Introduction
High educational level (HEL) seems to have an impact for slowing down cognitive degeneration in dementia cases by generating some sort of cognitive reserve1-5. EL has been defined as a factor involved in word retrieval performance assessed by the Boston naming test6-9. In the aging process, differences in EL could be significant in terms of cognitive activity and brain compensation capacity10-13. In contrast, a lower EL (LEL) has been associated with faster decline of memory, mental state, and verbal ability14,15.
The hemispheric asymmetry reduction in older adults (HAROLD) model16 attempts to explain changes in neural activation patterns found in older adults. This model proposes that the reduction in hemispheric lateralization is "reflective of a general aging phenomenon more than a task-specific occurrence"16. This reduction is visible when we compare activation in young adults with that of older adults during different tasks of working and verbal memory.
Following the HAROLD model, Springer et al.17 undertook a functional magnetic resonance imaging (fMRI) study comparing lateralization activity changes with age during a verbal episodic memory task. They came to the conclusion that older adults recruit more bilateral resources from prefrontal cortex as their EL rises. Similar to what has been observed in memory function, word retrieval efficiency also tends to decline with aging18-22. In a fMRI study, Wierenga et al.23 compared activation during a naming task between younger and older adults, controlling 15 years of schooling for both groups. Their results support the HAROLD model as the group of older adults showed more bilateral activations. The older group displayed more activation of the homolog Broca's area, which is extended up to the right inferior frontal gyrus (IFG BA 47), right superior temporal gyrus (STG), and right insula and left lingual gyrus. Left IFG is considered strategical for the "semantic working memory system," in charge of manipulating and monitoring retrieval, and maintaining semantic representations stored in temporal semantic cortices24. The authors discuss that difficulty for lexical access in elders could be due to frontal atrophy that could generate a compensatory mechanism recruiting a larger frontal network and increasing activity in right hemisphere but not necessarily beneficial for efficacy and performance.
However, HAROLD phenomenon has been mostly studied by comparing activity between young and older adults, and until now just a few studies have compared activation between two groups of older adults with different biological and sociocultural characteristics10, 25,26. In addition, researchers still need to confirm the origin of this phenomenon. Cabeza16 conceives it as an effect resulting from two possible causes: on the one hand, it could be a compensatory effect due to neural tissue loss in some areas of prefrontal cortex. In this case, activation increased in homolog structures could help to preserve the efficiency of cognitive functions2. On the other hand, this effect could result from a "dedifferentiation" effect in which the dominant hemisphere would lose its capacity to inhibit the other without actually contributing to function efficiency or even hampering its process.
Nevertheless, to the best of our knowledge, there are no studies that specifically analyze the effects of EL on the functional activation patterns in lexical access tasks among the older adult population. Using the HAROLD model, we have two aims for the present study: (1) measure lateralization neural patterns in healthy older adults with HEL and LEL during a picture-naming task and (2) test the compensation/de-differentiation hypothesis about functions of age-related asymmetry reductions by analyzing performance in a behavioral task in association with lateralization neural patterns.
Specifically, we predicted differences between EL groups, and these could rely on the semantic working memory system: IFG (BA 47) and temporal anterior cortices as a reflection of more and better strategies of people with HEL.
With respect to our second aim, high performers (Hi-P) and low performers (Low-P) would show different patterns of activation. Specifically, if hemispheric asymmetry reduction was present in Hi-P group, we would have an adaptive process of functional cognitive compensation; however, if these activation patterns appeared in low-performance group, then we would have a phenomenon that might imply functional disorganization or a disinhibition process creating interference in task performance efficiency.
Methods
Participants
Volunteers were accepted if they do not present hypertension or cardiovascular and thyroid diseases so hypertension would not be a factor of alteration in BOLD effect25,27. To confirm normal cognitive state, a neuropsychological profile was elaborated for each volunteer (PIEN "Test-Barcelona")28. 28 older adults, Spanish native-speakers, and right-handed were accepted for the study. HEL group n = 13 (7 females, mean age = 63.6 years, and mean schooling = 18.4 years). LEL group n = 15 (10 females, mean age = 65.5, and mean schooling = 6 years). All participants were informed about their rights and signed an informed consent letter elaborated following the standards of the Hospital General de México's Ethics Committee. Patients received only their neuropsychological and MRI structural results. This study has been supported by Research General Management at México´s General Hospital with register No. DI/11/403/04/126 and UNAM register No. PAPIIT IN200817.
