Brain and Cognition xxx (2011) xxx–xxx

Contents lists available at ScienceDirect

Brain and Cognition

journal homepage: www.elsevier .com/ locate /b&c

9 is Always on top: Assessing the automaticity of synaesthetic number-forms

Michelle Jarick ⇑, Michael J. Dixon, Daniel SmilekDepartment of Psychology, University of Waterloo, Ontario, Canada

a r t i c l e i n f o a b s t r a c t

Article history:Accepted 11 May 2011Available online xxxx

Keywords:SynaesthesiaNumber-formsAttentionSpatial-cueingMental number line

0278-2626/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.bandc.2011.05.003

⇑ Corresponding author. Address: Department oWaterloo, 200 University Ave. W., Waterloo, Ontario,

E-mail address: majarick@uwaterloo.ca (M. Jarick)

Please cite this article in press as: Jarick, M., et adoi:10.1016/j.bandc.2011.05.003

For number-form synaesthetes, digits occupy idiosyncratic spatial locations. Atypical to the mental num-ber line that extends horizontally, the synaesthete (L) experiences the numbers 1–10 vertically. We useda spatial cueing task to demonstrate that L’s attention could be automatically directed to locations withinher number-space – being faster to detect targets appearing in synaesthetically cued locations. Wesought to eliminate any influence of strategy on L’s performance by: (a) shortening the cue-target onsetto 150 ms, (b) making the cues counterpredictive, and (c) instructing L to use an opposing strategy. If L’sperformance was attributable to intentionally using the cue to predict target location, these manipula-tions should eliminate any cuing effects consistent with her synaesthesia. However, L showed an atten-tional bias compatible with her number-form, except when explicitly instructed of the opposing strategyand given enough time (800 ms). Therefore, we attribute L’s resilient cueing effects to the automaticity ofher number-form.

� 2011 Elsevier Inc. All rights reserved.

1. Introduction

Recent studies on synaesthesia suggest that for an increasingnumber of individuals, numbers can elicit a very lucid and con-scious experience of specific locations in space. These number-space mappings are referred to as ‘‘number-forms’’. Some synaes-thetes have described their number-forms as infinitely spiralinginto the distance, while others describe idiosyncratic number spacemappings that turn corners and form shapes that, to the non-synaesthete, defy logical structure. These experiences that werefirst described by Galton in 1880, are now recognized as a formof synaesthesia (termed ‘‘number-form’’ synaesthesia) and haverecently garnered a great deal of scientific attention (see the spe-cial section in Cortex on visuo-spatial forms and synaesthesia).One aspect of number-form synaesthesia that has not yet beenempirically demonstrated, but has been alluded to in the literature(Galton, 1880; Gertner, Henik, & Cohen Kadosh, 2009; Sagiv,Simner, Collins, Butterworth, & Ward, 2006; Seron, Pesenti, Noël,Deloche, & Cornet, 1992), is the automaticity with whichnumber-forms are experienced and affect behavior. The underlyingassumption is that digits automatically trigger a consciousexperience of a specific spatial location. However, a plausiblealternative that has yet to be ruled out is that these number spacemappings are voluntarily produced via top-down processes.

Synaesthetes frequently report that when they perceive a digit,they cannot inhibit themselves from also experiencing the

ll rights reserved.

f Psychology, University ofCanada N2L 1C3..

l. 9 is Always on top: Assessing

corresponding spatial location that the particular digit synaesthet-ically occupies. To date, researchers have taken the first steps to:(a) objectively confirm whether or not number-form synaesthetesdo indeed experience particular spatial locations upon seeing par-ticular digits (i.e., confirm the reality of number form synaesthesia;Hubbard, Ranzini, Piazza, & Dehaene, 2009; Jarick, Dixon, Maxwell,Nicholls, & Smilek, 2009a; Piazza, Pinel, & Dehaene, 2006; Sagivet al., 2006; Seron et al., 1992), (b) see whether number-formsaffect arithmetic (Ward, Sagiv, & Butterworth, 2009), and (c)determine the brain areas where number-forms might arise (Tang,Ward, & Butterworth, 2008).

A number of researchers have indicated that automaticity is ahallmark of various types of synaesthesia (Dixon, Smilek, Cudahy,& Merikle, 2000; Mattingley, Rich, Yelland, & Bradshaw, 2001). Ingrapheme-color synaesthesia, researchers have used cognitivetasks such as a variant of the Stroop task to show that synaesthetescannot prevent synaesthetic colors from occurring, and cannotignore these colors once they are produced (e.g., Cohen Kadosh &Henik, 2007; Cohen Kadosh, Tzelgov, & Henik, 2008). To date, how-ever, there has yet to be a conclusive demonstration that number-forms are activated automatically. Given that automaticity hasbeen identified as a hallmark of synaesthesia, we set out to builda stronger case for the automaticity of number-forms.

The automaticity of number-forms has been hinted at in a re-cent study by Gertner et al. (2009). These authors tested num-ber-form synaesthetes and non-synaesthetes on a numericalcomparison task (e.g., determining which of two digits is larger).In this task, two digits were either aligned in a horizontal (besideone another) or vertical (on top of one another) orientation andparticipants chose the number that was the largest in magnitude.

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

Fig. 1. L’s original 2-D drawing of her number-form. The arrow represents thedirection in which she prefers to view her number-space.

2 M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx

All of the synaesthetes showed a significant distance effect, suchthat response times increased as the distance between the digitsdecreased, but only when the digits appeared in the orientationconsistent with their number-forms. Non-synaesthetes, on theother hand, showed considerable flexibility in their assigning ofnumbers to space – they showed the distance effect for both orien-tations. When considering the distance effect, the fact that the spa-tial orientation of the presented digits seemed to matter to thesynaesthetes but was irrelevant for the non-synaesthetes sug-gested to Gertner et al., that the numbers were automatically acti-vating a conscious experience of space for the synaesthetes. That is,when the presented digits were aligned the synaesthetic space itexacerbated the distance effect, when they were misaligned it dis-rupted the distance effect. Although Gertner et al.’s (2009) findingsare certainly consistent with the idea that number-forms are elic-ited automatically, there remains the alternative possibility pro-posed by Gheri, Chopping, and Morgan (2008) that thesynaesthetes performed on the task as expected knowing that theywere in some way ‘special’ by having the condition. In other words,it is unclear in Gertner et al. whether the synaesthetes were usingtheir number-forms as a strategy to complete the task, or whethertheir number-forms were automatically activated and influencedtheir performance involuntarily.

A different way to assess the automaticity of number-forms insynaesthesia would be to evaluate the influence of number-formsynaesthesia on spatial attention. Although not all synaesthetes re-port consciously experiencing their number-forms in the spacearound them, for the few that do ‘project’ their numbers out inexternal space (like the synaesthete in the current study) numbersmight act as reflexive attentional cues. That is, seeing or hearing anumber could automatically direct their attention to the locationalong their synaesthetic number-form. Importantly, one couldmake the strong conclusion that number-form synaesthesia occursautomatically if number-forms were to (1) involuntarily (i.e.,reflexively) cause synaesthetes to attend to specific spatial loca-tions, (2) orient attention rapidly, and (3) do so independently ofcognitive control (i.e., top-down processes).

Research on human attention has shown that automatic or‘reflexive’ orienting of attention is commonly triggered by an exog-enous stimulus (e.g., external abrupt onset cue, such a peripheralflash of light). Exogenous cues have been shown to orient attentionto the target location despite explicit instructions to ignore the cue(Jonides, 1981), and despite the cue being counterpredictive ofwhere the target is going to be (e.g., Warner, Juola, & Koshino,1990). Importantly, this type of automatic orienting has been ex-tended more recently to occur for stimuli that are not classic exog-enous cues, such as arrows and eye-gaze (e.g., Friesen & Kingstone,1998; Kuhn & Kingstone, 2009). For instance, Friesen andKingstone (1998) have show that the central presentation of aschematic face looking either left or right could automatically ori-ent attention to the ‘looked at’ locations. The authors supportedtheir claim that eye-gaze causes reflexive shifts of covert attentionusing multiple pieces of evidence. Pertinent to the current studywere that the cueing effects emerged rapidly (within 105 ms be-tween cue and target), orienting of attention to the cued locationsoccurred despite knowing that the cues were non-predictive of thetarget locations, and lastly, a response time benefit occurred to de-tect the cued targets without showing a response time cost at theuncued location. These same characteristics have been widely usedwithin the spatial attention literature to distinguish between voli-tional and reflexive shifts of exogenous attention, demonstratingthat reflexive orienting occurs quickly (Cheal & Lyon, 1991), evenwhen the cue is non-predictive of target location (Jonides, 1981),is short-lived (Müller & Rabbitt, 1989), and results in RT benefitswithout costs (Posner & Synder, 1975). Thus, if a cue survived allof the above criteria it indicated the presence of a specialized

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

attentional system in the brain dedicated to that stimulus (e.g.,eye gaze; Friesen & Kingstone, 1998).

