Priming in a shape task but not in a category task undercontinuous flash suppression

Guido Hesselmann $

Visual Perception LaboratoryDepartment of Psychiatry and PsychotherapyCharitendashUniversitatsmedizin Berlin Germany

Natasha DarcyVisual Perception Laboratory

Department of Psychiatry and PsychotherapyCharitendashUniversitatsmedizin Berlin Germany

Karin Ludwig

Visual Perception LaboratoryDepartment of Psychiatry and PsychotherapyCharitendashUniversitatsmedizin Berlin Germany

Department of PsychologyHumboldt-Universitat zu Berlin Berlin Germany

Philipp Sterzer

Visual Perception LaboratoryDepartment of Psychiatry and PsychotherapyCharitendashUniversitatsmedizin Berlin Germany

Continuous flash suppression (CFS) is an interocularsuppression technique that uses high-contrast masksflashed to one eye to prevent conscious perception ofimages shown to the other eye It has become widely useddue to its strength and prolonged duration of suppressionand its nearly deterministic control of suppression onsetand offset Recently it has been proposed that action-relevant visual processing ascribed to the dorsal streamremains functional while processing in the ventral streamis completely suppressed when stimuli are invisible underCFS Here we tested the hypothesis that the potentiallydorsal-stream-based analysis of prime-stimulus elongationduring CFS affects the categorization of manipulable targetobjects In two behavioral experiments we found evidencefor priming in a shape task but none for priming in acategory task when prime stimuli were rendered invisibleusing CFS Our results thus support the notion that therepresentation of CF-suppressed stimuli is more limitedthan previously thought

Introduction

For decades visual processing at the verge ofconscious perception has been an intense area of

research in the scientific study of consciousness (De-haene amp Changeux 2011 Kouider amp Dehaene 2007)Psychophysical and neurophysiological research inhealthy observers has used various noninvasive experi-mental lsquolsquoblindingrsquorsquo methods to suppress visual input fromconsciousness and examine which features of images andobjects are processed on an unconscious level but canstill induce effects on subsequent behavior (BachmannBreitmeyer amp Ogmen 2007 Kim amp Blake 2005) Arecently introduced method called continuous flashsuppression (CFS) has become widely used1 due to itsstrength of suppression as well as its prolongedsuppression duration of up to several seconds andalmost perfect temporal control of suppression onsetand offset (Tsuchiya amp Koch 2005 Tsuchiya KochGilroy amp Blake 2006) CFS is an interocular suppres-sion technique that uses high-contrast dynamic maskson one eye to suppress perception of images presented tothe other eye It has been argued that CFS will act as alsquolsquogame changerrsquorsquo in consciousness research due to itsseemingly unlimited potential to present various kinds ofvisual stimuli unconsciously for extended periods of time(Sklar et al 2012 but see Hesselmann amp Moors 2015)However the level of visual processing which remainsintact under CFS is a controversial topic (Gayet Van

Citation Hesselmann G Darcy N Ludwig K amp Sterzer P (2016) Priming in a shape task but not in a category task undercontinuous flash suppression Journal of Vision 16(3)17 1ndash17 doi10116716317

Journal of Vision (2016) 16(3)17 1ndash17 1

doi 10 1167 16 3 17 ISSN 1534-7362Received July 27 2015 published February 16 2016

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 40 International LicenseDownloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

der Stigchel amp Paffen 2014 Hesselmann 2013 SterzerStein Ludwig Rothkirch amp Hesselmann 2014 YangBrascamp Kang amp Blake 2014)

Recently Breitmeyer (2014b 2015) reviewed most ofthe standard suppression methods and placed them in afunctional hierarchy of unconscious visual processingIn his review article (2015) Breitmeyer locates CFS inthe midrange of the functional hierarchy arguing thatit leaves semantic processing intact though previouslyit was believed to disrupt even early visual processingand thus ranked much lower on the scale in Breit-meyerrsquos earlier book chapter (2014b) Beyond thequestion of CFSrsquos absolute and relative placementwithin the functional hierarchy another hypothesisproposes that action-relevant visual processing ascribedto the dorsal stream remains functional while pro-cessing in the ventral stream is completely suppressedwhen stimuli are invisible under CFS (Lin amp He 2009Ludwig amp Hesselmann 2015) The finding of largelypreserved dorsal activity in a functional magneticresonance imaging (fMRI) study (Fang amp He 2005)inspired a series of behavioral priming experimentsshowing that brief presentations of manufacturedmanipulable objects (tools) that were rendered invisibleby CFS influenced subsequent responses to fully visibletool images in a two-alternative forced-choice (2AFCtool vs animal) category-discrimination task (AlmeidaMahon amp Caramazza 2010 Almeida Mahon Na-kayama amp Caramazza 2008) Specifically elongatedtool primes (eg hammer) elicited faster categorizationresponses for elongated tool targets compared withnonelongated animal primes (eg cow) Priming effectsfor animal targets were not significant

The conclusion of a CFS priming effect specific tothe action-relevant tool category was challenged by theresult that elongated stimuli of different categories(images of vegetables and geometric lines) had a similarfacilitatory influence on response times (SakurabaSakai Yamanaka Yokosawa amp Hirayama 2012)This finding suggests that such seemingly high-levelcategory effects may be explained by basic visualproperties such as elongation (Hebart amp Hesselmann2012) In their most recent set of experiments measur-ing priming as well as reaching trajectories Almeida etal (2014) argued against a shape-priming account andinstead proposed that lsquolsquothis dorsal-stream-basedanalysis of elongation along a principal axis is the basisfor how the dorsal visual object processing stream canaffect categorization of manipulable objectsrsquorsquo (p 319)Importantly however none of the experiments in-cluded the presentation of nonelongated tool targetsthus allowing for the possibility that participants basedtheir responses exclusively on stimulus elongation

In our first experiment we therefore aimed to rule outthis possibility and verify whether stimulus elongationcan indeed be extracted under CFS and thereby facilitate

the categorization of manipulable objects (Almeida etal 2014) We presented both elongated and non-elongated tools and animals and tested the predictionthat elongated prime stimuli irrespective of primecategory affect the speeded categorization of tooltargets irrespective of target elongation

Experiment 1

Methods and materials

Participants

We determined sample size based on a recent studyby Almeida et al (2014) In that study the effect ofinterest was the role of elongation in visual recognitionof manipulable objects Elongated tool primes (eghammer) as well as elongated animal primes (eg fish)elicited faster categorization responses for elongatedtool targets when compared with nonelongated(lsquolsquostubbyrsquorsquo) animal primes (eg cow) The mean effectswere 12 6 4 and 8 6 3 ms respectively and both weresignificant when tested against zero t(26) frac14 308 pfrac140005 and t(26)frac14 243 pfrac14 0022 (Almeida et al 2014experiment 1a) By contrast nonelongated tool primesdid not result in significantly faster categorizationresponses for elongated tool targets when comparedwith nonelongated animal primes (4 ms) t(26) 1 Inprevious studies these three response-time (RT)differences defined the pattern of the lsquolsquocategory-by-shapersquorsquo priming account

Using GPower 319 (Faul Erdfelder Lang ampBuchner 2007) we first calculated the associated effectsizes based on the mean and standard deviation ofdifference (dz frac14 058 and 051 respectively) We thendetermined that for the average effect size of dz frac14 055and type I error probability of afrac14005 a sample size ofN frac14 22 was required to achieve a power of 080 (t testfor matched pairs one-sided) Twenty-nine observersparticipated in Experiment 1 which was conducted atthe Department of Psychiatry and PsychotherapyCharitendashUniversitatsmedizin Berlin Germany withethics approval from the German Association ofPsychology (Deutsche Gesellschaft fur Psychologie)Participants were recruited from a student pool viaemail and paid 8 euroshr for their participation Fourparticipants were excluded from further analysesbecause they showed significant above-chance forced-choice discrimination performance for invisible stimuliin the first control experiment Two additional partic-ipants were excluded because they reported too fewprime stimuli as invisible in the main experiment (fordetails on the exclusion criteria see Exclusion ofparticipants later) All 23 remaining participants (14women nine men mean age 23 years range 18ndash34)

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 2

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

had normal or corrected-to-normal vision were un-aware of the purpose of the study and providedinformed written consent

Apparatus and setup

Participants were seated in a dark environment theonly light coming from the experimental monitor and asecond monitor and viewed the dichoptic images on a17-in CRT monitor (SAMTRON 98PDF effectivescreen diagonal 436 cm refresh rate 60 Hz) via amirror stereoscope To stabilize head position theparticipants placed their heads on a chin rest Theviewing distance from the eyes to the screen (includingdistances within the mirror system) was 66 cm Allstimuli were generated using an IBM-compatible PCwith an ATI FireGL V7100 graphics card andPsychToolbox 3 (Brainard 1997 Pelli 1997) runningon MATLAB R2007b (MathWorks Inc Natick MA)

Stimuli

Four categories of visual stimuli were used in thisexperiment images of manufactured manipulable ob-jects (tools) and images of animals either elongated ornonelongated Images from the tool category were thesame as in a previous study from our group (LudwigKathmann Sterzer amp Hesselmann 2015) There were10 exemplars in each category (see SupplementaryFigure S1 for all exemplars) All 40 grayscale images(400 3 400 pixels) were slightly smoothed (low-passfiltered) using a circular Gaussian filter with a standarddeviation of 30 cyclesimage in the frequency domainThe low-level image properties (luminance histogramsand rotational average of the Fourier spectra) werematched across all exemplars (of all categories) with theSHINE toolbox (Willenbockel et al 2010) In contrastto previous studies (Almeida et al 2008 Almeida et al2010 Almeida et al 2014 Sakuraba et al 2012) wedid not add random noise to the preprocessed imagesAs shown in Supplementary Figure S2A all elongatedstimuli (tools and animals) had a widthlength ratio04 while all nonelongated stimuli had a widthlengthratio 04 (width orthogonal to the longest extension)Note that similar ratio distributions can be found forthe tool stimuli used in the studies by Sakuraba et al(2012) and Almeida et al (2014) With respect tostimulus orientation we aimed for an approximatematching of orientations between the two elongated-stimulus categories Supplementary Figure S2B showsthe orientations of all stimuli used in Experiment 1

Interocular masking

We used CFS to render the prime stimuli invisible(Tsuchiya et al 2006 Tsuchiya amp Koch 2005) CFS

uses high-contrast dynamic images (masks) flashed toone eye to suppress images presented to the other eyefrom awareness The mask images consisted of coloredrectangles and circles whose sizes ranged from 4 to18 of the size of the CFS area which measured 81983 8198 The rectangles and circles were positioned atrandom locations on the mask image Twenty-five ofthese images were created and flashed in random orderat 10 Hz to the dominant eye (Figure 1A) A whitesquare frame was presented around the stimuli topromote stable binocular fusion during dichopticpresentation Throughout each trial a central redfixation cross was presented (0268) during thepresentation of the prime stimulus the fixation crosswas shown only to the dominant eye The dominant eyewas determined using a hole-in-card test (Miles 1930)