Procedure
FMRI NAMING TASK
Every participant was presented 120 images (45 described actions, 45 objects, and 45 were control condition). Each image was shown for 2500 ms randomly ordered in each presentation, inter-stimuli intervals were programmed randomly from 4400 to 8800 ms. Task was programmed using E-prime 2 Software, length about 13 min.
EXPERIMENTAL CONDITION
Participants were asked to state aloud the name of the object or action that they were watching. All images were black and white drawings taken from the International Picture-Naming Project (http://crl.ucsd.edu/experiments/ipnp/) and selected for Spanish language. In the same way, use frequency and capacity of stimuli were considered, for HEL group 15% of stimuli were changed for other less frequent.
CONTROL CONDITION
To subtract the visual and articulatory activations corresponding to lexical retrieval, the same images used for the experimental condition were distorted for the control condition. Participants were asked not to try to figure out its shape but utter the pseudoword "LOLE."
RECORDING OF ANSWERS IN FMRI TASK
Loud answers were monitored in situ and recorded using the Sound Forge Pro10 Software (Sony Creative Software Inc.) for its analysis out of line. For recording inside MRI, a non-metallic extension cord for the microphone was designed.
TRAINING SESSIONS
They took place during the neuropsychological interview. Each subject was presented with 200 images different from the ones shown in the experimental session to avoid learning effect. Participants were asked to answer aloud to confirm that they understood the task and that they were capable of perceiving and understanding the images.
ANSWER ONSET TIME ANALYSIS
Overt answers during the fMRI session were analyzed taking the beginning of the first audio wave form of every overt answer (speech) and subtracting 200 ms, considered as the time for articulatory codification process, to analyze the moment for lexical access only, following the Jescheniak et al.29 model of speech production. We considered this time as an onset for SPM analysis.
ASSESSMENT OF THE COMPENSATION-DEDIFFERENTIATION HYPOTHESIS OF THE HAROLD MODEL
Regarding our second objective, we reorganized data in terms of Hi-P and Low-P. The median of correct answers for the whole sample was defined (86.1% = 78 images correctly named out of 90). We obtained two groups: below median (from 68 to 78 correct answers) and surpass the median score (from 79 to 88 correct answers). Low-performance group (n = 10) was composed of 4 HEL and 6 LEL individuals. Hi-P group (n = 18) was integrated by 9 subjects with HEL and 9 with LELs.
MISTAKE CLASSIFICATION
We considered as errors the lack of answer (omission), the occasions when the uttered word had no relation with the object, or when the answers were not verbs in the case of action-related images.
FMRI ACQUISITION
T1-weighted gradient echo pulse sequence anatomical images (TR = 4000 ms, TE = 106 ms, FOV 230 mm, flip angle = 90°, and voxel size 0.7 × 0.7 × 5 mm3) and functional images (EPI sequence: data matrix: 64 × 64; FOV 192 mm2; TE 50 ms; TR 3800 ms) were obtained from a 1.5 T Siemens Avanto system equipped with a standard head coil. Functional T1-weighted images were collected covering the entire brain continuously, acquiring 36 interleaved slices (3 mm thick), and parallel to the anterior-posterior commissural plane (voxel size 3 mm3).
FMRI PREPROCESSING
Functional data were preprocessed using SPM8 (Wellcome Department of Cognitive Neurology, London, UK; see http://www.fil.ion.ucl.ac.uk/spm) following the standard procedure. Briefly, functional images for each subject were corrected for differences in slice acquisition times referring to the slices mean; realigned to correct head movement; spatially normalized to the stereotaxic space of Talairach and Tournoux using the Montreal Neurological Institute space30; and smoothed using an isotropic 8 mm FWHM Gaussian kernel.
