Left middle cerebral artery (MCA) stroke is known to result in significant speech and language deficits in adults, whereas infants and young children with left hemisphere lesions rarely develop persistent aphasic disorders. The resilience of speech and language function in the event of early left hemisphere injury (before 5-10 years of age) has been attributed to enhanced plasticity of the young brain. Recovery of language following brain injury is presumably related to reorganization of functional networks whereby intact brain regions assume new functions to compensate for damaged regions. This capacity for reorganization is thought to decrease with age, resulting in better functional outcomes the younger the age at injury.Reference Woods and Teuber 1
Several functional neuroimaging studies suggest that, in adults with left MCA stroke, right hemisphere language dominance is sometimes observed in the acute poststroke stage but rarely persists over time in those with good outcome.Reference Woods and Teuber 1 - Reference Xu, Zhang, Shang, Wang, Luo and Weng 7 In contrast, right hemisphere language dominance appears to be much more common in cases of large left MCA lesions caused by perinatal strokeReference Heller, Heier and Watts 8 - Reference Ilves, Tomberg and Kepler 14 or left hemisphere lesions accompanied by seizure disorder in the first 5 years of lifeReference Liégeois, Connelly and Cross 15 - Reference Müller, Rothermel and Behen 18 These findings are consistent with the notion of a “critical period” for left hemisphere language specializationReference Hurford 19 - Reference Vargha-Khadem, Isaacs, Papaleloudi, Polkey and Wilson 21 and evidence that left hemisphere language dominance begins very early in childhood and continues to strengthen into adolescence.Reference Holland, Plante, Weber Byars, Strawsburg, Schmithorst and Ball 22
However, the literature on language reorganization following early brain injury is not entirely consistent with the notion that language function is transferred to the “nondominant” hemisphere. In a large heterogeneous group of children with epilepsy and left hemisphere lesions of various etiologies with onset of epilepsy before 18 years of age, Anderson et alReference Anderson, Harvey and Saling 23 found that the majority exhibited left language dominance and that atypical language lateralization was not associated with age at onset or handedness. Liegeois et alReference Liégeois, Connelly and Cross 15 found that early lesions in classical language areas (ie. Broca’s and Wernicke’s areas) were not associated with right hemisphere language transfer but, unexpectedly, right hemisphere transfer was observed when damage involved other left hemisphere regions. Furthermore, some studies using cortical stimulationReference Duchowny, Jayakar and Harvey 24 and intracarotid amobarbital procedureReference DeVos, Wyllie, Geckler, Kotagal and Comair 25 have also reported perilesional activation and/or recruitment of novel left hemisphere regions in patients with early left hemisphere lesions. Finally, Raja Beharelle et alReference Raja Beharelle, Dick and Josse 26 found a positive correlation between left frontal lateralization and language function in individuals with a history of left perinatal stroke, even though the group as a whole showed increased right hemisphere lateralization relative to controls.
Very few studies have contrasted language reorganization following perinatal and later childhood stroke. Ilves et alReference Ilves, Tomberg and Kepler 14 found a right hemisphere shift in language dominance following perinatal stroke, but continued left hemisphere dominance following stroke in later childhood. However, childhood stroke was associated with poorer language outcome than perinatal stroke, making it difficult to determine if differences in language dominance were due to the timing of the stroke or the degree to which language developed normally. Moreover, because participants were tested as children, this study does not provide insight into differences in mature language representation following perinatal or childhood stroke.
A growing body of evidence suggests that higher level language and cognitive deficits may emerge gradually as children with early brain injuries grow older, and that earlier age at lesion is actually associated with poorer long-term outcome.Reference Dennis 27 , Reference Nass 28 Specifically, although younger children do to tend to recover more quickly from acute-stage language deficits then older children or adults, many go on to develop significant problems with more sophisticated language skills as they get older, such as narrative discourse, written expression, and verbal fluency.Reference Dennis 29 - Reference Woods and Carey 31 Deficits in these more subtle, high-level language abilities can be very disabling for stroke survivors in the workplace, at school, and in social situations. Thus, enhanced plasticity is not necessarily associated with more positive outcome if compensatory mechanisms are ineffective or disrupt the development of higher level skills.Reference Kolb 32 , Reference Vargha-Khadem 33 For this reason, it is important to obtain long-term language outcome measures when examining the issue of language reorganization following early brain injury.
