Title Neural mechanisms underlying conscious and unconsciousattentional shifts triggered by eye gaze.

Author(s) Sato, Wataru; Kochiyama, Takanori; Uono, Shota; Toichi,Motomi

Citation NeuroImage (2015), 124(Part A): 118-126

Issue Date 2015-09-04

URL http://hdl.handle.net/2433/200278

Right

© 2015. This manuscript version is made available under theCC-BY-NC-ND 4.0 licensehttp://creativecommons.org/licenses/by-nc-nd/4.0/; The full-text file will be made open to the public on 4 September 2016in accordance with publisher's 'Terms and Conditions for Self-Archiving'.; This is not the published version. Please cite onlythe published version.; この論文は出版社版でありません。引用の際には出版社版をご確認ご利用ください。

Type Journal Article

Textversion author

Kyoto University

NeuroImage 1

Neural mechanisms underlying conscious and unconscious attentional shifts

triggered by eye gaze

Wataru Sato1 , 2 , Takanori Kochiyama1, Shota Uono2, and Motomi Toichi2 , 3

1 The Hakubi Project , Primate Research Inst i tute, Kyoto Universi ty, Aichi 484-8506,

Japan. 2 The Organizat ion for Promoting Developmental Disorder Research, 40

Shogoin-Sannocho, Sakyo, Kyoto 606-8392, Japan. 3 Faculty of Human Health

Science, Graduate School of Medicine, Kyoto Universi ty, 53 Shogoin-Kawaharacho,

Sakyo-ku, Kyoto 606-8507, Japan.

Abstract

Behav iora l s tud ies have shown tha t eye gaze t r iggers a t ten t iona l sh i f t s bo th wi th

and wi thou t consc ious awareness . However, the neura l subs t ra tes of consc ious

and unconsc ious at tent ional shif ts t r iggered by eye gaze remain unclear. To

investigate this issue, we measured brain activi ty using event-related funct ional

magnetic resonance imaging while part icipants observed averted or straight

eye-gaze cues presented supral iminally or subl iminally in the central visual f ie ld

and then localized a subsequent target in the peripheral visual f ie ld. Reaction t imes

for localizing the targets were shorter under both supral iminal and subliminal

condit ions when eye-gaze cues were direct ionally congruent with the target

locations than when they were directionally neutral . Conjunction analyses revealed

that a bi lateral cort ical network, including the middle temporal gyri , inferior

parietal lobules, anterior cingulate cort ices, and superior and middle frontal gyri ,

was act ivated more in response to averted eyes than to straight eyes under both

supraliminal and subliminal condit ions. Interaction analyses revealed that the r ight

inferior parietal lobule was specifical ly active when part icipants viewed averted

eyes relat ive to straight eyes under the supral iminal condit ion; the bi lateral

subcort ical regions, including the superior col l iculus and amygdala, and the middle

NeuroImage 2

temporal and inferior frontal gyri in the r ight hemisphere were activated in response

to averted versus straight eyes under the subliminal condit ion. These results suggest

commonali t ies and differences in the neural mechanisms underlying consc ious and

unconsc ious at tent ional shif ts t r iggered by eye gaze.

Keywords

amygdala; a t tentional shift ; conscious awareness; eye gaze; fMRI; subliminal

presentat ion

Introduction

The eyes of other individuals automatical ly t r igger mult iple psychological

act ivi t ies in observers (Kendon, 1967). For example, the perception of averted eyes

may aler t observers to cr i t ical information about the environment , such as

dangerous animals , and al low rapid react ions to such s t imuli . At the same t ime,

averted eyes may signal the intent ion to share at tent ional focus with others and

thereby create social coordinat ion.

Consis tent with these ideas, several behavioral s tudies have revealed that eye

gaze can t r igger at tentional shifts under both conscious and unconscious condit ions

(Al-Janabi and Finkbeiner, 2012; Bailey et al . , 2014; Sato et al . , 2007; Sato et al . ,

2010; Xu et al . , 2011). In these studies, researchers presented eye-gaze cues in the

central visual f ie ld ei ther supral iminal ly or subl iminal ly using a cueing paradigm

(cf . Posner, 1980). The results consistently showed that part icipants’ react ion t imes

(RTs) for processing targets were shorter when the targets were preceded by cues

that were direct ional ly congruent with the target locat ions than when they were

preceded by directionally incongruent cues under both presentat ion condit ions. This

cueing effect was observed under both supral iminal and subl iminal presentat ion

condit ions when the cues were not predict ive of target locations (Bailey et al . , 2014;

Sato et a l . , 2007; Sato et a l . , 2010; Xu et a l . , 2011; however, see Al-Janabi and

Finkbeiner, 2012) and when concurrent information load was high (Xu et al . , 2011).

NeuroImage 3

These data suggest that a t tent ional shif ts are automatical ly t r iggered by eyes,

fol lowing a common pat tern with and without conscious awareness . At the same

t ime, some of these s tudies found different pat terns of a t tent ional shif ts across

supral iminal and subl iminal presentat ion condit ions. Specif ical ly, reduced cueing

effects caused by eye cues were found under subl iminal , but not supral iminal ,

condi t ions among individuals with aut is t ic spectrum disorders compared with

typical ly developing controls (Sato et a l . , 2010) and under supral iminal , but not

subliminal , condit ions in older adults relat ive to younger adults (Bailey et al . , 2014).

These dissociat ions suggest the involvement of different mechanisms in conscious

and unconscious gaze-tr iggered at tent ional shif ts . In summary, behavioral data

suggest that both conscious and unconscious viewing of eyes can tr igger at tent ional

shif ts , with cer tain commonali t ies and differences evident across condit ions.

Neuroimaging s tudies have explored the neural mechanisms underlying

at tent ional shif ts t r iggered by consciously viewed eye gaze. A number of s tudies

reported that the observat ion of averted eyes compared with s t raight eyes or other

control s t imuli , e l ic i ted more act ivat ion in several brain regions, including the

poster ior superior temporal sulcus/middle temporal gyrus (Calder et a l . , 2002;

Engell and Haxby, 2007; Hoffman and Haxby, 2000; Hooker et al . , 2003; Pelphrey

et al . , 2003; Puce et al . , 1998; Sato et al . , 2008; Wicker et al . , 1998), fusiform gyrus

(Calder et al . , 2002; George, et al . , 2001; Pelphrey et al . , 2003; Wicker et al . , 1998),

and inferior parietal lobule (Calder et al . , 2002; Hoffman and Haxby, 2000; Pelphrey

et al . , 2003; Sato et al . , 2008; Wicker et al . , 1998), and middle/ inferior frontal gyrus

(Calder et a l . , 2002; Hooker et a l . , 2003; Wicker et a l . , 1998). Several s tudies

invest igated the brain act ivat ion in response to averted versus s t raight eyes in the

framework of the cueing paradigm (Callejas et al . , 2013; Cazzato et al . , 2012; Engell

et al . , 2010; Greene et al . , 2009; Hietanen et al . , 2006; Kingstone et al . , 2004; Sato

et al . , 2009; Tipper et al . , 2008). Although foci , methodologies, and results were not

ident ical across these s tudies and disagreements persis t , several of these s tudies

NeuroImage 4

(Greene et a l . , 2009; Sato et a l . , 2009; Tipper et a l . , 2008) were consis tent in

report ing that the temporal , par ietal , and frontal regions were involved in

at tentional shif ts t r iggered by eye gaze. For example, Tipper et al . (2008) presented

eye-gaze cues using a cueing paradigm and found act ivat ion in the dis t r ibuted

temporal , par ietal , and frontal regions, including the superior temporal gyrus,

infer ior par ietal lobule, and middle and infer ior f rontal gyri , during at tent ional

shif ts e l ic i ted by eye-gaze cues. Several neuroimaging s tudies also reported that

these regions were act ive when at tent ional shif ts were automatical ly t r iggered by

non-social cues, such as per ipheral sudden onset s t imuli and central ly presented

symbols (e.g . , Rosen et al . , 1999; for a review, see Grosbras et al . , 2005). Several

s tudies reported that s imilar brain regions were act ivated for a t tent ional shif ts ,

regardless of whether they were t riggered by eye gaze or non-social cues (Greene et

al . , 2009; Sato et al . , 2009; Tipper et al . , 2008; however, see Hietanen et al . , 2006).

Based on this evidence, i t has been proposed that these regions const i tute the

at tent ional neural network (Corbet ta and Shulman, 2002; Grosbras et a l . , 2005) .

Taken together, these f indings suggest that conscious at tentional shif ts induced by

gaze are implemented by the act ivat ion of the temporo–parieto–frontal cort ical

a t tent ional network.

However, quest ions about whether the neural mechanisms underpinning

at tentional shifts t r iggered by consciously and unconsciously perceived gaze could

be common or different remain unanswered. No study has examined this issue.