FMRI ANALYSIS
Bold events for each word type (verbs, nouns, and control response) were modeled as pseudodelta functions coinciding with the stimulus onset and convolved with the synthetic hemodynamic response function. Brain responses associated with each experimental condition were estimated according to the general linear model for an event-related design at each voxel. Only correct trials were modeled to identify brain regions involved in successful naming (see mistake classification above). In the first level analysis, T-statistical parametric maps for every voxel were obtained for each subject (intra-subject effects) by applying linear contrasts to the parameter estimations for the events of interest (fixed effects). Subsequently, in a second level analysis, between-subjects activations were calculated for each condition (intragroup and between groups) by employing a two-sample t-test (random effects). Signals from cluster maxima (p = 0.001 K threshold superior to 10 voxels) were extracted, activation maxima refer to the Talairach space32. To correct multiple comparisons and to control for false positives and false negatives too, we applied the non-parametric morphology-based hypothesis testing (MBHT)31, this procedure makes possible to detect moderate activation level regions, whose cannot be detected by conventional approaches, such as Family Wise Error (FWE) or High Dynamic Range (HDR), but which are spatially extensive, by explicitly relating the magnitude of the signal in each voxel to that of its neighbors31.
Results
Behavioral results
Participants' performance was analyzed with a two-way ANOVA, which included the two groups (HEL and LEL) as the "inter-subject" factor and the image type (objects or actions) as the "intra-subject" factor. This analysis did not show any significant differences by group (F < 1), neither by the effect of interacting with the type of image (object, action or control type) (F < 1).
Statistical parametric mapping results
To compare word type activation patterns, correct answers to object images were called "nouns" and correct answers to action images were called "verbs." A third category called "words" included "nouns + verbs." This was created for contrast analysis to analyze whole lexical access function.
EL
ACTIVATION WITHIN EACH GROUP
Comparison inside HEL group did not show significant differences (p < 0.001) in any of the three contrasts (words, verbs, or nouns > control). Comparison within LEL group presented significant activation for words > control condition in right postcentral gyrus (BA 1), as well as in left posterior tail of caudate nucleus and left cingulate gyrus (BA 31) (Fig. 1). Only the latter showed significant results for the correction test of multiple comparisons (p < 0.054). Results for all significant activations are exposed on table 1. Verbs > control and nouns > control contrast did not show any significant differences within the LEL group either.
Group | Contrast | Hemisphere | Brain region | Talairach coordinates (X/Y/Z) | Cluster size | Z score (peak voxel) | MBHT (independent t-test) |
---|---|---|---|---|---|---|---|
LEL | Words-CTRL | Left | Cingulate gyrus (BA31) | -18/-44/-22 | 14 | 3.15 | 0.054* |
LEL | Words-CTRL | Right | Caudate tail | 24/-38/-18 | 67 | 3.99 | 0.127 |
LEL | Words-CTRL | Right | Post central gyrus (BA1) | 32/-32/-64 | 10 | 3.27 | 0.211 |
HEL-LEL | Words-CTRL | Right | Cingulate Gyrus (BA24) | 12/-2/-30 | 11 | 3.48 | 0.141 |
LEL-HEL | Words-CTRL | Right | Postcentral Gyrus (BA3) | 30/-32/-62 | 29 | 3.52 | 0.133 |
LEL-HEL | Words-CTRL | Right | Precentral Gyrus (BA4) | 36/-24/-58 | 29 | 3.24 | 0.276 |
LEL-HEL | Nouns-CTRL | Left | Precuneus | -28/-70/-50 | 51 | 3.56 | 0.11 |
LEL-HEL | Nouns-CTRL | Right | Precuneus | 2/-54/-52 | 74 | 3.91 | 0.139 |
LEL-HEL | Nouns-CTRL | left | Superior Parietal (BA7) | -18/-56/-60 | 21 | 3.32 | 0.204 |
LEL-HEL | Nouns-CTRL | Right | Superior Parietal (BA7) | 40/-62/-48 | 25 | 3.33 | 0.094 |
All regions presented are significant at p < 0.001 uncorrected from SPM8. Even when we can observe a bilateral distribution for neural activity between groups, non-parametric correction for multiple comparisons (MBHT) keep left cingulate gyrus (AB 31) as the only significant activation within LEL group. LEL: Low educational level, HEL: High educational level.
