Atten Percept Psychophys DOI 10.3758/s13414-014-0623-5
The hard-won benefits of familiarity in visual search: naturally familiar brand logos are found faster Xiaoyan Angela Qin & Wilma Koutstaal & Stephen A. Engel
# Psychonomic Society, Inc. 2014
Abstract Familiar items are found faster than unfamiliar ones in visual search tasks. This effect has important implications for cognitive theory, because it may reveal how mental representations of commonly encountered items are changed by experience to optimize performance. It remains unknown, however, whether everyday items with moderate levels of exposure would show benefits in visual search, and if so, what kind of experience would be required to produce them. Here, we tested whether familiar product logos were searched for faster than unfamiliar ones, and also familiarized subjects with previously unfamiliar logos. Subjects searched for preexperimentally familiar and unfamiliar logos, half of which were familiarized in the laboratory, amongst other, unfamiliar distractor logos. In three experiments, we used an N-back-like familiarization task, and in four others we used a task that asked detailed questions about the perceptual aspects of the logos. The number of familiarization exposures ranged from 30 to 84 per logo across experiments, with two experiments involving across-day familiarization. Preexperimentally familiar target logos were searched for faster than were unfamiliar, nonfamiliarized logos, by 8 % on average. This difference was reliable in all seven experiments. However, familiarization had little or no effect on search speeds; its average effect was to improve search times by 0.7 %, and its effect was significant in only one of the seven experiments. If priming, mere exposure, episodic memory, or relatively modest familiarity were responsible for familiarity’s effects on search, then performance should have improved following familiarization. Our results suggest that the search-related advantage of familiar logos does not develop easily or rapidly.
X. A. Qin : W. Koutstaal : S. A. Engel (*) Department of Psychology, University of Minnesota, Minneapolis, MN 55455, USA e-mail: [email protected]
Keywords Visual search . Familiarity . Brand logos . Semantic representations . Perceptual learning . Practice effects
The need for visual search arises frequently in everyday life; we routinely scan shopping shelves for preferred products, articles for specific references, and web pages for preferred content, to pick just a few examples. Because of these applications and more specialized ones in medical, military, and security domains, the factors that influence search efficiency have received intense study. One factor that affects search speed is the amount of experience observers have had with the items used in the task. Target familiarity, distractor familiarity, and familiarity differences between targets and distractors all decrease search times (e.g., Flowers & Lohr, 1985; Lubow & Kaplan, 1997; Malinowski & Hübner, 2001; Rauschenberger & Chu, 2006; Richards & Reicher, 1978; Shen & Reingold, 2001; Wang, Cavanagh, & Green, 1994). Understanding how familiarity speeds search has important implications for cognitive theory, because it may reveal how mental representations of commonly encountered items are changed by experience in order to optimize performance. Familiarity also plays an important role in visual search in applied settings, such as baggage scanning by security personnel and supermarket product choice by consumers. The goal of this study was to understand the everyday familiarity that people have for commonly encountered objects, such as a favorite article of clothing, brand of soda, or model of car. Most past research examining familiarity’s effects on visual search had used stimuli from domains in which subjects possessed extensive visual expertise, such as letters, digits, faces, and symbols. Thus, it remained an open question whether subjects would show advantages for stimuli with which they had a more moderate amount of experience.
Atten Percept Psychophys
We used brand logos to examine this level of familiarity. Logos provide a set of diverse, moderately complex items, which contain a natural manipulation of real-world familiarity; depending on marketing budgets and strategies, we are exposed to some brand logos much more often than others. In general, however, we are less expert at viewing logos than at viewing the highly overlearned categories of letters, digits, symbols, and faces that had been used in past research. A final benefit of studying logos was that the results might not only shed light on how natural experience changes the representation of everyday objects, but also inform marketing research. It also remains unknown precisely how much and what kind of exposure people require in order to locate familiar items more quickly. One kind of exposure that has been shown definitively to improve search times is practice with the items while performing the search task itself (e.g., Baluch & Itti, 2010; Le Dantec, Melton, & Seitz, 2012; Sireteanu & Rettenbach, 1995, 2000; Williams, Henderson, & Zacks, 2005; but see Tong & Nakayama, 1999, for an exception). These experiments confirmed that subjects can learn to improve performance on particular items in a particular task; such specificity is a hallmark of much of laboratory-based perceptual learning (for reviews, see Lu, Yu, Watanabe, Sagi, & Levi, 2010; Sagi, 2011; Sasaki, Nanez, & Watanabe, 2010). Real-world familiarity of the kind that logos and other everyday items possess, however, is not acquired within a single task context. Thus, while there is little doubt that practicing visual search for logos would improve visual search for logos, such a demonstration might be unenlightening about the nature of everyday familiarity. One study (Mruczek & Sheinberg, 2005) tested whether exposing subjects to individually presented items would produce the benefits of familiarity, but even here, targets were presented in a search task (in a zero-distractor condition). What kind of everyday exposure to items might improve visual search? It is possible that relatively brief, uninformative encounters with an item could produce large familiarity effects. For example, items that are “primed” by prior viewings, items that are very recently encountered, and items to which one has been “merely exposed” all show advantages in cognitive tasks, despite small numbers of exposures (e.g., Alter & Oppenheimer, 2009; Henson, 2003; Stafford & Grimes, 2012; Zajonc, 2001). These results suggest that seeing a logo just a few times before performing visual search could improve performance. On the other hand, larger amounts or richer kinds of experience might be required to produce familiarity effects. Particularly for logos, familiarity effects may depend on factors such as observers’ gaining affective associations due to advertising, identifying the logos in many highly diverse visual contexts, or simply having experienced a very large number of exposures. If such factors are important, laboratory exposure would be unlikely to produce familiarity effects for logos.