However, the aforementioned behavioral data showing commonali t ies across

conscious and unconscious gaze-tr iggered at tentional shif ts suggest commonali t ies

in the neural substrates . Several neuroimaging s tudies also reported common

patterns in neural act ivation regarding the conscious and unconscious processing of

facial s t imuli (J iang and He, 2006; Morris e t a l . , 2007; Prochnow et a l . , 2013).

Based on these data , we hypothesized that the temporo–parieto–frontal at tentional

network would be involved in both conscious and unconscious gaze-tr iggered

NeuroImage 5

at tent ional shif ts .

Addit ional ly, based on the behavioral data , we expected to f ind several

differences between conscious and unconscious gaze-tr iggered at tent ional shif ts .

Neuroimaging s tudies have provided indirect evidence related to this issue,

report ing that the amygdala was involved in the processing of subl iminal ly

presented facial s t imuli (e .g . , Morris e t a l . , 1998; Whalen et a l . , 1998) and

specif ical ly act ivated in response to subl iminal ly presented fearful versus neutral

eyes (Whalen et a l . , 2004) . A neuroimaging s tudy reported that the act ivi ty of the

amygdala of a pat ient with damage to the entrance of the cort ical visual areas

changed depending on the direct ion of unseen eyes (Burra et a l . , 2013) . An

intracranial e lectroencephalography s tudy reported that amygdala act ivat ion in

response to eyes was rapid, indicat ing that i t can occur pr ior to or s imultaneously

with the conscious awareness of faces (Sato et a l . , 2011, 2013). Several

neuroimaging s tudies also found that emotional facial expressions, which are

integrat ively processed with gaze direct ion (e .g. , Sato et a l . , 2004), were

unconsciously processed through the subcort ical visual pathway to the amygdala,

which includes the superior col l iculus and pulvinar (e.g . , Morris et al . , 2001; for a

review, see Tamietto and de Gelder, 2010). Further, the visual pathways involved in

processing conscious and unconscious emotional facial expressions differed (e .g. ,

Vuil leumier et a l . , 2002; for a review, see Vuil leumier and Pourtois , 2007). These

data suggest the involvement of subcort ical s t ructures in the unconscious

processing of eyes. Although direct evidence is lacking, based on these s tudies

together with behavioral data suggest ing specif ic mechanisms for unconscious

gaze-tr iggered at tentional shifts , we hypothesized that subcort ical s t ructures would

be specif ical ly related to unconscious at tent ional shif ts t r iggered by gaze.

To test these hypotheses, we measured brain act ivi ty using rapid event-related

functional magnetic resonance imaging (fMRI) while part icipants observed averted

and st raight eyes presented supraliminally or subliminally in the central visual f ield

NeuroImage 6

and then localized a subsequent target in the peripheral visual f ield. We performed

cognit ive conjunction analysis with interact ion masking (Price and Friston, 1997) to

ident ify commonali t ies in brain activi ty in response to averted versus st raight eyes

across presentat ion condit ions. We also examined differences in brain act ivi ty by

analyzing interact ions between gaze direct ion and presentat ion condi t ion.

Methods

Part icipants

Twenty-seven volunteers (3 women and 24 men; mean ± SD age, 25.0 ± 4.6

years) par t ic ipated in the experiment . All par t ic ipants were r ight-handed, as

assessed by the Edinburgh Handedness Inventory (Oldfield, 1971), and had normal

or corrected-to-normal visual acuity. After the experimental procedures were ful ly

explained, a l l part icipants provided informed consent regarding their part icipation.

This s tudy was approved by the local inst i tut ional e thics committee.

Experimental design

The fMRI analysis rel ied on a within-subject two-factor ial design, including

presentat ion condit ion (subl iminal or supral iminal) and direct ional condi t ion

(aver ted o r s t ra igh t ) . Cue–target congruence ( i .e . , congruence between the cue’s

direct ion and the target’s locat ion: congruent , neutral , incongruent) was also

included in the behavioral data analysis .

Stimuli

The eye-gaze s t imuli were almost ident ical to those used in a previous

behavioral s tudy (Uono et a l . , 2009) (Fig. 1) . We selected the cue s t imuli f rom a

standard set (Ekman and Friesen, 1976). Photographs of two models (one female and

one male) showing a neutral facial expression were selected and manipulated. To

manipulate gaze direct ion, the ir ises and pupils of the eyes were extracted from the

original photographs and inserted at the r ight or lef t s ide of the eyeball using Adobe

Photoshop 5.0. We cropped the photographs in an el l ipt ical shape, 2 .7° wide and

NeuroImage 7

3.8° high, to exclude hair and background.

A mosaic image was created from a neutral facial expression by dividing the

photos into a 50 × 40 gr id and randomly reordering the pieces , rendering the

result ing photograph unrecognizable as a face. The let ter “T” (0.6° wide × 0.6° high),

presented 5.7° to the lef t or r ight of the center of the screen, was used as a target

s t imulus.

Presentat ion apparatus

The events were control led by Presentat ion Software version 10.0

(Neurobehavioral Systems, Albany, CA, USA) implemented on a computer using

Microsoft Windows. The s t imuli were projected from a l iquid crystal projector

(DLA-G150CL; Victor Electronics, Brussels , Belgium) at a refresh rate of 75 Hz to

a mirror posi t ioned in a scanner in front of the part ic ipants .

Figure 1. I l lustrat ions of s t imulus presentat ions.

NeuroImage 8

Procedure

The part ic ipants completed a total of 240 t r ia ls presented in two runs of 120

t r ials last ing 427.5 s . Each run corresponded to one of the presentat ion condit ions

(supral iminal or subl iminal) , and the order of the condit ions was counterbalanced

across part ic ipants. We tested the presentat ion condit ions in separate runs to prevent

the part icipants from suspecting that eye-gaze cues were being presented under the

subliminal condit ion (i .e . , to make the part ic ipants believe that they were engaged

in two different tasks), fol lowing several previous behavioral s tudies (Bailey et al . ,

2014; Sato et a l . , 2007, 2010) . Each run consis ted of an equal number of t r ia ls

represent ing each of the gaze-direct ion condit ions ( i .e . , 40 t r ia ls each of

aver ted-left , ave r ted-right , and s t ra igh t eye-gaze cond i t ions ) and cue–target

congruence condit ions ( i .e . , 40 t r ia ls each of congruent , neutral , and incongruent

condi t ions; 20 congruen t and 20 incongruen t t r i a l s fo r each aver ted -lef t and

aver ted -r ight condi t ion and 40 neutral t r ia ls for the s t ra igh t cond i t ion), and the

order of these condit ions was randomized within each run. A break of approximately

1 min was inserted after the f i rst run. Ten practice t r ials preceded the experimental

t r ia ls .

A f ixat ion point ( i .e . , a small black “+”) was presented for 500 ms at the center

of the screen at the beginning of each t ria l (Fig. 1). The gaze cue was then presented

at the same locat ion. Under the supraliminal condit ion, the gaze cue was presented

for 200 ms, and no masking fol lowed. Under the subliminal condit ion, the gaze cue

was presented for 13 ms and was fol lowed by the presentat ion of the mask in the

same locat ion for 187 ms. Then, a target was presented in ei ther the lef t or r ight

visual f ie ld (5.0° from the center) 100 ms af ter the gaze cue (under the supral iminal

condi t ion) or mask (under the subl iminal condi t ion) disappeared; the target

remained unt i l a response was made or 1 ,700 ms elapsed. As in previous s tudies

(Friesen and Kingstone, 1998; Sato et a l . , 2007), par t ic ipants were instructed to

indicate as quickly as possible whether targets appeared on the lef t or r ight s ide of

NeuroImage 9

the monitor by pressing a key on the switch box using their lef t or r ight index f inger

to indicate left - or r ight-side locations, respectively. Part ic ipants were told that the

s t imuli preceding the targets were not predict ive.

Our experimental design was based on a rapid event-related paradigm in which

the eff ic iency of the design depended on the temporal pat tern of s t imulus or t r ia l

presentat ions (Dale, 1999; Friston et al . , 1999). We maximized the eff iciency with

which we detected different ial act ivat ion for aver ted and s traight eyes while also

maximizing the efficiency with which we est imated the evoked response under each

condit ion. To ensure the la t ter, a nul l event was included, which occurred at a

probabi l i ty of 25% of al l events . Accordingly, inter- t r ia l intervals var ied among

2,500, 5 ,000, 7 ,500, 10,000, and 12,500 ms. The eff iciency of the contrast est imates

was evaluated using the inverted t race of the covariance matr ix of the contrasts

(Dale, 1999; Friston et al . , 1999; cf . Morita et al . , 2008). Of the 200,000 randomly

generated design matrices, we selected the two most eff icient in each run under the

supral iminal and subl iminal condi t ions.

