Semantic category priming from the groundside of objects shown in nontarget locations and at unpredictable times

Abstract

Previous research demonstrated that familiar objects that are suggested, but not consciously perceived, on the groundside of the contours of a figure activate their semantic category during perceptual organization, at least when the figure appears at fixation at an expected time. Here, we investigate whether evidence for such semantic activation extends to stimuli presented at unpredictable times in peripheral locations. Participants categorized words shown centrally as denoting natural or artificial objects (Experiments 1 and 2a) or positive or negative concepts (Experiment 2b). Prior to the word, two distractor silhouettes appeared above and below fixation; both depicted novel figures. On experimental trials, portions of well-known (familiar) objects were suggested on the groundside of the borders of one (Experiment 1) or both (Experiment 2a and 2b) silhouettes. In Experiment 1, reaction times were slower when targets words were preceded by experimental distractor silhouettes regardless of whether the object suggested on the groundside of their borders was in the same or a different category as the object denoted by the word. Overall slowing may have occurred because (a) semantic category access by objects suggested on the groundside of experimental distractor silhouettes was sufficient to require filtering but not category-specific priming, (b) more competition for object status slowed processing of experimental compared to control silhouettes, or (c) featural differences increased the difficulty of processing the experimental versus the control silhouettes. The use of two identical experimental silhouettes in Experiment 2a allowed a semantic category priming effect to emerge, showing that the categories of objects suggested on the groundside of silhouette borders can be activated at unpredictable times in nontarget locations and in more than one location of the visual field. Experiment 2a suggested that (a) better explains the results of Experiment 1 than (b and c). Experiment 2b further ruled out explanations (b and c) as reasons for the Experiment 1 results by showing that the same pattern is not obtained when the semantic category of the objects suggested on the groundside of the experimental silhouettes borders is not task-relevant and does not require filtering. Thus, spatial prime-target congruence and temporal certainty are not necessary for priming by objects suggested on the groundside of figures. Implications for our understanding of the complex processes involved in perceptual organization are considered.

Introduction

Figure assignment is an elemental component of visual perception, although how it occurs is not yet fully understood. The shape of a figure is determined by its bounding contours. We take figure assignment to be object detection in its most fundamental instantiation; hence, we will often refer to figures as objects. The region outside an object's bounding contour (groundside) is perceived as shapeless near the contour and seems to simply continue behind the object as a background (see Figure 1). An abiding question is to what extent the groundside of a border is processed in the course of detecting where an object lies with respect to a border even though the ground is perceived as shapeless? 

Figure 1

 

(A & B) The borders shared by the black and white regions are assigned to the inner black regions; as a consequence, the inner regions are perceived as shaped figures/objects. The figural priors operating in A and B include symmetry, small area, convexity, and enclosure; familiar configuration also plays a role in A. The outer, white regions are perceived as shapeless grounds that merely continue behind the object at their shared border. Figure reproduced from Peterson & Skow-Grant (2003).

Semantic category priming arising from shapes suggested but not consciously perceived on the groundside of object contours

In both behavioral and psychophysiological experiments, Peterson and colleagues (Cacciamani, Mojica, Sanguinetti, & Peterson, 2014; Peterson, Cacciamani, Mojica, & Sanguinetti, 2012; Sanguinetti, Allen, & Peterson, 2014) showed that portions of well-known objects that were suggested but not consciously perceived on the groundside of the contour of novel silhouette objects nevertheless activated their semantic category. They demonstrated this behaviorally using a priming procedure in which participants categorized a target word as denoting a natural or an artificial object. Shortly before the target word, a black silhouette depicting a novel object centered on the same location as the upcoming target word was exposed briefly (see Figure 2). Unbeknownst to the participants, portions of well-known objects were suggested on the outside of the novel silhouette's contour. These well-known objects were not consciously perceived; myriad object priors (e.g., symmetry, small area, enclosure, and surroundedness) favored perceiving the inside as the object. Consequently, the object prior of familiar configuration that favored the outside lost the competition for object assignment and the outside was perceived as a shapeless ground (e.g., Peterson & Skow, 2008; Salvagio, Cacciamani, & Peterson, 2012; Trujillo, Allen, Schnyer, & Peterson, 2010). Peterson et al. (2012; Cacciamani et al., 2014) found that reaction times (RTs) to categorize the target word as denoting a natural or an artificial object were faster when the object suggested (but not consciously perceived) on the outside of the preceding novel silhouette object's contour was in the same category as the object denoted by the target word versus a different category (i.e., same category RTs were faster than different category RTs). These category priming effects obtained in the silhouette-target word paradigm showed that during the processes leading to figure assignment (aka object detection), category-level semantics are accessed for objects suggested on the side of a border ultimately determined to be a shapeless ground.¹ Cacciamani et al. (2014) ruled out an explanation of the category priming effects in terms of silhouette features. 

Figure 2

 

Schematic of two trial types used by Peterson et al. (2012). Participants' task was to categorize a target word shown at fixation as denoting an artificial or natural object. Prior to each target word, a silhouette was presented with a portion of a familiar object suggested on the groundside of the silhouette. The suggested objects could be from the same or different category as the target word.

The initial semantic category priming effect raised a number of questions. First, given that the silhouettes and the words were both presented at fixation, are such effects observed only when the silhouette appears in a location given priority as dictated by task set (i.e., the same location as the to-be-categorized target word)? If so, although the results would remain informative regarding what kind of processing can occur for regions of the visual field consciously experienced as shapeless, their generalizability would be limited and their implications for the normal course of object detection could be questioned. In order to understand the types of processing that generally occur during object detection, it is important to know whether the semantics of objects that might be perceived on both sides of borders are activated at multiple locations in the visual field or just at the target location. Lachter, Forster, and Ruthruff (2004) asked a similar question regarding word recognition and found that when the location of a lexical decision target was known and participants were encouraged to focus on that location, semantically related masked primes were effective only when they appeared in the attended location as opposed to an unattended location. Second, in the previous research using the silhouette-target word paradigm, the target word appeared at a fixed temporal interval after the trial began. Therefore, participants knew when the target would appear as well as where it would appear. Naccache, Blandin, and Dehaene (2002) observed numerical category priming effects from a prime shown shortly before a target in the same spatial location only when the target presentation time was predictable, not when it was unpredictable. Perhaps the temporal predictability of the target words in previous experiments using the silhouette-target word paradigm allowed participants to maximize category membership processing capacities in time as well as in space. If so, then semantic category priming would not be found when the time at which the target appears varies. It is important to address these questions in order to understand whether the semantic category priming effects reveal what commonly occurs during perceptual organization, or only what occurs when and where participants' categorization capacities are allocated. 

Assessing semantic category priming in the revised paradigm

In the present experiments, we investigated the generality of semantic category priming from objects suggested but not consciously perceived on the groundside of novel silhouettes. To do so, we presented the priming silhouettes in a location different from that of the target words (peripheral vs. foveal presentations, respectively). We also introduced a variable temporal interval before the onset of target word, thereby making it temporally unpredictable. Two peripheral priming silhouettes were used rather than one in order to balance the display (see Figure 3). If semantic category priming does not depend on the specific conditions used previously, then RTs to categorize the target word as natural or artificial should be faster when the object suggested on the outside of the preceding novel silhouette's contour is from the same category as the object denoted by the target word versus a different category (same-category vs. different-category trials, respectively). 

