Abstract

Recent neuroimaging and transcranial magnetic stimulation studies indicate that the occipital cortex of congenitally blind humans is functionally relevant for nonvisual tasks. There are suggestions that the underlying cortical reorganization is restricted by a critical period. These results were based on comparison between early and late blind groups, thereby facing the problem of great variability among individuals within each group. Using functional magnetic resonance imaging, we studied bilingual congenitally blind individuals during use of 2 languages: one acquired early (Hebrew), the other later in life (English, at ∼10 years). The subjects listened to chimeric words consisting of superimposed Hebrew and English nouns. They were instructed to either covertly generate a verb to the heard noun or repeat the noun, in either Hebrew or English. Lateralized activation during verb generation (vs. repeat) was found in classical language areas, in congruence with previous studies in sighted subjects. Critically, in our study, the blind participants typically also had robust left lateralized occipital differential activation during verb generation (vs. repeat), in both languages. This suggests that the critical period for plasticity persists beyond 10 years or that the visual cortex of the blind might be engaged in abstract levels of language processing, common to the 2 languages.

Introduction

What happens to the brain when the eyes cannot see? Previous studies showed that in circumstances in which the visual input is totally absent from birth (i.e., congenital blindness), the occipital cortex undergoes a radical form of reorganization (Sadato and others 1996, 1998; Cohen and others 1997, 1999; Buchel and others 1998; Roder and others 2002; Noppeney and others 2003; Lambert and others 2004). This can be observed, for example, in an increased functional magnetic resonance imaging (fMRI) signal in the occipital cortex during performance of verbal tasks, such as verb generation or verbal memory (Burton, Snyder, Conturo, and others 2002; Burton, Snyder, Diamond, and Raichle 2002; Amedi and others 2003; Raz and others 2005), which is unique to the blind and is not seen in sighted controls. Furthermore, using transcranial magnetic stimulation (TMS), a causal relationship between this cortical activation and behavioral performance has been recently demonstrated. Thus, transient disruption of occipital cortex function by repetitive TMS during performance of an auditory verb generation task leads to an increased error rate in blind subjects but not in sighted peers (Amedi and others 2004). This strongly suggests that the functional role of the occipital cortex is dramatically altered.

It therefore seems that the occipital cortex is much more adaptable than previously considered. However, it is unclear to what extent is this plasticity limited by a critical period, beyond which no such changes can be observed. The concept of critical period has been first described by Hubel and Wiesel (1970) in their classical study of the development of ocular dominance columns in the kitten's visual cortex following monocular deprivation. But is there a similar limitation in the case of the total alteration of (visual) cortical function that seems to occur following blindness? To address this, researchers typically compared the occipital activation elicited in early blind versus late blind patients during performance of various nonvisual tasks (Buchel and others 1998; Cohen and others 1999; Burton, Snyder, Conturo, and others 2002; Burton, Snyder, Diamond, and Raichle 2002; Sadato and others 2002). For example, using fMRI, Burton, Snyder, Diamond, and Raichle (2002) found that, on average, the early blind group showed significant and vastly distributed occipital cortex activation during an auditory-based verb generation task. In the late blind subjects (those turning blind at the average age of 19 years), the occipital activation was still evident, although to a lesser extent. The third group, consisting of sighted subjects, showed no significant occipital activation while performing the same task. This suggests that the degree of cortical plasticity might decline with time, but it is not strictly limited to the period before adolescence as the critical period hypothesis might suggest. On the other hand, studies investigating the degree of occipital activation in the blind during performance of tactile tasks suggest that there is a period of susceptibility for the cross-modal plasticity, in which the occipital cortex can become responsive to tactile input (Cohen and others 1999; Sadato and others 2002). For instance, Sadato and others (2002) reported that V1 was active during a tactile discrimination task in blind subjects who lost their sight before 16 years of age, whereas no such activation was evident in blind subjects who lost their sight after 16 years of age. Cohen and others (1999) reported that repetitive TMS to the occipital pole disrupted a Braille letter identification task in both congenitally blind and early-onset blind subjects but not in late-onset blind subjects (age of blindness onset >14 years). One way to explain the conflicting evidence regarding the critical period from various studies is that while Sadato and Cohen examined tactile tasks, Burton's study involved auditory input and performance of a verbal task. It is possible (though maybe not very plausible) that the temporal “window of opportunity” for cortical reorganization is task and modality dependent.

The group approach, used in the previous studies to tackle the issue of the critical period, suffers from a serious liability: the variability in the pattern of occipital activation among individuals within a group is immense. Thus, unless large numbers of subjects participate in the study (which is typically not the case due to obvious limitations), the conclusions made necessarily require caution.

Although previous studies have repeatedly shown that verb generation (in the mother language) elicits robust occipital activation in early blind subjects, there is considerable variation among the individuals in terms of the extent and magnitude of the evoked activation (see e.g., Fig. 6 in Amedi and others 2003). In this study, we circumvent this problem by examining the occipital activation in bilingual early blind subjects, during verb generation in either the mother tongue or using the second language, acquired later in life (at ∼10 years). This design allows us to test whether fMRI activation during use of the second language is found only in classical language areas (as in sighted subjects). Such a finding would suggest that indeed there is a critical period (<10 years) for the recruitment of occipital cortex in the blind toward other nonvisual high-level functions. If on the other hand, activation in the occipital cortex during use of the second language would not be different from that elicited by the mother tongue, this may suggest that the occipital cortex is mutable also in adulthood (or at least until the age of 10). An alternative interpretation would be that in the blind, the occipital cortex is engaged in abstract levels of language processing, common to both languages. This might mean that the “visual” cortex is recruited by the native language and is later available for use by the second language.

Materials and Methods

Subjects

The participants in this study were 9 congenitally blind volunteers, 6 women and 3 men, aged 21–42 years. The mean age of the group was 28.4 ± 6.7 years (mean ± standard deviation). Eight of the subjects lost their sight at birth, and one lost her sight at the age of 2 years. An expert examined the subjects to assess the cause of blindness and presence of any light perception. All subjects were blind due to dysfunction at the level of the eye or early optic nerve causing major retinal damage, and their blindness was not due to a progressive neurological disease. There was no history of neurological or psychiatric illness in any of the subjects. Seven of the subjects did not have any form of light perception. Two subjects could only report the presence of a strong light but could not localize it or recognize any pattern. Seven of the subjects were right handed, as assessed using the Edinburgh test. All were proficient Braille readers. See Table 1.

Table 1

Blind subjects characteristics

Subject Age and sex Cause of blindness Light perception Handedness Age of blindness onset RT difference (ms) Hebrew VG–English VG 
LL 34 M Retinopathy of prematurity Faint Right 30 
RM 21 F Retinopathy of prematurity None Right 196 
LD 22 F Leber congenital amaurosis Faint Ambidextrous 402 
KD 22 F Microphthalmia None Right 168 
NZ 42 F Retinoblastoma None Right 165 
FA 28 M Retinopathy of prematurity None Left 501 
OC 27 F Retinopathy of prematurity None Right 54 
DR 31 F Rubella None Right 194 
SR 29 M Rubella None Right 0
 
203
 
Subject Age and sex Cause of blindness Light perception Handedness Age of blindness onset RT difference (ms) Hebrew VG–English VG 
LL 34 M Retinopathy of prematurity Faint Right 30 
RM 21 F Retinopathy of prematurity None Right 196 
LD 22 F Leber congenital amaurosis Faint Ambidextrous 402 
KD 22 F Microphthalmia None Right 168 
NZ 42 F Retinoblastoma None Right 165 
FA 28 M Retinopathy of prematurity None Left 501 
OC 27 F Retinopathy of prematurity None Right 54 
DR 31 F Rubella None Right 194 
SR 29 M Rubella None Right 0
 
203
 

All subjects were native Hebrew speakers, who had learned English in elementary school at the age of 10 years (9.9 ± 0.6) and received further English instruction during their middle school and high school years. All had attained adequate proficiency, passing a standardized English examination at the end of high school. Seven out of the 9 subjects had an additional 2–4 years of English education beyond high school (one subject is an English teacher in elementary school, another subject lived one year in the USA participating in an English speaking program, and another subject is a proficient translator).