Procedure

In total the experiment lasted approximately 90 minWritten instructions familiarized participants with allstimuli used in the experiment Prior to the mainexperiment a training session acquainted participantswith the procedure and task In the main experimentvisible targets were preceded by either a suppressedprime stimulus or a blank On each prime-present trialthe prime stimulus was presented together with theCFS masks for 200 ms (Figure 1A) In prime-absenttrials a blank stimulus and the CFS masks werepresented These trials allowed us to compare visibilityratings in prime-present and prime-absent trials underfull suppression RTs in prime-absent trials were notanalyzed After the prime (or blank) stimulus thevisible target stimulus was presented to both eyes for2000 ms

All combinations of prime and target category(tools elongated and nonelongated animals elongatedand nonelongated) were presented yielding 16 prime-present and four prime-absent conditions Thirty-twotrials per condition were presented resulting in a totalof 640 trials Trial order was randomized Prime andtarget exemplars were chosen pseudorandomly on eachtrial so that prime and target exemplars were neveridentical There were self-paced breaks after blocks of64 trials Each block started with an instruction screenillustrating the assignments of the response keys whichwere counterbalanced across participants Participantsused the left and right arrow keys to indicate the targetcategory (2AFC tool versus animal irrespective ofshape) and they were instructed to respond as quicklyand as accurately as possible No feedback wasprovided After the speeded response participants wereasked to provide their subjective visibility rating for theprime stimulus according to the perceptual awarenessscale (PAS Ramsoy amp Overgaard 2004) Participantswere made familiar with the levels of the PAS during

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 3

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

had normal or corrected-to-normal vision were un-aware of the purpose of the study and providedinformed written consent

Apparatus and setup

Participants were seated in a dark environment theonly light coming from the experimental monitor and asecond monitor and viewed the dichoptic images on a17-in CRT monitor (SAMTRON 98PDF effectivescreen diagonal 436 cm refresh rate 60 Hz) via amirror stereoscope To stabilize head position theparticipants placed their heads on a chin rest Theviewing distance from the eyes to the screen (includingdistances within the mirror system) was 66 cm Allstimuli were generated using an IBM-compatible PCwith an ATI FireGL V7100 graphics card andPsychToolbox 3 (Brainard 1997 Pelli 1997) runningon MATLAB R2007b (MathWorks Inc Natick MA)

Stimuli

Four categories of visual stimuli were used in thisexperiment images of manufactured manipulable ob-jects (tools) and images of animals either elongated ornonelongated Images from the tool category were thesame as in a previous study from our group (LudwigKathmann Sterzer amp Hesselmann 2015) There were10 exemplars in each category (see SupplementaryFigure S1 for all exemplars) All 40 grayscale images(400 3 400 pixels) were slightly smoothed (low-passfiltered) using a circular Gaussian filter with a standarddeviation of 30 cyclesimage in the frequency domainThe low-level image properties (luminance histogramsand rotational average of the Fourier spectra) werematched across all exemplars (of all categories) with theSHINE toolbox (Willenbockel et al 2010) In contrastto previous studies (Almeida et al 2008 Almeida et al2010 Almeida et al 2014 Sakuraba et al 2012) wedid not add random noise to the preprocessed imagesAs shown in Supplementary Figure S2A all elongatedstimuli (tools and animals) had a widthlength ratio04 while all nonelongated stimuli had a widthlengthratio 04 (width orthogonal to the longest extension)Note that similar ratio distributions can be found forthe tool stimuli used in the studies by Sakuraba et al(2012) and Almeida et al (2014) With respect tostimulus orientation we aimed for an approximatematching of orientations between the two elongated-stimulus categories Supplementary Figure S2B showsthe orientations of all stimuli used in Experiment 1

Interocular masking

We used CFS to render the prime stimuli invisible(Tsuchiya et al 2006 Tsuchiya amp Koch 2005) CFS

uses high-contrast dynamic images (masks) flashed toone eye to suppress images presented to the other eyefrom awareness The mask images consisted of coloredrectangles and circles whose sizes ranged from 4 to18 of the size of the CFS area which measured 81983 8198 The rectangles and circles were positioned atrandom locations on the mask image Twenty-five ofthese images were created and flashed in random orderat 10 Hz to the dominant eye (Figure 1A) A whitesquare frame was presented around the stimuli topromote stable binocular fusion during dichopticpresentation Throughout each trial a central redfixation cross was presented (0268) during thepresentation of the prime stimulus the fixation crosswas shown only to the dominant eye The dominant eyewas determined using a hole-in-card test (Miles 1930)

Procedure

In total the experiment lasted approximately 90 minWritten instructions familiarized participants with allstimuli used in the experiment Prior to the mainexperiment a training session acquainted participantswith the procedure and task In the main experimentvisible targets were preceded by either a suppressedprime stimulus or a blank On each prime-present trialthe prime stimulus was presented together with theCFS masks for 200 ms (Figure 1A) In prime-absenttrials a blank stimulus and the CFS masks werepresented These trials allowed us to compare visibilityratings in prime-present and prime-absent trials underfull suppression RTs in prime-absent trials were notanalyzed After the prime (or blank) stimulus thevisible target stimulus was presented to both eyes for2000 ms

All combinations of prime and target category(tools elongated and nonelongated animals elongatedand nonelongated) were presented yielding 16 prime-present and four prime-absent conditions Thirty-twotrials per condition were presented resulting in a totalof 640 trials Trial order was randomized Prime andtarget exemplars were chosen pseudorandomly on eachtrial so that prime and target exemplars were neveridentical There were self-paced breaks after blocks of64 trials Each block started with an instruction screenillustrating the assignments of the response keys whichwere counterbalanced across participants Participantsused the left and right arrow keys to indicate the targetcategory (2AFC tool versus animal irrespective ofshape) and they were instructed to respond as quicklyand as accurately as possible No feedback wasprovided After the speeded response participants wereasked to provide their subjective visibility rating for theprime stimulus according to the perceptual awarenessscale (PAS Ramsoy amp Overgaard 2004) Participantswere made familiar with the levels of the PAS during

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 3

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

training On each trial the four levels of the PAS werevertically presented on the screen (lsquolsquono impressionrsquorsquolsquolsquoweak glimpsersquorsquo lsquolsquoalmost clearrsquorsquo lsquolsquofully visiblersquorsquo) andparticipants used the up and down arrow keys to movea cursor and select their rating The cursor randomlypreselected one of the PAS ratings on each trial Bypressing the space bar participants confirmed theirrating and initiated the next trial which started after adelay of 500ndash750 ms

Prime contrast

One of the features of CFS is its large interindividualvariability in terms of suppression strength andcontrast dependency In a recent study using a 370-slong CFS display dominance durations varied 130-foldbetween participants (Yamashiro et al 2014) CFS

suppression levels have also been shown to changeacross trials so that contrasts need to be adjusted inorder to maintain stimulus invisibility (Ludwig SterzerKathmann Franz amp Hesselmann 2013) To controlfor this variability our approach was to individuallyadjust contrast levels prior to the main experiment andrun objective awareness checks afterwards We adjust-ed prime contrasts following a staircase procedure asfollows After a stimulus presentation that conformedto that of the main experiment participants had topress a key according to whether the stimulus had beenvisible or not Based on this response the stimuluscontrast was decreased or increased following alogarithmic scale in the next trial (one-up-one-downstaircase) Each participant completed two staircases of20 trials The stimulus contrast in the main experimentwas set to the highest stimulus contrast that the

Figure 1 Stimulation and task (A) The CFS priming paradigm Using a mirror stereoscope prime stimuli were presented for 200 ms to

participantsrsquo nondominant eye (ND) while colored Mondrian masks were flashed at 10 Hz to the dominant eye (D) Targets were

presented for 2000 ms to both eyes In each trial of the main experiment participants had to report whether the visible target was a

manipulable object (tool) or an animal (category task) or whether it was elongated or nonelongated (shape task) and then rate the

visibility of the suppressed prime using the PAS Shown is a trial with an elongated tool (as prime) and an elongated animal (as

target) (B) Category task Illustrated is the prediction that prime elongation (irrespective of category) affects the recognition of tools

(irrespective of elongation) (C) Shape task Illustrated is the prediction that prime shape affects shape discrimination of target stimuli

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 4

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

participant always judged as invisible in the pretestThis adjustment of the prime contrast was performedto ensure maximal stimulus strength even under fullsuppression The resulting Michelson contrast was 0276 003 (mean 6 SEM)

It is important to note that Almeida et al (2014)used the same prime contrast for all participants andexcluded participants based on awareness checks Intheir experiment about half of the participants(experiment 1a 48) were excluded because they werenot classified as objectively unaware of or wereclassified as subjectively aware of the primes One canargue however that the need to exclude 48 ofparticipants indicates a high level of stimulus visibilityat the population level prior to exclusion In statisticalterms restricting the analysis to a small subset ofparticipants may violate the principle of randomsampling and capitalize on chance differences betweenparticipants (Schmidt 2015) although it is not clearwhat percentage defines a problematically small subset(Note that we excluded 15 of participants inExperiment 1 and 7 in Experiment 2 based onawareness checks)

First control experiment

The first control experiment served to estimate primeawareness and followed immediately after the mainexperiment In this experiment no targets werepresented after the presentation of the suppressedprimes There were no prime-absent trials Participantswere instructed to indicate the category of the primes(tool vs animal 2AFC) and provide prime visibilityratings Participants were instructed to guess the primecategory if the primes were invisible A total of 100trials were presented Importantly the order ofexperiments ensured that potential training effects onprime processing during the course of the mainexperiment could be detected in the first controlexperiment (Ludwig et al 2013) Furthermore thesame level of suppression (ie all primes fullysuppressed) was used in both experiments to avoid anunderestimation of prime awareness based on the firstcontrol experiment and thus an overestimation ofunconscious prime processing during the main exper-iment (Lin amp Murray 2014)

Second control experiment

In the second control experiment we sought to testto what degree participants were able to discriminatethe primes when they were presented as in the main andfirst control experiment (ie briefly and with lowcontrast) but without the CFS masks No targets werepresented after the presentation of the primes and asbefore participants were instructed to indicate the

category of the primes and provide prime visibilityratings A total of 80 trials were presented

Familiarity and manipulability ratings

After the three experimental sessions participantsrated all 40 stimuli on two dimensions FollowingSalmon McMullen and Filliter (2010) we used a 5-point Likert scale for familiarity (lsquolsquorate your familiaritywith the object or the degree to which you come intocontact with or think about the concept on a day-to-day basisrsquorsquo) and manipulability (lsquolsquorate the manipulabil-ity of the object according to how easy it is to grasp anduse the object with one handrsquorsquo)