Figure 3

 

Trial types from Experiment 1. (A) Control trial. The silhouette-distractor display contains two control silhouettes. (B) An experimental trial. The silhouette presented in one location of the silhouette-distractor display suggests a portion of a familiar object on the groundside. Here, an umbrella is suggested on the groundside of the bottom distractor silhouette. The target word, scissors, denotes an artificial object and the suggested object (umbrella) is also artificial; hence, this is a same-category trial. The silhouettes and target words are not shown to scale in this figure (see Methods).

The current design also introduces a within-subjects control condition using fully novel silhouettes that do not suggest a familiar object on the groundside of their contours. Previous experiments with the silhouette-target word paradigm used a between-subjects control condition without any silhouettes prior to the target (Peterson et al., 2012). 

Summary of experiments and results

In Experiment 1, evidence of semantic category priming (same-category RTs faster than different-category RTs) was not found. Instead, categorization RTs were slower for words preceded by experimental distractor silhouettes (both same- and different-category) rather than control distractor silhouettes. In Experiment 2a and 2b, we considered three hypotheses for this slowing: For experimental versus control distractor silhouettes, (a) there is greater need to filter out distracting nontarget semantic activation arising from the familiar configurations suggested on the groundside, (b) there is more competition for figural status, and (c) despite attempts to equate the experimental and control silhouettes for stimulus features, perhaps unknown differences at the stimulus level make the experimental silhouettes more difficult to process than the control silhouettes. 

To test these hypotheses, we used the same distractor silhouette in the top and bottom locations in Experiment 2a, which would predict more (or at least the same) slowing according to the competition or featural difference hypotheses. On the other hand, if slowing was due to filtering out nontarget semantic information, presenting two experimental silhouettes may increase semantic processing such that a semantic category priming effect would emerge. Experiment 2a showed a semantic category priming effect (same-category RTs faster than different-category RTs), and suggested that the addition of a second experimental silhouette resulted in increased semantic activation, such that it could not be filtered to the degree that it was in Experiment 1. Moreover, Experiment 2a shows that spatial prime-target congruence and target onset predictability are not necessary for semantic category priming by objects suggested (but not consciously perceived) on the groundside of silhouettes. Experiment 2b further explored the possibility that competition and stimulus feature differences delay processing of distractors by changing the target word categorization task; here, abstract concept words were categorized as denoting positive or negative concepts. This task change should not change the level of competition or stimulus features in the experimental versus control silhouettes; hence, the pattern of results observed in Experiment 1 would be expected according to these hypotheses. With this task change, however, no RT differences between experimental and control trial RTs were obtained, supporting the hypothesis that the results of both Experiments 1 and 2a were due to semantic processing of the objects suggested on the groundside of the contours of the experimental silhouettes. The experiments reported here reveal that semantic activation by objects that lose the competition for figural status may occur commonly across the visual field in the course of perceptual organization. 

Experiment 1

In Experiment 1, participants categorized a target word at fixation as denoting either a natural or an artificial object. Before the target word was displayed, two distractor silhouettes were presented above and below fixation (see Figure 3). On control trials, both silhouettes were control silhouettes depicting novel objects on the inside of their borders and suggesting portions of novel objects on their groundsides as well. On experimental trials, again both silhouettes depicted novel objects on the inside of their borders, but only one of the distractor silhouettes was a control silhouette; the other was an experimental silhouette that suggested a portion of a familiar object on its groundside. On different-category experimental trials, the object suggested on the groundside of the experimental silhouette was from a category different from that of the object denoted by the target word (i.e., a silhouette with a natural object suggested on the groundside preceded a word denoting an artificial object and vice versa). On same-category experimental trials, the object suggested on the groundside of the experimental silhouette was from the same natural or artificial category as the object denoted by the target word (but was never the same object denoted by the word). On all trials, the time at which the distractor display and subsequent target word appeared was variable (from 1,000–1,400 ms after the trial began). 

All participants completed all three trial types (control, same-category, and different-category), randomly interspersed. If the results reported by Peterson et al. (2012; Cacciamani et al., 2014) generalize to the conditions tested here, categorization responses on same-category trials should be faster than on different-category trials. The introduction of a within-subject control condition will allow us to assess word categorization performance when distractors do not suggest portions of familiar objects on the groundside of their contours. 

Methods

Participants

The participants were 67 University of Arizona students, with reported normal or corrected-to-normal vision, who took part in the experiment for course credit.² The data from two participants were not analyzed due to system malfunctions. The data from nine participants who reported being aware of familiar objects on the groundside of the experimental silhouettes were not analyzed because our design requires that participants are unaware of the familiar objects suggested on the groundside of the experimental silhouettes (see Procedure section for how these participants were identified). These exclusions were all for reasons stated a priori; the data from these participants were never viewed. Exclusions based upon data analyses (outliers, poor accuracy) are reported in the Results section. The final number of participants after all exclusions was 48. 

Apparatus and stimuli

All experiments were run on an Intel®Core™ i7-4790 CPU running at 3.60 GHz computer with an AOC G2460PG 24 Class Nvidia G-Sync LCD gaming monitor (24.91° × 14.25°) running at 100 Hz. Experiments were run using MATLAB (2015a; MathWorks, Natick, MA) and the Psychophysics Toolbox extensions (Brainard, 1997; Kleiner et al., 2007). Participants viewed the screen from a distance of 120 cm with their chins in a chinrest. Responses were collected from an Adesso EasyTouch 220 mechanical numeric keypad connected through the ps/2 port. All buttons except the button on the bottom of the keypad and the two buttons on the extreme left and right of the middle row were removed. The button situated at the bottom of the keypad was used to navigate through instructional slides; this was the only functional button during instructions. The middle row buttons were response buttons; they were labeled “artificial” and “natural” by index cards placed on either side of the keypad, to indicate the response mapping. The other buttons were not scanned for responses. 

Silhouettes

A total of 112 black silhouettes were used. All of the experimental silhouettes were used in previous research (Cacciamani et al., 2014; Peterson et al., 2012; Peterson & Kim, 2001; Peterson & Skow, 2008; Salvagio et al., 2012; Sanguinetti et al., 2014; Sanguinetti & Peterson, 2016; Sanguinetti, Trujillo, Schnyer, Allen, & Peterson, 2016; Trujillo et al., 2010); some additional control silhouettes were created for this experiment. All silhouettes were created so that the object/figure priors of enclosure, small area, and symmetry around a vertical axis favored the inner black region as the figure and the surrounding white area as the ground. Consequently, the silhouettes were highly likely to be perceived as depicting closed, bounded, novel black objects. 

Control silhouettes (N = 80) were designed so that the left and right borders suggested novel objects on both the inside and the outside. Experimental silhouettes (N = 32) were designed so that their left and right borders suggested novel objects on the inside and portions of (the same) familiar meaningful object (natural or artificial) on the outside (groundside). Sample experimental and control silhouettes are shown in Figures 2 and 3. Half the experimental silhouettes suggested a portion of a natural object on their groundsides; the other half suggested a portion of an artificial object on their groundsides. Half of each category (natural or artificial) was used in same- and different-category trials. Contour length, horizontal width, and number of black pixels were equated between the three silhouette groups (control, same-category, different-category). The objects suggested on the groundside of the silhouettes (and their paired target word) are listed in Table A1. 

