Abstracts

Rationale: To study language processing, reverse speech has often been used as a control task in noninvasive brain-mapping studies utilizing functional magnetic resonance imaging (fMRI), positron emission tomography, and electroencephalography. The rationale is that reverse speech is comparable to forward speech in terms of auditory characteristics, while omitting the linguistic components. This finds some support in fMRI studies indicating that reverse speech results in less blood-oxygen-level-dependent (BOLD) signal intensity in perisylvian regions than forward speech [Brain Lang, 2010; 14:90-100]. However, some authors argue reverse speech as an invalid control since it is not always incomprehensible as language [Nature, 1999; 398:760]. We are interested in incorporating a control task into our study of language using invasive electrocorticography (ECoG). We attempted to validate a reverse speech control task using ECoG in patients with focal epilepsy.Methods: We studied six adolescent and adult patients with focal epilepsy who underwent extraoperative ECoG prior to resective surgery. All patients received preoperative MRI and an auditory naming task during ECoG recording [Neuroimage, 2008; 41:1120-31]. In pseudorandom order, 118 audible question stimuli were presented, recorded, and integrated with ECoG; including 30 in reverse. Questions were designed to elicit 1 or 2 word answers; e.g. Q: What flies in the sky? Patients were to answer I don t know when they could not answer correctly. Only the 30 reverse speech and the 30 corresponding forward speech trials were analyzed. ECoG traces were time-locked to stimulus-onset as well as stimulus-offset. Augmentation of gamma-oscillations (80 120Hz) above baseline was determined at each temporal lobe site and transformed into time-frequency matrices. Sites of significant gamma augmentation beginning within 300msec after stimulus-onset [J Neurosci, 2010; 30:16643-50] and ending prior to 300msec after stimulus-offset were defined as auditory sites. Sites activated otherwise were defined as non-auditory language.Results: Across subjects, 24 temporal lobe electrode sites over superior or middle temporal gyri showed significant gamma-augmentation, including 17 auditory and 7 non-auditory language sites. Reverse speech trials elicited a larger peak gamma-augmentation response than did the corresponding forward speech trials at nearly all auditory sites (median: 116% vs 64%; p=0.009 by Wilcoxon Signed Rank Test) and most non-auditory language sites (median: 80% vs 54%; p=0.063). Area under the curve of gamma-amplitude was larger during reverse speech compared to forward speech at nearly all auditory sites (p <0.001) and most non-auditory language sites (p=0.091).Conclusions: We failed to validate reported BOLD-fMRI evidence of weaker activation of temporal cortices during reverse compared to forward speech. Our findings directly measuring electrical brain activity suggest the opposite: reverse speech more strongly activates temporal cortices. This may be associated with increased attention to reverse speech. Further characterization of the reverse speech task is required.