Regional Specificity of Format-SpecificPriming Effects in Mirror Word ReadingUsing Functional Magnetic ResonanceImaging

Lee Ryan1 and David Schnyer2

1Cognition and Neuroimaging Laboratories, Department of

Psychology, University of Arizona, Tucson, AZ 85721-0068,

USA and 2Memory Disorders Research Center, Boston VA

Healthcare System and Boston University School of Medicine,

Boston, MA, USA.

The speed and accuracy with which subjects can read words isenhanced or ‘‘primed’’ by a prior presentation of the same words.Moreover, priming effects are generally larger when the physicalform of the words is maintained from the first to the secondpresentation. We investigated the neural basis of format-specificpriming in a mirror word-reading task using event-related functionalmagnetic resonance imaging (fMRI). Participants read words thatwere presented either in mirror-image (M) orientation or in normal(N) orientation and were repeated either in the same or thealternate orientation, creating 4 study--test conditions, N-N, M-N,N-M, and M-M. Priming of N words resulted in reductions infMRI signal in multiple brain regions, even though reading times(RTs) were unchanged. Priming of M words showed a pattern ofRTs consistent with format-specific priming, with greater reduc-tions when the prime matched the form of the test word. Priming-related reductions in fMRI activity were evident in all regionsinvolved in mirror-image reading, regardless of the orientationof the prime. Importantly, reductions in several posterior regions,including fusiform, superior parietal, and superior temporal regionswere also format specific. That is, signal reductions in these re-gions were greatest when the visual form of the prime and targetmatched (M-M compared with N-M). The results indicate that,although there are global neural priming effects due to stimulusrepetition, it is also possible to identify regional brain changes thatare sensitive to the specific perceptual overlap of primes andtargets.

Introduction

Previous experiments have indicated that the speed and

accuracy with which subjects identify words is enhanced

or ‘‘primed’’ by a prior presentation of the same words. More-

over, priming effects are generally larger when the physical

form of the word remains the same from the first to the second

presentation. This is true for study--test manipulations of sen-

sory modality (e.g., visual vs. auditory; Clarke and Morton

1983; Graf and others 1985; Roediger and Blaxton 1987;

Schacter and Graf 1989), symbolic form (pictures vs. words;

Weldon and Roediger 1987), script orientation (backwards or

upside down vs. normal; Kolers 1975, 1979; Masson 1986;

Graf and Ryan 1990), and sometimes type font or case (e.g.,

upper vs. lower case, handwritten vs. typed; Graf and Ryan

1990; Marsolek and others 1992, 1996; Gibson and others

1993; Curran and others 1996; Wiggs and Martin 1998; but see

Scarborough and others 1977; Tardiff and Craik 1989; Rajaram

and Roediger 1993, for failures to find type-specific priming).

Graf and Ryan (1990) have used a transfer appropriate pro-

cessing (TAP) framework (cf., Morris and others 1977) to

account for format-specific priming. TAP is based on the notion

that remembering is best understood in terms of the cognitive

operations that are engaged by different study and test activi-

ties (Kolers and Ostry 1974; Kolers 1975, 1979). Reading a word

or sentence, for example, requires a particular set of sensory-

perceptual and semantic-analyzing operations. Engaging these

operations has the same effect as practicing a skill—it increases

the fluency and efficiency with which they can be carried

out subsequently. Performance on a priming test is facilitated

to the extent that it engages the same set of cognitive oper-

ations as used on the preceding study task. The greater the

overlap in processes from study to test the greater the fa-

cilitation. Importantly, priming is also dependent upon how

practiced these processes are—cognitive operations that are

executed with a high degree of efficiency and skill will

show little, if any, priming, whereas uncommon or unskilled

operations will show greater facilitation after a single practice

episode (Graf and Ryan 1990; Ostergaard 1998, 1999).

Several accounts of priming offer hypotheses as to the brain

regions that may mediate these effects. One such account,

described by Tulving and Schacter (1990; Schacter 1994),

postulates that priming is mediated by changes in the per-

ceptual representation system (PRS). The PRS is described as a

collection of domain-specific modules or subsystems that

operate on perceptual information about the form and struc-

ture, but not the meaning and associative properties, of words

and objects. Schacter and others (e.g., Schacter 1994; Schacter

and Buckner 1998) argue that priming on most standard tasks

such as word identification or fragment completion primarily

reflects increased efficiency in the perceptual analysis of the

word and, by extension, will be mediated by posterior cortical

regions. Recent imaging studies using positron emission tomog-

raphy and functional magnetic resonance imaging (fMRI) are

consistent with this notion. Collectively, these studies have

demonstrated that priming is accompanied by blood flow

decreases in bilateral posterior perceptual processing areas in

extrastriate occipital cortex, using such tasks as word stem

completion, visual word-fragment completion, verb generation,

and object classification (Buckner and others 1995; Martin and

others 1995; Blaxton and others 1996). Interestingly, primary

visual regions have not shown subsequent priming-related

reductions in blood flow, suggesting that early visual areas as

defined by retinotopic organization may not be involved in

perceptual priming (Halgren and others 1997; Buckner and

others 1998).

It is also clear, however, that frontal regions, particularly

left prefrontal cortical areas, are sometimes affected by prim-

ing, depending upon the specific demands of the task. Priming-

related decreases in blood flow in left prefrontal regions have

been demonstrated in various tasks including verb generation

(Raichle and others 1994), abstract versus concrete classification

Cerebral Cortex

doi:10.1093/cercor/bhl009

� The Author 2006. Published by Oxford University Press. All rights reserved.

For permissions, please e-mail: journals.permissions@oxfordjournals.org

Cerebral Cortex Advance Access published June 5, 2006

of words (Demb and others 1995; Wagner and others 2000),

object picture classification (Wagner and others 1997; Buckner

and others 1998), and word stem completion (Buckner and

others 1995). One possible commonality across these tasks is

that they all require some sort of semantic elaboration or

conceptual analysis, and conceptually based priming may there-

fore be mediated by prefrontal regions, especially left inferior

prefrontal cortex.

In combination, neuroimaging findings are consistent with

a TAP view of priming, suggesting that brain regions showing

priming-related reductions in activity are specific to the cog-

nitive requirements of the task, such as perceptual priming

mediated by posterior cortical regions, with separate and

distinct anterior regions involved in conceptual priming.

Few imaging studies, however, have directly manipulated the

task demands in a within-subject design. In one such study,

Wagner and others (2000) focused on the role of anterior

and posterior left prefrontal regions in conceptual priming.

Participants initially classified a series of words in a perceptual

decision task (upper or lower case) or a conceptual decision

task (abstract or concrete). During subsequent scanning, par-

ticipants made abstract/concrete judgments for words from 3

sources; words that had been previously encountered during

the conceptual decision task, words that had been encountered

during the perceptual decision task, and words that had not

been encountered in either decision task. The results indicated

that the left anterior inferior frontal gyrus was activated by

the abstract/concrete decision task, and this activity was

not reduced as a consequence of prior exposure to the words

in the perceptual decision task. By contrast, activation was

significantly reduced when the exact same items and task

demands were repeated, suggesting that priming and related

blood flow reduction in this region depended crucially on the

overlap of semantic operations from study to test episodes.

The present study was designed to further illustrate the

generality of a TAP view of priming using fMRI. Rather than

varying the conceptual component of the task demands as

in Wagner and others (2000), we varied the overlap in per-

ceptual operations involved in identifying words presented

in 2 visually distinct formats, using a standard word-reading

paradigm. Although priming during word reading may be

mediated primarily by perceptual processes (Blaxton 1989),

according to TAP, both perceptual and semantic processing

should play at least some role in priming because both oper-

ations are automatically engaged when reading a word. The

degree to which conceptual and perceptual processes are

the basis for priming will vary depending on the difficulty of

the task and—critical to the present study—the overlap of

specific perceptual and semantic processes engaged at study

and test.