Target words

The target words denoted natural or artificial objects. Words presented on experimental trials always named an object different from the one suggested on the groundside of the border of the silhouette with which they were paired but were matched to the basic-level names of those objects in length and frequency assessed using SUBTL norms (Brysbaert & New, 2009; see also Peterson et al., 2012). Target words shown after half of the experimental-natural and experimental-artificial silhouettes denoted objects in the same category as the object suggested on the groundside of the silhouette distractor preceding it (same-category trials); target words shown after the other half of the experimental-natural and experimental-artificial silhouettes denoted different category objects (different-category trials). The target words used in the same- and different-category conditions were matched for word length and frequency. On control trials, target words denoted the basic-level objects suggested on the groundside of 16 of the experimental silhouettes; eight denoted natural objects and eight denoted artificial objects. These words matched the target words used on experimental trials in word length and frequency. The basic-level names of the objects suggested on the groundside of the borders of the other 16 experimental silhouettes served as target words on practice trials.³ The pairings between silhouettes and target words can be found in Table A1. Target words for control and practice trials can be found in Table A2. 

Procedure

Participants were tested individually. Upon arriving at the laboratory, they completed a consent form; both the research and consent form were approved by the Human Subjects Protection Program at the University of Arizona. Instructions for the experiment were presented on the monitor; participants followed along as the experimenter read them aloud. They were instructed to categorize a target word shown centrally as naming a natural or an artificial object by pressing the appropriately labeled button as quickly as possible while maintaining high accuracy. They were instructed to ignore the preceding peripheral images. 

Each trial (see Figure 3) began with a display containing a fixation cross (0.65° × 0.65°) in the center of the screen; this was shown for a variable duration between 1,000 and 1,400 ms (in 100 ms increments). After the fixation display, a silhouette-distractor display containing two silhouettes centered 2.95° above and below the center of the screen was presented for 80 ms⁴. The silhouettes were 3.77° high (H) and 2.42°–5.37° wide (W). On control trials (N = 16), neither distractor silhouette suggested a portion of a familiar object on its groundside. On experimental trials, one of the distractor silhouettes was an experimental silhouette (balanced between top and bottom positions); the object suggested on the groundside of the experimental silhouette was either from the same category (natural/artificial) as the object named by the target word (N = 16 trials) or from a different category (N = 16 trials). On experimental trials, novel silhouettes (N = 32) were paired with experimental silhouettes (N = 32) with contour length, number of black pixels, and width equated across silhouette type. The placement of the distractors was modeled after Wyble, Folk, and Potter (2013), who showed that distractor pictures presented at roughly the same eccentricity prior to a central target were processed for semantic category. 

After the silhouette-distractor display, a blank white display was presented for 30 ms, followed by a black target word presented for 50 ms in the center of the white screen backdrop (distractor-to-target stimulus onset interval = 110 ms). The target word (0.41°H × 0.57°–3.02°W, Times New Roman font) was equally likely to denote a natural or an artificial object. The luminance of the black fixation cross, silhouette distractors, and target words (RGB = 0, 0, 0) was 0.1 foot-lamberts (fL); the luminance of the white backdrop (RGB = 255, 255, 255) was 72.64 fL. A blank screen was presented for 500 ms after response (or time out) before the fixation cross for the next trial appeared. Categorization RTs were measured from the onset of the word. Responses slower than 3,000 ms were not recorded; instead, the trial was labeled a “timeout” trial. Participants received feedback on incorrect trials in the form of an auditory tone played for 500 ms. 

Participants completed 16 practice trials that contained only control silhouettes in the silhouette-distractor display (two different silhouettes per trial). The 16 silhouettes shown on practice trials were presented twice, once in the first eight trials and again in the second eight trials. Stimuli used in the practice trials were not repeated in the experimental trials. During instructions and practice trials, participants had an opportunity to ask questions regarding their task. 

Following the practice trials, participants completed 48 experimental trials (16 per each of three conditions—control, experimental same-category, and experimental different-category). Trial type was randomly intermixed. Each stimulus appeared once only. Halfway through the experiment, participants were instructed to take a break; the break lasted as long as participants wanted. The trial sequence began again when the participants pressed the 0 (zero) button at the bottom of the keypad. Participants used either their thumb or index finger on each hand to respond. The mapping of the two buttons onto artificial or natural responses was balanced across participants. 

The design of these experiments requires that observers are unaware of the objects suggested on the groundside of the borders of the silhouettes. To assess whether participants were aware of those objects, after the experimental trials, we showed them a sample black silhouette suggesting a portion of a familiar object on its groundside that was not used during the experiment. We first confirmed that participants saw the portion of the familiar object suggested on the groundside of the sample silhouette, and then asked if they had seen any portions of familiar objects on the groundside of the black silhouettes they viewed during the experiment. If they said they had, we excluded their data from analysis. It is important to note that participants were not required to identify any of the portions of objects they claimed to have perceived, they merely had to indicate that they had seen portions of familiar objects. Thus, our rejection criterion is liberal. It is important to use a liberal rejection criterion in order to be confident that we retain data only from participants who were unaware of the familiar objects suggested on the groundside of the experimental silhouettes. This is important because we are investigating whether the semantics of objects that might be present in the display but are not assigned figural status and are instead experienced as shapeless grounds can nevertheless influence word categorization. This postexperimental questionnaire procedure was very similar to one used previously (e.g., Cacciamani et al., 2014; Peterson et al., 2012; Sanguinetti et al., 2014; Trujillo et al., 2010). 

The first question probing awareness of the objects suggested on the groundside of the experimental silhouette borders assessed recall, which is less sensitive than recognition. Accordingly, following the first question, participants were given a list of object names and asked to circle the name of any object they had seen on the white side of the silhouettes during the experiment. All participants were asked to complete this recognition task, even those who reported they were unaware of any objects suggested on the groundside. They were told that all of the items on the list were presented as word targets, and to circle only the words that named objects/shapes they saw. The list included the names of all of the target words presented on the experimental and practice trials (64 items). Thus, half were the names of objects suggested on the groundside of experimental silhouettes during the experiment (recall these words appeared following control silhouettes, not experimental silhouettes); the other half were presented only as words during the experiment (on experimental trials) and thus constituted foils for the names of the objects suggested on the groundside of silhouettes. We set a criterion a priori that in order to be classified as aware by their performance on the word list, participants had to circle at least four items and of those items, at least 75% had to have been the names of objects suggested on the groundside of the experimental distractor silhouettes. This method ensured that they circled more ground object names than foils, which was our criterion to be certain that they were not just doing word recognition on the checklist.⁵ Only one additional participant, in Experiment 1, was classified as being aware of the objects presented on the groundside by performance on this checklist (after originally reporting not seeing any objects on the groundside of the silhouettes). 