The goal of this case series study is to expand on the existing research literature by comparing patterns of successful language reorganization following left MCA stroke in early childhood and adulthood. Using verb generation and picture-word matching tasks in the functional magnetic resonance imaging (fMRI) scanner, we studied four young adults with a history of left MCA stroke before age 7 and five older adults with left MCA stroke after the age of 40, all of whom demonstrated good language outcome on standardized tests. Both of these tasks are among the most commonly used fMRI tasks for studying language.Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Price and Crinion 6 , Reference Pizzamiglio, Galati and Committeri 34 - Reference Schlaggar, Brown, Lugar, Visscher, Miezin and Petersen 36 Moreover, because language tasks rely on a different constellation of language-related abilities, use of multiple tasks is important for obtaining a more comprehensive view of language representation.Reference Barnett, Marty-Dugas and McAndrews 37 , Reference Gaillard, Balsamo and Xu 38 Here, we have chosen one task that has been shown to be strongly left lateralizing (verb generation) and one that has been found to activate more bilaterally (picture-word matching).Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Price and Crinion 6 , Reference Rijntjes and Weiller 35 , Reference Schlaggar, Brown, Lugar, Visscher, Miezin and Petersen 36 Ours is the first fMRI study to directly compare language lateralization following childhood and adult stroke. Based on neuroimaging findings from control children and children with seizure disorders and left hemisphere lesions of mixed etiology,Reference Vikingstad, Cao, Thomas, Johnson, Malik and Welch 11 , Reference Liégeois, Connelly and Cross 15 , Reference Holland, Plante, Weber Byars, Strawsburg, Schmithorst and Ball 22 , Reference Müller, Rothermel, Behen, Muzik, Mangner and Chugani 39 we expected to find stronger recruitment of bilateral regions in the childhood stroke group compared with the adult stroke group. However, because our participants sustained left hemisphere stroke after some initial language acquisition (after 2 years of age), we expected to see less pronounced right hemisphere language representation than has been reported in children with large perinatal left hemisphere stroke.Reference Tillema, Byars and Jacola 10 , Reference Staudt, Grodd, Niemann, Wildgruber, Erb and Krägeloh-Mann 12 - Reference Ilves, Tomberg and Kepler 14
Methods
Participants
Childhood Stroke
Participants were recruited from the Children’s Stroke Program at The Hospital for Sick Children in Toronto. Inclusion criteria for the pediatric stroke group were: single left MCA infarct sustained after the neonatal period (28 days) but before 13 years of age; native English speaker; and current age of at least 17 years. Four young adults (17-21 years old) with a history of left MCA stroke between 2 and 7 years of age met all criteria and participated in the study. Demographic and neurological information is presented in Table 1. All participants sustained lesions of the left basal ganglia and perisylvian cortex was spared in all except subject C3. This pattern of damage (lesioned basal ganglia and/or thalamus with cortical sparing) is much more common than isolated cortical lesions following MCA infarcts in children and adolescents because of the underlying mechanism of the stroke, which is typically a transient inflammation of the MCA stem.Reference Cardenas, Rho and Kirton 40 - Reference DeVeber, Kirton and Booth 43 However, because the young brain typically has better overall perfusion and collateral blood flow than the older brain, children and adolescents are less likely to sustain large cortical infarcts in addition to the subcortical damage. All participants demonstrated aphasic deficits in the acute period but showed substantial recovery and subsequent development of spoken language skills (see the following section for performance on standardized language tests). All participants were native right-handers, though participants C2 and C3 switched to left-hand dominance following the stroke. None had global intellectual impairment, epilepsy, congenital heart disease, sickle cell disease, malignancy, recurrent stroke, or other neurological comorbidities.
