Disconfounding Perceptual Precedence from Spatial Attention

Does local/global perceptual bias tell us anything about local/global attention?

Does local/global perceptual bias tell us anything about local/global selective attention?

Serge Caparos, Karina J. Linnell, Andrew J. Bremner, Jan W. de Fockert, and Jules Davidoff

Department of Psychology, Goldsmiths, University of London, UK

E-mail:

Telephone: 0044 20 7919 7222

Fax: 0044 20 7919 7873

Word count: 2480

Key words: Spatial attention; Perceptual style; Visual perception; Cross cultural; Local global

ABSTRACT [125 words]

A localperceptual bias has been linked to a lesser ability to attend globally. We examined this proposed link in the Himba, a remote Namibian population, who have demonstrated a strong local bias compared to British observers. If local perceptual bias is related to a lesser ability to attend globally, the Himba should be little distracted by global information when performing a local-selection task (on a local/global hierarchical figure) but substantially distracted by local information when performing a global-selection task. However, the Himba performed better than British observers in both local- and global-selection tasks, suggesting that they possess agreaterability to exercise control over attentional selection in response to task demands. We conclude that local/global perceptual bias must be distinguished from local/global selective attention.

INTRODUCTION

Local/global perceptual bias refers to the tendency to prioritize the processing of either local elements of a figure or its global configuration(see Figure 1). Such biases have constituted a longstanding interest in cross-cultural research (e.g., Davidoff, Fonteneau, & Fagot, 2008; Deregowski, 1989; Miyamoto, Nisbett, & Masuda, 2006; Rivers, 1905; Segall, Campbell, & Herskovits, 1966), with a recent focus on differences between Westerners and East-Asians (e.g., Nisbett, Peng, Choi, & Norenzayan, 2001). Compared to Westerners, East-Asians show a global bias: they prioritize the perceptual processing of global information when categorizing objects (Norenzayan, Smith, Kim, & Nisbett, 2002), when detecting changes in visual scenes (Miyamoto et al., 2006), and when memorizing visual information (Masuda & Nisbett, 2001).

The relative biases of East-Asians and Westerners have been commonly attributed to their greater abilities to attend, respectively,to global or local information(e.g., Kitayama, Duffy, Kawamura, & Larsen, 2003; McKone, Davies, Fernando, Aalders, Leung, Wickramariyaratne, & Platow, 2010; Üskül, Kitayama, & Nisbett, 2008). Thus, the facility for local/global selective attention, orthe tailoring of the breadth of spatial attention in response to local/global selection tasks, is argued to be linked to a predisposition to prioritize processing of local/global aspects of stimuli independently of task demands. Accordingly, East-Asians should be worse than Westerners in tasks requiring local selection, whereas in tasks requiring global selection Westerners should be worse than East-Asians. There is surprisingly little evidence for this view. Previousfindings have only shown that East-Asians perform better with global than local targets in tasks where local and global processing is equally beneficial (e.g., McKone et al., 2010); this result does not however show a lesser ability to attend locally when this is required by the task, it justsuggests a preference for global processing when given the choice.

To our knowledgeonly Kitayama et al. (2003)have presented evidencesuggesting that East-Asians and Westerners have, respectively, global and local advantages in selective attention. They used the ‘framed-line test’ in whichtask demands are varied so that a line has to be processed eitherin reference to, or ignoring, a surrounding frame, favoring, respectively, global and local attention.Theyfound a local-selection advantage in Westerners and a global-selection advantage in East-Asians. However, this was not replicated by Koh and Milne (2012) or Zhou, Gotch, Zhou, and Liu (2008) who reportedglobal-selection advantages for both Westerners and East-Asians.

To further investigatethe relationship between local/global perceptual bias and local/global selective attention, we examined whether the Himba, a remote traditional population in northern Namibiawho show a local perceptual bias compared to Westerners, revealcorresponding patterns of advantage and disadvantage on tasks requiring local and global selection. The Himbashow a strong local bias compared to British observers when making subjective similarity matchesregarding hierarchical figures (Caparos, Ahmed, Bremner, De Fockert, Linnell, & Davidoff, 2012; Davidoff, Fonteneau, & Fagot, 2008). They also perform better than British observers on tasks favoring local selection (Caparos et al., 2012; De Fockert, Davidoff, Fagot, Parron, & Goldstein, 2007; De Fockert, Caparos, Linnell, & Davidoff, 2011). Their performance on tasks favoringglobal selection has never been tested but,according to Kitayama et al. (2003) andMcKone et al. (2010), their local perceptual bias should lead to a disadvantage in global selection.