COMPARISON BETWEEN GROUPS
Significant activations for words > control condition (HEL > LEL) occurred in the right anterior cingulate gyrus (BA 24). Then, for the opposite contrast LEL > HEL (words > control condition) significant activations appeared in paracentral gyrus (BA 3,4). However, none of these activations reached the threshold of relevance for the multiple contrast correction testing (MBHT).
The verbs condition did not show significant activations in any of the tested contrasts, not even if the significance threshold was lowered to p < 0.005. Likewise, when comparing nouns > control condition (HEL > LEL), no significant activations were found. The opposite contrast, LEL > HEL, did show significant activations in the superior parietal (BA 7) and the bilateral precuneus (Table 1), even though none of them successfully passed the MBHT correction testing.
TASK PERFORMANCE
Regarding our second objective, participants' data were reorganized according to their performance in the naming task: Hi-P and Low-P groups, regardless of their EL. We took into account the same first level analysis for each subject.
ACTIVATION WITHIN EACH GROUP (Table 2)
Group | Contrast | Hemisphere | Brain region | Talairach coordinates (X/Y/Z) | Cluster size | Z score (peak voxel) | MBHT (independent t-test) |
---|---|---|---|---|---|---|---|
Hi-P | Words - CTRL | Right | Thalamus | 4/-4/-6 | 15 | 3.66 | 0.724 |
Hi-P | Words - CTRL | Right | Transverse temporal gyrus (BA41) | 34/-34/-12 | 22 | 3.47 | 0.051* |
Hi-P | Words-CTRL | left | Cerebellum culmen | 0/-40/-0 | 21 | 3.42 | 0.038* |
Hi-P | Verbs - CTRL | Left | Claustrum | -22/-18/-18 | 30 | 4.03 | 0.229 |
Low-P | Verbs - CTRL | Left | Caudate nucleus | -20/-18/-22 | 16 | 3.67 | 0.000 |
Hi-P | Nouns - CTRL | Right | Transverse temporal gyrus (BA41) | 34/-34/-12 | 22 | 3.47 | 0.801 |
Hi-P | Nouns - CTRL | Left | Fusiform gyrus (BA37) | -42/-52/-10 | 29 | 3.79 | 0.294 |
Hi-P | Nouns - CTRL | Left | Cerebelum culmen | 0/-40/-0 | 21 | 3.42 | 0.845 |
Statistical parametric activations for task performance groups (within group effects) uncorrected. We show with an asterisk the activations that remain significative after MBHT correction for multiple comparisons, which is transverse temporal gyrus and left cerebellum culmen for words contrast within Hi-P group; and Caudate nucleus for verbs contrast within Low-P group. All regions presented are significant at p < 0.001 uncorrected; MBHT (independent t-test) = MBHT for multiple comparisons correction (methods section). MBHT: morphologic-based hypothesis test, Hi-P: high performers, Low-P: low performers.
For words > control contrast, Hi-P group showed significant activations in the thalamus, as well as in the right temporal transversal gyrus (BA 41), and the left cerebellum culmen (Fig. 2). Temporal gyrus and left cerebellum maintain significativeness on MBHT correction. Low-P group did not show any significant differences in this contrast. For the verbs > control condition (Table 2), Hi-P group showed significant activations in regions of the left claustrum, but this could not surpass the MBHT correction. Low-P group showed activation of the caudate nucleus, and this one did reach statistical significance after being corrected (Fig. 2).
Regarding nouns > control contrast, Hi-P group showed significant activations in the right transversal temporal gyrus (BA 41), as well as in the left fusiform gyrus (BA 37) and cerebellum culmen, none of these three activations surpassed MBHT correction. Within the Low-P group, no differences for nouns > control condition were found.
