Cognitive Neuropsychology

ISSN: 0264-3294 (Print) 1464-0627 (Online) Journal homepage: https://www.tandfonline.com/loi/pcgn20

The role of sensorimotor learning in the

perception of letter-like forms: Tracking the causes
of neural specialization for letters

Karin H. James & Thea P. Atwood

To cite this article: Karin H. James & Thea P. Atwood (2009) The role of sensorimotor learning
in the perception of letter-like forms: Tracking the causes of neural specialization for letters,
Cognitive Neuropsychology, 26:1, 91-110, DOI: 10.1080/02643290802425914

To link to this article: https://doi.org/10.1080/02643290802425914

Published online: 07 May 2009.

Submit your article to this journal

Article views: 884

View related articles

Citing articles: 18 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=pcgn20
COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1), 91 – 110

The role of sensorimotor learning in the perception

of letter-like forms: Tracking the causes
of neural specialization for letters

Karin H. James and Thea P. Atwood

Department of Psychological and Brain Sciences, Indiana University, Bloomington, IN, USA

Functional specialization in the brain is considered a hallmark of efficient processing. It is therefore

not surprising that there are brain areas specialized for processing letters. To better understand the
causes of functional specialization for letters, we explore the emergence of this pattern of response
in the ventral processing stream through a training paradigm. Previously, we hypothesized that the
specialized response pattern seen during letter perception may be due in part to our experience in
writing letters. The work presented here investigates whether or not this aspect of letter proces-
sing—the integration of sensorimotor systems through writing—leads to functional specialization
in the visual system. To test this idea, we investigated whether or not different types of experiences
with letter-like stimuli (“pseudoletters”) led to functional specialization similar to that which exists
for letters. Neural activation patterns were measured using functional magnetic resonance imaging
(fMRI) before and after three different types of training sessions. Participants were trained to recog-
nize pseudoletters by writing, typing, or purely visual practice. Results suggested that only after
writing practice did neural activation patterns to pseudoletters resemble patterns seen for letters.
That is, neural activation in the left fusiform and dorsal precentral gyrus was greater when participants
viewed pseudoletters than other, similar stimuli but only after writing experience. Neural activation
also increased after typing practice in the right fusiform and left precentral gyrus, suggesting that
in some areas, any motor experience may change visual processing. The results of this experiment
suggest an intimate interaction among perceptual and motor systems during pseudoletter perception
that may be extended to everyday letter perception.

Keywords: Letters; fMRI; Pseudoletters; Writing; Development; Neural activation.

Functional specialization, the propensity for neural neural processing (Downing, Jiang, Shuman, &
systems to be active more to one category of stimu- Kanwisher, 2001; Gauthier, Williams, Tarr, &
lus than to a seemingly similar category, has been Tanaka, 1997; Tanaka & Gauthier, 1997).
thought to reflect efficient, sometimes expert, Functional specialization has been shown to exist

Correspondence should be addressed to Karin H. James, Department of Psychological and Brain Sciences, Indiana University,
1101 East 10th Street, Bloomington IN, 47405, USA (E-mail: khjames@indiana.edu).
The authors would like to acknowledge Thomas James for helpful comments on the manuscript and for scholarly discussion on
this topic.

# 2008 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business 91
http://www.psypress.com/cogneuropsychology DOI:10.1080/02643290802425914
JAMES AND ATWOOD

for several stimulus categories including specialization in high-level visual areas concerns
faces (Kanwisher, 2000; Kanwisher, Chun, & the eccentricity biases associated with different
McDermott, 1996), body parts (Downing et al., object categories (Hasson, Harel, Levy, &
2001), places (Kanwisher, 2000), biological Malach, 2003; Hasson, Levy, Behrmann,
motion (Grossman & Blake, 2002), words (Cohen Hendler, & Malach, 2002; Levy, Hasson,
et al., 2000), and individual letters (Flowers et al., Avidan, Hendler, & Malach, 2001; Levy, Hasson,
2004; James, James, Jobard, Wong, & Gauthier, Harel, & Malach, 2004; Malach, Levy, &
2005; Garrett et al., 2000; Longcamp, Anton, Hasson, 2002). Based on results from a variety of
Roth, & Velay, 2003). In one survey of functional functional magnetic resonance imaging (fMRI)
specialization of 20 stimulus categories in the studies, the authors suggested that functional
visual cortex, Downing, Chan, Peelen, Dodds, and specialization is a result of different resolution
Kanwisher (2006) found that there were a small demands associated with different object cat-
number of regions that were actually category selec- egories. Face perception is associated with centre-
tive, indicating that this is not a typical response biased visual areas for the detailed discrimination
profile for the ventral visual stream (Downing required, while perception of buildings is associated
et al., 2006). Functional specialization was revealed with periphery-biased visual areas for the large-
only for faces, body parts, and places—categories of scale integration involved. Words and letters,
stimuli with which we are very efficient at proces- because of their small size, represent the extreme
sing. Another category of stimuli (that was not case of object perception requiring high resolution
tested in the aforementioned work) that we learn and foveation.
to recognize extremely efficiently is composed of Another proposal based on stimulus processing
the written characters of our native writing requirements contends that specialized processing
system. Written characters are a particularly inter- in the ventral visual processing stream may also be
esting category of stimuli because the recency of due to the level of categorical analysis that is
such stimuli in our evolutionary history suggests required for a given task. For example, during
that there has been no change in our innate neural face perception one often has to identify the face
architecture that would support stimulus-specific of an individual (e.g., Al Gore). Such discrimi-
efficiency of processing. Nonetheless, with sufficient nation within a homogeneous class (e.g., among
exposure and training, individuals become extre- other faces) requires consideration of not only
mely efficient at processing their native characters the fine-grain, metric differences of features (e.g.,
in many formats over the course of just a few lip thickness), but also the second-order relations
years. Functional specialization for the written char- among face features (e.g., interocular distance;
acter has been shown in a number of studies (e.g., Diamond & Carey, 1986). However, face identifi-
Flowers et al., 2004; Garrett et al., 2000; James cation at this level does not require coarse dis-
et al., 2005). However, why this specialized neural crimination, for example whether a feature is
response pattern occurs is still a topic of speculation. present or absent—that is, we do not have to
First we consider a set of theories that are not discern whether or not Al Gore has a nose in
necessarily mutually exclusive, but that intend to order to distinguish him from other people.
offer explanations as to why functional specializ- Letter perception, in contrast, relies on the use
ation for letters occurs. We then show evidence of coarse information like feature presence/
that suggests that motor learning may play a part absence (e.g., an oblique stroke in “N”) and first-
in the emergence of neural specialization for order relations (e.g., the oblique stroke is in
letters in the ventral stream. between the two vertical strokes in “N”), as well
There are several theories that have been pro- as second-order relations (the angle of the
posed to explain why different regions become oblique line is also important to distinguish an
specialized for processing particular categories of “N” from an “H”). Analyses of the stimulus prop-
stimuli. One attempt to account for functional erties of characters across different languages have

92 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

also shown that a high level of redundancy is intro- the key factors determining the use of different
duced in their creation, such that one can perceive substrates for different categories. This theory
only part of a character and be able to distinguish it gains support from findings of recruitment of the
from the other alternatives (Changizi & Shimojo, face-selective areas for other types of objects like
2004). Therefore, letter identification involves cars, dogs, birds, fingerprints, and novel objects,
processing of different types of stimulus infor- which are claimed to require a common perceptual
mation when compared with other stimulus demand of fine-grained discrimination and
categories. common holistic manner of processing that
Other theories of ventral visual stream organiz- results from extensive experience (Gauthier et al.,
ation and specialization couch the problem in 2000; Xu, 2005).
terms of a continuous object-form topography, These ideas allow speculation as to why the
such that neighbouring neural substrates tend to ventral visual stream may demonstrate functional
represent features that are more similar to each specialization for letters, but how would this
other (Carlson, Schrater, & He, 2003; Cox & develop? We propose that functional specializ-
Savoy, 2003; Haxby et al., 2001; Ishai, ation for letters may be caused by the way that
Ungerleider, Martin, Schouten, & Haxby, 1999; we learn to recognize letters and, more specifically,
O’Toole, Jiang, Abdi, & Haxby, 2005). An that specialization for letters may reflect the sen-
important idea of this theory is that there is no sorimotor integration that is required when we
absolute neural specialization for certain object learn to write letters ( James & Gauthier, 2006;
categories. Instead, there is relative preference of Longcamp et al., 2003). Sensorimotor experience
certain substrates for certain categories based on in the form of learning to print and write letters
features, and object information is widely distribu- allows the interplay between motor production
ted and overlapping across the high-level visual and visual perception to broaden the stored rep-
areas. Support for this theory comes from finding resentation of letters. That is, motor construction
that activity in category-specific areas (e.g., face- of forms may lead to motor programmes that are
selective regions in the fusiform gyrus) contains stored with visual information. Because, during
diagnostic information for the categorization of writing, these programmes are variable, they may
nonpreferred categories (Haxby et al., 2001), and serve to augment visual information, allowing
activity patterns over a wide region of the inferior very different-looking exemplars to be categorized
temporal cortex correlate with physical properties as the same letter. Grouping visually dissimilar
of objects (O’Toole et al., 2005). It should be letters into a single category occurs before children
noted, however, that object representations in are exposed visually to a large variety of handwrit-
the occipito-temporal cortex are probably not ing styles or fonts, but after they learn to print. In
completely distributed, as some regions (e.g., the support of the idea that sensorimotor experience
face- and scene-selective areas) still enjoy a may lead to functional specialization in the visual
greater discriminatory power for their preferred processing stream, recent work has shown that
categories than other objects (O’Toole et al., when participants view letters, a sensorimotor
2005; Spiridon & Kanwisher, 2002). network becomes engaged ( James & Gauthier,
Another theory maintains that different parts of 2006). That is, areas in the ventral visual proces-
the ventral visual processing stream are suited for sing stream become active during visual perception
different processes. The observed object selectivity but, more interestingly, so do motor and premotor
is a result of the prolonged recruitment of different regions of the brain (perhaps part of the dorsal
substrates to fulfil specific recognition demands for stream of visual processing). This finding was
different objects (Bukach, Gauthier, & Tarr, 2006; interpreted as showing that neural circuits for
Gauthier, Skudlarski, Gore, & Anderson, 2000; writing letters (premotor cortex) were automati-
Tarr & Gauthier, 2000). Accordingly, “recog- cally activated upon seeing letters, implying that
nition demand” and “experience” are therefore motor experiences were stored and reactivated

