaviness. However, in a second set of trial blocks (completed after the first heaviness reports were obtained), we predicted that participants would differentiate the rods in terms of heaviness because their expectation to do so in the second set would lead them to attend to information about rod heaviness. We nonetheless expected that participants would use the readily avail able visual (geometric) information about rod length to inform their reach-ability estimates in the second set of trial blocks, and therefore those estimates would not differ for the two rods.
Figure 1. A pictorial representation of the apparatus used in Experiments 1 and 2. Participants instructed an experimenter to raise and lower the marker. The rod was wielded in the participant’s right hand in the reachwith-rod-present conditions while they reported, with the rod fully visible (Experiment 1) or occluded by a curtain that hung from the ceiling of the laboratory (Experiment 2, as shown). The weights on both rods were hidden by construction paper wrapped around the effective portion of the rod.
810
THOMAS AND RILEY
Everything below this weight was considered the rod’s “handle” resulting in an effective rod length of 32.1 cm (i.e., given the length of the handle, the rod could have extended reach by a maximum of 32.1 cm). Brown construction paper was wrapped around the distal portion of each rod, covering it from the end of the handle to the end of the rod, and red duct tape covered the distal end of the rod. In addition to the 50 g weight near the handle, which was present on both rods, the heavy rod had an additional 50 g weight attached to the distal end (i.e., the center of the attached weight was 29.5 cm from the top of the handle). The change in the first moment of inertia (/j) between the “light” and “heavy” rods was approximately 71% of the heavier rod’s / t3. The weights on both rods were covered by the brown paper so they were not visible to participants. A third aluminum rod with the same length and diameter as the other two rods, but without any additional attached weights, was used as a standard for the heaviness mag nitude estimation trials. The properties of the rod set are listed in Table 1. Procedure. Each participant stood 285 cm from the vertical surface in a viewing area (50 X 50 cm) that was centered relative to the vertical surface. At the beginning of each trial in a given condition, the marker was set either at its highest position (264 cm from the laboratory floor) or its lowest position (on the laboratory floor). In each perceptual report condition, participants completed three ascending trials (in which the object was initially set at its lowest position and then raised during reporting) and three de scending trials (in which the object was initially set at its highest position and then lowered during reporting). Ascending and de scending trials alternated. All participants completed the conditions in blocked fashion and necessarily did so in the same order. There were two sets of trial blocks. Each set began with a block of reach-with-rod-present trials. Half of the participants wielded the light rod in this block, and half wielded the heavy rod. Participants held out their right arm so that it was parallel to their body and the rod was placed in their hand, aligned with the distal end of the handle (see Figure 1). Participants were instructed to wield the rod comfortably in their hand about their wrist, elbow, or shoulder as long as they did not explicitly practice reaching upward with the rod or raise their elbow above shoulder height. Wielding movements were not mea sured in either experiment. However, the experimenter observed that participants tended to wield the rod by producing roughly circular motions with the tip of the rod. Participants reported the maximum height they would be able reach if they were to walk over to the front of the vertical surface and use the rod to reach for the marker. To provide reports, participants instructed the exper imenter to raise or lower the marker until it was at the maximum height the participant believed he or she could reach. The exper imenter adjusted the height of the marker from behind the vertical surface and was not visible to the participant. Participants were able to fine tune the height of the marker on a given trial until they were satisfied with the report. After each trial, participants were asked to close their eyes while the marker was set for the next trial. At the conclusion of this condition, participants closed their eyes and the rod was placed behind the vertical surface and out of view. They then performed the reach-with-hand condition, in which participants reported the maximum height they could reach if they were to walk over to the front of the vertical surface and reach up with the fingertips of their right hand.
Next, in the reach-with-rod-absent conditions, participants per formed the same perceptual task as in the reach-with-rod-present conditions except the rod remained out of view. Participants were instructed to imagine that they were still holding the rod and report the maximum height they would be able to reach if they were to walk over to the vertical surface, and use it to reach for the marker. The order of conditions resulted in a delay of a few minutes (i.e., the duration of the trial block for the reach-with-hand condition) between the reach-with-rod-present and the reach-with-rod-absent conditions. In the heaviness conditions, the participant was handed the standard rod in their right hand and told that they would refer to it to report the heaviness of the rod they wielded in the first block of trials. The participant was told that the standard rod was assigned a value of 100 that did not correspond to any actual units of heaviness, and that if they thought the rod held earlier was twice as heavy as the standard, to give it a value of 200 or a value of 50 if they thought it was half as heavy. However, they were told that they could give the rod any value they wanted. Prior to this point in the experiment, participants had no knowledge that they were to report the heaviness of the rod. Participants then repeated those four conditions in the second set of trial blocks. If in the first set of trial blocks they performed reach-with-rod-present trials with the light rod, they then per formed the reach-with-rod-present trials with the heavy rod in the second set of trial blocks. The second set continued with reachwith-hand, reach-with-rod-absent, and heaviness blocks. After completing the first set of trial blocks, it is likely that participants would expect to be instructed to report heaviness in the second set of trial blocks, and thus may have attended explicitly to rod heaviness in the second set of trial blocks. Our primary hypothesis about whether attending to rod length would support perception of rod heaviness was most appropriately evaluated using the data from the first block of trials. For these reasons we included the set of trial blocks as a factor in analyses of heaviness reports (see Figure 2 for a schematic of the experimental procedure). There were six trials per condition, yielding a total of 48 trials in the experiment. After the completion of all trials, the experimenter measured participants’ standing height and maximum reaching heights when reaching with their right hand and when reaching with the heavy and light rods held in their right hand. At no prior point in the experiment did any participant approach the surface or attempt to reach for the marker.