Because we tested the effect of presentat ion condit ion in separate runs, we also

evaluated the eff ic iency of this par t icular design. We compared three different

experimental designs: (1) the current design (each presentation condit ion was tested

separately in each of two runs), (2) mixed (both presentat ion condi t ions were tested

together in two runs of blocked t r ia ls) , and (3) randomized (both presentat ion

condit ions were tested together in each run, occurr ing as inter-mixed randomized

t r ia ls) . Eff iciency was evaluated using the same method as the aforementioned

simulat ion ( i .e . , 200,000 random generat ions of design matrices and an evaluation

of the inverted t race of the covariance matrix of the contrasts) . The contrasts of

interest consis ted of the different ial act ivat ion for averted versus s t raight eyes

within and across presentat ion condit ions. To take into account the effects of

temporal f i l ter ing (Friston et al . , 2000), we also computed the modified eff iciency

by applying a high-pass f i l ter of 128-s cut-off and a f i rs t-order autoregressive model

NeuroImage 10

(autoregressive coeff icient : 0 .3) . The resul ts of the s imulat ion showed that the

current design had the maximum efficiency (current > mixed > randomized) for both

types of measures.

To ensure that the subl iminal cue s t imuli were presented without the

part ic ipants’ conscious awareness , we assessed the subject ive thresholds of

part ic ipants after MRI image acquisi t ion with a procedure similar to those used in

previous behavioral s tudies (Sato et al . , 2007, 2010). The part icipants completed a

total of 30 t r ia ls , 24 of which were s imilar to the t r ia ls under the subl iminal

condit ion during image acquisi t ion, except that the gaze cues were presented for 13,

27, 40, and 53 ms in each of six t r ia ls . We also included six t r ials with no gaze cue

( i .e . , mask image only) as the basel ine condit ion ( to consider cases of false-posi t ive

responses) . The order of t r ials was randomized. Part ic ipants were asked, “Did you

see the gaze? If so, report the direction of the gaze.” Part icipants responded ei ther

“Yes” or “No;” in the case of a “Yes” response, they reported the gaze direct ion that

they had seen.

MRI acquisi t ion

Image scanning was performed on a 3-T scanning system (MAGNETOM Trio A,

Tim System; Siemens, Malvern, PA, USA) using a 12-channel head coil . A forehead

pad was used to stabi l ize the head posit ion. The functional images consisted of 40

consecutive sl ices parallel to the anterior–posterior commissure plane, covering the

whole brain. A T2*-weighted gradient-echo echo-planar imaging sequence was used

with the fol lowing parameters: repet i t ion t ime (TR) = 2,500; echo t ime (TE) = 30

ms; f l ip angle = 90°; matr ix s ize = 64 × 64; voxel s ize = 3 × 3 × 4 mm. After the

acquisi t ion of funct ional images, a T1-weighted high-resolut ion anatomical image

was obtained using a magnet izat ion-prepared rapid-acquisi t ion gradient-echo

sequence (TR = 2,250 ms; TE = 3.06 ms; f l ip angle = 9°; f ield of view = 256 × 256

mm; voxel s ize = 1 × 1 × 1 mm).

Behavioral data analysis

NeuroImage 11

The median correct RT under each condit ion was calculated for each

part ic ipant . To sat isfy normali ty assumptions for the subsequent analyses, these

data were subjected to log t ransformation. Then, the log-transformed RTs were

analyzed using a 3 (cue–target congruence: congruent , neutral , or incongruent) × 2

(presentat ion condit ion: subliminal or supral iminal) repeated-measures analysis of

variance (ANOVA). For s ignif icant interact ions, the s imple effects of cue–target

congruence were analyzed, based on our interests , using one-tai led t-s tat is t ics .

Prel iminary analyses were conducted for error percentages. The error rates were

small (<5%), and we found no evidence of a speed–accuracy t rade-off . Hence, we

report only the RT resul ts .

In terms of threshold assessment , we conducted a ser ies of paired t- tests

comparing the “Yes (seen)” responses under the no-gaze condit ion with 13-, 27-, 40-,

or 53-ms presentat ion t imes. We also examined whether correct responses under the

13- , 27- , 40- , or 53-ms presentat ion condi t ions exceeded the level of chance

(random select ion; i .e . , 25%) using one-sample t- tests .

Resul ts of a l l tes ts were considered s tat is t ical ly s ignif icant a t p < .05.

Image analysis

Image and stat ist ical analyses were performed using the stat ist ical parametric

mapping package SPM8 (ht tp: / /www.fi l . ion.ucl .ac .uk/spm), implemented in

MATLAB R2009a (MathWorks Inc. , Natick, MA, USA). Functional images of each

run were real igned using the f i rst scan as a reference to correct for head movements.

Data from al l 27 part ic ipants required only small motion correct ion (<2 mm). Then,

T1 anatomical images were coregistered to the f i rst scan of the functional images.

Fol lowing this , the coregis tered T1 anatomical image was normalized to the

Montreal Neurological Inst i tute space using the unif ied segmentat ion-spat ial

normalizat ion approach (Ashburner and Fris ton, 2005). The parameters from this

normalizat ion process were then applied to each of the functional images. Finally,

these spat ial ly normalized functional images were resampled to a voxel s ize of 2 ×

NeuroImage 12

2 × 2 and smoothed with an isotopic Gaussian kernel of 8-mm ful l -width at

half-maximum to compensate for anatomical var iabi l i ty among part ic ipants .

We used random-effects analyses to ident i fy s ignif icant ly act ivated voxels a t

the populat ion level (Holmes and Fris ton, 1998). Firs t , we performed a

s ingle-subject analysis (Fris ton et a l . , 1995). The p resen ta t ion o f each cond i t ion

was embedded in a se r i e s o f de l t a func t ions . The task- re la ted regressor was

mode led by convo lv ing i t wi th a canonica l hemodynamic response func t ion . We

used a high-pass f i l ter composed of a discrete cosine basis funct ion with a cutoff

period of 128 to el iminate the art i factual low-frequency trend. To correct the global

f luctuat ion related to motion ar t i facts , global scal ing was conducted. Serial

autocorrelat ion, assuming a f i rs t -order autoregressive model , was est imated from

the pooled act ive voxels with a restr ic ted maximum l ikel ihood procedure and was

used to whiten the data and the design matr ix (Fris ton et a l . 2002) .

Ini t ia l ly, the contrast between aver ted and s traight eyes was tested for each

presentat ion condit ion. Voxels were ident i f ied as s ignif icant ly act ivated i f they

reached a height threshold of p < .01 (uncorrected), with an extent threshold of 30

voxels (240 mm3). These analyses were conducted as exploratory analyses for the

fol lowing s tat is t ical tes ts of commonali t ies and differences.

Next , to test for commonali t ies in brain act ivi ty in response to aver ted versus

s t raight eyes across presentat ion condit ions, we performed a conjunct ion analysis

using interaction masking (Price and Friston, 1997) as in a previous s tudy (Sato et

a l . , 2009). For this analysis , we conducted a main-effect analysis of direct ional

condit ion (aver ted versus straight) using T-sta t ist ics . To search for brain areas that

showed similar act ivi ty across presentat ion condi t ions (supral iminal and

subl iminal) , the main effect was exclusively masked by the F- tests of interactions.

Voxels showing s ignif icant interact ions between effects a t a threshold of p < .05

(uncorrected) were el iminated from the s tat ist ical parametric map of the main effect .

For the main effect contrast , voxels were identif ied as s ignif icantly act ivated if they

NeuroImage 13

reached the height threshold of p < .01 (uncorrected) with the extent threshold of

100 cont iguous voxels (800 mm3), which roughly corresponded to p < .05

(corrected) determined by Monte Carlo simulat ions (e .g. , 77 voxels; Ramasubbu et

al . , 2014), to produce the best balance between Type I and Type II errors (Lieberman

and Cunningham, 2009) .

Finally, to test for differences in brain act ivi ty for ave r t ed versus st raight eyes

across presentat ion condit ions, interact ions between direct ion and presentat ion

condit ion were analyzed. We analyzed the specific instances in which higher levels

of act ivi ty were more st rongly associated with one presentat ion condi t ion than with

another. Thresholds were ident ical to those used in the aforementioned commonali ty

analysis .

The brain s t ructures were anatomical ly labeled using Talairach Client

(ht tp: / /www.talairach.org/) (Lancaster et al . , 2007) and the Automated Anatomical

Label l ing at las (Tzourio-Mazoyer et a l . , 2002) provided by the MRIcron software

(ht tp: / /www.mccauslandcenter.sc .edu/mricro/mricron/) . We also ident i f ied cer tain

brain regions based on visual inspect ion of anatomical MRI with reference to a

s tandard at las (Mai et a l . , 1997).

Results

Threshold assessment

The mean ± SE percentages of “Yes (seen)” responses were 5.6 ± 3.0, 6.2 ± 2.4,

4 .9 ± 2.3, 18.5 ± 4.1, and 40.7 ± 7.0% under the no-gaze, 13-, 27-, 40-, and 53-ms

presentat ion condit ions, respectively. We found significant differences between the

percentage of “Yes (seen)” responses under the no-gaze, 40-, and 53-ms presentat ion

condit ions ( t(26) > 2.90, p < .01) but not under the no-gaze, 13- , and 27-ms

condi t ions ( t(26) < 0.29, p > .10) .