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 93

JAMES AND ATWOOD

upon subsequent encounters. In addition, ventral (Bai & Bertenthal, 1992; Campos et al., 2000; for
visual areas—specifically, the left fusiform review see Wexler & Boxtel, 2005). For example,
gyrus—were engaged when participants wrote visual recognition of novel objects is facilitated by
letters (without seeing them), but not when they actively moving the objects compared to watching
drew shapes ( James & Gauthier, 2006). Thus, the same movement performed by another person
coactivation of brain regions that process both (Harman et al., 1999; James et al., 2001, 2002).
visual and motor information during either per- Such behavioural facilitation supports the idea that
ceptual or motor interaction with familiar letter the dorsal motor and ventral visual systems are inter-
stimuli was shown, suggesting a sensorimotor rep- acting during object processing.
resentation of letters. In addition, several neuroimaging studies have
Additional evidence for coactivation of the found that motor systems are automatically activated
visual processing areas and motor regions comes upon visual perception of some objects (Chao &
from a study showing common repetition suppres- Martin, 2000; Gerlach, Law, Gade, & Paulson,
sion profiles in ventral and dorsal stream regions 2002; Grezes & Decety, 2002; James & Gauthier,
(Mahon et al., 2007). In this study, repetition sup- 2006; Longcamp et al., 2003; Longcamp, Anton,
pression (RS), the tendency for neural activity to Roth, & Velay, 2005a; Longcamp, Zerbato-
be reduced upon repeated stimulus presentation Poudou, & Velay, 2005b; Mecklinger, Gruenewald,
(Grill-Spector, Henson, & Martin, 2006), was Besson, & von Cramon, 2002). This literature
shown to emerge for “tools” in areas in the suggests that motor systems are active when we
ventral visual processing stream as well as in visually perceive objects that we regularly interact
dorsal stream structures. In addition, the authors with motorically—for example, tools and utensils
use a functional connectivity analysis to show (Chao & Martin, 2000; Mahon et al., 2007;
that areas in the left middle temporal gyrus and Mecklinger et al., 2002). Interestingly, these objects
the left inferior parietal lobule both show RS for also invoke, by their appearance alone, specific ways
tool stimuli. Here, “tools” were defined as objects to interact with them. That is, they contain “affor-
whose function (how to interact with them) was dances” (Gibson, 1979) that can be used to specify
apparent in their structure—arguably, objects motor interactions (potentially independent of our
with “affordances” (Gibson, 1979). Interestingly, experience), and these affordances are visually per-
objects that did not have a systematic relationship ceptible. By investigating the effects of prior motor
between structure and function, termed “arbitrarily experience on visual recognition of objects without
manipulated” objects (e.g., book, envelope), over- affordances, one can more directly attribute the
lapped in their RS functions with “tools” in right recruitment of motor systems to experience.
hemisphere ventral (middle fusiform) and dorsal Letters, by their appearance alone, do not
stream (right caudal inferior parietal lobule) “suggest” how we must interact with them—in
systems, but not in left hemisphere systems. This this way, they are similar to the arbitrarily manip-
work suggests that motor information (how to ulable objects in the Mahon et al. (2007) studies.
interact with objects) is associated with visual Perceiving letters, however, does seem to invoke
object processing. the associated history of letter-specific motoric
This idea is supported by behavioural work as interactions. For instance, in a series of studies,
well. There is now a substantial body of evidence Freyd and colleagues (Babcock & Freyd, 1988;
that motor experience—that is, our history of inter- Freyd, 1983) found that the way that a subject
actions with some objects—can facilitate visual is taught to write a letter-like symbol directly
recognition (Harman, Humphrey, & Goodale, affects their subsequent recognition of that
1999; James, Humphrey, & Goodale, 2001; James symbol. In addition, writing experience can alter
et al., 2002) and mental rotation (Wexler, Kosslyn, the perception of movement illusions in written
& Berthoz, 1998; Wohlschläger & Wohlschläger, symbols (Tse & Cavanagh, 2000), and knowledge
1998), as well as the development of spatial maps of cursive stroke directions affects anticipated

94 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

letter identity (Orliaguet, Kandel, & Bois, 1997). letter identification and in learning to read—this
Longcamp et al. (2005b) have demonstrated that disorder is highly comorbid with dyslexia (e.g.,
children recognize letters more efficiently after Portwood, 2000). Furthermore, a patient has
being trained to print letters versus being been reported with agraphia (inability to write)
trained to type letters. This latter study suggests with alexia that has resulted from damage to the
that motor experience may be generative—the left premotor cortex (Anderson, Damasio, &
motor experience that is important for visual rec- Damasio, 1990). Although this patient cannot
ognition is through constructing the form of the read or identify letters and cannot write letters or
letter, not by the simple motor act of typing. A words, she can draw complex shapes, and she can
study conducted by Cunningham and Stanovich write numbers. This case provides some evidence
(1990) found similar results. They used three that damage to the motor system can affect not
different modalities to teach children how to only writing, but also visual processing of letters.
spell words. Children were presented with a It also shows that the motor deficit can be very cir-
word both visually and orally and were asked to cumscribed to one particular category of stimuli.
spell words by writing, arranging a set of tiles, The neuroimaging, patient, and behavioural
or typing the word. After spending 30 minutes results outlined above have revealed an interesting
training over the course of four days, children aspect of letter recognition—while we ultimately
were asked to spell out each of the words on a spend a lot of time reading words, it is the isolated
sheet of paper, by using a computer keyboard letter that we need to learn first, and we learn this
and by arranging tiles. Results demonstrated stimulus by seeing and writing. The work presented
that there were significantly more correct here investigates whether this aspect of letter
responses for those words learned in the writing processing—the integration of sensorimotor
condition versus the typing and tile conditions, systems—leads to functional specialization in the
and this effect emerged regardless of the format visual system. The left fusiform gyrus of the literate
in which the children spelled the words in the adult is already specialized for processing letters
test session. more than other, similar characters (Flowers et al.,
Furthermore, research on individuals with lit- 2004; Garrett et al., 2000; James & Gauthier,
eracy disabilities has also suggested a link among 2006; James et al., 2005). Therefore, it is difficult
motor and visual systems in letter processing. to assess the effects of writing experience on the
Writing movements can facilitate letter recog- development of this specialization. We therefore
nition in patients with pure alexia—the inability conducted a training study that directly compared
to identify letters and words (Bartolomeo, the effects of writing experience, typing experience,
Bachoud-Levi, Chokron, & Degos, 2002; Seki, and visual-only experience on visual “letter” proces-
Yajima, & Sugishita, 1995). Although these sing. Because adults have already learned letters,
patients cannot recognize a letter visually, if they however, we trained them on a group of letter-like
are allowed hand movements while they are characters, referred to here as “pseudoletters”. In
looking at the letter, they will often trace out the this way we investigated the effects of motor experi-
shape of the letter as if writing it—and this move- ence on visual recognition and on neural processing.
ment (that is unseen by them) facilitates their
visual recognition of letters. In addition, some
Method
dyslexic adults are delayed in motor tasks; it is
possible that in such cases motor difficulties Participants
affected their letter-learning ability (Stoodley, A total of 18 participants gave informed consent
Fawcett, Nicolson, & Stein, 2005). Similarly, chil- according to the guidelines of the Indiana
dren exhibiting developmental dyspraxia, a dis- University Human Participants Review Board
order that can manifest in reduced fine motor and were paid for their participation. All
skills (including writing), often have difficulty in were undergraduate or graduate students enrolled