Results and Discussion Each participant’s perceptual reports were averaged across the six trials per condition. Those data were screened for outliers that were 2.5 SD less or greater than the median. One outlier was found and that participant’s data were excluded from any further analy ses. Actual maximum reaching heights were submitted to a one-way, within-subjects analysis of variance (ANOVA) with reach-withhand, reach-with-heavy-rod, and reach-with-light-rod as the con ditions. There was a significant effect of condition, F(2, 66) = 3 A pilot study ( N = 4 naive participants) found that participants per ceived the heavy rod to be significantly heavier than the light rod (p < .05).
SELECTION AND USAGE OF INFORMATION
811 L ig h t r o d ■ H e avy ro d
P re s e n t
H and
Absent
Figure 3. Mean perceptual reports (cm) for the reach-with-rod-present, reach-with-hand, and reach-with-rod-absent conditions X heavy and light rods in Experiment 1. Error bars represent within-subjects standard errors.
Figure 2. A schematic of the experimental design and procedure for Experiments 1 and 2. The order of the rod presentation was counterbal anced across participants.
2396.81, p < .001, tip = .99. Bonferroni-corrected post-hoc t tests revealed that the reach-with-light-rod (M = 240.8 cm) and reachwith-heavy-rod (M = 241.3 cm) were greater than the reach-withhand (M = 209.1 cm), p < .001. The reach-with-light-rod and reach-with-heavy-rod did not differ, p = .682. The differences in reaching heights between the reach-with-hand condition and the two rod conditions (31.7 cm and 32.3 cm, respectively) was close to the 32.1 cm effective length of the rod. Mean perceived maximum reaching height reports were com pared in a two-way, repeated-measures ANOVA with the factors of rod (light and heavy) and condition (reach-with-rod-present, reach-with-hand, and reach-with-rod-absent). There was a signif icant effect of condition, F (l, 66) = 243.80, p < .001, = .88. According to Bonferroni-corrected post-hoc t tests, the reach-withrod-present (M = 219.0 cm) and reach-with-rod-absent (M = 219.8 cm) conditions did not significantly differ (p = .972), but both were significantly greater than the reach-with-hand condition (M = 201.9 cm), both p < .001. The main effect of rod [F( 1, 66) = 0.01, p = .906, rip = .01] and rod X condition interaction [F(2, 66) = 2.14, p = .216, rip = .06] were not significant. There was no difference between the mean reach-with-hand reports when participants reported the reach-ability of the light rod (M = 201.5 cm) compared to the heavy rod (M = 202.2 cm), t(33) = 0.62, p = .542 (Figure 3). Because of the carryover effects that were anticipated to occur when participants reported the heaviness of the rod in the second set of trial blocks, we compared mean heaviness reports in a two-way ANOVA with a between-subjects factor of rod (light and heavy) and a within-subjects factor of set (Set 1 and Set 2). There was a significant main effect of rod F (l, 32) = 10.08, p = .003, T)p = .24, with the heavy rod condition (M = 201.3) eliciting greater reports than the light rod condition (M = 142.5). There was a significant main effect of set, F (l, 32) = 6.39, p = .017, Tip = .17, with greater reports in Set 2 (M = 187.6) than Set 1 (M = 156.2). There was also a significant interaction, F (l, 32) = 7.23, p = .017, Tip = .18. A simple-effects test revealed that the Set 1
heavy rod condition (M = 168.8) did not differ from the light rod condition (M = 143.5), p = .263. However, in Set 2, the heavy rod condition (M = 233.8) significantly differed from the light rod condition (M = 141.5), p < .001 (Figure 4). A second set of simple-effects tests revealed that the Set 1 heavy rod condition differed from the Set 2 heavy rod condition, p = .001. The Set 1 light rod condition did not differ from the Set 2 light rod condition, p = .910. The results confirmed our predictions. The reach-ability reports in the reach-with-rod-present and -absent conditions were different from the reach-with-hand condition, and there was no difference between the light and heavy rods. When participants could see the rods, they seem to have ignored potential information available through dynamic touch about the rods’ length. There was no difference in remembered heaviness between the light and heavy rods when participants reported the rods’ heaviness in the first set of trial blocks. However, once participants were explicitly aware that they would be asked to report heaviness, they became sensi tive to the rods’ differential mass and rotational inertia in the second set of trial blocks (indicating participants simply did not lack the ability to distinguish some property of the rods by dy namic touch).
Figure 4. Mean magnitude estimated heaviness reports for Set 1 and Set 2 X heavy and light rods in Experiment 1. Error bars represent withinsubjects (set) and between-subjects (rod) standard errors.
THOMAS AND RILEY
812 Experiment 2
Experiment 2 was identical to Experiment 1, except the rods were occluded from view in the reach-with-rod-present conditions. We expected participants to indicate a higher reach with the heavy rod by virtue of its greater rotational inertia specifying greater reach-ability. We also expected that attention to dynamic touch information about length would support accurate discrimination of the two rods’ weights on the first set of trial blocks.