The mean ± SE percentages of correct responses were 4.3 ± 1.7, 2 .5 ± 1.5, 12.3

± 3.2, and 30.9 ± 6.0% under the 13-, 27-, 40-, and 53-ms presentat ion condit ions,

NeuroImage 14

respect ively. The percentages of correct responses were significant ly lower than

chance under the 13-, 27-, and 40-ms presentat ion condit ions ( t(26) > 4.01, p < .001)

but not under the 53-ms presentat ion condi t ion ( t(26) = 0.98, p > .10) .

The resul ts confirmed that the subliminal cue st imuli under the current ( i .e . ,

13-ms) condit ion were presented without el ici t ing part icipants’ conscious

awareness.

RT

For the correct RT (Table 1) af ter log-transformation, the ANOVA revealed a

s ignif icant main effect of cue–target congruence and a s ignif icant interact ion of

cue–target congruence × presentat ion condi t ion (F(2,52) = 21.62 and 10.89,

respect ively, a l l p < .001) . The main effect of presentat ion condit ion was not

s ignif icant (F(1,26) = 2.25, p < .1) .

Follow-up analyses of the interact ion indicated that the RTs for congruent cues

were s ignif icant ly shorter than those for neutral cues under both supral iminal and

subl iminal presentat ion condi t ions ( t(104) = 4.51 and 1.83, p < .001 and .05,

respect ively) . The RTs were also s ignif icant ly shorter for congruent than for

incongruent cues and for neutral than for incongruent cues under the supral iminal

condit ion ( t(104) = 7.56 and 3.05, p < .001 and .005, respectively) , but not under the

subl iminal condi t ion ( t(104) < 1.35, p > .10) .

Neural act iv i ty under each presentat ion condi t ion

The contrast between aver ted and st raight eyes was tested for each presentat ion

condit ion (Table 2) . Under each presentat ion condit ion, s ignif icant act ivi t ies were

detected in the bi la teral par ietal and r ight f rontal regions, which part ia l ly

overlapped between presentat ion condit ions. Act ivi t ies in the bi la teral temporal

NeuroImage 15

NeuroImage 16

regions and anterior cingulate cort ices were found under the subliminal presentat ion

condit ion, and this was also the case under the supraliminal condit ion when a more

l iberal height threshold was used (p < .05, uncorrected) . Signif icant act ivi ty was

found in the brainstem under only the subl iminal condi t ion; the cluster included

several other adjacent regions, including the amygdala, with a more l iberal height

threshold (p < .05, uncorrected) .

Commonali t ies in neural act iv i ty

The conjunct ion analysis using interact ion masking (Price and Fris ton, 1997)

revealed that the aver ted versus s t raight eyes s ignif icant ly act ivated temporal ,

par ietal , and frontal cort ical regions commonly under the subl iminal and

supral iminal presentat ion condit ions. These regions included the middle temporal

gyrus (covering the superior temporal sulcus) in the lef t hemisphere, the infer ior

par ietal lobules and anter ior c ingulate cort ices in both hemispheres , and the

superior and middle frontal gyri in the r ight hemisphere (Table 3; Fig. 2) .

Figure 2. Stat is t ical parametr ic maps indicat ing brain regions that were

s ignif icant ly act ivated in response to averted eyes compared with s t raight eyes

under both supral iminal and subliminal presentat ion condit ions. Areas of act ivation

are rendered on spat ial ly normalized brains . L, lef t hemisphere; R, r ight

hemisphere.

NeuroImage 17

NeuroImage 18

NeuroImage 19

Figure 3. Stat ist ical parametric maps indicating brain regions that were

significantly more act ivated in response to averted versus straight eyes under the

supral iminal than under the subliminal presentat ion condit ion. Areas of act ivat ion

are rendered on spat ial ly normalized brains . R, r ight hemisphere.

Figure 4. Stat ist ical parametric maps indicating brain regions that were

signif icantly more act ivated under the subliminal than under the supral iminal

presentat ion condit ion in response to averted versus straight eyes. Areas of

act ivat ion are rendered on spatial ly normalized brains ( left ) and overlaid on the

mean normalized st ructural MRI from all part icipants in this study at the locations

of the amygdala (middle) and superior col l iculus (r ight) . L, lef t hemisphere; R, r ight

hemisphere.

NeuroImage 20

Differences in neural act ivi ty

Interact ion analysis revealed s ignif icant ly higher act ivi ty in the r ight infer ior

par ietal lobule in response to aver ted versus s t raight eyes under the supral iminal

condi t ion than under the subl iminal condi t ion (Table 4, Fig. 3) .

Significant activations for ave r t ed versus straight eyes under the subliminal

condit ion relat ive to the supral iminal condit ion were found in broad subcort ical

regions, including the superior col l iculus (the top of the midbrain; Schneider and

Kas tne r , 2005) and amygdala (Table 4, Fig. 4) . The contrast also revealed

significant act ivations in the temporal and frontal cort ices in the r ight hemisphere,

including the middle temporal gyrus (covering the superior temporal sulcus) ,

precentral gyrus, and middle and infer ior frontal gyri (Fig. 4) .

Discussion

Our behavioral results showed that congruent cues induced shorter RTs than did

neutral cues under both supraliminal and subl iminal presentat ion condit ions. These

resul ts are consistent with those of previous studies (e .g . , Ba i ley et al . , 2014) and

indicate that a t tentional shif ts can be t riggered both by supraliminally and

subl iminal ly presented eyes.

More impor tan t ly, our con junc t ion ana lyses o f fMRI da ta r evea led tha t

widespread cor t i ca l reg ions were more ac t iva ted in response to aver ted than to

s t raight eyes under both the supral iminal and subl iminal presentat ion condit ions.

These regions included the middle temporal gyrus, inferior parietal lobule, anterior

c ingulate cortex, and superior and middle frontal gyri . The act ivat ion of these

regions is consistent with the resul ts of several previous neuroimaging studies that

investigated brain act ivation associated with the observat ion of averted eyes (e.g . ,

Pe lphrey e t a l . , 2003) and at tentional shifts t r iggered by averted eyes (e.g . , Greene

et a l . , 2009). However, these previous s tudies tested only the brain act ivi t ies

associated with the processing of consciously perceived eyes. Our resul ts extend

NeuroImage 21

these previous f indings and indicate that the t emporo–par ie to–f ron ta l a t tent ional

network is involved in both conscious and unconscious gaze-tr iggered at tent ional

shif ts .

Fur the rmore , our in te rac t ion ana lyses revea led tha t severa l b ra in reg ions

were more invo lved in a t t en t iona l sh i f t s t r iggered by sub l imina l than

supra l imina l eye-gaze cues . These reg ions inc luded broad bi la teral subcort ical

regions, including the superior col l iculus and amygdala, as wel l as the middle

temporal gyrus, precentral gyrus, and middle and infer ior f rontal gyri in the r ight

hemisphere. The ac t iva t ion o f the amygda la i s consis tent with previous

neuroimaging evidence showing that the amygdala was act ive in response to

sub l imina l ly p resen ted faces (Morr i s e t a l . , 1998 ; Whalen e t a l . , 1998) and eyes

(Bur ra e t a l . , 2013 ; Whalen e t a l . , 2004) . The co-ac t iva t ion o f the superior

coll iculus and amygda la co r robora tes neuro imaging da ta showing tha t these two

reg ions were ac t iva ted and func t iona l ly coup led dur ing p rocess ing o f unseen

fac ia l s t imul i (Morr i s e t a l . , 2001; c f . Pessoa and Adolphs , 2010; Tamie t to and de

Ge lder, 2010) . The involvement of several cort ical regions is consis tent with

numerous neuroimaging studies report ing that both subcort ical and cort ical regions

were involved in the unconsc ious processing of fac ia l s t imul i (Duan et a l . , 2010;

Jiang and He, 2006; Morris et al . , 2007; Phil l ips et al . , 2004; Prochnow et al . , 2013;

Yang et a l . , 2012) . Note, however , that our f indings do not indicate that these

regions are involved in only the unconscious processing of eye gaze. Several

previous neuroimaging and neuropsychological s tudies reported that these brain

regions, including the amygdala (Akiyama et al . , 2007; Okada et al . , 2008; Sato et

a l . , 2009), superior /middle temporal gyrus (Akiyama et a l . , 2006; Greene et a l . ,

2009; Sato et al . , 2009; Tipper et al . , 2008), precentral gyrus (Greene et al . , 2009;

Tipper et al . , 2008), and middle/ inferior frontal gyrus (Sato et al . , 2009; Tipper et

a l . , 2008) , were also involved in at tent ional shif ts t r iggered by supral iminal ly

presented eye cues. Taken together with these data , our resul ts suggest that these

NeuroImage 22

subcort ical and cort ical a t tent ional networks are involved in at tent ional shif ts

t r iggered by eye gaze and are more s t rongly act ivated by unconsciously than by

consciously perceived eye-gaze s t imuli .

