Abstract
Background: It is generally accepted that most people have left-hemispheric language dominance, though the actual incidence of atypical language distribution in non–right-handed subjects has not been extensively studied. The authors examined language distribution in these subjects and evaluated the relationships between personal handedness, family history of sinistrality, and a language laterality index (LI) measured with fMRI.
Methods: The authors used whole-brain fMRI to examine 50 healthy, non–right-handed subjects (Edinburgh Handedness Inventory quotient between −100 and 52) while they performed language activation and nonlinguistic control tasks. Counts of active voxels (p < 0.001) were computed in 22 regions of interest (ROI) covering both hemispheres and the cerebellum. LI were calculated for each ROI and each entire hemisphere using the formula [L − R]/[L + R].
Results: Activation was predominantly right hemispheric in 8% (4/50), symmetric in 14% (7/50), and predominantly left hemispheric in 78% (39/50) of the subjects. Lateralization patterns were similar for all hemispheric ROI. Associations were observed between personal handedness and LI (r = 0.28, p = 0.046), family history of sinistrality and LI (p = 0.031), and age and LI (r = −0.49, p < 0.001).
Conclusions: The incidence of atypical language lateralization in normal left-handed and ambidextrous subjects is higher than in normal right-handed subjects (22% vs 4–6%). These whole-brain results confirm previous findings in a left-handed cohort studied with fMRI of the lateral frontal lobe. Associations observed between personal handedness and LI and family history of handedness and LI may indicate a common genetic factor underlying the inheritance of handedness and language lateralization.
The incidence of left-handedness in the general population is about 13% during teenage years and declines gradually with age, reaching about 6% in the seventh and eighth decades of life.1,2⇓ Multiple factors are believed to affect handedness, including maternal handedness and family history of sinistrality,3,4⇓ sex,2 age,1 testosterone level,5 and history of early brain injury.6 In addition, left-handedness predisposes to certain psychiatric conditions,7,8⇓ choice of profession,9,10⇓ epilepsy,6 and decreased life expectancy.11-15⇓⇓⇓⇓ Left-handed individuals are believed to have a higher incidence of atypical (right dominant or mixed) language representation, though most of the evidence for this assertion comes from studies of patients with neurologic disease, in whom pathology might jointly determine both handedness and language lateralization.6,16-21⇓⇓⇓⇓⇓⇓
Recent noninvasive imaging studies demonstrate that approximately 95% of normal right-handed subjects have left-hemispheric dominance for language.16,17,22⇓⇓ These studies also show that the nondominant (right) hemisphere plays an important role in language processing, though to a varying degree in different people. In contrast, the spectrum of language lateralization patterns in normal left-handed and ambidextrous individuals is not well established. In one fMRI study, normal left-handed subjects had a higher incidence of atypical language dominance (24%) than comparable right-handed subjects (4%).17 Language dominance in this study was based on activation in one lateral frontal region; however, and it is not known whether this result holds for the brain as a whole. In contrast, another fMRI study found no relationship between degree of right-handedness and language lateralization in a normal right-handed cohort.16 Neither of these studies included ambidextrous people, about whom little is known. No studies have assessed whether other genetic factors, such as sex and family history of left-handedness, are related to language lateralization in left-handed individuals.
We estimated the incidence of atypical (symmetrical or right-hemispheric) language organization in normal left-handed and ambidextrous subjects. We used a language mapping protocol that has been extensively studied in normal right-handed subjects and that produces strongly left-lateralized activation in prefrontal, temporal, and parietal cortical regions known to be involved in language processing.16,23-26⇓⇓⇓⇓ This activation protocol and the language lateralization index derived from it were previously validated by direct comparison with results of Wada language testing in a series of patients with epilepsy.18 We used whole-brain activation measures, permitting conclusions to be drawn both about overall hemispheric lateralization and about lateralization within selected regions of interest. Furthermore, we sought to determine whether personal handedness patterns, family history of left-handedness, age, and sex are related to language lateralization in non–right-handed individuals.