Patient demographics and clinical profiles
F=female; L=left; M=male.
Adult Stroke
Participants were recruited from the Stroke Program at Toronto Western Hospital. To be included in the adult stroke group, participants must have sustained a single left MCA infarct sustained after 30 years of age and be a native English speaker. Exclusion criteria were neurological comorbidities, sickle cell disease, moyamoya disease, current seizure disorder, congenital or acquired heart disease, malignancy, bilateral lesions, and stroke recurrence. Five older adults (42-57 years of age) who sustained a left MCA stroke after the age of 40 also participated in the study. All sustained damage to the left perisylvian cortex with subcortical sparing. Participants were all at least 1 year poststroke at the time of the study. All are right-handed native English speakers and experienced aphasic deficits in the acute stroke period. All participants made a good recovery of spoken language skills, as evidenced by standardized test performance (see the following section). Exclusion criteria were recurrent stroke, dementia, psychiatric disturbance, epilepsy, sickle cell disease, cardiac disease, or other neurological disorders. The demographic and neurological profiles of these patients are presented in Table 1.
Controls
A control group of ten right-handed healthy adults (mean age, 50.4 years; standard deviation, 3.4 years) also participated. All are native English speakers with no neurological or psychiatric history.
Neuropsychological Assessment
Before the scanning session, patients and controls underwent a brief neuropsychological assessment consisting of the two-subtest version of the Wechsler Abbreviated Scale of Intelligence; the Boston Diagnostic Aphasia Exam, short form; letter and category fluency; the Wechsler Test of Adult Reading; and the Pyramids and Palm Trees Test of semantic memory. Neuropsychological test scores are presented in Table 2.
Neuropsychological profiles, experimental task performance, and LIs for patients
Unless otherwise indicated, all psychometric test scores have been translated into standard scores (mean, 100; standard deviation, 15).
BDAE=Boston Diagnostic Aphasia Examination, short form; WASI=Wechsler Abbreviated Scale of Intelligence; WTAR=Wechsler Test of Adult Reading.
fMRI Tasks
Verb Generation
Participants were asked to silently generate a verb that is associated with a particular concrete noun (e.g. dog – BARKS). The baseline condition involved silent viewing of symbol strings (e.g. ++$#$), each five items long; subjects were asked to fixate on the stimuli but make no response. Stimuli were presented at the rate of one every 4 seconds in blocks of five, for a total of six blocks. Thus, there were 30 trials for each of the verb generation and baseline conditions. A practice session with ten different stimuli was carried out before the scanning session. After the scanning session, participants were asked to perform the verb generation task again outside of the scanner, with the same stimuli, to ensure that they were performing correctly. This postscanning test session provided an estimate of response accuracy, though we recognize that this introduces the possibility of practice effects.
Picture-Word Matching
Participants were asked to press the left or right mouse button to indicate which of two words matched a black-and-white line drawing. There were a total of 30 drawings depicting common items from a variety of categories including animals, foods, furniture, vehicles, and tools. In the baseline condition, participants were asked to indicate which symbol string matched the large symbol in the center of the stimulus. Stimuli were presented at the rate of one every 4 seconds in blocks of five, for a total of six blocks (30 trials for each condition). A practice session was performed with each subject before entering the scanner. Tasks were run in a standard order for all participants.
This project was approved by Research Ethics Boards at the University Health Network and The Hospital for Sick Children, and all data were collected and handled in accordance with ethical requirements.
Data Analysis
Image Acquisition
Images were obtained on a 1.5 T EchoSpeed MRI (GE Medical Systems, Milwaukee, WI). The head of each participant was stabilized in a standard quadrature head coil with a pillow to minimize motion artifact. A mirror was adjusted to ensure that participants could easily see the projection screen on which the stimuli were displayed. A laptop computer with E-PRIME software generated the visual stimuli and recorded response accuracy and reaction time. A magnet-compatible three-buttoned mouse was used to record responses. T1-weighted structural images (120 axial slices; slice thickness, 11.5 mm; field of view, 200 mm) followed by T2*-weighted functional images (spiral acquisition) sensitive to blood oxygenation level-dependent contrast were obtained for each participant (25 axial slices; slice thickness, 4.4 mm; repetition time, 2 s; flip angle , 82°; echo time, 40 ms; field of view, 200 mm).