Our in te rac t ion ana lyses a l so revea led tha t the r igh t in fe r io r pa r ie ta l lobu le

was spec i f i ca l ly invo lved in a t t en t iona l sh i f t s t r iggered by supra l imina l ly

p resen ted eye-gaze cues . The ac t iva t ion o f the r igh t in fe r io r pa r i e ta l lobu le i s

consis tent with previous neuroimaging data showing that this region was act ive

during the processing of gaze direct ion (e .g. , Wicker e t a l . , 1998). The g rea te r

invo lvement o f the in fe r io r pa r i e ta l r eg ion in consc ious compared wi th

unconsc ious p rocess ing i s a l so cons i s t en t wi th the resu l t s o f a p rev ious

neuro imaging s tudy o f a b ra in -damaged pa t ien t , which found tha t th i s r eg ion was

ac t iva ted when the pa t i en t consc ious ly pe rce ived fac ia l s t imul i bu t no t when he

unconsc ious ly p rocessed these s t imul i (Vui l l eumier e t a l . , 2001) . Severa l

neuro imaging (Beck e t a l . , 2001) , s t imula t ion (Beck e t a l . , 2006 ; Tseng e t a l . ,

2010) , and neuro imaging + s t imula t ion (Zare t skaya e t a l . , 2010) s tud ies in

normal pa r t i c ipan t s a l so found tha t the consc ious p rocess ing o f f aces was

assoc ia ted wi th ac t iva t ion in the r igh t pa r i e ta l r eg ion . Taken toge the r wi th the

p rev ious f ind ings , our resu l t s sugges t tha t the r igh t in fer io r pa r ie ta l lobule p lays

an impor tan t ro le in gaze - t r iggered a t t en t iona l sh i f t s invo lv ing consc ious

awareness .

Our resul ts have several implications. First , the results explain behavio ra l da ta

r egard ing gaze- t r iggered a t t en t iona l sh i f t s based on the neurocogn i t ive

a rch i t ec tu re . P rev ious behav io ra l s tud ies have shown tha t supra l imina l ly and

sub l imina l ly p resen ted eyes t r iggered a t t en t iona l sh i f t s fo l lowing a common

pa t t e rn (e .g . , a t tent ional shif ts are t r iggered by unpredict ive cues under both

presentat ion condi t ions; Sato et a l . , 2007) . Our resu l t s sugges t tha t such

a t t en t iona l sh i f t s a re e l i c i t ed because bo th supra l imina l ly and sub l imina l ly

p resen ted eyes ac t iva te the cor t i ca l a t t en t iona l ne twork . Th i s concep t i s

NeuroImage 23

cons i s t en t wi th f ind ings o f neuro imaging s tud ies demons t ra t ing tha t the co r t ica l

a t t en t iona l ne twork i s ac t ive dur ing au tomat ic a t t en t iona l sh i f t s i r r e spec t ive o f

cue s t imul i ( e .g . , Sa to e t a l . , 2009) . At the same t ime , behavio ra l s tud ies showed

tha t supra l imina l ly and sub l imina l ly p resen ted eyes resu l ted in d i ffe ren t pa t t e rns

t r igger ing a t t en t iona l sh i f t s , such as specif ic impairments of gaze-tr iggered

at tentional shif ts under subliminal presentat ion condit ions among individuals with

aut ism spectrum disorders (Sato et a l . , 2010) and the opposi te pat tern in older

par t ic ipants (Bai leys et a l . , 2014). Our resul ts suggest that impa i rments o f

unconsc ious gaze - t r iggered a t t en t iona l sh i f t s in ind iv idua l s wi th au t i sm

spec t rum d i sorders may re f l ec t r educed ac t iv i ty in certain b ra in reg ions , such as

the amygda la , and those impa i rments o f consc ious gaze - t r iggered a t t en t iona l

sh i f t s in o lde r pa r t i c ipan t s may re f l ec t r educed ac t iv i ty in the in fe r io r pa r i e ta l

lobu le . These ideas a re cons i s t en t wi th ana tomica l s tud ies showing tha t g ray

mat te r vo lumes were reduced in the amygda la , bu t no t in the in fe r io r pa r i e ta l

lobu le , in au t i s t i c ind iv idua l s (Via e t a l . , 2011) and age- re la ted reduct ions were

obse rved in the neocor t i ces , inc lud ing the pa r i e ta l r eg ions , bu t no t in the

amygda la (Good e t a l . , 2001) . These exp lana t ions sugges t tha t the behav io ra l

impa i rments re la ted to gaze- t r iggered a t ten t iona l sh i f t s in these popu la t ions may

be modi f i ed by the t r ea tment o f spec i f i c b ra in reg ions (c f . S inha e t a l . , 2015) .

Second , our resul ts extend the theories of neurocogn i t ive mechan i sms fo r

a t t en t iona l sh i f t s to in tegra te unconsc ious componen t s . The major i ty o f

t r ad i t iona l theor ies have on ly ana lyzed a t t en t iona l sh i f t s wi th consc ious

awareness (e .g . , Corbe t t a and Shu lman , 2002) . Al though a few resea rchers

p roposed neura l mechan i sms under ly ing bo th consc ious and unconsc ious

a t t en t iona l sh i f t s ( e .g . , Mulckhuyse and Theeuwes, 2010) , they lacked ev idence .

Our resu l t s p rov ide empi r i ca l suppor t sugges t ing tha t the t r ad i t iona l theory o f

t emporo–par ie to–f ron ta l co r t i ca l ne twork fo r consc ious a t t en t iona l sh i f t s ( e .g . ,

Corbe t t a and Shu lman , 2002) can accommodate unconsc ious a t t en t iona l sh i f t s

NeuroImage 24

with the add i t ion o f some componen t s spec i f i c to consc ious and unconsc ious

a t t en t iona l sh i f t s .

Final ly, our f indings have implicat ions for the relat ionship between a t t en t ion

and consc iousness . Al though t r ad i t iona l theor ies have pos i t ed tha t a t t en t ion and

consc iousness a re t igh t ly coup led (c f . Posner, 1994) , i t has a l so been p roposed

tha t these two componen t s may re f l ec t two d i s t inc t neura l p rocesses (Koch and

Tsuch iya , 2007) . Our da ta suppor t th i s p roposa l by iden t i fy ing the neura l

subs t ra tes o f a t t en t ion tha t a re and a re no t accompanied by consc ious awareness .

Several l imitat ions of the present s tudy should be acknowledged. Firs t , the

differences in presentat ion durat ion between subl iminal ly and supral iminal ly

presented eyes may at least par t ia l ly explain the present resul ts . Although i t is

probably not possible to explain the commonali ty of the resul ts observed across

presentat ion condi t ions and those specif ic to the subl iminal condi t ion, the long

presentat ion of the averted gaze may be relevant to the supral iminal-specif ic

act ivat ion in the r ight infer ior par ietal lobule . This issue could be eff ic ient ly

invest igated using subliminal methods with longer st imulus presentat ion durations,

such as the continuous f lash suppression technique (Tsuchiya and Koch, 2005; Xu et

a l . , 2011) .

Second, al though we found increased act ivi ty for averted versus st raight eyes

under both supral iminal and subliminal condit ions, these results may be at tr ibutable

to the task we used. Although, as mentioned above, our resul ts under the

supral iminal condi t ion are consis tent with those of several previous s tudies (e .g. ,

Greene et a l . , 2009), other s tudies reported different pat terns of brain act ivi t ies

using different tasks; for example, st raight rather than averted eyes combined with

emotional expressions (Sato et al . , 2004) and social messages (Pelphrey et al . , 2004)

el ic i ted st ronger activi ty in the amygdala and superior temporal sulcus, respectively.

To account for such discrepancies , we speculate that the relat ive s ignif icance of

averted and st raight eyes can change depending on the si tuation (cf . Wicker et a l . ,

NeuroImage 25

1998) and that the eye-gaze st imuli in the cueing task we used may have increased

the significance of averted eyes relat ive to that of st raight eyes. To test such ideas ,

further invest igations using a variety of different tasks are necessary to determine

the neural mechanisms involved in the conscious and unconscious processing of eye

gaze.

Final ly, a l though the current fMRI resul ts suggest that brain regions are

rapidly act ivated in response to subl iminally presented eyes, the exact t iming of the

brain act ivat ion remains unclear. Previous electrophysiological s tudies

invest igat ing supral iminal ly presented gaze have consis tent ly reported that the

temporal regions around the middle temporal gyrus show higher act ivi ty in response

to aver ted versus straight eyes in a component peaking at around 200 ms (Caruana

et al . , 2014; Hietanen et al . , 2008; McCarthy et al . , 1999; Puce et al . , 2000; Sato et

al . , 2008; Uono et al . , 2014; Watanabe et al . , 2001; however, see Conty et al . , 2007).