We presented participants with a continuous reading task,

in which a list of single words were first presented either

in normal orientation (N) or in mirror-image orientation (M)

and then were repeated later on in the list either in the same

orientation (N-N and M-M) or in different orientations (N-M

and M-N). In a similar behavioral paradigm, Graf and Ryan

(1990) showed that the reading accuracy with words that were

presented mirror image and upside down differed depending

on the orientation of the prime. Primes presented in the same

unusual orientation produced a greater increase in reading

accuracy than primes presented in normal upright orientation.

Graf and Ryan (1990) also provided evidence that orientation

manipulations of this type induced a letter-by-letter reading

strategy, rather than the usual whole word--reading approach

(for a discussion of word-reading strategies, see also Shallice

1988). In the case of reading M words, letters must be rotated

in order to identify them, and the normal direction of reading

is disrupted (words must be read from right to left rather than

left to right). When M words are primed with identical words

in M orientation, there is considerable overlap in perceptual

processes required in order to identify the unique visual form

of the word, as well as overlap in the semantic analysis of word

meaning. According to TAP, we would expect to see decreases

in activation in left prefrontal regions associated with process-

ing word meaning and in posterior regions including temporal,

parietal, and fusiform regions that are associated with the per-

ceptual analysis of visual form. In contrast, when M words are

primed with words that were presented previously in N orien-

tation, there should be considerably less repetition of percep-

tual processes carried out during reading. Hence, priming

should be mediated to a significantly lesser degree by posterior

regions involved in perceptual analysis and should rely instead

on frontal regions involved in semantic processing.

According to a strict TAP view of priming, the same general

hypothesis should hold for N words as well—priming should

result in greater reductions in posterior regions when the

prime and target are the same, as opposed to when the prime

and target differ in physical form. However, Graf and Ryan

(1990) found that primes presented in unusual orientations

produced greater priming overall than primes in normal orien-

tation, even when the target word was presented in normal

orientation. If activation reductions track the magnitude of

behavioral priming, we might expect to find greater reductions

in activity when N words are preceded by M primes compared

with N primes. It is unclear, however, what to expect in terms

of cortical regions mediating this enhanced priming effect.

Method

ParticipantsVolunteers included 12 right-handed individuals, aged 21--30 years, who

were paid $40 for their participation. Participants were screened to

exclude drug and/or alcohol abuse, neurological disorder and serious

head injury, psychiatric illness, and contraindications to undergoing

magnetic resonance imaging (MRI).

MaterialsFour word lists were created from 800 words, 5--8 letters in length, with

medium to high word frequency (20--300 occurrences per million;

Kucera and Francis 1967). Words were randomly assigned to 4 lists,

ensuring that lists were similar in distribution of word lengths and word

frequencies. Within a list, each word was repeated once with varying

lags of 8--18 intervening words. Thus, the first 8 words were novel, and

then words began to repeat in pseudorandom order, until all words

in the list were presented twice. First and second presentations of

words were randomly distributed throughout all 4 lists in an event-

related design; therefore, any skill learning that may contribute to mirror

reading is balanced across all conditions. On the first presentation, half

of the words were presented in N orientation, whereas half of the items

were presented in M orientation. On the second presentation, words

were presented either in the same orientation from first to second

presentation or the orientation was reversed from first to second

presentation, creating 4 study--test orientation conditions (N-N, M-N, M-

M, and N-M). Interspersed randomly throughout all 4 lists were 400

stimuli that provided a visual motor control (!+!+!+!+). Examples of the

stimuli are depicted in Figure 1. Thus, each list contained a total of 500

items; 100 control stimuli plus 50N-N, 50M-N, 50 N-M, and 50M-M trials.

Page 2 of 11 Format-Specific Priming in Mirror Word Reading d Ryan and Schnyer

ProcedureStimuli were presented in the scanner on magnetic resonance (MR)

Vision 2000 goggles (Resonance Technology Inc., Northridge, CA) that

were mounted to the head coil so that they rested comfortably over the

participant’s eyes. All stimuli were presented in lower case letters

centered on the screen in bright green 80-point font (MS Word Arial

Bold) on a black background. Each list item was visible for up to 2500

ms. Piloting in normal young participants indicated that these param-

eters resulted in an optimal level of mirror word reading (98% of items

were correctly read). Items were preceded by an arrow for 500 ms that

oriented participants to the location of the beginning of the word (see

Fig. 1). An orienting stimulus was added because pilot data indicated

that this substantially increased the number of words that could be

successfully read in M orientation. On visual motor control trials, the

orienting stimulus was also presented and appeared equally often on the

left and right. Presentation rate of the stimuli was self-paced. A button

response to an item advanced the presentation to the next list item. If

a participant could not identify a word within the 2500-ms time limit,

the next item appeared. Response timeswere collected for all trials using

a computer mouse modified for use in the scanner and placed in the

participant’s right hand. All 4 lists were separated by a 2- to 3-min break.

Prior to scanning, participants were given a short practice run that

included words in N orientation, M orientation, and control items. They

were instructed that they should read each word as quickly as possible,

pressing the mouse button in their hand when they successfully read

the word. They were informed that each word would be preceded by an

arrow located at the beginning of the word and that moving their eyes to

the arrow position would help them to read the words quickly and more

accurately. Participants pushed a single mouse button (right hand index

finger) when they successfully read a word in either orientation. When

the control item appeared, participants were instructed to press another

button (right hand ring finger) as quickly as possible. Participants were

told that some of the words would be difficult to read in the allotted time

and that they should go on to the next itemwhen it appeared. They were

also informed that sometimes the words may be repeated, but their

primary task was to read each word as quickly as possible.

Imaging Acquisition and AnalysisImages were acquired on a GE Horizon 1.5-T whole-body echo-speed

MRI system, using single-shot spiral acquisition (Glover and Lee 1995).

Images (17 sections, 5 mm, skip 1 mm) were collected obliquely and

aligned on the AC-PC plane covering approximately the whole head

(time repetition [TR] = 2000, time echo [TE] = 40, flip angle = 90). The

first functional scan was collected with a total of 498 repetitions taking

16 min and 30 s to complete. For subsequent scans, the repetitions were

adjusted downward depending on the reading speed of the particular

participant. Afterward, T1-weighted images were obtained (256 3 256,

TE = 10, TR = 500, field of view [FOV] = 22) using the same slice

selection as the functional data set, and a high resolution spoiled

gradient recalled (SPGR) series covering the whole brain (1.5-mm

sections, 256 3 256, Flip = 30, TE = 6000, TR = 22, FOV = 25 cm) were

also collected in order to locate anatomical regions of activation and to

overlay functional images for reregistration in standard Talairach and

Tournoux (1988) coordinate space.

Regions of Activation

Images were reconstructed offline and then corrected for minor head

movement using a 3-dimensional volume registration algorithm (AFNI,

Cox 1996). Data were normalized by scaling the whole brain signal

intensity to a fixed value of 1000 to allow data to be combined across

participants. Then the linear slope was removed on a voxel-by-voxel

basis, and spatial filtering was accomplished using a Hanning filter with

a 1.5-voxel radius. After normalization, detrending, and filtering, data

were analyzed using software developed and validated by Burock and

others (1998) for rapid presentation event-related fMRI designs (for

details of methods, see Dale and Buckner 1997; Dale 1999). Epochs of

16 s poststimulus onset and 6 s prestimulus baseline were modeled as

a linear combination of a time-invariant hemodynamic response (HDR)

with Gaussian noise. An estimate of the HDR and variance with the mean

signal intensity removed for each condition was modeled using si-

multaneous least squares fitting of the original MR signal across the

epochs. Only words that were successfully read by the individual were

included in the event-related analysis.

