unconscious cross-modal priming of auditory sound ......comb, 2007; kouider & dehaene, 2009; kouider...

Unconscious Cross-Modal Priming of Auditory Sound Localizationby Visual Words

Ulrich AnsorgeUniversity of Vienna

Shah KhalidUniversity of Vienna and University of Osnabrück

Bernhard LabackAustrian Academy of Sciences, Vienna, Austria

Little is known about the cross-modal integration of unconscious and conscious information. In thecurrent study, we therefore tested whether the spatial meaning of an unconscious visual word, such as up,influences the perceived location of a subsequently presented auditory target. Although cross-modalintegration of unconscious information is generally rare, unconscious meaning stemming from only 1particular modality could, in principle, be available for other modalities. Also, on the basis of knowninfluences and dependencies of meaning on sensory information processing, such an unconsciousmeaning-based effect could impact sensory processing in a different modality. In 3 experiments, thisprediction was confirmed. We found that an unconscious spatial word, such as up, facilitated positiondiscrimination of a spatially congruent sound (here, a sound from above) as compared to a spatiallyincongruent sound (here, from below). This was found even though participants did not recognize themeaning of the primes. The results show that unconscious processing extends to semantic–sensoryconnections between different modalities.

Keywords: cross-modal priming, multisensory priming, unconscious processing, visual auditory integra-tion, auditory localization

Human information processing draws on both conscious andunconscious processing modes, but how these two modes interactis not entirely understood (Dehaene & Changeux, 2011). One wayof investigating this question is by studying the capabilities and limitsof unconscious processing (Ansorge, Kunde, & Kiefer, 2014; Atkin-son, Thomas, & Cleeremans, 2000; Dijksterhuis & Aarts, 2010). Inperception, unconscious processing can be studied with masked prim-ing. In vision, masking denotes a method of reducing prime visibilityby the presentation of a masking stimulus (Breitmeyer & Ogmen,2006). In masked priming, an unconscious prime is presented prior toa task-relevant target. Participants respond to the consciously per-ceived target, and the unconscious (masked) prime influences theresponses (Dehaene et al., 1998). The prime’s imperceptibility,together with the priming effect, offers an excellent tool for study-ing unconscious processing.

Past studies demonstrated unconscious priming with primes andtargets in the same modality—for example, visual primes andtargets (Dehaene & Changeux, 2011)—but little is known aboutunconscious priming across modalities (Lamy, Mudrik, & Deouell,2008). Cross-modal priming research is mostly restricted to con-scious priming (cf. Baars, 1988; Dehaene & Naccache, 2001).However, unconscious word and number priming are exceptions.A handful of studies have shown that unconscious visual words (ornumbers) prime auditory target words (or numbers; Faivre,Mudrik, Schwartz, & Koch, 2014; Grainger, Diependaele, Spinelli,Ferrand, & Farioli, 2003; Kiyonaga, Grainger, Midgley, & Hol-comb, 2007; Kouider & Dehaene, 2009; Kouider & Dupoux, 2001;Nakamura et al., 2006).

Yet, unconscious cross-modal priming should not be restrictedto interactions between visual and auditory words or numbersbecause semantic memory also connects (word) meaning withsensory information from different modalities (Barsalou, 1999;Chen & Spence, 2011; Glaser & Glaser, 1989). Take the exampleof perceived space (Regier & Carlson, 2001). All perceived stimuliare located in space and, accordingly, sensory representationsinclude a spatial index specifying stimulus location. Yet, to usesuch a sensory spatial index for various purposes such as actions,attention shifts, or judgments, a semantic (or conceptual) spatialreference frame has to be imposed onto the spatial index (Logan,1995). For instance, if you locate two objects in the sky—a planeand a missile approaching the plane from below—both theseobjects are above you in a conceptual spatial frame with an originthat is centered on you. However, the missile would be below theplane, so to direct your attention from the plane to the missile, you

Ulrich Ansorge, Faculty of Psychology, University of Vienna; ShahKhalid, Faculty of Psychology, University of Vienna, and Institute ofCognitive Science, University of Osnabrück; Bernhard Laback, AcousticsResearch Institute, Austrian Academy of Sciences, Vienna, Austria.

We declare shared first authorship of Ulrich Ansorge and Shah Khalid.This study was supported by Deutsche Forschungsgemeinschaft Grant KH341/1-1 to Shah Khalid. Thanks to Annabella Puhr and Maximilian Steinfor help with the data collection.

Correspondence concerning this article should be addressed to UlrichAnsorge, Faculty of Psychology, University of Vienna, Liebiggasse 5,A-1010 Vienna, Austria. E-mail: [email protected]

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Journal of Experimental Psychology:Learning, Memory, and Cognition

© 2015 American Psychological Association

2015, Vol. 41, No. 6, 0000278-7393/15/$12.00 http://dx.doi.org/10.1037/xlm0000217

1

AQ: auAQ: 1

AQ: 2

AQ: 3

tapraid5/zfv-xlm/zfv-xlm/zfv00116/zfv3316d15z xppws S�1 11/10/15 13:35 Art: 2015-1890APA NLM

mailto:[email protected]://dx.doi.org/10.1037/xlm0000217

would have to shift the origin of the conceptual spatial frame fromyourself as a center onto the plane as its center. Obviously, theusage of sensory spatial indices for such purposes as locating astimulus relative to you or to another object requires the projectionof a particular conceptual spatial frame onto the object in questionso as to specify, for example, the deictic relation between twoobjects (e.g., as above, below, right, or left of one another).Because of such close connections between the sensory and thesemantic domain (Glaser & Glaser, 1989; Logan, 1995) and be-cause unimodal unconscious semantic priming by visual words hasbeen demonstrated (Kiefer, 2002; McCauley, Parmelee, Sperber,& Carr, 1980), unconscious cross-modal priming between a primeword’s meaning in one modality and the perception of sensoryfeatures of a nonword target in a different modality should bepossible once there is a semantic connection between a word’smeaning and a sensory feature. Here, we used the spatial discrim-ination of auditory target locations to study this hypothesis.

Experiment 1

In Experiment 1, we studied unconscious cross-modal primingbetween seen words and heard locations. In the first stage of theexperiment, participants localized an auditory stimulus as comingfrom above or below. We used a taxing localization task in whichparticipants had to discriminate between relatively small elevationdifferences based on participant-specific head-related transferfunctions (HRTFs). Prior to this auditory target, participants saw aconscious or unconscious visual word. This prime had a spatialmeaning that was equally likely to be congruent or incongruent tothe target’s position. For example, the word up was congruent ifthe following target was a sound from above, but it was incongru-ent if the target was from below. Participants were informed thatprimes were not informative to the sound localization task. How-ever, word primes were presented in the participants’ line of sight.This could be sufficient for a semantic priming effect of uncon-scious task-irrelevant word primes (Naccache & Dehaene, 2001).We used a relatively difficult auditory task so that performancewas not perfect. In this way, we hoped to provide sufficient roomfor a subliminal priming effect.

In the second stage of the experiment, we evaluated the visibilityof the primes. Here, the task of the participants was to judge theprime–target relation as congruent versus incongruent. As com-pared to a pure prime discrimination, this task has the advantage ofasking for the participants’ awareness of exactly that characteristicof the stimuli (i.e., the congruence vs. incongruence between primeand target) that was expected to lead to a behavioral priming effect(cf. Desender, Van Opstal, & Van den Bussche, 2014). The taskalso has a disadvantage. It is more demanding than a simple primediscrimination and thus may not be as sensitive as other conceiv-able visibility tests. Therefore, across experiments, the visibilitytest varied, and simpler tasks were used in Experiments 2 and 3.

Method

Participants. Twenty-two participants were tested. Here andin the following experiments, participants had normal or corrected-to-normal vision and were paid 7 €/hr. Also, informed consent wasobtained from all participants, and the participants were treated inaccordance with American Psychological Association standards

and the rules of the Declaration of Helsinki. One participant had tobe excluded because of too high a number of correct prime–targetcongruence judgments in masked conditions. He had a d= of 1.1 inthe masked condition that exceeded the group mean (d= � 0.2) bymore than 2 standard deviations (SD � 0.3). Six additional par-ticipants had to be excluded because of chance performance(around 50%) in the sound localization task, indicating that theywere unable to localize the targets. The remaining 15 participants(12 female, three male, Mage � 28.2 years, age range: 24–42years) were analyzed.