Exclusion of trials

In the main experiment only data from trials withcorrect responses were analyzed We discarded RTsthat were 15 interquartile ranges above the thirdquartile or below the first quartile with respect to theindividual distribution of all correct RTs (Tukey 1977)To further minimize the effect of outliers trials withanticipatory responses (RT 100 ms) were excluded aswell (Whelan 2008) Following this definition 758 ofcorrect trials were RT outliers and were removed fromfurther analysis As an alternative approach we firstexcluded trials with RTs 150 ms and 3000 ms in anext step trials with RTs more than three standarddeviations above or below the mean were excludedbased on the individual RT distributions Followingthis lsquolsquomean 63 SDrsquorsquo definition 213 of correct trialswere RT outliers We reasoned that strong primingeffects should be robust to using either of these outlierdefinitions

Exclusion of participants

In the first control experiment 2AFC discriminationperformance for stimuli rated as invisible (PASfrac141) wasdetermined for each participant and submitted tobinomial tests (chance level 50 a 005) Fourparticipants were excluded because they showedsignificant above-chance performance (65 58 80and 63 all ps 005) Next we sought to excludeparticipants with too few prime stimuli rated assubjectively invisible in the main experiment but at thesame time we aimed to keep the number of participantsas large as possible and hence statistical power as highas possible To this end we first selected participantswho gave more lsquolsquoweak glimpsersquorsquo ratings (PASfrac14 2) thanlsquolsquono impressionrsquorsquo ratings (PAS frac14 1) in the mainexperiment This was the case for six participants Nextwe tested for each of these participants whetherincluding the lsquolsquoweak glimpsersquorsquo trials in the analysis ofthe first control experiment would yield significant

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 5

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

above-chance performance if this was not the case weincluded the participant in the main analysis butdefined prime invisibility for the participant based onlsquolsquono impressionrsquorsquo and lsquolsquoweak glimpsersquorsquo PAS ratings (ieratings 1 and 2) as well as chance-level performance inthe discrimination task (for a similar approach seeLudwig et al 2015) Following this procedure two ofthe six selected participants were excluded (59 and62 all ps 005) For all remaining 23 participantsSupplementary Figure S3A plots forced-choice dis-crimination performance in the first control experi-ment

Data analysis

To check for the predicted pattern of results alreadydescribed (also see Figure 1B) RT data were firstanalyzed using paired t tests following Almeida et al

(2014) Next we submitted the data from prime-presenttrials to a repeated-measures ANOVA with thefollowing factors Target category (two levels animalsvs tools) Target elongation (two levels elongated vsnonelongated) Shape congruency (two levels congru-ent vs incongruent) and Category congruency (twolevels congruent vs incongruent) ANOVAs werecalculated using IBM SPSS (Version 220)3

Results

Prime visibility and prime discrimination

Figure 2A summarizes subjective prime visibilityratings in prime-present and prime-absent trials of themain experiment as well as in the first and secondcontrol experiments In the main experiment lsquolsquoinvisi-blersquorsquo ratings (PASfrac14 1) were most frequent both in

Figure 2 Prime visibility ratings in Experiment 1 (A) and Experiment 2 (B) Plotted are PAS ratings in prime-present (red bars) and

prime-absent (gray bars) trials of the main experiment and in the first (green bars) and second (blue bars) control experiment There

were no prime-absent trials in Experiment 2 Plotted is the mean 6 standard error of the mean

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 6

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

participant always judged as invisible in the pretestThis adjustment of the prime contrast was performedto ensure maximal stimulus strength even under fullsuppression The resulting Michelson contrast was 0276 003 (mean 6 SEM)

It is important to note that Almeida et al (2014)used the same prime contrast for all participants andexcluded participants based on awareness checks Intheir experiment about half of the participants(experiment 1a 48) were excluded because they werenot classified as objectively unaware of or wereclassified as subjectively aware of the primes One canargue however that the need to exclude 48 ofparticipants indicates a high level of stimulus visibilityat the population level prior to exclusion In statisticalterms restricting the analysis to a small subset ofparticipants may violate the principle of randomsampling and capitalize on chance differences betweenparticipants (Schmidt 2015) although it is not clearwhat percentage defines a problematically small subset(Note that we excluded 15 of participants inExperiment 1 and 7 in Experiment 2 based onawareness checks)

First control experiment

The first control experiment served to estimate primeawareness and followed immediately after the mainexperiment In this experiment no targets werepresented after the presentation of the suppressedprimes There were no prime-absent trials Participantswere instructed to indicate the category of the primes(tool vs animal 2AFC) and provide prime visibilityratings Participants were instructed to guess the primecategory if the primes were invisible A total of 100trials were presented Importantly the order ofexperiments ensured that potential training effects onprime processing during the course of the mainexperiment could be detected in the first controlexperiment (Ludwig et al 2013) Furthermore thesame level of suppression (ie all primes fullysuppressed) was used in both experiments to avoid anunderestimation of prime awareness based on the firstcontrol experiment and thus an overestimation ofunconscious prime processing during the main exper-iment (Lin amp Murray 2014)

Second control experiment

In the second control experiment we sought to testto what degree participants were able to discriminatethe primes when they were presented as in the main andfirst control experiment (ie briefly and with lowcontrast) but without the CFS masks No targets werepresented after the presentation of the primes and asbefore participants were instructed to indicate the

category of the primes and provide prime visibilityratings A total of 80 trials were presented

Familiarity and manipulability ratings

After the three experimental sessions participantsrated all 40 stimuli on two dimensions FollowingSalmon McMullen and Filliter (2010) we used a 5-point Likert scale for familiarity (lsquolsquorate your familiaritywith the object or the degree to which you come intocontact with or think about the concept on a day-to-day basisrsquorsquo) and manipulability (lsquolsquorate the manipulabil-ity of the object according to how easy it is to grasp anduse the object with one handrsquorsquo)

Exclusion of trials

In the main experiment only data from trials withcorrect responses were analyzed We discarded RTsthat were 15 interquartile ranges above the thirdquartile or below the first quartile with respect to theindividual distribution of all correct RTs (Tukey 1977)To further minimize the effect of outliers trials withanticipatory responses (RT 100 ms) were excluded aswell (Whelan 2008) Following this definition 758 ofcorrect trials were RT outliers and were removed fromfurther analysis As an alternative approach we firstexcluded trials with RTs 150 ms and 3000 ms in anext step trials with RTs more than three standarddeviations above or below the mean were excludedbased on the individual RT distributions Followingthis lsquolsquomean 63 SDrsquorsquo definition 213 of correct trialswere RT outliers We reasoned that strong primingeffects should be robust to using either of these outlierdefinitions

Exclusion of participants

In the first control experiment 2AFC discriminationperformance for stimuli rated as invisible (PASfrac141) wasdetermined for each participant and submitted tobinomial tests (chance level 50 a 005) Fourparticipants were excluded because they showedsignificant above-chance performance (65 58 80and 63 all ps 005) Next we sought to excludeparticipants with too few prime stimuli rated assubjectively invisible in the main experiment but at thesame time we aimed to keep the number of participantsas large as possible and hence statistical power as highas possible To this end we first selected participantswho gave more lsquolsquoweak glimpsersquorsquo ratings (PASfrac14 2) thanlsquolsquono impressionrsquorsquo ratings (PAS frac14 1) in the mainexperiment This was the case for six participants Nextwe tested for each of these participants whetherincluding the lsquolsquoweak glimpsersquorsquo trials in the analysis ofthe first control experiment would yield significant

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 5

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

above-chance performance if this was not the case weincluded the participant in the main analysis butdefined prime invisibility for the participant based onlsquolsquono impressionrsquorsquo and lsquolsquoweak glimpsersquorsquo PAS ratings (ieratings 1 and 2) as well as chance-level performance inthe discrimination task (for a similar approach seeLudwig et al 2015) Following this procedure two ofthe six selected participants were excluded (59 and62 all ps 005) For all remaining 23 participantsSupplementary Figure S3A plots forced-choice dis-crimination performance in the first control experi-ment

Data analysis

To check for the predicted pattern of results alreadydescribed (also see Figure 1B) RT data were firstanalyzed using paired t tests following Almeida et al

(2014) Next we submitted the data from prime-presenttrials to a repeated-measures ANOVA with thefollowing factors Target category (two levels animalsvs tools) Target elongation (two levels elongated vsnonelongated) Shape congruency (two levels congru-ent vs incongruent) and Category congruency (twolevels congruent vs incongruent) ANOVAs werecalculated using IBM SPSS (Version 220)3

Results

Prime visibility and prime discrimination

Figure 2A summarizes subjective prime visibilityratings in prime-present and prime-absent trials of themain experiment as well as in the first and secondcontrol experiments In the main experiment lsquolsquoinvisi-blersquorsquo ratings (PASfrac14 1) were most frequent both in

Figure 2 Prime visibility ratings in Experiment 1 (A) and Experiment 2 (B) Plotted are PAS ratings in prime-present (red bars) and

prime-absent (gray bars) trials of the main experiment and in the first (green bars) and second (blue bars) control experiment There

were no prime-absent trials in Experiment 2 Plotted is the mean 6 standard error of the mean

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 6

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

above-chance performance if this was not the case weincluded the participant in the main analysis butdefined prime invisibility for the participant based onlsquolsquono impressionrsquorsquo and lsquolsquoweak glimpsersquorsquo PAS ratings (ieratings 1 and 2) as well as chance-level performance inthe discrimination task (for a similar approach seeLudwig et al 2015) Following this procedure two ofthe six selected participants were excluded (59 and62 all ps 005) For all remaining 23 participantsSupplementary Figure S3A plots forced-choice dis-crimination performance in the first control experi-ment

Data analysis

To check for the predicted pattern of results alreadydescribed (also see Figure 1B) RT data were firstanalyzed using paired t tests following Almeida et al

(2014) Next we submitted the data from prime-presenttrials to a repeated-measures ANOVA with thefollowing factors Target category (two levels animalsvs tools) Target elongation (two levels elongated vsnonelongated) Shape congruency (two levels congru-ent vs incongruent) and Category congruency (twolevels congruent vs incongruent) ANOVAs werecalculated using IBM SPSS (Version 220)3

Results

Prime visibility and prime discrimination

Figure 2A summarizes subjective prime visibilityratings in prime-present and prime-absent trials of themain experiment as well as in the first and secondcontrol experiments In the main experiment lsquolsquoinvisi-blersquorsquo ratings (PASfrac14 1) were most frequent both in

Figure 2 Prime visibility ratings in Experiment 1 (A) and Experiment 2 (B) Plotted are PAS ratings in prime-present (red bars) and

prime-absent (gray bars) trials of the main experiment and in the first (green bars) and second (blue bars) control experiment There

were no prime-absent trials in Experiment 2 Plotted is the mean 6 standard error of the mean