In the present experiments, we explored the strength of the target familiarity effect for brand logos and attempted to generate it using several different, non-search-exposure procedures for initially unfamiliar logos. We used two different tasks to familiarize observers with the logos: an N-back-like task, and a set of questions about detailed perceptual aspects of the logos. In all experiments, the testing phase consisted of a visual search task in which familiarized and nonfamiliarized logos served as the targets and a heterogeneous set of nonfamiliarized unfamiliar logos acted as distractors. To anticipate our results, preexperimentally familiar logos manifested a robust advantage over unfamiliar logos, in both search speed and search efficiency. However, familiarization in the laboratory produced little, and usually no, benefits for visual search, despite the use of procedures that varied the amount of exposure, the amount of attention given to the exposure, and the amount of time given for consolidation of the exposure’s effects.
Experiment 1 In our first experiment, we tested whether familiar logos are searched for faster than unfamiliar ones. Past research has tested whether familiarity can affect search times for items in overlearned domains, such as letters and faces. Familiarity generally speeds search times, though when it is attached to distractors it can also slow responses (McGugin, McKeeff, Tong, & Gauthier, 2011). Only one study, to our knowledge, has examined search and familiarity in domains in which subjects are less expert, and found that familiar targets can be found faster than unfamiliar ones (Mruczek & Sheinberg, 2005). However, in this study familiarity was gained through practice with the visual search task itself, and it is well known that practice in search can speed search (Le Dantec et al., 2012; Mruczek & Sheinberg, 2005; Sireteanu & Rettenbach, 1995, 2000). Thus one contribution of Experiment 1 is to test whether the diverse but modest familiarity with brand logos acquired in everyday life can speed visual search. In past work, displays were also often simplified by placing a target in a homogeneous field of distractors. Although this type of experimental design offers good control over the physical parameters of the displays, the observed effects may not generalize well to real-world search behaviors (e.g., Hershler & Hochstein, 2009; Mruczek & Sheinberg, 2005). Using more complex everyday stimuli and more heterogeneous displays addresses this limitation (Mruczek & Sheinberg, 2005; Tong & Nakayama, 1999), and we adopted that approach here. Experiment 1 also tested whether familiarization with logos through simple visual exposure would be able to improve search times. Many theories of the benefits of familiarity, including those based upon priming, recency of exposure, or
Atten Percept Psychophys
“mere” exposure, could predict that such familiarization should speed search. For example, items to which one has been exposed just a few times in the recent past, referred to as “primed,” are processed more rapidly (e.g., Maccotta & Buckner, 2004), which could lead to an advantage in visual search (cf. object-based priming within visual search tasks, as in Ásgeirsson & Kristjánsson, 2011; Kristjánsson & Campana, 2010). Alternatively, items to which one has been exposed a small number of times may gain not only perceptual fluency, but also positive affect (Stafford & Grimes, 2012; Zajonc, 2001). Preference for recently exposed logos could also possibly speed search. To familiarize subjects with a set of logos, they viewed them repeatedly, while performing an N-back-like task. Next, subjects performed a visual search task using both these logos and a second set of logos. Both sets contained preexperimentally familiar as well as preexperimentally unfamiliar logos. Thus subjects searched for four different kinds of logos obtained by crossing two factors: preexperimentally familiar and unfamiliar logos that were familiarized in the laboratory and preexperimentally familliar and unfamiliar logos that were not familiarized.
“unfamiliar” (rating less than 2.58) or “familiar” (rating higher than 4.55). To verify the reliability of ratings and the validity of these categories, we computed for the familiar and unfamiliar logos used in our experiments, the average number of logos that would be misclassified by our norming population. We did this by simply looking up the ratings for those logos for each member of our norming group and comparing them to the cutoffs. Of the 40 logos used in our experiments, on average fewer than four would be misclassified for individual subjects by our cutoffs. These 40 logos were further divided into two sets of ten familiar and two sets of ten unfamiliar logos; one set of each familiarity category was used as the experimentally familiarized set, whereas the other set was not familiarized. The familiarized and nonfamiliarized sets were counterbalanced across subjects. In addition, we approximately balanced the color, shape, and the number of component letters (if any) for the four sets of ten target logos. The mean familiarity ratings were 1.62 and 1.68 for the two unfamiliar sets (SDs = 0.57 and 0.50, respectively), and 4.79 and 4.75 for the two familiar sets (SDs = 0.10 and 0.13, respectively). A different set of 120 unfamiliar logos (mean familiarity rating = 1.73, SD = 0.57) was used as distractors in the visual search task.
Method
Procedure
Subjects
Subjects were tested individually. After providing informed consent, they completed a demographic inquiry form asking about their age, gender, color vision, visual acuity, and number of years residing in the United States. Written instructions were provided before each task.