Regions of activation associated with reading were identified in each

individual by contrasting all M words with the control condition and,

separately, all N words with the control condition, using a t-statistic

weighted for an ideal HDR modeled as a gamma function with 2.25 s

onset time and a tau of 1.25 s (Dale and Buckner 1997). Maps of active

voxels were created by calculating the covariance between the

estimated signal response and the ideal HDR function. The statistical

threshold was determined using AlphaSim, which estimates the statis-

tical power for fMRI data by Monte Carlo simulation (AFNI, Cox 1996).

This calculation indicated that a 3-voxel cluster size using an individual

voxel threshold of P < 0.0001 would result in statistical maps with

a corrected significance level of P < 0.05 (corrected for multiple voxel

comparisons).

HDR Differences

Image sets were translated into Talairach coordinates (Talairach and

Tournoux 1988) using AFNI software. Statistical maps from individuals

were then overlaid on an averaged brain image to create a group image,

indicating the overlap of significantly active voxels across participants.

This group image was then used to determine common regions of

interest (ROI) for further analysis. Time series data for all active voxels

within each ROI were imported into SPSS for statistical analysis. For each

individual, average HDR estimates (0--16 s poststimulus onset) were

calculated for each ROI by averaging all significantly active voxels within

that region. Because we were interested in the amplitude of the

activation, all analyses were conducted on the mean HDR amplitude

of signal, which was averaged across time points 2, 4, 6, and 8 s

poststimulus onset, and baselined to an average of 4 prestimulus onset

Figure 1. Example of the displays used to present normal and mirror-image text.Screens with arrows were visible for 500 ms and alerted the participant to theorientation of the next word. Text screens were visible until the participant presseda key or for 2500 ms if they were unable to identify the word. Randomly interspersedwith text screens were visual spatial control items (!+!+!+!+) which were precededequally often by arrows on the right and left side of the screen.

Cerebral Cortex Page 3 of 11

points (–6, –4, –2, and 0 s). (To ensure that prestimulus baseline

differences could not account for differences in activation amplitudes

for the critical test conditions, we compared the averaged baseline

values at –6, –4, –2, and 0 s using a 1-way analysis of variance [ANOVA].

Baseline values did not differ between M words, N words, and the

control condition [F < 1].)

Results

Behavioral Results

Mean reading times (RTs) were calculated for 6 groups of

words: first presentation of N words (N baseline), first presen-

tation of M words (M baseline), and the second presentation

of words in the 4 priming conditions (N-N, M-N, M-M, and N-M).

Mean RTs, listed in Table 1, were calculated for each individual

excluding words for which they did not make a response

(approximately 2% of M words, less than 0.5% of N words). Not

surprisingly, on first presentation, M words took significantly

longer to read than N words, paired t11 = 7.68, P < 0.0001.

Priming was assessed for N and M words separately using 2

repeated-measures ANOVAs comparing the 3 conditions (base-

line, N primed, and M primed). For words presented in N

orientation, there was no measurable priming effect; N baseline,

N-N, and M-N conditions did not differ, F < 1. For words pre-

sented in M orientation, a previous presentation of a word

significantly decreased RTs, F1,11 = 86.77, mean square error

(MSe) = 239.34, P < 0.0001. Follow-up paired t-tests indicated

significant priming for both N-M and M-M conditions compared

with baseline, t11 values = 9.31 and 8.53, respectively, P value <

0.0001. Importantly, there was also a difference in reading speed

between N-M and M-M conditions. M-M words were read

significantly faster than N-M words, t11 = 4.66, P < 0.001,

indicating a format-specific priming effect.

Regions Involved in N and M Reading

Regions showing significant activation during N and M reading

compared with the control condition are listed in Table 2. There

was considerable overlap in active regions associated with

reading N and M words. Areas of activation were present in

the M condition (right superior temporal, right middle frontal,

and right fusiform gyri) that showed no activation during N

reading. In addition, the mean HDR was significantly greater

when reading M words compared with N words in bilateral

superior parietal and left superior temporal regions.

Activation Reductions Associated with Primingof M Words

ROIs were analyzed separately using a repeated-measures

ANOVA comparing the mean HDR amplitude in 3 conditions

(M baseline, N-M, and M-M) in order to determine whether

there was evidence for priming-related reductions in signal

amplitude. Participant was treated as a random factor in all

analyses. Significant regions were followed up with paired

t-tests comparing baseline with N-primed words and baseline

with M-primed words. Alpha for follow-up paired t-tests was

P < 0.05. Table 2 lists the results of these analyses identify-

ing regions that showed significant decreases in activation

in the 2 priming conditions. Among those ROIs identified as

active during the mirror-reading task, there were no repetition

related increases revealed. For words presented in M orienta-

tion, all brain regions that were active during the first pre-

sentation of M words showed significant decreases in activation

after repetition, regardless of the orientation of the prime.

In order to assess the effect of format-specific priming, HDR

‘‘priming’’ scores were calculated for each region by subtract-

ing the mean HDR amplitude at each time point during the

Table 1Mean RTs and standard deviations (SDs) for words in 2 baseline orientations,

mirror image (M) and normal (N), and 4 priming conditions (N-N, M-N, M-M, N-M)

N orientation M orientation

Mean SD Mean SD

Baseline 457.67 59.59 Baseline 841.17 203.95N-N 450.92 60.52 M-M 727.15 169.44M-N 454.00 61.68 N-M 782.33 200.93

Table 2Regions active during the 2 reading conditions, mirror-image (M) and normal (N) orientations, compared with the visual control condition, with mean cluster size (and standard deviations [SDs]),

Talairach and Tournoux (1988) coordinates, and corresponding Brodmann areas (BAs)

Anatomical region Mean cluster sizein voxels (SD)

Regions active during readingM and N words compared withcontrol conditionb

Talairach coordinatesa

and BAsRegions showing significantreductions in activation levelcompared with the appropriatebaseline condition (N or M)

Format-specificprimingb

M N M[ N x y z BA N-N M-N N-M M-M M-M[ N-M

L occipital/temporal and fusiform 184 (39) Yes Yes — �41 �7 �1 19 — — \0.0001 \0.0001 YesL superior temporal 28 (5) Yes Yes Yes �52 �35 12 22 \0.01 \0.01 \0.0001 \0.001 —L superior parietal 129 (39) Yes Yes Yes �30 �62 47 40/7 — — \0.01 \0.01 YesL anterior inferior frontal 36 (16) Yes Yes — �36 33 1 45/47 \0.01 — \0.05 \0.05 —L posterior prefrontal 88 (22) Yes Yes — �40 10 27 44/6 — — \0.01 \0.0001 —R occipital/temporal and fusiform 138 (34) Yes — Yes 41 �68 �1 19 — — \0.001 \0.001 YesR superior temporal 35 (11) Yes — Yes 57 �37 12 22 — — \0.05 \0.05 YesR superior parietal 92 (32) Yes Yes Yes 22 �65 52 40/7 \0.01 \0.05 \0.01 \0.01 YesR anterior inferior frontal 32 (18) Yes Yes — 35 27 0 45/47 \0.05 — \0.05 \0.01 —R middle frontal 40 (13) Yes — Yes 40 27 20 45/46 — — \0.01 \0.01 —R superior frontal 82 (39) Yes Yes — 36 27 39 8 — — \0.01 \0.01 —

Note: Also listed are those regions that showed significantly greater activation during M reading compared with N reading and the P values for regions showing significant signal reductions in the 4

priming conditions. Finally, format-specific priming regions listed are those in which signal reductions were greater when the prime and target were visually identical (M-M) than when they differed

(N-M).aCoordinates are expressed in millimeters in the Talairach and Tournoux brain atlas (1988): x, Medial--lateral axis (negative is left); y, anterior--posterior axis (negative, posterior); z, dorsal--ventral

axis (negative, ventral). BAs correspond to those defined in the Talairach and Tournoux brain atlas.bRefer to Results for specifics of analyses and P values.