However, to date , no electrophysiological s tudy has tested at tent ional shif ts

t r iggered by subl iminal ly presented gaze in comparison with at tent ional shif ts

induced by supral iminal ly presented gaze. Future electrophysiological s tudies are

needed to fur ther understand the commonali t ies and differences in the

neuro-cognit ive mechanisms involved in conscious and unconscious gaze-tr iggered

at tent ional shif ts .

In conc lus ion , our conjunct ion analyses revealed that averted versus s t raight

eyes act ivated the cort ical a t tent ional network under both supral iminal and

subl iminal condit ions. Interact ion analyses revealed that averted versus s t raight

eyes different ial ly act ivated the cort ical and subcort ical regions across supraliminal

and subl iminal presentat ion condit ions. These resul ts suggest commonali t ies and

differences in the neural mechanisms underlying consc ious and unconsc ious

at tent ional shif ts t r iggered by eye gaze.

Acknowledgments

NeuroImage 26

This s tudy was supported by funds from the Benesse Corporat ion, the Japan

Society for the Promotion of Science Funding Program for Next Generat ion

World-Leading Researchers (LZ008), and the Organization for Promoting Research

in Developmental Disorders .

Confl ict of interest statement

The authors declare no competing f inancial or other interests .

References

Akiyama, T. , Ka to , M. , Muramatsu , T. , Sa i to , F. , Umeda , S . , Kash ima , H. , 2006 .

Gaze bu t no t a r rows : A d i s soc ia t ive impa i rment a f t e r r igh t super io r

t empora l gyrus damage . Neuropsycho log ia 44 , 1804–1810 .

Akiyama, T. , Ka to , M. , Muramatsu , T. , Umeda , S . , Sa i to , F. , Kash ima , H. , 2007 .

Uni la te ra l amygda la le s ions hamper a t t en t iona l o r i en t ing t r iggered by gaze

d i rec t ion . Cereb . Cor tex 17 , 2593–2600 .

Al - Janab i , S . , F inkbe iner, M. , 2012 . Effec t ive p rocess ing o f masked eye gaze

requ i res vo l i t iona l con t ro l . Exp . Bra in Res . 216 , 433–443 .

Ashburner, J . , F r i s ton , K .J . , 2005 . Uni f ied segmenta t ion . Neuro image 26 ,

839–851 .

Ba i ley, P.E . , S lessor, G. , Rendel l , P.G. , Benne t t s , R . J . , Campbe l l , A . , Ruffman , T. ,

2014 . Age d i ffe rences in consc ious ve r sus subconsc ious soc ia l pe rcep t ion :

The in f luence o f f ace age and va lence on gaze fo l lowing . Psycho l . Ag ing 29

491–502 .

Beck , D .M. , Muggle ton , N . , Walsh , V. , Lav ie , N . , 2006 . R ight pa r ie t a l cor tex

p lays a c r i t i ca l ro le in change b l indness . Cereb . Cor tex 16 , 712–717 .

Beck , D .M. , Rees , G. , F r i th , C .D. , Lav ie , N . , 2001 . Neura l cor re la tes o f change

de tec t ion and change b l indness . Na t . Neurosc i . 4 , 645–650 .

Bur ra , N . , Herva i s -Ade lman , A. , Kerze l , D . , Tamie t to , M. , de Ge lder, B . , Pegna ,

NeuroImage 27

A.J . , 2013 . Amygda la ac t iva t ion fo r eye con tac t desp i te comple te cor t ica l

b l indness . J . Neurosc i . 33 , 10483–10489 .

Ca lder, A .J . , Lawrence , A .D. , Keane , J . , Sco t t , S .K. , Owen , A.M. , Chr i s to ffe l s , I . ,

Young , A .W. , 2002 . Read ing the mind f rom eye gaze . Neuropsychologia 40 ,

1129–1138 .

Ca l le jas , A. , Shu lman , G.L . , Corbe t t a , M. , 2013 . Dorsa l and vent ra l a t t en t ion

sys tems under l ie soc ia l and symbol ic cue ing . J . Cogn . Neurosc i . 26 , 63–80.

Caruana , F. , Can ta lupo , G. , Lo Russo , G. , Mai , R . , Sa r to r i , I . , Avanz in i , P. , 2014 .

Human cor t i ca l ac t iv i ty evoked by gaze sh i f t obse rva t ion : an in t rac ran ia l

EEG s tudy. Hum. Bra in Mapp . 35 , 1515–1528 .

Cazza to , V. , Macaluso , E . , Cros te l l a , F. , Agl io t i , S .M. , 2012 . Mapping re f l ex ive

sh i f t s o f a t t en t ion in eye-cen te red and hand-cente red coord ina te sys tems .

Hum. Bra in Mapp . 33 , 165–178 .

Conty, L . , N 'Diaye , K . , Ti jus , C . , George , N . , 2007 . When eye c rea tes the con tac t !

ERP ev idence fo r ea r ly d i s soc ia t ion be tween d i rec t and aver ted gaze mot ion

p rocess ing . Neuropsycho log ia 45 , 3024–3037 .

Corbe t t a , M. , Shu lman , G.L . , 2002 . Cont ro l o f goa l -d i rec ted and s t imulus -d r iven

a t t en t ion in the b ra in . Na t . Rev. Neurosc i . 3 , 201–215 .

Da le , A .M. , 1999 . Opt imal exper imenta l des ign fo r even t - re la ted fMRI . Hum.

Bra in Mapp . 8 , 109–114 .

Duan , X . , Da i , Q . , Gong , Q . , Chen , H . , 2010 . Neura l mechan i sm of unconsc ious

pe rcep t ion o f su rpr i sed fac ia l express ion . Neuro image 52 , 401–407 .

Ekman , P. , F r i esen , W.V. , 1976 . P ic tu res o f Fac ia l Affec t . Ca l i fo rn ia : Consu l t ing

Psycho log i s t .

Engel l , A .D. , Haxby, J .V. , 2007 . Fac ia l express ion and gaze-d i rec t ion in human

super io r t empora l su lcus . Neuropsycho log ia 45 , 3234–3241 .

Enge l l , A .D. , Nummenmaa , L . , Oos te rhof , N .N. , Henson , R .N. , Haxby, J .V. ,

Ca lder, A .J . , 2010 . Di ffe ren t i a l ac t iva t ion o f f ron topar ie ta l a t t en t ion

NeuroImage 28

ne tworks by soc ia l and symbol ic spa t ia l cues . Soc . Cogn . Affec tNeurosc i .

5 , 432–440 .

F r iesen , C .K. , Kings tone , A . , 1998 . The eyes have i t ! Ref lex ive o r ien t ing i s

t r iggered by nonpred ic t ive gaze . Psychon . Bu l l . Rev. 5 , 490–495 .

F r i s ton , K .J . , Glase r, D .E . , Henson , R.N. , Kiebe l , S . , Ph i l l ips , C . , Ashburner, J . ,

2002 . C lass ica l and Bayes ian in fe rence in neuro imaging: Appl ica t ions .

Neuro image 16 , 484–512 .

Fr i s ton , K .J . , Holmes , A .P. , Po l ine , J .B . , Grasby, P. J . , Wi l l i ams , S .C . ,

F rackowiak , R.S . J . , Turner, R . , 1995 . Ana lys i s o f fMRI t ime-se r i es

r ev i s i t ed . Neuro image 2 , 45–53 .

Fr i s ton , K .J . , Josephs , O . , Zarahn , E . , Holmes , A .P. , Rouque t te , S . , Po l ine , J . ,

2000 . To smooth o r no t to smooth? Bias and e ff i c i ency in fMRI t ime-se r i es

ana lys i s . Neuro image 12 , 196–208 .

F r i s ton , K. J . , Zarahn , E . , Josephs , O . , Henson , R.N. , Da le , A .M. , 1999 .

Stochas t i c des igns in even t - re la ted fMRI . Neuro image 10 , 607–619 .

George , N . , Dr ive r, J . , Dolan , R. J . , 2001 . Seen gaze-d i rec t ion modula tes fus i form

ac t iv i ty and i t s coupl ing wi th o the r b ra in a reas dur ing face p rocess ing .

Neuro image 13 , 1102–1112 .

Good , C .D. , Johns rude , I .S . , Ashburner, J . , Henson , R .N. , F r i s ton , K .J . ,

F rackowiak , R.S . , 2001 . A voxel -based morphomet r i c s tudy o f age ing in

465 normal adu l t human b ra ins . Neuro image 14 , 21–36 .

Greene , D .J . , Mooshag ian , E . , Kap lan , J .T. , Za ide l , E . , I acoboni , M. , 2009 . The

neura l co r re la tes o f soc ia l a t t en t ion : Automat ic o r i en t ing to soc ia l and

nonsoc ia l cues . Psycho l . Res . 73 , 499–511 .