Subjects and methods.
Participants.
Fifty healthy, non–right-handed adults were recruited on a voluntary basis using written informed consent procedures. Each subject underwent a screening interview; subjects who reported a history of possible or actual neurologic or psychiatric problems were excluded from participation. Neurologic examinations were not performed. All subjects had normal T1-weighted brain MRI results. Non–right-handedness was defined as a handedness quotient (HQ), based on the Edinburgh Handedness Inventory, between −100 and 50.27 One subject identified herself as left-handed for writing and eating but not for other functions and was included in the study despite an HQ of 52. Family history of sinistrality was assessed using a family handedness inventory. The family history was considered positive if at least one member of the immediate family (parent, sibling, or grandparent) was reported to be left-handed. English was the primary language of all the subjects. Age, sex, and educational history were also recorded. Demographic characteristics of the cohort are presented in table 1. All procedures were approved by the Human Research Review Committee of the Medical College of Wisconsin.
Table 1 Demographic characteristics of the 50 subjects
Image acquisition.
Imaging was performed on a 1.5 Tesla GE Signa (General Electric, Milwaukee, WI) scanner using a three-axis local gradient coil (Medical Advances, Inc., Milwaukee, WI). High-resolution, T1-weighted anatomical reference images were obtained using a three-dimensional spoiled-gradient-cho sequence. fMRI used a T2*-weighted, gradient-echo, echoplanar sequence (echo time [TE] 40 ms, repetition time [TR] 3,000 ms or 4,000 ms, field of view [FOV] 240 mm, 64 × 64 pixel matrix). Echoplanar images were acquired as contiguous 7-mm sagittal slices covering the whole brain, including the cerebellum. Voxel size was 3.75 × 3.75 × 7 mm.
Stimuli and activation tasks.
The stimuli and activation tasks, their rationale, and the typical patterns of activation and lateralization observed in right-handed subjects have been described previously.16,18,24-26⇓⇓⇓⇓ Stimuli were 16-bit, digitally synthesized tones and sampled male speech sounds presented binaurally at precise intervals using a computer playback system. In the control task (tone decision), subjects heard trains of three to seven 150-ms tones of either 500 or 750 Hz frequency. The subjects were instructed to press a button each time they heard a train containing two high-pitched (750 Hz) tones. In the language task (semantic decision), subjects heard names of animals (e.g., horse) and were instructed to press a button for animals they considered to be both “found in the United States” and “used by humans.” The tasks were matched for average stimulus intensity, stimulus duration, and frequency of positive targets.
fMRI data analysis.
All image analysis was done with the AFNI software package (available at: http://afni.nimh.nih.gov/afni).28 Motion artifacts were minimized by within-subject registration of all raw echoplanar image volumes to the first steady-state volume (fifth volume in the first run). Registration was performed with the “3dvolreg” module in AFNI, which uses an iterative least-squares procedure to minimize the variance in voxel intensity differences between images (see http://afni.nimh.nih.gov/afni for a complete description of the registration algorithm). Image registration also generated estimates of all rotation and translation parameters for each volume. To identify task-related changes in MRI signal, we used a correlation approach as described previously.29 The predicted response was estimated using a gamma function to model the hemodynamic response to each task trial. Analysis of covariance (ANCOVA) was then used to identify voxels with observed blood oxygenation level–dependent (BOLD) responses that matched the predicted response, incorporating the rotation and translation vectors as covariates of no interest. Voxels exceeding a threshold r value corresponding to p value <0.001 were counted for each subject in 22 regions of interest (ROI; 11 per hemisphere). These primary ROI included medial and lateral frontal regions, a medial temporal region, lateral anterior and lateral posterior temporal regions, a thalamo–striato–capsular region, angular gyrus, posterior cingulate gyrus, an occipital region, a cerebellar region, and a general region that included brain areas not encompassed in the other regions. ROI were defined using an average activation map from 80 normal right-handed subjects.16,26⇓ Each cluster of activated voxels in the left hemisphere and right cerebellum of this activation map constituted an ROI. Right hemisphere and left cerebellum ROI were created by reflecting these ROI symmetrically across the