Page 4 of 11 Format-Specific Priming in Mirror Word Reading d Ryan and Schnyer

second presentation from the corresponding amplitude during

the first presentation. Priming activation scores for fusiform,

superior temporal, superior parietal, and anterior inferior

frontal regions were then compared using a repeated-measures

ANOVA with 2 factors, priming condition (M-M, N-M) and he-

misphere (Left, Right). The right middle frontal, right superior

frontal, and left posterior prefrontal regions were analyzed

using a paired t-test (M-M vs. N-M) because no homologous

contralateral hemisphere activation was observed. Regions

showing a format-specific effect should show significantly larger

priming effects (reductions in activation) in the M-M condition

compared with the N-M condition.

Frontal regions (right superior, right middle, bilateral ante-

rior inferior frontal gyri, and left posterior prefrontal cortex)

failed to show a format-specific priming effect. Although all

these regions showed significant priming, the magnitude of the

priming did not differ for primes presented in N or M orien-

tation, and priming was similar in right and left hemispheres

(all P values > 0.05). However, 3 regions (fusiform gyrus,

superior parietal cortex, and superior temporal lobes) showed

format-specific reductions in activity. These regions are de-

picted in Figure 2. The left and right fusiform gyri showed

a significant effect of prime type, F1,11 = 21.29, MSe = 0.106,

P < 0.001, with a larger decrease in activation for M-M words

compared with N-M words. There was no difference in priming

in left and right hemispheres (F < 1), and prime type did not

interact with hemisphere (F < 1). Left and right parietal regions

also showed a significant effect of prime type, with M-M de-

creases being larger than N-M decreases, F1,9 = 6.41, MSe = 0.44,

P < 0.05. Again, priming did not differ by hemisphere, F1,9 = 1.38,not significant (NS), and there was no interaction between

prime type and hemisphere, F1,9 = 1.32, NS. Finally, the superior

temporal lobe showed a significant difference between M-M

and N-M conditions, but in contrast to the bilateral effects in

fusiform and parietal cortices, format-specific priming was

found only in the right hemisphere. ANOVA indicated a signif-

icant interaction between prime type and hemisphere, F1,8 =9.94, P < 0.05. Follow-up paired t-tests showed a significant

difference between the 2 prime types in the right hemisphere

but not in the homologous left hemisphere region. The main

effects of prime type and hemisphere did not approach

significance, F value < 1.

Figure 2. Examples of regions showing mean amplitude reductions in the HDR due to repetition of words. y axis refers to the normalized signal with mean removed, averagedacross all time points from 2 s postonset of the stimulus to 16 s postonset. Graphs show the mean amplitude response, expressed in percent signal change, for the first presentationof words in mirror-image (M) orientation and the mean amplitude responses for 2 priming conditions. M words were either primed with words in the same orientation (M-M) orprimed with words in normal orientation (N-M). The right fusiform, left fusiform, and right superior temporal gyri show a pattern of amplitude reduction consistent with format-specific priming; reduction in the M-M priming condition is significantly greater than the reduction observed in the N-M condition. Other regions showing this pattern but notdepicted here were the right and left superior parietal cortices. In contrast, left posterior prefrontal cortex shows similar reductions for both priming conditions compared with thebaseline reading condition. This pattern was true for all frontal regions and for the left superior temporal gyrus.

Cerebral Cortex Page 5 of 11

Ruling Out Time-On-Screen Effects in Format-SpecificPriming

It is important to address one possible confound to the results

described above. Because word-reading trials were self-paced, it

is possible that differences in signal amplitude were simply

reflecting the differences in the time each word remained on

the screen and that posterior brain regions are more sensitive

to visual presentation times than anterior regions. Previous

research has demonstrated that visual presentation time can be

an important variable in determining the level of neuroimaging

signal in visual cortex. However, it is unclear if presentation

time is critical for item differences of 100 ms or less when the

items are presented for as long as 1000 ms (Maccotta and others

2001). Nevertheless, we explored this possibility by identifying

the 4 subjects with the smallest M-M/N-M differences in RTs

(average difference 36 ms). RTs for these subjects were further

matched by discarding the longest items in the mirror-reading

condition, so that normal and mirror RTs were equated within

±5 ms. Signal amplitude for each condition was then compared

with 4 other subjects with the greatest M-M/N-M differences

in RTs (average difference 92 ms). Hemodynamic estimates

for each subgroup were recalculated and were averaged across

the 5 posterior regions showing format-specific priming.

Results are displayed in Figure 3. The results of the analysis

are inconsistent with the notion that the critical M-M/N-M

difference is due merely to differences in viewing time. In fact,

the results show that the format-specific effect was numerically

greater after response times were matched in the 2 critical

conditions.

Activation Reductions Associated with Priming of NWords

In spite of the fact that we observed no appreciable RT prim-

ing for N words, we analyzed the HDR responses in regions

involved in N reading similarly to the analyses described above,

in order to determine whether or not activation reductions

might provide a more sensitive measure of stimulus repetition

than RTs. Results of these analyses are also summarized in Table

2. Several regions showed significant decreases in HDR ampli-

tude corresponding to repeated items. Regions that showed

significant reductions in both priming conditions (N baseline

compared with N-N, and N baseline compared with M-N)

included the left superior temporal gyrus and the right parietal

region. In these 2 regions, the magnitude of the priming

effect was similar regardless of the orientation of the prime.

Two other regions showed significant priming, but only in the

N-N condition; these included the left anterior inferior frontal

gyrus and the right anterior inferior frontal gyrus.

Discussion

To summarize, the purpose of the present study was to de-

monstrate that format-specific priming is mediated by regions

specifically involved in processing aspects of perceptual fea-

tures, consistent with a TAP account of priming. In contrast

to our predictions, we found that priming for mirror-image

reading occurred in all regions associated with the baseline

task, regardless of the orientation of the prime, M or N (for

a similar result, see Poldrack and Gabrieli 2001). Importantly,

however, the priming-related reduction in activation was

greater in some brain regions when the visual form of the

prime and target matched (M-M condition), compared with

when they did not match (N-M condition). Although not

specific to the format, those regions showing format sensitivity

in the pattern of activation reduction have been implicated in

perceptual processing of word form and include bilateral

fusiform and right superior temporal gyri and bilateral superior

parietal cortex. The results are discussed in detail below in the

context of several theoretical views of priming.

Before turning to the results, the issue of explicit contami-

nation should be addressed briefly. Explicit contamination

(participants recognizing some or all of the target items as

repeated) is an issue in almost all priming studies. We tried to

decrease this likelihood by providing instructions to the par-

ticipants that emphasized the importance of reading the words

aloud, as quickly as possible. We also informed participants that

some of the words will be repeated during the long list, but they

should ignore this and focus on reading the words quickly.