Results

The data from three participants who failed to attain word categorization accuracy of 85% were discarded. For each participant, RTs on correct trials (93.36% of all trials) were analyzed. Individual participants' RTs on correct trials that were more than two standard deviations above or below their mean RT for a condition (control, same-category, different-category) were trimmed (7.25% of correct trials); these trimmed trials are referred to as within-participant outlier trials in subsequent results sections. Timeout trials (0.13% of all trials) were also excluded. Finally, the data from five participants whose mean RT on correct trials was more than two standard deviations above or below a between-participants condition mean were eliminated. All of these exclusion criteria were determined a priori. The data from 48 participants were submitted to analysis. 

Reaction times

A one-way ANOVA was performed on condition (control, same-category, different-category) to explore semantic category priming effects. Mauchly's test of sphericity indicated that the assumption of sphericity was violated, χ²(2) = 14.273, p < 0.01; therefore, Huynh-Feldt estimates (ϵ = 0.812) of sphericity were used to correct the degrees of freedom. The ANOVA showed a significant main effect of condition F(1.624, 76.349) = 10.583, p < 0.001, η² = 0.184 (see Figure 4). Planned pairwise comparisons did not produce evidence of semantic category priming: mean RTs on same- and different-category trials did not differ from each other, F(1, 47) = 1.969, p = 0.167. Interestingly, RTs on both same-category trials (634.28 ms), F(1, 47) = 18.277, p < 0.001, d = 0.307 and different-category trials (645.72 ms), F(1, 47) = 15.707, p < 0.001, d = 0.446 were significantly slower than RTs on control trials (612.17 ms), indicating a slowing of RTs on both types of experimental trials relative to control. 

Figure 4

 

Experiment 1 Mean RTs as a function of condition. Error bars represent standard error of the mean. *p < 0.001.

Given that only one experimental silhouette was shown on experimental trials, the question arises whether the top versus bottom location of the experimental silhouette made a difference. We examined that question by conducting a 2 (experimental distractor location: top, bottom) × 2 (condition: same-category, different-category) repeated measures ANOVA. The control condition could not be included in this analysis as there were control silhouettes in both locations. There was no main effect of experimental distractor location, F(1, 47) = 0.805, p = 0.374, and no interaction between location and condition, F(1, 47) = 2.767, p = 0.103; hence, the experimental silhouettes did not affect performance differentially when they were presented in the top versus the bottom location. 

Accuracy

Accuracy was near ceiling. A one-way ANOVA performed on condition showed no significant differences between conditions (control: 0.938; same-category: 0.944; different-category: 0.923), F(2, 94) = 1.685, p = 0.191. 

Discussion

In Experiment 1, there was not a significant difference between RTs on same- and different-category trials; thus, evidence of semantic category priming from the object suggested but not perceived on the groundside of the peripherally presented distractors was not found. Despite this, RTs were significantly slower on both types of experimental trials (same- and different-category) than on control trials. We consider three possible reasons for this finding. 

The first possibility is that the semantic category of the portions of familiar objects suggested on the groundside of the experimental silhouettes was processed, and it took time to filter out this semantic activation, creating a delay in categorizing the semantic category of the target word. Filtering of distractor information incurs a processing cost that can delay responses to a target (e.g., Kahneman, Triesman, & Burkell, 1983; Luck & Hillyard, 1994; Treisman, Kahneman, & Burkell, 1983). If the objects suggested on the groundside of the silhouettes activated their semantic category, this raises the question of why semantic category priming effects did not emerge. Perhaps due to their peripheral location (and therefore, the lower resolution of the experimental silhouette borders), semantic activation by the objects suggested on their groundsides was sufficient to require filtering, but insufficient to produce semantic-category priming. Participants also knew that the target word would be presented centrally, not peripherally. This factor may have also reduced semantic activation from the objects suggested on the groundside of the experimental distractors compared to previous research where the distractors and the target words were in the same central location. Finally, the presence of a second control silhouette on experimental trials may have reduced the semantic activation relative to trials on which only one experimental silhouette was presented as in the previous research. According to divisive normalization (e.g., Carandini & Heeger, 2012; Reynolds & Heeger, 2009), activity to a stimulus can be reduced by surrounding stimuli; thus, the addition of a second control silhouette may have lowered the response evoked by the experimental silhouette. 

The second possible reason why RTs may have been longer on both types of experimental trials than on control trials in Experiment 1 is that there is more competition for object status across the borders of experimental than control distractor silhouettes; this follows because familiar configuration is an object prior that competes with the myriad priors that favor assigning object status on the inside (see Cacciamani, Scalf, & Peterson, 2015; Peterson & Skow, 2008; Salvagio et al., 2012). Given that competition resolution takes time (Brooks & Palmer, 2011; Peterson & Enns, 2005; Peterson & Lampignano, 2003), it may take longer to assign object status to the inside of the experimental than the control silhouettes and this may delay word categorization responses on experimental trials. Note that, unlike the filtering explanation, this potential explanation does not necessarily entail semantic activation initiated by the objects suggested on the groundside of the borders of experimental silhouettes. Competition for object status can occur solely between potential shapes/objects without semantic activation, at least when those objects have no meaning (e.g., Peterson & Enns, 2005; Peterson & Lampignano, 2003). 

The third possible reason why RTs may have been longer on both types of experimental trials than on control trials in Experiment 1 is that, despite our efforts to match the experimental and control silhouettes on visual features (see Methods), featural differences that necessitate more processing for experimental silhouettes could remain and could lead to a delay before word categorization can proceed. 

Experiment 2

Experiment 2 was conducted to determine which of the potential explanations described above was likely responsible for the results of Experiment 1 while also increasing the possibility of observing semantic category priming from peripherally presented silhouettes. In two experiments (Experiment 2a and 2b), we changed the trial design so that on both control and experimental trials, identical silhouettes were presented in both the top and bottom distractor locations. On experimental trials, two experimental silhouettes suggesting the same objects on their groundsides should produce increased semantic activation on either a probability summation account, or a divisive normalization account (by increasing semantic activation through presentation of a second experimental silhouette and elimination of a control which may suppress the semantic activation; Reynolds & Heeger, 2009). 

In Experiment 2a, the participants' task remained the same: to categorize a target word shown centrally as denoting a natural object or an artificial object. If processing of the semantics of the objects suggested on the groundside of the experimental silhouettes in Experiment 1 was responsible for the slowed RTs on experimental trials, as in the filtering explanation, then we might observe semantic category priming in Experiment 2a instead of, or in addition to, slowing on experimental trials. This is because semantic category activation by the objects suggested repeatedly on the outside of two identical experimental silhouettes might be high enough to have a differential effect in the same- versus different-category conditions: that is faster same-category RTs than different-category RTs. Obtaining such a result in Experiment 2a would show that semantic activation occurs for objects suggested on the groundside of experimental silhouettes shown in the periphery at an unpredictable time when participants are set to respond to a centrally presented target word (albeit a boost in semantic activation is necessary to observe it). Such a result would be consistent with the filtering explanation for the results of Experiment 1. In contrast, if the slowed RTs on experimental trials in Experiment 1 were due only to increased shape-based competition for figural status from the groundsides of the experimental silhouettes (and not to concomitant semantic activation), or only to featural differences between the experimental and control silhouettes, then the same pattern of results should be observed in Experiment 2a with two identical experimental silhouettes as was observed in Experiment 1 with one experimental silhouette. 