The Tel-Aviv Sourasky Medical Center Ethics Committee approved the experimental procedure, and written informed consent was obtained from all subjects according to the Declaration of Helsinki.

Magnetic Resonance Image Acquisition

The blood oxygen level–dependent (BOLD) fMRI measurements were performed in a whole-body 1.5-T, Signa Horizon, LX8.25 General Electric scanner. The fMRI protocol was based on a multislice gradient echo-planar imaging and a standard head coil. The functional data were obtained under the optimal timing parameters: time repetition = 3 s, time echo = 55 ms, flip angle = 90°, imaging matrix = 80 × 80, field of view = 24 cm. The 27 slices with slice thickness 4 mm and 1 mm gap were oriented in the axial position. The scan covered the whole brain. The fMRI images were superimposed on T1-weighted 3-dimensional SPGR images (3D-fast spoiled GRASS, spatial resolution: 1*1*1 mm) that were acquired immediately after the experimental run.

Experimental Setup

All stimuli (superimposed Hebrew and English nouns) were aurally presented. The words were recorded from the voice of a male narrator that does not have a foreign accent in either languages (speaking Hebrew in a Hebrew accent and English in an English accent). The nouns were recorded by a digital recording device using GoldWave Shareware software (GoldWave, Inc., Newfoundland, Canada). Using Matlab, extraneous noise before and after each word was eliminated, and the root mean square of the signal amplitude was equalized for all words. The auditory stimulus sequences were played on a CD player. The auditory signals were heard binaurally through a commercially available noise shielding headphones (Slimline noise guard headset, Newmatic Sound System, USA). All responses (verb generation or word repeat) were performed covertly during the scan.

Stimuli and Experimental Design

Each stimulus was composed of a Hebrew noun and a superimposed English noun (it sounded like the same person speaking in 2 different languages at the same time). In all, 40 chimeric stimuli were heard during the scan. Each chimeric stimulus was heard 4 times during the experiment, once in each of the 4 conditions. Thus, the only difference between the conditions was the instruction before each block. There were 4 different conditions: 1) Hebrew repetition task, in which the subjects were instructed to pay attention to the Hebrew noun and to covertly repeat it, 2) English repetition task, in which the subjects were instructed to pay attention to the English noun and to covertly repeat it, 3) Hebrew verb generation task, in which the subjects were instructed to pay attention to the Hebrew noun and to covertly generate an appropriate Hebrew verb to the heard noun, and 4) English verb generation task, where the subjects were instructed to pay attention to the English noun and to covertly generate an appropriate English verb (e.g., think of the word “open” in response to “door”), see Figure 1. All nouns were of everyday objects (familiarity score of 600–700 according to the Medical Research Council [MRC] Psycholinguistic Database www.psy.uwa.edu.au/MRCDataBase/uwa_mrc.htm). We refrained from using nouns that can also be used as verbs (i.e., responding “to phone” when hearing the noun “phone”).

Figure 1.

Experimental design. The experiment was conducted using a block design paradigm. In each block, 4 auditory chimeric stimuli were presented, 1 every 3 s. Each stimulus was composed of an English noun (e.g., “suitcase”) and a superimposed Hebrew noun (e.g., “tomato” in Hebrew: “agvania”). There were 4 different conditions using the same chimeric stimuli in each condition. These were Hebrew REP—in which the subjects had to repeat the Hebrew noun in Hebrew; English REP—in which the subject repeated the English noun in English; Hebrew VG—requiring generation of an appropriate Hebrew verb to the Hebrew noun; English VG—requiring generation of an appropriate English verb to the English noun. All responses were performed covertly.

Figure 1.

Experimental design. The experiment was conducted using a block design paradigm. In each block, 4 auditory chimeric stimuli were presented, 1 every 3 s. Each stimulus was composed of an English noun (e.g., “suitcase”) and a superimposed Hebrew noun (e.g., “tomato” in Hebrew: “agvania”). There were 4 different conditions using the same chimeric stimuli in each condition. These were Hebrew REP—in which the subjects had to repeat the Hebrew noun in Hebrew; English REP—in which the subject repeated the English noun in English; Hebrew VG—requiring generation of an appropriate Hebrew verb to the Hebrew noun; English VG—requiring generation of an appropriate English verb to the English noun. All responses were performed covertly.

All blocks lasted 12 s followed by a 9 s period of rest. In the rest period, subjects simply lay still in the scanner. This rest condition served as a hemodynamic baseline condition. Four chimeric stimuli were presented in each block, one every 3 s. Each condition was repeated 10 times using different stimuli (40 blocks in total, 10 for each condition). The order of block presentation was counterbalanced across conditions within each subject. All subjects performed the same sequence of conditions. A short (∼1 s) auditory instruction was given before the beginning and at the end of all blocks.

Two of the subjects (LL and LD) did a second control experiment, in which the same 4 conditions (verb generation and repetition, in either English or Hebrew) were used. However, in this control experiment single nouns in one language were presented in each block (instead of superimposed nouns from both languages, as in the original experiment).

Procedure

A week before the scan, a practice session took place, in which the subjects practiced listening to the chimeric stimuli (that were later heard during the scan) until they could repeat either word (in both languages) in each chimeric stimulus. The subjects practiced on a larger set of stimuli and performed the verb generation task on a different set of stimuli than the one used in the experiment. Verb generation during the scan was therefore for “novel” nouns, thus eliminating the possibility of an automatic response (see Raichle and others 1994). The practice session assured that the participants could perform the task accurately and fast enough to respond within the time interval between heard stimuli. Reaction times (RTs) and behavioral performance were measured.

Behavioral performance (percent correct) in all tasks was assessed immediately after the scan using the same procedure as the one performed during the experiment. To eliminate the possibility that verb generation immediately after the scan was based on mere memory of the associated verbs during the scan (rather than an active search for the appropriate verb), we conducted another test 2 months after the scan. The subjects performed exactly the same trials as they did in the experiment, presented in a different order. RT and percent correct were measured in each trial.

Data Analysis

Behavioral Data

Performance level was assessed as percentage of semantically correct responses out of the total number of trials. RT was measured using the DMDX software, measuring the time from the onset of the heard word (headphones) to the first vocal response of the subject.

Imaging Data

Data analysis was performed using the Brain Voyager 4.96 software package (Brain Innovation, Maastricht, The Netherlands). Before statistical analysis, head motion correction, slice scan time correction, and high-pass temporal smoothing in the frequency domain were applied in order to remove drifts and to improve the signal to noise ratio. A general linear model (GLM) was used to generate statistical parametric maps (modeling the hemodynamic response function using parameters as in Boynton and others 1996). Across-subject statistical parametric maps (Fig. 2) were calculated using hierarchical random-effect model analysis (Friston and others 1999). This was done after the voxel activation time courses of all subjects were transformed into Talairach space (Talairach and Tournoux 1988), Z-normalized, and concatenated. Significant levels were calculated taking into account the probability of a false detection for any given cluster (Forman and others 1995) by a Monte Carlo simulation (AlphaSim by B. Douglas Ward, Medical College of Wisconsin, a software module in Cox 1996) using the combination of individual voxel probability threshold and a minimum cluster size of 10 functional voxels. The minimum significance level, corrected for any given cluster, was P < 0.05. The retinotopic borders displayed on the Talairach normalized unfolded brain of the blind (across-subject statistical parametric maps, Fig. 2) were roughly estimated using the rotating wedge technique (Engel and others 1997) on one sighted subject, who was scanned in the same scanner using the same sequence. The Talairach normalized volumetric time course of activation of the sighted subject was superimposed on a blind subject's Talairach normalized brain. Then the approximate retinotopic borders were assessed using the phase information.

Figure 2.