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 6

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

prime-present (7700) and in prime-absent trials(7874) intermediate (PASfrac14 2 or 3) and full visibilityratings (PAS frac14 4) were less frequent Median PASratings were highly similar across prime categories(elongated tools 128 nonelongated tools 124elongated animals 126 nonelongated animals 126)In the first control experiment which we conducted toestimate objective prime unawareness primes werepredominantly rated as invisible (8809) Forced-choice discrimination performance for invisible primeswas at chance level (4887) t(22) frac14141 p frac14 0172In the second control experiment we sought to test towhat degree participants were able to discriminate theprimes when they were presented at the individuallyadjusted contrast levels but without the CFS masksPrimes were predominantly rated as clearly visible(PASfrac14 4 8565) and 2AFC discrimination of visibleprimes was near ceiling level (9727)

RTs in trials with invisible primes

2AFC discrimination of target category (tool vsanimal) was at 9757 Table 1 summarizes meancorrect RTs in Experiment 1 Since this was the mainfinding in previous work (Almeida et al 2014) we firsttested whether elongated tool primes as well aselongated animal primes elicited faster categorizationresponses for elongated tool targets (657 and 658 msrespectively) when compared with nonelongated animalprimes (662 ms) While the observed RT differenceswere numerically in the predicted direction (5 and 4 msie facilitatory) they were not significant (one-sidedpaired t tests) t(22)frac14 056 pfrac14 0290 and t(22)frac14 052pfrac14 0305

If there were a facilitatory effect of prime-stimuluselongation (irrespective of category) on the categori-zation of tools then we should expect a significantTarget category 3 Target elongation 3 Shape congru-ency interaction The ANOVA showed that thisinteraction was not significant F(1 22)frac14 067 pfrac140421 gp

2frac14 003 No other main effects or interactionswere significant (see Supplementary Table S1 for fullANOVA results) Overall this pattern of results wasrobust against using the lsquolsquomean 63 SDrsquorsquo outlierdefinition which included more trials with long RTs

Familiarity and manipulability ratings

Average familiarity ratings for the different sets ofstimuli (Supplementary Figure S1 one out of 29participants did not fill out the rating questionnaire)were as follows elongated tools 381 nonelongatedtools 441 elongated animals 290 nonelongatedanimals 311 An exploratory analysis of manipula-bility ratings (elongated tools 464 nonelongated tools468 elongated animals 196 nonelongated animals206) with factors Category and Shape revealed thattools were rated as significantly more manipulable thananimals F(1 27) frac14 30954 p 0001 gp

2frac14 092 theeffect of shape was marginally significant F(1 27)frac14421 pfrac14 0050 gp

2frac14 014 while the Category 3 Shapeinteraction was not F(1 27) 1 Shape did notinfluence manipulability ratings of tools t(27)frac14100pfrac14 0322

Discussion

The results from Experiment 1 provide no evidencein support of the hypothesis that elongated primestimuli that are rendered invisible by CFS facilitate thecategorization of manipulable objects due to a lsquolsquodorsalCFS biasrsquorsquo (Almeida et al 2014) In contrast toprevious CFS studies (Almeida et al 2008 Almeida etal 2010 Sakuraba et al 2012) we observed nopriming effects in the tool-versus-animal categorizationtask How can this finding be explained First thisfailure to reproduce previous results may be due torandom sampling (Francis 2013)mdashie even truephenomena sometimes do not yield significant resultsmdashand low statistical power in our experiment (Button etal 2013) Beyond effect size and significance howeverthe descriptive analysis of our RT data does not matchwell with the predicted pattern of RT differences

Second there are a number of methodologicaldifferences between our experiment and previousstudies Among them is the set of visual stimuli Onecould argue that the tool stimuli used in our experiment(Supplementary Figure S1) were less manipulable thanthe tool stimuli used in previous studies But partici-pants rated all our tool images including the non-elongated exemplars as highly manipulableFurthermore the results from our recent fMRI study

Target

Prime

Elongated tool Nonelongated tool Elongated animal Nonelongated animal

Elongated tool 657 6 20 652 6 21 658 6 19 662 6 22

Nonelongated tool 665 6 22 655 6 19 658 6 19 659 6 21

Elongated animal 666 6 23 654 6 23 660 6 22 646 6 21

Nonelongated animal 656 6 27 655 6 23 662 6 25 658 6 23

Table 1 RTs (mean 6 SEM ms) in Experiment 1 across participants (N frac14 23)

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 7

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

(Ludwig et al 2015) which used the same set of toolstimuli showed that tool-selective areas in parietalcortex were activated similarly by elongated andnonelongated tool exemplars We therefore rule outlow manipulability of our tool stimulus set as apotential confound An important basic visual featureof our stimuli is their elongatedness While the widthlength ratios of our stimuli form a bimodal distributionsimilar to the ratio distributions found in previouspriming studies (Supplementary Figure S2) one couldargue that the distributions should be not just similarbut equal in terms of mean and variance Thishowever was not even the case for two previous studiesthat both showed significant priming effects (Almeidaet al 2014 Sakuraba et al 2012) Of note wedeliberately decided to use more exemplars per stimuluscategory than in previous studies to minimize potentialconfounds of stimulus sampling In studies with fewstimulus exemplars it may well be that the observedeffects are driven by only a subset of stimuli this wouldlimit the generalization to future studies involvingdifferent samples of stimuli and would remain unde-tected unless statistical tests were calculated includingthe stimulus exemplar as a random factor (JuddWestfall amp Kenny 2012)

In terms of interocular masking the types of CFSmasks and methods for dichoptic stimulation weredifferent between studies We used a mirror stereoscopein combination with colored Mondrian masks similarto the masks in the original CFS study (Tsuchiya ampKoch 2005) the studies that reported significant toolpriming effects under CFS used anaglyph glasses incombination with monotone random-noise images asmasks (Almeida et al 2008 Almeida et al 2010Almeida et al 2014 Sakuraba et al 2012) The mirrorstereoscope gives no cross talk between the eyes butthe anaglyph glasses produce considerable stereoscopiccross talk (Baker Kaestner amp Gouws 2015) Whilethere has been no systematic investigation of thepotential influence of the various dichoptic stimulationmethods on CFS the influence of the spatiotemporalmask properties on the strength of CFS suppressionhas already been investigated Specifically Yang andBlake (2012) showed that Mondrian masks composedof only their high-spatial-frequency components yield-ed much weaker elevation of test-probe detectionthresholds (ie lower suppression strength) thanMondrian masks that were filtered to contain only theirlow-spatial-frequency components Using a similarprobe detection technique (Tsuchiya et al 2006)unpublished data from our lab show that random-noisemasks result in lower detection thresholds thanMondrian masks (Supplementary Figure S4) Onecould therefore argue that the suppression level in ourexperiment was deeper than in previous CFS primingexperiments which might explain the absence of

priming in our case By carefully adjusting prime-stimulus contrasts individually for each participanthowever we ensured maximal stimulus strength evenunder full CFS suppression On a final note behavioraland neural effects under CFS appear to be ratherrobust against smaller changes in the experimentalsetup for example significant cortical responses toinvisible stimuli have also been found for Mondrian-like masks (Jiang amp He 2006 Sterzer Haynes amp Rees2008) We conclude that the divergent results are notlikely to be due to methodological differences

Most critically the conceptual difference betweenour experiment and previous studies might be theperformed task As pointed out in the Introduction alltool targets in previous experiments were elongatedwhile all animal targets were nonelongated Thus it ispossible that participants performed a shape-discrimi-nation task (elongated vs nonelongated) rather thanthe instructed category-discrimination task (tool vsanimal) In our first experiment participants wereunable to use this shape strategy as targets could beeither elongated or nonelongated irrespective ofcategory (Supplementary Figure S1) In Experiment 2we sought to investigate this further by testing a newsample of participants with both shape- and category-discrimination tasks in a within-subject design

Experiment 2

Methods and materials

Methods including apparatus and setup stimuli andinterocular masking were the same as in Experiment 1except for the differences described in the followingThe main difference between the experiments was theinclusion of a shape-discrimination task in Experiment2 (see Procedure)

Participants

Thirty-one observers participated in Experiment 2All 27 participants who were included (16 women 11men mean age 23 years range 18ndash32 for details onthe exclusion criteria see later) had normal orcorrected-to-normal vision were unaware of thepurpose of the study and provided informed writtenconsent

Stimuli

Based on the debriefing of participants in Experi-ment 1 we removed a subset of nonelongated stimulithat were considered less lsquolsquoroundishrsquorsquo exemplars 13 1819 and 20 were removed from the nonelongated-tool

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 8

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

category and exemplars 39 and 40 were removed fromthe nonelongated-animal category We also removedexemplars 8 and 10 from the elongated-tool categoryfinally to remove the same number of exemplars fromeach category additional exemplars of similar stimulusorientation were excluded (marked in red inSupplementary Figure S1 Supplementary Figure S2Bshows the orientations of all stimuli used in Experiment2)

Procedure

In total the experiment lasted approximately 90 minThe blank prime condition was removed from thedesign yielding a total of 4 3 4frac14 16 conditions Thesewere presented in two blocks of 384 trials each theblocks differed only with respect to the task instructionwhich was either a shape-discrimination task (elongat-ed vs nonelongated) or a category-discrimination task(tool vs animal) The first 64 trials of each block werediscarded as training trials Each miniblock of 64 trialsstarted with an instruction screen illustrating theassignments of the response keys The order of blocksand the response-key assignments were counterbal-anced across participants

Prime contrast

Prime contrast was individually adjusted to ensuremaximal stimulus strength even under full suppressionThe resulting Michelson contrast was 065 6 005(mean 6 SEM)

First control experiment

In the first control experiment two blocks of 60trials each were presented to estimate prime awarenessbased on objective performance in both tasks (iecategory and shape discrimination)

Second control experiment

In the second control experiment in which primeswere shown without the CFS masks only a singlecategory-discrimination block of 60 trials was present-ed We reasoned that above-chance category discrim-ination of visible primes would imply above-chanceshape discrimination of the same primes

Exclusion of participants

One participant was excluded from further analysesbecause he pressed the same response button on eachlsquolsquoinvisible primersquorsquo trial in the first control experiment(hence we assume that he was not guessing asinstructed) furthermore his mean RTs (1 s) were

almost twice as long as the grouprsquos mean RTs in themain experiment Another participant was excludedbecause he gave very few ratings of PAS frac14 1 (34)and PAS frac14 2 (25) ratings in the main experimentwhile he rated the suppressed primes as fully invisible(PAS frac14 1 99) during the first control experiment(hence we assume that he misunderstood the in-struction and reported the visibility of the unsup-pressed target stimulus during the main experimentless plausible is the alternative scenario that he mighthave been lsquolsquoimmunersquorsquo to CFS suppression during themain experiment) In the first control experiment2AFC discrimination performance for stimuli rated asinvisible (PAS frac14 1) was determined for each partici-pant and submitted to binomial tests separately forthe shape- and category-discrimination tasks (chancelevel 50 a 0025) Two further participants wereexcluded because they showed significant above-chance performance one was above chance in theshape-discrimination task (63) and the other wasabove chance in the category-discrimination task(68)