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 95

JAMES AND ATWOOD

at Indiana University. All participants were matching tasks more difficult in the behavioural
right-handed and reported normal or corrected- portion of the experiment. All orientations and
to-normal vision, had English as their first font types were used in the matching task, both
language, and had no known history of neurologi- within the scanning environment and outside.
cal or psychiatric disorders. A total of 10 females
and 8 males participated, and they were between General procedure
the ages of 21 to 31 years with a median age of There were four sessions in total, and each session
23.5 years. was separated by one day of rest: First, there was a
pretraining imaging session (a “pretrain scan”),
Stimuli which helped to determine the participant’s
Stimuli were 96
96 pixels and were presented in initial blood-oxygen-level-dependent (BOLD)
isolation, in the centre of a computer screen. activation to the novel (pseudoletter) alphabet as
Stimuli included a group of 18 capital letters (H, well as to letters and simple shapes. Participants
A, F, C, S, U, K, N, T, B, D, G, R, J, L, P, Z, Y), then took part in two training sessions, separated
18 shapes and symbols (e.g., clover, heart, percen- by one day. The training sessions consisted of
tage sign, pound sign, treble clef, star, etc.), 18 training procedures, followed by behavioural
pseudoletters (studied), and another set of 18 testing. We included testing in these training ses-
pseudoletters (unstudied; see Figure 1). Each of sions to (a) assess whether or not the training had
these groups had four more subsets—an Arial any effect from Day 1 to Day 2 and (b) to motiv-
font group, a serif-type font group, a cursive-type ate the participants to learn the stimuli. All
font group, and a rotation group (consisting of testing procedures involved visual tasks only.
each stimulus rotated 0, 45, 90, and 180 After training, participants took part in a final
degrees). Stimuli that lacked noticeable difference posttraining imaging session (a “posttrain scan”)
when rotated were not used—for example, the to determine whether any change in BOLD acti-
letter “O” and the shape of a square were not vation patterns had occurred as a result of the
used because each of these symbols, when rotated training sessions.
180 degrees, are no different from their counter-
parts oriented at 0 degrees rotation. Different Scanning protocol. Both pre- and posttraining
formats of the stimuli were used to add variation scanning sessions proceeded in the same
to the stimulus sets and to attempt to make manner. All stimuli were back-displayed with a
Mitsubishi XL30 projector onto a screen that
was viewed through a mirror from the bore of
the Siemens Trio 3T scanner. Stimuli were pre-
sented with SuperLab Pro 2.0.4 software with
Dell Inspiron 6000 laptops. Each scanning
session consisted of six runs. The first five runs
were functional scans that measured activation
to our stimulus conditions, and the final run was
a high-resolution anatomical scan. Conditions
were presented in a blocked design: Each block
contained 16 different presentations of the stimu-
lus condition. Each run consisted of the same
stimulus blocks, but the order of the blocks was
randomized across runs. Order of runs was
varied across participants. Each run began and
Figure 1. A: Example of pseudoletter stimuli. B: Font types: sans ended with a 16-s fixation cross and, in addition,
serif, serif, cursive. consisted of 16 blocks of experimental trials: 4

96 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

stimulus types (letters, studied pseudoletters, the order of the presentation of the tests varied
unstudied pseudoletters, and shapes)
2 change per training session (1 or 2). A training session
conditions (font or rotation)
2 repetitions. consisted of a training exposure, followed by two
Within each block, participants were required to recognition tests, followed by another training
perform a one-back matching task by pressing a exposure, until a total of four training exposures
button with their right index finger when two were completed.
stimuli presented consecutively possessed the
same category identity (an identity that is shared Training exposure conditions. For the writing con-
regardless of font type or orientation). That is, dition, participants were given a pad of paper
participants had to detect whether an “A” was and a pen and were asked to copy pseudoletters
an “A” regardless of font or orientation. Stimuli presented on the computer screen to the best of
in each block were pseudorandomized such that their ability. For the typing condition, partici-
there were at least two repetitions (one-back pants were asked to find and type the pseudoletter
match) of a stimulus in each block. The one- that was presented on the screen. A keyboard was
back perceptual match task was used to maintain modified for this purpose (pictures of the pseudo-
attention. Participants pressed a button with letters were affixed to a regular keyboard). When
their right index finger when an image was the participant typed the key, the presentation
repeated irrespective of font or orientation. This of the pseudoletter on the screen disappeared
task was described to the participant prior to and was not replaced. Thus, the screen was
scanning. Stimuli were presented in the centre blank after the participant typed their response.
of the screen for 500 ms, followed by a Gaussian For the visual condition, participants were asked
noise mask presented for 500 ms. There were 16 to look at the presented stimulus and to try to
presentations of stimuli within each block, and memorize its form. All participants were notified
the ratio of repetitions to nonrepeats was 1 : 7 that they would be tested on these stimuli after
(making this a very easy task). The stimuli that training. No feedback was given for any of the
were presented within a block were randomly training conditions. In each training task, each
selected such that not every Rotation
Stimulus of the 18 stimuli were presented a total of three
combination was presented, the only requirement times in their proper orientation, for a total of
being that an average of two repetitions occurred 54 stimulus presentations per training task. This
in each block. Each block was separated by a 10- was repeated four times per training day (for a
s fixation cross. total of eight repetitions). Each stimulus was
presented for the same amount of time (4
Training Sessions 1 and 2 seconds) for each training condition. This
Both training sessions were run in the same resulted in the visual condition having a greater
manner, although the order of the behavioural amount of visual exposure to each stimulus,
tests varied. All stimulus presentations were run because they were not required to divert their
on a Dell Optiplex GX 280 desktop computer, gaze to write or type. However, in the typing con-
via SuperLab Pro 2.0.4 software, and response dition, participants would find and then see the
time and accuracy data were collected via a com- pseudoletter on the keyboard. In the writing con-
puter keyboard. Participants were randomly dition, they would write and then see the written
assigned to one of three conditions: a writing con- pseudoletter. For each training group the stimulus
dition, a typing condition, or a visual condition. was presented on the screen for the entire trial.
Each condition consisted of four training Although we did not measure the amount of
exposures and six behavioural tests. The training time that each person required to type or write,
exposure varied according to the condition both conditions were easily completed in the
(writing, typing, or visual). The six behavioural time allotted, and one did not appear to take
tests were constant for all participants, although longer than any other.

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 97

JAMES AND ATWOOD

Behavioural testing. The tests did not change from was one that had been previously studied, or if
participant to participant, although the order of the stimulus was new and had not been studied.
the tests did vary. A total of six tests were pre- Again, this test consisted completely of the trained
sented in one training session. Three of these pseudoletters, as well as 18 novel pseudoletter
were standard visual search tests. The participant stimuli (the “new” stimuli). Font and orientation
was presented with a target (trained pseudoletter) were changed randomly throughout this pro-
stimulus, followed by a Gaussian noise mask, fol- cedure; thus, not all levels of orientation and
lowed by a field of stimuli. The fields were 2
2 font were necessarily used. Participants pressed
(small array), 3
3 (medium array), or 6
6 the “/” key if the pseudoletter had been previously
(large array). The participant’s task was to studied and “z” if the pseudoletter was new. The
quickly decide whether the target stimulus was novel pseudoletters presented in the old – new
presented in the field of stimuli. The participant test decision were presented only within the con-
pressed the “/” key if the target was present and fines of this task. Of the three tasks, this was the
“z” if the target was absent. Stimuli were presented only test that required explicit recollection of the
in roman font and in an upright orientation. The studied pseudoletter stimuli.
visual search tests were included as a measure of Because this type of training exposure is vir-
learning that did not require explicit recognition tually untested, we were unsure how it would
of the target stimulus. However, visual search affect performance. For this reason, we included
has been shown to be sensitive to learning effects several behavioural tests that would potentially
and even as an effective measure of automaticity allow us to determine how our training affected
in processing (Czerwinski, Lightfoot, & Shiffrin, different types of performance (e.g., implicit vs.
1992). We were therefore curious to see whether explicit recognition).
or not visual search ability became more efficient
with our training paradigm. Imaging parameters. Imaging was performed using
Two tests required matching of stimuli. The a 3-T Siemens Magnetom Trio whole-body MRI
participant was presented with a target stimulus, system and a phased-array eight-channel head coil,
followed by a Gaussian noise mask, and then a located at the Indiana University Psychological
second stimulus, followed by a brief fixation and Brain Sciences department. The field of view
cross. The participant’s task was to decide was 22
22
12.5 cm, with an in-plane resol-
whether the two stimuli were the same or different ution of 64
64 pixels and 25 slices per volume
(serial match task). The first match test incorpor- that were 4 mm thick with a 1.0-mm gap among
ated font changes between the two stimuli; the them. These parameters allowed us to collect
second match test changed orientation of the data from the entire brain. The resulting voxel
first and second stimulus presentations. These size was 3.4 mm
3.4 mm
5.0 mm. Images
two matching tasks consisted completely of the were acquired using an echo-planar technique
trained pseudoletters. Participants pressed the “/” (echo time, TE ¼ 30 ms; time to repetition,
key if the stimuli were the same and the “z” if TR ¼ 2,000 ms; flip angle ¼ 708) for BOLD
the stimuli were different. This task was included based imaging. High-resolution T1-weighted
to assess implicit processing of the pseudoletters anatomical volumes were acquired using a 3D
and how this may change over time. In addition, Turbo-flash acquisition. Functional data under-
this task was a close approximation of the one- went slice time correction, 3D motion correction,
back task that was performed in the imaging linear trend removal, and Gaussian spatial blurring
environment, except that during this task all (FWHM 4 mm) using the analysis tools in Brain
levels of orientation changes and font changes VoyagerTM . Individual functional volumes were
were used. coregistered to anatomical volumes with an
The final test was an old/new recognition test: intensity-matching, rigid-body transformation
Participants decided whether a presented stimulus algorithm. Individual anatomical volumes were