The crucial aspect of our study was to compare Himba and British participants’ ability to attend selectively to local features and global configurations. We used Navon hierarchical figures (Navon, 1977; see Figure 1) which are widely used to measure local/global processing.Participants were instructedeither to identify the local elements of the figure while ignoring its global shape (‘local-selection’ task) or to identify the global shape of the figure while ignoring its local elements (‘global-selection’ task). On any given trial, local elements and global shape could be either consistent or inconsistent (see Figure 1). We measured the effect of the consistency of information between attended and unattended levels on the speed and accuracy of participants’identification responses. If local/globalperceptual bias leads to local/global selective attention then, compared to the Himba, British participants should be more affected by global inconsistent information in the local-selection task; in contrast, Himba participants should be more affected by local inconsistent information in the global-selection task.

Perceptual biases, as well as differing among participant groups, can also derive from characteristics of the stimuli (Martin, 1979). We employedconditions which varied the relative saliency of the local elements and global configurations. The ‘local-salient’ figures contained larger local elements, whose centers were spaced further apart, and the ‘global-salient’ figures contained more elements which were smaller and more densely packed. Previous research demonstrates that such figures support local and global selection, respectively (e.g., Martin, 1979). With regard to theseeffects of stimulus characteristics (local- vs. global-salient), we hypothesized that they would operate similarly for the participant groups. See, for example, the equal effects for Westerners and Himba of varying stimulus characteristics in the Ebbinghaus illusion despite markedly different magnitudes of illusion (De Fockert et al., 2007).

METHOD

Participants

The two populations comprised: (1) 54 British (32 females, mean age 21 years, range 18-35) and (2) 58 Himba (28 females, mean estimated age 23 years, range 16-45, monolingual in Otjiherero). The Himba are semi-nomadic herders, having little contact with Western artifacts. On average, they had visited even a small town only three times in their lifetime. None of them had ever been involved in experimental research. The British were undergraduate students from Goldsmiths, University of London and were either paid or received course credits. The Himba were rewarded in kind.

Stimuli

A target figure occurred on each trial, equally often at the center of the screen, 1.5° (of visual angle) above the center of the screen, or 1.5° below it. The target was a hierarchical Navon-like figure (Navon, 1977; see Figure1). The local elements (crosses or squares) were equally spaced from each other and were arranged to create a larger square or cross – the global shape. Two conditions manipulated the saliency of local and global elements by varying the size, and distance between the centers of, the local elements (see Figure 1). The global shape (4.4° in height and width) could be made of a few large local elements (local-salientcondition; 12 local elements formed a global square or 9 local elements formed a global cross; the local elements subtended 0.9° in height and width) or it could be made of many small local elements (global-salient condition; 20 local elements formed a global square or 13 local elements formed a global cross; the local elements subtended 0.6° in height and width). In addition to the saliency conditions,there were two conditions of consistency. The local elements and the global shape could be either consistent (e.g., local crosses making a global cross) or inconsistent (e.g., local crosses making a global square).

Figure 1. Illustration ofthe eight possible target figures: two shapes (cross and square) across two conditions of saliency (local-salient and global-salient) and two conditions of consistency (consistent and inconsistent).

Procedure

We tested the Himba participants inside a tent placed in a shaded area and the British participants in a quiet and moderately lit testing room. The experiments were run using a script constructed in E-Prime (Schneider, Eschman, & Zuccolotto, 2002). Stimuli were presented on a 20-in CRT screen at a viewing distance of 70 cm.

Participants were asked to fixate a central cross thatoccurredfor 1000 ms at the start of each trial and was followed by a blank screen for 500 ms. A target figure then occurred for 250 ms. There were two possible tasks with regard to this target figure: (1) a global-selection task - participants were asked to ignore the local elements and to identify the global shape with a left-button press on a button box for ‘square’ and a right-button press for ‘cross’; (2) a local selection task -participants were asked to ignore the global shape and to identify the local elements with a left-button press for ‘square’ and a right-button press for ‘cross’.