Building on studies of reflexive orienting in non-synaestheticindividuals, we recently modified the Posner cueing paradigm(replicating Fischer, Castel, Dodd, & Pratt, 2003) in an attempt todemonstrate that a given number (1, 2, 8, or 9) could direct a num-ber-form synaesthete’s (L’s) attention to the spatial locations occu-pied by those numbers (Jarick, Dixon, Maxwell, et al., 2009a). Ourfindings clearly demonstrated that L’s attention was shifted tothe locations cued by the numbers, even though she was awarethat the numbers did not predict the location of the target. In otherwords, for L, attention shifts were triggered by the number cues,even though they were irrelevant to the task (i.e., 50% valid, 50% in-valid). Furthermore, L’s attention was not only directed to the loca-tions along her number-form, but the cues successfully directedher attention at every stimulus-onset-asynchrony (SOA) betweencue and target (350, 400, 500, 600, and 700 ms). Critically, L dem-onstrated strong cueing effects at even the shortest cue-target SOAof 350 ms. However, the SOA of 350 ms is still well above the105 ms SOA used in Friesen and Kingstone (1998) to draw conclu-sions about automatic orienting. This leaves open the possibility inJarick et al. that L’s cueing effects might have been the influencedby strategic (as opposed to automatic) processes.

The current study aims to provide more convincing evidencethat numbers automatically or reflexively orient synaesthetes’attention by showing that not only do numbers elicit the experi-ence of spatial locations very quickly (within 150 ms), they do soeven if the cues are counterpredictive (� 85% of the time targetsfall in the uncued location), and despite being explicitly instructedto use an opposing strategy. In the present experiments we againstudied L – a number-form synaesthete whose numbers 1–9 areorthogonal to the standard left-to-right ‘‘mental number line’’(see Fig. 1). L’s numbers ascend upwards from 1 at the bottom to9 at the top. Following the digit 9, her number-forms change direc-tion with the numbers 10–20 running horizontally from left toright, and the numbers 21–40 running from right to left. The atyp-ical structure of L’s number-form has been empirically confirmedusing two tasks: a SNARC-type task and spatial cueing task (Jarick,Dixon, Maxwell, et al., 2009a). Where non-synaesthetes showed asignificant SNARC effect when the response codes were alignedhorizontally (consistent with the mental number line) but not ver-tically, the synaesthete L demonstrated a significant SNARC effect

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx 3

when the response codes were aligned vertically (correspondingwith her number-forms) but not horizontally. This same vertical–horizontal dissociation was replicated using a spatial cueing taskadopted from Fischer et al. (2003). L was centrally presented withsingle digit cues (e.g. 1, 2, 8, or 9) and was very quick to detect tar-gets when they appeared in locations consistent with her number-form. That is, L showed significant cueing-effects when the targetsappeared at the bottom of the screen following low numbers andthe top of the screen following high numbers. Importantly, sheshowed no cueing-effects when targets appeared either on the leftor right of the screen (presumably since these horizontal targetlocations were misaligned with this part of her vertically alignednumber form). She did, however, show cueing-effects for left andright targets when we subsequently cued her attention with thedigits 10, 11, 19, and 20, which for her formed a horizontal seg-ment of her number form.

In the current study we modified the previous cueing task to as-sess the involuntary and automatic nature of L’s number-forms. Inthis task, one of four digits (1, 2, 8, or 9) was centrally presented onthe computer screen flanked by two empty boxes, either at the topand bottom of the display (aligned orientation) or to the left and tothe right (misaligned orientation). Following a variable delay, a tar-get (white circle) appeared in one of the boxes and participantswere instructed to respond to the targets’ appearance as quicklyas possible by pressing a button on a keypad. The SOA betweenthe cue and target was either long (500 ms) or extremely short(150 ms) – a value well below the 350 ms SOA used previouslyand where strategic influences are believed to be precluded(Friesen & Kingstone, 1998). Thus, if we find that L’s number-formsinfluence her spatial attention within this very short time window,numbers must direct her attention to locations specified by hernumber-forms very rapidly, satisfying another attribute of auto-matic orienting (Cheal & Lyon, 1991). To further reduce the poten-tial of strategic effects from influencing L’s response times, we hadtargets fall in synaesthetically ‘‘correct’’ (valid) locations veryrarely – only 14.2% of trials. On the vast majority of trials (85.8%)targets fell in her synaesthetically ‘‘incorrect’’ (invalid) locations.Therefore, if L were to use a strategy to optimize her performanceit would benefit her to direct her attention to a location in spacethat was opposite to the synaesthetic location cued by her num-bers. More concretely, upon seeing a low number in our design,targets were six times as likely to occur at the top of the screen(in the synaesthetically invalid location) compared to the validlocation. Thus if it were possible, the best strategy for L would beto prevent her attention from being directed to her synaesthetical-ly-cued location (i.e., suppress her number-form), and to direct herattention to the top of the screen when seeing a low number andthe bottom of the screen when seeing a high number. As such,by loading up on invalid trials we sought to assess whether auto-matic cueing effects would still emerge in the face of strategic taskdemands to orient her attention to locations that were opposite tothe synaesthetic locations cued by the numbers.

1 Although eye-movements were not monitored using an eye-tracking system,participants were strongly advised to fixate on the central cue and the experimenterwas present in the room to enforce this. It is also well-known that participants willavoid making spontaneous eye-movements to the target location during simpledetection tasks such as ours here (Friesen & Kingstone, 1998; Posner, 1980).

2. Experiment 1

2.1. Methods

2.1.1. ParticipantsA healthy 21-year-old female number-form synaesthete and 12

naïve non-synaesthetic, age-matched University of Waterloo stu-dents (nine females; mean age of 20.5 years) volunteered to partic-ipate for an honorarium. Non-synaesthetes were selected based ona prescreen questionnaire where they selected ‘‘I do not organizenumbers in a spatial arrangement’’ and ‘‘I do not see colors fornumbers or letters.’’ These questions were enough to qualify them

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

for participation without informing them of purpose of the task.Following participation, each participant was debriefed on the pur-pose of the experiment and educated about number-forms andsynaesthesia more generally. Upon learning about synaesthesia,controls were asked again whether or not they had similar experi-ences described by the debriefing. All participants reported no hintof having synaesthesia and were intrigued to learn more about it.Both L and controls reported having normal or corrected-to-normalvision and were right-handed. The University of Waterloo Officefor Research Ethics approved all experimental procedures and par-ticipants gave written consent before participating.

2.1.2. Stimuli and materialsStimuli were presented on a LG 17 in. LCD flatscreen computer

display controlled by a Mac mini. We used the same stimuli as inJarick, Dixon, Maxwell, et al. (2009a). The fixation was a centraldot (diameter 0.1� of visual angle) flanked by two hollow boxes(1� in width and height) that were 10� in eccentricity (approx. 5�from central fixation). For the horizontal condition the boxes ap-peared to the left and right of the display, whereas for the verticalcondition they were situated at the bottom and top of the display.The cues were the digits 1, 2, 8, and 9 in Arial font (subtending 2�).The target was a white circle (0.7�) that appeared inside one of thetwo boxes. All stimuli were white presented on a black back-ground. Manual button-presses were made on a response padequipped to record response times with 1-ms accuracy (CedrusRB 530). SuperLab 4.0 was used to present the stimuli and recordthe response times of each participant.