Apparatus, stimuli, and procedure. Visual stimuli were pre-sented on a 17-in. (43-cm) color flat screen display with a refreshrate of 59.1 Hz via an NVIDIA GeForce GT 220 (with 1,024 MB)graphics adapter. The participants sat at a distance of 57 cm fromthe screen in a quiet, dimly lit room, with their head resting in achin rest to ensure a constant viewing distance and a straight-aheadgaze direction. Target responses were registered through the keysC and M labeled as left and right of a standard keyboard. Partic-ipants operated the keys with their left and right index fingers. Thetargets were two spatial sound stimuli generated using a virtualacoustics technique. A 300-ms broadband white Gaussian noisetoken, temporally shaped with a Tukey window with a 10-msrise–fall time, was filtered with participant-specific directionaltransfer functions (DTFs). DTFs were derived from HRTFs, whichwere measured in a prior session, using the experimental setup andpostprocessing techniques described in Appendix A. Targets’ spa-tial positions were above (with an elevation angle of �30° in themedian plane) and below (with an elevation angle of �30° in themedian plane). Targets were presented using Sennheiser 380 Proheadphones (with �0.1% total harmonic distortion). Visual primestimuli were 20 German words denoting directions or positions onthe vertical axis. All primes were high-frequency words, with morethan 60 instances in 1 million words. The 10 spatial up primeswere the words oben (on top), darüber (above), hinauf, aufwärts,empor (upward), hoch, gehoben, erhöht (elevated), aufsteigend,and steigend (rising), with a mean word length of Ø � 6.6 letters(range of four to 11 letters). The 10 spatial down primes were thewords unten (down), darunter (below), hinab, abwärts, herab(downward), niedrig (low), gesenkt (lowered), abfallend, sinkend(declining), and tief (deep), with a mean word length of Ø � 6.3letters (range of four to nine letters). Up and down words thus hada comparable length and frequency, and they were easily andequally discriminable by their spatial category membership (en-sured by empirical pretesting; see Ansorge & Bohner, 2014). (Forfurther details, refer to Ansorge, Khalid, & König, 2013.)

Each prime word was shown equally often (16 times per each oftwo stages of the experiment). For the creation of the prime–targetpairs, each of the equally often presented sound locations wasrandomly combined with each of the 20 spatial primes. The re-sulting prime–target pairs were equally likely to be congruent andincongruent.

Figure 1 shows the sequence of events in a trial. The fixationcross, prime, and masks were all presented at the screen center.They were black (�1 cd/m2) against a gray background (24cd/m2). A trial started with the fixation cross presented for 750 ms.In masked trials, next a forward mask was presented. It consistedof 10 randomly drawn uppercase letters and was shown for 200ms. Then, the prime word occurred for 34 ms (in the short stimulusonset asynchrony [SOA] condition) or for 68 ms (in the long SOA

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

2 ANSORGE, KHALID, AND LABACK

AQ: 4

AQ: 5

AQ: 6

F1


condition). Two different SOAs were used because, due to a lackof fitting prior studies, we were not sure how quickly the primingeffect would develop and dissipate. The prime was always inlowercase letters. A backward mask was presented after the prime.The backward mask also consisted of 10 uppercase letters, whichwere drawn independently of the letters of the forward mask. Thebackward mask was shown for 34 ms (leading to an SOA of 68 msin the short SOA condition) or for 68 ms (leading to an SOA of136 ms in the long SOA condition). Finally, the target was pre-sented. The intervals between successive events were always 0 ms.In unmasked trials, everything was the same, with the exceptionthat both forward and backward masks were omitted and replacedby blank screens.

The experiment consisted of two blocked stages, beginning witha sound localization task followed by a prime–target congruencediscrimination task. During the localization task, participantsjudged the target stimulus as coming from above or below. Inpsychophysical terms, this corresponds to a sound source positiondiscrimination task. Half of the participants were instructed topress the right key for positions above and the left key for positionsbelow. The other half of the participants got the reversed stimulus–response (S–R) mapping. After each incorrect response or if aresponse exceeded 1,250 ms, participants received feedback abouttheir error or their too-slow responses. Feedback lasted for 750 ms.Thus, keeping high accuracy and a fast response was mildlyrewarded (i.e., saved 750 ms per trial that would otherwise be usedfor the feedback).

The second stage was a prime visibility task. For half of theparticipants, this task required that the right key be pressed in trialsin which the prime was congruent to the target and that the left keybe pressed if the prime was incongruent to the target. The otherhalf of the participants got the reverse S–R assignment. Before theprime visibility test, the levels of the variable prime–target con-gruence were carefully explained to the participants, with relevantexamples in the instructions. In the prime visibility stage, the

temporal structure of events was identical to the localization dis-crimination stage so that the processing requirements with respectto the targets were the same in the two tasks. As a prime visibilitytest, the task of discriminating congruent from incongruent trialsrequires processing of the prime and target and thus tests thecritical dimension responsible for the priming effect in the local-ization task. No feedback was given on the correctness of re-sponses.

Each stage consisted of 320 trials. In total, this involved 40 trialsof each combination of the two levels of the variables of PrimeVisibility (masked, unmasked) � 2 Prime–Target SOAs (short, 68ms; long, 136 ms) � 2 Prime–Target Congruence Relations (con-gruent, incongruent). The conditions were presented in a com-pletely randomized order. Prior to both stages, participants werecarefully instructed about their upcoming task. Before the local-ization task, participants were also informed about the fact thatthe primes were uninformative about the targets, and before theprime visibility task, participants were informed about the fact thatthe specific target locations were uninformative about the congru-ence versus incongruence of the primes. Also, the participantspracticed the task for a minimum of 32 trials and, if they wanted,they could extend the practice by another 20 trials. We used thesame stimuli for the test trials as for data collection. In addition tothe written instructions on the display, if necessary (i.e., if therewere questions), before and during practice, the task was explainedverbatim in more detail. Together with two short breaks at regularintervals within stages and one break between the stages, theexperiment took about 1 hr.

Results

Sound localization. Out of all correct responses, 3.2% wereexcluded because their reaction times (RTs) deviated by more than2 standard deviations from a respective condition’s and individualparticipant’s mean RT. See Appendix B and Figure 2 for meancorrect RTs, error rates (ERs), and prime visibility indices. It canbe observed that participants’ performance in the congruent con-ditions was better than in the incongruent conditions at both shortand long SOAs and in the masked as well as unmasked conditions.This was also confirmed by formal analyses.

An initial analysis of variance (ANOVA) with the between-participants variable of S–R mapping did not show a significantmain effect of mapping nor any significant interaction with thisvariable. Data were therefore collapsed across mappings for arepeated-measures ANOVA of the means of the correct RTs withthe within-participant variables of congruence (congruent, incon-gruent), prime–target SOA (short, long), and prime visibility(masked, unmasked). Bonferroni adjustments for multiple compar-isons and an alpha level of .05 were applied here and throughoutthe study.

A significant main effect of congruence, F(1, 14) � 6.52, p �.02, Cohen’s f � 0.69, was found, reflecting faster RTs in con-gruent (607 ms, SD � 38) than incongruent (619 ms, SD � 37)conditions. No other significant main effect or interaction wasfound, all Fs � 1.0. An additional pair sampled t test confirmedsignificantly faster responses in the masked congruent condition(M � 600 ms, SD � 36) than in the masked incongruent condition(M � 616 ms, SD � 42), t(14) � 2.65, p � .02. A similar t test forthe unmasked congruent condition (M � 614 ms, SD � 43) and

Figure 1. Depicted are schematic sequences of a masked trial on the leftand an unmasked trial on the right. The up and down words represent thevisual primes, and the speaker symbol at the end of the trial represents thesound targets. The arrow depicts the flow of time. Stimulus durations werenot shown on the screen. Figures are not drawn to scale. For further details,refer to the Method section.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

3UNCONSCIOUS AND CONSCIOUS CROSS-MODAL PRIMING

F2


incongruent condition (M � 623 ms, SD � 41) was not significant,t(14) � 0.92, p � .38.

A repeated-measures ANOVA of the ERs with the same vari-ables showed a lower mean accuracy in incongruent (ER � 15.0%,SD � 7.2) than congruent (ER � 12.2%, SD � 5.5) conditions.This was reflected in an almost significant main effect of congru-ence, F(1, 14) � 4.43, p � .054, Cohen’s f � 0.56. There was alsoa significant main effect of prime visibility, F(1, 14) � 5.61, p �.03, Cohen’s f � 0.64, with higher ERs for unmasked (15.3%,SD � 7.0) than masked primes (11.9%, SD � 6.0). No othersignificant main effect or interaction was found, all Fs � 1.0.

Prime visibility. To test whether participants failed to see themasked primes but identified the unmasked primes, we computedd=, an index of stimulus visibility (Reingold & Merikle, 1988).Individual d= was computed separately for masked and unmaskedprimes and for primes at shorter and longer SOAs. For calculationof d=, congruent trials counted as signals and incongruent trialscounted as noise. Accordingly, correct (i.e., congruent) judgmentsin congruent trials figured as hits, and incorrect (i.e., incongruent)judgments in congruent trials figured as false alarms (FAs).