Image Analysis
Functional images were preprocessed and analyzed with the Statistical Parametric Mapping-2 software package. 44 Images were realigned to correct for motion, corrected for within-frame time of acquisition, spatially normalized, and smoothed with a 7.6-mm full-width half-max Gaussian kernel. Functional maps were overlaid onto T1 anatomical scans. Activation maps were generated by contrasting each task condition with a baseline condition (viewing symbol strings for verb generation, symbol matching for picture-word matching) at a corrected threshold of p=0.05. Random-effects analysis was used for the control data. A family-wise error corrected threshold of p<0.05 was used for controls and patients.
One global indicator that weReference Barnett, Marty-Dugas and McAndrews 37 and othersReference Cao, Vikingstad and George 3 , Reference Raja Beharelle, Dick and Josse 26 , Reference Catani, Jones and Ffytche 45 , Reference Rutten, Ramsey, van Rijen and van Veelen 46 have used to describe the relative contribution of left and right hemispheres to language task I is the Laterality Index, defined as (L-R)/(L+R), where L represents the number of activated voxels in the left hemisphere and R represents the number of activated voxels in the right hemisphere. Lateralization indices (LIs) were calculated by group for each task, using the entire supratentorial region for each hemisphere, using voxels that were greater for the task condition than the control condition. Only voxels that exceed threshold (in this case, family-wise error of p<0.05) are entered in to the equation. Thus, LIs can range from −1.0 to +1.0; by convention, values between −0.2 and +0.2 represent bilateral language distribution; values between −0.2 and −1.0 represent right hemisphere dominance; and values between +0.2 and +1.0 represent left hemisphere dominance. Values between ±0.5 and ±1.0 are considered to reflect strong hemispheric dominance .Reference Cao, Vikingstad and George 3 , Reference Catani, Jones and Ffytche 45 , Reference Rutten, Ramsey, van Rijen and van Veelen 46
Results
Behavioral Data
Controls performed at ceiling on both tasks (mean, 99.2%; standard deviation, 1.15; corrected for both). In the verb generation task, response accuracy was defined as generating an appropriate verb within the allotted time period. In the picture-word matching task, response accuracy was defined as the selection of the correct match. As shown in Table 2, all patients performed above 85% correct on both tasks, with the exception of adult patient A5 (76% correct on verb generation).
LIs
LIs for controls and the two patient groups are shown in Figures 1 and 2. Controls showed left lateralized activation for both tasks, although much more pronounced for verb generation. The adult stroke patients also showed left lateralization for both tasks, with no statistically significant difference from controls in a non-parametric Mann-Whitney U test (Verb Generation: t(13)=−1.92, p=NS; Picture Naming: t(13)=−1.35, p=NS). In contrast, the childhood stroke patients exhibited significantly weaker LIs than the adult group for verb generation (t(7)=2.76, p<0.01), but there was no significant group difference for picture-word matching (t(7)=0.687, p=NS).
Activation patterns for verb generation>baseline in controls, adult stroke (A1-A5), and childhood stroke (C1-C4). LIs are presented below each subject’s activation map.
Activation patterns for picture-word matching>baseline in controls, adult stroke (A1-A5,) and childhood stroke (C1-C4). LIs are presented below each subject’s activation map.