2.1.3. ProcedureParticipants were seated comfortably at a distance of 57 cm in

front of the computer monitor and response pad. Typical trialevents were as follows: fixation for 500 ms, replaced by a cue(either 1, 2, 8, or 9) for 150 ms, followed by a target. Targets wereeither presented immediately after the cue (150 ms SOA), or after350 ms (500 ms cue-target SOA). Targets remained on screen untilthe participant responded or 3500 ms elapsed. On ‘‘catch’’ trials, notarget was presented and the boxes remained empty for 3500 ms.Participants were asked to focus centrally on the fixation dot andthe digit cue for the duration of the experiment.1 They were in-structed to detect the target circle that appeared in one of the boxesas quickly and accurately as possible by pressing the central buttonon the response pad, and to withhold their responses on trials whenno target appeared (catch trials). These ‘‘catch’’ trials were insertedto make sure that participants were performing the task accuratelyand were not responding blindly.

All participants completed the horizontal (misaligned for L) con-dition first followed by the vertical (aligned) condition second,(yoked to the order in which L was tested). In each condition partic-ipants were given 10 practice trials (four valid, four invalid, twocatch) followed by six blocks of experimental trials with a 1-minbreak in between. Each block contained 16 valid trials (two repeti-tions of each digit at each SOA), 96 invalid trials (12 repetitions ofeach digit at each SOA), and 16 catch trials (four repetitions of eachdigit with no target). Thus, in terms of the overall experiment theproportion of invalid trials amounted to 85.8% compared to only14.2% valid trials. That is, the cues in this design were predictive ofthe target location on 85.8% of the trials (but note for the synaes-thete, they were predictive of the ‘‘wrong’’ location). In addition tothe experimental trials, we included 12.5% catch trials. All trials

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

4 M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx

were randomized within each block. Overall participants completeda total of 768 trials for each condition (horizontal and vertical), withthe entire experiment lasting approximately 50 min in duration.

2.2. Results and discussion

L’s performance on the ‘catch’ trials was nearly perfect (99%accurate). Only control participants that performed greater than80% catch trials were included (mean of controls was 93.5%), whichexcluded three participants. All response times were submitted toa non-recursive outlier rejection procedure that adjusts the cut-offcriterion for an outlier depending on the size of the sample (VanSelst & Jolicoeur, 1994). Thus, observations greater than ±2.5 stan-dard deviations were removed for the invalid trials (n = 288) andobservations greater than ±2.47 standard deviations were removedfor the valid trials (n = 48). This procedure resulted in few trialsbeing discarded for L (0.31% valid, 0.06% invalid) and controls(0.76% valid, 0.09% invalid). The remaining response times for Land the group of controls were submitted to a three-way analysisof variance (ANOVA) involving orientation (horizontal or vertical),SOA (150 or 500 ms), and validity (valid or invalid). Individualthree-way ANOVAs were also performed on each of the nine con-trols separately and Bonferroni corrected to an alpha of .005. Themean response times, standard deviations, and cueing-effect size(invalid minus valid RTs) for L and controls can be seen in Table 1for the horizontal and Table 2 for the vertical orientation. Note thatvalidity in the horizontal condition was in reference to the mentalnumber line (low digits on the left and higher digits on the right),while validity in the vertical condition was in reference to L’s num-ber-form (low digits on the bottom and higher digits on the top).

For both L and controls, the ANOVAs showed a significant main ef-fect of SOA [F(1, 1289) = 207.45, p < .001 and F(1, 8) = 45.7, p < .001,respectively], where response times decreased as the delay betweencue and target increased. This finding is common with a variety ofSOAs and is known to reflect the Variable Foreperiod effect describedin Vallesi, Shallice, and Walsh (2007), where response times havebeen shown to decrease as time to prepare for the upcoming targetincreases. SOA did not interact with any other variables.

For the horizontal (misaligned for L) condition, high digits pre-dicted the target to appear on the left and low digits predicted tar-gets to appear on the right (i.e., 85.8% invalid locations), so that theproper strategy would be in opposition with the mental numberline. For the vertical (aligned for L) condition, high digits predictedthe target to appear on the bottom and low digits predicted targetsto appear on the top (i.e., 85.8% invalid locations), so that the prop-er strategy would be in opposition to L’s number-forms. If L were touse a strategy for this task she should use the same one as controls,and thus show a similar pattern of response times as controls

Table 1Experiment 1 – horizontal condition: No explicit instruction. Response times (RT), standardindividual non-synaesthetic control.

SOA 150

Valid Invalid Cueing effect

RT (SD) RT (SD) RT (In–V)

L 388 (82) 380 (60) �8C1 402 (49) 392 (50) �10C2 444 (70) 478 (89) 34C3 383 (44) 375 (45) �8C4 475 (83) 466 (85) �9C5 353 (35) 354 (48) 1C6 360 (51) 353 (48) �7C7 410 (73) 413 (71) 3C8 284 (50) 287 (45) 3C9 376 (49) 371 (54) �5Avg. 387 (56) 388 (59) 1

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

regardless of orientation. Given the predictive nature of the cues(85.8% invalid, 14.2% valid), the best strategy for L and controlswould be to direct attention to the ‘‘invalid’’ target locations:left/bottom following high digits (8, 9) or right/top following lowdigits (1, 2). The ANOVA for L revealed a significant two-way inter-action between orientation and validity, F(1, 1289) = 65.34,p < .001, showing that L only demonstrated cueing-effects (fasterRTs for valid targets instead of invalid targets) in the vertical orien-tation but not in the horizontal orientation. This was not the casefor the non-synaesthetic controls who showed no significant inter-action, F(1, 8) = 1.703, p = .228, n.s, nor a main effect of validity,F(1, 8) = 1.088, p = .327, n.s. In fact, the individual ANOVAs revealedthat not even one control showed this significant interaction be-tween orientation and validity (F’s ranged from .001 to 3.741 andlowest p-value = .053, which is below the Bonferroni-correctedalpha < .005).

To more directly compare L’s cueing-effects to those of controls,we performed the Revised Standardized Difference Test (RSDT;Crawford & Garthwaite, 2005). This test is designed to assesswhether a single case (e.g., L) shows differences between condi-tions (valid vs. invalid) that are larger than the differences shownby a small control sample. We performed this analysis for the longand short SOAs for both the vertical and horizontal orientationsand applied the appropriate Bonferroni correction (alpha of .05/4 = .0125). We predicted that L would show significantly largervalidity effects for vertically aligned targets but not for horizontallymisaligned targets. Indeed, for the horizontal targets, L’s responsetime differences between valid and invalid targets for both 150and 500 ms SOAs were not significantly different from controls,RDST t(8) = �.59, p = .568 and t(8) = �.38, p = .356. However forthe vertical (aligned) targets, L’s cueing effects (invalid minus validRTs) at the 150 ms SOA condition were significantly larger thanthose of controls, RSDT t(8) = �9.19, p < .001 (one-tailed). More-over, these same cueing effects were still larger for L at the500 ms SOA, RSDT t(8) = �2.7, p = .01 (one-tailed). Recall that if Lwere applying a performance optimization strategy (i.e., strategyopposite to her number-forms), L should have shown faster perfor-mance on invalid trials, rather than showing her standard cueingeffects where valid trials are responded to faster than invalid trials.

What our data suggest is that neither L nor the controls em-ployed the optimal strategy, or possibly any strategy at all. Thatis, even though the ANOVA revealed a significant cueing-effectfor L in the vertical condition, it was not in the direction that wouldhave optimized performance. Rather L continued to respond withreference to her number-form, such that she was much quickerto detect targets in their synaesthetically valid locations (14.2%of the trials) than invalid locations (85.8% of the trials). Further-more, the absence of an interaction between SOA and validity

deviations (SD) and cueing effects (invalid minus valid trials) for L compared to each

500

Valid Invalid Cueing effect

RT (SD) RT (SD) RT (In–V)

319 (52) 314 (43) �5352 (52) 342 (53) �10404 (85) 386 (79) �18342 (44) 336 (55) �6420 (84) 427 (91) 7312 (41) 327 (48) 15316 (32) 310 (42) �6326 (76) 337 (64) 11266 (41) 268 (41) 2317 (59) 316 (51) �1339 (57) 339 (58) 0

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

Table 2Experiment 1 – vertical condition: No explicit instruction. Response times (RT), standard deviations (SD) and cueing effects (invalid minus valid trials) for L compared to eachindividual non-synaesthetic control.