Method Participants. Thirty-nine undergraduates from the University of Cincinnati participated in this experiment for course credit. Written informed consent was obtained prior to data collection. Participants were required to be no taller than 173 cm for the same reason as Experiment 1. The average height of participants was 166.4 cm (SD = 6.4 cm). The sample consisted of 38 females and one male, and their average age was 19.9 years (SD = 4.3 years). Materials and apparatus. The materials and apparatus from Experiment 1 were again used in Experiment 2. A gray curtain that fully occluded the wielded rods was also utilized. The curtain had a slit that participants put their hand through and was hung from the ceiling. Procedure. The procedure was the same as Experiment 1, except for in the reach-with-rod-present conditions. In these con ditions, participants put their hand through the slit in the curtain, which necessitated that they hold their hand parallel to their body (see Figure 1). The rod was placed in the participant’s hand so that the distal end of the handle was even with the bottom of the hand. Participants were instructed to wield the rod comfortably about the shoulder, elbow, and wrist as long as they did not practice reaching with the rod, raise their elbow above shoulder height, or hit the screen with the rod. Except for the requirement to place the hand through the screen, the wielding procedure was identical to Ex periment 1. All other instructions and procedures were the same, including the amount of time the rods were wielded.
Results and Discussion Actual maximum reaching heights were submitted to a one-way, within-subjects ANOVA with participants reach-with-hand, reachwith-light-rod, and reach-with-heavy-rod as the conditions. There was a significant effect of condition, F(2, 74) = 2997.40, p < .001, r\2 p = .99. Bonferroni-corrected post-hoc t tests revealed that the reach-with-light-rod (M = 239.4 cm) and reach-with-heavyrod (M = 239.1 cm) were greater than the reach-with-hand values (M = 208.3 cm), both p < .001. The reach-with-light-rod and reach-with-heavy-rod did not differ from each other, p = .530. Mean perceived maximum reaching height reports were com pared in a two-way, repeated-measures ANOVA with the factors of rod (light and heavy) and condition (reach-with-rod-present, reach-with-hand, and reach-with-rod-absent). There was a main effect of rod, F (l, 41) = 18.36, p < .001, r\2 = .31; reports made in the heavy rod condition (M = 209.5 cm) were greater than the light rod condition (M = 205.1 cm). There was a significant effect of condition, F(2, 82) = 108.98, p < .001, t)2 = .73. According to Bonferroni-corrected post-hoc t tests, all conditions significantly differed, with the reach-with-rod-absent condition (M = 214.2 cm)
greater than the reach-with-rod-present condition (M = 210.2 cm), and both of those were greater than the reach-with-hand condition (M = 197.5 cm), all p < .001. There was a significant rod X condition interaction, F(2, 82) = 5.24, p = .007, -t\2p = .11 (Figure 5). Post hoc simple-effects tests revealed that the light and heavy rod conditions differed in the reach-with-rod-present and -absent conditions (all p < .003), but the rod reports did not differ in the reach-with-hand conditions (p = .276); this was expected because the rod variable was essentially undefined during reach-with-hand conditions. As in Experiment 1, mean heaviness reports were compared in a two-way ANOVA, with the between-subjects factor of rod (heavy and light) and the within-subjects factor of set (Set 1 and Set 2). There was a significant main effect of rod, F (l, 40) = 25.67, p < .001, T)p = .39, with the heavy rod condition (M = 239.9) being greater than the light rod condition (M = 158.9). There was not a significant main effect of set, F (l, 40) = .69, p = .412, r\2 = .02. The interaction was not significant, F (l, 40) = 2.22, p = .144, T)p = .05 (Figure 6). Both present and absent reports of overhead reach with the rod were greater than reports in the reach-with-hand condition. Reach ability reports with the heavy rod were greater than reports with the light rod in both the present and absent conditions. Also, reports of remembered heaviness were significantly greater for the heavy than the light rod in both Set 1 and Set 2. The Set 1 heaviness results indicate that the intention-attention-exploration linkage for perceived reach-ability with the rods appeared to support memory of the rods’ heaviness by dynamic touch, because of the partial overlap in the variables that determine both perceived heaviness and perceived reach-ability (i.e., /, and /3 are involved in the perception of both properties).
General Discussion In Experiment 1, reach-ability reports were not affected by the differences in the rods’ inertial properties, indicating that partici pants attended to visual (geometric) information about the reach ability of the two rods when it was available. In Experiment 2, when visual information about the rods’ length was unavailable, perceived reach-ability differed for the light and heavy rods in a
Light rod
240
I Heavy rod
?