98 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

normalized to the stereotactic space of Talairach of interest (that is, we did not extract BOLD
and Tournoux (1988) using an eight-parameter signal change to shapes).
affine transformation, with parameters selected
by visual inspection of anatomical landmarks.
Results and discussion
Applying the same affine transformation to
the coregistered functional volumes placed the Behavioural results
functional data in a common brain space, allowing Two separate 2
3 mixed measures ANOVAs
comparisons across participants. Voxel size of (one for reaction times and one for accuracy)
the normalized functional volumes was standar- were performed for each test type with training
dized at 1 mm
1 mm
1 mm using trilinear day (1 or 2) as a within-subjects variable and train-
interpolation. ing condition (write, type, or visual) as a between-
subjects variable.

fMRI data analysis procedures. The functional data Reaction times. For each test type there was no sig-
were further analysed with a random effects nificant effect of training group: Match 1 (font
general linear model (GLM) using Brain change), F(2, 15) ¼ 1.4, ns; Match 2 (orientation
Voyager’sTM multisubject GLM procedure. The change), F(2, 15) ¼ 0.93, ns; Visual Search 1,
GLM analysis allows for the correlation of predic- F(2, 15) ¼ 1.6, ns; Visual Search 2, F(2, 15) ¼
tor variables or functions with the recorded acti- 1.03, ns; recognition, F(2, 15) ¼ 0.85, ns. Thus,
vation data (criterion variables) across scanning the type of training did not affect speed of response
sessions. The predictor functions were based on in any of these measures. However, each group did
the blocked stimulus presentation paradigm of improve their performance over training days in
the particular run being analysed and represent the match tasks and in the large-array visual
an estimate of the predicted haemodynamic search task: Match 1, F(1, 15) ¼ 23.4, p ,
response during that run. Regions of interest .0001; Match 2, F(1, 15) ¼ 13.9, p , .002;
were determined based on group statistical para- Visual Search 2, F(1, 15) ¼ 8.7, p , .01. There
metric maps (SPMs) that were considered above was no improvement with training in the small-
threshold if they met the following criteria in our array visual search task, F(1, 15) ¼ 1.8, ns; and
random-effects analysis: (a) significant at only a trend towards significance in the recog-
p , .001, uncorrected, with a cluster threshold of nition task, F(1, 15) ¼ 3.9, p , .06. There were
270 contiguous 1-mm isometric voxels; (b) peak no significant interactions between training day
statistical probability within a cluster at least and training condition.
p , .0001, uncorrected.
To localize regions of the brain that were Accuracy. There were no significant effects on accu-
engaged during letter processing, we performed a racy as a function of training, or training group (all
letters versus fixation contrast in the group data Fs , 2.8, ns). This is a surprising result given the
combined across the pretrain and posttrain scans. expectation that motor training would facilitate
Results of this contrast are presented in Figure 2 both reaction time and accuracy. Accuracy,
and produced 12 regions of interest (ROIs). We however, was quite high in these tests prior to
then extracted percentage BOLD signal change any training, which may have contributed to the
values for all participants within these ROIs for insignificant change in performance. In addition,
both pretrain and posttrain scans. Using peak acti- it is not unusual to show reaction time effects
vation from each participant as our dependent without accuracy effects in many cognitive tasks
measure, we then performed an omnibus analysis (e.g., Harman et al., 1999; Prinzmetal, McCool,
of variance (ANOVA) for each ROI. However, & Park, 2005). Another possible explanation for
we only compared trained and untrained pseudo- the absence of accuracy differences is the small
letters with one another, these being our stimuli sample size: In each condition there were only 6

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 99

JAMES AND ATWOOD

Figure 2. Regions of interest (ROIs) resulting from the contrast of letters greater than fixation baseline combined across pretrain and posttrain
scans. Statistical parametric maps (SPMs) depict averaged group data (pre-and postscan data are collapsed for each participant) viewed at an
uncorrected statistical threshold of p , .0001 (see Table 1 for Talairach coordinates, cluster size, t-values, and p-values for each ROI). Data
are depicted in radiological coordinates. Top row, left: bilateral inferior occipital gyrus; centre: bilateral posterior fusiform gyrus; right: bilateral
middle fusiform gyrus. Middle row, left: bilateral middle occipital gyrus; centre: bilateral precentral gyrus; right: left dorsal precentral gyrus
(seen dorsal to the precentral activation). Bottom row: left medial precentral gyrus (seen medial to bilateral precentral activation). To view a
colour version of this figure, please see the online issue of the Journal.

100 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

participants. Studies that have shown effects of Localizing regions of interest

motor training on behavioural performance have We compared letters to fixation to localize regions
used much larger sample sizes (e.g., Longcamp of the brain that were involved in letter processing.
et al., 2005a, 2005b). More days of training may This contrast resulted in 12 ROIs (Figure 2): bilat-
have facilitated changes in accuracy as well. Prior eral inferior occipital gyrus (see Table 1 for all
studies on training participants on novel objects Talairach coordinates), bilateral posterior fusiform
have required 7 hours of training to result in a gyrus,bilateral middle fusiform gyrus, bilateral
behavioural change (Gauthier et al., 1997). Thus, middle occipital gyrus, bilateral precentral gyrus,
the behavioural methods that we used may not left dorsal precentral gyrus, and left medial precen-
have been sensitive enough to reveal the under- tral gyrus. Inferential statistics for each region of
lying neural changes that were occurring. interest are reported below, and because figures
represent the data in terms of difference scores
(trained minus untrained pseudoletters), we also
fMRI results include all raw data (% BOLD signal change
Behavioural results from imaging session. mean peaks) in Table 2.
Behavioural performance in the one-back percep-
tual matching task was not analysed beyond com-
puting descriptive statistics, because of ceiling ROI results
effects. Note that there was an average of only two Left inferior occipital gyrus (Figure 3). A 3 (training
repetitions of a stimulus per block; therefore, this group, TG: writing, typing, visual)
2 (scan day,
task was very easy for the participants. Mean SD: pretrain scan, posttrain scan)
2 (pseudolet-
responses in the letter task for each group was ter, PL: untrained or trained) mixed-model
at ceiling (99 + 0.5% for motor group; ANOVA revealed three significant interactions.
98.3 + 1.1% for typing group, and 97.5 + 2.1% The two 2-way interactions, one between TG
for visual group), as was performance for studied and SD, F(2, 15) ¼ 5.3, p , .01, and the second
pseudoletters (motor group, 97.3 + 0.5%; type between TG and PL, F(2, 15) ¼ 8.5, p , .01,
group, 98.0 + 2.8%; visual group, 96.3 + 1%), are better interpreted by looking at the significant
unstudied pseudoletters (motor group, 94.4 + 3-way interaction, F(2, 15) ¼ 3.8, p , .05. The
2.3%; type group, 98.9 + 1.7%; visual group significant 3-way interaction was due to no differ-
99 + 1.3%), and simple shapes (motor group ences between the untrained and trained pseudo-
99.9 + 0.5%; type group, 94.5 + 2.7%; visual letters in the pretrain scan session for any of the
group, 95.5 + 1.5%). groups—writing, t(5) ¼ 1.3, ns; type, t(5) ¼ 1.4,

Table 1. Talairach coordinates, cluster size, peak t value, and significance level of each region of interest

Region of interest Talairach coordinates (peak) (x, y, z) Cluster size (1-mm voxels) t(17) peak p , (uncorrected)

Left inferior occipital gyrus –29, –93, – 1 353 8.2 .000003

Right inferior occipital gyrus 27, –92, – 1 496 11.3 .000004
Left posterior fusiform gyrus –43, –66, – 12 657 9.4 .000002
Right posterior fusiform gyrus 38, –66, – 15 275 7.0 .00001
Left middle fusiform gyrus –43, –58, – 10 680 8.2 .000003
Right middle fusiform gyrus 36, –59, – 13 560 6.3 .00002
Left middle occipital gyrus –46, –73, – 9 290 9.3 .000002
Right middle occipital gyrus 46, –73, – 5 487 8.7 .000002
Left precentral gyrus –45, –4, 34 675 7.4 .000001
Right precentral gyrus 40, –4, 32 409 7.9 .000001
Left dorsal precentral gyrus –46, –8, 51 275 5.6 .0001
Left medial precentral gyrus –34, –9, 50 270 5.8 .0001