midline.16,26⇓ Counts from these primary ROI were then grouped into five larger “combination” ROI that were used to calculate regional laterality indexes (LI). These combination ROI included a total hemisphere ROI (all regions in one cerebral hemisphere), a frontal ROI (medial and lateral frontal regions), a temporal ROI (anterior and posterior lateral temporal regions and medial temporal region), a medial hemispheric ROI (medial frontal region, medial temporal region, and posterior cingulate region), and a lateral hemispheric ROI (lateral frontal region, both lateral temporal regions, and angular gyrus). LI were calculated for each of these ROI in each subject as the normalized ratio 100*[l − R]/[L + R], where L and R represent the number of suprathreshold voxels in left and right homologous ROI. This approach yields LI that range between strongly left dominant (100) and strongly right dominant (−100). We also categorized subjects as left-hemisphere language dominant (LI > 20), symmetrical (20 ≥ LI ≥ −20), or right-hemisphere language dominant (LI < −20).16,26⇓ Such categorization schemes, when applied to a continuous measure such as the LI, are inevitably somewhat arbitrary. As pointed out by several authors, for example, the variability in reported incidence of “bilateral” language dominance in prior studies is largely a function of how these category boundaries have been defined.16,30,31⇓⇓ The boundaries used here were adopted for consistency with our previous study2 and with similar category boundaries adopted by other laboratories,1,21⇓ with the aim of simplifying comparisons between studies. In a previous cohort of 100 right-handed subjects studied using the same tasks and analysis methods, 94% of subjects showed left-dominant activation patterns, and the median hemispheric LI was 66.1.16
Simple regression techniques were used to assess relationships between the continuous variables LI, HQ, and age. Student’s t-tests were used to assess differences on these measures related to sex and family history of sinistrality.
To illustrate the average pattern of language-related activation in the current subject group, the individual subject activation maps were averaged in standard stereotaxic space using methods described previously.25,26⇓ In brief, voxel-wise r values were converted to t statistics and resampled in stereotaxic space using AFNI software. The resulting t-maps were smoothed with a 4-mm root-mean-square Gaussian filter and averaged. The resulting group map was thresholded at an estimated uncorrected p < 0.0001.
Results.
Task performance.
All subjects learned the tasks easily and performed well above chance levels. The percentage of correct responses averaged 89.2 (SD = 7.4) for the semantic task and 97.5 (SD = 6.5) for the tone task. These means were nearly identical to and did not differ significantly from those of a right-handed cohort studied previously.16
Brain activation.
As shown in figure 1, increases in BOLD signal during the language task were observed in supratentorial and cerebellar regions similar to patterns previously reported in right-handed subjects.16,18,24-26⇓⇓⇓⇓ These regions included prefrontal cortex of the inferior, middle, and superior frontal gyri; posterior cingulate gyrus and retrosplenial cortex; anterior superior temporal sulcus and middle temporal gyrus; posterior inferior temporal gyrus; fusiform and anterior parahippocampal gyri; anterior hippocampus; angular gyrus; and posterior cerebellum. On the whole, the activation pattern in the current subject group closely replicates that observed in several previous studies, confirming the robustness of the activation paradigm.25,26⇓ As illustrated in figure 1, the average activation pattern of the subjects in the current study was, however, somewhat less lateralized and showed more activation in the right hemisphere compared with an average pattern from 50 normal right-handed subjects.26
Figure 1. Group average activation maps, showing areas with higher blood oxygenation level–dependent signal during the semantic task in the semantic-tones task contrast. The top row shows results for the 50 non–right-handed subjects in the current study, and the bottom row shows results for a group of 50 right-handed subjects reported by Frost et al.16,26⇓ Activation data, thresholded at an uncorrected p < 0.0001, are shown for five representative axial sections through the standard coordinate space of Talairach and Tournoux.42 The left side of the brain is on the reader’s left. Horizontal and vertical green lines represent the x and y axes of the coordinate space. Numbers beside each slice indicate the z-axis position of the slice. Activations in the non–right-handed subjects (top row) are relatively less left lateralized than activations in the right-handed subjects.