225
ao
210
QJ
> 7=
TO TO
195
JZ
© 2015 American Psychological Association 0096-1523/15/$ 12.00 http://dx.doi.org/10.1037/xhp0000050
The Selection and Usage of Information for Perceiving and Remembering Intended and Unintended Object Properties Brandon J. Thomas and Michael A. Riley University of Cincinnati The current study was designed to investigate the linkage between intention, attention, and information selection and detection. Two experiments determined whether the intention to perceive maximum overhead reaching height with the use of handheld rods with different rotational inertia resulted in the ability to remember the rods’ heaviness after they were removed from view. Participants were partially successful at distinguishing the heaviness of rods but only when visual information was restricted and reaching height was perceived using the rods by dynamic touch. The results support an ecological approach to attention, and suggest that information picked up for perception can support the memory of other properties of the environment if their informational bases are related. Keywords: direct perception, affordance, remembered affordance, attention, nonspecifying information
In the ecological approach to perception-action (E. J. Gibson, 1969; Gibson & Pick, 2000; J. J. Gibson, 1966, 1979) attention is defined as an active process of selecting and obtaining goal relevant information. Many informational variables are potentially available in the ambient energy arrays to which animals’ percep tual systems are sensitive. When performing a particular percep tual task, a perceiver attends to and selects one or some of those variables for detection, but ignores others. Michaels and Carello (1981) described attention, thus construed, as the control o f infor mation detection. This formulation suggests that an organism’s intention is a driving force of attention, and implicates purposive, exploratory (i.e., information-generating or -seeking) behaviors as major parts of the selection process involved in attention. An organism that intends to perceive a certain property, affordance, or event will behave to select information relevant to that perceptual goal (Turvey, Carello, & Kim, 1990). The current study was designed to further test how intention, attention, and the selection and detection of information affect perception. Specifically, the study was conducted to address the question of whether an intention-attention-exploration linkage for detecting one object property could support remembering object properties other than the intended property if the informational variables that specify the properties in question correlated with each other. The preceding perspective on intention, attention, and explora tion has been supported by the results of several empirical studies. For example, in Arzamarski, Isenhower, Kay, Turvey, and Mi chaels (2010) participants were instructed to perceive by dynamic touch (Carello & Turvey, 2004) either the width or length of unseen handheld objects. Participants accomplished those two
distinct perceptual goals by selectively attending to the informa tion that uniquely specified the to-be-perceived object dimensions, and different intentions were accompanied by attention toward and detection of different informational variables (see also Michaels & Isenhower, 2011; Riley, Wagman, Santana, Carello, & Turvey, 2002; Stephen, Arzamarski, & Michaels, 2010). Similarly, partic ipants can discriminate differences between two unseen rods held by the same hand (Turvey, Carello, Fitzpatrick, Pagano, & Kadar, 1996) and the full and partial length of a single unseen rod (Carello, Santana, & Burton, 1996) by selectively exploiting dif ferent informational variables. Thus, the intention of perceivers constrains the information attended and detected. Michaels, Weier, and Harrison (2007) found similar results when participants perceived the affordances (action capabilities) of multiple tools in a number of different behavioral tasks. Perception of tool suitability was accomplished by attending to task-specific properties of the tools and information about each tool was gen erated by differential exploratory movement patterns. Further more, participants tended to rely on visual information when it was available, even when concurrently available dynamic touch infor mation could have supported perception of the intended affor dance. In other words, the perception of affordances was achieved by selectively attending to information in the optic array, even though the affordances were also specified by information avail able for pickup in another energetic medium by a different per ceptual system (i.e., modality). This is consistent with other studies that have found that vision is consistently preferred over other modalities for the perception of affordances and object properties when information in other energetic media is available (Bongers, Michaels, & Smitsman, 2004). A basic principle of J. J. Gibson’s (1979) ecological approach is that information (patterns within ambient energy arrays) lawfully specifies properties of the animal-environment system. A number of studies suggest that individuals sometimes use variables that merely correlate with specifying variables— so-called nonspecify ing variables (Gilden & Proffitt, 1994; Jacobs & Michaels, 2001; Jacobs, Runeson, & Michaels, 2001; Runeson & Vedeler, 1993;
This article was published Online First April 13, 2015. Brandon J. Thomas and Michael A. Riley, Center for Cognition, Action, and Perception, Department of Psychology, University of Cincinnati. Correspondence concerning this article should be addressed to Brandon J. Thomas, Mail Location 0376, Department of Psychology, University of Cincinnati, Cincinnati, OH 45221-0376. E-mail: [email protected]
807
THOMAS AND RILEY
808
Withagen & van Wermeskerken, 2009).1 Perceptual accuracy is typically lower when a perceiver is using a nonspecifying variable than it could be if the perceiver was using a specifying variable. The degree of accuracy depends on the magnitude of the correla tion of the nonspecifying variable with the property in question. With feedback, perceivers can improve perceptual accuracy by shifting to a reliance on specifying variables (i.e., those specific to the intended perceptual task: Jacobs et al., 2001; Michaels, Arzamarski, Isenhower, & Jacobs, 2008; Michaels & de Vries, 1998; Smith, Flach, Dittman, & Stanard, 2001). This process is called attunement (Michaels et al., 2008) or the education o f attention (J. J. Gibson, 1966). The relevance of that work for the present study is that a single informational variable can correlate with multiple properties in addition to uniquely specifying some other property. If a perceiver intends to perceive a certain property, such as the length of a handheld rod, then the perceiver’s intention-attention-exploration linkage might incidentally support the perception of properties that merely correlate with the information detected about length. The main hypothesis that motivated this project was that an intention to perceive a certain property constrains attention and that the infor mation picked up for that intention can be used to report other properties of the environment if the two properties’ informational bases are related, but not if they are uncorrelated. To investigate these issues we leveraged the extensive empirical literature on perception of hand-held objects using the haptic subsystem of dynamic touch. Research on dynamic touch has provided an understanding of the different informational variables used to perceive, for example, both the length and heaviness of hand-held rods. Perceived length is a power function of the rods’ first moment of inertia raised to a positive value and the third moment of inertia raised to a negative value,
lp = KW*,
(i)
where Lp is perceived length, 7, is the first moment of inertia, and 73 is the third moment of inertia (Fitzpatrick, Carello, & Turvey, 1994)2. Perceived heaviness is a function of the object’s mass, inertia ellipsoid volume (V), and inertia ellipsoid symmetry (5), with the latter two quantities given by V=4Tt/ 3(I ] X /2 X /3r 1/2
(2)
S = 2 • /3/ ( / | + /2).