The participants were not able to discriminate the maskedprimes’ prime–target pairs with better-than-chance accuracy. Forthe masked primes at the shorter SOA, d= was 0.08 (SD � 0.4),t(14) � 0.77, p � .30, and at the longer SOA it was 0.16 (SD �0.31), t(14) � 1.97, p � .07. The unmasked primes’ prime–targetpairs were successfully discriminated with the shorter SOA, d= �1.31 (SD � 2.0), t(14) � 2.51, p � .02, and the longer SOA, d= �1.33 (SD � 2.3), t(14) � 2.21, p � .04.

Congruence and prime visibility correlations. We alsotested whether individual mean RT congruence effects (incongru-ent RT � congruent RT) were correlated with individual primevisibility indices in the masked condition, as would be expected ifprime visibility accounted for the RT congruence effect of themasked primes. We found no significant correlation betweenmasked primes’ RT congruence effects and d=, Pearson’sr(13) � �.16, p � .60.

Discussion

Results showed a cross-modal priming effect. Participants werebetter at localizing the sounds preceded by congruent than byincongruent visual primes. This was found in particular for uncon-scious primes, meaning that our participants probably activelysuppressed processing of the nonpredictive conscious primes (Ki-noshita, Mozer, & Forster, 2011). In addition, the primes were notvisible in the masked conditions but were visible in the unmaskedconditions. Thus, it seems as if the cross-modal priming effect didnot depend on the participants’ awareness of the primes. Thisconclusion was also supported by a second observation: No sig-nificant correlation was found between the RT congruence effectand the prime visibility indices of the masked primes.

Yet, how could an unconscious visual word have influencedprocessing of the auditory targets? We think that semantic primingcould have influenced processing of the location of the auditorytargets via a joint representation of semantic and sensory informa-tion. According to this explanation, both a visual prime’s meaningand an auditory target’s location would have activated units in ajoint semantic network connecting word meaning with associatedsensory features in other modalities (cf. Chen & Spence, 2011).For instance, congruent spatial word meanings could have preac-tivated units that were subsequently used for processing of audi-tory locations. Because of the word prime’s preactivation of a unit,the critical threshold activity of such a unit that was also used forsuccessful processing of auditory positions would have requiredless additional activation by the congruently primed than by theincongruently primed target sounds.

An alternative explanation of the unconscious priming effect interms of response priming is less likely. First, S–R mapping had noeffect. Thus, the primes did not lead to an orthogonal compatibilityeffect, with faster responses to spatially compatible (e.g., the primeabove before a right response) than spatially incompatible primes(the prime above before a left response; see Proctor & Cho, 2006).Second, the visual primes were very different from the auditorytargets. Response priming on the basis of prime–target fusion(Norris & Kinoshita, 2008) or prime–target confusion (cf. Damian,2001; Kunde, Kiesel, & Hoffmann, 2003) was therefore alsounlikely.

Experiment 2

In Experiment 2, we included neutral nonspatial primes inaddition to the spatial primes. The neutral primes denoted shapes(e.g., angular) or surface properties (e.g., brittle). In the auditorytarget discrimination task, this allowed us to test (a) whethercongruent spatial primes facilitated target discrimination as com-pared to neutral primes and/or (b) whether incongruent spatialprimes delayed target discrimination relative to neutral primes.

Figure 2. Results of Experiment 1: mean reaction times (RTs) in milli-seconds (upper panel) and error rates (ERs) as a percentage (lower panel)on the ordinate plotted as a function of congruence, stimulus onset asyn-chrony (SOA), and prime visibility on the abscissa. Standard errors areindicated by bars. Details about the mean RTs and ERs in each of theconditions as well as results of the analysis of the prime’s visibility aresummarized in Appendix B. Congr. � congruent; Incongr. � incongruent.The dark symbols represent the short SOA, and the light symbols representthe long SOA. The circles represent the masked condition, and the trianglesrepresent the unmasked condition.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 7

AQ: 8


Also, in Experiment 2’s prime visibility test, we asked directlyfor the prime’s meaning category—that is, whether the prime wasa spatial word or a nonspatial word.1 This test might not be asstraightforward as the prime–target congruence discrimination thatwe used in Experiment 1. However, the present prime visibilitytest asks for prime meaning only and is therefore maybe easier andthus more sensitive for the participants’ residual awareness of theprimes than the visibility test of Experiment 1.

Method

Participants. Sixteen participants were tested. Four partici-pants had to be excluded by the same criterion of chance perfor-mance in the sound localization task as in Experiment 1. Theremaining 12 participants (10 female, two male, Mage � 28.4years, age range: 23–44 years) were analyzed.

Apparatus, stimuli, and procedure. These were the same asin Experiment 1, except that in addition to the 20 spatial words, 20spatially neutral German words about the surface and shape prop-erties of objects were included as primes. These neutral wordswere hart (hard), fest (tight), glatt (smooth), eben (firm), eckig(angular), kantig (edgy), stabil (stable), spröde (brittle), steif(stiff), brüchig (fragile), weich (soft), lose (loose), rau (rough),rissig (cracked), rund (round), warm (warm), kalt (cold), biegsam(bendable), klebrig (sticky), and haftend (adhesive), with a meanword length of Ø � 5.2 letters (range of three to seven letters) andmoderate frequencies between 1.0 � 10�7 (haftend) and 1.8 �10�4 (fest). Further, because the variable SOA had no significantinfluence in Experiment 1, only the shorter SOA of 68 ms wasused. Thus, the number of trials was kept the same, although alarger number of primes was used. Moreover, here and in Exper-iment 3, we increased the number of practice trials. At the begin-ning, participants practiced the target discrimination task until theirperformance was at least 75% correct. Finally, in the prime dis-crimination task, participants were asked to categorize the primewords as either spatial or neutral (nonspatial), and before this taskit was carefully explained to the participants which words wereused and which of them were of a spatial meaning and which werespatially neutral. The S–R mapping was balanced across partici-pants. Half of the participants pressed the left key for the spatialwords and the right key for neutral words, and the other half of theparticipants got the reverse response assignment. The experimenttook about 1 hr.

Results

Sound localization. Out of all correct responses, 3.5% wereexcluded because RTs deviated by more than 2 standard deviationsfrom a respective condition’s and individual participant’s meanRT. For mean values, see Appendix C and Figure 3. As can beseen, the participants performed better in the congruent than in theincongruent conditions in both the masked and unmasked condi-tions. This was confirmed by our formal analyses (see below). Theneutral conditions lay between the congruent and incongruentconditions.

As in Experiment 1, an initial ANOVA with the between-participants variable of S–R mapping did not show a significantmain effect of mapping nor any significant interaction with thisvariable. Data were therefore again collapsed across mappings for

a repeated-measures ANOVA of the means of the correct RTs,with the within-participant variables of congruence (congruent,neutral, incongruent) and prime visibility (masked, unmasked).

A significant main effect of prime visibility, F(1, 11) � 5.09,p � .04, Cohen’s f � 0.69, was found, reflecting faster RTs inmasked (626 ms, SD � 62) than unmasked (644 ms, SD � 79)conditions. Importantly, there was also a significant main effect ofcongruence, F(1, 11) � 5.59, p � .01, Cohen’s f � 0.72. Theparticipants performed faster in the prime–target congruent condi-tion (617 ms, SD � 64) than in the neutral (635 ms, SD � 74),t(11) � 2.27, p � .04, and incongruent (653 ms, SD � 79)conditions, t(11) � 2.67, p � .02, but not in the neutral ascompared to the incongruent condition, t(11) � 1.77, p � .10. Noother significant effect or interaction was found, all Fs � 1.0.Additional t tests showed that when the spatial words were pre-sented as primes prior to the targets, participants were not signif-icantly faster in the masked congruent condition (612 ms, SD �57) than in the masked neutral condition (629 ms, SD � 77),t(11) � 2.02, p � .07, but they were significantly faster in themasked congruent than the masked incongruent condition (636 ms,SD � 57), t(11) � 2.61, p � .02. The t tests also showed that inthe unmasked congruent condition (622 ms, SD � 73), participants

1 Discrimination between up versus down primes would possibly havebeen an even more suitable prime visibility test, but as we also usednonspatial primes, these primes would have required a third response in alocation discrimination task (i.e., up primes vs. down primes vs. neutralprime). Instead of increasing the task difficulty of the prime visibility testby asking for three alternative responses, we therefore opted for thetwo-alternative choice procedure for the prime visibility test and used aspatial prime discrimination task as a visibility test in Experiment 3.