LIs for individual participants are presented in Table 2 and language maps are shown in Figures 3 and 4. In the adult stroke group, three of the five patients showed strong left lateralization on verb generation (+0.64, +0.74), whereas the other two showed moderate left lateralization (+0.21, +0.34). On picture-word matching, two of the adult stroke patients showed strong left lateralization (+0.55, +0.59), two showed bilateral activation (+0.10, +0.15) and one showed strong right lateralization (-0.56). In the childhood stroke group, two patients showed moderate left lateralization on verb generation (+0.22, +0.31) and two patients showed bilateral activation (+0.07, −0.18). On picture-word matching, three of the childhood stroke patients showed bilateral activation (+0.11, +0.13, +0.14) and one showed moderate right lateralization (−0.45).
Box plots depicting the variability in LIs for controls, adult stroke (A1-A5), and childhood stroke (C1-C4) on the verb generation task.
Box plots depicting the variability in LIs for controls, adult stroke (A1-A5), and childhood stroke (C1-C4) on the picture-word matching task.
Regions of Significant Activation: Controls
As Table 3 indicates, control participants performing the verb generation task activated a network of left hemisphere regions including inferior frontal, middle frontal, superior frontal, inferior parietal, superior temporal, and cingulate. Broca’s area emerged as the clear point of maximum activation. A few right hemisphere regions were also activated (inferior frontal, superior frontal, superior temporal, cingulate), although to a much lesser extent than their left hemisphere homologues. In the picture-word matching task, controls activated left hemisphere regions including inferior and medial frontal, middle temporal, cingulate, dorsomedial thalamus, and parahippocampal gyrus. Maximum activations occurred in left inferior and medial frontal areas, but to a lesser extent than in the verb generation task. Bilateral activation was observed in several regions including superior temporal, cuneus, and fusiform/lingual gyrus.
Areas activated by control subjects during verb generation and picture-word matching
Talairach coordinates (x, y, z) are reported.
Regions of Significant Activation: Adult Stroke
In the verb generation task (Table 4, Fig. 1), all five of the adult stroke patients showed left inferior frontal activation but, unlike controls, this was not the region of maximum activation. Rather, the strongest activations were seen in other regions of left frontal cortex (middle, medial, and superior frontal) and, in some cases, left temporal regions. Of note, only two of the adult stroke patients showed activation of Broca’s area (BA 44/45); the other three demonstrated activation other areas within the left inferior and middle frontal gyri (BA 46/9). In contrast to controls, none of the adult stroke patients activated the left angular gyrus and only two patients showed left inferior parietal activation. Moreover, right frontal regions and bilateral occipital/fusiform regions were activated in the adult stroke patients to a greater extent than the controls. Overall, activation in the adult stroke patients was more diffuse, with no consistent region of maximum activation across the group. In the picture-word matching task (Table 4, Fig. 2), adult stroke patients again showed more diffuse activation and more bilateral activation than controls. Nonetheless, strong left frontal activations were observed in all patients.
Areas activated by adult stroke patients during verb generation and picture-word matching
Brodmann areas are shown in parentheses. Talairach coordinates (x, y, z) are reported.
Regions of Significant Activation: Childhood Stroke
In the verb generation task (Table 5, Fig. 1), the childhood stroke patients activated many of the same regions as controls and the adult stroke patients, but activations tended to be bilateral rather than left dominant. Moreover, in contrast to the other two groups, maximum activations in the childhood stroke patients were seen in anterior cingulate (BA 24) and occipitotemporal regions, not in left frontal regions. Of note, all four childhood stroke patients activated the left and right inferior frontal gyrus to a similar extent, whereas all but one of the adult patients (A5) showed stronger left than right inferior frontal activation. None of the childhood stroke patients activated left angular gyrus and only two activated left inferior parietal cortex, though two showed right inferior parietal activation. In the picture-word matching task (Table 5, Fig. 2), maximum activations were seen in posterior regions (particularly right occipitotemporal) and, in two patients, right cingulate gyrus. Frontal activations were bilateral or predominantly right-sided. This was in contrast to the strong left frontal activations shown by controls and all of the adult stroke patients.
Areas activated by childhood stroke patients during verb generation and picture-word matching
Brodmann areas are shown in parentheses. Talairach coordinates (x, y, z) are reported.