In Experiment 2b, the participants' task was to categorize different words as denoting a positive [+] or a negative [−] abstract concept (e.g., Safe [+], Crisis [−]) rather than a natural or artificial object. The objects suggested but not perceived on the groundside of the experimental silhouette contours were not strongly valenced; hence, the semantic categories of the objects suggested on the groundside of the silhouettes (natural or artificial) were orthogonal to the target categories (positive or negative). Experiment 2b offers a different approach to distinguishing among the three possible explanations for the results of Experiment 1. If the results were due to stimulus differences between experimental and control silhouettes or to shape-based competition for object status that is independent of the semantic category of the object suggested on the groundside of the contours of the experimental silhouettes, the pattern of results found in Experiment 1 should be obtained in Experiment 2b because neither of these explanations attribute the slowing to access to semantic categories by the objects suggested on the groundside of the experimental silhouettes. In contrast, if the results of Experiment 1 were due to filtering of task-related semantic category activation arising from the groundside of the experimental distractor silhouettes rather than the target words, then the same pattern should not be observed in Experiment 2b: There is no need to filter the semantic activation from the objects suggested on the groundside of the experimental silhouettes because they are unrelated to the subject's positive/negative categorization task. 

Methods

Participants

The participants in Experiment 2a were 69 University of Arizona students with reported normal or corrected-to-normal vision who took part in the experiment for course credit. System malfunctions resulted in the loss of the data from two participants. The data from 10 participants who indicated during postexperimental questions that they saw some objects on the outside of the silhouette borders were not analyzed, as per our a priori criterion.⁶ Exclusions based upon data analyses (outliers, poor accuracy) are reported in the Results section. The final number of participants after all exclusions was 52. 

The participants in Experiment 2b were 62 University of Arizona students with reported normal or corrected-to normal vision, who took part in the experiment for course credit. Prior to data analysis, the data from one participant were removed because of system malfunctions; the data from nine participants were removed because they indicated during postexperimental questions that they saw familiar objects on the outside of the silhouette borders, and the data from one participant were removed because of inability to understand instructions.⁷ These reasons for removal were all stated a priori. Exclusions based upon data analyses (outliers, poor accuracy) are reported in the Results section. The final number of participants after all exclusions was 48. 

Stimuli, apparatus, and procedure

The stimuli and procedure used in Experiment 2a were the same as in Experiment 1, with one change. On both experimental and control trials, the silhouette-distractor displays now contained the same silhouette in both distractor locations. Because of this change, fewer control silhouettes were used in this experiment (N = 32; 16 used in practice trials, 16 used in control trials); the subset used was selected such that, at the mean level, they still matched the experimental silhouettes in contour length, number of black pixels, and width. Note that although fewer control silhouettes were used in Experiment 2 than in Experiment 1, they were carefully matched (contour length, horizontal width, and number of black pixels) to the experimental silhouettes and, as in Experiment 1, on experimental trials, individual silhouettes were not repeated across trials. As in Experiment 1, the number of control trials was equal to that of the same- and different-category trials. The apparatus and procedure were the same as in Experiment 1. 

Experiment 2b followed the methods and procedures of Experiment 2a with a few changes. Note that given the use of a positive/negative word categorization task, the experimental distractors could no longer be considered same- and different-category distractors. Instead, we refer to them as experimental-natural and experimental-artificial distractors. In the former type of trials, a natural object is suggested on the groundside of the experimental distractors; in the latter type of trials, an artificial object is suggested on the groundside of the experimental distractors. All three silhouette types (control, experimental-natural, experimental-artificial) were followed equally often by a positive or negative target word. 

The targets were 96 words denoting positive or negative abstract concepts culled from a corpus of 192 words selected from the Affective Norms for English Words (ANEW; Bradley & Lang, 1999) and the Evaluative Lexicon (Rocklage & Fazio, 2015). None of the words in this initial corpus denoted a concrete object. The rated valences of the meaning of all words were either above seven (positive) or below three (negative). The Experiment 2b target words were chosen based on the results of a pilot study in which participants viewed 192 words presented one at a time as “positive” or “negative” as quickly as possible; each participant viewed each word once in a random order. Words that were not categorized correctly by at least 85% of participants were discarded (N = 40). Next, six sets of 16 words (eight positive and eight negative) were formed. Sets were matched on word length, accuracy, and mean RT. From these six sets, three groups of 32 words were created, each containing two sets (see Table A3). The three groups of words were balanced across participants in the three experimental conditions (control, experimental-artificial, experimental-natural). 

One key difference between Experiment 2b and Experiments 1 and 2a was that more target words were available for Experiment 2b (96 vs. 48), which allowed twice as many trials in Experiment 2b. This increase in power allows increased sensitivity in detecting any potential effects of silhouette type in this task. The 96 trials in Experiment 2b were split into two blocks of 48 trials. Within each group of words, one set of 16 was presented in each of the three conditions in the first block of 48 trials and the other sets were presented in the second block; which set was presented first was counterbalanced across participants. Target words were presented only once. Every silhouette was presented once in the first block and then repeated, with a different target word, in the second block. Twelve practice trials preceded the experimental trials in Experiment 2b; as in the preceding experiments, words and silhouettes presented on control trials were not shown on experimental trials. The target words used on practice trials are shown in Table A4. 

Results

The same rejection criteria stated a priori were applied to Experiment 2 as to Experiment 1. 

Experiment 2a

The data from three participants were removed before analysis because their word-categorization accuracy was below 85%. The data from an additional two participants were eliminated because their mean RT in at least one condition was more than two standard deviations from the condition mean across participants. The analyses were based on 52 participants. Only RTs recorded on correct trials were included (94.59% of all trials). Within-participant outlier trials (7.50% of correct trials), and timeout trials (0.08% of all trials) were excluded from RT analyses. 

Experiment 2b

All participants had word-categorization accuracies above 85%. The data from three participants were eliminated because their mean RT in at least one condition was more than two standard deviations from the condition mean across participants. The analyses were based on 48 participants. Only RTs recorded on correct trials were included (96.48% of all trials). Within-participant outlier trials (4.84% of correct trials), and timeout trials (0.04% of all trials) were excluded from RT analyses. 