Lateralized cortical activation during verb generation in either language. Statistical parametric maps of activation (n = 9) using random-effect GLM analysis. The data are presented on a full Talairach-normalized unfolded brain of the left hemisphere (LH) and right (RH) hemisphere as well as a lateral and medial inflated view of both hemispheres. Color scale denotes significance (P < 0.05, corrected for multiple comparisons). The posterior green line indicates the approximate V1/V2 border, whereas the anterior dotted line denotes the estimated anterior border of retinotopic areas in the sighted (see Materials and Methods for details). (A) Hebrew VG > Hebrew REP. Contrasting activation during Hebrew VG versus that elicited during the Hebrew REP task shows robust, lateralized, and highly significant activation in 2 main foci; one in the left occipital cortex, including the calcarine sulcus (i.e., V1) and extrastriate visual areas (V2–V4). The other is found in the prefrontal cortex, including the IFS, the IFG (inferior frontal gyrus), and the PCS (precentral sulcus), see Table 2. (B) English VG > English REP. A second test, contrasting the activation elicited during the English VG versus the English REP conditions. This contrast shows a very similar pattern of activation—including the occipital areas and the prefrontal cortex (P < 0.05, corrected), see Table 3. Anatomical markers: LS, lateral sulcus; CS, central sulcus; STS, superior temporal sulcus; CoS, collateral sulcus.

Figure 2.

Lateralized cortical activation during verb generation in either language. Statistical parametric maps of activation (n = 9) using random-effect GLM analysis. The data are presented on a full Talairach-normalized unfolded brain of the left hemisphere (LH) and right (RH) hemisphere as well as a lateral and medial inflated view of both hemispheres. Color scale denotes significance (P < 0.05, corrected for multiple comparisons). The posterior green line indicates the approximate V1/V2 border, whereas the anterior dotted line denotes the estimated anterior border of retinotopic areas in the sighted (see Materials and Methods for details). (A) Hebrew VG > Hebrew REP. Contrasting activation during Hebrew VG versus that elicited during the Hebrew REP task shows robust, lateralized, and highly significant activation in 2 main foci; one in the left occipital cortex, including the calcarine sulcus (i.e., V1) and extrastriate visual areas (V2–V4). The other is found in the prefrontal cortex, including the IFS, the IFG (inferior frontal gyrus), and the PCS (precentral sulcus), see Table 2. (B) English VG > English REP. A second test, contrasting the activation elicited during the English VG versus the English REP conditions. This contrast shows a very similar pattern of activation—including the occipital areas and the prefrontal cortex (P < 0.05, corrected), see Table 3. Anatomical markers: LS, lateral sulcus; CS, central sulcus; STS, superior temporal sulcus; CoS, collateral sulcus.

Tables 2 and 3 are based on the BOLD signal intensity of the peak voxel in a smoothed volume after convolution with Gaussian kernel of 8 mm (full width at half maximum).

Table 2

Brain regions active during Hebrew VG > Hebrew REP control

Area (nearest BA) Talairach coordinates
 
t-Value in peak voxel Adjusted P value (corrected) 
MFG (BA 6) −39 −4 48 7.6 0.00006 
MFG (BA 9) −53 39 5.1 0.001 
DFG (BA 32) −8 45 7.2 0.00009 
IFG (BA 45) −42 18 4.1 0.005 
IFG (BA 44) −57 18 14 2.9 0.02 
IFG/INS (BA 47/13) −37 10 3.3 0.01 
PCG (BA 44) −44 51 11 3.8 0.01 
IPL (BA 7) −29 −58 45 7.9 0.00005 
IPL (BA 40) −38 −43 45 4.8 0.001 
CaS (BA 17) −17 −1 −89 4.4 0.003 
CaS (BA 17) −9 −89 −2 3.9 0.005 
LG (BA 18) −17 −73 −8 5.8 0.0004 
LG (BA 18) −17 −74 −7 5.6 0.0005 
FG (BA 18) −35 −75 −13 6.8 0.0001 
MTG (BA 21) −60 −45 7.7 0.00006 
CG (BA 32) −6
 
8
 
45
 
4.7
 
0.002
 
Area (nearest BA) Talairach coordinates
 
t-Value in peak voxel Adjusted P value (corrected) 
MFG (BA 6) −39 −4 48 7.6 0.00006 
MFG (BA 9) −53 39 5.1 0.001 
DFG (BA 32) −8 45 7.2 0.00009 
IFG (BA 45) −42 18 4.1 0.005 
IFG (BA 44) −57 18 14 2.9 0.02 
IFG/INS (BA 47/13) −37 10 3.3 0.01 
PCG (BA 44) −44 51 11 3.8 0.01 
IPL (BA 7) −29 −58 45 7.9 0.00005 
IPL (BA 40) −38 −43 45 4.8 0.001 
CaS (BA 17) −17 −1 −89 4.4 0.003 
CaS (BA 17) −9 −89 −2 3.9 0.005 
LG (BA 18) −17 −73 −8 5.8 0.0004 
LG (BA 18) −17 −74 −7 5.6 0.0005 
FG (BA 18) −35 −75 −13 6.8 0.0001 
MTG (BA 21) −60 −45 7.7 0.00006 
CG (BA 32) −6
 
8
 
45
 
4.7
 
0.002
 

Note: MFG, middle frontal gyrus; DFG, medial frontal gyrus; IFG, inferior frontal gyrus; INS, insula; PCG, precentral gyrus; IPL, inferior parietal lobe; CaS, calcarine sulcus; LG, lingual gyrus; FG, fusiform gyrus; MTG, middle temporal gyrus; CG, cingulate gyrus; BA, Brodmann area.

Table 3

Brain regions active during English VG > English REP control

Area (nearest BA) Talairach coordinates
 
t-Value in peak voxel Adjusted P value (corrected) 
SFG (BA 6) −4 64 5.1 0.0009 
MFG (BA 6) −51 42 6.2 0.0002 
MFG (BA 6) −48 48 6.0 0.0003 
MFG (BA 6) −34 −2 52 3.8 0.005 
DFG (BA 6) −4 63 5.0 0.0009 
IFG (BA 45) −49 17 3.8 0.01 
IFG/INS (BA 45/13) −32 22 2.9 0.02 
PCG (BA 4) −50 −7 51 5.8 0.0004 
PCG (BA 6) −51 44 4.8 0.002 
SPL (BA 7) −23 −63 55 4.4 0.003 
CaS (BA 17) −5 −84 3.7 0.006 
CaS (BA 17) −6 −85 3.5 0.008 
LG (BA 18) −22 −73 −8 4.1 0.003 
PH (BA 35) −17
 
−37
 
−5
 
4.8
 
0.001
 
Area (nearest BA) Talairach coordinates
 
t-Value in peak voxel Adjusted P value (corrected) 
SFG (BA 6) −4 64 5.1 0.0009 
MFG (BA 6) −51 42 6.2 0.0002 
MFG (BA 6) −48 48 6.0 0.0003 
MFG (BA 6) −34 −2 52 3.8 0.005 
DFG (BA 6) −4 63 5.0 0.0009 
IFG (BA 45) −49 17 3.8 0.01 
IFG/INS (BA 45/13) −32 22 2.9 0.02 
PCG (BA 4) −50 −7 51 5.8 0.0004 
PCG (BA 6) −51 44 4.8 0.002 
SPL (BA 7) −23 −63 55 4.4 0.003 
CaS (BA 17) −5 −84 3.7 0.006 
CaS (BA 17) −6 −85 3.5 0.008 
LG (BA 18) −22 −73 −8 4.1 0.003 
PH (BA 35) −17
 
−37
 
−5
 
4.8
 
0.001
 

Note: IFG, inferior frontal gyrus; INS, insula; MFG, middle frontal gyrus; PCG, precentral gyrus; PH, parahippocampal gyrus; SFG, superior frontal gyrus; SPL, superior parietal lobe; BA, Brodmann area.

The activation time course of individual subjects (Fig. 3) was obtained from statistically significant clusters in each region of interest (ROI), by using the GLM analysis with correction for multiple comparisons in each subject (applying the fixed false detection rate [FDR] technique [at q < 0.05] and a minimal cluster size of 10 functional voxels for each subject). The averaged signal change during stimulus presentation (averaging over 6–12 s after block onset) was also calculated (Fig. 3). Data from individual subjects were then averaged across all subjects.