LI derived from the total hemisphere ROI varied from strongly left dominant (LI = 89.5) to strongly right dominant (LI = −57.5; figure 2). Using dominance classification criteria with the total hemisphere LI, the majority of subjects (78%) had left language dominance, 14% had symmetrical activation, and 8% had right dominance. We observed a similar distribution of dominance patterns in the four other combination ROI (table 2). The frontal ROI yielded a higher proportion of atypical (right dominant or symmetrical) LI than did the temporal ROI, but the overall difference in dominance patterns between these two ROI did not reach significance (χ2 [df = 2] = 3.98, p = 0.137). Similarly, the medial ROI produced a higher proportion of symmetrical LI than did the lateral ROI, but the overall difference in dominance patterns between these two ROI did not reach significance (χ2 [df = 2] = 3.03, p = 0.219). In addition, we noted strong correlations between LI derived from nonoverlapping combination ROI, i.e., temporal and frontal (r = 0.82, p < 0.001) and lateral and medial (r = 0.9, p < 0.001). LI derived from primary ROI were also similar, with correlations ranging from r = 0.35 (p = 0.013) between the anterior and posterior lateral temporal LI to r = 0.86 (p < 0.001) between the lateral frontal and angular gyrus LI. Table 2 shows categorical LI distributions for eight of the primary ROI. With the exception of the cerebellar region, these distributions were also similar to that of the entire hemisphere. As expected, the dominance patterns in the cerebellar region were partially reversed from those of the supratentorial regions, with 58% of subjects showing right cerebellar dominance and only 22% showing left dominance.
Figure 2. Frequency distribution of the total hemisphere language laterality index (LI) in normal left-handed and ambidextrous subjects. Bars represent 20-point intervals (lowest interval −60 to −40, highest interval 80 to 100).
Table 2 Incidence of right language dominant, symmetrical, and left language dominant subjects based on language laterality indexes computed by region of interest (ROI)
Men and women in the study were matched as a group on age (mean age for men = 26.4 vs 27.6 for women; p = 0.522), HQ (mean HQ for men = −24.1 vs −35.3 for women; p = 0.453), and proportion reporting familial left-handedness (42% for men vs 38% for women; p = 0.817). Despite elimination of these potential confounders, there was no difference between LI of men and women using the total hemisphere ROI (mean LI for men = 42.1, mean LI for women = 44.3, t = −0.228, p = 0.82), nor were there any sex differences using LI from the other ROI (all p > 0.1).
The total hemisphere LI declined with increasing age (r = −0.49, p < 0.001). Figure 3 shows the scatterplot depicting total hemisphere LI as a function of age. Age was also inversely correlated with LI from the lateral, medial, frontal, and temporal combination ROI (all p ≤ 0.001). The total hemisphere LI was correlated with HQ (r = 0.28, p = 0.046). Figure 4 shows the scatterplot of total hemisphere LI as a function of HQ. HQ was also correlated with lateral (r = 0.31, p = 0.026) and frontal LI (r = 0.30, p = 0.033), but not with temporal (r = 0.16, p = 0.28) or medial LI (r = 0.21, p = 0.15). To determine whether familial left-handedness was associated with a shift toward right-brain language processing, LI of subjects with (FH+) and without (FH−) left-handed relatives were compared using a one-tailed t-test. FH+ subjects had a significantly lower LI than FH− subjects (mean LI for FH+ = 32.2, mean LI for FH− = 50.6; p = 0.031). The proportion of subjects with atypical language dominance was 35% (7/20) among FH+ subjects vs 13% (4/30) among FH− subjects (χ2 [1] = 3.28, p = 0.07). The multiple regression equation (LI = 110.7 + 0.112*(HQ) − 2.181*[age] − 13.07*[FH]) relating total hemisphere LI to HQ, age, and FH (where FH+ = 1 and FH− = 0) was significant at p = 0.001 and accounted for 30.7% of the variance in LI.