Discussion
This is the first fMRI study to compare patterns of language lateralization following left MCA stroke in childhood and adulthood. Our study was also unique in that all participants were tested as adults, making it possible to directly compare “mature” patterns of language lateralization after childhood and adult MCA stroke in the language-dominant hemisphere. Controls showed left lateralized activation during verb generation and, to a lesser extent, during picture-word matching. Activation patterns for patients, particularly the childhood stroke group, were generally more diffuse than controls, and involved more bilateral and posterior regions. Adult stroke patients showed left lateralization for both tasks, with no statistically significant difference from controls, consistent with previous studies.Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Cao, Vikingstad and George 3 , Reference Hillis, Wityk and Tuffiash 47 Of note, however, the most robust activation for the adult stroke patients during verb generation was in left inferior and medial frontal regions, but outside of BA, which was the site of maximal activation in controls. Childhood stroke patients exhibited significantly weaker lateralization than the adult group for verb generation, but there was no significant group difference for picture-word matching. Maximum activations for the childhood stroke patients involved occipitotemporal cortex and anterior cingulate, in contrast to the strong left frontal activations of the adult group and controls. Our findings suggest that reorganization of language function may be more likely to involve bilateral recruitment following left MCA stroke in childhood than in adulthood. In our small sample, bilateral reorganization was most apparent on a task that is strongly left lateralized in controls (verb generation). Moreover, more diffuse left hemisphere engagement was seen in both the stroke groups compared with controls.
Age at Stroke and Language Lateralization
Our findings of bilateral language representation following childhood stroke stand in contrast to studies of adult stroke and to studies of perinatal stroke. The general consensus in the adult stroke literature is that left hemispheric reorganization (particularly involving perilesional regions) is associated with better long-term language outcome than right hemisphere dominance or bilateral representation.Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Cao, Vikingstad and George 3 , Reference Xu, Zhang, Shang, Wang, Luo and Weng 7 Conversely, following perinatal left MCA stroke, right hemisphere language dominance appears to be very common, even in patients with good outcome.Reference Heller, Heier and Watts 8 - Reference Tillema, Byars and Jacola 10 , Reference Szaflarski, Allendorfer and Byars 48 This right hemisphere language shift can occur even when perinatal stroke damage is restricted to the left periventricular regionReference Staudt, Grodd, Niemann, Wildgruber, Erb and Krägeloh-Mann 12 , Reference Staudt, Lidzba, Grodd, Wildgruber, Erb and Krägeloh-Mann 13 ; however, there is some evidence that, when individual differences are examined, left inferior frontal involvement in language tasks is still advantageous in this population.Reference Raja Beharelle, Dick and Josse 26 Our study suggests that left MCA stroke during childhood may be associated with bilateral language representation—essentially, a middle-ground between the adult and perinatal groups. This finding is consistent with evidence that language lateralization beings early in lifeReference Dehaene-Lambertz, Dehaene and Hertz-Pannier 49 and increases throughout childhood.Reference Schlaggar, Brown, Lugar, Visscher, Miezin and Petersen 36 It is also consistent with studies showing equal likelihood of language deficits in children with a history of right or left perinatal brain injury,Reference Nass 28 , Reference Dennis 29 , Reference Vargha-Khadem, O’Gorman and Watters 50 - Reference Trauner, Eshagh, Ballantyne and Bates 53 but a strong predominance of language disorders following left hemisphere injury in older children.Reference Dennis 29 , Reference Woods and Carey 31 , Reference Nass and Trauner 54 , Reference Van Hout 55 Thus, with increasing lateralization, the likelihood of recruiting prepotent right hemisphere regions for language following left hemisphere injury appears to quickly decrease.