Grosbras , M.H. , La i rd , A .R. , Paus , T. , 2005 . Cor t i ca l r eg ions invo lved in eye

movements , sh i f t s o f a t t en t ion , and gaze pe rcep t ion . Hum. Bra in Mapp. 25 ,

140–154

Hie tanen , J .K . , Leppanen , J .M. , Nummenmaa , L . , As t ika inen , P. , 2008 .

NeuroImage 29

Visuospa t ia l a t t en t ion sh i f t s by gaze and a r row cues : An ERP s tudy. Bra in

Res . 1215 , 123–136 .

Hie tanen , J .K . , Nummenmaa , L . , Nyman, M.J . , Pa rkko la , R . , Hamala inen , H . ,

2006 . Automat ic a t t en t ion o r i en t ing by soc ia l and symbol ic cues ac t iva t es

d i ffe ren t neura l ne tworks : An fMRI s tudy. Neuro image 33 , 406–413 .

Hoffman , E .A. , Haxby, J .V. , 2000 . Dis t inc t r epresen ta t ions o f eye gaze and

iden t i ty in the d i s t r ibu ted human neura l sys tem fo r f ace pe rcep t ion . Na t .

Neurosc i . 3 , 80–84 .

Holmes , A .P. , F r i s ton , K . J . , 1998 . Genera l i sab i l i ty, r andom effec t s and

popu la t ion in fe rence . Neuro image 7 , S754 .

Hooker, C . I . , Pa l le r, K .A. , Gi te lman , D .R. , Pa r r i sh , T.B . , Mesulam, M.M. , Reber,

P. J . , 2003 . Bra in ne tworks fo r ana lyz ing eye gaze . Bra in Res . Cogn . Bra in

Res . 17 , 406–418 .

J i ang , Y. , He , S . , 2006 . Cor t ica l r e sponses to inv i s ib le faces : Ddissoc ia t ing

subsys tems fo r f ac ia l - in fo rmat ion p rocess ing . Cur r. B io l . 16 , 2023–2029 .

Kendon , A . , 1967 . Some func t ions o f gaze -d i rec t ion in soc ia l in te rac t ion . Ac ta

Psycho l . 26 , 22–63 .

Kings tone , A . , Tipper, C . , R i s t i c , J . , Ngan, E . , 2004 . The eyes have i t ! : An fMRI

inves t iga t ion . Bra in Cogn . 55 , 269–271 .

Koch , C. , Tsuch iya , N . , 2007 . At ten t ion and consc iousness : Two d i s t inc t b ra in

p rocesses . Trends Cogn . Sc i . 11 , 16–22 .

Lancas te r, J .L . , Tordes i l l a s -Gut ié r rez , D . , Mar t inez , M. , Sa l inas , F. , Evans , A . ,

Z i l l e s , K . , Mazz io t t a , J .C . , Fox , P.T. , 2007 . B ias be tween MNI and

Ta la i rach coord ina tes ana lyzed us ing the ICBM-152 b ra in templa te . Hum.

Bra in Mapp . 28 , 1194–1205 .

L ieberman , M.D. , Cunn ingham, W.A. , 2009 . Type I and Type I I e r ro r concerns in

fMRI resea rch : r e -ba lanc ing the sca le . Soc . Cogn . Affec tNeurosc i . 4 ,

423–428 .

NeuroImage 30

Mai , J .K . , Assheuer, J . , Pax inos , G. , 1997 . At las o f the human bra in . San Diego :

Academic Press .

McCar thy, G. , Puce , A . , Be lger, A . , Al l i son , T. 1999 . Elec t rophys io log ica l

s tud ies o f human face pe rcep t ion . I I : Response p roper t i e s o f f ace -spec i f ic

po ten t i a l s genera ted in occ ip i to tempora l cor tex . Cereb . Cor tex 9 , 431–444 .

Mor i t a , T. , I t akura , S . , Sa i to , D .N. , Nakash i t a , S . , Harada , T. , Koch iyama, T. ,

Sada to , N . , 2008 . The ro le o f the r igh t p re f ron ta l co r tex in se l f -eva lua t ion

o f the face : A func t iona l magne t i c r e sonance imaging s tudy. J . Cogn.

Neurosc i . 20 , 342–355 .

Morr i s , J .S . , de Ge lder, B . , Weiskran tz , L . , Dolan , R . J . , 2001 . Di ffe ren t i a l

ex t ragen icu los t r i a te and amygda la responses to p resen ta t ion o f emot ional

f aces in a co r t i ca l ly b l ind f i e ld . Bra in 124 , 1241–1252 .

Morr i s , J .S . , Öhman, A . , Dolan , R . J . , 1998 . Consc ious and unconsc ious

emot iona l l ea rn ing in the human amygda la . Na tu re 393 , 467–470 .

Morr i s , J .P. , Pe lphrey, K.A. , McCar thy, G. , 2007 . Face p rocess ing wi thou t

awareness in the r igh t fus i fo rm gyrus . Neuropsycho log ia 45 , 3087–3091 .

Mulckhuyse , M. , Theeuwes , J . , 2010 . Unconsc ious a t ten t iona l o r ien t ing to

exogenous cues : A rev iew of the l i t e ra ture . Ac ta Psycho l . (Amst ) 134 ,

299–309 .

Okada , T. , Sa to , W. , Kubota , Y. , Usu i , K. , Inoue , Y. , Mura i , T. , Hayash i , T. ,

To ich i , M. , 2008 . Invo lvement o f media l t empora l s t ruc tu res in re f l ex ive

a t t en t iona l sh i f t by gaze . Soc . Cogn . Affec t . Neurosc i . 3 , 80–88 .

Oldf ie ld , R .C . , 1971 . The assessment and ana lys i s o f handedness : The Ed inburgh

inven to ry. Neuropsycho log ia 9 , 97–113 .

Pe lphrey, K .A. , S ingerman , J .D . , Al l i son , T. , McCar thy, G. , 2003 . Bra in

ac t iva t ion evoked by pe rcep t ion o f gaze sh i f t s : The in f luence o f con tex t .

Neuropsycho log ia 41 , 156–170 .

Pe lphrey, K. A . , Vio la , R . J . , McCar thy, G. 2004 . When s t rangers pass : P rocess ing

NeuroImage 31

of mutua l and aver ted soc ia l gaze in the super io r tempora l su lcus . Psycho l .

Sc i . 15 , 598–603 .

Pessoa , L . , Adolphs , R . , 2010 . Emot ion p rocess ing and the amygda la : F rom a ' l ow

road ' to 'many roads ' o f eva lua t ing b io log ica l s ign i f i cance . Na t . Rev.

Neurosc i . 11 , 773–783 .

Ph i l l ips , M.L . , Wi l l iams , L .M. , He in ing , M. , Herba , C .M. , Russe l l , T. , Andrew,

C . , Bu l lmore , E .T. , Brammer, M.J . , Wi l l i ams , S .C . , Morgan , M. , Young ,

A .W. , Gray, J .A . , 2004 . Di ffe ren t i a l neura l r esponses to over t and cover t

p resen ta t ions o f f ac ia l express ions o f f ea r and d i sgus t . Neuro image 21 ,

1484–1496 .

Posner, M. I . , 1980 . Or ien t ing o f a t t en t ion . Q . J . Exp . Psycho l . 32 , 3–25 .

Posner, M.I . , 1994 . At ten t ion : The mechan i sms o f consc iousness . P roc . Na t l .

Acad . Sc i . U . S . A . 91 , 7398–7403 .

P r ice , C . J . , F r i s ton , K .J . , 1997 . Cogn i t ive con junc t ion : A new approach to b ra in

ac t iva t ion exper iments . Neuro image 5 , 261–270 .

P rochnow, D. , Kossack , H . , Brunhe im, S . , Mul le r, K . , Wi t t sack , H .J . ,

Markowi t sch , H . J . , Se i t z , R . J . , 2013 . P rocess ing of sub l imina l f ac ia l

express ions o f emot ion : A behav iora l and fMRI s tudy. Soc . Neurosc i . 8 ,

448–461 .

Puce , A . , Al l i son , T. , Bent in , S . , Gore , J .C . , McCar thy, G. , 1998 . Tempora l cor tex

ac t iva t ion in humans v iewing eye and mouth movements . J . Neurosc i . 18 ,

2188–2199 .

Puce , A . , Smi th , A . , Al l i son , T. , 2000 . , ERPs evoked by v iewing fac ia l

movements . Cogn . Neuropsycho l . 17 , 221–239 .

Ramasubbu, R . , Konduru , N . , Cor tese , F. , Bray, S . , Gax io la -Va ldez , I . , Goodyear,

B . , 2014 . Reduced in t r ins ic connec t iv i ty o f amygda la in adul t s wi th major

depress ive d i so rder. F ron t . Psych ia t ry 5 , 17 .