Figure 3. Scatterplot of total hemispheric laterality index (LI) as a function of age. Linear regression with 95% mean prediction interval. Total hemisphere LI = 109.79 − 2.46*Age; R2 = 0.24.
Figure 4. Scatterplot of total hemispheric laterality index (LI) as a function of personal handedness quotient (HQ); linear regression with 95% mean prediction interval. Total hemisphere LI = 48.84 + 0.19*HQ; R2 = 0.08.
Discussion.
We observed atypical language dominance in 22% of our normal, non–right-handed subjects (14% symmetrical and 8% right-dominant). This pattern resembles previously reported Wada test data obtained from left-handed patients with epilepsy with no history of early brain injury, which suggests an atypical dominance rate of about 30% in this patient group (15% bilateral and 15% right-hemispheric language representation).6 These Wada test results have been considered representative of the general population with regard to language lateralization in left-handed subjects. Our results are also similar to findings from an earlier fMRI study of normal right- and left-handed subjects, which showed an atypical dominance rate of 24% in normal sinistrals.17 Recently, language laterality data have been reported for a large sample of normal volunteers studied with transcranial Doppler during performance of language tasks.22 These authors showed a 15% incidence of right-hemisphere language dominance (defined as any LI < 0) among ambidextrous subjects and a 27% incidence among strongly left-handed subjects. These rates are all considerably higher than the atypical dominance rate of 4 to 6% found in normal right-handers,16,17,22⇓⇓ supporting the general claim that handedness and language lateralization are linked.
Language lateralization patterns are often described as typical (left-hemispheric) or atypical (symmetrical and right-hemispheric), but categorization schemes of this kind are probably an oversimplification. Although language-related activation in normal right-handed subjects is predominantly left hemispheric, almost all subjects activate right hemisphere areas to some extent in fMRI and PET language studies.16,17,32,33⇓⇓⇓ Quantitative studies with large subject samples suggest the existence of a continuum of language lateralization patterns ranging from strongly left dominant to strongly right dominant.16,17,22,26,33⇓⇓⇓⇓ The quantitative relationship between language lateralization and the Edinburgh Handedness Inventory handedness quotient (HQ) has been assessed in several previous studies. In one study, no significant correlation was found between the language LI used in the current study and HQ among normal right-handed subjects with HQ in the range of 50 to 100.16 In contrast, a simple linear relationship between HQ and the likelihood of having right-hemispheric language dominance (i.e., LI < 0) was observed using transcranial Doppler measures of language lateralization.22 This discrepancy in results is likely due, in part, to the more restricted HQ range examined in the fMRI study.16 The current results confirm those of the transcranial Doppler study in that language LI was significantly correlated with HQ among normal subjects with HQ in the range of −100 to 52.22 This relationship was of modest strength, however, accounting for only about 8% of the variance in total hemisphere LI (see figure 4).
The correlation observed here between HQ and total hemisphere LI may be relatively small because of the task used for language activation and the methods used to compute the LI. In the transcranial Doppler studies and in several previous fMRI studies, the language activation task involved “word generation,” which is known to produce activation primarily in frontal regions.17,21,23,34,35⇓⇓⇓⇓ Indeed, in the only previous fMRI study of language lateralization in left-handed subjects, the language LI was derived entirely from activation in a lateral frontal ROI.1 In contrast, the total hemisphere LI used in the current study is derived from activation in temporal, parietal, and cingulate cortices as well as frontal regions. If handedness is more closely linked to lateralization of frontal language areas than to these other regions, then inclusion of other regions in the LI could weaken the correlation with HQ. The data from the current study support this model, in that HQ was significantly correlated with the frontal LI but not with LI from other regions, and was more strongly correlated with the frontal LI than with the temporal LI.