Figure 3. Results of Experiment 2: mean reaction times (RTs) in milli-seconds (upper panel) and error rates (ERs) as a percentage (lower panel)on the ordinate plotted as a function of congruence (congruent, neutral, andincongruent) and prime visibility on the abscissa. Standard errors areindicated by bars. Details about the mean RTs and ERs in each of theconditions as well as results of the analysis of the prime’s visibility aresummarized in Appendix C. The dark symbols represent the maskedcondition, and the light symbols represent the unmasked condition.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


Fn1

F3


were not faster than in the unmasked neutral condition (641 ms,SD � 74), t(11) � 1.82, p � .10, but again they were significantlyfaster in the unmasked congruent than the unmasked incongruentcondition (670 ms, SD � 104), t(11) � 2.50, p � .03.

A repeated-measures ANOVA of the ERs with the same vari-ables showed a significant main effect of congruence, F(1, 11) �8.47, p � .01, Cohen’s f � 0.89. ERs were lower in the congruent(15.2%, SD � 10.2) than in the neutral (18.1%, SD � 10.9),t(11) � 2.41, p � .04, and in the incongruent (20.5%, SD � 13.2)condition, t(11) � 3.52, p � .01, as well as in the neutral ascompared to the incongruent condition, t(11) � 2.13, p � .05.There was also a significant two-way interaction, F(1, 11) � 6.26,p � .01, Cohen’s f � 0.75. A follow-up ANOVA for the unmaskedcondition showed a significant main effect of congruence, F(1,11) � 10.44, p � .01, Cohen’s f � 0.98. ERs were significantlylower in the congruent (12.7%, SD � 10.0) than in the neutral(18.0%, SD � 10.7), t(11) � 2.61, p � .02, and in the incongruent(24.0%, SD � 15.6) condition, t(11) � 3.53, p � .01, as well as inthe neutral than the incongruent condition, t(11) � 3.01, p � .01.A similar ANOVA for the masked condition did not show asignificant congruence effect, F � 1.0.

Prime visibility. As in Experiment 1, individual d= was com-puted separately for masked and unmasked primes. For the calcu-lation of d=, spatial words counted as signals and neutral wordscounted as noise. Accordingly, correct judgments of spatial wordsfigured as hits, and incorrect judgments of neutral words figured asFAs.

The participants were not able to discriminate the maskedprimes with better-than-chance accuracy: Mean d= was 0.01 (SD �0.29), t(11) � 0.14, p � .89. In the unmasked condition, theparticipants successfully discriminated the primes: Mean d= was2.82 (SD � 1.03), t(11) � 9.48, p � .001.

Congruence and prime visibility correlations. Once again,we found no significant correlation between RT congruence ef-fects and prime visibility scores for the masked primes, Pearson’sr(10) � �.13, p � .68.

Discussion

Results confirmed the congruence effect already found in Ex-periment 1. This time, we found the congruence effect in themasked and in the unmasked conditions. In fact, the congruenceeffect was numerically stronger in the unmasked than in themasked condition. The latter finding indirectly supports our sup-pression interpretation: Because in half of all trials of Experiment2 visible nonspatial word primes were used and because theseprimes would never interfere with target processing, participantsmight have suppressed all of the priming words less. Also, bycomparison to the neutral condition, we were able to show thatfacilitation in congruent conditions as compared to interference inincongruent conditions contributed to a stronger degree to the netRT congruence effect, but the evidence for this stronger facilitationeffect was restricted to the unmasked priming conditions in whichthe difference between congruent and neutral conditions was sig-nificant. In contrast, in the masked priming conditions, the onlysignificant difference was that between the congruent and theincongruent conditions.

Importantly, we also replicated the chance performance in ourvisibility test of the masked primes, although this test was based on

a categorization of the primes only. Which test was used to assesprime visibility is therefore not crucial for the conclusion that themasked primes were not seen. Using the arguably more demandingjudgment about the prime–target congruence (as in Experiment 1)and the maybe simpler judgment about prime meaning (as in thepresent experiment) as prime visibility tests led to the same con-clusion: The primes could not be seen. Again, the lack of asignificant correlation between the RT congruence effect andprime visibility measures in the masked condition supported theconclusion that this congruence effect was independent of theparticipants’ awareness of the primes.

Following Experiment 1, we speculated that the congruenceeffect might have reflected a facilitation of the representation oftarget sound locations by the congruent meaning of the maskedprime words. Yet, in Experiments 1 and 2, prime word meaningsmight have also fitted to corresponding response triggers that wererepresented as part of our participants’ task sets (Kunde et al.,2003). For example, participants might have represented an S–Rrule, with an up S (a sound from above) as an action trigger for arightward response. The up prime might have fitted to such anaction trigger, thereby activating a response. In congruent condi-tions, such a prime-activated response would have been the sameas the finally required response, but in incongruent conditions, theprime-activated response would have been different from the re-quired target response. Therefore, it is possible that the congruenceeffect reflected prime-activated responses rather than prime-activated sensory representations. Whether action triggeringthrough the masked primes could have accounted for the congru-ence effect was tested in the final experiment.

Experiment 3

In Experiment 3, we changed the target discrimination task.Participants had to distinguish between two sound types by press-ing one button for white noise sounds (from above or from below)and an alternative button for click train sounds (from above orbelow). As a consequence, up and down primes were neutral withrespect to the response triggers. If the congruence effect reflectedresponse triggering by the primes, the congruence effect shouldtherefore be eliminated. However, if the congruence effect re-flected a more or less tight fit between the prime word’s meaningand the sound location, a congruence effect should persist, withfaster responses in congruent conditions (i.e., with up–up anddown–down prime–target pairs) than in incongruent conditions(i.e., with up–down and down–up prime–target pairs).

We also took the opportunity for yet another prime visibilitytest. To obtain a fuller picture of the participants’ residual aware-ness of the primes, in Experiment 3 we conducted yet anothercomplementary prime visibility test. We asked our participants todiscriminate between up primes and down primes. If we find aprime–target congruence effect and if this congruence effect istruly based on subliminal input, participants should perform atchance level in the discrimination of the masked primes’ differentspatial meanings also.

Method

Participants. Sixteen participants were tested. Two partici-pants had to be excluded by the same criterion of chance-level

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 9


performance in the sound localization task as in Experiments 1 and2. The remaining 14 participants (12 female, two male, Mage �27.9 years, age range: 23–37 years) were analyzed.

Apparatus, stimuli, and procedure. These were the same asin Experiment 1, except for the following. First, four target soundswere used: two 300-ms broadband white noise sounds (one fromabove, one from below), as in Experiments 1 and 2, and two300-ms click train sounds (again, one from above and one frombelow). The click train consisted of 200 clicks per second. Thebroadband noise and click train are characterized by their noisyand tonal sound percept, respectively. All sounds were createdindividually for each participant using the same procedure as inExperiment 1. Here, as in Experiment 2, task practice (until 75%correct responses were achieved) and only the short SOA of 68 mswere used. The participants were asked to categorize the targets astonal or noisy. In the prime discrimination task, the participantswere asked to categorize the primes as up or down words. The S–Rmapping was balanced across the participants.

Results

Sound categorization. Out of all correct responses, 3.9%were excluded because RTs deviated by more than 2 standarddeviations from a respective condition’s and individual partici-pant’s mean RT. As can be seen in Appendix D and Figure 4, theoverall ERs were quite low, showing that the listeners had nodifficulty discriminating the two sound types. With respect to theRT, participants once again responded faster in the congruent

conditions than in the incongruent conditions. This was againconfirmed by formal analyses.

As in Experiments 1 and 2, an initial ANOVA with the between-participants variable of S–R mapping did not show a significantmain effect of mapping nor any significant interaction with thisvariable. Data were therefore collapsed across mappings for arepeated-measures ANOVA of the means of the correct RTs, withthe within-participant variables of congruence (congruent, incon-gruent), target type (noise sound, click train sound), and primevisibility (masked, unmasked).

We found a significant main effect of congruence, F(1, 13) �6.60, p � .02, Cohen’s f � 0.72, reflecting faster RTs in congruent(604 ms, SD � 61) than incongruent (615 ms, SD � 71) condi-tions. No other significant main effect or interaction was found, allFs � 1.0. An additional t test confirmed that participants weresignificantly faster in the masked congruent condition (M � 601ms, SD � 65) than in the masked incongruent condition (M � 614ms, SD � 76), t(13) � 2.35, p � .03. A similar t test of theunmasked congruent condition (M � 607 ms, SD � 68) andincongruent condition (M � 615 ms, SD � 67) was not significant,t(13) � 1.07, p � .30.