The local- and global-selection tasks both consisted of 120 trials. They were blocked and performed in a counterbalanced order across participants. In both task blocks, consistent and inconsistent trials occurred equally often in a randomized order. Different participants performed the local-salient and global-salient conditions. Most (90%) of the Himba did not know the meaning of the words ‘square’and/or ‘cross’ (in Otjiherero). Regardless of whether they knew the meaning of the words, we used drawings to explain the task. When it was clear that the participants understood the instructions, they performed ten practice trials and the task started.

RESULTS

For each participant, we analyzed the speed and accuracy of target identification (see Figure 2). Reaction times (RTs)were examined only for trials in which participants had correctly identified the target (3.3% trials were excluded by this criterion) and only when RTs were longer than 200 ms and fell within three standard deviations of the overall mean for the participant (0.8% trials were excluded by this criterion). An estimate of interference was obtained by subtracting consistent RTs/errors from inconsistent RTs/errors (see Figure 3). Interference, in both RTs and errors, was then compared across conditions of: (1) Task (local vs. global selection), (2) Saliency (local-salientvs. global-salient figure), and (3) Culture (British vs. Himba) using three-way mixed-design ANOVAs.

Figure 2. RTs (2a) and errors (2b) as a function of saliency (local- or global-salient), task (local- or global-selection), consistency (consistent or inconsistent irrelevant information), and culture (British or Himba). Error bars show +/- SEM.

The RT-interference analysis showed a significant main effect of Culture (F(1,108) = 10.5, p = 0.002, ηp2= 0.089); interference was significantly lower in Himba (27 ms, SEM=3.4) than British observers (47ms, SEM=5.2). Importantly, there was no significant Culture x Task interaction (F1); interference was lower in the Himbain both local and global tasks. It is unlikely that the lower interference in the Himba can be explained by their overall longer RTs.[1] Indeed, a subset of Himba and British participants pair-matched on their overall RTs still showed lower interference in Himba (24 ms, SEM=5.5) than British observers (46 ms, SEM=7.0, F(1,46) = 5.5, p = 0.023, ηp2= 0.107).

Other significant effects on RT interference were consistent with what one would expect from the task and stimulus parameters (Martin, 1979). There was a significant interaction between Saliency and Task (F(1,108) = 6.5, p = 0.013, ηp2= 0.056); in the local-salient condition, interference from the global shape was marginally smaller than interference from local elements (effect of Task approaching significance; t(55) = 1.8, p = 0.068, d= 0.298) whereas, in the global-salient condition, interference from the global shape was marginally larger than interference from local elements (effect of Task approaching significance; t(55) = 1.9, p = 0.062, d= 0.341). Note that the interaction between Saliency and Task did not vary across British/Himba observers (theSaliency x Task x Culture interaction was not significant; F < 1). Thus, whereas interference from local elements and global shape clearly varied as a function of Saliency, this effect was the same in the Himba and the British observers. No other effects were reliable.

Figure 3. Interference in RTs (inconsistent minus consistent RTs; 3a) and in errors (inconsistent minus consistent error percentages; 3b) as a function of saliency (local- or global-salient), type of interference (local elements or global shape) and culture (British or Himba). Error bars show +/- SEM.

The equivalent analysis of interference in errors generated no significant effect (all ps > 0.1; an unsurprising finding given that interference in errorswas below 3% in all conditions; see Figure 3). Nevertheless, in order to ensure that the results from the RT-interference analysis did not result from a trade-off with errors, we ran an additional analysis on inverse efficiency (i.e., RTs divided by accuracy). The same pattern of main effects and interactionswas obtained as in the RT-interference analysis. Most notably, the main effect of Culture on interference was significant (F(1,108) = 10.5, p = 0.002, ηp2= 0.089) confirming that interference was lower in Himba than British observers.