SOA 150 500

Valid Invalid Cueing effect Valid Invalid Cueing effect

RT (SD) RT (SD) RT (In–V) RT (SD) RT (SD) RT (In–V)

L 428 (75) 510 (72) 82 345 (51) 421 (55) 76C1 380 (64) 371 (53) �9 334 (61) 339 (60) 5C2 415 (71) 409 (58) �6 322 (37) 339 (49) 17C3 348 (49) 352 (55) 4 305 (44) 323 (62) 18C4 479 (105) 452 (82) �27 364 (53) 411 (86) 47C5 341 (50) 333 (47) �8 323 (38) 331 (45) 8C6 362 (60) 366 (58) 4 315 (49) 321 (58) 6C7 368 (43) 379 (63) 11 332 (67) 336 (71) 4C8 275 (29) 280 (40) 5 276 (38) 276 (44) 0C9 382 (57) 377 (57) �5 324 (55) 314 (49) �10Avg. 372 (59) 369 (57) �3 322 (49) 332 (58) 11

M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx 5

paired with the significant RSDT results shows that this validity ef-fect occurred at both short (150 ms) and long (500 ms) SOAs be-tween cue and target. Earlier we mentioned that if L were to usea strategy for detecting the targets in the vertical condition, sheshould orient her attention to the top when cued by a low digitand to the bottom when cued by a high digit (opposite to her syn-aesthetic number-space). Our findings suggest that L did not (orpotentially could not) use this strategy and her number-forms con-tinued to bias her attention to the synaesthetically valid locations.

These finding are consistent with our previous work (Jarick,Dixon, Stewart, Maxwell, & Smilek, 2009b; Jarick, Dixon, Maxwell,et al., 2009a) using a non-predictive design (50% valid, 50% invalidtargets). As in that study, L demonstrated cueing-effects only whenthe targets were aligned with her number-forms (vertical condition)and showed no cueing effects when target locations were misa-ligned (horizontal condition). Importantly, despite the fact that weloaded up on invalid trials so that moving her attention to her syn-aesthetic location would yield a performance cost on the vastmajority of trials, L still responded in accordance with her synaes-thesia. Critically, not only did L show cueing-effects with the SOAof 500 ms, but she showed cueing-effects with only 150 ms be-tween cue and target onset as well. L’s ability to rapidly detect va-lid targets presented 150 ms after cue onset implies that thenumber cues must have oriented her attention automatically.

3. Experiment 2

What is evident from Experiment 1 is that L’s number-form caninfluence her attention very quickly (within 150 ms) in spite of thecues being counterpredictive (cues predicted the opposite targetlocations). Importantly, this was only observed in the vertical con-dition when the targets were aligned with her number-form, suchthat L responded significantly faster to detect the targets whenthey appeared in her synaesthetically valid locations. Although Ldemonstrated synaesthetic cueing effects at the shortest SOA of150 ms – suggestive of automatic orienting – there is still the pos-sibility that conscious controlled processes might have played arole. In Experiment 2 we sought to obtain even stronger evidencethat L’s number-forms can automatically influence the allocationof attention.

The procedure used in Experiment 2 was based on a methoddeveloped in the memory and perception literature (see Debner& Jacoby, 1994; Jacoby, Lindsay, & Toth, 1992; Jacoby & White-house, 1989) to unambiguously isolate the influence of automaticprocesses. Evident from this literature is the notion that most tasksare not ‘‘process pure’’ measure of automaticity and are likely al-ways contaminated by controlled processing. In order to controlfor this contamination, Jacoby, Toth, and Yonelinas (1993)

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

developed the exclusion task in which controlled and automaticprocesses are put in opposition. For instance, in Jacoby et al.(1993) participants were shown a study list of words followed bya stem-completion task in which participants were explicitly toldto complete the stems with words not seen in the previous studylist. In this condition, the automatic influences from reading thewords in the study list were pitted against conscious recollectionof the study words. That is, if participants completed the stem withthe word presented in the study list above a baseline chance level,it could not be due to conscious recollection of the item (or elsethey should have omitted the study word and chose a new one).The only explanation left would be that the study words automat-ically influenced participants’ word choices for completing thestems. This exclusion task is now viewed as the gold standard fordetermining the influence of automatic processes on perceptionand memory.

In Experiment 2 we applied a variant of the exclusion task toevaluate whether L’s cueing effects shown in Experiment 1 (evenat the 150 ms SOA) were due to automatic influences of her num-ber-form without the possibility of contamination by controlled,top-down strategies. Hence, in Experiment 2 we explicitly instructedL to use a performance optimization strategy that was based on thefact that for the vast majority of trials, one could correctly predictwhere the target would be located based on the identity of the digit(e.g., in the vertical condition, low numbers predict targets on thetop and high numbers predict targets on the bottom of the screen).As in Experiment 1, this strategy, if implemented, would be in directcontrast with the organization of L’s number-forms. Our goal in thisdesign was to pit conscious controlled strategy against automatic-ity. That is, in order to use the opposing strategy dictated by theexperimenter, L would have to voluntarily direct her attention toher synaesthetically invalid locations. There are only two ways inwhich she could accomplish this: (1) by suppressing her number-forms and using the strategy by focusing on the predictability ofthe cue, or (2) allow the cues to involuntarily direct her attention to-wards the synaesthetically valid locations, and then re-direct herattention to the invalid locations. Both of these processes shouldtake time and require cognitive effort.

Therefore, if L is able to suppress her number-form and followthe opposing strategy predicted by the cues, we should find theopposite pattern of results from Experiment 1. That is, L should re-spond more quickly to detect synaesthetically invalid targets forboth long and short SOAs. Alternatively, if L’s attention is involun-tarily directed to synaesthetically valid locations and she has to re-direct her attention to the opposite locations to comply with thestrategy, this re-directing of attention should take time and poten-tial cueing effects could be diminished. However if as Experiment 1suggests, L’s number-forms influence her attention rapidly, thenwe might see evidence of this automatic influence at the shortest

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

6 M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx

SOA of 150 ms in the vertical condition that is resistant to top-down control. If this is the case, L should continue to respond fast-est to the synaesthetically valid targets in this short amount oftime. Given a longer SOA however, L should have ample time toemploy one of the above strategies, either allowing her to reversethe direction of the cueing effects (i.e., faster for invalidly cued tar-gets), or eliminating cueing effects altogether. In contrast, L shouldbe able to implement the strategy and show cueing effects for thehorizontal condition seeing as though her number-forms (beingvertical) do not conflict with the target locations – if not for theshort SOA, she should for the longer SOA.

Since we did not find cueing effects in the non-synaestheticcontrol group consistent with the cue-target predictability inExperiment 1, we reasoned that perhaps 500 ms was still not en-ough time for non-synaesthetes to show cueing effects in this typeof task (consistent with Fischer et al., 2003 and Jarick, Dixon,Maxwell, et al., 2009a). Thus, we increased the long SOA from500 ms in Experiment 1 to 800 ms in Experiment 2. Now withthe longer SOA of 800 ms and the explicit instruction to followthe strategy predicted by the number cues, we expect to findcueing effects with controls consistent with the cue-target predict-ability in both vertical and horizontal orientations. Yet, we stillpredict that the SOA of 150 ms will be too short to allow time forstrategy use, and therefore controls will likely not show any re-sponse time differences between valid and invalid cues in this case.

3.1. Methods

3.1.1. ParticipantsThe same number-form synaesthete (L) and 12 naïve2 non-syn-

aesthetic University of Waterloo students (10 females; average ageof 21 years) volunteered to participate for an honorarium. All partic-ipants reported having normal or corrected-to-normal vision, wereright-handed, and were native English speakers. The University ofWaterloo Office for Research Ethics approved all experimental pro-cedures and participants gave written consent before participating.