COMPARISON BETWEEN GROUPS (Table 3)
Group | Contrast | Hemisphere | Brain region | Talairach coordinates (X/Y/Z) | Cluster size | Z score (peak voxel) | MBHT (independent t-test) |
---|---|---|---|---|---|---|---|
Hi-P - Low-P | Words-CTRL | Right | STG (BA22) | 60/-4/-6 | 15 | 3.61 | 0.132 |
Hi-P - Low-P | Words-CTRL | Left | Medial temporal gyrus (BA21) | -56/-14/-10 | 61 | 3.9 | 0.138 |
Hi-P - Low-P | Words-CTRL | Left | IFG (BA47) | -40/-26/-2 | 19 | 3.82 | 0.03** |
Low-P - Hi-P | Verbs-CTRL | Right | Precentral gyrus (BA6) | 44/-16/-26 | 28 | 4.23 | 0.244 |
Low-P - Hi-P | Verbs-CTRL | Right | Medial frontal gyrus (BA8) | 14/-30/-38 | 14 | 3.66 | 0.005*** |
Hi-P - Low-P | Verbs-CTRL | Left | IFG (BA47) | -40/-26/-2 | 70 | 4.37 | 0.928 |
Hi-P - Low-P | Verbs-CTRL | Left | Middle temporal gyrus (BA21) | -54/-16/-10 | 57 | 4.94 | 0.738 |
Hi-P - Low-P | Verbs-CTRL | Right | Middle temporal gyrus (BA22) | 60/-4/-8 | 25 | 4.3 | 0.006*** |
We show with an asterisk the activations that remain significative after MBHT correction for multiple comparisons. That is, left IFG for the Hi-P<Low-P contrast (words); right medial frontal gyrus for the Low-P>Hi-P contrast (verbs); and right middle temporal lobe for Hi-P<Low-P contrast (verbs). All regions presented are significant at p < 0.001 uncorrected; MBHT (independent t-test) = MBHT for multiple comparisons correction, MBHT: morphologic-based hypothesis test, IFG: inferior frontal gyrus, Hi-P: high performers, Low-P: low performers.
For the words > control contrast, Hi-P > Low-P activity was shown in the right STG (BA 22), left IFG (BA 47), and left medial temporal gyrus (BA 21). Right BA 47, as well as AB 22, kept their significance after the MBHT correction (Fig. 3). Low-P > Hi-P contrast did not present any significant activation for words > control condition. For the verbs > control condition (Table 3), Low-P > Hi-P contrast showed significant clusters in the right precentral gyrus (BA 6) and frontal medial gyrus (BA 8), this last zone kept its significance after MBHT correction (Fig. 3). On the opposite contrast (Hi-P > Low-P), activations were present in right STG (BA 22), as well as in left medial temporal lobe (BA 21) and left IFG (AB 47). MBHT correction only allowed significance for the right STG (BA 22) (Fig. 3). For the nouns > control condition, SPM did not show significant activations in any contrast.
Discussion
The first objective of this study was to analyze the effect of EL on brain activation patterns during lexical access in healthy elderly, that is to say, whether EL had direct repercussions on brain lateralization reorganizing trough aging. Brain metabolic activation was measured in older adults with HEL and LEL during a picture-naming task. Following the HAROLD model, our hypothesis assumed that there would be less lateralization in people with HELs compared to the LEL group. We also expected to find differences between both groups in areas such as the IFG (BA 47) and the anterior temporal cortex as a result of better and more numerous working memory strategies employed by people with HELs.
Our results showed that both groups of healthy older adults - without cardiovascular diseases or hypertension - different only for their number of regular schooling years show no differences in their brain lateralization patterns when compared to one another. We observed only one structure where differences are significant within the LEL group, corresponding to the left posterior cingulate gyrus (BA 31). These activations are interesting because they are consistent with those found in studies carried out in illiterate people and individuals with LEL.33-37 Castro-Caldas et al.38 as well as Petersson et al.15 showed differences in areas of the posterior cingulate cortex, as well as in areas of the splenium of the corpus callosum and the inferior parietal cortex15. In the same way, when Springer et al.17, analyzed different ELs, they found differences in the activation of the right posterior cingulate cortex and the temporoparietal region. The posterior cingulate gyrus (BA 31) is one of the structures more consistently active during tasks involving the semantic system39. The posterior cingulate cortex, as well as the cortex adjacent to splenium of the corpus callosum, has strong connections with the hippocampus through the cingulate bundle40. The posterior cingulate cortex could be acting as an interface between semantic retrieval and the episodic codification systems. In this sense, our results support the hypothesis that people with a LEL could be showing how they have preferred a lexical access route based on semantic representations and episodic memory to assure its efficiency14. These results could describe that people who have had only a few years of training in a schooling system could be developing many other strategies to link semantic representations and their lexical tags; however, we cannot certainly say that those results would be related to any alteration of the hemispheric dominance of language functions. That is to say that having few schooling years during childhood is not, in any way, a factor which would condemn cognitive functions in aging; neither has it seemed to determine changes on functional lateralization of language.