A group of 36 subjects were recruited from the University of Minnesota (nine males, 27 females; mean age = 20.5 years). All of the subjects in all experiments reported in this article reported that they had lived in the United States for at least 10 years, had normal or corrected-tonormal vision, and received course credit or monetary compensation for their participation. All experimental sessions lasted about 60 min, and at the end of the session subjects were debriefed and thanked. Stimuli The logo images used in all of the experiments reported here were initially found on the Internet. All logos were edited in Photoshop CS2 to remove trademarks and were resized to ensure equivalent image size and resolution. Each logo was scaled to fit within a 60×48 pixel rectangle. All logos were normed for familiarity using ratings by subjects (N = 92) from the same general population as the main experimental subjects. The logos were rated on a 5-point scale, from 1, not familiar at all, to 5, extremely familiar. We used these norming data to select twenty familiar and twenty unfamiliar logos as targets in our experiments. The norming data were very reliable, which allowed us to use simple cutoffs in the average ratings to classify logos as either
Familiarization task The subjects were asked to detect repeated presentations of the logos in a stream of singly presented logos (see Fig. 1a and b). We made no mention of subsequent testing or visual search; that is, the experimental familiarization or encoding phases in this experiment and in all subsequent experiments were incidental. The familiar and unfamiliar sets of logos were each presented in 40 separate alternating blocks. One half of the subjects received experimental familiarization of the familiar logos in odd-numbered blocks, and experimental familiarization of the unfamiliar logos in evennumbered blocks; the other half of the subjects received the opposite pattern. Each experimental familiarization block consisted of 17 trials. Given that ten logos were in each set, this resulted in seven repeats in a block. The repeated items could occur anywhere within a block; that is, varying numbers of interleaved items could occur between the first and second presentations of a logo within a block. Subjects were instructed to respond to any repeated logos within a given block; since the logos also repeated across blocks, subjects were asked to “restart” their repetition count for each block.
Atten Percept Psychophys
a
b
Fig. 1 Familiarization phase in Experiments 1–2b. (a) Experimental familiarization for familiar logos. (b) Experimental familiarization for unfamiliar logos. In Experiment 1, subjects were experimentally familiarized with both familiar and unfamiliar logos; in Experiments 2a and 2b,
only unfamiliar logos were familiarized. In Experiment 2b, the responses were “1” or “2” instead of “yes” or “no,” to indicate the first or second occurrence of a logo
In each trial, a logo appeared in the center of a white screen for 1.5 s. Subjects were situated at about 60 cm from the monitor. A chinrest was used to control viewing distance and to prevent unnecessary head movements. Subjects pressed one of two keys (labeled “yes” or “no”) to indicate whether or not the currently displayed logo was a (within-block) repeat. Subjects completed 12 practice trials before starting the actual task. By the end of the 30-min familiarization phase, each subject was exposed to each logo approximately 30 times.
rested their chins on a chinrest. On each trial, a black fixation cross appeared in the center of the screen to alert subjects that a search array was about to be presented. After 1 s, a “cue” logo appeared in the position of the fixation cross, indicating the to-be-searched-for target in the subsequently presented array. After another second, three, six, or nine logos appeared in randomly selected positions in an invisible 5×5 grid. The central position of this grid was always occupied by the cued logo, leaving 24 possible locations for the search array. The logo positions were jittered slightly within the boxes that comprised the grid, in order to avoid perfect or repeated alignments of the logo positions. Given that the unfamiliar target logos were not well known to subjects beforehand, these logos may have been disadvantaged in terms of subjects’ ability to keep them in mind when they were targets during the search task. To minimize any working memory effects on search performance for logos of differing familiarities, the cue logo remained on the screen throughout the search. To differentiate the cue from the to-besearched-for target, a black frame outlined the cue at the time when the search array appeared. Subjects were asked to press the space bar as soon as they thought they had located the designated target logo (Fig. 2). Accuracy and speed were equally emphasized. To assess visual search accuracy, after the subject pressed the space bar, a letter array containing 24 letters (A–X) replaced the search array, with each letter corresponding to one of the possible locations of the logos in the search array. Subjects were required to indicate where the target logo had appeared by pressing the letter corresponding to its position. After subjects made this location judgment (i.e., had indicated the letter that they thought corresponded to the location of the target in the prior search array), the next trial started automatically with the 1-s fixation. Three practice trials were provided
Search task To assess the effects of preexperimental familiarity and the laboratory-based perceptual familiarization, subjects performed a visual search task about 5 min after the familiarization task. Subjects were asked to detect a cued target logo from within a varying number of distractors. The targets were drawn from the sets of ten familiar and (formerly) unfamiliar logos that were presented in the familiarization phase, as well as the sets of ten familiar and ten unfamiliar logos that had not been presented before. The distractors were 120 unfamiliar logos that also had never appeared in the familiarization task. Thus, we created four visual search task conditions: preexperimentally familiar and unfamiliar logos that had been familiarized, and preexperimentally familiar and unfamiliar logos that had not been familiarized. Subjects searched for the target in displays that contained three, six, and nine total items. A total of 240 trials were separated into two 120-trial blocks. For each block, each of the 40 target logos appeared in a display of each set size one time. Every logo that was used as a distractor appeared five times in the first block and five times in the second block. The pairings of the targets and distractors and the order of all trials were randomized for each subject. The search task is illustrated in Fig. 2. Subjects were again seated at about 60 cm distance from the computer screen and
Atten Percept Psychophys
1 sec fixation
1 sec cue
Target Detection
Location Judgment
Fig. 2 The visual search task. In the target detection phase, subjects pressed the space bar to indicate that they had found the previously cued target logo. In the location judgment phase, they typed the letter that corresponded to the location of the target logo on the screen. This figure
provides an example of a set size 3 trial. Note that 24 letters (a–x) were presented on the screen during the target logo location judgment. The letters and the logos here are not drawn to scale
to acquaint subjects with the task, and subjects were provided a brief break after the first 120 trials.