As in Experiment 1 we then selected participantswho gave more lsquolsquoweak glimpsersquorsquo ratings (PASfrac14 2) thanlsquolsquono impressionrsquorsquo ratings (PAS frac14 1) in the mainexperiment (two participants) and tested them forabove-chance performance in the first control experi-ment for both visibility ratings Following this proce-dure neither was excluded their lsquolsquoweak glimpsersquorsquo trialsin the main experiment were included in the analysisFor the resulting 27 participants SupplementaryFigure S3B plots forced-choice discrimination perfor-mance in the first control experiment

Results

Prime visibility and prime discrimination

Figure 2B summarizes subjective prime visibilityratings in Experiment 2 In the main experimentlsquolsquoinvisiblersquorsquo ratings (PAS frac14 1) were most frequent(8318) intermediate (PASfrac14 2 or 3) and full visibility(PASfrac14 4) ratings were less frequent Median PASratings were highly similar across prime categories(elongated tools 107 nonelongated tools 111elongated animals 107 nonelongated animals 111)

In the first control experiment primes were pre-dominantly rated as invisible (PASfrac14 1 8827)Forced-choice discrimination performance for invisibleprimes was at chance level in the shape task (5096)t(26)frac14 073 pfrac14 0475 and marginally significant in thecategory task (5214) t(26) frac14 205 pfrac14 0050 In thesecond control experiment primes were predominantlyrated as clearly visible (PASfrac14 4 9383) and 2AFCcategory discrimination was near ceiling level for visibleprimes (9518)

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 9

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

RTs in trials with invisible primes

2AFC discrimination of targets was at 9183 in theshape task and 9600 in the category task The lowerdiscrimination performance in the shape task was dueto the fact that some of the participants falselycategorized some nonelongated tool exemplars aselongated As this false classification was nearlysystematic for these participants it effectively removedtrials with these stimuli as targets from their dataanalysis Of all correct trials 672 were RT outliersBased on the alternative lsquolsquomean 63 SDrsquorsquo definition219 were RT outliers

Table 2 summarizes mean correct RTs in Experiment2 separately for the category- and shape-discriminationtasks Overall RTs were similar for the two tasks (607vs 615 ms)

For the category task we first tested whetherelongated tool primes as well as elongated animalprimes elicited faster categorization responses forelongated tool targets (602 and 607 ms respectively)when compared with nonelongated animal primes (602ms) The observed RT differences were not in thepredicted direction and not significant (one-sidedpaired t tests) t(26) frac14001 p frac14 0498 and t(26) frac14082 p frac14 0209

If there were a facilitatory effect of prime-stimuluselongation on the categorization of manipulable objects(Figure 1B) then we should expect a significant Targetcategory 3 Target elongation 3 Shape congruencyinteraction in the category-discrimination task TheANOVA showed that this interaction was again notsignificant F(1 26) frac14 063 p frac14 0435 gp

2 frac14 002 Themain effect of Target category was significant (seeSupplementary Table S2 for full ANOVA results)because it took participants longer to categorize tooltargets as compared to animal targets (614 vs 600 ms)Overall this pattern of results was robust against thelsquolsquomean 63 SDrsquorsquo outlier definition which included moretrials with long RTs

If there were shape priming (ie congruent primeshape facilitates target shape discrimination Figure1C) then we should expect an influence of Shapecongruency in the shape-discrimination task as either amain or interaction effect While we found no maineffect (see Supplementary Table S2 for full ANOVAresults) interactions involving Shape congruency weresignificantmdashnamely Category congruency 3 Shapecongruency F(1 26)frac14 562 p frac14 0025 gp

2frac14 018 andTarget category 3 Target elongation 3 Shape congru-ency F(1 26)frac14 588 pfrac14 0023 gp

2frac14 018mdashindicatingthat prime shape indeed affected RTs in the shape-discrimination task The three-way interaction wasrobust to the inclusion of more trials with long RTswhen using the lsquolsquomean 63 SDrsquorsquo outlier definition F(126)frac14 626 pfrac14 0019 gp

2 frac14 019 and we furtherexplored it by testing the Target elongation 3 Shapecongruency interaction separately for tool and animaltargets (Figure 3) For tool targets the two-wayinteraction was significant F(1 26)frac14 430 pfrac14 0048gp

2frac14 014 because the Shape congruency effect waslarger for elongated targets (589 vs 600 ms) than fornonelongated targets where it was numerically re-versed (647 vs 635 ms) Post hoc tests revealed that thefacilitatory effect was significant t(26)frac14215 pfrac140041while the reversed effect was not t(26) frac14142 p frac140166 Using the lsquolsquomean 63 SDrsquorsquo outlier definition(Figure 3B) post hoc tests were significant for thefacilitatory effect (606 vs 623 ms) t(26) frac14 210 p frac140046 and the reversed effect (682 vs 661 ms) t(26) frac14214 p frac14 0042 For animal targets the two-wayinteraction was not significant F(1 26)frac14 009 pfrac140769 gp

2 001Finally we observed a significant effect of Target

elongation in the shape task because it took partici-pants longer to respond to nonelongated targets ascompared to elongated targets (626 vs 604 ms) andthis main effect was mainly driven by nonelongatedtools

Target

Prime

Elongated tool Nonelongated tool Elongated animal Nonelongated animal

Category task

Elongated tool 602 6 17 616 6 21 607 6 18 602 6 15

Nonelongated tool 616 6 19 626 6 18 618 6 20 624 6 20

Elongated animal 608 6 19 601 6 20 611 6 21 607 6 21

Nonelongated animal 587 6 19 591 6 19 590 6 18 607 6 20

Shape task

Elongated tool 585 6 15 601 6 15 593 6 15 600 6 16

Nonelongated tool 640 6 15 649 6 19 630 6 16 645 6 18

Elongated animal 617 6 13 609 6 14 607 6 12 617 6 13

Nonelongated animal 607 6 13 610 6 14 621 6 14 604 6 13

Table 2 RTs (mean 6 SEM ms) in Experiment 2 across participants (N frac14 27)

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 10

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

Stimulus orientation

When preparing our stimuli we aimed for anapproximate matching of stimulus orientations be-tween the two elongated categories Due to the fact thatthe matching was only approximate (SupplementaryFigure S2B) and since we randomly chose prime andtarget exemplars on each trial (with the exception thatexemplars could not be identical) there was somevariability of absolute differences between prime andtarget orientation across the experimental conditions(398ndash518 Table 3) Note that absolute orientationdifferences were defined to be within the range 08ndash898

We reasoned that the visual processing of stimulusorientation might precede any higher level shapeanalysis and that it thus could have influenced theresults in the shape task2 Specifically we focused ourexploratory control analysis on the conditions with anRT effect involving elongated targetsmdashie the effect ofshape congruency observed for tool stimuli but not foranimal stimuli (indicated by blue bars in Figure 3) Welimited the analysis to trials with elongated primes andcalculated Pearson correlation coefficients betweenabsolute differences between prime and target orienta-

tion on the one hand and RT on the other handseparately for each participant If stimulus orientationhad a modulatory influence on the RT effect then weshould expect a difference between the coefficients fortrials with tool targets and the coefficients for trialswith animal targets Coefficients turned out to beslightly larger for tool targets than for animal targets(tools rfrac14 0052 animals rfrac140042 two-sided paired ttest) t(26) frac14 213 p frac14 0043

Familiarity and manipulability ratings

Average familiarity ratings for the four sets ofstimuli (Supplementary Figure S1 subset) were asfollows elongated tools 363 nonelongated tools 388elongated animals 236 nonelongated animals 270An exploratory analysis of manipulability ratings(elongated tools 442 nonelongated tools 436elongated animals 213 nonelongated animals 249)revealed that tools were rated as significantly moremanipulable than animals F(1 30)frac14 5967 p 0001gp

2frac14 067 nonelongated stimuli were rated as slightlymore manipulable than elongated stimuli and the main

Figure 3 RTs in the shape task of Experiment 2 Plotted are RTs (in milliseconds) illustrating the Target category (TC) 3 Target

elongation (TE)3 Shape congruency (SC) interaction (A) Analysis based on the interquartile range outlier definition by Tukey (1977)

(B) Analysis based on the lsquolsquomean 63 SDrsquorsquo outlier definition which included more trials with long RTs Target stimuli were either tools

or animals Blue bars denote RT data from trials with elongated targets and gray bars denote RT data from trials with nonelongated

targets Shape congruency of targets and primes was either congruent (C) or incongruent (IC) irrespective of category Plotted are the

mean 6 CousineaundashMorey corrected error bars for within-subject designs (Cousineau 2005 Morey 2008)

Target

Prime

Elongated tool Nonelongated tool Elongated animal Nonelongated animal

Elongated tool 51 6 1 43 6 1 43 6 1 46 6 1

Nonelongated tool 44 6 2 46 6 1 44 6 2 39 6 1

Elongated animal 43 6 1 43 6 2 52 6 1 43 6 1

Nonelongated animal 47 6 1 42 6 1 44 6 1 43 6 1

Table 3 Absolute orientation differences between prime and target (mean 6 SEM 8) in Experiment 2 across participants (Nfrac14 27)

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 11

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

effect of Shape turned out to be significant F(1 30)frac14513 pfrac14 0031 gp

2frac14 015 as did the Category3 Shapeinteraction F(1 30)frac14 942 pfrac14 0005 gp

2 frac14 024because shape influenced tool manipulability ratingsless than manipulability ratings of animals

Discussion

As in Experiment 1 we found no evidence support-ing the notion of elongation-based unconscious prim-ing in a category task as proposed by Almeida et al(2014) However using the same stimuli and setup inthe same group of participants but with a shape-discrimination task (elongated vs nonelongated) in-stead of category discrimination the RT results fromExperiment 2 provided some evidence for priming

We found facilitatory priming by shape-congruentversus shape-incongruent primes for elongated tooltargets and a reversed (albeit not robust) RT differencefor nonelongated tool targets Or put differently wefound that elongated primes could elicit faster shapediscrimination of tool targets elongated and non-elongated alike Our RT data thus suggest that someshape information can still be extracted under CFS andaffect manual target responses in a shape-discrimina-tion task Numerically both effects were relativelysmall (11 and 12 ms) but well within the range ofsignificant RT differences (8ndash12 ms) reported byAlmeida et al (2014) Our RT differences are alsocomparable with recently reported numerical primingeffects under CFS (Hesselmann Darcy Sterzer ampKnops 2015 Hesselmann amp Knops 2014) Twofurther studies however found no evidence forunconscious priming and suggest an even lower boundfor priming effects under CFS (Izatt Dubois Faivre ampKoch 2014 Peremen amp Lamy 2014)