3.1.2. Stimuli and designWe used the same stimuli and design as Experiment 1, with the

exception of the long SOA. We used a longer SOA of 800 ms pairedwith the short SOA of 150 ms between cue and target onset to givecontrols and L ample time to implement a strategy.

3.1.3. ProcedureThe procedure was identical to that of Experiment 1, except

with different instructions. Instead of the participants being naïveto the experimental design, we explicitly informed L and controlsof the disproportional trial types in both orientation conditions.For example, we informed participants in the vertical orientationthat, ‘‘approximately 85% the targets will appear at the top follow-ing a low digit (1 or 2) and at the bottom following a high digit (8or 9)’’. We further encouraged them to use that strategy to improvetheir performance during the experiment.

3.2. Results and discussion

L performed perfectly on the catch trials (100% accurate in bothconditions) and we included only those control participants thatperformed greater than 80% or better on these trials. No controlswere excluded on this basis. Controls’ catch trial means were

2 Although some studies have trained controls to compare with each synaesthete(e.g., Meier & Rothen, 2009), we believed that the best chance of having participantsperform similar to L would be to make them explicitly aware of digit cue-targetpredictability and instruct them to use that strategy to optimize their performance,rather than asking them to visualize an arbitrary array of numbers.

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

98.7% for the horizontal condition and 95.2% for the vertical condi-tion. All response times were submitted to the observationweighted non-recursive outlier procedure (Van Selst & Jolicoeur,1994). Thus, observations greater than ±2.47 standard deviationswere removed for the valid trials (n = 48) and observations greaterthan ±2.5 standard deviations were removed for invalid trials(n = 288). This procedure resulted in few trials being discardedfor L (0.62% valid, 0.04% invalid) and controls (0.12% valid, 0.09% in-valid). The remaining response times for L and the group of con-trols were submitted to a three-way analysis of variance(ANOVA) involving orientation (horizontal or vertical), SOA (150or 800 ms), and validity (valid or invalid). The mean response timesfor L across the two orientations and SOAs are shown in Fig. 2 andcontrols in Tables 3 and 4. Like Experiment 1, both L and controlsshowed the Variable Foreperiod effect F(1, 1298) = 216.38,p < .0001 and F(1, 11) = 21.98, p = .001, respectively. That is, partic-ipants were faster to detect the targets following the longer SOA of800 ms than 150 ms.

The ANOVA for L revealed a significant two-way interaction be-tween orientation and validity F(1, 1298) = 149.77, p < .001. As canbe seen in Fig. 2, when the targets were aligned horizontally, shewas able to take advantage of the strategy that was given to her– targets in ‘‘invalid’’ locations (with respect to the standard num-ber line) were responded to faster than targets in ‘‘valid’’ locations.This pattern occurred at both short and long SOAs (both p-val-ues < .0001). By contrast, in the vertical condition when target loca-tions were aligned with her number forms (i.e., presented either onthe top or the bottom of the screen), yet presented 85.8% of thetime in invalid locations, she was unable to take advantage of thestrategy. Indeed at the long SOA of 800 ms, she showed no differ-ence between valid and invalid trials (p = .472), and at the shortSOA she showed faster valid than invalid trials (p = .032) – a datapattern that is opposite to the strategy predicted by the cues, butconsistent with the hypothesis that numbers automatically orient

Fig. 2. Number-form synaesthete’s response times (RTs) to detect targets in thehorizontal, misaligned orientation (A) and vertical, aligned orientation (B) for bothExperiments 1 (no strategy) and 2 (strategy given). The error bars representconfidence intervals and the asterisks symbolize significance – �p < .01 and��p < .001.

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

Table 3Experiment 2 – horizontal condition: Explicit instruction given to use a strategy. Response times (RT), standard deviations (SD) and cueing effects (invalid minus valid trials) for Lcompared to each individual non-synaesthetic control.

SOA 150 800

Valid Invalid Cueing effect Valid Invalid Cueing effect

RT (SD) RT (SD) RT (In–V) RT (SD) RT (SD) RT (In–V)

L 354 (39) 296 (25) �58 355 (41) 262 (21) �93C1 376 (51) 362 (40) �14 341 (52) 305 (39) �36C2 340 (40) 340 (37) 0 302 (22) 311 (36) 9C3 334 (36) 332 (30) �2 307 (36) 291 (32) �16C4 336 (52) 326 (52) �10 311 (59) 317 (56) 6C5 332 (52) 321 (44) �11 286 (31) 288 (37) 2C6 332 (24) 335 (30) 3 303 (37) 293 (36) �10C7 358 (40) 356 (47) �2 322 (42) 319 (45) �3C8 434 (37) 423 (49) �11 345 (46) 352 (53) 7C9 310 (43) 304 (37) �6 322 (52) 294 (37) �28C10 346 (34) 340 (36) �6 260 (23) 277 (37) 17C11 339 (33) 331 (27) �9 317 (39) 310 (33) �7C12 317 (44) 330 (35) 13 309 (40) 312 (48) 3Avg. 346 (41) 342 (40) �5 310 (40) 306 (41) �5

Table 4Experiment 2 – vertical condition: Explicit instruction given to use a strategy. Response times (RT), standard deviations (SD) and cueing effects (invalid minus valid trials) for Lcompared to each individual non-synaesthetic control.

SOA 150 800

Valid Invalid Cueing effect Valid Invalid Cueing effect

RT (SD) RT (SD) RT (In–V) RT (SD) RT (SD) RT (In–V)

L 402 (61) 425 (75) 22 320 (48) 325 (46) 5C1 392 (53) 370 (49) �22 375 (58) 319 (47) �56C2 356 (41) 352 (44) �4 395 (55) 387 (55) �8C3 354 (38) 341 (45) �13 322 (41) 330 (40) 8C4 384 (109) 387 (118) 3 362 (96) 378 (87) 16C5 336 (36) 336 (34) 0 340 (42) 340 (47) 0C6 330 (34) 337 (24) 7 328 (37) 319 (41) �9C7 377 (52) 374 (36) �3 378 (53) 368 (66) �10C8 401 (47) 412 (37) 11 452 (44) 435 (62) �17C9 333 (49) 299 (42) �34 326 (45) 311 (41) �15C10 355 (52) 356 (56) 1 333 (56) 329 (57) �4C11 325 (30) 317 (30) �8 361 (46) 345 (40) �16C12 334 (36) 327 (37) �7 341 (53) 346 (46) 5Avg. 356 (48) 351 (46) �6 359 (52) 351 (52) �9

M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx 7

her attention to specific spatial locations dictated by her synaes-thetic number-form.

The ANOVA for the group of non-synaesthetes revealed only asignificant main effect due to orientation, F(1, 11) = 43.14, p < .001and a significant SOA by orientation interaction, F(1, 11) = 15.33,p = .002. Unlike L, controls showed no significant two-way interac-tion between orientation and validity, F(1, 11) = 1.538, p = .241, n.s.Indeed this interaction was not apparent in any of the 12 controls(F-values ranged from .049 to 3.44, lowest p-value = .064, which isbelow the Bonferroni alpha of .004). We are not surprised at thisnull two-way interaction since there is no reason to believe thatcontrols should differ in their strategy use between horizontaland vertical orientations. What we would expect if controls wereusing the strategy dictated by the cues, however, is a main effectof validity regardless of orientation. Yet, the ANOVA only showeda marginal main effect of validity, F(1, 11) = 3.83, p = .076, n.s. Visualinspection of Tables 3 and 4 suggest that there is much variabilityacross the controls, with cueing-effects ranging from �56 (invalidfaster than valid) to +17 (valid faster than invalid). In the horizontalcondition, the RTs in about half of the controls trended towards theappropriate strategy, while the other half seemed capable at the150 ms SOA but not at the 800 ms SOA. In the vertical condition,this pattern increased to 8 of the 12 that showed a trend in theexpected direction, while the other four seemed to show minimalcueing effects. In fact, the individual ANOVAs revealed that four

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

of the controls did show a significant main effect of validity (allp-values < .05). Thus, the evidence that even some controls wereable to employ the appropriate strategy suggests that the strategymanipulation was successful. The individual differences in theirability to use the strategy could potentially be a product of interfer-ence from pre-existing mental number lines. While Dehaene, Bos-sini, and Giraux (1993) has demonstrated that the left-to-rightmental number line is apparent in the majority of the population,there is evidence to suggest that a vertical number line also exists(Gevers, Lammertyn, Notebaert, Vertguts, & Fias, 2006; Ito & Hatta,2004; Santens & Gevers, 2008; Schwartz & Keus, 2004). As we willnote in the Discussion, the strength of individual number lines islikely to fall along a continuum of numerical-spatial ability, fromno spatial experience at all to a very lucid number line arrangement(i.e., number-form synaesthesia).