ANOVA showed a main effect of familiarization, F(1, 34) = 11.14, p < .002, d = 1.14, but also an interaction between preexperimental familiarity and experimental familiarization, F(1, 34) = 8.99, p < .005, d = 1.03. Focused follow-up oneway ANOVAs revealed that for the preexperimentally familiar logos, the familiarization effect was significant, F(1, 34) = 34.96, p < .001, d = 2.03, but not for unfamiliar logos, F < 1, d = −0.03. We also performed analyses on the amount that search times increased for each additional item in the display, which has been termed “search efficiency” and is calculated as the slope of the function that relates RT to set size. ANOVAs showed no significant difference in the slopes between preexperimentally familiar familiarized versus nonfamiliarized conditions, nor between the preexperimentally unfamiliar familiarized versus nonfamiliarized conditions, both Fs < 1, ds = 0.04 and −0.21, respectively.
Results Performance on the N-back-like familiarization task was high. The overall response accuracy was 86 % (SD = 5 %), indicating that subjects were generally attentive to the task. Accuracy was also high in the visual search task, as assessed by the subjects’ location judgment accuracy (i.e., indicating where the target logos had been presented), with subject error rates ranging from 0 % to 6 % of trials. In these and all subsequent analyses in all experiments, trials were excluded if they featured incorrect location judgment responses and/or reaction times (RTs) greater than three standard deviations from the mean in each condition for each subject. In Experiment 1, 2.71 % of trials were excluded across all subjects. We also excluded outlier subjects, whose average RTs in more than one experimental condition were more than three standard deviations beyond the cross-subject mean RTs those conditions. One subject was thus excluded in this experiment. These same outlier detection procedures were used in all following experiments. RTs in the search task showed a clear effect of set size. Figure 3a displays the mean RTs across subjects as a function of set size. For all four conditions, RTs increased as set size increased, which was statistically reliable in this and all other experiments (all Fs > 24.69, all ps < .001). Subjects searched for preexperimentally familiar logos faster than for unfamiliar ones (Fig. 3a). This difference was visible at all set sizes and was statistically reliable. To formally test for differences between conditions, we ran a 2 (preexperimentally familiar vs. unfamiliar) × 2 (familiarized vs. nonfamiliarized) × 3 (set size) repeated measures analysis of variance (ANOVA). This analysis revealed a significant familiarity effect, such that preexperimentally familiar logos were searched for faster than were unfamiliar logos, F(1, 34) = 156.07, p < .001, d = 4.28. Familiarizing logos in the lab speeded RTs, but only for preexperimentally familiar logos. As is shown in Fig. 3a, familiarized preexperimentally familiar targets were searched for faster than were their nonfamiliarized counterparts. However, familiarized preexperimentally unfamiliar targets did not show an advantage over other unfamiliar targets. The
Discussion Experiment1 revealed that preexperimentally familiar logos were located faster than were unfamiliar logos, by 37.76 ms on average. This result extends prior work showing speeded search for familiar items (e.g., Flowers & Lohr, 1985; Hershler & Hochstein, 2009) to stimuli from nonexpert domains. However, the advantage here took the form of overall faster responses, without the change in the slopes of the search functions (i.e., a difference in search efficiency) that had been visible in past work. This is the only experiment (out of seven experiments in this article) in which we did not find a difference in slopes between preexperimentally familiar and unfamiliar targets, suggesting that it may reflect a Type II error (Table 1 shows the search slopes for all experiments). Familiarizing subjects with preexperimentally familiar logos further enhanced performance for these targets, by 28.80 ms on average. Importantly, this finding shows that within our paradigm, exposure (here achieved through the Nback-like familiarization task) unrelated to visual search can improve how quickly items are found. These results could be due to the logos being primed, gaining positive affect, or other mechanisms that depend upon recent exposure (e.g., Maccotta & Buckner, 2004; Simons, Koutstaal, Prince, Wagner, & Schacter, 2003; Stafford & Grimes, 2012; Zajonc, 2001).
Atten Percept Psychophys
a
b
c
d
e
f
g
h
Fig. 3 Results. (a) Experiment 1. In all graphs, the mean reaction times (RTs) for preexperimentally familiar logos are plotted in darker lines (blue online), and those for unfamiliar logos are lighter (red online). The mean RTs for logos familiarized in the laboratory are plotted as
dotted lines, and the lines for logos that were not familiarized are solid. (b–c) Results of Experiments 2a and 2b. (d–e) Results of Experiments 3a and 3b. (f) Results of Experiment 4a, Day 1. (g) Results of Experiment 4a, Day 2. (h) Results of Experiment 4b
However, such mechanisms cannot account for preexperimental familiarity, because no effects of laboratorybased exposure were found for unfamiliar logos; the two conditions differed by −0.69 ms on average. This result suggests that the benefits of preexperimental familiarity are not
due to mechanisms that depend upon recent exposure, such as priming or the mere exposure effect. Rather, the effects of exposure seem to require some amount of preexisting familiarity. We speculate that familiar logos may have possessed more complete or clear representations that
Atten Percept Psychophys Table 1 Mean RTs and search slopes of familiar and unfamiliar (experimentally nonfamiliarized) logos in all experiments RT (ms)
Slope (ms/item)
Exp.
