Cortical lateralization during verb generation: a combined ERP and fMRI study

Abstract

Lateralization of scalp-recorded event-related potentials (ERPs) and functional MRI (fMRI) activation was investigated using a verb generation task in 10 healthy right-handed adults. ERPs showed an early transient positivity in the left inferior temporal region (500–1250 ms) following auditory presentation of the stimulus noun. A sustained slow cortical negativity of later onset (1250–3000 ms) was then recorded, most pronounced over left inferior frontal regions. fMRI data were in agreement with both ERP effects, showing left lateralized activation in inferior and superior temporal as well as inferior frontal cortices. Lateralized ERP effects occurred during the verb generation task but not during passive word listening or during word- and nonword repetition. Thus, ERPs and fMRI provided convergent evidence regarding language lateralization, with ERPs revealing the temporal sequence of posterior to anterior cortical activation during semantic retrieval.

Introduction

The development of noninvasive imaging techniques to assess hemispheric lateralization of language is required both for the presurgical evaluation of eloquent cortex and for understanding the normal development and specialization for speech and language. A number of tasks have been used to investigate language lateralization with functional imaging. The most robust lateralizing effects, particularly prominent in the inferior frontal cortex, were observed during word retrieval tasks, such as generating verbs in response to nouns Benson et al., 1999, Frith et al., 1991, Petersen et al., 1988, Raichle et al., 1994, Rutten et al., 2002, Thiel et al., 1998, Warburton et al., 1996.

Although functional MRI (fMRI) provides excellent spatial resolution, the time course of cortical activation cannot be studied with precision. Moreover, not all subjects are able to tolerate the scanning environment, in particular young children. This highlights the necessity for complementary, noninvasive methods for the assessment of language lateralization. Event-related potentials (ERPs) immediately preceding speech onset can be recorded from the intact scalp (McAdam and Whitaker, 1971), corresponding to similar responses recorded directly from the exposed language cortex in patients undergoing neurosurgery (Fried et al., 1981). Unfortunately, scalp-recorded prespeech potentials are often contaminated by extracerebral artifacts, caused by muscle activity, breathing, and tongue movements Grözinger et al., 1980, Szirtes and Vaughan, 1977. In order to avoid or minimize artifact contamination, ERPs can also be recorded during covert language processing and with deferred speech output (reviewed in Rugg et al., 1986). Lateralized slow ERP shifts have been recorded during phonological judgment tasks Rugg, 1984a, Rugg, 1984b and synonym generation Altenmüller et al., 1993, Thomas et al., 1997. These slow cortical potential (SCP) shifts take the form of a sustained negativity in frontocentral regions that is typically greater over the left than the right hemisphere. The correlation between language lateralization derived from SCPs and functional imaging, however, has not been studied, although a few reports have shown the feasibility of comparing magnetoencephalographic (MEG) localization with functional imaging Dale et al., 2000, Dhond et al., 2001, Eulitz et al., 1994, Singh et al., 2002.

ERP lateralization is commonly assessed by comparing homologous electrodes over each hemisphere. In contrast, lateralization using functional imaging methods is mostly determined by calculating laterality indices based on the number of activated voxels in a specified region of interest of each hemisphere (Hertz-Pannier et al., 1997). These laterality indices, however, are dependent on the statistical threshold used. More recently, a direct statistical method for comparison of fMRI activation in homotopic voxels of each hemisphere was proposed (Liégeois et al., 2002). The computation of statistical lateralization maps (SLMs) successfully revealed lateralization of language in a group of pediatric surgical patients who also underwent invasive assessments of language representation. This method of computing SLMs, based on a direct comparison of fMRI activation in homotopic voxels, lends itself to the comparison with ERP lateralization based on a direct comparison of homologous electrodes.

The main purpose of the current study was to determine the degree to which ERPs/SCPs and fMRI provide convergent evidence on language lateralization in the same group of healthy adult volunteers using a covert verb generation task. In order to achieve this, the fMRI verb generation task (Liégeois et al., 2002) was adapted for use with ERPs, by using the same time interval for covert verb generation to allow comparison between the two methods. ERP lateralization was defined a priori based on the following features: (a) negative polarity ERPs/SCPs over frontocentral regions Altenmüller et al., 1993, Thomas et al., 1997, and (b) positive ERPs from posterior temporoparietal regions, as observed by Kutas and Hillyard, 1980, Kutas and Hillyard, 1982, and Snyder et al. (1995).

Two additional aims were (1) to explore regional differences in the time course of ERP/SCP activation; and (2) to determine the degree to which ERP/SCP lateralization effects are task dependent. For the ERP task, a tone cue was used to signal the end of the covert word retrieval phase. Such S1–S2 deferred stimulus response tasks are known to induce large amplitude SCPs over the frontal cortex (termed contingent negative variation, CNV, Walter et al., 1964). Through inclusion of control ERP tasks, which placed different demands on working memory and phonological processing, the task specificity of ERP/SCP lateralization was examined.