From a broader perspective our finding that primeshape can escape CFS suppression and affect subse-quent manual responses to target stimuli appears tosupport the notion that interocularly suppressed stimuliare processed at the level of features and coarse featureconfigurations (Gayet et al 2014) CFS is a method ofinterocular suppression and thus closely related tobinocular rivalry the mechanisms of which have beenextensively investigated in the last decades (Blake ampLogothetis 2002 Sterzer 2013) Although it is not yetclear whether CFS should be regarded as a variant ofbinocular rivalry that induces particularly strongsuppression (Shimaoka amp Kaneko 2011) or whetherCFS is supported by mechanisms distinct frombinocular rivalry (Tsuchiya et al 2006) a reasonabledefault view would be to expect only little high-levelprocessing during CFS (Breitmeyer 2014b) given thelimited stimulus processing during binocular suppres-sion (Blake 1988 Cave Blake amp McNamara 1998

Kang Blake amp Woodman 2011 Zimba amp Blake1983) However the current CFS literature sketches anincreasingly fuzzy picture of the extent to which high-level unconscious visual processing is possible underCFS (Heyman amp Moors 2014 Sterzer et al 2014Yang et al 2014) Future studies investigating low-level (eg shape priming) and high-level (eg categorypriming) effects under CFS should therefore aim toidentify the relevant methodological differences be-tween previous studies that might explain the hetero-geneous results

General discussion

In two experiments we found no evidence in supportof the hypothesis that elongated prime stimuli uncon-sciously presented under CFS facilitate the task ofcategorizing manipulable objects (ie category-by-shape priming) We did however observe that primeshape affects the shape discrimination of subsequentlypresented tool stimuli In the following we will discussour results in the shape task address the notion of adorsal-stream processing bias under CFS and outlinethe limitations of the current study

Priming in the shape task

We found that shape-congruent primes differentiallyaffected the shape-discrimination responses for elon-gated and nonelongated tool targets This pattern ofresults is difficult to reconcile with a simple primingmodel based on shape congruency alone and it meritsfurther exploration

From these data one may tentatively conclude thatelongated stimuli facilitated the shape task for tooltargets elongated and nonelongated alike While this isa pattern rather to be expected in the category task (inline with Almeidarsquos category-by-shape priming hy-pothesis) it could potentially be generated in the shapetask as well when assuming a fast and automaticcategory-recognition process that influences targetprocessing in the shape task (Fabre-Thorpe 2011Mouchetant-Rostaing Giard Delpuech Echallier ampPernier 2000) Obviously the question then remainswhy we did not observe exactly this pattern in thecategory task Speculatively the category task mighthave desensitized perceptual pathways related to theprocessing of prime elongation and thus diminishedcategory-by-shape priming by contrast the shapetaskmdashrequiring only attention to basic visual fea-turesmdashsensitized perceptual pathways and boostedcategory-by-shape priming Attentional sensitizationhas previously been proposed as a mechanism to

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 12

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

explain enhanced and decreased priming effects ob-served under different task sets (Kiefer 2012 KieferAdams amp Zovko 2012) According to this modelattentional sensitization of unconscious processing bytask sets is achieved by enhancing the sensitivity oftask-relevant pathways and reducing the sensitivity oftask-irrelevant pathways (Kiefer amp Martens 2010)

Alternatively our results in the shape task might bedue to differences in stimulus orientation To the bestof our knowledge only one previous CFS primingstudy directly addressed this issue (Sakuraba et al2012 experiment 5) They used differently oriented(08 458 908 1358) prime stimuli and found significantpriming effects for elongated tool targets only whengeometric lines were used as primes not when otheroriented stimuli (less elongated diamond shapes andnonelongated lsquolsquoPacmanrsquorsquo stimuli) were used Theauthors concluded lsquolsquoIt is unlikely that the orientationsof elongated shapes have a significant impact on thepriming effects found in previous experimentsrsquorsquo (p3952) Based on this conclusion we did not expectsignificant effects of stimulus orientation in ourexperiments but we nevertheless aimed for anapproximate matching of stimulus orientation be-tween our elongated stimulus categories (ie toolsand animals) However Sakuraba et al (2012) did notreport whether their shape-priming effect was furthermodulated by the orientation differences betweenprimes and targets We addressed this question in acontrol analysis that was limited to conditions withelongated stimuli It revealed that the correlationbetween orientation differences and RT was larger fortrials with tool targets than for trials with animaltargets We conclude that stimulus orientation mighthave modulated the results in the shape task althoughmore experiments (eg with a small set of discretestimulus orientations) will be needed to betterunderstand the magnitude and direction of thismodulation

Are priming effects observed under CFSmediated by the dorsal visual pathway

In an influential fMRI study Fang and He (2005)reported largely preserved neural activity in the dorsalvisual pathway in response to CFS-suppressed imagesof tools while activity in the ventral stream was fullysuppressed In the context of the duplex vision model(Milner amp Goodale 1995 2006) the results of Fangand He were interpreted as evidence that CFS could beused specifically to lsquolsquoisolatersquorsquo or lsquolsquobiasrsquorsquo dorsal visualprocessingmdashie selectively disrupt conscious objectidentification mediated by the ventral stream whileleaving nonconscious (possibly visuomotor) processesmediated by the dorsal stream intact (Lin amp He 2009)

We have recently published a review of the accumu-lated evidence in relation to this lsquolsquodorsal CFS biasrsquorsquoand conclude that given the currently available dataa dorsal processing bias under CFS cannot beassumed (Ludwig amp Hesselmann 2015) At the neurallevel for example follow-up fMRI studies did notconfirm the finding of preserved dorsal activity underCFS (Fogelson Kohler Miller Granger amp Tse 2014Hesselmann Hebart amp Malach 2011 Hesselmann ampMalach 2011) and showed that CFS-suppressedcategory information (face vs house) could bedecoded from cortex in the ventral stream (Sterzer etal 2008)

As mentioned before both Almeida et al (2014) andSakuraba et al (2012) assume that their tool-primingeffects originate from dorsal visual processes onlyWhile the observed RT differences are interpreted ascategory-by-shape priming effects in the former studyand as shape-priming effects in the latter both studiesexclusively used elongated tools as target stimuli It isinteresting to note that the tool stimuli used by Fangand He (2005) were also exclusively elongated (personalcommunication) Our recent fMRI study howeverdoes not support the notion that CFS-suppressedstimulus elongation (within the tool category) ispreferentially processed in the dorsal visual pathwaysince we found that both dorsal and ventral regions ofinterest were informative of stimulus shape ie facesversus elongated tools (Ludwig Kathmann Sterzer ampHesselmann 2015)

A new and potentially promising avenue of researchcould be to directly investigate the dorsal- and ventral-stream contributions to the priming effects bysystematically varying prime contrast and extractingcontrast-response curves for the observed RTs(Breitmeyer 2014a Tapia amp Breitmeyer 2011) Thecontrast-response curves would then be compared tothe known contrast-response functions for magno-and parvocellular cells (Kaplan amp Shapley 1986) Thisnovel approach however faces two major challengesFirst the mapping between the magno- and parvo-cellular systems and the two visual pathways is notexclusive (Merigan amp Maunsell 1993 Skottun 2014)Second fitting nonlinear functions to RT effects withupper bounds of approximately 8ndash12 ms wouldrequire an enormous number of trials given the noisein RT data

Limitations

The post hoc examination of widthlength ratios ofstimulus exemplars in each category used in Experi-ment 2 showed that the distribution of ratios wasbroader in the case of nonelongated tools as comparedto the other three categories (Supplementary Figure

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 13

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

explain enhanced and decreased priming effects ob-served under different task sets (Kiefer 2012 KieferAdams amp Zovko 2012) According to this modelattentional sensitization of unconscious processing bytask sets is achieved by enhancing the sensitivity oftask-relevant pathways and reducing the sensitivity oftask-irrelevant pathways (Kiefer amp Martens 2010)

Alternatively our results in the shape task might bedue to differences in stimulus orientation To the bestof our knowledge only one previous CFS primingstudy directly addressed this issue (Sakuraba et al2012 experiment 5) They used differently oriented(08 458 908 1358) prime stimuli and found significantpriming effects for elongated tool targets only whengeometric lines were used as primes not when otheroriented stimuli (less elongated diamond shapes andnonelongated lsquolsquoPacmanrsquorsquo stimuli) were used Theauthors concluded lsquolsquoIt is unlikely that the orientationsof elongated shapes have a significant impact on thepriming effects found in previous experimentsrsquorsquo (p3952) Based on this conclusion we did not expectsignificant effects of stimulus orientation in ourexperiments but we nevertheless aimed for anapproximate matching of stimulus orientation be-tween our elongated stimulus categories (ie toolsand animals) However Sakuraba et al (2012) did notreport whether their shape-priming effect was furthermodulated by the orientation differences betweenprimes and targets We addressed this question in acontrol analysis that was limited to conditions withelongated stimuli It revealed that the correlationbetween orientation differences and RT was larger fortrials with tool targets than for trials with animaltargets We conclude that stimulus orientation mighthave modulated the results in the shape task althoughmore experiments (eg with a small set of discretestimulus orientations) will be needed to betterunderstand the magnitude and direction of thismodulation

Are priming effects observed under CFSmediated by the dorsal visual pathway

In an influential fMRI study Fang and He (2005)reported largely preserved neural activity in the dorsalvisual pathway in response to CFS-suppressed imagesof tools while activity in the ventral stream was fullysuppressed In the context of the duplex vision model(Milner amp Goodale 1995 2006) the results of Fangand He were interpreted as evidence that CFS could beused specifically to lsquolsquoisolatersquorsquo or lsquolsquobiasrsquorsquo dorsal visualprocessingmdashie selectively disrupt conscious objectidentification mediated by the ventral stream whileleaving nonconscious (possibly visuomotor) processesmediated by the dorsal stream intact (Lin amp He 2009)

We have recently published a review of the accumu-lated evidence in relation to this lsquolsquodorsal CFS biasrsquorsquoand conclude that given the currently available dataa dorsal processing bias under CFS cannot beassumed (Ludwig amp Hesselmann 2015) At the neurallevel for example follow-up fMRI studies did notconfirm the finding of preserved dorsal activity underCFS (Fogelson Kohler Miller Granger amp Tse 2014Hesselmann Hebart amp Malach 2011 Hesselmann ampMalach 2011) and showed that CFS-suppressedcategory information (face vs house) could bedecoded from cortex in the ventral stream (Sterzer etal 2008)