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 101

JAMES AND ATWOOD

Table 2. Peak BOLD percentage signal change averaged across all participants in each ROI

Motor Typing Visual

Pretrain Posttrain Pretrain Posttrain Pretrain Posttrain

ROI New PL Tr. PL New PL Tr. PL New PL Tr. PL New PL Tr. PL New PL Tr. PL New PL Tr. PL

L. inf. occipital gyrus 0.95 0.92 0.95 1.10 0.95 0.99 0.87 1.31 0.98 0.93 0.95 0.73
(0.08) (0.11) (0.08) (0.17) (0.30) (0.25) (0.23) (0.23) (0.31) (0.22) (0.24) (0.13)
R. inf. occipital gyrus 1.13 1.18 1.19 1.33 1.15 1.13 1.14 1.31 0.89 0.83 0.90 0.88
(0.10) (0.10) (0.08) (0.15) (0.28) (0.23) (0.12) (0.15) (0.18) (0.25) (0.27) (0.09)
L. post. fusiform 0.11 0.10 0.12 0.35 0.09 0.11 0.11 0.14 0.12 0.10 0.14 0.13
(0.06) (0.05) (0.09) (0.07) (0.04) (0.06) (0.06) (0.13) (0.09) (0.08) (0.11) (0.11)
R. post. fusiform 0.57 0.59 0.59 0.67 0.58 0.56 0.57 0.85 0.45 0.41 0.45 0.44
(0.05) (0.05) (0.04) (0.07) (0.14) (0.11) (0.06) (0.17) (0.09) (0.12) (0.14) (0.04)
L. middle fusiform 0.56 0.53 0.58 0.67 0.58 0.56 0.57 0.58 0.45 0.41 0.45 0.44
(0.04) (0.09) (0.04) (0.07) (0.14) (0.11) (0.06) (0.21) (0.09) (0.12) (0.14) (0.04)
R. middle fusiform 0.65 0.67 0.65 0.67 0.63 0.63 0.65 0.67 0.65 0.63 0.65 0.63
(0.14) (0.25) (0.22) (0.11) (0.18) (0.17) (0.22) (0.25) (0.14) (0.16) (0.14) (0.17)
L. middle occipital 1.26 1.21 1.33 1.19 1.18 1.20 1.25 1.26 1.18 1.16 1.23 1.22
(0.14) (0.17) (0.10) (0.18) (0.18) (0.19) (0.13) (0.13) (0.18) (0.09) (0.13) (0.13)
R. middle occipital 1.57 1.57 1.57 1.58 1.56 1.53 1.56 1.56 1.54 1.55 1.57 1.55
(0.10) (0.10) (0.10) (0.10) (0.10) (0.17) (0.11) (0.10) (0.10) (0.15) (0.10) (0.15)
L. precentral 0.53 0.51 0.53 0.83 0.51 0.52 0.51 0.72 0.50 0.50 0.45 0.50
(0.07) (0.09) (0.06) (0.16) (0.10) (0.08) (0.10) (0.18) (0.06) (0.07) (0.14) (0.07)
R. precentral 0.36 0.31 0.34 0.42 0.37 0.31 0.37 0.36 0.35 0.38 0.36 0.39
(0.03) (0.07) (0.08) (0.12) (0.05) (0.07) (0.05) (0.03) (0.03) (0.04) (0.03) (0.02)
L. dorsal precentral 0.71 0.73 0.69 0.86 0.74 0.74 0.75 0.73 0.71 0.73 0.71 0.74
(0.24) (0.29) (0.21) (0.12) (0.20) (0.20) (0.14) (0.17) (0.24) (0.25) (0.24) (0.24)
L. medial precentral 0.35 0.35 0.35 0.39 0.35 0.33 0.34 0.34 0.34 0.34 0.31 0.40
(0.95) (0.92) (0.95) 1.10) (0.95) (0.99) (0.87) 1.31) (0.98) (0.93) (0.95) (0.73)

Note: Standard deviations in parentheses. BOLD ¼ blood-oxygen-level-dependent. ROI ¼ region of interest. PL ¼ pseudoletter.
Tr. ¼ trained. R. ¼ right. L. ¼ left. inf. ¼ inferior. post. ¼ posterior.

ns; visual, t(5) ¼ 1.7, ns—with significant differ- motor group (M ¼ 1.15), t(5) ¼ 4.1, p , .001.
ences between the two PL types in the postscan The writing and typing groups were not signifi-
session for two (writing and typing) of the three cantly different from one another, t(5) ¼ 1.2, ns.
groups: writing, t(5) ¼ 2.6, p , .05; type,
t(5) ¼ 2.3, p , .05; visual, t(5) ¼ 0.65, ns. Thus,
Left posterior fusiform gyrus (see Figure 4). The
in this early visual area, the PL training had an
3
2
2 ANOVA revealed several significant
effect on percentage BOLD signal change in the
differences in this region. First, there were main
motor and typing training groups.
effects of SD, F(1, 15) ¼ 17.4, p , .001, and
PL, F(1, 15) ¼ 5.6, p , .05; the main effect of
Right inferior occipital gyrus. The results of the group was not significant, F(1, 15) ¼ 1.3, ns.
ANOVA in this region revealed one significant The significant main effects must be interpreted
main effect, that of the group variable, F(2, in light of the three 2-way interactions and one
15) ¼ 11.9, p , .001. This effect was driven by 3-way interaction. The three 2-way interactions,
lower overall peak percentage signal change in between TG and SD, F(2, 15) ¼ 5.7, p , .01,
the visual group (M ¼ 0.87) than in the typing TG and PL, F(2, 15) ¼ 4.7, p , .05, and
group (M ¼ 1.2), t(5) ¼ 5.4, p , .001, and in the SD
PL, F(1, 15) ¼ 18.4, p , .0001, are better

102 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

Figure 3. A depiction of the significant three-way interaction

among pseudoletters (PLs), training group, and scan day in the
left inferior occipital gyrus. All interaction graphs depict peak Figure 4. A depiction of the significant three-way interaction
blood-oxygen-level-dependent (BOLD) activation of untrained among pseudoletters (PLs), training group, and scan day in the
pseudoletters subtracted from trained pseudoletters as a function of left posterior fusiform gyrus. There is no significant difference
training group and scan day. Error bars depict 95% confidence between trained and untrained pseudoletters prior to training and
intervals, and therefore overlap with the x-axis depicts no difference in activation to these two stimulus sets after typing
nonsignificance, while nonoverlap depicts a significant difference and visual training. There is, however, a large difference in
between trained and untrained pseudoletters (see text for blood-oxygen-level-dependent (BOLD) percentage signal change
inferential statistics and Table 2 for mean % BOLD signal (greater to trained than to untrained) after motor training.
change values). Note that there are no differences between trained
and untrained PLs before training (light grey bars), but
significant differences after training for the motor and typing training. When letters are compared to other
training group. visual stimuli, the left fusiform gyrus has been
shown to be engaged more during letter processing
than during processing of other, similar shapes
understood when we consider the three-way inter-
(Flowers et al., 2004; James et al., 2005;
action, F(2, 15) ¼ 13.2, p , .0001. The 3-way
Longcamp et al., 2003). Here we show that acti-
interaction revealed that for the writing training
vation before and after training with pseudoletters
group, there was a significant difference only in
changed activation patterns in this region,
the posttraining scan between trained and
suggesting that the sensitivity to letters in this
untrained pseudoletters, t(5) ¼ 4.02, p , .005,
region may be due to sensorimotor interactions
but not between trained and untrained PLs prior
as well.
to training, t(5) ¼ 1.9, ns. In the typing training
group, there were no significant differences
between trained and untrained PLs during the Right posterior fusiform gyrus (Figure 5). Another
pretrain scan, t(5) ¼ 0.21, ns, or during the post- 3
2
2 mixed-model ANOVA was run on the
train scan, t(5) ¼ 0.26, ns. For the visual training data from the right fusiform gyrus, revealing a
group, again, there were no significant differences main effect of SD, F(1, 15) ¼ 9.8, p , .005, and
between the trained and untrained PLs in the pre- of PL, F(1, 15) ¼ 9.2, p , .005. Three significant
train scan, t(5) ¼ 1.97, ns, or in the posttrain con- interactions emerged as well: one between TG
dition, t(5) ¼ 0.47, ns. Thus, in the left posterior and PL, F(1, 15) ¼ 6.8, p , .005, one between
fusiform gyrus, we find that percentage BOLD SD and PL, F(1, 15) ¼ 7.0, p , .01, and a three-
signal change increases as a function of training, way interaction, F(1, 15) ¼ 3.4, p , .05. The
but only when the training involves writing the three-way interaction was driven by no significant
pseudoletters, not after typing or purely visual differences in percentage BOLD signal change

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 103

JAMES AND ATWOOD

this region—a result also found in the right

inferior occipital gyrus.