Figure 3.

Time course and average histograms of the fMRI activation in specific ROIs. The time course of activation (upper panel) was constructed by pooling significant voxels (GLM test: all > rest; FDR at q < 0.05, corrected for multiple comparisons) on a subject-by-subject basis and then averaging the activation in each condition across all subjects (n = 9). The gray area indicates the duration of the experimental block (12 s). The histograms (lower panel) show the average percent signal change in each ROI (averaging over 6–12 s after block onset). Asterisks denote significant difference between VG- and REP-related activation (paired t-test, P < 0.05) in each language. Error bars denote SEM. (A) activation in the IFS, (B) activation in the Calcarine sulcus, (C) activation in the extrastriate regions V2–V4.

Figure 3.

Time course and average histograms of the fMRI activation in specific ROIs. The time course of activation (upper panel) was constructed by pooling significant voxels (GLM test: all > rest; FDR at q < 0.05, corrected for multiple comparisons) on a subject-by-subject basis and then averaging the activation in each condition across all subjects (n = 9). The gray area indicates the duration of the experimental block (12 s). The histograms (lower panel) show the average percent signal change in each ROI (averaging over 6–12 s after block onset). Asterisks denote significant difference between VG- and REP-related activation (paired t-test, P < 0.05) in each language. Error bars denote SEM. (A) activation in the IFS, (B) activation in the Calcarine sulcus, (C) activation in the extrastriate regions V2–V4.

ROI Selection

ROI selection (used in Fig. 3) was done on an individual basis: 1) Inferior frontal sulcus (IFS): Voxel selection for the IFS was based on an anatomical marker (IFS, obviously). The IFS of the individual subjects was easily identified using their high-resolution SPGR slices. 2) Calcarine sulcus: Voxels in this ROI were collected according to an anatomical marker; the calcarine sulcus including its upper and lower banks (2 mm for each side). The calcarine sulcus of the individual subjects was easily identified using their high-resolution SPGR slices. The approximate retinotopic borders (defined in a sighted subject and superimposed on a blind subject's Talairach normalized brain, as described above) were not used for the selection of significant voxels within the calcarine sulcus' ROI. 3) Extrastriate visual areas V2–V4 were assessed on a sighted subject, using the rotating wedge technique (Engel and others 1997). Briefly, a rotating wedge, made of contrast reversing checkerboard stimuli, circled around a central fixation point, completing a full circle in 36 s. For each voxel, the response phase was calculated by finding the lag that maximizes the cross-correlation function of the hemodynamic response function with the time series data. This allowed coarse assessment of the borders between the early visual areas (i.e., V2–V4) due to their retinotopic organization with respect to polar angle. The resulting retinotopic borders were superimposed on each of the blind subjects, to get a crude approximation of the extrastriate visual areas.

Results

Behavioral Data

RTs were measured one week before the scan. The mean RT during the Hebrew verb generation task (1421 ± 86 ms) was significantly longer than during the Hebrew repetition task (1023 ± 71 ms); paired t-test, P < 0.0001. Similarly, English verb generation task required longer RTs than the English repetition task (1634 ± 104 vs. 1058 ± 64 ms, respectively; P < 0.0001).

This reflects the comparative ease of the repetition task with respect to the verb generation task, within each language. Presumably, the increase in RT reflects the need for lexical search and retrieval that are an integral part of verb generation but are unnecessary during mere word repetition. Verb generation in English took significantly longer times than in Hebrew (t-test, P < 0.01), reflecting the fact that the subjects were less fluent in English. No significant difference was found between the Hebrew and English repetition tasks (P = 0.4), suggesting that the repetition task indeed required automatic repetition of auditory sounds with little semantic processing. A 2-way analysis of variance (ANOVA) for repeated measures revealed that both factors were statistically significant (language: F1,8 = 25.585, P < 0.001; task: F1,8 = 79.063, P < 0.000), as well as an interaction effect (language × task: F1,8 = 5.879, P < 0.042).

Quantitatively similar RTs were also obtained in the RTs measured 2 months after the scan: Hebrew verb generation (1460 ± 113 ms); Hebrew repetition task (1027 ± 73 ms); English verb generation (1648 ± 116 ms); English repetition task (1021 ± 78 ms). Again, a 2-way ANOVA for repeated measures revealed that both language and task were statistically significant (language: F1,8 = 34.393, P < 0.000; task: F1,8 = 120.094, P < 0.000; language × task: F1,8 = 16.412, P < 0.004). The subjects' performance immediately after the scan was almost perfect (93% correct ± 1% standard error of mean [SEM]), as well as 2 months after the scan (97% correct ± 1% SEM).

We also calculated the difference between the RTs for verb generation in English and Hebrew (as assessed a week before the scan, see Table 1) to obtain a measure of English level knowledge, independent of individual differences in response latency that may stem from other motor or cognitive elements (which are mutual to both tasks).

Functional Imaging Data

We investigated the pattern of cortical activation in 9 congenitally blind subjects under 4 different conditions. Data were analyzed at both the single-subject level and group analysis. Using group analysis, we first show the extent of the cortical activation elicited by the covert verbal task (verb generation) compared with the repeat task, separately for each language (Fig. 2). We follow this by a subject-by-subject analysis, focusing on specific ROIs in the IFS, the calcarine sulcus, and extrastriate retinotopic areas (Fig. 3) and comparing the activation elicited in these ROIs during the different experimental conditions. Next, we show on an individual subject basis the degree to which fMRI activation during verb generation was similar for the first (Hebrew) and second language (English) throughout the cortical sheet (Figs 4–6).

Group Analysis

We start by presenting the group results of the cortical activation in the blind, using a random-effects GLM analysis (Friston and others 1999). Activation maps are presented on a full Talairach-normalized (Talairach and Tournoux 1988) unfolded cortical sheet (as well as the inflated cortical representation above that). To assess the activation elicited by an active use of semantic processes, we compared the fMRI signal during covert verb generation with that evoked by the word repetition conditions, within each language. The first flattened maps (Fig. 2A) depict voxels that are significantly more active during Hebrew verb generation (i.e., Hebrew VG) than during the Hebrew repetition (i.e., Hebrew REP) task, across 9 subjects. Two major regions of widespread activation were found; the first one was in the left occipital cortex, spanning the greater share of both the ventral and dorsal parts of the visual retinotopic areas (including V1) and expanding to the object-related areas in occipitotemporal cortex. The second focus of activation was found in the inferior prefrontal cortex including the IFS, the inferior frontal gyrus, the precentral sulcus, and the anterior cingulate gyrus. (For full details, see Table 2.)

The second map (Fig. 2B) presents voxels showing differential fMRI activation during English VG, in comparison with that elicited by the English REP task. Activations were found in similar regions as in the case of the analogous comparison, in Hebrew, though somewhat to a lesser extent (see Table 3). The activation included again 2 main foci: one in the occipital areas including the calcarine sulcus and the other in the inferior frontal cortex, with a clear tendency for left lateralization.

Finally, no cluster of statistically significant voxels was found (at P < 0.05) when contrasting the fMRI activation elicited during Hebrew verb generation with that evoked by English verb generation (not shown). To summarize, left lateralized activation is seen in both the frontal and occipital cortices when either the first or second language are actively engaged in the verb generation task.

ROI Analysis

To confirm that differential activation for VG in each language can also be seen on an individual subject basis, we use the ROI approach, focusing on 3 main regions of activation found in the group analysis (the IFS, the calcarine sulcus, and extrastriate retinotopic regions, all in the left hemisphere). The calcarine sulcus and the IFS were individually identified in each of the subjects by their distinct anatomical features. Crude approximation of the extrastriate visual areas, corresponding roughly to areas V2–V4, was based on mapping of the horizontal and vertical meridians in a sighted subject using the rotating wedge technique (see Materials and Methods).