(3)
The important point is that the variables f and / 3 influence both the perceived length and perceived heaviness of a wielded object (Shockley, Grocki, Carello, & Turvey, 2001). The intention to perceive the length of an unseen rod via dynamic touch might therefore partially support perception of the rod’s heaviness, even when perceivers lack that explicit intention from the outset, if perceivers are detecting information that expresses the influence of those mechanical quantities. In support of that possibility, it has been shown that training on a number of perception-action tasks transfers to performance of other similar tasks (Abeele & Bock, 2003; Hamilton, 1964; Rieser, Pick, Ashmead, & Garing, 1995; Stephen & Hajnal, 2011). For instance, Stoffregen, Yang, Giveans, Flanagan, and Bardy (2009) found that practice locomoting with a wheelchair resulted in more
accurate judgments of the minimum lintel that could be passed under, even though pass-under-ability was not explicitly practiced. Similarly, Wagman (2012) found that practice in overhead reach ing while standing on toes improved the perception of reaching with an object without standing on toes. Also, perceptual training on perceived length of rods by dynamic touch transfers to per ceived location of the center of percussion of the rods (Withagen & Michaels, 2007); despite that, Cooper, Carello, and Turvey (1999) found that reports of length and center of percussion were perceptually independent. The capacity for transfer across tasks suggests that individuals pick up informational variables that in cidentally share common constraints. Though the perception of different properties may be separable, their respective informa tional bases might correlate and thus support the perception of different properties. Again, the primary hypothesis of this study is that, by virtue of constraining information selection and detection, the intention to perceive a property will partially support the memory of properties that share an informational basis with the intended property, but not of other properties that do not. We tested this hypothesis using the remembered affordance paradigm (Wagman, Thomas, McBride, & Day, 2013) because that paradigm permits an evalu ation of what it is that perceivers fundamentally tune in to when they encounter some environmental circumstance (Thomas & Ri ley, 2014). The straightforward premise of this approach is that people will remember most accurately what it is they attend to in the first place. It is possible that individuals can accurately remem ber object properties that differ from the properties they initially intended to perceive if the informational variable detected corre lates with the “unintended” object property (i.e., if the detected information acts as a nonspecifying variable). This should, how ever, depend on the salience and availability of the type of infor mation to the perceiver, which we manipulated by making differ ent perceptual modalities available to the perceiver. For example, in accordance with the results of Michaels et al. (2007), perceivers might attend to the optic array and utilize simple geometric infor mation to perceive object length when the object is visible, but wield the object to utilize dynamic touch when vision is unavail able. The latter might incidentally support perception of object heaviness, but the former might not. Participants were tasked with reporting their maximum over head reaching height while wielding rods that were either visible (Experiment 1) or occluded (Experiment 2). Two different rods were used in each experiment. They were the same length, but one had a weight attached to the end, increasing its mass and rotational inertia relative to the other (Table 1); the attached weight was not visible (both rods appeared to have a uniform tubular shape). After participants estimated how high they could reach with the rods, the rods were removed from view. Participants then reported their
1 The values of nonspecifying variables are not arbitrary or random— they are lawfully determined and thus specific to some aspect of the animal-environment system— but the variables do not specify the intended property. 2 The current study assumed that information picked up to perceive maximum overhead reaching height using a handheld rod is the same as that picked up by participants in studies on the perception of length. In those studies, participants are typically instructed to report perceived object length as the distance they could reach with the object (Riley et al., 2002; Withagen & Michaels, 2007; Withagen & Michaels, 2004).
SELECTION AND USAGE OF INFORMATION
Table 1 Metric Properties o f the Three Rods Used in Experiments 1 and 2 Property
Light rod
Heavy rod
Standard rod
Length Mass Handle (diameter) Distal end (diameter) /, (g X cm2) h (g X cm2) 7, (g X cm2)
46 cm 11L4 g 1.3 cm 5.2 cm 36,766 35,739 1,780
46 cm 163.1 g 1.3 cm 5.2 cm 126,657 125,183 2,945
46 cm 50.5 g 1.3 cm 1.3 cm 28,971 28,401 604
Note. Moments of inertia are calculated with respect to the wrist as the origin.
remembered reach-ability with each rod, and then reported the remembered heaviness of each rod. They were not instructed to pay attention to the rods’ heaviness while they were wielding them, so participants had no explicit need to attend to heaviness until they were later asked to report it. We hypothesized that reports of both perceived and remembered reach-ability would be scaled to the visible rod length when the rods were visible (and thus in Experiment 1 there would be no significant effect of rod) but would be based on the rods’ inertial properties when the rods were occluded in Experiment 2 (in which case there would be a significant effect of the rod— when partic ipants had to rely on dynamic touch, the rod with the extra attached weight and consequently greater /, would be perceived to increase reach-ability more than the other rod). We additionally hypothe sized that remembered heaviness would be discerned when the rods were occluded in Experiment 2. Both length and heaviness are related to I, and I3, so the intention to detect information about how the rods’ length enhanced reach-ability should incidentally support perception of heaviness. We did not expect participants to easily discriminate the heaviness of the rods when the rods were visible in Experiment 1, since the intention to perceive reach ability visually would not encourage participants to detect infor mation related to the rods’ heaviness (cf., Michaels et al., 2007).