Nevertheless, it is still possible that at least some of the words

in the present study were consciously recognized by subjects, at

least after gaining access to the meaning of the word through

reading. Although the analyses comparing targets with primes

showed only decreases in activation across all brain regions, it

is premature to assume that fMRI deactivations are synonymous

with priming (for discussion, see Henson and Rugg 2003).

Because we did not assess explicit recognition for the target

words, the present study cannot rule out the possibility that

processes mediating recognition are mixed, at least to some

Format Specific Priming

0

0.05

0.1

0.15

0.2

0.25

Large Normal/Mirror

Difference

Matched Normal/Mirror

Difference

Percen

t S

ig

nal D

ifferen

ce (M

irro

r - N

orm

al)

Figure 3. The graph shows the percent signal change for format-specific priming,comparing 2 subgroups of subjects: subjects who had a large difference in RTs betweenM-M and N-M conditions (average 92 ms) and subjects for whom RTs were matched(within ± 5 ms) between M-M and N-M RTs. Signal differences between M-M and M-Nconditions are collapsed across all 5 ROIs that exhibited format-specific priming.

Page 6 of 11 Format-Specific Priming in Mirror Word Reading d Ryan and Schnyer

degree, with priming effects nor can these influences be

separated definitively.

Reading Mirror-Image Text

Compared with N words, reading M words resulted in increased

activation in several regions, primarily in the right hemisphere,

and recruited additional brain regions that were not evident in

normal reading, again in the right hemisphere. These included

posterior regions that demonstrate robust activity and repeti-

tion effects in studies involving object recognition and priming,

including bilateral lateral superior parietal lobe and fusiform

gyrus (Wagner and others 1997; Buckner and others 1998;

Koutstaal and others 2001). Fusiform cortex has also shown

increased activation when participants made subordinate judg-

ments regarding an object (is this a sparrow?) relative to su-

perordinate judgments (is this a bird?), possibly reflecting the

additional perceptual processing required to arrive at a more

fine-grained classification (Gauthier and others 1997). Reading

M words also resulted in increased activity in right hemisphere

regions, including the right superior temporal gyrus, right ante-

rior inferior frontal gyrus, and right middle frontal gyrus. The

increased activation in right anterior hemisphere regions may

be in response to the increased difficulty of reading M words

compared with N words, resulting in additional recruitment

of regions involved in attention, working memory, and task

monitoring. One possible confound contributing to the findings

of additional regions involved in mirror reading when compared

with normal reading is the fact that mirror words remained

on the screen longer than normally oriented words. Although

this may contribute to our findings, the presence of additional

right hemisphere regions for mirror reading is consistent with

at least 2 previous studies (Poldrack and others 1998; Poldrack

and Gabrieli 2001). Interestingly, these studies also demon-

strated that with practice (skill learning), many of these addi-

tional right hemisphere regions dropped away. An obvious

limitation of the present study is that, without varying further

aspects of the reading task in systematic ways, it is not possible

to determine the specific contribution of each of the regions

to mirror-image reading.

Priming Effects in Normal and Mirror-Image Reading

Priming, evident both in RTs and in the extent of priming-

related reduction in brain activity, was greater overall for words

tested in M orientation compared with N orientation. This is

consistent with Graf and Ryan (1990; see also Ostergaard 1998,

1999), who showed that tasks that are less practiced benefit

most from a single prior presentation. Furthermore, a previous

study of repetition priming for mirror presented words found

that although the reductions with repetition were widespread

across a brain network involved in mirror reading, the extent

of this priming effect was reduced after considerable training

in mirror reading (Poldrack and Gabrieli 2001). For our pur-

poses, using a task like mirror-image reading without extensive

practice slowed RTs and increased overall priming, thereby

increasing the sensitivity to more subtle manipulations, in this

case, the similarity between the visual form of the prime and

target.

For M words, there was a significant reduction in RT follow-

ing the presentation of primes in both M and N orientations,

and primes, regardless of their orientation, resulted in a re-

duction in activation in all brain regions that were evident in

M reading at baseline. In contrast, when words were tested in

N orientation, there was no measurable effect of a prior pre-

sentation on RTs. Reading words is a highly practiced skill, and

when the primes are presented with longer lag times (between

8 and 18 intervening words in the present experiment), it is

unlikely to produce a decrease in RT. Most other studies using

word identification present target words in some degraded

form (Ostergaard 1998, 1999), or at a very fast presentation rate

(Masson 1986; Graf and Ryan 1990), or they require the par-

ticipant to make a decision about the stimulus being presented

(Balota and Chumbley 1984), thereby increasing the complexity

of the task and, hence, the priming effect. Nevertheless, despite

the lack of a behavioral priming effect for N words, we found

significant fMRI signal reductions after repeated trials in sev-

eral regions involved in reading N words, including the left

superior temporal gyrus, right superior parietal cortex, and

anterior inferior frontal gyri bilaterally. The finding that re-

ductions occurred in some, but not other, brain regions under

these conditions may be important, but it may also reflect a

lack of sensitivity in this simple reading task. It would be

premature to interpret such a finding without replication

and without a priori predictions as to why some brain regions,

and not others, are affected by priming under these conditions.

However, the results indicate that activation reductions may

provide a more sensitive priming measure than RTs for some

highly practiced tasks such as normal reading, and this warrants

further investigation.

Format-Specific Priming Effects

In the mirror-image reading task, the RT data indicated a

clear pattern of format-specific priming. RTs decreased an

average of 14% for M words that were preceded by a prime

in the same (M) orientation, compared with a reduction of 7%

for M words that were preceded by a prime in the alternate

(N) orientation. Format-specific effects were evident not only

in RTs, but in the pattern of brain activity as well. Although

all regions associated with mirror-image reading showed re-

duced activity after repeated trials in both priming conditions

(M-M and N-M), in some brain regions, reductions were sig-

nificantly greater when the orientation of the prime and target

matched. There are 2 important points regarding the regions

in which format-specific activation reductions were observed.

First, the magnitude of priming-related reductions in left

and right frontal regions did not differ between the 2 prime

types. This result is consistent with previous studies that

suggest that anterior frontal regions mediate nonperceptual

processing such as the identification and semantic analysis of

format-invariant aspects of words. For example, left anterior

inferior frontal cortex has been implicated in access to and

evaluation of long-term semantic knowledge (Petersen and

others 1988; Demb and others 1995). Alternatively, it may fun-

ction as the neural substrate for executive control processes

involved in working with semantic knowledge (Kapur and

others 1994; Gabrieli and others 1998; Poldrack and others

1999; for an alternative view, see Thompson-Schill and others

1998, 1999). In contrast, left posterior prefrontal cortex is

reliably activated during tasks that require passive viewing of

words, reading words aloud, or making lexical decisions about

words and pseudowords (reviewed in Poldrack and others

1999). Based on imaging and neuropsychological findings

(Frost 1998), this region may contribute to the transformation

of lexical information into phonological codes. However, al-

though one study (Wagner and others 2000) demonstrated that

Cerebral Cortex Page 7 of 11

only the anterior left prefrontal region benefited from prior

conceptual processing, whereas the posterior region benefited

from prior perceptual processing, a number of studies have

found that repetition effects produce qualitatively similar

effects in both regions (Raichle and others 1994; Buckner and

others 1998, 2000). In addition, there is at least one study that

provides evidence to suggest that left prefrontal regions are

format invariant. Buckner and others (2000) found similar

priming-related reductions in activity in the left inferior frontal

gyrus during visual-to-visual and auditory-to-auditory priming

conditions using a word stem completion task, suggesting that

the region is not modality specific. Activity in these regions

should therefore be primed by the prior presentation of a word,

regardless of the perceptual form of the word.