We first screened for voxels activated by all 4 conditions (vs. rest) within each of the ROIs (using single-subject GLM analysis with FDR at q < 0.05, and using a minimal cluster size of 10 functional voxels to correct for multiple comparisons). Note that because the selection of the voxels was based on having fMRI activation in all conditions (vs. rest), no a priori preference was given to one condition over the others. In each of the ROIs (of the left hemisphere), the average time course of activation across all selected voxels was first computed on a subject-by-subject basis and then averaged across subjects (an equal weight was assigned to each subject's data, the average number of active voxels within the ROI was 1407 [range: 555–2978 voxels]). The time course of BOLD responses for each condition as well as the average percent signal change (averaged across an interval spanning 6–12 s after stimulus onset) are shown in Figure 3.

An ROI analysis was performed for the IFS, an area that is activated in sighted subjects performing the same verb generation task (Petersen and others 1988, 1989; McCarthy and others 1993; Herholz and others 1996; Gabrieli and others 1998; Rowan and others 2004). Figure 3A depicts the BOLD activation of the significant voxels in the IFS. The differences between VG and REP in both languages did not reach significance. No significant difference was found between activation in the IFS during VG in the 2 languages, in congruence with a study using the same experimental paradigm in bilingual sighted subjects (Pu and others 2001). The results are summarized in the ANOVA in Table 4 (IFS).

Table 4

Two-way ANOVA for repeated measures

Effect F df Significance Partial eta squared 
IFS     
    Task 5.109 0.054 0.390 
    Language 0.040 0.846 0.005 
    Task × language 1.771 0.220 0.181 
Calcarine sulcus     
    Task 11.795 0.009 0.596 
    Language 0.049 0.831 0.006 
    Task × language 0.049 0.370 0.101 
V2–V4     
    Task 12.569 0.008 0.611 
    Language 3.981 0.081 0.332 
    Task × language 0.722
 
8
 
0.420
 
0.083
 
Effect F df Significance Partial eta squared 
IFS     
    Task 5.109 0.054 0.390 
    Language 0.040 0.846 0.005 
    Task × language 1.771 0.220 0.181 
Calcarine sulcus     
    Task 11.795 0.009 0.596 
    Language 0.049 0.831 0.006 
    Task × language 0.049 0.370 0.101 
V2–V4     
    Task 12.569 0.008 0.611 
    Language 3.981 0.081 0.332 
    Task × language 0.722
 
8
 
0.420
 
0.083
 

Note: df, degrees of freedom.

The activation profile and time course of the calcarine clusters are shown in Figure 3B. The overall picture resembles that of the IFS activation during the VG conditions (for either language) was significantly higher than during the REP conditions in the same language. Significant difference was found between Hebrew VG and Hebrew REP (t8 = 3.521, P < 0.009), as well as between English VG and English REP (t8 = 2.683, P < 0.03). No significant difference was found between the 2 languages. Thus, the differences here are independent of language but highly related to the nature of the task. See ANOVA in Table 4 (calcarine sulcus).

Finally, an ROI analysis was performed for the extrastriate retinotopic visual areas, loosely termed V2–V4 (Fig. 3C). Significant difference was found between English VG and English REP (t8 = 4.318, P < 0.004). A similar trend was evident in the comparison between Hebrew VG and Hebrew REP, though it did not reach significance. No significant difference was found between the languages. These results are captured by a significant task effect (and the lack of an interaction effect) in the ANOVA presented in Table 4 (V2–V4). To summarize, we find that in the congenitally blind, the left hemisphere of the occipital cortex mimics the patterns of activation found in the classical language areas. This is true for both native and later acquired languages.

Individual Subject Data

Having found significant calcarine activation in the group analysis, we next tested whether the differential activation (for VG vs. REP) can be found on an individual basis (see Fig. 4). In 6 of the 9 subjects, significant calcarine activation was seen when contrasting activation during Hebrew VG and Hebrew REP (P < 0.05, corrected for multiple comparisons); 4 of the 9 subjects showed significant preference in the second language (English VG > English REP). Thus, unlike the sighted (Burton, Snyder, Diamond, and Raichle 2002), a sizable fraction of the blind show occipital activation during a language-related task in both languages, but there is a great degree of variation in the occipital activation among the blind subjects.

Figure 4.

Patterns of differential occipital activation (VG > REP) in individual subjects. Subject-by-subject activation in the calcarine sulcus during Hebrew VG versus Hebrew REP and during English VG versus English REP. The upper figure shows a sagittal view of the brain of one subject focusing on significant voxels (P < 0.05, corrected for multiple comparisons) within the calcarine sulcus. Corresponding views of the calcarine sulcus in the other subjects (indicated by initials) are shown below. Significant calcarine activation was found in 6 of the 9 subjects for the Hebrew VG versus Hebrew REP contrast and in 4 subjects for the English VG versus English REP contrast. Anatomical markers: Calc calcarine sulcus; PO, parieto-occipital sulcus; Cer, cerebellum.

Figure 4.

Patterns of differential occipital activation (VG > REP) in individual subjects. Subject-by-subject activation in the calcarine sulcus during Hebrew VG versus Hebrew REP and during English VG versus English REP. The upper figure shows a sagittal view of the brain of one subject focusing on significant voxels (P < 0.05, corrected for multiple comparisons) within the calcarine sulcus. Corresponding views of the calcarine sulcus in the other subjects (indicated by initials) are shown below. Significant calcarine activation was found in 6 of the 9 subjects for the Hebrew VG versus Hebrew REP contrast and in 4 subjects for the English VG versus English REP contrast. Anatomical markers: Calc calcarine sulcus; PO, parieto-occipital sulcus; Cer, cerebellum.

Language Congruency Maps

Finally, we studied the degree of overlap in the activation elicited by verb generation in the first and second languages. This cannot be appreciated from the “group” cortical maps, shown in Figure 2, as an overlap between activation in the 2 languages may be masked in the group analysis (e.g., if the area of convergence is at different Talairach positions across subjects). Therefore, we generated statistical parameter maps of activation, elicited by verb generation versus repetition, in each language on an individual subject basis.

Figure 5A depicts individual congruency maps of activation in the left hemisphere in the 4 subjects that showed significant preference for verb generation over the word repetition task in both languages (using single-subject GLM analysis at P < 0.05, and using a minimal cluster size of 10 functional voxels to correct for multiple comparisons). Both the unfolded cortical sheets as well as medial and lateral inflated views of the left hemisphere are shown. Orange clusters denote voxels that are preferentially activated by Hebrew VG versus Hebrew REP conditions. Cyan voxels are significantly activated by English VG compared with English REP conditions. Finally, voxels showing preference for both Hebrew and English verb generations (using the 2 independent tests above) are depicted in green. As can be observed, typically, subjects who have robust occipital activation during performance of both language tasks show overlap in the activation patterns for the 2 languages.

Figure 5.

Individual subjects' cortical maps showing the extent of language-related activation for both languages and their overlap, using either (A) chimeric or (B) single-word stimuli. The maps show full Talairach-normalized unfolded brains of the left hemisphere of 4 congenitally blind subjects (OC, KD, LL, and LD), as well as medial and lateral inflated views of the left hemispheres. (A) Data from the original experiment using superimposed Hebrew and English nouns (i.e., chimeric stimuli). (B) Data from the single-word control experiment (in either Hebrew or English) performed on 2 of the 4 subjects (LL and LD). Orange blobs indicate clusters with preferential Hebrew VG activation compared with Hebrew REP. Cyan blobs indicate voxels activated only by English VG versus English REP conditions. Green blobs indicate voxels activated independently by both tests. The depicted voxels are ones above a threshold using a single-subject GLM analysis, (P < 0.05, corrected for multiple comparisons). Overall, the individual maps look very similar, showing greater fMRI activation during verb generation (compared with word repetition) in both languages and in both experimental settings.

Figure 5.