809
Method Participants. Thirty-seven undergraduates from the Univer sity of Cincinnati participated in this experiment for course credit. Written informed consent was obtained prior to data collection. Given the overhead reaching task and the height constraints of the laboratory and apparatus, participants were required to be no taller than 173 cm. The average height of participants was 164.4 cm (SD = 5.1 cm). The sample consisted of 32 females and five males, and their average age was 20.0 years (SD = 3.9 years). Materials and apparatus. When providing reaching height reports, participants instructed an experimenter to use a pulley and string to raise or lower a marker consisting of 11 stacked washers (11 g mass, 3 cm diameter, 2 cm tall) until they felt it was at their maximum reaching height. Maximum reaching height was defined as the maximum height at which the marker could be touched with the fingertips of the right hand or the distal tip of the rod held in the right hand with the arm or rod fully extended above the head, without lifting either foot off of the floor or standing on tiptoes. The marker was suspended via a pulley system attached to the top of a planar surface (264 cm tall X 165 cm wide). To create a uniform background behind the marker, gray sheets were draped over the surface. A tape measure affixed to the back of the vertical surface was used to measure the height of the suspended marker once participants finalized their perceptual reports. The tape mea sure was not visible to participants. The apparatus is depicted in Figure 1. Two 46 cm long aluminum rods were used to extend partici pants’ reach. Both rods had a 50 g circular weight (5.2 cm diameter and 1.9 cm tall) attached 13.9 cm from the bottom of the rod.
Experiment 1 In Experiment 1, visual and dynamic touch information were both available to participants. We predicted that participants would perceive reach-ability using geometric, visual information. There fore, perceptual reports of reach-ability would not differ between the two wielded rods, since they were the same length despite one having the extra attached weight. We also expected that partici pants would initially be unable to differentiate the heaviness of the two rods, since their intention would not support the pickup of information about heaviness. However, in a second set of trial blocks (completed after the first heaviness reports were obtained), we predicted that participants would differentiate the rods in terms of heaviness because their expectation to do so in the second set would lead them to attend to information about rod heaviness. We nonetheless expected that participants would use the readily avail able visual (geometric) information about rod length to inform their reach-ability estimates in the second set of trial blocks, and therefore those estimates would not differ for the two rods.
Figure 1. A pictorial representation of the apparatus used in Experiments 1 and 2. Participants instructed an experimenter to raise and lower the marker. The rod was wielded in the participant’s right hand in the reachwith-rod-present conditions while they reported, with the rod fully visible (Experiment 1) or occluded by a curtain that hung from the ceiling of the laboratory (Experiment 2, as shown). The weights on both rods were hidden by construction paper wrapped around the effective portion of the rod.
810
THOMAS AND RILEY
Everything below this weight was considered the rod’s “handle” resulting in an effective rod length of 32.1 cm (i.e., given the length of the handle, the rod could have extended reach by a maximum of 32.1 cm). Brown construction paper was wrapped around the distal portion of each rod, covering it from the end of the handle to the end of the rod, and red duct tape covered the distal end of the rod. In addition to the 50 g weight near the handle, which was present on both rods, the heavy rod had an additional 50 g weight attached to the distal end (i.e., the center of the attached weight was 29.5 cm from the top of the handle). The change in the first moment of inertia (/j) between the “light” and “heavy” rods was approximately 71% of the heavier rod’s / t3. The weights on both rods were covered by the brown paper so they were not visible to participants. A third aluminum rod with the same length and diameter as the other two rods, but without any additional attached weights, was used as a standard for the heaviness mag nitude estimation trials. The properties of the rod set are listed in Table 1. Procedure. Each participant stood 285 cm from the vertical surface in a viewing area (50 X 50 cm) that was centered relative to the vertical surface. At the beginning of each trial in a given condition, the marker was set either at its highest position (264 cm from the laboratory floor) or its lowest position (on the laboratory floor). In each perceptual report condition, participants completed three ascending trials (in which the object was initially set at its lowest position and then raised during reporting) and three de scending trials (in which the object was initially set at its highest position and then lowered during reporting). Ascending and de scending trials alternated. All participants completed the conditions in blocked fashion and necessarily did so in the same order. There were two sets of trial blocks. Each set began with a block of reach-with-rod-present trials. Half of the participants wielded the light rod in this block, and half wielded the heavy rod. Participants held out their right arm so that it was parallel to their body and the rod was placed in their hand, aligned with the distal end of the handle (see Figure 1). Participants were instructed to wield the rod comfortably in their hand about their wrist, elbow, or shoulder as long as they did not explicitly practice reaching upward with the rod or raise their elbow above shoulder height. Wielding movements were not mea sured in either experiment. However, the experimenter observed that participants tended to wield the rod by producing roughly circular motions with the tip of the rod. Participants reported the maximum height they would be able reach if they were to walk over to the front of the vertical surface and use the rod to reach for the marker. To provide reports, participants instructed the exper imenter to raise or lower the marker until it was at the maximum height the participant believed he or she could reach. The exper imenter adjusted the height of the marker from behind the vertical surface and was not visible to the participant. Participants were able to fine tune the height of the marker on a given trial until they were satisfied with the report. After each trial, participants were asked to close their eyes while the marker was set for the next trial. At the conclusion of this condition, participants closed their eyes and the rod was placed behind the vertical surface and out of view. They then performed the reach-with-hand condition, in which participants reported the maximum height they could reach if they were to walk over to the front of the vertical surface and reach up with the fingertips of their right hand.