As in Experiment 1, we again performed the Revised Standard-ized Difference Test (RSDT) to directly compare L to controls in theresponse times to both orientations. Although we did not find asignificant three-way interaction between Orientation, SOA, andValidity for the ANOVA, the RDST procedure is actually the appro-priate test to indentify differences between one case and a smallsample in this respect. Our predictions were that L would showsignificantly larger validity effects for the vertically aligned targetsat the short (150 ms) SOA, but not at the longer (800 ms) SOA. In-deed, the RSDT for vertical targets revealed that L’s cueing effects

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

Fig. 3. Schematic ‘‘birds-eye’’ view of L’s preferred mental vantage point (MVP) atthe bottom and L’s revised MVP once she was explicitly instructed to use a strategywith high digits cueing her to the left and low digits cueing her to the right(Experiment 2).

8 M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx

were significantly greater than controls at the shortest SOA of150 ms, RSDT t(11) = �1.97, p < .05 (one-tailed), but not at the long-er SOA of 800 ms, RSDT t(11) = �.849, p = .207 (one-tailed). Thus, aspredicted L’s response times were quicker to detect the synaesthet-ically valid targets than invalid targets when the SOA was too shortto implement a strategy (within 150 ms). In other words, when thetargets were aligned with her number-forms and the SOA was toorapid for her to use a strategy, L was much quicker to detect targetsthat fell in the synaesthetically valid locations than the invalidlocations. This finding replicates the validity effect shown in Exper-iment 1. As such, it confirms that L’s number-forms are elicitedrapidly despite her deliberate intentions to use the strategy pro-vided. The fact that she was attempting to use the strategy canbe seen by her performance in the 800 ms SOA condition. In con-trast to Experiment 1 where she showed a significant cueing effect,in this experiment her attempts to use the strategy abolished anycueing effects. Thus, one must assume that in the vertical condi-tion, she adopted a cognitive set where she attempted to use thestrategy on all trials. With enough time to implement the strategy(in the 800 ms condition) she was able to overcome her synaes-thetic number-space mappings and diminish the cueing effectsfound previously. Without enough time to implement the strategy,her performance in the 150 ms SOA condition was dictated by theautomatic cueing of her attention by the numbers to their synaes-thetically valid locations.

Like Experiment 1, we predicted that L would not show validityeffects for the horizontal misaligned targets compared to controls.Contrary to our predictions, the RSDT for the horizontal (misa-ligned) condition revealed significantly larger cueing effects for Lcompared to controls for both the 800 ms SOA, t(11) = �3.85,p < .001 and for the 150 ms SOA, t(11) = �3.41, p < .001. Particu-larly surprising was L’s significantly faster response times to the in-valid trials compared to valid trials at the shortest SOA of 150 ms.Since this SOA seems too short to implement a performance opti-mization strategy, another explanation must be sought. Previousresearch with L has showed her remarkable ability to change themental vantage point (MVP) with which she views her synaesthet-ic forms (Jarick, Dixon, Stewart, et al., 2009b). Regarding her num-ber forms, L claims that although her typical mental vantage pointis in front of the number 1, so that the numbers 1–9 rise up in frontof her as shown in Fig. 3. This is her typical MVP, yet she reports(through informal interviews) that she can move around in hernumber-space, and take on a variety of MVPs. Indeed she did re-port having to switch MVPs in elementary school when presentedwith the horizontal number line in mathematics. Thus, given theinstruction that high digits would most often be followed by left-ward targets, and low digits by rightward targets, one way inwhich L could align her vertical number-form to this particularspatial arrangement is by mentally viewing her number form fromthe left side with the number 5 in front of her (the ‘‘revised’’ MVPin Fig. 3). Note, that from this MVP her vertical number-form be-comes a horizontal, right-to-left number-form that is now alignedto the specific strategy instructions. Viewed from this new MVP,low numbers would now rapidly cue her attention to her right,and high numbers to her left. As such, with this new MVP, whatwe have called invalid locations (for the standard left-to-rightnumber line) would be synaesthetically valid locations for L – a sit-uation that could account for the large cueing effects shown by L atthis very short cue-target SOA. This explanation presumes thatwhile she can adopt certain MVPs, others may be too incompatiblewith her preferred MVP (e.g., in front of the digit 9 in Fig. 3). Thus,while she can view her rising number form from the side via a 90�change in MVP (the revised MVP in Fig. 3), she cannot completelyreverse her MVP by 180� and look at her rising number form fromthe top. From informal conversations with L, if she were to switchher MVP in this way, high numbers would be on the bottom and

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

low numbers on the top, but the numbers would then appear in-verted and perhaps cause more interference than facilitation.Therefore, it is very possible that there are constraints on her po-tential MVPs, whereby she can tolerate looking at a vertical formfrom the side, but not a 180� perturbation of her habitual (or‘‘canonical’’) mental vantage point for viewing this form.

4. Overall conclusions

Overall, these experiments examined whether synaestheticnumber-forms could automatically (or reflexively) orient L’s spa-tial attention to the number locations, independent of any strategicinfluences imposed on by the task. To investigate this, we used aspatial-cueing paradigm where we: (1) limited the SOA betweencue and target to be extremely short �150 ms, (2) imposed an im-plicit strategy manipulation by having the cues be counterpredic-tive of the target locations (i.e., by increasing the proportion ofinvalid trials compared to valid trials), and (3) explicitly instructingL to direction her attention in opposition to her number-form. Re-search has repeatedly shown that this type of opposition logic isthe only process pure way to be certain that top-down processesare not ‘contaminating’ automatic effects (Jacoby et al., 1992).Thus, if L’s number-forms were elicited without voluntary control,then the number cues should direct her visual attention to the syn-aesthetically valid target locations at the shortest SOA of 150 ms.This was true in Experiment 1, where L demonstrated cueing-ef-fects consistent with her synaesthetic number-form (faster RTsfor valid trials) at both the 150 and 500 ms SOAs, but only whenthe targets were aligned vertically. The cueing effects at the150 ms SOA were even observed once L was explicitly informedabout the uneven trial proportions (85.8% invalid) in Experiment2. Although she was clearly instructed to use the cues to predictthe target locations, she was still faster to detect targets in the syn-aesthetically cued locations than at the uncued locations in the150 ms condition. In this situation, automaticity was placed indirect conflict with top-down processes, and the automatic processprevailed – thus providing the strongest evidence that the numbercues triggered L’s attention automatically. At the longer SOA of800 ms, however, it is evident that top-down, strategic processeswere having an effect on L’s attention but only enough to diminish

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx 9

the cueing-effects found in the 150 ms SOA. That is, once L had am-ple time between cue and target onset, it seemed as though shewas able to exert some control over where her attention was allo-cated, most likely by re-directing her attention from the cued (va-lid) locations to the invalid locations.