Familiar
Unfamiliar (Nonfamiliarized)
p
Familiar
Unfamiliar (Nonfamiliarized)
p
1 2a
620.26 549.10
658.01 601.77
The hard-won benefits of familiarity in visual search: naturally familiar brand logos are found faster Xiaoyan Angela Qin & Wilma Koutstaal & Stephen A. Engel
# Psychonomic Society, Inc. 2014
Abstract Familiar items are found faster than unfamiliar ones in visual search tasks. This effect has important implications for cognitive theory, because it may reveal how mental representations of commonly encountered items are changed by experience to optimize performance. It remains unknown, however, whether everyday items with moderate levels of exposure would show benefits in visual search, and if so, what kind of experience would be required to produce them. Here, we tested whether familiar product logos were searched for faster than unfamiliar ones, and also familiarized subjects with previously unfamiliar logos. Subjects searched for preexperimentally familiar and unfamiliar logos, half of which were familiarized in the laboratory, amongst other, unfamiliar distractor logos. In three experiments, we used an N-back-like familiarization task, and in four others we used a task that asked detailed questions about the perceptual aspects of the logos. The number of familiarization exposures ranged from 30 to 84 per logo across experiments, with two experiments involving across-day familiarization. Preexperimentally familiar target logos were searched for faster than were unfamiliar, nonfamiliarized logos, by 8 % on average. This difference was reliable in all seven experiments. However, familiarization had little or no effect on search speeds; its average effect was to improve search times by 0.7 %, and its effect was significant in only one of the seven experiments. If priming, mere exposure, episodic memory, or relatively modest familiarity were responsible for familiarity’s effects on search, then performance should have improved following familiarization. Our results suggest that the search-related advantage of familiar logos does not develop easily or rapidly.
X. A. Qin : W. Koutstaal : S. A. Engel (*) Department of Psychology, University of Minnesota, Minneapolis, MN 55455, USA e-mail: [email protected]
Keywords Visual search . Familiarity . Brand logos . Semantic representations . Perceptual learning . Practice effects
The need for visual search arises frequently in everyday life; we routinely scan shopping shelves for preferred products, articles for specific references, and web pages for preferred content, to pick just a few examples. Because of these applications and more specialized ones in medical, military, and security domains, the factors that influence search efficiency have received intense study. One factor that affects search speed is the amount of experience observers have had with the items used in the task. Target familiarity, distractor familiarity, and familiarity differences between targets and distractors all decrease search times (e.g., Flowers & Lohr, 1985; Lubow & Kaplan, 1997; Malinowski & Hübner, 2001; Rauschenberger & Chu, 2006; Richards & Reicher, 1978; Shen & Reingold, 2001; Wang, Cavanagh, & Green, 1994). Understanding how familiarity speeds search has important implications for cognitive theory, because it may reveal how mental representations of commonly encountered items are changed by experience in order to optimize performance. Familiarity also plays an important role in visual search in applied settings, such as baggage scanning by security personnel and supermarket product choice by consumers. The goal of this study was to understand the everyday familiarity that people have for commonly encountered objects, such as a favorite article of clothing, brand of soda, or model of car. Most past research examining familiarity’s effects on visual search had used stimuli from domains in which subjects possessed extensive visual expertise, such as letters, digits, faces, and symbols. Thus, it remained an open question whether subjects would show advantages for stimuli with which they had a more moderate amount of experience.
Atten Percept Psychophys
We used brand logos to examine this level of familiarity. Logos provide a set of diverse, moderately complex items, which contain a natural manipulation of real-world familiarity; depending on marketing budgets and strategies, we are exposed to some brand logos much more often than others. In general, however, we are less expert at viewing logos than at viewing the highly overlearned categories of letters, digits, symbols, and faces that had been used in past research. A final benefit of studying logos was that the results might not only shed light on how natural experience changes the representation of everyday objects, but also inform marketing research. It also remains unknown precisely how much and what kind of exposure people require in order to locate familiar items more quickly. One kind of exposure that has been shown definitively to improve search times is practice with the items while performing the search task itself (e.g., Baluch & Itti, 2010; Le Dantec, Melton, & Seitz, 2012; Sireteanu & Rettenbach, 1995, 2000; Williams, Henderson, & Zacks, 2005; but see Tong & Nakayama, 1999, for an exception). These experiments confirmed that subjects can learn to improve performance on particular items in a particular task; such specificity is a hallmark of much of laboratory-based perceptual learning (for reviews, see Lu, Yu, Watanabe, Sagi, & Levi, 2010; Sagi, 2011; Sasaki, Nanez, & Watanabe, 2010). Real-world familiarity of the kind that logos and other everyday items possess, however, is not acquired within a single task context. Thus, while there is little doubt that practicing visual search for logos would improve visual search for logos, such a demonstration might be unenlightening about the nature of everyday familiarity. One study (Mruczek & Sheinberg, 2005) tested whether exposing subjects to individually presented items would produce the benefits of familiarity, but even here, targets were presented in a search task (in a zero-distractor condition). What kind of everyday exposure to items might improve visual search? It is possible that relatively brief, uninformative encounters with an item could produce large familiarity effects. For example, items that are “primed” by prior viewings, items that are very recently encountered, and items to which one has been “merely exposed” all show advantages in cognitive tasks, despite small numbers of exposures (e.g., Alter & Oppenheimer, 2009; Henson, 2003; Stafford & Grimes, 2012; Zajonc, 2001). These results suggest that seeing a logo just a few times before performing visual search could improve performance. On the other hand, larger amounts or richer kinds of experience might be required to produce familiarity effects. Particularly for logos, familiarity effects may depend on factors such as observers’ gaining affective associations due to advertising, identifying the logos in many highly diverse visual contexts, or simply having experienced a very large number of exposures. If such factors are important, laboratory exposure would be unlikely to produce familiarity effects for logos.