As mentioned before both Almeida et al (2014) andSakuraba et al (2012) assume that their tool-primingeffects originate from dorsal visual processes onlyWhile the observed RT differences are interpreted ascategory-by-shape priming effects in the former studyand as shape-priming effects in the latter both studiesexclusively used elongated tools as target stimuli It isinteresting to note that the tool stimuli used by Fangand He (2005) were also exclusively elongated (personalcommunication) Our recent fMRI study howeverdoes not support the notion that CFS-suppressedstimulus elongation (within the tool category) ispreferentially processed in the dorsal visual pathwaysince we found that both dorsal and ventral regions ofinterest were informative of stimulus shape ie facesversus elongated tools (Ludwig Kathmann Sterzer ampHesselmann 2015)

A new and potentially promising avenue of researchcould be to directly investigate the dorsal- and ventral-stream contributions to the priming effects bysystematically varying prime contrast and extractingcontrast-response curves for the observed RTs(Breitmeyer 2014a Tapia amp Breitmeyer 2011) Thecontrast-response curves would then be compared tothe known contrast-response functions for magno-and parvocellular cells (Kaplan amp Shapley 1986) Thisnovel approach however faces two major challengesFirst the mapping between the magno- and parvo-cellular systems and the two visual pathways is notexclusive (Merigan amp Maunsell 1993 Skottun 2014)Second fitting nonlinear functions to RT effects withupper bounds of approximately 8ndash12 ms wouldrequire an enormous number of trials given the noisein RT data

Limitations

The post hoc examination of widthlength ratios ofstimulus exemplars in each category used in Experi-ment 2 showed that the distribution of ratios wasbroader in the case of nonelongated tools as comparedto the other three categories (Supplementary Figure

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 13

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

S2) We assume that this broader distribution of widthlength ratios resulted in the long RTs for nonelongatedtools in the shape task (and perhaps also in the categorytask) One might conclude that priming effects in theshape task could also have been modulated by thewidthlength ratio which was not modeled as a randomeffect in the statistical analysis Therefore futurestudies should further investigate the influence of thedistribution parameters of widthlength ratio distribu-tions on shape-priming effects under CFS On the otherhand previous CFS priming studies using tool andanimal stimuli also differed with respect to thesedistributions (Supplementary Figure S2) and showedpriming by shape congruency only suggesting somerobustness of priming to variations in widthlengthratio distributions Finally in our experiments sup-pressed prime stimuli were also presented as fullyvisible target stimuli thus it cannot be excluded thatthe observed priming effects might reflect consequencesof stimulus-response mappings that can act at severallevels (classification decision action) and potentiallylead to facilitating and reverse RT effects (HensonEckstein Waszak Frings amp Horner 2014 see alsoKunde Kiesel amp Hoffmann 2003) One way ofprotecting priming effects against alternate interpreta-tion as stimulus-response mapping effects would be touse naming tasks in combination with primes notincluded in the target stimulus set

Conclusion

We found evidence for priming in a shape task butno evidence for priming in a category task when primestimuli were rendered invisible using CFS Our resultsare in line with the notion that the visual processing ofCFS-suppressed stimuli is limited to the level offeatures and coarse feature configurations The degreeto which the priming effect in the shape task can bemodulated by other low-level (eg stimulus orienta-tion) and high-level (eg target category) informationawaits further investigation

Keywords continuous flash suppression interocularsuppression priming dual-stream model dorsal stream

Acknowledgments

GH and PS were supported by the German ResearchFoundation (grants HE 62441-2 STE 14307-1) KLwas supported by the Studienstiftung des deutschenVolkes (German National Academic Foundation) Wethank Max Pekrul and Leon Kerger for their help withdata acquisition and Maren Suilmann for her supportin revising the manuscript

GH and ND contributed equally to this articleCommercial relationships noneCorresponding author Guido HesselmannEmail guidohesselmanncharitedeAddress Visual Perception Laboratory Department ofPsychiatry and Psychotherapy CharitendashUniversitats-medizin Berlin Germany

Footnotes

1 For an up-to-date CFS reference list please refer tohttpwwwgestaltrevisionbeenresourcesreference-guides83-resourcesreference-guides343

2 We would like to thank one anonymous reviewerfor raising this point

3 Data from our study can be found at httpsfigsharecomsf2dda499bac6eb229f8b

References

Almeida J Mahon B Z amp Caramazza A (2010)The role of the dorsal visual processing stream intool identification Psychological Science 21 772ndash778

Almeida J Mahon B Z Nakayama K ampCaramazza A (2008) Unconscious processingdissociates along categorical lines Proceedings ofthe National Academy of Sciences USA 10515214ndash15218

Almeida J Mahon B Z Zapater-Raberov VDziuba A Cabaco T Marques J F ampCaramazza A (2014) Grasping with the eyes Therole of elongation in visual recognition of manip-ulable objects Cognitive Affective amp BehavioralNeuroscience 14(1) 319ndash335

Bachmann T Breitmeyer B G amp Ogmen H (2007)The experimental phenomena of consciousness Abrief dictionary Oxford UK Oxford UniversityPress

Baker D H Kaestner M amp Gouws A (2015)Comparison of crosstalk for several methods ofstereoscopic presentation Perception 44(4) 455ndash455

Blake R (1988) Dichoptic reading The role ofmeaning in binocular rivalry Perception amp Psy-chophysics 44(2) 133ndash141

Blake R amp Logothetis N K (2002) Visualcompetition Nature Reviews Neuroscience 3(1)13ndash21

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 14

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

Brainard D H (1997) The Psychophysics ToolboxSpatial Vision 10 433ndash436

Breitmeyer B G (2014a) Contributions of magno-and parvocellular channels to conscious and non-conscious vision Philosophical Transactions of theRoyal Society B Biological Sciences 369(1641)20130213

Breitmeyer B G (2014b) Functional hierarchy ofunconscious object processing In The visual(un)conscious amp its (dis)contents (ch 5 pp 89ndash102) Oxford UK Oxford University Press

Breitmeyer B G (2015) Psychophysical lsquolsquoblindingrsquorsquomethods reveal a functional hierarchy of uncon-scious visual processing Consciousness and Cogni-tion 35 234ndash250

Button K S Ioannidis J P Mokrysz C Nosek BA Flint J Robinson E S amp Munafo M R(2013) Power failure Why small sample sizeundermines the reliability of neuroscience NatureReviews Neuroscience 14(5) 365ndash376

Cave C B Blake R amp McNamara T P (1998)Binocular rivalry disrupts visual priming Psycho-logical Science 9 299ndash302

Cousineau D (2005) Confidence intervals in within-subject designs A simpler solution to Loftus andMassonrsquos method Tutorials in Quantitative Meth-ods for Psychology 1 42ndash45

Dehaene S amp Changeux J P (2011) Experimentaland theoretical approaches to conscious processingNeuron 70(2) 200ndash227

Fabre-Thorpe M (2011) The characteristics and limitsof rapid visual categorization Frontiers in Psy-chology 2 243

Fang F amp He S (2005) Cortical responses toinvisible objects in the human dorsal and ventralpathways Nature Neuroscience 10 1380ndash1385

Faul F Erdfelder E Lang A G amp Buchner A(2007) GPower 3 A flexible statistical poweranalysis program for the social behavioral andbiomedical sciences Behavior Research Methods39(2) 175ndash191

Fogelson S V Kohler P J Miller K J GrangerR amp Tse P U (2014) Unconscious neuralprocessing differs with method used to renderstimuli invisible Frontiers in Psychology 5 1ndash11

Francis G (2013) Replication statistical consistencyand publication bias Journal of MathematicalPsychology 57(5) 153ndash169

Gayet S Van der Stigchel S amp Paffen C L (2014)Breaking continuous flash suppression Competingfor consciousness on the pre-semantic battlefieldFrontiers in Psychology 5 460

Hebart M N amp Hesselmann G (2012) What visualinformation is processed in the human dorsalstream The Journal of Neuroscience 32(24) 8107ndash8109

Henson R N Eckstein D Waszak F Frings C ampHorner A J (2014) Stimulus-response bindings inpriming Trends in Cognitive Sciences 18(7) 376ndash384

Hesselmann G (2013) Dissecting visual awarenesswith FMRI The Neuroscientist 19(5) 495ndash508

Hesselmann G Darcy N Sterzer P amp Knops A(2015) Exploring the boundary conditions ofunconscious numerical priming effects with con-tinuous flash suppression Consciousness and Cog-nition 31 60ndash72

Hesselmann G Hebart M amp Malach R (2011)Differential BOLD activity associated with subjec-tive and objective reports during lsquolsquoblindsightrsquorsquo innormal observers The Journal of Neuroscience31(36) 12936ndash12944

Hesselmann G amp Knops A (2014) No conclusiveevidence for numerical priming under interocularsuppression Psychological Science 25(11) 2116ndash2119

Hesselmann G amp Malach R (2011) The linkbetween fMRI-BOLD activation and perceptualawareness is lsquolsquostream-invariantrsquorsquo in the humanvisual system Cerebral Cortex 21(12) 2829ndash2837

Hesselmann G amp Moors P (2015) Definitely maybeCan unconscious processes perform the samefunctions as conscious processes Frontiers inPsychology 6 584

Heyman T amp Moors P (2014) Frequent words donot break continuous flash suppression differentlyfrom infrequent or nonexistent words Implicationsfor semantic processing of words in the absence ofawareness PLoS One 9(8) e104719

Izatt G Dubois J Faivre N amp Koch C (2014) Adirect comparison of unconscious face processingunder masking and interocular suppression Fron-tiers in Psychology 5 659

Jiang Y amp He S (2006) Cortical responses toinvisible faces Dissociating subsystems for facial-information processing Current Biology 16 2023ndash2029

Judd C M Westfall J amp Kenny D A (2012)Treating stimuli as a random factor in socialpsychology A new and comprehensive solution toa pervasive but largely ignored problem Journal ofPersonality and Social Psychology 103(1) 54ndash69

Kang M S Blake R amp Woodman G F (2011)Semantic analysis does not occur in the absence of

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 15

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

awareness induced by interocular suppression TheJournal of Neuroscience 31(38) 13535ndash13545

Kaplan E amp Shapley R M (1986) The primateretina contains two types of ganglion cells withhigh and low contrast sensitivity Proceedings of theNational Academy of Sciences USA 83(8) 2755ndash2757

Kiefer M (2012) Executive control over unconsciouscognition Attentional sensitization of unconsciousinformation processing Frontiers in Human Neu-roscience 6 61

Kiefer M Adams S C amp Zovko M (2012)Attentional sensitization of unconscious visualprocessing Top-down influences on masked prim-ing Advances in Cognitive Psychology 8(1) 50ndash61

Kiefer M amp Martens U (2010) Attentional sensiti-zation of unconscious cognition Task sets modu-late subsequent masked semantic priming Journalof Experimental Psychology General 139(3) 464ndash489

Kim C-Y amp Blake R (2005) Psychophysical magicRendering the visible lsquolsquoinvisiblersquorsquo Trends in Cogni-tive Sciences 9 381ndash388