DISCUSSION

We compared Himba and British observers on local- and global-selection tasks. Consistent with previous findings (De Fockert et al., 2011), the Himba demonstrated a superior ability to select local information; compared to British observers, they were little distracted by inconsistent global information in the local-selection task. Importantly, however, the Himba also demonstrated a superior ability to select global information; compared to British observers they were little distracted by inconsistent local information in the global-selection task even with figures composed of salient local elements.Thus, despite exhibiting a particularly strong local perceptual bias (i.e., Caparos et al., 2012; Davidoff, Fonteneau, & Fagot, 2008), the Himba can nonetheless attend globally when necessary. These findings are strikingly inconsistent with the idea that local or globalperceptual bias is linked to the ability to attend locally or globally(e.g., Kitayama et al., 2003; McKone et al., 2010; Üskül et al., 2008). Instead, these data make it necessary to distinguish local/globalperceptual bias from local/global attention. We argue that the Himba showed more efficient selective attention than Westerners.

It might be suggested that the superior selective attention which we observed in the Himba is somehow related to theirunfamiliarity with geometric shapes and/or our equipment. However, such lack of familiarity couldnot improve performance at both local and global selection tasks. Thus, lack of familiarity with stimuli and/or our equipment could lead to greater alertness but that should improve global but not local selection (Van Vleet,Hoang-duc, DeGutis, & Robertson, 2010; Weinbach & Henik, 2011). Furthermore, practice and task expertise is known to improve rather than impair selective attention (Green & Bavelier, 2012; Kelley & Yantis, 2009; Wilson, MacLeod, Muroi, 2008); so, in that respect, the Himba should be poorer at selective attention. In any case, other findings suggest that effects of stimulus familiarity in selective attention may be small for the Himba (Davidoff, Fonteneau & Goldstein, 2008; De Fockert et al., 2011).Crucially, our findings that the Himba were better at selective attention than British observers, across tasks, were significant even in a subset of Himba and British participants who demonstrated the same proficiency (i.e., who were pair-matched on their overall RTs).

In conclusion, Himba observers arebetter than British observers at attending both locally and globally according to task requirements. One could speculate that this might be because they have superior attentional control compared to Westerners, consistent with the idea that remote peoples are better able to concentrate (Jung, 1933/2001). Whatever the mechanism, these data,taken together with the more local perceptual bias of the Himba, question the view that there is a straightforward relationship between local/global selective attention andlocal/globalperceptual bias.

ACKNOWLEDGEMENTS

We acknowledge support from the ESRC (2558227; awarded to JD, KJL, JWdF and AJB), and the ERC (241242; awarded to AJB). We also thank Cambridge Research Systems for sponsorship.

REFERENCES

Caparos, S., Ahmed, L., Bremner, A., De Fockert, J. W., Linnell, K. J., & Davidoff, J. (2012). Exposure to an urban environment alters the local bias of a remote culture. Cognition, 122, 80-85.

Davidoff, J., Fonteneau, E., & Fagot, J. (2008). Local and global processing: observations from a remote culture. Cognition, 108, 702-709.

Davidoff, J., Fonteneau, E., & Goldstein, J. (2008). Cultural differences in perception: observations from a remote culture. Journal of Cognition and Culture, 8, 189-209.

De Fockert, J. W., Caparos, S., Linnell, K. J., & Davidoff, J. (2011). Reduced distractibility in a remote culture. PLoS ONE,6, 1-8.

De Fockert, J. W., Davidoff, J., Fagot, J., Parron, C., & Goldstein, J. (2007). More accurate size contrast judgments in the Ebbinghaus Illusion by a remote culture. Journal of Experimental Psychology: Human Perception and Performance, 33, 738-742.

Deregowski, J. B. (1989). Real space and represented space. Behavioral and Brain Sciences. 12, 51-119.

Green, C. S., & Bavelier, D. (2012). Learning, attentional control, and action video games. Current Biology, 22, 197-206.

Jung, C. (1933/2001) Modern man in search of a soul. Zurich: Kegan Paul, Trench * Truebner/Routledge.

Kelley, T., & Yantis, S. (2009). Learning to attend: Effects of practice on information selection. Journal of Vision, 9, 1-18.

Kitayama, S., Duffy, S., Kawamura, T., & Larsen, J. T. (2003). Perceiving an object and its context in different cultures: a cultural look and new look. Psychological Science, 14, 201-206.