In the present experiments, we explored the strength of the target familiarity effect for brand logos and attempted to generate it using several different, non-search-exposure procedures for initially unfamiliar logos. We used two different tasks to familiarize observers with the logos: an N-back-like task, and a set of questions about detailed perceptual aspects of the logos. In all experiments, the testing phase consisted of a visual search task in which familiarized and nonfamiliarized logos served as the targets and a heterogeneous set of nonfamiliarized unfamiliar logos acted as distractors. To anticipate our results, preexperimentally familiar logos manifested a robust advantage over unfamiliar logos, in both search speed and search efficiency. However, familiarization in the laboratory produced little, and usually no, benefits for visual search, despite the use of procedures that varied the amount of exposure, the amount of attention given to the exposure, and the amount of time given for consolidation of the exposure’s effects.
Experiment 1 In our first experiment, we tested whether familiar logos are searched for faster than unfamiliar ones. Past research has tested whether familiarity can affect search times for items in overlearned domains, such as letters and faces. Familiarity generally speeds search times, though when it is attached to distractors it can also slow responses (McGugin, McKeeff, Tong, & Gauthier, 2011). Only one study, to our knowledge, has examined search and familiarity in domains in which subjects are less expert, and found that familiar targets can be found faster than unfamiliar ones (Mruczek & Sheinberg, 2005). However, in this study familiarity was gained through practice with the visual search task itself, and it is well known that practice in search can speed search (Le Dantec et al., 2012; Mruczek & Sheinberg, 2005; Sireteanu & Rettenbach, 1995, 2000). Thus one contribution of Experiment 1 is to test whether the diverse but modest familiarity with brand logos acquired in everyday life can speed visual search. In past work, displays were also often simplified by placing a target in a homogeneous field of distractors. Although this type of experimental design offers good control over the physical parameters of the displays, the observed effects may not generalize well to real-world search behaviors (e.g., Hershler & Hochstein, 2009; Mruczek & Sheinberg, 2005). Using more complex everyday stimuli and more heterogeneous displays addresses this limitation (Mruczek & Sheinberg, 2005; Tong & Nakayama, 1999), and we adopted that approach here. Experiment 1 also tested whether familiarization with logos through simple visual exposure would be able to improve search times. Many theories of the benefits of familiarity, including those based upon priming, recency of exposure, or
Atten Percept Psychophys
“mere” exposure, could predict that such familiarization should speed search. For example, items to which one has been exposed just a few times in the recent past, referred to as “primed,” are processed more rapidly (e.g., Maccotta & Buckner, 2004), which could lead to an advantage in visual search (cf. object-based priming within visual search tasks, as in Ásgeirsson & Kristjánsson, 2011; Kristjánsson & Campana, 2010). Alternatively, items to which one has been exposed a small number of times may gain not only perceptual fluency, but also positive affect (Stafford & Grimes, 2012; Zajonc, 2001). Preference for recently exposed logos could also possibly speed search. To familiarize subjects with a set of logos, they viewed them repeatedly, while performing an N-back-like task. Next, subjects performed a visual search task using both these logos and a second set of logos. Both sets contained preexperimentally familiar as well as preexperimentally unfamiliar logos. Thus subjects searched for four different kinds of logos obtained by crossing two factors: preexperimentally familiar and unfamiliar logos that were familiarized in the laboratory and preexperimentally familliar and unfamiliar logos that were not familiarized.
“unfamiliar” (rating less than 2.58) or “familiar” (rating higher than 4.55). To verify the reliability of ratings and the validity of these categories, we computed for the familiar and unfamiliar logos used in our experiments, the average number of logos that would be misclassified by our norming population. We did this by simply looking up the ratings for those logos for each member of our norming group and comparing them to the cutoffs. Of the 40 logos used in our experiments, on average fewer than four would be misclassified for individual subjects by our cutoffs. These 40 logos were further divided into two sets of ten familiar and two sets of ten unfamiliar logos; one set of each familiarity category was used as the experimentally familiarized set, whereas the other set was not familiarized. The familiarized and nonfamiliarized sets were counterbalanced across subjects. In addition, we approximately balanced the color, shape, and the number of component letters (if any) for the four sets of ten target logos. The mean familiarity ratings were 1.62 and 1.68 for the two unfamiliar sets (SDs = 0.57 and 0.50, respectively), and 4.79 and 4.75 for the two familiar sets (SDs = 0.10 and 0.13, respectively). A different set of 120 unfamiliar logos (mean familiarity rating = 1.73, SD = 0.57) was used as distractors in the visual search task.
Method
Procedure
Subjects
Subjects were tested individually. After providing informed consent, they completed a demographic inquiry form asking about their age, gender, color vision, visual acuity, and number of years residing in the United States. Written instructions were provided before each task.