Kouider S amp Dehaene S (2007) Levels of processingduring non-conscious perception A critical reviewof visual masking Philosophical Transactions of theRoyal Society B 362 857ndash875

Kunde W Kiesel A amp Hoffmann J (2003)Conscious control over the content of unconsciouscognition Cognition 88(2) 223ndash242

Lin Z amp He S (2009) Seeing the invisible The scopeand limits of unconscious processing in binocularrivalry Progress in Neurobiology 87(4) 195ndash211

Lin Z amp Murray S O (2014) Priming of awarenessor how not to measure visual awareness Journal ofVision 14(1)27 1ndash17 doi10116714127[PubMed] [Article]

Ludwig K amp Hesselmann G (2015) Weighing theevidence for a dorsal processing bias undercontinuous flash suppression Consciousness andCognition 35 251ndash259

Ludwig K Kathmann N Sterzer P amp HesselmannG (2015) Investigating category- and shape-selective neural processing in ventral and dorsalvisual stream under interocular suppression Hu-man Brain Mapping 36(1) 137ndash149

Ludwig K Sterzer P Kathmann N Franz V Hamp Hesselmann G (2013) Learning to detect butnot to grasp suppressed visual stimuli Neuro-psychologia 51(13) 2930ndash2938

Merigan W H amp Maunsell J H (1993) How

parallel are the primate visual pathways AnnualReview of Neuroscience 16 369ndash402

Miles W R (1930) Ocular dominance in humanadults Journal of General Psychology 3 412ndash429

Milner A D amp Goodale M A (1995) The visualbrain in action Oxford UK Oxford UniversityPress

Milner A D amp Goodale M A (2006) The visualbrain in action (2nd ed) Oxford UK OxfordUniversity Press

Morey R D (2008) Confidence intervals fromnormalized data A correction to Cousineau (2005)Tutorials in Quantitative Methods for Psychology 461ndash64

Mouchetant-Rostaing Y Giard M H Delpuech CEchallier J F amp Pernier J (2000) Early signs ofvisual categorization for biological and non-bio-logical stimuli in humans NeuroReport 11(11)2521ndash2525

Pelli D G (1997) The VideoToolbox software forvisual psychophysics Transforming numbers intomovies Spatial Vision 10(4) 437ndash442

Peremen Z amp Lamy D (2014) Comparing uncon-scious processing during continuous flash suppres-sion and meta-contrast masking just under thelimen of consciousness Frontiers in Psychology 5969

Ramsoy T Z amp Overgaard M (2004) Introspectionand subliminal perception Phenomenology and theCognitive Sciences 3 1ndash23

Sakuraba S Sakai S Yamanaka M Yokosawa Kamp Hirayama K (2012) Does the human dorsalstream really process a category for tools TheJournal of Neuroscience 32(11) 3949ndash3953

Salmon J P McMullen P A amp Filliter J H (2010)Norms for two types of manipulability (graspabilityand functional usage) familiarity and age ofacquisition for 320 photographs of objects Behav-ior Research Methods 42(1) 82ndash95

Schmidt T (2015) Invisible stimuli implicit thresh-olds Why invisibility judgments cannot be inter-preted in isolation Advances in CognitivePsychology 11 31ndash41

Shimaoka D amp Kaneko K (2011) Dynamicalsystems modeling of Continuous Flash Suppres-sion Vision Research 51(6) 521ndash528

Sklar A Y Levy N Goldstein A Mandel RMaril A amp Hassin R R (2012) Reading anddoing arithmetic nonconsciously Proceedings of theNational Academy of Sciences USA 109(48)19614ndash19619

Skottun B C (2014) The magnocellular system versus

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 16

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

awareness induced by interocular suppression TheJournal of Neuroscience 31(38) 13535ndash13545

Kaplan E amp Shapley R M (1986) The primateretina contains two types of ganglion cells withhigh and low contrast sensitivity Proceedings of theNational Academy of Sciences USA 83(8) 2755ndash2757

Kiefer M (2012) Executive control over unconsciouscognition Attentional sensitization of unconsciousinformation processing Frontiers in Human Neu-roscience 6 61

Kiefer M Adams S C amp Zovko M (2012)Attentional sensitization of unconscious visualprocessing Top-down influences on masked prim-ing Advances in Cognitive Psychology 8(1) 50ndash61

Kiefer M amp Martens U (2010) Attentional sensiti-zation of unconscious cognition Task sets modu-late subsequent masked semantic priming Journalof Experimental Psychology General 139(3) 464ndash489

Kim C-Y amp Blake R (2005) Psychophysical magicRendering the visible lsquolsquoinvisiblersquorsquo Trends in Cogni-tive Sciences 9 381ndash388

Kouider S amp Dehaene S (2007) Levels of processingduring non-conscious perception A critical reviewof visual masking Philosophical Transactions of theRoyal Society B 362 857ndash875

Kunde W Kiesel A amp Hoffmann J (2003)Conscious control over the content of unconsciouscognition Cognition 88(2) 223ndash242

Lin Z amp He S (2009) Seeing the invisible The scopeand limits of unconscious processing in binocularrivalry Progress in Neurobiology 87(4) 195ndash211

Lin Z amp Murray S O (2014) Priming of awarenessor how not to measure visual awareness Journal ofVision 14(1)27 1ndash17 doi10116714127[PubMed] [Article]

Ludwig K amp Hesselmann G (2015) Weighing theevidence for a dorsal processing bias undercontinuous flash suppression Consciousness andCognition 35 251ndash259

Ludwig K Kathmann N Sterzer P amp HesselmannG (2015) Investigating category- and shape-selective neural processing in ventral and dorsalvisual stream under interocular suppression Hu-man Brain Mapping 36(1) 137ndash149

Ludwig K Sterzer P Kathmann N Franz V Hamp Hesselmann G (2013) Learning to detect butnot to grasp suppressed visual stimuli Neuro-psychologia 51(13) 2930ndash2938

Merigan W H amp Maunsell J H (1993) How

parallel are the primate visual pathways AnnualReview of Neuroscience 16 369ndash402

Miles W R (1930) Ocular dominance in humanadults Journal of General Psychology 3 412ndash429

Milner A D amp Goodale M A (1995) The visualbrain in action Oxford UK Oxford UniversityPress

Milner A D amp Goodale M A (2006) The visualbrain in action (2nd ed) Oxford UK OxfordUniversity Press

Morey R D (2008) Confidence intervals fromnormalized data A correction to Cousineau (2005)Tutorials in Quantitative Methods for Psychology 461ndash64

Mouchetant-Rostaing Y Giard M H Delpuech CEchallier J F amp Pernier J (2000) Early signs ofvisual categorization for biological and non-bio-logical stimuli in humans NeuroReport 11(11)2521ndash2525

Pelli D G (1997) The VideoToolbox software forvisual psychophysics Transforming numbers intomovies Spatial Vision 10(4) 437ndash442

Peremen Z amp Lamy D (2014) Comparing uncon-scious processing during continuous flash suppres-sion and meta-contrast masking just under thelimen of consciousness Frontiers in Psychology 5969

Ramsoy T Z amp Overgaard M (2004) Introspectionand subliminal perception Phenomenology and theCognitive Sciences 3 1ndash23

Sakuraba S Sakai S Yamanaka M Yokosawa Kamp Hirayama K (2012) Does the human dorsalstream really process a category for tools TheJournal of Neuroscience 32(11) 3949ndash3953

Salmon J P McMullen P A amp Filliter J H (2010)Norms for two types of manipulability (graspabilityand functional usage) familiarity and age ofacquisition for 320 photographs of objects Behav-ior Research Methods 42(1) 82ndash95

Schmidt T (2015) Invisible stimuli implicit thresh-olds Why invisibility judgments cannot be inter-preted in isolation Advances in CognitivePsychology 11 31ndash41

Shimaoka D amp Kaneko K (2011) Dynamicalsystems modeling of Continuous Flash Suppres-sion Vision Research 51(6) 521ndash528

Sklar A Y Levy N Goldstein A Mandel RMaril A amp Hassin R R (2012) Reading anddoing arithmetic nonconsciously Proceedings of theNational Academy of Sciences USA 109(48)19614ndash19619

Skottun B C (2014) The magnocellular system versus

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 16

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016

the dorsal stream Frontiers in Human Neurosci-ence 8 786

Sterzer P (2013) Functional neuroimaging of binoc-ular rivalry In S M Miller (Ed) The constitutionof visual consciousness Lessons from binocularrivalry (pp 187ndash210) Amsterdam the NetherlandsJohn Benjamins

Sterzer P Haynes J D amp Rees G (2008) Fine-scaleactivity patterns in high-level visual areas encodethe category of invisible objects Journal of Vision8(15)10 1ndash12 doi10116781510 [PubMed][Article]

Sterzer P Stein T Ludwig K Rothkirch M ampHesselmann G (2014) Neural processing of visualinformation under interocular suppression Acritical review Frontiers in Psychology 5 1ndash12

Tapia E amp Breitmeyer B G (2011) Visual con-sciousness revisited Magnocellular and parvocel-lular contributions to conscious and nonconsciousvision Psychological Science 22(7) 934ndash942

Tsuchiya N amp Koch C (2005) Continuous flashsuppression reduces negative afterimages NatureNeuroscience 8 1096ndash1101

Tsuchiya N Koch C Gilroy L A amp Blake R(2006) Depth of interocular suppression associatedwith continuous flash suppression flash suppres-sion and binocular rivalry Journal of Vision 6(10)

6 1068ndash1078 doi1011676106 [PubMed][Article]

Tukey J W (1977) Exploratory data analysisReading MA Addison-Wesley

Whelan R (2008) Effective analysis of reaction timedata The Psychological Record 58 475ndash482

Willenbockel V Sadr J Fiset D Horne G OGosselin F amp Tanaka J W (2010) Controllinglow-level image properties The SHINE toolboxBehavior Research Methods 42(3) 671ndash684

Yamashiro H Yamamoto H Mano H UmedaM Higuchi T amp Saiki J (2014) Activity in earlyvisual areas predicts interindividual differences inbinocular rivalry dynamics Journal of Neurophys-iology 111(6) 1190ndash1202

Yang E amp Blake R (2012) Deconstructing contin-uous flash suppression Journal of Vision 12(3)81ndash13 doi1011671238 [PubMed] [Article]

Yang E Brascamp J Kang M S amp Blake R(2014) On the use of continuous flash suppressionfor the study of visual processing outside ofawareness Frontiers in Psychology 5 724

Zimba L D amp Blake R (1983) Binocular rivalry andsemantic processing Out of sight out of mindJournal of Experimental Psychology Human Per-ception and Performance 9(5) 807ndash815

Journal of Vision (2016) 16(3)17 1ndash17 Hesselmann Darcy Ludwig amp Sterzer 17

Downloaded From httpjovarvojournalsorgpdfaccessashxurl=dataJournalsJOV934914 on 05182016