Individual subjects' cortical maps showing the extent of language-related activation for both languages and their overlap, using either (A) chimeric or (B) single-word stimuli. The maps show full Talairach-normalized unfolded brains of the left hemisphere of 4 congenitally blind subjects (OC, KD, LL, and LD), as well as medial and lateral inflated views of the left hemispheres. (A) Data from the original experiment using superimposed Hebrew and English nouns (i.e., chimeric stimuli). (B) Data from the single-word control experiment (in either Hebrew or English) performed on 2 of the 4 subjects (LL and LD). Orange blobs indicate clusters with preferential Hebrew VG activation compared with Hebrew REP. Cyan blobs indicate voxels activated only by English VG versus English REP conditions. Green blobs indicate voxels activated independently by both tests. The depicted voxels are ones above a threshold using a single-subject GLM analysis, (P < 0.05, corrected for multiple comparisons). Overall, the individual maps look very similar, showing greater fMRI activation during verb generation (compared with word repetition) in both languages and in both experimental settings.

Control Experiment

A possible disadvantage of the chimeric stimuli used (i.e., superposition of heard nouns in the 2 languages) is that even though the task requires the subject's attention to be focused on the word in the relevant language, the unattended word (in the other language) might be still processed to some extent. Thus, the activation seen for the second language (English) could actually be caused by word processing in the native language (Hebrew). To rule out this possibility, we constructed a simplified version of the experiment, in which during a given condition, nouns were heard in only one language. This has the disadvantage that there are clear differences in the auditory components between language conditions (though not between the verb generation and repetition conditions within each language), but obviously it rules out the potential problem of “unattended” processing in the other language. Figure 5B depicts unfolded and inflated maps in 2 of the 4 subjects that were examined. It can be seen that the general pattern of activation is similar in both versions of the experiment. Figure 6 shows the results of the ROI analysis (using the same individually defined ROIs) in each of the 2 subjects, for both the chimeric stimuli (Fig. 6A) and the single words (Fig. 6B) of the control experiment. The results clearly indicate that the occipital preference for verb generation (over word repetition) can be seen in both languages and in both experimental conditions. We therefore conclude that the occipital cortex of the congenitally blind is genuinely active during language processing in both the native language and a later acquired language.

Figure 6.

Comparison of the individual fMRI activation in the left calcarine sulcus evoked by (A) chimeric and (B) single-word stimuli. The time course of activation was constructed by pooling statistically significant voxels (GLM test: all conditions > rest; FDR at q < 0.05, corrected for multiple comparisons) separately for each subject. The gray area indicates the duration of each experimental block (12 s). The histograms show the average percent signal change in each ROI (averaging over 6–12 s after block onset). Asterisks denote significant difference between VG- and REP-related activation (t-test, P < 0.05) in each language. Error bars denote SEM across repetitions of the same condition in the course of the experiment (15 or 10 repetitions in subjects LL and LD, respectively). The greater fMRI activation during verb generation (compared with word repetition) can be seen in both languages and in both experimental settings.

Figure 6.

Comparison of the individual fMRI activation in the left calcarine sulcus evoked by (A) chimeric and (B) single-word stimuli. The time course of activation was constructed by pooling statistically significant voxels (GLM test: all conditions > rest; FDR at q < 0.05, corrected for multiple comparisons) separately for each subject. The gray area indicates the duration of each experimental block (12 s). The histograms show the average percent signal change in each ROI (averaging over 6–12 s after block onset). Asterisks denote significant difference between VG- and REP-related activation (t-test, P < 0.05) in each language. Error bars denote SEM across repetitions of the same condition in the course of the experiment (15 or 10 repetitions in subjects LL and LD, respectively). The greater fMRI activation during verb generation (compared with word repetition) can be seen in both languages and in both experimental settings.

Discussion

In this research, we present evidence that active usage of the second language, acquired at the age of 10 years, generates fMRI activation in the occipital cortex of the congenitally blind, similar to that evoked by active use of the mother tongue. Furthermore, between-subject analysis revealed no significant difference between the fMRI activation elicited by the 2 languages. These results confirm previous reports, showing widespread occipitotemporal activation in blind subjects performing language-related tasks (Burton, Snyder, Conturo, and others 2002; Burton, Snyder, Diamond, and Raichle 2002; Roder and others 2002; Amedi and others 2003; Burton and others 2003; Gizewski and others 2003). Previously, differential occipital activation was found when verb generation activation was compared with nonverbal conditions: performing a one-back task on auditory noise stimuli (Amedi and others 2003) or passive listening to reverse words (Burton, Snyder, Diamond, and Raichle 2002). Our study extends these findings by demonstrating that similar activation patterns can be seen when contrasting the active production of the verbs with repetition of the same words. Thus, our contrast highlighted areas that are putatively engaged in lexical search and retrieval.

In accordance with other studies in the blind, that utilized tasks involving the mother tongue (Burton, Snyder, Diamond, and Raichle 2002; Amedi and others 2003), our results show left lateralized cortical activation. Specifically, when contrasting the fMRI activation during verb generation with that observed during word repetition, we found lateralized activation in both the left occipital cortex and the left IFS, replicating previous results (Burton, Snyder, Conturo, and others 2002; Burton, Snyder, Diamond, and Raichle 2002). In this study, we provide evidence that such a lateralized pattern of activation can also be seen in the blind when using a later acquired language.

Our results are congruent with studies in sighted subjects during performance of the same verb generation task using the native language (Petersen and others 1988, 1989; Gabrieli and others 1998; Rowan and others 2004) as well as the second language (Klein and others 1995, 1999; Chee and others 1999). The lack of difference between the activation elicited by the 2 languages in the IFS is also in congruence with the results obtained in late (sighted) second language learners (Pu and others 2001).

Possible Confounding Factors

One problem in comparing the activation elicited by 2 languages might be that words in each language have different auditory features, which are characteristic of that language (e.g., some phonemes that exist in Hebrew are not used in English and vice versa). To avoid this, we generated the chimeric stimuli, which enable presentation of identical auditory stimuli in all conditions. The only physical difference between the conditions was the instruction before each block. This assured that any difference in the fMRI activation during verb generation in Hebrew versus English would stem from the verbal elements of the task.

Another point is that the greater occipital activation during verb generation (in either language) compared with the repetition task may reflect the relative ease of the repetition task rather than the use of lexical search and retrieval processes. However, our previous results do not support this task difficulty hypothesis (Raz and others 2005). In that study, the most difficult verbal task and the easiest task elicited similar levels of calcarine activation, which were significantly lower than the midlevel performance task.

Another issue is the potential use of translation strategies: For example, it is possible that when a subject is required to perform the verb generation task in the less proficient language (second language, English), he naturally translates the noun to his mother language (first language, Hebrew), and, after generation of the verb in the first language, translates it back to the second language. In that case, the observed occipital activation may be due to processes evoked by the mother tongue and not by the later acquired language. This might also explain why we did not find any differences in the pattern of activation between the 2 languages.

Although we cannot entirely rule out this possibility, we think it is unlikely. For one thing, all participants in our study were proficient bilinguals. There is evidence from various cognitive experiments (priming, stroop effect, etc.) that proficient bilinguals can access concepts directly from the second language without having to perform an internal translation via the first language (Kroll and deGrott 1997). Finally, all the subjects reported at the end of the experiment that they did not use translation in performing the tasks.

The last issue of concern in the experiment is that the activation observed may be due to processing of the “unattended” noun, that is, activation assigned to the attended second language (English) may actually result from implicit processing of the simultaneously presented words in the unattended first language (Hebrew). To rule out this possibility, we constructed a control experiment, in which single nouns in one language were heard (rather than 2 superimposed nouns in both languages, as in the original experiment). The results from 2 subjects who performed this experiment (Figs 5 and 6) were remarkably similar to the ones obtained from the same subjects in the original experiment. We therefore conclude that the occipital activation during the use of the later acquired language (English) is genuinely related to processing in that language.