Second, posterior regions (bilateral fusiform, bilateral supe-

rior parietal, and right superior temporal cortices) showed

format-specific HDR reductions in response to priming. The

signal reduction in response to primes was greatest when the

prime was presented in the same visual format as the target.

Koutstaal and others (2001) obtained similar results in a recent

study of priming using pictures in a size judgment task. They

found that both anterior and posterior brain regions showed

decreases in fMRI activation for repeated versus novel pictures

of common objects. Additionally, when the object was primed

with the exact same picture, several posterior cortical regions

showed greater activation reductions compared with a condi-

tion where the object was primed with a similar but different

picture (such as 2 different umbrellas or 2 different coffee

mugs). These regions included the right precuneus and bilateral

fusiform, parahippocampal, occipital, and superior parietal

cortices. Regions sensitive to the format-specific priming in

both these studies have been shown to participate in various

aspects of visual form analysis of words and objects.

Theoretical Accounts of Format-Specific Priming

There is general agreement that repetition priming results in

regional fMRI signal reduction. A small number of studies have

now added an important qualification to this general finding

by demonstrating that the magnitude of the signal reduction is

at least partially dependent upon the degree of similarity be-

tween the cognitive processes engaged at study and at test

(for review, see Schacter and others 2004). For example, in

the present study and in Koutstaal and others (2001), when

the overlap of perceptual processes from study to test was

increased by making the target and prime identical, priming

increased behaviorally and was accompanied by greater reduc-

tions in activity in posterior cortical regions. These results are

at least partially consistent with a PRS system view of priming

described by Schacter (1994). However, the fact that anterior

regions of the brain also show priming-related signal reductions

in a simple visual task such as word identification is not pre-

dicted by a strict PRS view. Priming-related signal reductions

occur throughout the brain, and it is insufficient to posit that

priming on any task is mediated solely by increased efficiency

in the perceptual analysis of the stimuli. However, to the extent

that priming is enhanced by the overlap of specific perceptual

features, this aspect of priming will be mediated by PRS cortical

regions involved in perceptual analysis of the form and structure

of stimuli. In contrast to the PRS system, TAP provides a more

general framework that can account for a broader range of

region-specific priming effects, whether they occur in posterior

regions that mediate perceptual processing of visual form or

in frontal regions mediating conceptual and semantic processes

involved in categorization (as in Wagner and others 2000).

Depending upon the specific operations required for each

task and the overlap across 2 tasks, the specific regions showing

format-specific priming may differ significantly from our cur-

rent results, but the general framework of TAP will apply.

It is important to note, however, that the pattern of signal

reductions observed here cannot be explained fully by a purely

TAP account of priming. According to a strong TAP view, pri-

ming is presumed to occur because of the overlap in processes

engaged at study and test, and therefore only those processes

that are utilized during the study episode are capable of medi-

ating priming. By this view, we expected that N oriented primes

would provide little, if any, facilitation for the perceptual anal-

ysis of mirror-image words because there is no requirement

in normal reading to engage in letter rotation and the letter-

by-letter reading strategy utilized when reading words in this

unusual orientation. However, when M words were primed

with N words, all regions active during mirror-image reading

showed robust signal reductions, even in those regions that

were not activated during normal reading including right

middle frontal gyrus, right superior temporal gyrus, and right

fusiform gyrus. Dale and colleagues (Dale and others 2000) de-

monstrated that when words are repeated in a size judgment

task, priming effects are widespread across the entire cortical

network involved in word processing. In addition, they reported

that electrocortical repetition effects did not begin until well

after the initial perceptual analysis of a stimulus was complete

(200--260 ms poststimulus onset). How does one account

for such global priming effects? The results of Dale and others

(2000), combined with our own, suggest that there are 2

distinct and perhaps interactive bases for priming. Priming

may be mediated either by ‘‘top down’’ or ‘‘bottom-up’’ pro-

cessing of words, depending upon the specific circumstances

of the test. We suggest that a recent presentation of a word in

any form that fully engages the semantic representation of

that word will prime all aspects of a repeated presentation,

semantic and perceptual. Increasing the efficiency of semantic

processing will serve to disambiguate the perceptual processing

of the unfamiliar item as analysis progresses, and this effect

will be particularly evident when the task is unfamiliar or dif-

ficult. Recent exposure to the identical perceptual processes

required for the repeated item will provide an added benefit,

but only to specific processes involved in perceptual analysis.

Priming, then, may be driven by both semantic and perceptual

processes. Importantly, perceptual processes can benefit from

a recent exposure to semantic information, without exposure

to the specific perceptual processes involved in analyzing

the visual form.

One prediction that derives from the above discussion is

that the global priming effects we observe might only occur

under circumstances where the participant successfully acce-

sses the meaning of the word, regardless of howmuch time they

spend analyzing the visual form. On trials where participants

are unsuccessful in identifying the word, no semantic level

priming should occur. We were unable to assess this possibility

because we specifically designed the study to minimize the

number of words that were unsuccessfully read on the first

presentation. However, the issue warrants further investigation.

An alternative explanation, and also consistent with the pre-

sent pattern of results, is that there is a difference between the

processes that underlie behavioral priming such as enhanced

Page 8 of 11 Format-Specific Priming in Mirror Word Reading d Ryan and Schnyer

RTs and accuracy and those that alter the HDR; some

neuroimaging signal changes may be directly related to

behavioral priming, and some may reflect postprocessing

modulations (Dobbins and others 2004). Moreover, effects at

this level may have been evident in our normal reading results

where, despite the lack of behavioral priming, several brain

regions exhibited signal reductions. Tulving and others

(Tulving and others 1996) have suggested that reductions in

neural activity associated with repetition may be the result of

postidentification attentional modulation associated with the

detection of nonnovel stimuli. By this account, once perceptual

processing is completed, the stimulus is assessed for novelty

and then attentional resources are allocated for further analysis

depending on the novelty of the item (i.e., more resources are

allocated for novel items and less for nonnovel items). The

determination that an item is novel must be context dependent;

although frequently encountered words are not particularly

novel on their own, in the context of a specific episode (the

priming experiment), words that are repeated will be deemed

less novel than nonrepeated words. The degree of similarity

between semantic and perceptual processes across items

influences the degree of adjustment to signal levels throughout

the currently activated network. Hence, one might expect

greater reductions in allocation of resources when the visual

form of the word is maintained across repeated presentations.

However, it is difficult to explain why such format-specific

priming effects should only occur within certain brain regions

and not others. The notion that novelty assessment occurs after

perceptual analysis is complete is consistent with the findings of

late-onset brain changes associated with word identification

priming (Dale and others 2000) and may also provide a mech-

anism that accounts for the nonspecific or global priming ef-

fect observed in this and similar studies. At present, however,

Tulving’s hypothesis is not sufficiently elaborated to account

for region-specific priming effects unless one postulates an in-

teraction between global and process-specific priming effects.

Finally, the finding of format-specific signal reductions in

right, but not left, superior temporal lobe is consistent with the

suggestion that the right hemisphere is engaged in the specific

form analysis of a word or object. A further refinement of the

PRS account of format-specific priming is provided by Marsolek

(Marsolek and others 1992, 1994, 1996). Marsolek suggests that

the right and left hemispheres preferentially analyze specific

and abstract properties of the visual form, respectively. The

abstract visual-form subsystem, associated with posterior left

hemisphere regions, supports analysis of abstract categories of

forms by processing visual features that are relatively invariant

across different instances. In contrast, the specific visual-form

(SVF) subsystem, associated with posterior right hemisphere

regions, processes distinctive aspects of visual form that would

distinguish a specific instance. In word stem completion exper-

iments, for example (Marsolek and others 1992, 1994), greater

priming was obtained when prime and target word stems

appeared in the same letter case than in different letter cases.