Next, in the reach-with-rod-absent conditions, participants per formed the same perceptual task as in the reach-with-rod-present conditions except the rod remained out of view. Participants were instructed to imagine that they were still holding the rod and report the maximum height they would be able to reach if they were to walk over to the vertical surface, and use it to reach for the marker. The order of conditions resulted in a delay of a few minutes (i.e., the duration of the trial block for the reach-with-hand condition) between the reach-with-rod-present and the reach-with-rod-absent conditions. In the heaviness conditions, the participant was handed the standard rod in their right hand and told that they would refer to it to report the heaviness of the rod they wielded in the first block of trials. The participant was told that the standard rod was assigned a value of 100 that did not correspond to any actual units of heaviness, and that if they thought the rod held earlier was twice as heavy as the standard, to give it a value of 200 or a value of 50 if they thought it was half as heavy. However, they were told that they could give the rod any value they wanted. Prior to this point in the experiment, participants had no knowledge that they were to report the heaviness of the rod. Participants then repeated those four conditions in the second set of trial blocks. If in the first set of trial blocks they performed reach-with-rod-present trials with the light rod, they then per formed the reach-with-rod-present trials with the heavy rod in the second set of trial blocks. The second set continued with reachwith-hand, reach-with-rod-absent, and heaviness blocks. After completing the first set of trial blocks, it is likely that participants would expect to be instructed to report heaviness in the second set of trial blocks, and thus may have attended explicitly to rod heaviness in the second set of trial blocks. Our primary hypothesis about whether attending to rod length would support perception of rod heaviness was most appropriately evaluated using the data from the first block of trials. For these reasons we included the set of trial blocks as a factor in analyses of heaviness reports (see Figure 2 for a schematic of the experimental procedure). There were six trials per condition, yielding a total of 48 trials in the experiment. After the completion of all trials, the experimenter measured participants’ standing height and maximum reaching heights when reaching with their right hand and when reaching with the heavy and light rods held in their right hand. At no prior point in the experiment did any participant approach the surface or attempt to reach for the marker.
Results and Discussion Each participant’s perceptual reports were averaged across the six trials per condition. Those data were screened for outliers that were 2.5 SD less or greater than the median. One outlier was found and that participant’s data were excluded from any further analy ses. Actual maximum reaching heights were submitted to a one-way, within-subjects analysis of variance (ANOVA) with reach-withhand, reach-with-heavy-rod, and reach-with-light-rod as the con ditions. There was a significant effect of condition, F(2, 66) = 3 A pilot study ( N = 4 naive participants) found that participants per ceived the heavy rod to be significantly heavier than the light rod (p < .05).
SELECTION AND USAGE OF INFORMATION
811 L ig h t r o d ■ H e avy ro d
P re s e n t
H and
Absent
Figure 3. Mean perceptual reports (cm) for the reach-with-rod-present, reach-with-hand, and reach-with-rod-absent conditions X heavy and light rods in Experiment 1. Error bars represent within-subjects standard errors.
Figure 2. A schematic of the experimental design and procedure for Experiments 1 and 2. The order of the rod presentation was counterbal anced across participants.
2396.81, p < .001, tip = .99. Bonferroni-corrected post-hoc t tests revealed that the reach-with-light-rod (M = 240.8 cm) and reachwith-heavy-rod (M = 241.3 cm) were greater than the reach-withhand (M = 209.1 cm), p < .001. The reach-with-light-rod and reach-with-heavy-rod did not differ, p = .682. The differences in reaching heights between the reach-with-hand condition and the two rod conditions (31.7 cm and 32.3 cm, respectively) was close to the 32.1 cm effective length of the rod. Mean perceived maximum reaching height reports were com pared in a two-way, repeated-measures ANOVA with the factors of rod (light and heavy) and condition (reach-with-rod-present, reach-with-hand, and reach-with-rod-absent). There was a signif icant effect of condition, F (l, 66) = 243.80, p < .001, = .88. According to Bonferroni-corrected post-hoc t tests, the reach-withrod-present (M = 219.0 cm) and reach-with-rod-absent (M = 219.8 cm) conditions did not significantly differ (p = .972), but both were significantly greater than the reach-with-hand condition (M = 201.9 cm), both p < .001. The main effect of rod [F( 1, 66) = 0.01, p = .906, rip = .01] and rod X condition interaction [F(2, 66) = 2.14, p = .216, rip = .06] were not significant. There was no difference between the mean reach-with-hand reports when participants reported the reach-ability of the light rod (M = 201.5 cm) compared to the heavy rod (M = 202.2 cm), t(33) = 0.62, p = .542 (Figure 3). Because of the carryover effects that were anticipated to occur when participants reported the heaviness of the rod in the second set of trial blocks, we compared mean heaviness reports in a two-way ANOVA with a between-subjects factor of rod (light and heavy) and a within-subjects factor of set (Set 1 and Set 2). There was a significant main effect of rod F (l, 32) = 10.08, p = .003, T)p = .24, with the heavy rod condition (M = 201.3) eliciting greater reports than the light rod condition (M = 142.5). There was a significant main effect of set, F (l, 32) = 6.39, p = .017, Tip = .17, with greater reports in Set 2 (M = 187.6) than Set 1 (M = 156.2). There was also a significant interaction, F (l, 32) = 7.23, p = .017, Tip = .18. A simple-effects test revealed that the Set 1
heavy rod condition (M = 168.8) did not differ from the light rod condition (M = 143.5), p = .263. However, in Set 2, the heavy rod condition (M = 233.8) significantly differed from the light rod condition (M = 141.5), p < .001 (Figure 4). A second set of simple-effects tests revealed that the Set 1 heavy rod condition differed from the Set 2 heavy rod condition, p = .001. The Set 1 light rod condition did not differ from the Set 2 light rod condition, p = .910. The results confirmed our predictions. The reach-ability reports in the reach-with-rod-present and -absent conditions were different from the reach-with-hand condition, and there was no difference between the light and heavy rods. When participants could see the rods, they seem to have ignored potential information available through dynamic touch about the rods’ length. There was no difference in remembered heaviness between the light and heavy rods when participants reported the rods’ heaviness in the first set of trial blocks. However, once participants were explicitly aware that they would be asked to report heaviness, they became sensi tive to the rods’ differential mass and rotational inertia in the second set of trial blocks (indicating participants simply did not lack the ability to distinguish some property of the rods by dy namic touch).