However, there are some other important differences between our study of childhood stroke and those of perinatal stroke that may contribute to the different lateralization patterns. Tillema et alReference Tillema, Byars and Jacola 10 and Heller et alReference Heller, Heier and Watts 8 included children with active seizure disorders, whereas we excluded participants with epilepsy or other neurological comorbidities. The chronic impact of seizures on the brain may have altered lateralization patterns differently than static unilateral lesions alone. Moreover, several studies of perinatal stroke have either included children with very low IQs and/or poor language outcomeReference Tillema, Byars and Jacola 10 or they have not included any measures of language outcome.Reference Heller, Heier and Watts 8 , Reference Jacola, Schapiro and Schmithorst 9 , Reference Staudt, Grodd, Niemann, Wildgruber, Erb and Krägeloh-Mann 12 , Reference Staudt, Lidzba, Grodd, Wildgruber, Erb and Krägeloh-Mann 13 In our study, all participants had average IQs and good language outcome documented by standardized testing; thus, differences between participants with good and poor language outcome may also have contributed to discrepancies between our findings and others. The childhood stroke group reported by Ilves and colleaguesReference Ilves, Tomberg and Kepler 14 exhibited LIs that were more somewhat more left hemisphere dominant than ours, but lesion location information was not presented for these children, so it is difficult to compare findings directly. Finally, some studies of perinatal stroke have examined participants while they are still children,Reference Jacola, Schapiro and Schmithorst 9 , Reference Tillema, Byars and Jacola 10 , Reference Ilves, Tomberg and Kepler 14 , Reference Raja Beharelle, Dick and Josse 26 whereas our participants were young adults at the time of the study. Because of the potential differences between lateralization patterns in the developing brain versus the mature brain, it is difficult to directly compare our studies.
The Impact of Cortical Versus Subcortical Lesions on Lateralization Patterns
With respect to the lateralization differences between the adult and childhood stroke groups in our study, it is important to consider that four of the five patients in the childhood group had lesions restricted to subcortical regions with perisylvian sparing, yet all showed significant alterations in language representation. In contrast, all of the adult participants sustained lesions to canonical language areas in the left frontoparietal cortex, which one might expect to result in more pronounced alterations of language representation.
There are few data on aphasia following subcortical lesions in children,Reference Aram, Rose, Rekate and Whitaker 56 - Reference Gout, Seibel and Rouvière 60 but the literature examining subcortical aphasia following adult stroke is quite extensive.Reference Perani, Vallar, Cappa, Messa and Fazio 61 - Reference Nadeau and Crosson 65 The role of the left thalamus in language function (particularly word finding) has been firmly established,Reference Crosson 66 - Reference Kumar, Masih and Pardo 69 but there is considerable debate about the role of the basal ganglia. Aphasic deficits following left basal ganglia stroke in adults have been widely reported, but there does not appear to be a particular aphasic syndrome associated with such lesions.Reference Gout, Seibel and Rouvière 60 , Reference Nadeau and Crosson 65 , Reference Crosson 66 Several theories regarding the neurological mechanism of basal ganglia aphasia have attempted to explain symptom variability across patients. Perhaps the most compelling theory is that the underlying large vessel stenosis that causes the basal ganglia infarct also results in cortical hypoperfusion, which, in turn, disrupts language function. Hillis and colleagues found cortical hypoperfusion in patients with basal ganglia infarcts was strongly associated with aphasia and that symptoms resolved with perfusion normalization.Reference Hillis, Wityk and Tuffiash 47 , Reference Hillis, Kane and Tuffiash 70 , Reference Hillis, Barker, Beauchamp, Winters, Mirski and Wityk 71 Moreover, the regional location of the hypoperfusion predicted the specific aphasic deficits that were manifest in each patient.Reference Hillis, Barker and Wityk 72 Also, in a sample of children and young adults with basal ganglia stroke, Rowan et alReference Rowan, Vargha-Khadem and Calamante 73 found that language deficits were associated with subtle abnormalities in cortical language areas, detected only through voxel-based morphometry. Extrapolating these findings to the context of our study, one could speculate that large-vessel stenosis and basal ganglia infarction in childhood could disrupt cortical perfusion within the dominant hemisphere, thereby impacting the way in which cortical-subcortical language networks develop. Furthermore, because the diagnosis and treatment of pediatric stroke is often delayed,Reference Rafay, Pontigon and Chiang 74 children may be at increased risk for long periods of cortical hypoperfusion. Of note, all of the pediatric participants included in our study had documented stenosis of the left internal carotid artery and/or left MCA at the time of the stroke, making this a plausible explanation for our findings. We did not examine resting cortical perfusion in our participants, but this would be an interesting element to incorporate into future studies.