Right middle fusiform gyrus. There were no signifi-

cant differences among any of the conditions in
this region (all F values , 2.0). This result
reinforces the idea that the middle fusiform
gyrus, although letter sensitive, is not involved in
the changes due to training that are seen in the
more posterior fusiform.

Bilateral middle occipital gyrus. There were no

significant differences among any of the conditions
in these regions.

Figure 5. The significant three-way interaction among Left precentral gyrus (Figure 6). From the
pseudoletters (PLs), training group, and scan day in the right 3
2
2 ANOVA, we see significant main
posterior fusiform gyrus. There is no significant difference between effects of all three conditions: TG, F(1,
trained and untrained pseudoletters prior to training and no
difference after visual training. There is, however, a difference in
15) ¼ 8.8, p , .005; SD, F(1, 15) ¼ 8.0, p , .01;
blood-oxygen-level-dependent (BOLD) percentage signal change and PL, F(1, 15) ¼ 15.7, p , .001. There were
(greater to trained than to untrained) after motor and typing also two significant interactions, one between
training. TG and SD, F(1, 15) ¼ 3.7, p , .05, and one
between SD and PL, F(1, 15) ¼ 16.3, p , .001.
Simple effects demonstrated that the TG
SD
for new versus trained pseudoletters in the
interaction was due to the writing, t(5) ¼ 2.7,
pretraining scan session for any of the groups (all
t values , 1.9), contrasting with a difference
between the new and trained PL conditions in
the posttraining session after typing training,
t(5) ¼ 3.4, p , .005, and writing training,
t(5) ¼ 2.75, p , .05, but not after visual-only
training, t(5) ¼ 0.45, ns.

Left middle fusiform gyrus. Another 3
2
2

mixed-model ANOVA revealed a significant
main effect of TG in this region, F(1, 15) ¼ 8.6,
p , .005, with the visual training group producing
less percentage signal change than the writing
group, t(5) ¼ 2.8, p , .05, and less than the
typing group, t(5) ¼ 3.1, p , .05 (writing,
M ¼ 0.58, SD ¼ 0.07; typing, M ¼ 0.57,
SD ¼ 0.13; visual, M ¼ 0.43, SD ¼ 0.09). No
other effects were significant (all other F values Figure 6. The three-way interaction among pseudoletters (PLs),
training group, and scan day in the left precentral gyrus. There is
, 2.0). In contrast with the left posterior fusiform,
no significant difference between trained and untrained
the left middle fusiform was not affected by train- pseudoletters prior to training and, again, no difference after
ing in this paradigm. It is curious that the visual visual training. However, motor and typing training both result
training group displayed less activation overall in in greater activation to trained than to untrained PLs.

104 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

p , .05, and typing, t(5) ¼ 2.1, p , .05, groups due to the contrast that is used (letters vs. fixation).
both showing an increase in percentage signal Although the 3
2
2 ANOVA performed on
change after training, whereas the visual training the data from this region revealed a significant
group did not, t(5) ¼ 0.69, ns. The SD
PL main effect of SD, F(1, 15) ¼ 7.4, p , .01, with
interaction was due to an increase in percentage the posttraining scan resulting in a higher percen-
signal change after training in the trained pseudo- tage signal change overall (M ¼ 0.37, SD ¼ 0.05)
letters only, t(5) ¼ 3.7, p , .005, that was not than the pretraining scan (M ¼ 0.37, SD ¼ 0.06),
present prior to training, t(5) ¼ 0.25, ns. Taken there were no other significant effects (all
together, after training, the two groups that inter- Fs , 2.5).
acted with the stimuli using motor systems both
showed increases in percentage BOLD signal Left dorsal precentral gyrus (Figure 7). This region is
change. The second interaction suggests that in similar to that previously found to be engaged
all three groups, the trained pseudoletters were during letter perception in some studies ( James
responded to with greater percentage signal & Gauthier, 2006). The 3
2
2 ANOVA
change than were the untrained pseudoletters that was performed on the data from this region
only after training (M ¼ 0.49, SD ¼ 0.10 for revealed three effects. The first was a main effect
new PLs, and M ¼ 0.68, SD ¼ 0.19 for trained of PL, F(1, 15) ¼ 5.7, p , .03. Perceiving new
PLs after training sessions). The three-way inter- PLs resulted in a lower percentage signal change
action approached significance, F(1, 15) ¼ 2.9, (M ¼ 0.71, SD ¼ 0.12) than did trained PLs
p , .08, as reflected in Figure 6—the writing (M ¼ 0.76, SD ¼ 0.14). There were also two
and typing groups appeared to be significantly interactions, one 2-way interaction between TG
different from the visual group in terms of the and PL, F(2, 15) ¼ 3.5, p , .05, and a 3-way
change in BOLD response after training. Recent interaction among TG, SD, and PL, F(2,
research has found recruitment of the left 15) ¼ 3.8, p , .05. Simple effects analyses
precentral gyrus during visual letter processing revealed that there was an increase in percentage
( James & Gauthier, 2006; Longcamp et al., signal change between the new and trained PLs
2003, 2005a, 2005b). The hypothesis that has only after writing training, t(5) ¼ 2.7, p , .05
been brought forth by both groups is that left (all other t values , 2.0; see Figure 7).
motor regions are activated because of stored
motor programmes that result from experience in Left medial precentral gyrus. The results of our
writing letters. The current results support this localizer contrast also revealed a region in the
claim and extend it by showing that typing train- medial portion of the precentral gyrus that was
ing also results in activation in this region during involved in letter perception. A 3
2
2
visual presentation of pseudoletter stimuli. This ANOVA in this region revealed one significant
is the first direct evidence supporting the claim interaction among the conditions: that between
that neural engagement in motor areas during SD and PL, F(1, 15) ¼ 5.1, p , .05. This inter-
visual tasks is due to motor experience and not action was due to a significant difference between
specifically writing experience. new and trained PLs only after training,
t(5) ¼ 2.5, p , .05, but not before training,
Right precentral gyrus. Prior evidence for right pre- t(5) ¼ 0.33, ns. Thus, there was an effect of
central gyrus engagement during letter-processing training in this region, but it was not specific to
tasks is less compelling than that for the left pre- a given type of training.
central gyrus. The usual explanation for the lack
of activation is that the left hemisphere processes Summary
language stimuli more than does the right hemi- To summarize these results, our localizer contrast
sphere. We did find some right precentral engage- revealed 12 regions in the brain that were active
ment during letter tasks here, though—perhaps more to letter perception than to a fixation

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 105

JAMES AND ATWOOD

letter-selective regions. Interestingly, the training

conditions served to change the response patterns
in several of the regions.

GENERAL DISCUSSION

There is abundant evidence that there is neural

functional specialization for processing individual
letters (Flowers et al., 2004; James & Gauthier,
2006; James et al., 2005; Polk & Farah, 1998;
Polk et al., 2002), but, until now, there has not
been any work investigating how this specializ-
ation may develop. We now present evidence
Figure 7. The three-way interaction among pseudoletters (PLs), that our experience in writing letters may contrib-
training group, and scan day in the left dorsal precentral gyrus. ute to the development of functional specialization
There is no significant difference between trained and untrained for letters. We demonstrate that after writing
pseudoletters prior to training and, again, no difference after
practice, some brain regions that are engaged
typing or visual training. However, motor training resulted in
greater activation to trained than to untrained PLs. during letter processing are also engaged more to
trained pseudoletters than to untrained pseudolet-
ters, implying that these regions increase their
baseline. Of these 12 regions, 5 showed a three- response to these particular stimuli only after a
way interaction among our conditions, where specific type of motor interaction. In other brain
new versus trained pseudoletters were responded regions, activation changed both after writing
to differently as a function of scan day and training and after typing training, implying that these
group. In 3 of these 5 regions (the left inferior regions increase responses after any motor experi-
occipital gyrus, the right fusiform gyrus, and the ence. Interestingly, the regions that were found to
left precentral gyrus), writing and typing groups change response as a result of writing training were
showed an increase in percentage signal change the same regions as those found in previous work
when perceiving trained pseudoletters after train- to respond during letter perception and during
ing versus perceiving untrained PLs in the same letter writing (the left fusiform gyrus and the left
session. In 2 other regions (the left posterior fusi- dorsal precentral gyrus; James & Gauthier, 2006).
form gyrus and the left dorsal precentral gyrus) Although performance on some of our beha-
there was greater neural activation during trained vioural measures improved with training, there
PL perception than during untrained PL percep- was no difference in performance in behavioural
tion only after writing training. Interestingly, tasks as a function of different training conditions.
these latter regions are the same areas that have This null behavioural result is in marked contrast
been found previously to respond to letter percep- to our significant differences in BOLD activation
tion and letter writing (James & Gauther, 2006). among training groups. Showing a difference in
Thus, we have demonstrated that regions of the BOLD activation, but not a difference in beha-
brain that respond more to letters than to a fixation vioural measures, is not an uncommon pattern
cross are not necessarily letter specific. These of results (e.g., see James & Gauthier, 2006;
regions are also engaged during perception of McLaughlin, Osterhout, & Kim, 2004;
other letter-like stimuli after a certain type of Wilkinson & Halligan, 2003a, 2003b). Such a
experience. Here we tested how writing, typing, result can be attributed to the greater sensitivity
and visual experience with previously novel of measuring changes in BOLD response than
characters would affect neural responses in these response times or accuracy measures (see also