Although we cannot be certain of the strategy that L wasemploying to diminish the vertical cueing-effects at the 800 msSOA in Experiment 2, what is clear from our data is that L’s uniquenumber-form can influence her attention rapidly (within 150 ms)and seems impossible for her to suppress its’ influence in a shortperiod of time. These findings support Gertner et al. (2009) who ar-gued that number-forms are involuntarily activated in synaes-thetes, yet our data limits their claim and suggest that only atvery short time intervals, or if not instructed to use a strategy, doesthis inflexibility occur. Here in Experiment 2, we showed that Lcould flexibly utilize her number-forms to improve her perfor-mance in the horizontal condition, where the strategy did not di-rectly conflict with her preferred mental vantage point. Our dataindicates that L was able to turn her vertical number-form into aright-to-left number-form by changing her MVP. With this newMVP, L showed that low numbers automatically cued rightwardspatial locations and high numbers leftward locations. That thesecueing effects were automatic is supported by the findings thatthey emerged at a very short SOA, when cues were counterpredic-tive, and independent of top-down, strategic control. Thus, whereGertner et al. (2009) argued that synaesthetes have very concreteforms in which encountering visual numbers causes their uniqueforms to ‘‘pop out’’ (pp. 372) in an inflexible manner, we wouldsuggest that although the number forms may be inflexible, theymay be mentally viewed from different vantage points. Whilethere may be some flexibility with which these mental vantagepoints can be viewed, not all mental vantage points appear to bepossible – L could not adopt a vantage point that directly conflictedwith her preferred vantage point.

4.1. Individual differences

It is important to keep in mind that this research is heavilybased on the results on one synaesthete, and that synaesthetesnot only differ significantly from non-synaesthetes in terms oftheir number-form structure; they also differ from one anotherin terms of their experience. From multiple interviews with syn-aesthetes, one can observe qualitative differences from one synaes-thete to the next. One empirical demonstration of these individualdifferences has been shown by Dixon, Smilek, and Merikle (2004),where they report that some synaesthetes experience their synaes-thesia being projected out into space (termed projectors), whileothers experience their synaesthesia in their ‘‘minds eye’’ (termedassociators). With such differences between synaesthetes, it is nowonder why some studies could find no differences between syn-aesthetes and controls (e.g., Piazza et al., 2006), while others havefound strong differences between the two groups (e.g., Hubbardet al., 2009; Jarick, Dixon, Maxwell, et al., 2009a; Jarick, Dixon,Stewart, et al., 2009b; Simner, Mayo, & Spillar, 2009). In fact, Smi-lek, Callejas, Dixon, & Merikle, (2007) highlights this individualvariation among time–space synaesthetes, by showing strongeffects in only two of the four that they had tested. Another indi-vidual difference can be noted in this paper, where L claims to beable to ‘navigate’ within her number and time spaces, taking on avariety of MVPs. As expressed by Sagiv et al. (2006), it is remainsto be an open question whether or not synaesthetes who movearound in their synaesthetic space do so by altering the spatial-form itself or by flexibly modifying their mental vantage point.Our intuition is that both cases could occur. In L’s case here, shehas reported moving herself within the spatial form to view it dif-ferently, however, we have interviewed synaesthetes who report

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

being able to manipulate the form, such as ‘zooming in’ or ‘turningit to a certain month’. The fascinating individual differences re-ported by spatial-form synaesthetes definitely warrant futureinvestigations. Yet, the important message is that in order for thesynaesthesia literature to advance with an accurate understandingof the phenomenon, group studies need to control for individualdifferences between synaesthetes.

4.2. Limitations

Although we feel that our findings are important in highlightingwhat is possible in number-form synaesthesia, our results must beconsidered carefully, since we did concentrate on one synaesthete(L). While the case-study approach provides many advantages, italso provides some disadvantages. One obvious disadvantage isthe limited in the generalizability of our findings, in that we canonly confidently report that our findings occurred for one synaes-thete and do not include all number-form synaesthetes. Althoughit is ideal to report findings that will apply to all synaesthetes, evengroup studies are subject to this limitation due to the substantialvariability across individuals. A second disadvantage is that L hastaken part in previous studies (Jarick, Dixon, Maxwell, et al.,2009a; Jarick, Dixon, Stewart, et al., 2009b) and therefore is likelymore ‘practiced’ than naïve non-synaesthetes at the spatial-cueingtask (even though they were completed years apart). Some mightsuggest that awareness of the task purpose and her willingnessto help could have potential inflated her motivation to performwell. However, the need to perform well on this task should haveencouraged L to reverse her number-form, not respond in accor-dance with it. Indeed this exact behavior can be seen in Experiment2, where L was able to follow the strategy given in the Horizontaltask, but not the Vertical. Thus, although being practiced at the taskis a valid concern, we believe it is highly unlikely that L clued intothe task purpose and performed consistent with task demands. Weaccept that L could have been more motivated than non-synaes-thetes, but this is also the case for every synaesthesia study. Inspite of these disadvantages, we regard L as being a rare individualwhose abilities represent what is possible among number formsynaesthetes, rather than the average number-form synaesthete.L’s unique experience makes it easy to study her accuratelyand reliably. For instance, she (a) possess a vertically alignednumber-form that is directly orthogonal to the horizontal mentalnumber line, (b) is able to portray her three-dimensional num-ber-space onto a two-dimension computer screen without losingthe integrity of the structure, (c) projects her number-form intoexternal space (the prevalence of ‘projector’ number-form synaes-thetes is presently unknown), and (d) has the potential to modifyher mental vantage point to view her number-form from differentperspectives (L is currently the only synaesthete with objectivedata to verify this ability – Jarick, Dixon, Maxwell, et al., 2009a;Jarick, Dixon, Stewart, et al., 2009b). Therefore, for the purpose ofthe present study, we strongly believe that the advantagesoutweigh any disadvantages.

4.3. Conclusions

In sum, we focused on one synaesthete who has reported thatshe ‘‘projects’’ the numbers out in the space around her. Thus,she perceptually experiences numbers in very specific spatial loca-tions that unlike non-synaesthetes, is a very vivid and consciousexperience for her. Although the claims made in this study areheavily reliant on one case-study, we strongly believe that thisone case (L) is representative of individuals who belong on the ex-treme end of the sequence-space continuum. That said, there arelikely many individuals who fall in the middle of this continuum– some of whom might have synaesthesia (i.e., associators) and

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),

10 M. Jarick et al. / Brain and Cognition xxx (2011) xxx–xxx

others who might not. At the opposing end of the continuumwould presumably be individuals who have absolutely no spatialarrangement for sequences. Support for such a continuum comesfrom Brang, Teuscher, Ramachandran, and Coulson (2010) whoconducted a large-scale study of time–space synaesthesia in an at-tempt to meticulously characterize the phenomenon and they dis-covered that out of 183 ‘‘potential’’ synaesthetes, only 2.2% of thempassed their consistency test. Furthermore, those synaesthetes thatdid pass (‘‘verified’’ synaesthetes) were significantly less variable intheir spatial arrangements (three quarters were circular). However,Brang et al. (2010) noted that it was extremely difficult for somesynaesthetes to project their 3D representation onto a 2D platform,and that the month often encompassed a region of space and not aparticular point. This is a common limitation in studies of se-quence-space synaesthesia. Thus, although the authors estimatethat 2.2% of their sample could be classified as verified synaes-thetes, this estimate might be considered quite conservative.Nonetheless, their study indicates that those with vivid se-quence-space associations (at the extreme end of the continuum)are quite rare. In our case here, we have objectively verified L’s sub-jective reports of her extraordinary experience using two objectivetasks: SNARC-effect and spatial cueing tasks (Jarick, Dixon,Maxwell, et al., 2009a) and have run many consistency tests overthe years. Thus, we are certain that L is an extreme case thatwarrants this case-study investigation.

Here we extend our previous findings to show that L’s number-forms are not only highly consistent and strong enough to influ-ence her attention, but they occur extremely quickly and shifther attention rapidly. In addition, the influence of her number-forms on attention cannot be attributed to top-down processes,since even when attempting to use a strategy that dictates direct-ing attention to precisely the opposite spatial locations cued by hersynaesthesia, she still showed cueing-effects that are consistentwith her synaesthetic representations at the shortest SOA. Thisfast, involuntary directing of attention in spite of an opposing strat-egy bears the key attributes of automatic orienting. Even thoughthis is only one case, we provide empirical evidence to supportthe contention that automaticity occurs in at least one number-form synaesthete and could be a potential hallmark of number-form synaesthesia in general.