More importantly, this effect was greater when stems were

presented directly to the right hemisphere via the left visual

field than when stems were presented directly to the left

hemisphere via the right visual field (see also Vaidya and others

1998).

The specific regions of right posterior cortex mediating

these effects may depend on the type of material being tested.

In their picture-priming experiment, Koutstaal and others

(2001) found one region, the right fusiform gyrus, that showed

significantly greater format-specific reduction in activation

compared with the homologous left hemisphere region. Al-

though format-specific reductions were also found in other

posterior regions, these effects were bilateral. Using words

as the critical stimuli, we found a format-specific activation

reduction in the right, but not left, superior temporal gyrus,

consistent with a recent study using a masked priming para-

digm (Dehaene and others 2004). In contrast, both left and

right fusiform gyri showed similar degrees of format specificity.

Previous research has suggested that visual word forms are

processed in regions of fusiform gyrus (Rumsey and others

1997), whereas superior temporal gyrus may be more involved

in processing of phonological representations (Majerus and

others 2005). Our finding of changes in the right superior

temporal lobe suggest that participants utilized phonology as

a strategy for identifying mirror orientation words, essentially

‘‘sounding out’’ the words as they read (see also Ryan and others

2001). The results of the 2 studies suggest that the processing of

SVFs of stimuli, such as words and pictures, may be dependent

on the processing approach taken to identify them, and these

approaches are most likely mediated by different brain regions.

These findings also warrant further investigation.

In summary, we found that the degree of priming-related

signal reduction in brain regions depends at least partially on

the degree of overlap of cognitive processes engaged at study

and again at test. However, the complete pattern of signal re-

ductions obtained in the present study cannot fully be

accounted for by existing views of priming. The results raise

interesting and as yet unresolved questions regarding the

mechanisms underlying the priming effect. Functional imaging

methods clearly provide an important new tool for understand-

ing this phenomenon.

Notes

Conflict of Interest: None declared.

Address correspondence to Lee Ryan, PhD, Cognition and Neuro-

imaging Laboratories, Department of Psychology, P.O. Box 210068,

University of Arizona, Tucson, AZ 85721-0068, USA. Email: ryant@

u.arizona.edu.

References

Balota DA, Chumbley JI. 1984. Are lexical decisions a good measure of

lexical access? The role of word frequency in the neglected decision

stage. J Exp Psychol Hum Percept Perform 10:340--357.

Blaxton TA. 1989. Investigating dissociations among memory measures:

support for a transfer-appropriate processing framework. J Exp

Psychol Learn Mem Cogn 15:657--668.

Blaxton TA, Bookheimer SY, Zeffiro TA, Figlozzi CM, Gaillard WD,

Theodore WH. 1996. Functional mapping of human memory using

PET: comparisons of conceptual and perceptual tasks. Can J Exp

Psychol 50:42--56.

Buckner RL, Goodman J, Burock M, Rotte M, Koutstaal W, Schacter DL,

Rosen B, Dale AM. 1998. Functional-anatomical correlates of object

priming in humans revealed by rapid presentation event-related

fMRI. Neuron 20:285--196.

Buckner RL, Koutstaal W, Schacter DL, Rosen BR. 2000. fMRI evidence

for a role of frontal and inferior temporal cortex in conceptual

priming. Brain 123:620--640.

Buckner RL, Petersen SE, Ojemann JG, Miezin FM, Squire LR, Raichle ME.

1995. Functional anatomical studies of explicit and implicit memory

retrieval tasks. J Neurosci 15:12--29.

Burock MA, Buckner RL, Woldorff MG, Rosen BR, Dale AM. 1998.

Randomized event-related experimental design allows for extremely

Cerebral Cortex Page 9 of 11

rapid presentation rates using functional MRI. Neuroreport

9:3735--3739.

Clarke R, Morton J. 1983. Cross modality facilitation in tachistoscopic

word recognition. Q J Exp Psychol 104:268--294.

Cox RW. 1996. AFNI: software for analysis and visualization of functional

magnetic resonance neuroimages. Comput Biomed Res 29:162--173.

Curran T, Schacter DL, Bessenoff G. 1996. Visual specificity effects on

word stem completion: beyond transfer appropriate processing? Can

J Exp Psychol 50:22--33.

Dale AM. 1999. Optimal experimental design for event-related fMRI.

Hum Brain Mapp 8:109--114.

Dale AM, Buckner RL. 1997. Selective averaging of rapidly presented

individual trials using fMRI. Hum Brain Mapp 5:329--340.

Dale AM, Liu AK, Fischl BR, Buckner RL, Belliveau JW, Lewine JD,

Halgren E. 2000. Dynamic statistical parametric mapping: combining

fMRI and MEG for high-resolution imaging of cortical activity.

Neuron 26:55--67.

Dehaene S, Jobert A, Naccache L, Ciuciu P, Poline JB, Le BihanD, Cohen L.

2004. Letter binding and invariant recognition of masked words:

behavioral and neuroimaging evidence. Psychol Sci 15:307--313.

Demb JB, Desmond JE, Wagner AD, Vaidya CJ, Glover GH, Gabrieli JDE.

1995. Semantic encoding and retrieval in the left inferior prefrontal

cortex: a functional MRI study of task difficulty and process

specificity. J Neurosci 15:5870--5878.

Dobbins IG, Schnyer DM, Verfaellie M, Schacter DL. 2004. Cortical

activity reductions during repetition priming can result from rapid

response learning. Nature 428:316--319.

Frost R. 1998. Toward a strong phonological theory of visual word

recognition: true issues and false trails. Psychol Bull 123:71--99.

Gabrieli JDE, Poldrack RA, Desmond JE. 1998. The role of left prefrontal

cortex in language and memory. Proc Natl Acad Sci USA 95:906--913.

Gauthier I, Anderson AW, Tarr MJ, Skudlarski P, Gore JC. 1997. Levels of

categorization in visual recognition studies using functional mag-

netic resonance imaging. Curr Biol 7:645--651.

Gibson JM, Brooks JO, Friedman L, Yesavage JA. 1993. Typography

manipulations can affect priming of word stem completion in older

and younger adults. Psychol Aging 8:481--489.

Glover GH, Lee AT. 1995. Motion artifacts in fMRI: comparison of

2DFT with PR and spiral scan methods. Magn Reson Med 33:

624--635.

Graf P, Ryan L. 1990. Transfer-appropriate processing for implicit and

explicit memory. J Exp Psychol Learn Mem Cogn 16:978--992.

Graf P, Shimamura AP, Squire LR. 1985. Priming across modalities and

priming across category levels: extending the domain of preserved

function in amnesia. J Exp Psychol Learn Mem Cogn 11:386--396.

Halgren E, Buckner RL, Marinkovic K, Rosen BR, Dale AM. 1997. Cortical

localization of word repetition effects. Cogn Neurosci Soc Annu

Meet 4:34.

Henson RN, Rugg MD. 2003. Neural suppression, hemodynamic repe-

tition effects and behavioural priming. Neuropsychologia

41:263--270.

Kapur S, Rose R, Liddle PF, Zipursky RB, Brown GM, Stuss D, Houle S,

Tulving E. 1994. The role of the left prefrontal cortex in verbal

processing: semantic processing or willed actions. Neuroreport

5:2193--2196.