A group of 36 subjects were recruited from the University of Minnesota (nine males, 27 females; mean age = 20.5 years). All of the subjects in all experiments reported in this article reported that they had lived in the United States for at least 10 years, had normal or corrected-tonormal vision, and received course credit or monetary compensation for their participation. All experimental sessions lasted about 60 min, and at the end of the session subjects were debriefed and thanked. Stimuli The logo images used in all of the experiments reported here were initially found on the Internet. All logos were edited in Photoshop CS2 to remove trademarks and were resized to ensure equivalent image size and resolution. Each logo was scaled to fit within a 60×48 pixel rectangle. All logos were normed for familiarity using ratings by subjects (N = 92) from the same general population as the main experimental subjects. The logos were rated on a 5-point scale, from 1, not familiar at all, to 5, extremely familiar. We used these norming data to select twenty familiar and twenty unfamiliar logos as targets in our experiments. The norming data were very reliable, which allowed us to use simple cutoffs in the average ratings to classify logos as either
Familiarization task The subjects were asked to detect repeated presentations of the logos in a stream of singly presented logos (see Fig. 1a and b). We made no mention of subsequent testing or visual search; that is, the experimental familiarization or encoding phases in this experiment and in all subsequent experiments were incidental. The familiar and unfamiliar sets of logos were each presented in 40 separate alternating blocks. One half of the subjects received experimental familiarization of the familiar logos in odd-numbered blocks, and experimental familiarization of the unfamiliar logos in evennumbered blocks; the other half of the subjects received the opposite pattern. Each experimental familiarization block consisted of 17 trials. Given that ten logos were in each set, this resulted in seven repeats in a block. The repeated items could occur anywhere within a block; that is, varying numbers of interleaved items could occur between the first and second presentations of a logo within a block. Subjects were instructed to respond to any repeated logos within a given block; since the logos also repeated across blocks, subjects were asked to “restart” their repetition count for each block.
Atten Percept Psychophys
a
b
Fig. 1 Familiarization phase in Experiments 1–2b. (a) Experimental familiarization for familiar logos. (b) Experimental familiarization for unfamiliar logos. In Experiment 1, subjects were experimentally familiarized with both familiar and unfamiliar logos; in Experiments 2a and 2b,
only unfamiliar logos were familiarized. In Experiment 2b, the responses were “1” or “2” instead of “yes” or “no,” to indicate the first or second occurrence of a logo
In each trial, a logo appeared in the center of a white screen for 1.5 s. Subjects were situated at about 60 cm from the monitor. A chinrest was used to control viewing distance and to prevent unnecessary head movements. Subjects pressed one of two keys (labeled “yes” or “no”) to indicate whether or not the currently displayed logo was a (within-block) repeat. Subjects completed 12 practice trials before starting the actual task. By the end of the 30-min familiarization phase, each subject was exposed to each logo approximately 30 times.
rested their chins on a chinrest. On each trial, a black fixation cross appeared in the center of the screen to alert subjects that a search array was about to be presented. After 1 s, a “cue” logo appeared in the position of the fixation cross, indicating the to-be-searched-for target in the subsequently presented array. After another second, three, six, or nine logos appeared in randomly selected positions in an invisible 5×5 grid. The central position of this grid was always occupied by the cued logo, leaving 24 possible locations for the search array. The logo positions were jittered slightly within the boxes that comprised the grid, in order to avoid perfect or repeated alignments of the logo positions. Given that the unfamiliar target logos were not well known to subjects beforehand, these logos may have been disadvantaged in terms of subjects’ ability to keep them in mind when they were targets during the search task. To minimize any working memory effects on search performance for logos of differing familiarities, the cue logo remained on the screen throughout the search. To differentiate the cue from the to-besearched-for target, a black frame outlined the cue at the time when the search array appeared. Subjects were asked to press the space bar as soon as they thought they had located the designated target logo (Fig. 2). Accuracy and speed were equally emphasized. To assess visual search accuracy, after the subject pressed the space bar, a letter array containing 24 letters (A–X) replaced the search array, with each letter corresponding to one of the possible locations of the logos in the search array. Subjects were required to indicate where the target logo had appeared by pressing the letter corresponding to its position. After subjects made this location judgment (i.e., had indicated the letter that they thought corresponded to the location of the target in the prior search array), the next trial started automatically with the 1-s fixation. Three practice trials were provided
Search task To assess the effects of preexperimental familiarity and the laboratory-based perceptual familiarization, subjects performed a visual search task about 5 min after the familiarization task. Subjects were asked to detect a cued target logo from within a varying number of distractors. The targets were drawn from the sets of ten familiar and (formerly) unfamiliar logos that were presented in the familiarization phase, as well as the sets of ten familiar and ten unfamiliar logos that had not been presented before. The distractors were 120 unfamiliar logos that also had never appeared in the familiarization task. Thus, we created four visual search task conditions: preexperimentally familiar and unfamiliar logos that had been familiarized, and preexperimentally familiar and unfamiliar logos that had not been familiarized. Subjects searched for the target in displays that contained three, six, and nine total items. A total of 240 trials were separated into two 120-trial blocks. For each block, each of the 40 target logos appeared in a display of each set size one time. Every logo that was used as a distractor appeared five times in the first block and five times in the second block. The pairings of the targets and distractors and the order of all trials were randomized for each subject. The search task is illustrated in Fig. 2. Subjects were again seated at about 60 cm distance from the computer screen and
Atten Percept Psychophys
1 sec fixation
1 sec cue
Target Detection
Location Judgment
Fig. 2 The visual search task. In the target detection phase, subjects pressed the space bar to indicate that they had found the previously cued target logo. In the location judgment phase, they typed the letter that corresponded to the location of the target logo on the screen. This figure
provides an example of a set size 3 trial. Note that 24 letters (a–x) were presented on the screen during the target logo location judgment. The letters and the logos here are not drawn to scale
to acquaint subjects with the task, and subjects were provided a brief break after the first 120 trials.