Differences between the 2 Languages

Although no significant differences were found in the group analysis when contrasting directly the activation elicited by the 2 languages, the patterns of activation for the 2 languages were somewhat different. For example, Z-scores were generally smaller for English (Table 3) than for Hebrew verb generation (Table 2). The average RTs (across all subjects) were significantly longer during English verb generation than for its Hebrew counterpart. In addition, 6 out of the 9 subjects showed significantly greater V1 activation for the verb generation (compared with the repeat) in Hebrew, whereas in English, this was found only in 4 subjects. The 4 subjects showing significant activation during verb generation in English (compared with repeat) were ranked first, second, fourth, and eighth of the 9 subjects when testing their English fluency level (measured by the difference in RT for verb generation in the 2 languages). Overall, these findings suggest that both task and language factors influence the cortical activity in the calcarine sulcus.

Critical Period for Plasticity

Studies that have addressed the issue of the putative critical period for plasticity in the blind usually focus on comparison between 2 groups: one group that is deprived from birth, and a second group deprived only at an older age. The typical question is whether a novel function (or neuronal activation) that can be found in the congenitally deprived group can also be found in the later deprived individuals (that had a period of visual experience). If this function is not seen in the later-deprived group, this is taken as evidence for a critical period, in which changes in cortical function can occur.

Taking a slightly different approach to study the same question, instead of comparing 2 different groups of subjects, we studied only individuals who were congenitally blind. To address the issue of the critical period, the comparison we applied was between the activation evoked by 2 different languages acquired in 2 different time periods (i.e., mother tongue and second language). Our question was therefore whether neuronal populations that were never engaged in visual analysis (in the blind) and seem capable of processing input other than vision maintain their capability for plasticity across time.

Our comparison of activation elicited by the first and second languages in the congenitally blind corresponds to the usual comparison between the early and late blind groups that were typically used in previous studies. Both examine the degree to which plasticity can also be seen in the late phase. However, our methodology had certain advantages. First, because comparison was between the activation elicited by the 2 languages in the same subjects, variability between subjects in the 2 groups, which is substantial, is factored out. Second, because all subjects were congenitally blind with no visual experience, activation due to possible visual imagery, in the late blind, can be ruled out (Buchel and others 1998). Finally, the late and early blind often differ in their period of blindness (i.e., the time from blindness onset until the time of the study) and blindness development in time (gradual or abrupt). Thus, to determine the degree to which plasticity can also be seen in the late phase, we focused here only on the congenitally blind cohort, comparing the activation elicited by the 2 languages in the same subjects.

Our results are in congruence with studies comparing early blind, late blind, and sighted subjects, suggesting that nonvisually related activity can be seen in the occipital cortex of the blind, irrespective of the age of blindness onset. Burton and others (2003) compared fMRI activation during semantic versus phonologic tasks, for heard words. Only blind subjects showed such differential activation in the occipital cortex, but response magnitude in V1 remained relatively constant, irrespective of age of blindness onset (the average age of blindness onset in the late blind group was 18 years). Another study by Burton, Snyder, Diamond, and Raichle (2002) tested early blind, late blind (average age of blindness onset: 20 years), and sighted subjects in a covert verb generation task to heard nouns. Both groups of blind subjects showed a comparable spatial extent of V1 responses in the calcarine sulcus, as well as a similar magnitude of response (using ROI analysis). No such activation was found in sighted subjects. The authors concluded that the cortical plasticity persists after the age of 20 years, suggesting that adaptive changes to sensory deprivation are possible throughout life rather than during a limited period of susceptibility.

In contradistinction to these 2 studies, when examining the question of the critical period using tasks involving tactile stimuli, plasticity seems to be restricted to a clear critical period. Sadato and others (Sadato and others 2002) examined blind subjects with various ages of onset of blindness, in a tactile discrimination task (same/different judgment on a pair of 2 dot Braille letters). They found that V1 was activated in blind subjects who lost their sight before 16 years of age, whereas it was suppressed in subjects who lost their sight after 16 years of age. It should be noted, however, that 5 out of 6 of the late blind subjects in this study had some binocular visual sensitivity to light or hand movements or both. This retained visual function might have precluded adaptive changes to nonvisual modalities in the late blind group. Similarly, Cohen and others (1999) used repetitive TMS to induce transient disruption of cortical function during a Braille identification task. Stimulation of the occipital cortex disrupted the task in congenitally blind and early-onset blind subjects but not in late-onset blind subjects. The authors suggested that the susceptible period for this form of cross-modal plasticity does not extend beyond 14 years.

In order to dissociate between language processes and tactile aspects that are confounded in Braille reading, Burton (Burton and others 2004) conducted another study, using a tactile flutter vibration task (processing stimuli independent of language). Subjects were instructed to indicate whether a pair of consecutive vibrations of the index finger is at the same frequency or not (one-back, same/different judgment). They found that all early blind subjects had statistically significant fMRI activation in V1, whereas only 3 out of 9 late blind subjects and 3 of the 9 sighted subjects had significant activation in V1. The authors concluded that visual cortex plasticity is inversely proportional to the age of blindness onset.

It is currently unclear why there are such discrepancies regarding the length of the critical period in the blind. The general picture is that tasks requiring language processing typically activate the occipital cortex in both early and late blind subjects, suggesting that there is no critical period, whereas results from tactile discrimination tasks suggest the opposite (although for evidence for rapid plasticity in adult blindfolded subjects after 5 days, see Pascual-Leone and others 2005). One possibility is that the plasticity is governed by the task requirements. For example, if the development of tactile skills follows a different time schedule than that of language processing, it may be reflected by a different window of susceptibility. Clearly, further research is required to determine the factors that affect the degree of cortical plasticity over time.

Conclusions

Taken together, the results of the present study may be interpreted in one of 2 ways: One interpretation is that the critical period for plasticity persists (at least) until the age of 10 years because we find patterns of occipital activation in the blind during use of both native and later acquired languages. This, however, does not rule out the possibility that indeed there is a critical period for plasticity until the age of puberty (∼15 years), as has been implicated by previous work showing occipital activation during tactile discrimination in the blind (Cohen and others 1999; Sadato and others 2002). To resolve this question, it would be necessary to examine early blind subjects who acquired their second language during adulthood, at an age clearly beyond the range of the suggested critical period for susceptibility.

An alternative interpretation of our results is that the visual cortex of the blind might be engaged in abstract levels of language processing, common to the 2 languages (such as the semantics). This is in line with the idea of reorganization according to a reverse hierarchy principle, in which in the absence of visual input, normally low-level visual regions are recruited for high-level cognitive abstraction (Amedi and others 2003; Buchel 2003; Burton and others 2003; Raz and others 2005). Further research is needed to elucidate the exact nature of the hierarchy among the various occipitotemporal regions in the blind.

We thank N. Raz for her help throughout this study, A. Amedi for insightful suggestions, T. Orlov for the help with the 3-dimensional cortex reconstructions, and I. Rabinowitch and G. Jacobson for help with stimuli preparations. This study was funded by the McDonnell-Pew foundation grant #220020046. Conflict of Interest: None declared.