Reaction times

Experiment 2a

In Experiment 2a, where two identical distractor silhouettes preceded target words that were to be categorized as denoting natural or artificial objects, evidence of semantic category priming was observed. A one-way repeated measures ANOVA was performed on condition (control, same-category, different-category). Mauchly's test of sphericity indicated that, unlike Experiment 1, the assumption of sphericity had not been violated, χ²(2) = 0.900, p = 0.638; hence, degrees of freedom were not corrected. The ANOVA showed a significant main effect of condition, F(2, 102) = 12.381, p < 0.001, Display Formula\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\unicode[Times]{x1D6C2}}\)\(\def\bupbeta{\unicode[Times]{x1D6C3}}\)\(\def\bupgamma{\unicode[Times]{x1D6C4}}\)\(\def\bupdelta{\unicode[Times]{x1D6C5}}\)\(\def\bupepsilon{\unicode[Times]{x1D6C6}}\)\(\def\bupvarepsilon{\unicode[Times]{x1D6DC}}\)\(\def\bupzeta{\unicode[Times]{x1D6C7}}\)\(\def\bupeta{\unicode[Times]{x1D6C8}}\)\(\def\buptheta{\unicode[Times]{x1D6C9}}\)\(\def\bupiota{\unicode[Times]{x1D6CA}}\)\(\def\bupkappa{\unicode[Times]{x1D6CB}}\)\(\def\buplambda{\unicode[Times]{x1D6CC}}\)\(\def\bupmu{\unicode[Times]{x1D6CD}}\)\(\def\bupnu{\unicode[Times]{x1D6CE}}\)\(\def\bupxi{\unicode[Times]{x1D6CF}}\)\(\def\bupomicron{\unicode[Times]{x1D6D0}}\)\(\def\buppi{\unicode[Times]{x1D6D1}}\)\(\def\buprho{\unicode[Times]{x1D6D2}}\)\(\def\bupsigma{\unicode[Times]{x1D6D4}}\)\(\def\buptau{\unicode[Times]{x1D6D5}}\)\(\def\bupupsilon{\unicode[Times]{x1D6D6}}\)\(\def\bupphi{\unicode[Times]{x1D6D7}}\)\(\def\bupchi{\unicode[Times]{x1D6D8}}\)\(\def\buppsy{\unicode[Times]{x1D6D9}}\)\(\def\bupomega{\unicode[Times]{x1D6DA}}\)\(\def\bupvartheta{\unicode[Times]{x1D6DD}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bUpsilon{\bf{\Upsilon}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\(\def\iGamma{\unicode[Times]{x1D6E4}}\)\(\def\iDelta{\unicode[Times]{x1D6E5}}\)\(\def\iTheta{\unicode[Times]{x1D6E9}}\)\(\def\iLambda{\unicode[Times]{x1D6EC}}\)\(\def\iXi{\unicode[Times]{x1D6EF}}\)\(\def\iPi{\unicode[Times]{x1D6F1}}\)\(\def\iSigma{\unicode[Times]{x1D6F4}}\)\(\def\iUpsilon{\unicode[Times]{x1D6F6}}\)\(\def\iPhi{\unicode[Times]{x1D6F7}}\)\(\def\iPsi{\unicode[Times]{x1D6F9}}\)\(\def\iOmega{\unicode[Times]{x1D6FA}}\)\(\def\biGamma{\unicode[Times]{x1D71E}}\)\(\def\biDelta{\unicode[Times]{x1D71F}}\)\(\def\biTheta{\unicode[Times]{x1D723}}\)\(\def\biLambda{\unicode[Times]{x1D726}}\)\(\def\biXi{\unicode[Times]{x1D729}}\)\(\def\biPi{\unicode[Times]{x1D72B}}\)\(\def\biSigma{\unicode[Times]{x1D72E}}\)\(\def\biUpsilon{\unicode[Times]{x1D730}}\)\(\def\biPhi{\unicode[Times]{x1D731}}\)\(\def\biPsi{\unicode[Times]{x1D733}}\)\(\def\biOmega{\unicode[Times]{x1D734}}\)\(\eta \)² = 0.195. Planned comparisons showed that word categorization RTs were significantly faster on same-category trials (629.99 ms) than on different-category trials (650.24 ms), F(1, 51) = 11.245, p = 0.002, d = 0.262 (see Figure 5). This is the pattern expected if semantic category priming occurs under the conditions tested here, as it did in the conditions tested by Peterson et al. (2012; Cacciamani et al., 2014). Although RTs were significantly slower on different-category trials than on control trials (622.76 ms), F(1, 51) = 26.133, p < 0.001, d = 0.343, RTs on same-category trials and control trials were not significantly different, F(1, 51) = 1.585, p = 0.214. 

Figure 5

 

Reaction time results from Experiment 2a. Error bars represent standard error of the mean. *p < 0.003.

Experiment 2b

In Experiment 2b, where two identical distractor silhouettes preceded target words that were to be categorized as denoting positive or negative concepts, evidence of slowed RTs on experimental trials was not observed. An initial repeated measures two-way ANOVA, 2 (block: first, second) × 3 (condition: control, experimental-natural, experimental-artificial), was performed to explore any potential effects of block. Mauchly's test of sphericity indicated that the assumption of sphericity was violated, χ²(2) = 8.491, p = 0.014; therefore, Huynh-Feldt estimates (ϵ = 0.885) of sphericity were used to correct the degrees of freedom. No significant main effect of block was observed, F(1, 47) = 0.292, p = 0.591. Nor did block and condition interact significantly, F(1.770, 83.182) = 0.589, p = 0.537, indicating that the pattern of performance did not differ across blocks. As a result, subsequent analyses were collapsed across blocks. 

Positive/negative concept categorization RTs did not differ as a function of whether experimental-natural, experimental-artificial, or control distractor silhouettes preceded the target words. A one-way repeated measures ANOVA was performed on condition (control, experimental-natural, experimental-artificial) to explore potential competition effects, which would be evident as slower RTs on trials with experimental distractors rather than control distractors. No differences in performance between any of the three conditions were detected. The ANOVA did not yield significant results, F(2, 94) = 0.747, p = 0.477 (see Figure 6).⁸ Thus, no evidence in support of the competition interpretation of our previous results was obtained. 

Figure 6

 

Reaction time results from Experiment 2b. Error bars represent standard error of the condition mean.

Between-experiment analysis

A one-way between-experiments ANOVA was run on control RTs between Experiments 1, 2a, and 2b to compare baseline performance in the three experiments. The ANOVA showed a significant effect of experiment, F(2, 145) = 6.692, p = 0.002. Pairwise comparisons revealed that control trial RTs were faster in Experiment 2b (537.37 ms) than in Experiment 1 (612.76 ms), F(1, 94) = 8.387, p = 0.005, and Experiment 2a (622.76 ms), F(1, 98) = 11.807, p = 0.001. Control trial RTs in Experiments 1 and 2a did not differ, F(1, 98) = 0.518, p = 0.473. 

Accuracy

Experiment 2a

Accuracy was again near ceiling (94.77%). A one-way ANOVA performed on condition did not show any significant effects (control: 0.946; same-category: 0.957; different-category: 0.937), F(2, 102) = 1.713, p = 0.186. 

Experiment 2b

Accuracy was again near ceiling (96.53%). A one-way ANOVA showed a significant main effect of condition (control: 0.956; experimental-natural: 0.971; experimental-artificial: 0.970), F(2, 94) = 3.162, p = 0.047. Post hoc comparisons showed that performance was significantly more accurate on experimental-artificial trials than on control trials, F(1, 47) = 4.840, p = 0.033, d = 0.361. This difference was not statistically significant for experimental-natural trials, however, F(1, 47) = 4.017, p = 0.051. It is important to note that higher accuracy on experimental trials is not expected if processing of the experimental silhouettes interfered with word categorization responses; instead, the opposite pattern would be expected. This finding is anomalous, so we do not pursue it further. 