ANOVA showed a main effect of familiarization, F(1, 34) = 11.14, p < .002, d = 1.14, but also an interaction between preexperimental familiarity and experimental familiarization, F(1, 34) = 8.99, p < .005, d = 1.03. Focused follow-up oneway ANOVAs revealed that for the preexperimentally familiar logos, the familiarization effect was significant, F(1, 34) = 34.96, p < .001, d = 2.03, but not for unfamiliar logos, F < 1, d = −0.03. We also performed analyses on the amount that search times increased for each additional item in the display, which has been termed “search efficiency” and is calculated as the slope of the function that relates RT to set size. ANOVAs showed no significant difference in the slopes between preexperimentally familiar familiarized versus nonfamiliarized conditions, nor between the preexperimentally unfamiliar familiarized versus nonfamiliarized conditions, both Fs < 1, ds = 0.04 and −0.21, respectively.
Results Performance on the N-back-like familiarization task was high. The overall response accuracy was 86 % (SD = 5 %), indicating that subjects were generally attentive to the task. Accuracy was also high in the visual search task, as assessed by the subjects’ location judgment accuracy (i.e., indicating where the target logos had been presented), with subject error rates ranging from 0 % to 6 % of trials. In these and all subsequent analyses in all experiments, trials were excluded if they featured incorrect location judgment responses and/or reaction times (RTs) greater than three standard deviations from the mean in each condition for each subject. In Experiment 1, 2.71 % of trials were excluded across all subjects. We also excluded outlier subjects, whose average RTs in more than one experimental condition were more than three standard deviations beyond the cross-subject mean RTs those conditions. One subject was thus excluded in this experiment. These same outlier detection procedures were used in all following experiments. RTs in the search task showed a clear effect of set size. Figure 3a displays the mean RTs across subjects as a function of set size. For all four conditions, RTs increased as set size increased, which was statistically reliable in this and all other experiments (all Fs > 24.69, all ps < .001). Subjects searched for preexperimentally familiar logos faster than for unfamiliar ones (Fig. 3a). This difference was visible at all set sizes and was statistically reliable. To formally test for differences between conditions, we ran a 2 (preexperimentally familiar vs. unfamiliar) × 2 (familiarized vs. nonfamiliarized) × 3 (set size) repeated measures analysis of variance (ANOVA). This analysis revealed a significant familiarity effect, such that preexperimentally familiar logos were searched for faster than were unfamiliar logos, F(1, 34) = 156.07, p < .001, d = 4.28. Familiarizing logos in the lab speeded RTs, but only for preexperimentally familiar logos. As is shown in Fig. 3a, familiarized preexperimentally familiar targets were searched for faster than were their nonfamiliarized counterparts. However, familiarized preexperimentally unfamiliar targets did not show an advantage over other unfamiliar targets. The
Discussion Experiment1 revealed that preexperimentally familiar logos were located faster than were unfamiliar logos, by 37.76 ms on average. This result extends prior work showing speeded search for familiar items (e.g., Flowers & Lohr, 1985; Hershler & Hochstein, 2009) to stimuli from nonexpert domains. However, the advantage here took the form of overall faster responses, without the change in the slopes of the search functions (i.e., a difference in search efficiency) that had been visible in past work. This is the only experiment (out of seven experiments in this article) in which we did not find a difference in slopes between preexperimentally familiar and unfamiliar targets, suggesting that it may reflect a Type II error (Table 1 shows the search slopes for all experiments). Familiarizing subjects with preexperimentally familiar logos further enhanced performance for these targets, by 28.80 ms on average. Importantly, this finding shows that within our paradigm, exposure (here achieved through the Nback-like familiarization task) unrelated to visual search can improve how quickly items are found. These results could be due to the logos being primed, gaining positive affect, or other mechanisms that depend upon recent exposure (e.g., Maccotta & Buckner, 2004; Simons, Koutstaal, Prince, Wagner, & Schacter, 2003; Stafford & Grimes, 2012; Zajonc, 2001).
Atten Percept Psychophys
a
b
c
d
e
f
g
h
Fig. 3 Results. (a) Experiment 1. In all graphs, the mean reaction times (RTs) for preexperimentally familiar logos are plotted in darker lines (blue online), and those for unfamiliar logos are lighter (red online). The mean RTs for logos familiarized in the laboratory are plotted as
dotted lines, and the lines for logos that were not familiarized are solid. (b–c) Results of Experiments 2a and 2b. (d–e) Results of Experiments 3a and 3b. (f) Results of Experiment 4a, Day 1. (g) Results of Experiment 4a, Day 2. (h) Results of Experiment 4b
However, such mechanisms cannot account for preexperimental familiarity, because no effects of laboratorybased exposure were found for unfamiliar logos; the two conditions differed by −0.69 ms on average. This result suggests that the benefits of preexperimental familiarity are not
due to mechanisms that depend upon recent exposure, such as priming or the mere exposure effect. Rather, the effects of exposure seem to require some amount of preexisting familiarity. We speculate that familiar logos may have possessed more complete or clear representations that
Atten Percept Psychophys Table 1 Mean RTs and search slopes of familiar and unfamiliar (experimentally nonfamiliarized) logos in all experiments RT (ms)
Slope (ms/item)
Exp.
Familiar
Unfamiliar (Nonfamiliarized)
p
Familiar
Unfamiliar (Nonfamiliarized)
p
1 2a
620.26 549.10
658.01 601.77