106 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

Wilkinson & Halligan, 2003a for further discus- the motor information that would affect visual
sion of this view). Such sensitivity is important processing?
in revealing neural processing that can inform We would suggest that when children learn to
current cognitive theories and shape future write letters, they also learn a very efficient way
research endeavours. to group letters into “basic-level” categories
The imaging results of this work suggest that (Rosch, Mervis, Gray, Johnson, & Boyes-Braem,
after experience with certain types of stimuli—in 1976) and to perhaps ignore irrelevant changes
this case, pseudoletters—brain regions change among exemplars within that category. That is,
their responses to the learned versus unlearned the category of the letter “A”—presumably a
exemplars of that stimulus set. Because neural acti- basic, or “entry level” ( Jolicoeur, 1985) cat-
vation changes are based on training and not based egory—has many exemplars within it that look
on stimulus properties (trained and untrained very different (e.g., A, a, a, A, etc). To become effi-
pseudoletters are interchangeable), our results do cient readers, we must learn to identify exemplars
not support theories of ventral stream organization that are in the same category, despite irrelevant
that focus on eccentricity biases (e.g., Hasson changes in appearance. Children initially learn to
et al., 2003) or stimulus characteristics (e.g., visually recognize and categorize letters that are
Haxby et al., 2001) as organizing principles of in a single font type and always appear in the
ventral stream activation. We would argue that same orientation (the sans serif, capital block
the functional specialization that has been docu- alphabet). When children start to write letters
mented to letters (e.g., Flowers et al., 2004; themselves however, the motor information that
Garrett et al., 2000; James et al., 2005) in the produces that letter is variable. This variability,
left ventral stream may be based partially on our at first, may conflict with the intent to write a
motor experience with the stimuli. certain letter. With practice, the motor variability
We know that motor interactions with objects may serve to augment visual processing by broad-
play a crucial role in learning about the environ- ening a visual-motor representation. In this
ment, and it makes sense that the brain is orga- interpretation, motor experience changes stored
nized to reflect this interaction: We see representations and perhaps then also changes
coactivation of visual and motor regions during visual processing. In fact, recent research has
perception of a variety of different types of shown that preliterate children (4 –5-year-olds)
objects (e.g., Chao & Martin, 2000; Grezes & who are given printing practice with letters have
Decety, 2002; James & Gauthier, 2006). The a very different neural response to letters than do
present results support this general idea: that children who do not receive such practice. That
after motor experience, seeing a pseudoletter is, prior to practice, the ventral visual processing
recruits areas of the brain that are specialized for stream of both groups of children responds simi-
letter processing, comprising a sensorimotor larly to visually presented letters—neural responses
network. But what is it about writing experience to letters is the same as those to shapes. But after
that leads to this functional specialization? printing practice, the left fusiform gyrus shows a
Presumably, practised motor experiences set up heightened activation to letters compared to
representations of motor programmes in the simple shapes ( James, 2008).
frontal cortex, and these representations are acti- Another possibility as to why the sensorimotor
vated upon subsequent visual presentation of the systems interact may be due simply to coactivation.
object. Efferent copies of such information may Coactivation of multiple systems during learning
project to the ventral visual processing stream, could lead to “wiring together” of such systems,
affecting neural activation in these regions. This leading to future coactivations upon subsequent
process may facilitate recognition of the object encounters of the same input. More likely, and
by augmenting the visual information with not mutually exclusive with the former expla-
sensorimotor information. What then is it about nation, activation of sensorimotor systems may

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 107

JAMES AND ATWOOD

serve to potentiate a motor response. If seeing Witherington, D. (2000). Travel broadens the
the item previously required a motor response, mind. Infancy, 1, 149–219.
then perhaps seeing it again would activate a pro- Carlson, T. A., Schrater, P., & He, S. (2003). Patterns
gramme associated with the response to prepare of activity in the categorical representations
of objects. Journal of Cognitive Neuroscience, 15,
for action.
704– 717.
Our knowledge of the effects of learning
Changizi, M. A., & Shimojo, S. (2004). Character com-
experience on functional brain changes in plexity and redundancy in writing systems over
humans is in its beginning stages. The research human history. Proceedings of the Royal Society
presented here demonstrates that our sensorimotor Britain.
interactions with the environment are stored Chao, L. L., & Martin, A. (2000). Representation of
during learning. Furthermore, our motor systems man-made objects in the dorsal stream.
are active during visual presentations, suggesting NeuroImage, 12, 478– 484.
that the motor system is involved in visual proces- Cohen, L., Dahaene, S., Naccache, L., Lehéricy, S.,
sing at some level and may contribute to functional Dahaene-Lambertz, G., Hénaff, M. A., et al.,
specialization in the ventral stream. Activation of 2000. The visual word form area: Spatial and tem-
poral characterization of an initial stage of reading
the motor system during visual processing shown
in normal participants and posterior split-brain
here presumably reflects the behavioural findings
patients. Brain, 123, 291– 307.
that motor information affects visual processing Cox, D., & Savoy, R. (2003). Functional magnetic
of letters (e.g., Freyd, 1983; Orliaguet et al., resonance imaging (fMRI) “brain reading”:
1997; Tse & Cavanagh, 2000). Additional work Detecting and classifying distributed patterns of
is needed to uncover why motor interactions fMRI activity in human visual cortex. NeuroImage,
during learning influence visual processing and 19, 261 – 270.
whether this interaction is crucial for normal Cunningham, A. E., & Stanovich, K. E. (1990). Early
letter recognition ability to develop. spelling acquisition: Writing beats the computer.
Journal of Educational Psychology, 82, 159– 162.
Czerwinski, M., Lightfoot, N., & Shiffrin, R. M.
(1992). Automatization and training in visual
REFERENCES search. American Journal of Psychology, 105, 271– 315.
Diamond, R., & Carey, S. (1986). Why faces are and are
Anderson, S. W., Damasio, A. R., & Damasio, H. not special: An effect of expertise. Journal of
(1990). Troubled letters but not numbers: Domain Experimental Psychology: General, 115, 107– 117.
specific cognitive impairments following focal Downing, P. E., Chan, A. W.-Y., Peelen, M. V.,
damage in frontal cortex. Brain, 113, 749–766. Dodds, C. M., & Kanwisher, N. (2006). Domain
Babcock, M. K., & Freyd, J. J. (1988). Perception of specificity in visual cortex. Cerebral Cortex, 16,
dynamic information on static form. American 1453– 1461.
Journal of Psychology, 101, 111– 131. Downing, P. E., Jiang, Y., Shuman, M., &
Bai, D. L., & Bertenthal, B. I. (1992). Locomotor status Kanwisher, N. (2001). A cortical area selective for
and the development of spatial search skills. Child visual processing of the human body. Science, 293,
Development, 63, 215–226. 2470– 2473.
Bartolomeo, P., Bachoud-Levi, A.-C., Chokron, S., & Flowers, D. L., Jones, K., Noble, K., VanMeter, J.,
Degos, J. D. (2002). Visually- and motor- based Zeffiro, T. A., Wood, F. B., et al. (2004).
knowledge of letters: Evidence from a pure alexic Attention to single letters activates left extrastriate
patient. Neuropsychologia, 40, 1363– 1371. cortex. NeuroImage, 21, 829– 839.
Bukach, C. M., Gauthier, I., & Tarr, M. J. (2006). Freyd, J. J. (1983). Representing the dynamics of static
Beyond faces and modularity: The power of an form. Memory & Cognition, 11, 342– 346.
expertise framework. Trends in Cognitive Sciences, Garrett, A. S., Flowers, D. L., Absher, J. R., Fahey,
10, 159– 166. F. H., Gage, H. D., Keyes, J. W., et al. (2000).
Campos, J. J., Anderson, D. I., Barbu-Roth, M. A., Cortical activity related to accuracy of letter recog-
Hubbard, E. M., Hertenstein, M. J., & nition. NeuroImage, 11, 111– 123.