Task Differences and Lateralization Patterns
Another important finding to highlight from our study is that LIs and activation patterns varied considerably between the verb generation and picture-word matching tasks. This was true for controls and both patient groups. Verb generation is one of the most commonly used language task in neuroimaging studies and has been associated with strongly lateralized activation patterns in control participantsReference Pizzamiglio, Galati and Committeri 34 , Reference Schlaggar, Brown, Lugar, Visscher, Miezin and Petersen 36 and clinical populations.Reference Price and Crinion 6 , Reference Rijntjes and Weiller 35 Thus, we reasoned that this task would be particularly useful for contrasting intra- and inter-hemispheric reorganization of language function following left MCA stroke in childhood and adulthood. As hypothesized, the verb generation task was most informative in distinguishing between the activation patterns of the three groups, with controls showing stronger left lateralization than patients and the adult stroke group showing stronger left lateralization than the childhood stroke group. In comparison, the picture-word matching task was not as good at distinguishing among the three groups in our study, although the generation trend toward stronger left lateralization in control versus patients and in adult stroke participants versus childhood stroke participants was still observed. Other studies using similar tasks involving picture-word matching or overt picture naming indicate that the associated patterns of brain activation are less consistent—and less strongly lateralized—than those during verb generationReference Price and Crinion 6 , Reference Pizzamiglio, Galati and Committeri 34 ; thus, it is perhaps not surprising that our picture-word matching task was less powerful than the verb generation task for investigating our hypotheses. Ours is not the first study to document task differences in lateralization,Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Barnett, Marty-Dugas and McAndrews 37 and it is widely recognized that task differences across studies contribute significantly to inconsistencies in the literature.Reference Price and Crinion 6 , Reference Rijntjes and Weiller 35 , Reference Gaillard, Balsamo and Xu 38 As a result, it is critical to consider task demands when interpreting patterns of language lateralization, both in normal participants and clinical populations.
Future Directions
Our study is limited by small sample sizes and differences in lesion location across the adult and childhood stroke groups. We are currently recruiting participants with cortical left MCA stroke in childhood to address this confound. A cortical childhood stroke group would be an important contrast to both the adult stroke and subcortical childhood groups included in the present study because it would permit more definitive conclusions about the impact of lesion location versus age at lesion on reorganization patterns. Moreover, it would be informative to follow children with left MCA stroke over time with repeated fMRI sessions to examine how lateralization changes with recovery and development. Another limitation of our study is the significant difference in age at study between the pediatric and adult stroke groups. As a result, factors related to neural development (such as brain atrophy related to aging and continued brain maturation in young adults) may have affected our findings. The pediatric and adult stroke groups also differ in number of years since stroke, with the adult group having fewer years to develop compensatory strategies and brain networks. This may have also affected our findings, although based on previous research,Reference Calvert, Brammer, Morris, Williams, King and Matthews 2 , Reference Cao, Vikingstad and George 3 , Reference Price and Crinion 6 one might expect that a shorter length of time since stroke would make it more likely that the adult stroke participants would show such strong left lateralization. Longitudinal studies that follow childhood stroke survivors throughout the lifespan are needed to address important questions related to the impact of an early lesion on the aging process. Stroke is the ideal population in which to study structure-function relationships and age-related changes in plasticity because the lesions are static, focal, and can occur at any age. However, because of the relatively low incidence of pediatric stroke,Reference deVeber 75 opportunities to directly contrast childhood and adult stroke are rare. Multicenter studies may allow for larger sample sizes, more tightly matched subject groups, and statistical analysis of individual differences.