The same ANOVA with ERs showed only a significant two-wayinteraction of congruence and target type, F(1, 13) � 8.27, p �.01, Cohen’s f � 0.80. Subsequent t tests showed that neither thenoise sounds’ congruent condition (2.7%, SD � 2.1) was signifi-cantly different from its incongruent (1.6%, SD � 1.3) condition,t � 1.55, p � .15, nor that the click train sounds’ congruentcondition (1.8%, SD � 1.5) was significantly different from itsincongruent (3.2%, SD � 2.8) condition, t � 1.71, p � .11. Noother significant main effect or interaction was found, all Fs � 1.0.

Prime visibility. The participants were not able to discrimi-nate the masked primes with better-than-chance accuracy: Mean d=was �0.01 (SD � 0.20), t(13) � 0.11, p � .92. However, theysuccessfully discriminated the unmasked primes: Mean d= was2.45 (SD � 1.04), t(13) � 8.79, p � .001.

Congruence and prime visibility correlations. As in Exper-iments 1 and 2, we found no significant correlation between the RTcongruence effect and prime visibility for the masked primes,Pearson’s r(12) � .29, p � .31.

Discussion

The results confirmed a significant RT congruence effect of themasked primes, although response triggering was ruled out and theprimes were presented subliminally. When taken together, theseresults are in line with an influence of subliminal prime meaningon the sensory representation of the sound locations. As in Exper-iment 1, active suppression probably prevented the congruenceeffect in the unmasked condition (Kinoshita et al., 2011).

Analysis Across the Three Experiments

We also conducted a final analysis covering data from all threeexperiments. This was done to get a better estimate of the trueeffect size of the congruence effect. For this analysis, we consid-ered the conditions that were the same across Experiments 1, 2,and 3: the spatial primes and noise targets with the short SOA.

Figure 4. Results of Experiment 3: mean reaction times (RTs) in milli-seconds (upper panel) and error rates (ERs) as a percentage (lower panel)on the ordinate plotted as a function of congruence (congruent vs. incon-gruent), target type, and prime visibility on the abscissa. Standard errors areindicated by bars. Details about the mean RTs and ERs in each of theconditions as well as results of the analysis of the prime’s visibility aresummarized in Appendix D. Note. Congr. � congruent; Incongr. � incon-gruent. The dark symbols represent noise, and the light symbols representclick train sounds. The circles represent the masked condition, and thetriangles represent the unmasked condition.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


F4


Results

A repeated-measures ANOVA, with the within-participantsvariables of prime–target congruence (congruent, incongruent) andprime visibility (masked, unmasked) and the between-participantsvariable of experiment (Experiment 1, Experiment 2, Experiment3), was conducted on RTs. It showed a main effect of prime–targetcongruence, F(1, 40) � 12.80, p � .01, Cohen’s f � 0.58. MeanRTs were shorter in congruent (M � 613 ms, SD � 59) than inincongruent conditions (M � 632 ms, SD � 67). We also found asignificant main effect of prime visibility, F(1, 40) � 6.48, p �.02, Cohen’s f � 0.41. Mean RTs were shorter in masked (M �614 ms, SD � 60) as compared to unmasked conditions (M � 631ms, SD � 68). No other significant effect or interaction was found,all Fs � 2.7.

Discussion

In the current analysis across experiments, with its larger n, weachieved a more trustworthy and lower measure of the true effectsize of the congruence effect than in the individual experiments.The analysis also showed that masked conditions were slightlyeasier than unmasked conditions (see also Experiment 2). Maybethe visible primes elicited some form of additional suppression.This would explain why RTs were longer in unmasked thanmasked conditions and why unmasked primes tended to elicit aquantitatively weaker congruence effect than masked primes, al-though this was not reflected in a significant interaction betweenprime visibility and congruence and although this was not true ofExperiment 2. We will get back to this question at the end of theGeneral Discussion.

General Discussion

Based on theories connecting semantic word representationswith corresponding sensory representations within one joint net-work (Barsalou, 1999; Chen & Spence, 2011; Logan, 1995), wehypothesized that subliminal prime words in the visual modalityshould impact sensory discrimination between auditory targets. Inthe current experiments, this prediction was borne out with maskedvisual word primes and auditory targets. Our participants had tojudge the vertical direction of an auditory target relative to theirown position. In congruent conditions, the masked word prime’sspatial meaning fitted to the location of the auditory target. Forexample, the word up was presented before an auditory target fromabove. In the incongruent conditions, the masked word prime’sspatial meaning did not fit to the location of the auditory target. Forexample, the word up was presented before an auditory target frombelow. In line with the predictions, we found a congruence effect:faster responses in congruent than in incongruent conditions (inExperiments 1 to 3). The results support the existence of a cross-modal link between masked visual primes and auditory targets.

Importantly, all these cross-modal congruence effects werefound where the participants could not recognize the maskedprimes. This was shown in our prime visibility (or prime aware-ness) tests. Certainly, each of these visibility tests had its draw-backs as well as its advantages. For example, our visibility test inExperiment 1 asked for prime-target congruence discrimination.This is a straightforward question when one wants to demonstrate

a dissociation between congruence effects and a lacking awarenessfor the underlying congruence relation, but the task is more de-manding than asking for the prime meaning alone (as we did inExperiments 2 and 3). Therefore, the prime visibility test ofExperiment 1 might not have been optimally sensitive for theparticipants’ residual awareness of the primes, and easier taskswere used in Experiments 2 and 3. In Experiment 2, we asked theparticipants to only discriminate if the prime did or did not have aspatial meaning, and in Experiment 3, we directly asked theparticipants to only decide if the prime was an up or a down word.Both of these prime visibility tasks were probably easier than thecongruence discrimination in Experiment 3, but arguably have thedrawback of not asking an entirely correct question to demonstratethe dissociation. However, as we varied the visibility tests betweenexperiments and found that the specifics of the visibility tests wereimmaterial for our conclusion that the masked primes were notseen, we are relatively sure that the cross-modal congruence ef-fects were independent of the participants’ awareness of theprimes. This conclusion was also supported by lacking correlationsbetween masked primes’ visibility scores and RT congruenceeffects.

This leads us to the question as to how this awareness-independent influence of the masked primes was brought about. Apossible answer can be given on the basis of Logan’s (1995)theory. First of all, our participants had to give a deictic judgment:They had to decide whether an auditory target was presented fromabove or from below. This required that the participants locate thetargets relative to one another or to their own (i.e., the partici-pant’s) position. According to Logan’s theory, this can only bedone by the application of a conceptual spatial reference frame tothe stimuli by which the spatial relation between the targets orbetween the participant’s own body (or here: his or her ears) andeach target is determined. In this situation, the prime words couldhave worked in one of the following ways. First, each prime wordcould have facilitated the usage of one particular spatial frame, anup frame or a down frame, corresponding to the prime’s meaning.If that was the case, a congruent prime would have preactivated aconceptually fitting frame for the target, whereas an incongruentprime would have preactivated a conceptually nonfitting frame forthe target, with the resulting necessity to adjust the spatial frame tothe target in a time-consuming manner only in the incongruentconditions. Second, it is also possible that the participants had setup a joint conceptual spatial frame for all deictic judgments as atop-down template and in anticipation of the auditory targets. As aconsequence of such a top-down template, participants could thenhave inadvertently processed the prime word’s meaning in accor-dance with the template, just as this is known to be the case forother forms of top-down contingent processing, such as motoractivations or attention shifts (Folk, Remington, & Johnston, 1992;Kunde et al., 2003). Note that the spatial frame is a semanticconcept, so it is easy to understand why a prime word’s meaningcould have specified one particular spatial relation within theframe. Also, because the same concept of a spatial frame had to beused for the judgment of the deictic relation between targetsbetween target and participant, an incongruent prime specifying aspatial relation different from that to the target could have delayedthe correct judgment about target position. In general agreementwith this possibility, application of a top-down template to processunconscious stimuli is possible once the top-down template has

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 10

AQ: 11


been set up in advance of conscious targets (for a review, seeAnsorge et al., 2014).

From a more general view, the awareness-independent cross-modal priming effect that we found is in line with views ofword understanding that assume a tight link between wordmeaning and sensory processing, such as the embodied cogni-tion view (e.g., Barsalou, 1999) or theories that reckoned thatreaders construct situation models out of the words they read(e.g., Kintsch & van Dijk, 1978; Zwaan & Radvansky, 1998).For instance, according to the embodied cognition view, whenunderstanding an abstract word, a reader has to simulate (partof) the fitting past sensory and motor experiences that layaground of understanding word meaning in the first place(Barsalou, 1999; Glenberg & Kaschak, 2002). Because embod-ied processing is assumed to be an obligatory consequence ofreading a word and would link abstract word meaning andperceptual processing, it would be clear why even subliminalvisual words could prime sensory processing in a differentmodality: Awareness independence is one hallmark of the au-tomatic or obligatory processing (e.g., Posner & Snyder, 1975).