Nevertheless, the sample size was a limitation for our study. Possibly, if it would have been larger, we could have obtained statistical significance in activation areas that did not pass the multiple comparison correction. Nonetheless, it is important to insist on our variable control: stimuli were balanced in terms of frequency for EL, and participants were selected according to their cardiovascular health, to control for physiological as well as cognitive variables for isolating, as much as possible, the experimental variable. These results allow us to discuss and reconsidering the EL as a trustworthy variable. Nevertheless, years of schooling are an extremely wide concept and could fail to define cognitive development, academic progress, quantity, and quality of information storage. Besides, there is the blood supply variable; this could be much more significant to preserve brain tissue avoiding the reduction of asymmetry; however, more studies in this field could clarify this question. Finally, when considering task performance regardless EL, our results show a larger number of significantly active regions. This finding, alone, suggests that EL as a variable, defined just as a number of schooling years at developmental age does not allow us to observe differential effects in naming tasks for elders. Thus, we show that task performance is a more reliable and valid measure to build hypotheses that may provide helpful data for finding the factors that contribute to a successful aging process.
Our second goal was to evaluate the compensation/de-differentiation hypothesis associated with the asymmetry reduction in elderly. We assumed that if this lack of lateralization had a beneficial function, it would then appear in the group showing better performance, or on the contrary, if such reduction was a response without an adaptive benefit to face the aging process, then it should appear in the group presenting a larger number of mistakes. For this reason, participants' results were regrouped following their number of mistakes on the naming task, regardless of their EL. Our results show that the Hi-P group has significant activity in right temporal gyrus (BA 41), as well as in cerebellum culmen with a left side tendency. It is interesting to notice that the right cerebellum has been reported as being a structure with significant activity in different language tasks41. Functional connectivity fMRI studies have shown that cerebellum presents opposite lateralization for language than the one observed in the cortex42-46. Even if Hi-P group's activations appear in the culmen, coordinates locate the activation peak already within the left hemisphere, as well as in right superior and posterior temporal lobes; this could imply some reorganization of lateralization in cortical and subcortical structures, probably to preserve the efficiency of the lexical retrieval function.
In addition, when we subtract the Low-P group activation to the Hi-P group activation for the words > control contrast (verbs and nouns), the IFG (BA 47) was more significantly activated, this could mean that the group with better performance could be recruiting a significantly larger number of resources from the left IFG than the Low-P group. This region is strongly related with semantic processing working memory24, although not necessarily with lexical storage. Taking together, this data could be understood as the IFG was in charge of monitoring and keeping of verbal working memory span of semantic representations stored in the temporoparietal structures47. Wierenga et al.23 results commented above, compared activation between young and older adults during a picture-naming task. They reported that both groups performed well, and their findings showed that older adults present more activation in the right Broca's area (BA 44,45) and in the right IFG (BA 47) than young adults. Together, their findings and our results could suggest that the inverse lateralization pattern is not only an effect observed between younger and older people, but also between two groups with similar ages. As our results showed the left hemisphere keeps its dominance in some key structures for task execution and control, such less specialized, as the IFG, but not necessarily in others as the temporal plane of Wernicke's area that carry out sensory integration and lexico-semantic storage29,39,48.
