Introduction : The theory of a bipolar organization of language was revised in the light of neuroimaging results and language seems to be processed in a large scale cognitive network. The organization and function of this neural network may be modified by several factors including gender and age. There is some controversy about sex differences in language according to behavioural and neuroimaging studies :

• a small but significant difference favouring females was observed for language abilities.
• a greater bi hemispheric functional Magnetic Resonance Imaging (fMRI)activation was described in females during language tasks, but Frost et al found no differences during a language comprehension task according to gender.

Age differences were mainly studied in childhood, but no consistent functional imaging data on aging are available.

Objective : To study the influence of gender and age on cortical fMRI activations during a silent semantic task.

Subjects : 48 right-handed native French-speaking subjects were included in 4 groups of 12 according to age and gender : young female (YF), Young Male (YM), Old Female (OF), and Old Male (OM). Old subjects had no Alzheimer's disease or other dementia according to their MMS score (> 26) and NINCDS-ADRDA criteria.

Method :

• fMRI : Data were collected on a 1.5 T General Electric Signa MR scanner.

Functional images covered the whole brain and were scanned using a single-shot echoplanar imaging sequence (FOV :

24 cm ¡thickness : 5 mm - 64 x 64 matrix - TE : 60 ms ¡ TR : 3s ¡ flip angle : 90°). The paradigm followed a box car design and 3 periods of activation (30 s duration) alternated with 3 periods of fixation (30s duration).

During the active condition subjects had to mentally generate words corresponding to the semantic category (animals, furniture, or fruit) displayed through MR-compatible special goggles.During the control condition, they were requested to concentrate on their breathing. Data were analyzed using the FSFAST package. MR data were motion corrected using AFNI and then smoothed with a FWHM of 5 mm. The voxel intensities were rescaled such that the in-brain grand mean was 1000 for all subjects. The hemodynamic response was modeled as a gamma function (delay : 2.25 s - dispersion :1.25 s). The offset and linear drift in the BOLD signal were removed by estimating those components simultaneously with the amplitude of the hemodynamic response. A t-test was then performed on the amplitude of the hemodynamic response to determine which voxels had significant paradigm-related activity.

• anatomical MR : Results were displayed on the inflated cortical surface for each subject. A group analysis was performed using a surface coordinate system and random model effect and General linear model using age and gender as regressors.