Handedness is at least partly genetically determined.3,36⇓ If language lateralization is related to handedness, then both language lateralization and handedness may be influenced by a common genetic substrate. To date, genetic models of cerebral dominance for language have been restricted to explaining the association of handedness and language dominance mainly on a theoretical basis, without much empirical data.3,4,36,37⇓⇓⇓ In her right-shift gene theory, Annett postulates that a single gene influences not whether a person is right- or left-handed, but whether there is a bias toward right-handedness or not.3,4⇓ According to this theory, approximately 20% of the normal population does not have the right-shift gene, and such persons have a 50% chance of becoming left-handed.36 This model is supported by empirical observations in the normal population, which show that approximately 9.5% of children of right-handed parents are left-handed, and this rate increases to 20.5% if one parent is left-handed and to 26.1% if both parents are left-handed.36,37⇓ Recently, the same author suggested that the right-shift gene influences both cerebral language dominance and handedness.3 If this theory is correct, homozygotes with the right-shift gene should be predisposed to very strong left-hemispheric language dominance, heterozygotes to symmetrical language representation, and persons lacking the right-shift gene to symmetrical or right-hemispheric dominance.
If handedness and language dominance are influenced by a common genetic factor, then rightward shift of language functions should be greater, on average, in persons with a family history of sinistrality. A previous fMRI study of right-handed subjects, however, found no significant difference in language LI between subjects with and without a positive family history.16 Similarly, a transcranial Doppler language dominance study showed a significant effect of familial sinistrality on HQ but no effect on language lateralization.22 In contrast to these results, we observed a small but significant rightward shift in the total hemisphere LI in subjects with a positive family history of sinistrality. These results provide the first functional imaging support for a connection between familial sinistrality and language lateralization, and may indicate a common genetic factor underlying the hemispheric organization of language and motor functions. It should be emphasized, however, that the effect of family history on LI was small and therefore needs to be confirmed in larger studies.
We found no evidence of a relationship between sex and language lateralization. This negative result joins a growing body of evidence from imaging and transcranial Doppler studies—now collectively involving some 600 normal volunteers—that have also shown no differences between language dominance patterns of men and women.16,17,22,26,32⇓⇓⇓⇓ Another finding of this study was a significant negative correlation between age and language LI. This effect was in the same direction and of similar magnitude (r = −0.49) as the significant age/LI correlation observed by Springer et al. (r = −0.23).16 Although similar declines in lateralization with age were seen in prior functional PET studies of visual perception and memory using older subjects,38-41⇓⇓⇓ it is somewhat surprising to find such a robust effect of age on LI in these samples of young and middle-aged adults. Together, these results suggest that cognitive functions become somewhat less lateralized with increasing age, possibly reflecting compensation for age-related loss of functional capacity.40
Understanding the quantitative relationships between language lateralization, handedness, and the various demographic and genetic factors that influence these asymmetries of function in the normal population is of clinical relevance for two reasons. First, these relationships might be useful for predicting the risk of postoperative language disturbance in patients undergoing brain surgery for adult-onset disease. Second, such knowledge could lead to an improved understanding of the biological basis of language lateralization, which might eventually result in novel therapeutic strategies for patients with impaired language processing.
Acknowledgments
Supported by National Institute of Neurological Diseases and Stroke grants RO1 NS35929 and RO1 NS33576, and by National Institute of Mental Health grant PO1 MH51258.
Footnotes
-
Presented in part at the 53rd Annual Meeting of the American Academy of Neurology; Philadelphia, PA; May 2001.
- Received December 13, 2001.
- Accepted March 30, 2002.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.