Figure 4. Mean magnitude estimated heaviness reports for Set 1 and Set 2 X heavy and light rods in Experiment 1. Error bars represent withinsubjects (set) and between-subjects (rod) standard errors.
THOMAS AND RILEY
812 Experiment 2
Experiment 2 was identical to Experiment 1, except the rods were occluded from view in the reach-with-rod-present conditions. We expected participants to indicate a higher reach with the heavy rod by virtue of its greater rotational inertia specifying greater reach-ability. We also expected that attention to dynamic touch information about length would support accurate discrimination of the two rods’ weights on the first set of trial blocks.
Method Participants. Thirty-nine undergraduates from the University of Cincinnati participated in this experiment for course credit. Written informed consent was obtained prior to data collection. Participants were required to be no taller than 173 cm for the same reason as Experiment 1. The average height of participants was 166.4 cm (SD = 6.4 cm). The sample consisted of 38 females and one male, and their average age was 19.9 years (SD = 4.3 years). Materials and apparatus. The materials and apparatus from Experiment 1 were again used in Experiment 2. A gray curtain that fully occluded the wielded rods was also utilized. The curtain had a slit that participants put their hand through and was hung from the ceiling. Procedure. The procedure was the same as Experiment 1, except for in the reach-with-rod-present conditions. In these con ditions, participants put their hand through the slit in the curtain, which necessitated that they hold their hand parallel to their body (see Figure 1). The rod was placed in the participant’s hand so that the distal end of the handle was even with the bottom of the hand. Participants were instructed to wield the rod comfortably about the shoulder, elbow, and wrist as long as they did not practice reaching with the rod, raise their elbow above shoulder height, or hit the screen with the rod. Except for the requirement to place the hand through the screen, the wielding procedure was identical to Ex periment 1. All other instructions and procedures were the same, including the amount of time the rods were wielded.
Results and Discussion Actual maximum reaching heights were submitted to a one-way, within-subjects ANOVA with participants reach-with-hand, reachwith-light-rod, and reach-with-heavy-rod as the conditions. There was a significant effect of condition, F(2, 74) = 2997.40, p < .001, r\2 p = .99. Bonferroni-corrected post-hoc t tests revealed that the reach-with-light-rod (M = 239.4 cm) and reach-with-heavyrod (M = 239.1 cm) were greater than the reach-with-hand values (M = 208.3 cm), both p < .001. The reach-with-light-rod and reach-with-heavy-rod did not differ from each other, p = .530. Mean perceived maximum reaching height reports were com pared in a two-way, repeated-measures ANOVA with the factors of rod (light and heavy) and condition (reach-with-rod-present, reach-with-hand, and reach-with-rod-absent). There was a main effect of rod, F (l, 41) = 18.36, p < .001, r\2 = .31; reports made in the heavy rod condition (M = 209.5 cm) were greater than the light rod condition (M = 205.1 cm). There was a significant effect of condition, F(2, 82) = 108.98, p < .001, t)2 = .73. According to Bonferroni-corrected post-hoc t tests, all conditions significantly differed, with the reach-with-rod-absent condition (M = 214.2 cm)
greater than the reach-with-rod-present condition (M = 210.2 cm), and both of those were greater than the reach-with-hand condition (M = 197.5 cm), all p < .001. There was a significant rod X condition interaction, F(2, 82) = 5.24, p = .007, -t\2p = .11 (Figure 5). Post hoc simple-effects tests revealed that the light and heavy rod conditions differed in the reach-with-rod-present and -absent conditions (all p < .003), but the rod reports did not differ in the reach-with-hand conditions (p = .276); this was expected because the rod variable was essentially undefined during reach-with-hand conditions. As in Experiment 1, mean heaviness reports were compared in a two-way ANOVA, with the between-subjects factor of rod (heavy and light) and the within-subjects factor of set (Set 1 and Set 2). There was a significant main effect of rod, F (l, 40) = 25.67, p < .001, T)p = .39, with the heavy rod condition (M = 239.9) being greater than the light rod condition (M = 158.9). There was not a significant main effect of set, F (l, 40) = .69, p = .412, r\2 = .02. The interaction was not significant, F (l, 40) = 2.22, p = .144, T)p = .05 (Figure 6). Both present and absent reports of overhead reach with the rod were greater than reports in the reach-with-hand condition. Reach ability reports with the heavy rod were greater than reports with the light rod in both the present and absent conditions. Also, reports of remembered heaviness were significantly greater for the heavy than the light rod in both Set 1 and Set 2. The Set 1 heaviness results indicate that the intention-attention-exploration linkage for perceived reach-ability with the rods appeared to support memory of the rods’ heaviness by dynamic touch, because of the partial overlap in the variables that determine both perceived heaviness and perceived reach-ability (i.e., /, and /3 are involved in the perception of both properties).
General Discussion In Experiment 1, reach-ability reports were not affected by the differences in the rods’ inertial properties, indicating that partici pants attended to visual (geometric) information about the reach ability of the two rods when it was available. In Experiment 2, when visual information about the rods’ length was unavailable, perceived reach-ability differed for the light and heavy rods in a
Light rod
240
I Heavy rod
?
225
ao
210
QJ
> 7=
TO TO
195
JZ