Discussion

Experiment 2a replicated findings from previous studies using the paired silhouette-target word paradigm (Cacciamani et al., 2014; Peterson et al., 2012): Word categorization responses were faster on same-category than different-category trials. This pattern of results indicates that in the course of object assignment, category-level semantics are activated for objects that are suggested but not consciously perceived on the groundside of the contours of a perceived object (i.e., the distractor silhouettes). At least when participants are set to semantically categorize a target word, this semantic access results in faster categorization responses when the word denotes an object in the same category compared to a different category. These results are in accordance with experiments that demonstrated that semantic category priming from pictures can affect responses to words (Carr, McCauley, Sperber, & Parmelee, 1982), and extend that research by showing that the pictorial prime can be an object suggested on the groundside of perceived objects. Presenting two experimental silhouettes on each trial in Experiment 2a yielded sufficient semantic activation to allow us to observe semantic category priming effects. The results of Experiment 2a show that in order to observe semantic category priming from objects suggested but not consciously perceived on the groundside of silhouette borders, it is not necessary to present the silhouette and the word in the same location, nor is it necessary to present the word at a fixed time after the trial begins. Hence, the semantic category priming effects observed by Peterson et al. (2012; Cacciamani et al., 2014) are generalizable in time and spatial location. 

The results of Experiment 1 (slower RTs on experimental compared to control trials) were hypothesized to arise either due to filtering of semantic information activated by the objects suggested on the groundside of the experimental silhouettes, an increased level of competition for figural status by the experimental silhouettes, or featural differences between the experimental and control silhouettes. Experiment 2a did not reveal the same pattern of results. Although different-category RTs were slower than control RTs, same-category RTs were not slower than control RTs (and were faster than on different-category experimental trials). If the slowed RTs on experimental trials in Experiment 1 were due to increased competition or to featural differences, the addition of a second experimental silhouette should have produced the same pattern of results to at least the same degree. This was not the case, providing some evidence against those explanations for the pattern of results observed in Experiment 1. Moreover, note that the size of the difference between different-category RTs and control RTs was not larger in Experiment 2 (d = 0.343) than in Experiment 1 (d = 0.446), showing that the addition of a second experimental silhouette did not yield slower word categorization on experimental trials in Experiment 2a than in Experiment 1. 

The use of control silhouettes shows that, at least under the conditions tested here, the semantic category priming effect is driven by slowing on different-category trials rather than speeding on same-category trials. Perhaps the need to filter semantic activation from the experimental distractors prevented the observation of speeding on same-category trials compared to control trials. Future experiments will have to address this question, but the absence of speeding on same-category trials compared to control trials indicates that the semantic category priming does not occur at the level of response: were that the case, both response speeding on same-category trials and response slowing on different-category trials should have been observed. 

In Experiment 2b, where the semantic category of the objects suggested on the groundside of the experimental distractor silhouettes was not related to the positive/negative categories into which the target words were to be categorized, experimental trial RTs were not slower than control trial RTs. This finding shows that, if any featural differences do remain between experimental and control silhouettes despite our attempts to equate them, they are not sufficient to produce the pattern of results obtained in Experiment 1. The results of Experiment 2b also show that greater competition for figural status per se, without activation of the semantics of the objects suggested on the groundside of experimental versus control silhouettes is not sufficient to produce the pattern of results obtained in Experiment 1. To reject the competition explanation, one must assume that competition for figural status occurs even when it is not task relevant. Cacciamani et al. (2015) provide evidence that competition for figural status occurs even when task irrelevant (see General discussion for further explication). 

RTs were faster on control trials in Experiment 2b than in Experiments 1 and 2a. This was expected based on pilot experiments that measured baseline RTs for target words in all three experiments when no silhouettes were present: Peterson et al. (2012; experiment 1a) observed mean natural/artificial categorization RTs of approximately 670 ms for the target words used in Experiments 1 and 2a under pilot conditions similar to those used for the positively and negatively valenced words. We observed mean categorization times of approximately 630 ms for the positive/negative words used in Experiment 2b. The between-experiments difference observed in control RTs in the present study (∼40–50 ms) is the same order of magnitude as the difference observed in the pilot studies (∼50 ms). This is important because it indicates that the difference in baseline performance between Experiment 2b and 1 and 2a is not due to differential effects of distractor processing. In addition, with faster RTs in Experiment 2b, there is ample room to observe a slowing of RTs on experimental trials, but this did not occur. 

General discussion

The goal of the current study was to investigate the generalizability of semantic category priming effects from portions of familiar objects that are suggested but not perceived on the groundside of silhouette contours. It is important to do so in order to determine whether semantic activation commonly occurs during perceptual organization or whether it occurs only under a restricted set of conditions. Experimental silhouettes are silhouettes that are perceived as novel, enclosed objects, but suggest a portion of a familiar object on the outside (groundside) of their contours. Previous studies using the silhouette-target word paradigm (Cacciamani et al., 2014; Peterson et al., 2012) found semantic category priming effects (same-category RTs faster than different-category RTs) when an experimental silhouette prime was presented in the same spatial location as a word target, and when the time at which the word appeared after trial onset was predictable. In order to test the generalizability of semantic category priming from objects suggested on the groundside of a silhouette border, we presented the silhouettes in a peripheral location from the target, and made its onset time unpredictable. We also included control trials as a within-subject factor. 

In Experiment 1, two different novel distractors were presented in peripheral locations. On experimental trials, one of the novel distractors was an experimental silhouette and the other was a control silhouette that suggested a novel object on both the inside and the outside of its borders. On control trials, both novel distractors were control silhouettes. The participants' task was to categorize a centrally presented target word as denoting either a natural or an artificial object. Evidence of semantic category priming was not found. Instead, RTs were slower on both same- and different-category experimental trials than on control trials, and were not significantly different from each other. We proposed three different explanations for this pattern of results: (a) greater need to filter out distracting nontarget semantic activation arising from the familiar configurations suggested on the groundside of experimental silhouettes, (b) more competition for figural status by experimental silhouettes, and (c) despite attempts to equate the experimental and control silhouettes for stimulus features, it could still be the case that unknown differences at the stimulus level make the experimental silhouettes more difficult to process than the control silhouettes. 

In order to adjudicate between these hypotheses, and to increase the possibility of observing semantic category priming, we presented identical silhouettes in each location in Experiment 2a. If the slowing of RTs on experimental trials in Experiment 1 was due to competition or stimulus featural differences, then the same pattern of results is expected with two identical silhouettes. This was not observed. Instead, a significant semantic category priming effect emerged (faster same-category trial RTs than different-category trial RTs). Thus, in Experiment 2a, we obtained evidence of semantic category priming, demonstrating that semantic category priming effects in the silhouette-target word paradigm are not dependent upon the prime being presented in the same spatial location as the target word or a predictable target onset. Experiment 2a supported the hypothesis that semantic activation from the objects suggested on the groundside of the experimental silhouettes was not fully filtered out, and that when two of the same experimental silhouettes were presented, there was sufficient semantic access for a semantic category priming effect to emerge. 

Experiment 2b investigated the three potential explanations for the results of Experiment 1 further by changing the word categorization task to categorizing a target word as denoting a positive or negative abstract concept. With this change, no RT differences were observed between control and experimental trials. These results show that neither feature differences between the experimental and control silhouettes nor differences in competition that do not entail semantic activation can produce the pattern of results obtained in Experiment 1. Thus, taken together, the results of Experiments 2a and 2b favor the filtering explanation for the results of Experiment 1. The filtering explanation entails that semantic category was activated for the objects suggested on the groundside of the single experimental distractor silhouette shown in Experiment 1 to a degree sufficient to require filtering but not to a degree sufficient to produce category-specific priming. 