In contrast to these potential explanations in terms of semanticor conceptual priming, several alternative explanations of thecross-modal congruence effect were ruled out. According to onepotential explanation, the masked primes might have activatedresponses via an orthogonal Simon effect (Proctor & Cho, 2006).If an orthogonal Simon effect would have created the congruenceeffect, we would have expected a masked congruence effect in theS–R compatible conditions (in which targets from above requireda right-hand response and targets from below required a left-handresponse) but not in the S–R incompatible conditions (in whichtargets from above required a left-hand response and targets frombelow required a right-hand response). However, the between-participants variable of S–R mapping did not significantly interactwith the congruence effect in any of the experiments. Also, forExperiments 1 and 2, it could be argued that the masked primecould have directly triggered a response based on a fit of theprimes’ meanings to the action triggers (Kunde et al., 2003). Inthese experiments, the participants had to give one response totargets from above and an alternative response to targets frombelow. However, this possibility was ruled out in Experiment 3 inwhich primes did not fit to the action triggers—that is, the primes’meanings were not related to the responses.

In addition to an account of cross-modal connections betweenrepresentations of word meaning and sensory stimulus charac-teristics in a single joint network, it may also be possible thatthe primes exerted their influence via an attention shift (Gibson& Kingstone, 2006; Logan, 1994). According to this explana-tion, the masked word would have triggered an attention shiftwith a direction corresponding to its spatial meaning, so that incongruent conditions attention would have been directed towardthe target’s spatial position and in incongruent conditions itwould have been directed away from the target’s position. Inline with this possibility, previous research demonstrated (a)intermodal attentional effects of visual cues during the discrim-ination of auditory targets (Driver & Spence, 1998; Eimer &Schröger, 1998) as well as (b) attentional effects on the basis ofsubliminal visual cues (Ansorge, Kiss, & Eimer, 2009; McCor-mick, 1997). Although it is thus theoretically possible thatmasked primes led to an attention shift, this remains to be

proven in future studies because, to our knowledge, so far therehas not been a single study demonstrating attention shifts basedon masked visual words.

Before we conclude, we need to quickly discuss the size ofthe congruence effect. The cross-modal congruence effects (in-congruent RT � congruent RT) of 12 and 11 ms in the currentExperiments 1 and 3 (see Figures 2 and 4) were quantitativelysmall as compared to the cross-modal congruence effect of 36ms in Experiment 2 and to the intramodal congruence effect of25 ms that was found when using the same spatial words asprimes and targets in intramodal priming experiments (Ansorgeet al., 2013). Two interconnected reasons for the small congru-ence effect might have been (a) the nonpredictive nature of theprimes and (b) the intermixed presentation of the masked andunmasked primes. Concerning the first reason, because congru-ent and incongruent primes were equally likely and did thus notpredict target identity, participants would have been well ad-vised to ignore the primes. Also, it might be that the participantsdid not register the prime–target contingencies in the maskedconditions (but see, e.g., Reuss, Desender, Kiesel, & Kunde,2014), but at least in the unmasked conditions, participantsshould have noticed the nonpredictive nature of the primes. Asa consequence, participants might have set up a mind-set toactively suppress the unmasked primes as much as possiblegiven the constraint that they had to gaze at the experimentalscreen presenting the prime words. In line with this possibility,other studies sometimes observed zero congruence effects withunmasked nonpredictive primes (Kinoshita et al., 2011). Con-cerning the second reason, in the current study, the participantscould have applied their suppressive mind-set for the unmaskedprimes inadvertently to the masked primes, too, beca-use the masked primes were presented randomly intermixedwith the unmasked primes. In other words, if the participantsprepared to suppress the visible primes, masked primes couldhave also triggered their own suppression by their fit to asuppressive mind-set, at least in some of the trials, and in a waysimilar to their ability to trigger other forms of inhibitoryprocesses (Experiment 5 of Ansorge, 2004). Therefore, in Ex-periments 1 and 3, suppression would have counteracted thecongruence effect of the clearly seen unmasked primes, but thecongruence effect of the masked prime could also have sufferedfrom this suppression to some extent. This would explain boththe fact that the congruence effect tended to be small and thefact that the congruence effect tended to be even smaller withunmasked than masked primes.

However, a notable exception from this pattern was Experiment2. In Experiment 2, a quantitatively stronger overall RT congru-ence effect was found, and this was mostly due to the unmaskedprimes rather than the masked primes (see Appendix C). We thinkthat this finding neatly dovetails with our explanation in terms ofactive prime suppression counteracting the congruence effect. InExperiment 2, half of the primes were spatially neutral and thuscompletely unrelated to target classification. As a consequence, onaverage, less suppression of the primes was necessary in Experi-ment 2. Yet, clearly, such a context effect would have had a betterchance to express itself in the conditions with unmasked primes,too, so that the RT congruence effect of the unmasked primes wasaffected to a larger degree (in this case, it was boosted due to a

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 12


release from inhibition) than the congruence effect of the maskedprimes.

The size of the masked congruence effect of 24 ms in the presentExperiment 2 (see Appendix C) is also interesting for yet anotherreason: This cross-modal congruence effect was almost as strongas the congruence effect of masked intermodal visual word–wordpriming of 27 ms in Experiment 2 of Ansorge et al. (2013). Thissimilarity of the congruence effects suggests that almost all of themasked words’ congruence effects could be due to modality-unspecific (or amodal) semantic priming: The fact that (a) only inthe study of Ansorge et al. (2013) but not in the present study couldthe visual words have primed the target modality, the letters, or thewords of the target displays and that (b) these potential sources fordifferences did not affect the size of the congruence effect suggeststhat a principle that was shared between both studies, such assemantic priming, could have accounted for the congruence effectof the masked primes. This conclusion, however, is tentative. Theresults could be different where visual and auditory targets have tobe discriminated from one another or where a different temporalinterval between the prime and target is used. Therefore, answer-ing the question if the priming effect in the current study wasentirely amodal has to await a whole series of future experimentsand is beyond the scope of the present study.

Conclusion

We have shown that the meaning of an unconscious visual wordcan cross-modally prime the perceived location of an auditorytarget. This unconscious cross-modal priming effect appears toreflect semantic processing.

References

Ansorge, U. (2004). Top-down contingencies of nonconscious primingrevealed by dual-task interference. The Quarterly Journal of Experimen-tal Psychology A: Human Experimental Psychology, 57, 1123–1148.http://dx.doi.org/10.1080/02724980343000792

Ansorge, U., & Bohner, G. (2014). Investigating the association betweenvalence and elevation with an implicit association task that requiresupward and downward responses. Universitas Psychologica, 12, 1453–1471. http://dx.doi.org/10.11144/Javeriana.UPSY12-5.iave

Ansorge, U., Khalid, S., & König, P. (2013). Space-valence priming withsubliminal and supraliminal words. Frontiers in Psychology, 4, 81.http://dx.doi.org/10.3389/fpsyg.2013.00081

Ansorge, U., Kiss, M., & Eimer, M. (2009). Goal-driven attentional captureby invisible colors: Evidence from event-related potentials. Psycho-nomic Bulletin & Review, 16, 648–653. http://dx.doi.org/10.3758/PBR.16.4.648

Ansorge, U., Kunde, W., & Kiefer, M. (2014). Unconscious vision andexecutive control: How unconscious processing and conscious actioncontrol interact. Consciousness and Cognition, 27, 268–287. http://dx.doi.org/10.1016/j.concog.2014.05.009

Atkinson, A. P., Thomas, M. S., & Cleeremans, A. (2000). Consciousness:Mapping the theoretical landscape. Trends in Cognitive Sciences, 4,372–382. http://dx.doi.org/10.1016/S1364-6613(00)01533-3

Baars, B. (1988). A cognitive theory of consciousness. New York, NY:Cambridge University Press.

Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and BrainSciences, 22, 577–609.