Kolers PA. 1975. Memorial consequences of automatized encoding.

J Exp Psychol Hum Learn Mem 1:689--701.

Kolers PA. 1979. A pattern-analyzing basis of recognition. In: Cermak LS,

Craik FIM, editors. Levels of processing in human memory. Hillsdale,

NJ: Erlbaum. p 363--384.

Kolers PA, Ostry DJ. 1974. Time course of loss of information regarding

pattern analyzing operations. J Verb Learn Verb Behav 13:599--612.

Koutstaal W, Wagner AD, Rotte M, Maril A, Buckner RL, Schacter DL.

2001. Perceptual specificity in visual object priming: functional

magnetic resonance imaging evidence for a laterality difference in

fusiform cortex. Neuropsychologia 39:184--199.

Kucera M, Francis W. 1967. Computational analysis of present-day

American English. Providence, RI: Brown University Press.

Maccotta L, Zachs JM, Buckner RL. 2001. Rapid self-paced event-related

functional MRI: feasibility and implications of stimulus- versus

response-locked timing. Neuroimage 14:1105--1121.

Majerus S, Van der Linden M, Collette F, Laureys S, Poncelet M,

Degueldre C, Delfiore G, Luxen A, Salmon E. 2005. Modulation of

brain activity during phonological familiarization. Brain Lang

92:320--331.

Marsolek CJ, Kosslyn SM, Squire LR. 1992. Form-specific visual priming

in the right cerebral hemisphere. J Exp Psychol Learn Mem Cogn

18:492--508.

Marsolek CJ, Schacter DL, Nicholas CD. 1996. Form-specific visual

priming for new associations in the right cerebral hemisphere. Mem

Cogn 24:539--556.

Marsolek CJ, Squire LR, Kosslyn SM, Lulenski ME. 1994. Form-specific

explicit and implicit memory in the right cerebral hemisphere.

Neuropsychology 8:588--597.

Martin A, LaLonde FM, Wiggs CL, Weisberg J, Ungerleider LG, Haxby JV.

1995. Repeated presentation of objects reduces activity in ventral

occipitotemporal cortex: a fMRI study of repetition priming. Soc

Neurosci Abstr 21:1497.

Masson MEJ. 1986. Identification of typographically transformed words:

instance-based skill acquisition. J Exp Psychol Learn Mem Cogn

12:479--488.

Morris CD, Bransford JE, Franks JJ. 1977. Levels of processing versus

transfer appropriate processing. J Verb Learn Verb Behav

16:519--533.

Ostergaard A. 1998. The effects on priming of word frequency, number

of repetitions, and delay depend on the magnitude of priming. Mem

Cogn 26:40--60.

Ostergaard A. 1999. Priming deficits in amnesia: now you see them, now

you don’t. J Int Neuropsychol Soc 5:175--190.

Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME. 1988. Positron

emission tomographic studies of the cortical anatomy of single-word

processing. Nature 331:585--589.

Poldrack RA, Desmond JE, Glover GH, Gabrieli JDE. 1998. The neural

basis of visual skill learning: an fMRI study of mirror reading. Cereb

Cortex 8:1--10.

Poldrack RA. Gabrieli JDE. 2001. Characterizing the neural mechanisms

of skill learning and repetition priming: evidence from mirror

reading. Brain 125:67--82.

Poldrack RA, Wagner AD, Prull MW, Desmond JE, Gover GH, Gabrieli

JDE. 1999. Distinguishing semantic and phonological processing in

the left inferior prefrontal cortex. Neuroimage 10:15--35.

Raichle ME, Fiez JA, Videen TO, MacLeod AM, Pardo JV, Fox PT, Petersen

SE. 1994. Practice-related changes in human brain functional

anatomy during nonmotor learning. Cereb Cortex 4:8--26.

Rajaram S, Roediger HL. 1993. Direct comparison of four implicit

memory tests. J Exp Psychol Learn Mem Cogn 19:765--776.

Roediger HL, Blaxton TA. 1987. Effects of varying modality, surface

features, and retention interval on priming in word-fragment

completion. Mem Cogn 15:379--388.

Rumsey J, Horwitz B, Donohue B, Nace K, Maisog J, Andreason P. 1997.

Phonological and orthographic components of word recognition.

Brain 120:739--759.

Ryan L, Ostergaard A, Norton L, Johnson J. 2001. Explicit and implicit

word stem completion: declines in performance across the lifespan

and the role of search processes and familiarity. Mem Cogn

29:678--690.

Scarborough DL, Cortese C, Scarborough HS. 1977. Frequency and

repetition effects in lexical memory. J Exp Psychol Hum Percept

Perform 3:1--17.

Schacter DL. 1994. Priming and multiple memory systems: perceptual

mechanisms of implicit memory. In: Schacter DL, Tulving E, editors.

Memory systems 1994. Cambridge, MA: MIT Press. p 233--268.

Schacter DL, Buckner RL. 1998. Priming and the brain. Neuron

20:185--195.

Schacter DL, Dobbins IG, Schnyer DM. 2004. Specificity of priming: a

cognitive neuroscience perspective. Nat Neurosci Rev 5(11):853--862.

Schacter DL, Graf P. 1989. Modality specificity of implicit memory for

new associations. J Exp Psychol Learn Mem Cogn 15:3--12.

Shallice T. 1988. From neuropsychology to mental structure. Cambridge,

MA: Cambridge University Press.

Talairach J, Tournoux P. 1988. Co-planar stereotaxic atlas of the human

brain. New York: Thieme.

Page 10 of 11 Format-Specific Priming in Mirror Word Reading d Ryan and Schnyer

Tardiff T, Craik FIM. 1989. Reading a week later: perceptual and

conceptual factors. J Mem Lang 28:107--125.

Thompson-Schill SI, D’Esposito M, Kan IP. 1999. Effects of repetition and

competition on activity in left prefrontal cortex during word

generation. Neuron 23:513--522.

Thompson-Schill SI, Swick D, Farah MJ, D’Esposito M, Kan IP, Knight RT.

1998. Verb generation in patients with focal frontal lesions: a neuro-

psychological test of neuroimaging findings. Proc Natl Acad Sci USA

95:15855--15860.

Tulving E, Markowitsch HJ, Craik FIM, Habib R, Houle S. 1996. Novelty

and familiarity activations in PET studies of memory encoding and

retrieval. Cereb Cortex 6:71--79.

Tulving E, Schacter DL. 1990. Priming and human memory systems.

Science 247:301--306.

Vaidya CJ, Gabrieli JDE, Verfaellie M, Fleischman D, Askari N. 1998. Font-

specific priming following global amnesia and occipital lobe damage.

Neuropsychology 12:183--192.

Wagner AD, Desmond JE, Demb JB, Glover GH, Gabrieli JDE. 1997.

Semantic repetition priming for verbal and pictorial knowledge:

a functional MRI study of left inferior prefrontal cortex. J Cogn

Neurosci 9:714--726.

Wagner AD, Koutstaal W, Maril A, Schacter DL, Buckner RL. 2000. Task-

specific repetition priming in left inferior prefrontal cortex. Cereb

Cortex 10:1176--1184.

Weldon MS, Roediger HL. 1987. Altering retrieval demands reverses the

picture superiority effect. Mem Cogn 15:269--280.

Wiggs CL, Martin A. 1998. Properties and mechanisms of perceptual

priming. Curr Opin Neurobiol 8:227--233.

Cerebral Cortex Page 11 of 11