References

Amedi
A
Floel
A
Knecht
S
Zohary
E
Cohen
LG
Transcranial magnetic stimulation of the occipital pole interferes with verbal processing in blind subjects
Nat Neurosci
 , 
2004
, vol. 
7
 (pg. 
1266
-
1270
)
Amedi
A
Raz
N
Pianka
P
Malach
R
Zohary
E
Early ‘visual’ cortex activation correlates with superior verbal memory performance in the blind
Nat Neurosci
 , 
2003
, vol. 
6
 (pg. 
758
-
766
)
Boynton
GM
Engel
SA
Glover
GH
Heeger
DJ
Linear systems analysis of functional magnetic resonance imaging in human V1
J Neurosci
 , 
1996
, vol. 
16
 (pg. 
4207
-
4221
)
Buchel
C
Cortical hierarchy turned on its head
Nat Neurosci
 , 
2003
, vol. 
7
 (pg. 
657
-
658
)
Buchel
C
Price
C
Frackowiak
RS
Friston
K
Different activation patterns in the visual cortex of late and congenitally blind subjects
Brain
 , 
1998
, vol. 
121
 (pg. 
409
-
419
)
Burton
H
Diamond
JB
McDermott
KB
Dissociating cortical regions activated by semantic and phonological tasks: a fMRI study in blind and sighted people
J Neurophysiol
 , 
2003
, vol. 
90
 (pg. 
1965
-
1982
)
Burton
H
Sinclair
RJ
McLaren
DG
Cortical activity to vibrotactile stimulation: an fMRI study in blind and sighted individuals
Hum Brain Mapp
 , 
2004
, vol. 
23
 (pg. 
210
-
228
)
Burton
H
Snyder
AZ
Conturo
TE
Akbudak
E
Ollinger
JM
Raichle
ME
Adaptive changes in early and late blind: a fMRI study of Braille reading
J Neurophysiol
 , 
2002
, vol. 
87
 (pg. 
589
-
607
)
Burton
H
Snyder
AZ
Diamond
JB
Raichle
ME
Adaptive changes in early and late blind: a fMRI study of verb-generation to heard nouns
J Neurophysiol
 , 
2002
, vol. 
88
 (pg. 
3359
-
3371
)
Chee
MWL
Tan
EDL
Thiel
T
Mandarin and English single word processing studied with functional magnetic resonance imaging
J Neurosci
 , 
1999
, vol. 
19
 (pg. 
3050
-
3056
)
Cohen
LG
Celnik
P
Pascual-Leone
A
Corwell
B
Faiz
L
Dambrosia
J
Honda
M
Sadato
N
Gerloff
C
Catala
MD
Hallett
M
Functional relevance of cross-modal plasticity in blind humans
Nature
 , 
1997
, vol. 
389
 (pg. 
180
-
183
)
Cohen
LG
Weeks
RA
Sadato
N
Celnik
P
Ishii
K
Hallett
M
Period of susceptibility for cross-modal plasticity in blind
Ann Neurol
 , 
1999
, vol. 
45
 (pg. 
451
-
460
)
Cox
RW
AFNI: software for analysis and visualization of functional magnetic resonance neuroimages
Comput Biomed Res
 , 
1996
, vol. 
29
 (pg. 
162
-
173
)
Engel
SA
Glover
GH
Wandell
BA
Retinotopic organization in human visual cortex and the spatial precision of functional MRI
Cereb Cortex
 , 
1997
, vol. 
7
 (pg. 
181
-
192
)
Forman
SD
Cohen
JD
Fitzgerald
M
Eddy
WF
Mintun
MA
Noll
DC
Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster-size threshold
Magn Reson Med
 , 
1995
, vol. 
33
 (pg. 
636
-
647
)
Friston
KJ
Holmes
AP
Worsley
KJ
How many subjects constitute a study?
Neuroimage
 , 
1999
, vol. 
10
 (pg. 
1
-
5
)
Gabrieli
JDE
Poldrack
RA
Desmond
JE
The role of left prefrontal cortex in language and memory
Proc Natl Acad Sci USA
 , 
1998
, vol. 
95
 (pg. 
906
-
913
)
Gizewski
ER
Gasser
T
de Greiff
A
Boehm
A
Forsting
M
Cross-modal plasticity for sensory and motor activation patterns in blind subjects
Neuroimage
 , 
2003
, vol. 
19
 (pg. 
968
-
975
)
Herholz
K
Wienhard
TK
Pietrzyk
U
von Stockhausen
HM
Karbe
H
Kessler
J
Bruckbauer
T
Halber
M
Heiss
WD
Individual functional anatomy of verb generation
Neuroimage
 , 
1996
, vol. 
3
 (pg. 
185
-
194
)
Hubel
DH
Wiesel
TN
The period of susceptibility to the physiological effects of unilateral-eye closure in kittens
J Physiol (Lond)
 , 
1970
, vol. 
206
 (pg. 
419
-
436
)
Klein
D
Milner
B
Zatorre
RJ
Meyer
E
Evans
AC
The neural substrates underlying word generation: a bilingual functional-imaging study
Proc Natl Acad Sci USA
 , 
1995
, vol. 
92
 (pg. 
2899
-
2903
)
Klein
D
Milner
B
Zatorre
RJ
Zhao
V
Nikelski
J
Cerebral organization in bilinguals: a PET study of Chinese-English verb generation
Neuroreport
 , 
1999
, vol. 
10
 (pg. 
2841
-
2846
)
Kroll
JF
deGrott
AMB
Kroll
JF
deGroot
AMB
Lexical and conceptual memory in bilingual: mapping form to meaning in two languages
Tutorials in bilingualism: psycholinguistic perspectives
 , 
1997
Malwah, NJ
Lawrence Erlbaum Associates
(pg. 
169
-
199
)
Lambert
S
Sampaio
E
Mauss
Y
Scheiber
C
Blindness and brain plasticity: contribution of mental imagery? An fMRI study
Brain Res Cogn Brain Res
 , 
2004
, vol. 
20
 (pg. 
1
-
11
)
McCarthy
G
Blamire
AM
Rothman
DL
Gruetter
R
Shulman
RG
echo-planar magnetic resonance imaging studies of frontal cortex activation during word generation in humans
Proc Natl Acad Sci USA
 , 
1993
, vol. 
90
 (pg. 
4952
-
4956
)
Noppeney
U
Friston
KJ
Price
CJ
Effects of visual deprivation on the organization of the semantic system
Brain
 , 
2003
, vol. 
126
 (pg. 
1620
-
1627
)
Pascual-Leone
A
Amedi
A
Frengi
F
Merabet
LB
The plastic human brain cortex
Annu Rev Neurosci
 , 
2005
, vol. 
28
 (pg. 
377
-
401
)
Petersen
SE
Fox
PT
Posner
MI
Mintun
M
Raichle
ME
Positron emission tomographic studies of the cortical anatomy of single-word processing
Nature
 , 
1988
, vol. 
31
 (pg. 
585
-
589
)
Petersen
SE
Fox
PT
Posner
MI
Mintun
M
Raichle
ME
Positron emission tomographic studies of the processing of single words
J Cogn Neurosci
 , 
1989
, vol. 
1
 (pg. 
153
-
170
)
Pu
Y
Liu
HL
Spinks
JA
Mahankali
S
Xiong
J
Feng
CM
Tan
LH
Fox
PT
Gao
JH
Cerebral hemodynamic response in Chinese (first) and English (second) language processing revealed by event-related functional MRI
Magn Reson Imaging
 , 
2001
, vol. 
19
 (pg. 
643
-
647
)
Raichle
ME
Fiez
JA
Videen
TO
MacLeod
AK
Pardo
JV
Fox
PT
Petersen
SE
Practice-related changes in human brain functional anatomy during nonmotor learning
Cereb Cortex
 , 
1994
, vol. 
4
 (pg. 
8
-
26
)
Raz
N
Amedi
A
Zohary
E
V1 activation in congenitally blind humans is associated with episodic retrieval
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
1459
-
1468
)
Roder
B
Stock
O
Bien
S
Neville
H
Rosler
F
Speech processing activates visual cortex in congenitally blind humans
Eur J Neurosci
 , 
2002
, vol. 
16
 (pg. 
930
-
936
)
Rowan
A
Liegeois
F
Vargha-Khadem
F
Gadian
D
Connelly
A
Baldeweg
T
Cortical lateralization during verb generation: a combined ERP and fMRI study
Neuroimage
 , 
2004
, vol. 
22
 (pg. 
665
-
675
)
Sadato
N
Okada
T
Honda
M
Yonekura
Y
Critical period for cross-modal plasticity in blind humans: a functional MRI study
Neuroimage
 , 
2002
, vol. 
16
 (pg. 
389
-
400
)
Sadato
N
Pascual-Leone
A
Grafman
J
Deiber
MP
Ibanez
V
Hallett
M
Neural networks for Braille reading by the blind
Brain
 , 
1998
, vol. 
121
 (pg. 
1213
-
1229
)
Sadato
N
Pascual-Leone
A
Grafman
J
Ibanez
V
Deiber
MP
Dold
G
Hallett
M
Activation of the primary visual cortex by Braille reading in blind subjects
Nature
 , 
1996
, vol. 
380
 (pg. 
526
-
528
)
Talairach
J
Tournoux
P
Co-planar stereotaxic atlas of the human brain
 , 
1988
New York
Thieme