Breitmeyer, B., & Ogmen, H. (2006). Visual masking: Time slices throughconscious and unconscious vision (Vol. 41). New York, NY: Oxford

University Press. http://dx.doi.org/10.1093/acprof:oso/9780198530671.001.0001

Chen, Y. C., & Spence, C. (2011). Crossmodal semantic priming bynaturalistic sounds and spoken words enhances visual sensitivity. Jour-nal of Experimental Psychology: Human Perception and Performance,37, 1554–1568. http://dx.doi.org/10.1037/a0024329

Damian, M. F. (2001). Congruity effects evoked by subliminally presentedprimes: Automaticity rather than semantic processing. Journal of Ex-perimental Psychology: Human Perception and Performance, 27, 154–165. http://dx.doi.org/10.1037/0096-1523.27.1.154

Dehaene, S., & Changeux, J. P. (2011). Experimental and theoreticalapproaches to conscious processing. Neuron, 70, 200–227. http://dx.doi.org/10.1016/j.neuron.2011.03.018

Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience ofconsciousness: Basic evidence and a workspace framework. Cognition,79, 1–37. http://dx.doi.org/10.1016/S0010-0277(00)00123-2

Dehaene, S., Naccache, L., Le Clec’H, G., Koechlin, E., Mueller, M.,Dehaene-Lambertz, G., . . . Le Bihan, D. (1998). Imaging unconscioussemantic priming. Nature, 395, 597–600. http://dx.doi.org/10.1038/26967

Desender, K., Van Opstal, F., & Van den Bussche, E. (2014). Feeling theconflict: The crucial role of conflict experience in adaptation. Psycho-logical Science, 25, 675– 683. http://dx.doi.org/10.1177/0956797613511468

Dijksterhuis, A., & Aarts, H. (2010). Goals, attention, and (un)conscious-ness. Annual Review of Psychology, 61, 467–490. http://dx.doi.org/10.1146/annurev.psych.093008.100445

Driver, J., & Spence, C. (1998). Cross-modal links in spatial attention.Philosophical Transactions of the Royal Society B: Biological Sciences,353, 1319–1331. http://dx.doi.org/10.1098/rstb.1998.0286

Eimer, M., & Schröger, E. (1998). ERP effects of intermodal attention andcross-modal links in spatial attention. Psychophysiology, 35, 313–327.http://dx.doi.org/10.1017/S004857729897086X

Faivre, N., Mudrik, L., Schwartz, N., & Koch, C. (2014). Multisensoryintegration in complete unawareness: Evidence from audiovisual con-gruency priming. Psychological Science, 25, 2006–2016. http://dx.doi.org/10.1177/0956797614547916

Folk, C. L., Remington, R. W., & Johnston, J. C. (1992). Involuntarycovert orienting is contingent on attentional control settings. Journal ofExperimental Psychology: Human Perception and Performance, 18,1030–1044. http://dx.doi.org/10.1037/0096-1523.18.4.1030

Gibson, B. S., & Kingstone, A. (2006). Visual attention and the semanticsof space: Beyond central and peripheral cues. Psychological Science, 17,622–627. http://dx.doi.org/10.1111/j.1467-9280.2006.01754.x

Glaser, W. R., & Glaser, M. O. (1989). Context effects in Stroop-like wordand picture processing. Journal of Experimental Psychology: General,118, 13–42. http://dx.doi.org/10.1037/0096-3445.118.1.13

Glenberg, A. M., & Kaschak, M. P. (2002). Grounding language in action.Psychonomic Bulletin & Review, 9, 558–565. http://dx.doi.org/10.3758/BF03196313

Grainger, J., Diependaele, K., Spinelli, E., Ferrand, L., & Farioli, F. (2003).Masked repetition and phonological priming within and across modali-ties. Journal of Experimental Psychology: Learning, Memory, and Cog-nition, 29, 1256–1269. http://dx.doi.org/10.1037/0278-7393.29.6.1256

Kiefer, M. (2002). The N400 is modulated by unconsciously perceivedmasked words: Further evidence for an automatic spreading activationaccount of N400 priming effects. Cognitive Brain Research, 13, 27–39.http://dx.doi.org/10.1016/S0926-6410(01)00085-4

Kinoshita, S., Mozer, M. C., & Forster, K. I. (2011). Dynamic adaptationto history of trial difficulty explains the effect of congruency proportionon masked priming. Journal of Experimental Psychology: General, 140,622–636. http://dx.doi.org/10.1037/a0024230

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 13

AQ: 14


http://dx.doi.org/10.1080/02724980343000792http://dx.doi.org/10.11144/Javeriana.UPSY12-5.iavehttp://dx.doi.org/10.3389/fpsyg.2013.00081http://dx.doi.org/10.3758/PBR.16.4.648http://dx.doi.org/10.3758/PBR.16.4.648http://dx.doi.org/10.1016/j.concog.2014.05.009http://dx.doi.org/10.1016/j.concog.2014.05.009http://dx.doi.org/10.1016/S1364-6613%2800%2901533-3http://dx.doi.org/10.1093/acprof:oso/9780198530671.001.0001http://dx.doi.org/10.1093/acprof:oso/9780198530671.001.0001http://dx.doi.org/10.1037/a0024329http://dx.doi.org/10.1037/0096-1523.27.1.154http://dx.doi.org/10.1016/j.neuron.2011.03.018http://dx.doi.org/10.1016/j.neuron.2011.03.018http://dx.doi.org/10.1016/S0010-0277%2800%2900123-2http://dx.doi.org/10.1038/26967http://dx.doi.org/10.1038/26967http://dx.doi.org/10.1177/0956797613511468http://dx.doi.org/10.1177/0956797613511468http://dx.doi.org/10.1146/annurev.psych.093008.100445http://dx.doi.org/10.1146/annurev.psych.093008.100445http://dx.doi.org/10.1098/rstb.1998.0286http://dx.doi.org/10.1017/S004857729897086Xhttp://dx.doi.org/10.1177/0956797614547916http://dx.doi.org/10.1177/0956797614547916http://dx.doi.org/10.1037/0096-1523.18.4.1030http://dx.doi.org/10.1111/j.1467-9280.2006.01754.xhttp://dx.doi.org/10.1037/0096-3445.118.1.13http://dx.doi.org/10.3758/BF03196313http://dx.doi.org/10.3758/BF03196313http://dx.doi.org/10.1037/0278-7393.29.6.1256http://dx.doi.org/10.1016/S0926-6410%2801%2900085-4http://dx.doi.org/10.1037/a0024230

Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text compre-hension and production. Psychological Review, 85, 363–394. http://dx.doi.org/10.1037/0033-295X.85.5.363

Kiyonaga, K., Grainger, J., Midgley, K., & Holcomb, P. J. (2007). Maskedcross-modal repetition priming: An event-related potential investigation.Language and Cognitive Processes, 22, 337–376. http://dx.doi.org/10.1080/01690960600652471

Kouider, S., & Dehaene, S. (2009). Subliminal number priming within andacross the visual and auditory modalities. Experimental Psychology, 56,418–433. http://dx.doi.org/10.1027/1618-3169.56.6.418

Kouider, S., & Dupoux, E. (2001). A functional disconnection betweenspoken and visual word recognition: Evidence from unconscious prim-ing. Cognition, 82, B35–B49. http://dx.doi.org/10.1016/S0010-0277(01)00152-4

Kunde, W., Kiesel, A., & Hoffmann, J. (2003). Conscious control over thecontent of unconscious cognition. Cognition, 88, 223–242. http://dx.doi.org/10.1016/S0010-0277(03)00023-4

Lamy, D., Mudrik, L., & Deouell, L. Y. (2008). Unconscious auditoryinformation can prime visual word processing: A process-dissociationprocedure study. Consciousness and Cognition, 17, 688–698. http://dx.doi.org/10.1016/j.concog.2007.11.001

Logan, G. D. (1994). Spatial attention and the apprehension of spatial relations.Journal of Experimental Psychology: Human Perception and Performance, 20,1015–1036. http://dx.doi.org/10.1037/0096-1523.20.5.1015

Logan, G. D. (1995). Linguistic and conceptual control of visual spatialattention. Cognitive Psychology, 28, 103–174. http://dx.doi.org/10.1006/cogp.1995.1004

Majdak, P., Walder, T., & Laback, B. (2013). Effect of long-term trainingon sound localization performance with spectrally warped and band-limited head-related transfer functions. The Journal of the AcousticalSociety of America, 134, 2148 –2159. http://dx.doi.org/10.1121/1.4816543

McCauley, C., Parmelee, C. M., Sperber, R. D., & Carr, T. H. (1980). Earlyextraction of meaning from pictures and its relation to conscious iden-tification. Journal of Experimental Psychology: Human Perception andPerformance, 6, 265–276. http://dx.doi.org/10.1037/0096-1523.6.2.265

McCormick, P. A. (1997). Orienting attention without awareness. Jour-nal of Experimental Psychology: Human Perception and Perfor-mance, 23, 168 –180. http://dx.doi.org/10.1037/0096-1523.23.1.168

Naccache, L., & Dehaene, S. (2001). Unconscious semantic priming ex-tends to novel unseen stimuli. Cognition, 80, 215–229. http://dx.doi.org/10.1016/S0010-0277(00)00139-6

Nakamura, K., Hara, N., Kouider, S., Takayama, Y., Hanajima, R., Sakai,K., & Ugawa, Y. (2006). Task-guided selection of the dual neuralpathways for reading. Neuron, 52, 557–564. http://dx.doi.org/10.1016/j.neuron.2006.09.030

Norris, D., & Kinoshita, S. (2008). Perception as evidence accumulationand Bayesian inference: Insights from masked priming. Journal ofExperimental Psychology: General, 137, 434–455. http://dx.doi.org/10.1037/a0012799

Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive control.In R. L. Solso (Ed.), Information processing and cognition (pp. 55–85).Hillsdale, NJ: Erlbaum.