Acknowledgments

This research was supported by the Natural Sciences and Engi-neering Research Council of Canada (NSERC), with a research Grantto Mike Dixon and a postgraduate scholarship to Michelle Jarick.We sincerely thank the synaesthete (L), for which this researchwould not have been possible. We also thank Jody Woodcockand Alisha Maxfield for their help collecting the control data.

References

Brang, D., Teuscher, U., Ramachandran, V. S., & Coulson, S. (2010). Temporalsequences, synesthetic mappings, and cultural biases: The geography of time.Consciousness and Cognition, 19, 311–320.

Cheal, M., & Lyon, D. R. (1991). Central and peripheral precuing of forced-choicediscrimination. Quarterly Journal of Experimental Psychology, 43A, 859–880.

Crawford, J. R., & Garthwaite, P. H. (2005). Testing for suspected impairments anddissociations in single-case studies in neuropsychology: Evaluations ofalternatives using Monte Carlo simulations and revised tests for dissociations.Neuropsychology, 19, 318–331.

Cohen Kadosh, R., & Henik, A. (2007). Can synaesthesia inform cognitive science?Trends in Cognitive Science, 11, 177–184.

Cohen Kadosh, R., Tzelgov, J., & Henik, A. (2008). A colorful walk on the mentalnumber line: Striving for the right direction. Cognition, 106, 564–567.

Debner, J. A., & Jacoby, L. L. (1994). Unconscious perception: Attention, awareness,and control. Journal of Experimental Psychology: Learning, Memory, and Cognition,20, 304–317.

Dixon, M. J., Smilek, D., Cudahy, C., & Merikle, P. M. (2000). Five plus two equalsyellow. Nature, 406, 365.

Please cite this article in press as: Jarick, M., et al. 9 is Always on top: Assessingdoi:10.1016/j.bandc.2011.05.003

Dixon, M. J., Smilek, D., & Merikle, P. M. (2004). Not all synaesthetes are createdequal: Projector versus associator synaesthetes. Cognitive, Affective, andBehavioral Neuroscience, 4, 355-343.

Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental representation of parity andnumber magnitude. Journal of Experimental Psychology: General, 122, 371–396.

Fischer, M. H., Castel, A. D., Dodd, M. D., & Pratt, J. (2003). Perceiving numbers causesspatial shifts of attention. Nature Neuroscience, 6, 555–556.

Friesen, C. K., & Kingstone, A. (1998). The eyes have it! Reflexively orienting istriggered by nonpredictive gaze. Psychonomic Bulletin and Review, 5, 490–495.

Galton, F. (1880). Visualized numerals. Nature, 21, 252–256.Gheri, C., Chopping, S., & Morgan, M. J. (2008). Synaesthetic colours do not

camouflage form in visual search. Proceedings of Biological Science, 275, 841–846.Gertner, L., Henik, A., & Cohen Kadosh, R. (2009). When 9 is not on the right:

Implications from number-form synaesthesia. Consciousness and Cognition, 18,366–374.

Gevers, W., Lammertyn, J., Notebaert, W., Vertguts, T., & Fias, W. (2006). Automaticresponse activation of implicit spatial information: Evidence from the SNARCeffect. Acta Psychologica, 122, 221–233.

Hubbard, E. M., Ranzini, M., Piazza, M., & Dehaene, S. (2009). What information iscritical to elicit interference in number-form synesthesia. Cortex, 45,1200–1216.

Ito, Y., & Hatta, T. (2004). Spatial structure of quantitative representation ofnumbers: Evidence from the SNARC effect. Memory and Cognition, 32, 662–673.

Jacoby, L. L., Lindsay, D. S., & Toth, J. P. (1992). Unconscious influences revealed:Attention, awareness, and control. American Psychologist, 47, 802–809.

Jacoby, L. L., Toth, J. P., & Yonelinas, A. P. (1993). Separating conscious andunconscious influence of memory: Measuring recollection. Journal ofExperimental Psychology: General, 122, 139–154.

Jacoby, L. L., & Whitehouse, K. (1989). An illusion of memory: False recognitioninfluenced by unconscious perception. Journal of Experimental Psychology:General, 118, 126–135.

Jarick, M., Dixon, M. J., Maxwell, E. C., Nicholls, M. E. R., & Smilek, D. (2009a). The upsand downs (and lefts and rights) of synaesthetic number forms: Validation fromspatial cueing and SNARC-type tasks. Cortex, 45, 1190–1199.

Jarick, M., Dixon, M. J., Stewart, M. T., Maxwell, E. C., & Smilek, D. (2009b). Adifferent outlook on time: Visual and auditory month names elicit differentmental vantage points for a time-space synaesthete. Cortex, 45, 1217–1228.

Jonides, J. (1981). Voluntary versus automatic control over the mind’s eye’smovement. In J. Long & A. Baddeley (Eds.), Attention and performance IX(pp. 187–203). Hillside, NJ: Erlbaum.

Kuhn, G., & Kingstone, A. (2009). Look away! Eyes and arrows engage oculomotorresponses automatically. Attention, Perception, and Psychophysics, 71, 314–327.

Mattingley, J. B., Rich, A. N., Yelland, G., & Bradshaw (2001). Unconscious primingeliminates automatic binding of colour and alphanumeric form in synaesthesia.Nature, 410, 533–534.

Meier, B., & Rothen, N. (2009). Training grapheme-colour associations produces asynaesthetic Stroop effect, but not a conditioned synaesthetic response.Neuropsychologia, 47, 1208–1211.

Müller, H. J., & Rabbitt, P. M. A. (1989). Reflexive and voluntary orienting of visualattention: Time course of activation and resistance to interruption. Journal ofExperimental Psychology: Human Perception and Performance, 15, 315–330.

Piazza, M., Pinel, P., & Dehaene, S. (2006). Objective correlates of an usual subjectiveexperience: A single-case study of number-form synaesthesia. CognitiveNeuropsychology, 23, 1162–1173.

Posner, M. I. (1980). Orienting of attention. Quarterly Journal of ExperimentalPsychology, 32, 3–25.

Posner, M. I., & Synder, C. R. R. (1975). Attention and cognitive control. In R. L. Solso(Ed.), Information processing and cognition: The loyola symposium. LawrenceErlbaum.

Sagiv, N., Simner, J., Collins, J., Butterworth, B., & Ward, J. (2006). What is therelationship between synaesthesia and visuo-spatial number forms? Cognition,101, 114–128.

Santens, S., & Gevers, W. (2008). The SNARC effect does not imply the mentalnumber line. Cognition, 108, 263–270.

Schwartz, W., & Keus, I. M. (2004). Moving the eyes along the mental number line:Comparing SNARC effects with saccadic and manual responses. Perception andPsychophysics, 66, 651–664.

Seron, X., Pesenti, M., Noël, M., Deloche, G., & Cornet, J. (1992). Images of numbers,or ‘‘when 98 is upper left and sky blue’’. Cognition, 44, 159–196.

Simner, J., Mayo, N., & Spillar, M. (2009). A foundation for savantism? Visuo-spatialsynaesthetes present with cognitive benefits. Cortex, 45, 1246–1260.

Smilek, D., Callejas, A., Dixon, M. J., & Merikle, P. M. (2007). Ovals of time: Time-space associations in synaesthesia. Consciousness and Cognition, 16, 507–519.

Tang, J., Ward, J., & Butterworth, B. (2008). Number forms in the brain. Journal ofCognitive Neuroscience, 20, 1547–1556.

Vallesi, A., Shallice, T., & Walsh, V. (2007). Role of the prefrontal cortex in theforeperiod effect: TMS evidence for dual mechanisms in temporal preparation.Cerebral Cortex, 17, 466–474.

Van Selst, M., & Jolicoeur (1994). A solution to the effect of sample size on outlierelimination. The Quarterly Journal of Experimental Psychology Section A, 47,631–650.

Ward, J., Sagiv, N., & Butterworth, B. (2009). The impact of visuo-spatial numberforms on simple arithmetic. Cortex, 45, 1261–1265.

Warner, C. B., Juola, J. F., & Koshino, H. (1990). Voluntary allocation versusautomatic capture of visual attention. Perception and Psychophysics, 48,243–251.

the automaticity of synaesthetic number-forms. Brain and Cognition (2011),