In contrast, the Low-P group shows their most significant activity on the left caudate nucleus. This structure has been said to have important implications on the ability to inhibit unselected words during lexical retrieval49. Gil Robles et al.50 carried out an intrasurgical electric stimulation study on six patients. According to their findings, caudate functions may be specific to language because there were no facial or limb motor effects during stimulation. Consistently with these results, studies on diseases affecting basal ganglia, such as Parkinson, Huntington, and HIV, have shown that verbal fluency tasks can predict the integrity of the basal ganglia51,52. According to the semantic memory hybrid neuronal model proposed by Hart et al.53, the caudate nucleus interacts with the area localized anterior to the supplementary motor area (preSMA) and with the thalamus for semantic retrieval process. PreSMA region seems to be involved in beginning and ending the specific semantic category search, while thalamus sends information to the cortex and modulates the concepts activation, and the caudate seems to be responsible for taking the right decisions, helping for thalamocortical transmission and correct word selection depending on the search's intention; i.e., the caudate suppresses the competitor items by decreasing or inhibiting thalamocortical interaction. Caudate involvement in this process seems to depend on task difficulty53, what could be related to our study as there is higher activation within the group with more mistakes.
In the same sense, when we subtract the Hi-P group activation to the Low-P group activation, the Low-P group also showed significant cortical activity. Significant activations were found for verbs > control contrast in the right frontal medial gyrus (BA 8) corresponding to homologous PreSMA, which has been linked to the thalamus and the caudate head as an essential part to carry out word generation and category research processes53. Therefore, these structures form a searching, attention, and selection circuit for which the left medial frontal cortex is a key part to initiate, control, monitor, and terminate the search once the selection process is over. Our results showed that likewise the Hi-P group, participants with higher mistake rates (Low-P) show a HAROLD phenomenon but in the right frontal medial area (BA 8). Hence, as with the Hi-P group, we can see reorganization of the lateralization patterns, but it seems that when this happens in structures that are essential for the task, function efficiency starts to fail. This could mean that lateralization reduction does not have a compensatory effect.
Regarding grammatical categories, it is remarkable that participants made most mistakes with verbs. In fact, the most common error was having a verb image named with a noun, despite the training that participants received before the task. Verb generation and action image naming have been strongly linked with left medial frontal gyrus activations54-56. Besides, aphasia studies describe that verb naming is the main difficulty for patients with frontal injuries, as opposed to patients with posterior injuries54,57-59. Even more, Mesulam56 affirms that naming deficit can give us clear signs of cognitive impairment and dementia in early stages. Hence, we can hypothesize that the differentiation by grammatical categories could offer clues about the location of those deficits and the systems affected. For this reasons, more research concerning naming and mild cognitive impairment could help find better ways to make early diagnosis and design better rehabilitation programs.
In summary, our results showed a reduction of lexical access functional lateralization among older adults when compared according to their task performance. However, such reduction shows different patterns for each performance level. Older adults that achieve better performance for lexical access function keep a left lateralization pattern for key structures in semantic processing working memory like the IFG (BA 47), even if they show more interaction with homologous temporal and subcortical structures implicated with semantic processing. In contrast, older adults who made more mistakes during the task showed a pattern of greater activation in right frontal areas, homologous to those described as being central for initiation and persistence of lexical retrieval while keeping typical lateralization of subcortical structures. These results were in accordance to the HAROLD model, because we found a reduction in lateralization within elderly groups and allow us to provide data supporting the dedifferentiation hypothesis because the group presenting the inverse lateralization pattern in key structures for lexical access is the one integrated by people who made more mistakes during the task. This finding allows us to think of new questions about the functioning of different paths for lexical retrieval and the way these paths may get modified during the aging process.
Conclusions
EL, taken as the number of schooling years, according to our data, is not a determining factor for the reorganization of hemispheric asymmetry patterns. On the other hand, our study supports the de-differentiation hypothesis of the HAROLD model. That is, people who showed optimal performance for naming task, keep their left dominance of highly specialized structures for this function, as the IFG (BA 47). In contrast, people with lower scores in the same task show greater activation of homolog cortical structures involved in the initiation, searching, and ending lexical retrieval processes (right medial frontal gyrus, BA 8).