Proctor, R. W., & Cho, Y. S. (2006). Polarity correspondence: A generalprinciple for performance of speeded binary classification tasks. Psy-chological Bulletin, 132, 416–442. http://dx.doi.org/10.1037/0033-2909.132.3.416

Regier, T., & Carlson, L. A. (2001). Grounding spatial language in per-ception: An empirical and computational investigation. Journal of Ex-perimental Psychology: General, 130, 273–298. http://dx.doi.org/10.1037/0096-3445.130.2.273

Reingold, E. M., & Merikle, P. M. (1988). Using direct and indirectmeasures to study perception without awareness. Perception &Psychophysics, 44, 563–575. http://dx.doi.org/10.3758/BF03207490

Reuss, H., Desender, K., Kiesel, A., & Kunde, W. (2014). Uncon-scious conflicts in unconscious contexts: The role of awareness andtiming in flexible conflict adaptation. Journal of ExperimentalPsychology: General, 143, 1701–1718. http://dx.doi.org/10.1037/a0036437

Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in languagecomprehension and memory. Psychological Bulletin, 123, 162–185.http://dx.doi.org/10.1037/0033-2909.123.2.162

Appendix A

Method for Obtaining Directional Transfer Functions

We describe only those aspects of the method that are relevantfor the current application. Head-related transfer functions(HRTFs) were measured individually for each listener in a sessionprior to the experiment and on a different day. Twenty-two loud-speakers were mounted on a vertical circular arc with a spacing of5°. The listener was seated in the center point of the arc, with adistance between the center point and each speaker of 1.2 m.Microphones were inserted into the listener’s ear canals, and theiroutput signals were directly recorded. The test signal was a1,729-ms exponential frequency sweep from 0.05 to 20 kHz. Theacoustic influence of the equipment was removed by equalizing

the HRTFs with the transfer functions of the equipment. Forderiving the directional transfer functions (DTFs), first the mag-nitude of the common transfer function (CTF) was calculatedby averaging the log-amplitude spectra of all HRTFs for eachindividual listener. The phase spectrum of the CTF was set tothe minimum phase corresponding to the amplitude spectrum.The DTFs were finally obtained by filtering the HRTFs with theinverse complex CTF. Finally, the impulse responses of the DTFswere windowed with an asymmetric Tukey window (fade-in of 0.5ms and fade-out of 1 ms) to a 5.33-ms duration. More details onthis method can be found in Majdak, Walder, and Laback (2013).

(Appendices continue)

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.



http://dx.doi.org/10.1037/0033-295X.85.5.363http://dx.doi.org/10.1037/0033-295X.85.5.363http://dx.doi.org/10.1080/01690960600652471http://dx.doi.org/10.1080/01690960600652471http://dx.doi.org/10.1027/1618-3169.56.6.418http://dx.doi.org/10.1016/S0010-0277%2801%2900152-4http://dx.doi.org/10.1016/S0010-0277%2801%2900152-4http://dx.doi.org/10.1016/S0010-0277%2803%2900023-4http://dx.doi.org/10.1016/S0010-0277%2803%2900023-4http://dx.doi.org/10.1016/j.concog.2007.11.001http://dx.doi.org/10.1016/j.concog.2007.11.001http://dx.doi.org/10.1037/0096-1523.20.5.1015http://dx.doi.org/10.1006/cogp.1995.1004http://dx.doi.org/10.1006/cogp.1995.1004http://dx.doi.org/10.1121/1.4816543http://dx.doi.org/10.1121/1.4816543http://dx.doi.org/10.1037/0096-1523.6.2.265http://dx.doi.org/10.1037/0096-1523.23.1.168http://dx.doi.org/10.1016/S0010-0277%2800%2900139-6http://dx.doi.org/10.1016/S0010-0277%2800%2900139-6http://dx.doi.org/10.1016/j.neuron.2006.09.030http://dx.doi.org/10.1016/j.neuron.2006.09.030http://dx.doi.org/10.1037/a0012799http://dx.doi.org/10.1037/a0012799http://dx.doi.org/10.1037/0033-2909.132.3.416http://dx.doi.org/10.1037/0033-2909.132.3.416http://dx.doi.org/10.1037/0096-3445.130.2.273http://dx.doi.org/10.1037/0096-3445.130.2.273http://dx.doi.org/10.3758/BF03207490http://dx.doi.org/10.3758/BF03207490http://dx.doi.org/10.1037/a0036437http://dx.doi.org/10.1037/a0036437http://dx.doi.org/10.1037/0033-2909.123.2.162

Appendix B

Mean Reaction Times, Error Rates, and Prime Visibility Indices of Experiment 1

Conditions Reaction times (ms) Error rates (%) Prime visibility

Visibility SOA Congruence Mean RTsNet

congruence Mean ERsNet

congruence

Mean d’(congruent andincongruent)

t testagainst zero

Significance(two-tailed)

Masked Short Congruent 600 19 9.0 3.3 .08 .77 .45Incongruent 619 12.3

Long Congruent 600 14 13.0 .3 .16 1.97 .07Incongruent 614 13.3

Unmasked Short Congruent 623 3 14.3 1.4 1.31 2.51 .03Incongruent 626 15.7

Long Congruent 604 15 12.7 6.0 1.33 2.21 .04Incongruent 619 18.7

Note. SOA � stimulus onset asynchrony; RT � reaction time; ER � error rate. Mean RTs and error rates are compared in the congruent and incongruentconditions, and the net congruence effect is calculated as the mean performance in the incongruent condition minus the mean performance in the congruentcondition.

Appendix C



Visibility Congruence Mean RTsNet


congruence


t testagainst zero


Masked Congruent 612 24 17.7 .0 .01 .14 .89Neutral 629 18.2Incongruent 636 17.7

Unmasked Congruent 622 48 12.7 11.3 2.82 9.48 .00Neutral 641 18.0Incongruent 670 24.0

Note. RT � reaction time; ER � error rate. Mean RTs and error rates are compared in the congruent and incongruent conditions, and the net congruenceeffect is calculated as the mean performance in the incongruent condition minus the mean performance in the congruent condition.

(Appendices continue)

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 15

AQ: 16


Appendix D



Visibility Sound Congruence Mean RTsNet


congruence


t testagainst zero


Masked Noise Congruent 604 12 2.7 �.9 �.01 .11 .92Incongruent 616 1.8

Click Congruent 59,814 2.3 1.3Incongruent 612 3.6

Unmasked Noise Congruent 619 6 2.7 �1.3 2.45 9.79 .00Incongruent 625 1.4

Click Congruent 595 10 1.3 1.6Incongruent 605 2.9

Note. RT � reaction time; ER � error rate. Mean RTs and error rates are compared in the congruent and incongruent conditions, and the net congruenceeffect is calculated as the mean performance in the incongruent condition minus the mean performance in the congruent condition.

Received January 13, 2015Revision received September 17, 2015

Accepted October 8, 2015 �

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.


AQ: 17


Unconscious Cross-Modal Priming of Auditory Sound Localization by Visual WordsExperiment 1MethodParticipantsApparatus, stimuli, and procedure

ResultsSound localizationPrime visibilityCongruence and prime visibility correlations

Discussion

Experiment 2MethodParticipantsApparatus, stimuli, and procedure

ResultsSound localizationPrime visibilityCongruence and prime visibility correlations

Discussion

Experiment 3MethodParticipantsApparatus, stimuli, and procedure

ResultsSound categorizationPrime visibilityCongruence and prime visibility correlations

Discussion

Analysis Across the Three ExperimentsResultsDiscussion

General DiscussionConclusion

ReferencesAppendix AMethod for Obtaining Directional Transfer FunctionsAppendix BMean Reaction Times, Error Rates, and Prime Visibility Indices of Experiment 1Appendix CMean Reaction Times, Error Rates, and Prime Visibility Indices of Experiment 2Appendix DMean Reaction Times, Error Rates, and Prime Visibility Indices of Experiment 3

unconscious cross-modal priming of auditory sound ......comb, 2007; kouider & dehaene, 2009; kouider...

Documents