Rosen , A .C. , Rao , S .M. , Caffa r ra , P. , Scagl ion i , A . , Bobholz , J .A . , Woodley, S . J . ,

NeuroImage 32

Hammeke , T.A. , Cunningham, J .M. , P r ie to , T.E . , B inder, J .R . , 1999 . Neura l

bas i s o f endogenous and exogenous spa t i a l o r ien t ing . A func t iona l MRI

s tudy. J . Cogn . Neurosc i . 11 , 135–152 .

Sa to , W. , Koch iyama, T. , Uono , S . , Matsuda , K . , Usu i , K . , Inoue , Y. , Toich i , M. ,

2011 . Rap id amygda la gamma osc i l la t ions in r esponse to eye gaze . PLoS

One 6 , e28188 .

Sa to , W. , Koch iyama, T. , Uono , S . , Matsuda , K . , Usu i , K . , Inoue , Y. , Toich i , M. ,

2013 . Rapid and mul t ip le - s t age ac t iva t ion o f the human amygda la fo r

p rocess ing fac ia l s igna l s . Commun. In tegr. B io l . 6 , e24562 .

Sa to , W. , Koch iyama, T. , Uono , S . , Yosh ikawa, S . , 2008 . Time course o f super io r

t empora l su lcus ac t iv i ty in response to eye gaze : A combined fMRI and

MEG s tudy. Soc . Cogn . Affec t . Neurosc i . 3 , 224–232 .

Sa to , W. , Koch iyama , T. , Uono , S . , Yosh ikawa , S . , 2009 . Commonal i t i e s in the

neura l mechan i sms under ly ing au tomat ic a t t en t iona l sh i f t s by gaze ,

ges tu res , and symbols . Neuro image 45 , 984–992 .

Sa to , W. , Okada , T. , To ich i , M. , 2007 . At ten t iona l sh i f t i s t r iggered by gaze

wi thou t awareness . Exp . Bra in Res . 183 , 87–94 .

Sa to , W. , Uono , S . , Okada , T. , To ich i , M. , 2010 . Impa i rment o f unconsc ious , bu t

no t consc ious , gaze - t r iggered a t t en t ion o r ien t ing in Asperger ' s d i so rder.

Res . Aut i sm Spec t r. Disord . 4 , 782–786 .

Sa to , W. , Yosh ikawa , S . , Koch iyama, T. , Matsumura , M. , 2004 . The amygda la

p rocesses the emot iona l s ign i f icance o f fac ia l express ions : An fMRI

inves t iga t ion us ing the in te rac t ion be tween express ion and face d i rec t ion .

Neuro image 22 , 1006–1013 .

Schneider, K .A. , Kas tne r, S . , 2005 . Visua l r esponses o f the human super io r

co l l i cu lus : A h igh- reso lu t ion func t iona l magne t i c r e sonance imaging s tudy.

J . Neurophys io l . 94 , 2491–2503 .

S inha , S . , McGovern , R .A. , She th , S .A. , 2015 . Deep b ra in s t imula t ion fo r severe

NeuroImage 33

au t i sm: From pa thophys io logy to p rocedure . Neurosurg . Focus 38 , E3 .

Tamie t to , M. , de Ge lder, B . 2010 . Neura l bases o f the non-consc ious pe rcep t ion

o f emot iona l s igna l s . Na t . Rev. Neurosc i . 11 , 697–709 .

Tipper, C .M. , Handy, T.C . , Giesbrech t , B. , Kings tone , A . , 2008 . Bra in responses

to b io log ica l r e l evance . J . Cogn . Neurosc i . 20 , 879–891 .

Tseng , P. , Hsu , T.Y. , Muggle ton , N .G. , Tzeng , O .J . , Hung , D .L . , Juan , C .H. , 2010 .

Pos te r io r pa r ie ta l co r tex media tes encoding and main tenance p rocesses in

change b l indness . Neuropsycho log ia 48 , 1063–1070 .

Tsuch iya , N. , Koch , C . , 2005 . Cont inuous f l ash suppress ion reduces nega t ive

a f t e r images . Na t . Neurosc i . 8 , 1096–1101 .

Tzour io -Mazoyer, N . Landeau , B. , Papa thanass iou , D . , Cr ive l lo , F. , E ta rd , O . ,

De lc ro ix , N . , Mazoyer, B . , Jo l io t , M. , 2002 . Automated ana tomica l l abe l ing

o f ac t iva t ions in SPM us ing a macroscop ic ana tomica l pa rce l l a t ion o f the

MNI MRI s ing le - sub jec t b ra in . Neuro image 15 , 273–289 .

Uono , S . , Sa to , W. , Koch iyama, T. , 2014 . Commonal i t i es and d i ffe rences in the

spa t io tempora l neura l dynamics assoc ia ted wi th au tomat ic a t t en t iona l

sh i f t s induced by gaze and a r rows . Neurosc i . Res . 87 , 56–65 .

Uono , S . , Sa to , W. , To ich i , M. , 2009 . Dynamic fea r fu l express ions enhance

gaze- t r iggered a t ten t ion o r ien t ing in h igh and low anx ie ty ind iv idua ls . Soc .

Behav. Pe rs . 37 , 1313–1326 .

Via , E . , Radua , J . , Cardoner, N . , Happe , F. , Mata ix -Cols , D . , 2011 . Meta -ana lys i s

o f g ray mat te r abnormal i t i e s in au t i sm spec t rum d i sorder : should Asperger

d i so rder be subsumed under a b roader umbre l l a o f au t i s t ic spec t rum

d i sorder? Arch . Gen . Psych ia t ry 68 , 409–418 .

Vui l l eumier, P. , Armony, J .L . , Cla rke , K . , Husa in , M. , Dr ive r, J . , Dolan , R.J . ,

2002 . Neura l r e sponse to emot iona l f aces wi th and wi thou t awareness :

even t - re la ted fMRI in a pa r i e ta l pa t i en t wi th v i sua l ex t inc t ion and spa t i a l

neg lec t . Neuropsycho log ia 40 , 2156–2166 .

NeuroImage 34

Vui l l eumier, P. , Pour to i s , G. , 2007 . Dis t r ibu ted and in te rac t ive b ra in mechan i sms

dur ing emot ion face pe rcep t ion : ev idence f rom func t iona l neuro imaging .

Neuropsycho log ia 45 , 174–194 .

Vui l l eumier, P. , Sag iv, N . , Haze l t ine , E . , Po ldrack , R .A. , Swick , D . , Rafa l , R .D. ,

Gabr ie l i , J .D . , 2001 . Neura l f a te o f seen and unseen faces in v i suospa t i a l

neg lec t : A combined even t - re la ted func t iona l MRI and event - re la ted

po ten t i a l s tudy. P roc . Na t l . Acad . Sc i . U . S . A . 98 , 3495–3500 .

Watanabe , S . , Kak ig i , R . , Puce , A . , 2001 . Occ ip i to tempora l ac t iv i ty e l i c i t ed by

v iewing eye movemen ts : A magnetoencepha lograph ic s tudy. Neuro image 13 ,

351–363 .

Whalen , P. J . , Kagan , J . , Cook , R.G. , Dav is , F.C . , Kim, H. , Po l i s , S . , McLaren ,

D .G. , Somerv i l le , L .H. , McLean , A .A. , Maxwel l , J .S . , Johns tone , T. , 2004 .

Human amygda la respons iv i ty to masked fea r fu l eye whi tes . Sc ience 306 ,

2061 .

Whalen , P. J . , Rauch , S .L . , E tcoff , N .L . , McInerney, S .C . , Lee , M.B. , Jen ike ,

M.A. , 1998 . Masked p resen ta t ions o f emot iona l fac ia l express ions

modula te amygda la ac t iv i ty wi thou t exp l i c i t knowledge . J . Neurosc i . 18 ,

411–418 .

Wicker, B . , Miche l , F. , Henaff , M.A. , Dece ty, J . , 1998 . Bra in reg ions invo lved in

the pe rcep t ion o f gaze : A PET s tudy. Neuro image 8 , 221–227 .

Xu , S . , Zhang , S . , Geng , H . , 2011 . Gaze- induced jo in t a t ten t ion pe rs i s t s under

h igh pe rcep tua l load and does no t depend on awareness . Vis ion Res . 51 ,

2048–2056 .

Yang , J . , Cao , Z . , Xu , X . , Chen , G. , 2012 . , The amygda la i s invo lved in a ffec t ive

p r iming e ffec t fo r f ea r fu l f aces . Bra in Cogn . 80 , 15–22 .

Zare t skaya , N . , Th ie l scher, A . , Logothe t i s , N .K. , Bar te l s , A . , 2010 . Dis rupt ing

pa r ie ta l func t ion p ro longs dominance dura t ions in b inocu la r r iva l ry. Cur r.

B io l . 20 , 2106–2111 .