We note that Experiment 2a also shows that semantic activation for objects suggested but not perceived on the groundside of experimental silhouettes is not limited to one silhouette, but can be processed simultaneously or in parallel for at least two silhouettes. This is because the addition of a second peripheral experimental silhouette in Experiment 2a allowed us to observe semantic category priming whereas the single peripheral experimental silhouette in Experiment 1 was insufficient. It has not previously been shown that semantic access can occur for objects suggested but not perceived on the groundside of more than a single object. These results imply that, at least during the 80 ms the peripheral distractors were presented, perceptual organization and concomitant activations are not restricted to one location in the visual field. This is a new finding with respect to isolated silhouettes, but it is consistent with evidence that context affects figure–ground assignment (Lass, Bennet, & Sekuler, 2016; Peterson & Salvagio, 2008; Rauschenberger, Peterson, Mosca, & Bruno, 2004). 

We have attributed the semantic category priming effects to semantic activation by the objects suggested but not perceived on the groundside of the experimental distractor silhouette borders, and to response priming from the silhouette borders themselves. The borders of silhouettes suggesting natural versus artificial objects on their groundsides necessarily differ because borders of natural objects are more curvilinear than borders of artificial objects (e.g., Kurbat, 1997; Zusne, 1975; cf. Cacciamani et al., 2014). We do not believe this “borders only” interpretation can account for our effects because the results of Experiment 1, although consistent with the hypothesis that semantic activation has occurred, show no evidence of a border-dependent pattern in the RTs. Such a pattern should have been evident if the results simply reflected responses to the borders of the silhouettes. In addition, the borders-only hypothesis would predict that RTs should be faster on same-category trials than on control trials in Experiment 2a, but they were not. The overall pattern of our results, along with those of Cacciamani et al. (2014) argue against the borders only hypothesis. 

Open questions

An open question is whether attention capture by the experimental silhouettes contributed to the pattern of results obtained in Experiment 1. This remains a possible explanation in addition to filtering, but at this point, we cannot distinguish between attention capture and filtering. We do not view capture and filtering as exclusive processes (cf. Keehn, Westerfield, & Townsend, 2018), but filtering need not entail attentional capture to the location of the experimental silhouette(s). In future experiments, we plan to test whether attentional capture occurs using eye movements to assess when capture occurs (cf. Beck & Vickery, 2018), and thereby, trials on which semantic activation was sufficient to produce capture. 

Another open question is why Lachter et al. (2004) did not observe effects of semantically related masked primes when they appeared in an unattended, nontarget location, as opposed to an attended, target location, whereas we observed semantic category priming from objects suggested but not perceived on the groundside of distractor silhouettes shown in nontarget locations. There are many differences between the Lachter et al. (2004) studies and ours, including the task required of the participants (e.g., lexical decision vs. word categorization), and the method of rendering the prime unconscious (masking vs. ground assignment). Recent research (Kiefer, 2012; Kiefer & Martens, 2010; Martens, Ansorge, & Kiefer, 2011) shows that task set plays a large role in governing when effects of masked primes are evident in RTs. This principle is similar to what we observed in our experiments. We point out that RTs are only not the only way to assess what kind of processing occurred for primes and they are far from the most sensitive. 

The question of whether semantics activation by objects suggested but not consciously perceived on the groundside of the borders of the experimental distractor silhouettes occurs regardless of task set is relevant to this set of experiments and to conclusions regarding perceptual organization in general. The results of experiments conducted by Cacciamani et al. (2015) suggest the answer is “yes.” They found that when the subjects' task was to detect one of two lowercase letters in a rapid serial visual presentation (RSVP) stream of keyboard characters at fixation, neural activity on the groundside of peripherally presented silhouettes was reduced for experimental compared to control silhouettes. This is the pattern they predicted if the familiar configuration prior was activated, competed for figural status, lost the competition, and was suppressed, despite the fact that the peripherally presented silhouettes were conceptually irrelevant to the RSVP task. We point out, however, that, as we argued previously in considering potential explanations for the pattern of results obtained in Experiment 1, it is not clear whether semantic activation is entailed by evidence of cross-border competition for figural status. 

Conclusion

The current study demonstrated that even when participants are prepared to respond to a target at fixation, semantic category is accessed by objects that are suggested but not consciously perceived on the groundside of distractor silhouettes located in the periphery and appearing at unpredictable times. These results imply that semantic activation by objects that are suggested but not consciously perceived on the groundside of items in the visual field occurs more commonly in the course of perceptual organization than conventional theory implies. 

Acknowledgments

MAP acknowledges support from ONR N00014-14-1-067. CSF presented these results at the 2016 Meeting of the Vision Sciences Society VSS 2016, supported by ONR N00014-14-1-067. We thank Steven Most, Eve Isham, and the members of the Visual Perception and Cognition Laboratory at the University of Arizona for their helpful comments on previous drafts of the manuscript. 

Commercial relationships: none. 

Corresponding author: Colin S. Flowers. 

Email: . 

Address: Department of Psychology, University of Arizona, Tucson, Arizona, USA. 

Footnotes

¹  By inference, category-level semantics are accessed for the side of the contour perceived as the figure/object as well, but this was not tested because it would be difficult to know whether semantic category access occurred before or after figure assignment.

Footnotes

²  Based on a procedure stated a priori, we did not analyze the data from eight additional participants whose responses on a post-experiment questionnaire indicated they had learned/started using English as their primary language after age eight. For the purposes of this study, they were identified as non-fluent in English. We did not exclude such participants from taking part in the experiment because had we done so, they would not have had the same opportunity as other students to fulfill the requirements for their introductory psychology course. This is the recommended procedure for all experimenters in the Psychology Department at the University of Arizona.

Footnotes

³  Peterson et al. (2012) showed that baseline categorization time—time to categorize the words when they were not preceded by a silhouette—was equal for the words used in the three conditions tested here.

Footnotes

⁴  We note that although the presentation time of the silhouettes is variable, their appearance can predict when the target word will appear. This is constant across both control and experimental trials.

Footnotes

⁵  Participants who circled fewer than four words or did not reach 75% accuracy for identifying objects suggested on the groundside of the experimental silhouette borders were included in analyses. Of these participants, a majority did not circle any items on the checklist (Experiment 1: 75%, Experiment 2: 82.4%, Experiment 3: 88.3%).

Footnotes

⁶  Based on an a priori criterion, the data from an additional seven participants were not analyzed because they were not fluent in English (see Footnote 2)

Footnotes

⁷  Based on an a priori criterion, the data from an additional 20 participants were not analyzed because they were not fluent in English (see Footnote 2).

Footnotes

⁸  A one-way ANOVA (control, experimental-natural, experimental-artificial) run on only block one data to mirror analyses of Experiments 1 and 2 showed the same results

Appendix

Table A1

 

Target words used in experiments 1 and 2a with experimental silhouette pairings. Italicized items represent artificial objects suggested on the groundside and their paired word.

 

Table A2

 

Target words used on control and practice trials of Experiments 1 and 2a. Italicized items represent artificial target words.

 

Table A3

 

Target words used in Experiment 2b.

 

Table A4

 

Target words used in practice trials for Experiment 2b. Italicized items represent positive valenced target words.