108 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)

 SENSORIMOTOR LEARNING AND LETTERS

Gauthier, I., Skudlarski, P., Gore, J. C., & Anderson, James, K. H., Humphrey, G. K., & Goodale, M. A.
A. W. (2000). Expertise for cars and birds recruits (2001). Manipulating and recognizing visual
brain areas involved in face recognition. Nature objects: Where the action is. Canadian Journal of
Neuroscience, 3, 191– 197. Experimental Psychology, 55, 111– 120.
Gauthier, I., Williams, P., Tarr, M. J., & Tanaka, J. James, K. H., Humphrey, G. K., Vilis, T., Corrie, B.,
(1997). Training “greeble” experts: A framework Baddour, R., & Goodale, M. A. (2002). “Active”
for studying expert object recognition process. and “passive” learning of three-dimensional object
Vision Research, 38, 2401– 2428. structure within an immersive virtual reality environ-
Gerlach, C., Law, I., Gade, A., & Paulson, O. B. ment. Behavior Research Methods, Instruments &
(2002). The role of action knowledge in the compre- Computers, 34, 383– 390.
hension of artifacts—a PET study. NeuroImage, 18, James, K. H., James, T. W., Jobard, G., Wong, A., &
928– 937. Gauthier, I. (2005). Letter processing in the visual
Gibson., J. J. (1979). The ecological approach to visual system: Different activation patterns for single
perception. Boston: Houghton Mifflin. letters and strings. Cognitive, Affective, &
Grezes, J., & Decety, J. (2002). Does visual Behavioral Neuroscience, 5, 452– 466.
perception of object afford action? Evidence from Jolicoeur, P. (1985). The time to name disoriented
a neuroimaging study. Neuropsychologia, 40, natural objects. Memory & Cognition, 13, 289– 303.
212– 222. Kanwisher, N. (2000). Domain specificity in face
Grill-Spector, K., Henson, R., & Martin, A. (2006). perception. Nature Neuroscience, 3, 759– 763.
Repetition and the brain: Neural models of stimulus- Kanwisher, N., Chun, M. M., & McDermott, J. (1996).
specific effects. Trends in Cognitive Science, 10, 18–24. fMRI in individual subjects reveals loci in extrastriate
Grossman, E. D., & Blake, R. (2002). Brain activity cortex differentially sensitive to faces and objects.
evoked by inverted and imagined biological motion. Investigative Ophthalmology & Visual Science, 37, S193.
Vision Research, 41, 1475– 1482. Levy, I., Hasson, U., Avidan, G., Hendler, T., & Malach,
Harman, K. L., Humphrey, G. K., & Goodale, M. A. R. (2001). Center-periphery organization of human
(1999). Active manual control of object views facili- object areas. Nature Neuroscience, 4, 533–539.
tates recognition. Current Biology, 9, 1315– 1320. Levy, I., Hasson, U., Harel, M., & Malach, R. (2004).
Hasson, U., Harel, M., Levy, I., & Malach, R. (2003). Functional analysis of the periphery effect in
Large-scale mirror-symmetry organization of human building related areas. Human Brain
human occipito-temporal object areas. Neuron, 37, Mapping, 22, 15 – 26.
1027– 1041. Longcamp, M., Anton, J. L., Roth, M., & Velay, J. L.
Hasson, U., Levy, I., Behrmann, M., Hendler, T., & (2003). Visual presentation of single letters activates
Malach, M. (2002). Eccentricity bias as an organiz- a premotor area involved in writing. NeuroImage, 19,
ing principle for human high order object areas. 1492– 1500.
Neuron, 34, 479– 490. Longcamp, M., Anton, J. L., Roth, M., & Velay, J. L.
Haxby, J. V., Gobbini, M. I., Furey, M. L., Ishai, A., (2005a). Premotor activations in response to visually
Schouten, J. L., & Pietrini, P. (2001). Distributed presented single letters depend on the hand used to
and overlapping representations of faces and write: A study on left-handers. Neuropsychologia,
objects in ventral temporal cortex. Science, 293, 43, 1801– 1809.
2425– 2430. Longcamp, M., Zerbato-Poudou, M., & Velay, J. L.
Ishai, A., Ungerleider, L. G., Martin, A., Schouten, (2005b). The influence of writing practice on letter
J. L., & Haxby, J. (1999). Distributed representation recognition in preschool children: A comparison
of objects in the human ventral visual pathway. between handwriting and typing. Acta Psychologica,
Proceedings of the National Academy of Sciences, USA, 119, 67 – 79.
96, 9379– 9384. Mahon, B. Z., Milleville, S., Negri, G. A. L., Rumiati, R.
James, K. H. (2008). Sensori-motor experience leads to I., Martin, A., & Caramazza, A. (2007). Action-related
changes in visual processing in the developing brain. properties of objects shape object representations in
Manuscript submitted for publication. the ventral stream. Neuron, 55, 507–520.
James, K., & Gauthier, I. (2006). Letter processing Malach, R., Levy, I., & Hasson, U. (2002). The topo-
automatically recruits a sensory-motor brain graphy of high-order human object areas. Trends in
network. Neuropsychologia, 44, 2937– 2949. Cognitive Sciences, 6, 176–184.

COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1) 109

JAMES AND ATWOOD

McLaughlin, J., Osterhout, L., & Kim, A. (2004). occipito-temporal cortex? An fMRI study. Neuron,
Neural correlates of second-language word learning: 35, 1157– 1165.
Minimal instruction produces rapid change. Nature Stoodley, C. J., Fawcett, A. J., Nicolson, R. I., & Stein,
Neuroscience, 7, 703– 704. J. F. (2005). Impaired balancing ability in dyslexic
Mecklinger, A., Gruenewald, C., Besson, M. N., & von children. Experimental Brain Research, 167,
Cramon, D. Y. (2002). Separable neuronal circuits 370– 380.
for manipulable and non-manipulable objects in Talairach, J., & Tournoux, P. (1988). Co-planar stereo-
working memory. Cerebral Cortex, 12, 1115– 1123. taxic atlas of the human brain. New York: Thieme.
Orliaguet, J.-P., Kandel, S., & Bois, L.-J. (1997). Visual Tanaka, J. W., & Gauthier, I. (1997). Expertise
perception of motor anticipation in cursive hand- in object and face recognition. In R. L. Goldstone,
writing: Influence of spatial and movement infor- D. L. Medin, & P. G. Schyns (Eds.), Psychology of
mation on the perception of forthcoming letters. learning and motivation: Vol. 36. Perceptual mechan-
Perception, 26, 905–912. isms of learning (pp. 83 – 125). San Diego, CA:
O’Toole, A. J. O., Jiang, F., Abdi, H., & Haxby, J. V. Academic Press.
(2005). Partially distributed representations of Tarr, M. J., & Gauthier, I. (2000). FFA: A flexible
objects and faces in ventral temporal cortex. Journal fusiform area for subordinate-level visual processing
of Cognitive Neuroscience, 17, 580– 590. automatised for expertise. Nature Neuroscience, 3,
Polk, T. A., & Farah, M. J. (1998). The neural develop- 764– 769.
ment and organization of letter recognition: Tse, P. E., & Cavanagh, P. (2000). Chinese and
Evidence from functional neuroimaging, compu- Americans see opposite apparent motions in a
tational modeling, and behavioral studies. Chinese character. Cognition, 74, B27– B32.
Proceedings of the National Academy of Sciences, USA, Wexler, M., & Boxtel, J. J. A. (2005). Depth perception
95, 847– 852. by the active observer. Trends in Cognitive Science, 9,
Polk, T. A., Stallcup, M., Aguirre, G. K., Alsop, D. C., 431– 438.
D’Esposito, M., Detre, J. A., et al. (2002). Neural Wexler, M., Kosslyn, S. M., & Berthoz, A. (1998).
specialization for letter recognition. Journal of Motor processes in mental rotation. Cognition, 68,
Cognitive Neuroscience, 14, 145– 159. 77 – 94.
Portwood, M. (2000). Developmental dyspraxia. Wilkinson, D., & Halligan, P. W. (2003a). The rel-
London: David Fulton Publishing. evance of behavioral measures for functional-
Prinzmetal, W., McCool, C., & Park, S. (2005). imaging studies of cognition. Nature Reviews,
Attention: Reaction time and accuracy reveal differ- Neuroscience, 5, 67 – 73.
ent mechanisms. Journal of Experimental Psychology: Wilkinson, D., & Halligan, P. W. (2003b). Stimulus
General, 134, 73 – 92. symmetry affects the bisection of figures but not
Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M., & lines: Evidence from event-related fMRI.
Boyes-Braem, P. (1976). Basic objects in natural NeuroImage, 20, 1756– 1764.
categories. Cognitive Psychology, 8, 382–439. Wohlschläger, A., & Wohlschläger, A. (1998). Mental
Seki, K., Yajima, M., & Sugishita, M. (1995). The effi- and manual rotation. Journal of Experimental
cacy of kinesthetic reading treatment for pure alexia. Psychology: Human Perception and Performance, 24,
Neuropsychologia, 33, 595– 609. 397– 412.
Spiridon, M., & Kanwisher, N. (2002). How distribu- Xu, Y. (2005). Revisiting the role of the fusiform face area
ted is visual category information in human in visual expertise. Cerebral Cortex, 15, 1234–1242.

110 COGNITIVE NEUROPSYCHOLOGY, 2009, 26 (1)