Introduction Two different speech coding systems exist in the cochlea: one represents spectral information encoded tonotopically according to the space theory, 1 while the other codes for temporal information, as represented by the firing pattern of the primary auditory neurons. Spectral information is known to play an important role in speech recognition, such as the formant frequencies in vowel recognition. 2 In contrast, the role of temporal information in speech recognition has not been fully elucidated, despite many previous animal experiments 3–7 which have confirmed the production of temporal information in the cochlea. One reason for this lack of information is the lack of appropriate methods for investigating the temporal coding system on the auditory cortex in humans. In the present study, a train of clicks of ultra-short duration were constructed from original speech sounds, which would be expected to contain only temporal information with stable spectral structures. We used this sound stimulation to examine the role of temporal information in speech recognition subjec- tively by a recognition test, and objectively by positron emission tomography (PET) in normal subjects. Materials and Methods Method for constructing stimulation sounds: Original speech sound waves were recorded by a digital audio-tape (DAT) recorder and transferred to a computer (Macintosh Quadra 950, Apple Inc.). We identified the zero-crossing points where the original speech sound waves crossed the baseline from nega- tive to positive and then measured each of the inter- vals between two zero-crossing points. The stimulation sound was constructed by arranging biphasic clicks with silent intervals, which had been measured above as the zero-crossing intervals. The duration of each click was 41.6 s. The computer program for this processing was written using Symantec C++ 7.0. Figure 1 shows the original speech sound waveform [e] and the stimulation sound made from this waveform. Thus, the stimulation sounds were composed of click sequences. Recognition tests: To examine whether the stimula- tion speech sounds could be easily recognized, 10 healthy volunteers (all male, mean age 32 years, range 26–42) underwent recognition tests for these stimu- lation sounds. In test 1, the stimulation sounds made from five vowels were presented randomly twice per vowel, Auditory and Vestibular Systems, Lateral Line 1 1 1 1 1 p © Rapid Science Publishers Vol 8 No 9–10 7 July 1997 2395 TO elucidate the temporal coding system for speech recognition, we synthesized stimulation sounds which do not contain formant information but do contain temporal information by transforming original sound wave to click sequences. Using this stimulation sound, we performed a recognition test and used PET to examine the cortical activities in normal subjects listening to this sound. The results of the recognition test showed a good perception of the sounds made from sequential speech. The PET study demonstrated signifi- cant activation of the superior temporal gyri while listening to the stimulation speech sounds. Our results imply that these stimulation sounds were processed semantically in the auditory cortices. The temporal processing system is thought to make an important contribution to speech recognition. Key words: Auditory cortex; Click sequences; PET; Speech recognition; Temporal coding The role of the temporal coding system in the auditory cortex on speech recognition Hisayoshi Kojima, CA Shigeru Hirano, Kazuhiko Shoji, Yasushi Naito, Iwao Honjo, Yoko Kamoto, 1 Hidehiko Okazawa, 1 Koichi Ishizu, 1 Yoshiharu Yonekura, 2 Yasuhiro Nagahama, 3 Hidenao Fukuyama, 3 Junji Konishi 1 Department of Hearing and Speech Science, 1 Nuclear Medicine, and 3 Neurology, Graduate School of Medicine, Kyoto University, Sakyo- ku, Kyoto 606; 2 Biomedical Imaging Research Center, Fukui Medical School, Shimoaizuki, Matuoka, Fukui 910-11, Japan CA Corresponding Author NeuroReport 8, 2395–2398 (1997)