Accumulation System: Distributed Neural Substrates of Perceptual Decision Making Revealed by fMRI Deconvolution.

Neural substrates of evidence accumulation have been a central issue in decision-making studies because of the prominent success of the accumulation model in explaining a wide range of perceptual decision making. Since accumulation-shaped activities have been found in multiple brain regions, which are called accumulators, questions regarding functional relations among these accumulators are emerging. This study employed the deconvolution method of functional magnetic resonance imaging (fMRI) signals from human male and female participants during object-category decision tasks, taking advantage of the whole-brain coverage of fMRI with improved availability of temporal information of the deconvolved activity. We detected the accumulation activity in many non-category-selective regions (NCSRs) over the frontal, parietal, and temporal lobes as well as category-selective regions (CSRs) of the categorization task. Importantly, the frontal regions mostly showed activity peaks matching the decision timing (classified as "type-A accumulator"), while activity peaks of the parietal and temporal regions were behind the decision (classified as "type-B accumulator"). The CSRs showed activity peaks whose timing depended on both region and stimulus preference, plausibly reflecting the competition among the alternative choices (classified as "type-C accumulator"). The results suggest that these functionally heterogeneous accumulators form a system for evidence accumulation in which the type-A accumulator regions make decisions in a general manner while the type-B and type-C accumulator regions are employed depending on the modality and content of decision tasks. The concept of the accumulation system may provide a key to understanding the universality of the accumulation model over various kinds of decision tasks.SIGNIFICANCE STATEMENT Perceptual decision making, such as deciding to walk or stop on seeing the signal colors, has been explained theoretically by the accumulation model, in which sensory information is accumulated to reach a certain threshold for making decisions. Neural substrates of this model, however, are still under elucidation among candidate regions found over the brain. We show here that, taking advantage of the whole-brain coverage of functional magnetic resonance imaging (fMRI) with improving availability of temporal information by deconvolution method, the accumulation is conducted by a system comprising many regions in different abstraction levels and only a part of these regions in the frontal cortex make decisions. The system concept may provide a key to explaining the universality of the accumulation model.

Stochastic process underlying emergent recognition of visual objects hidden in degraded images.

When a degraded two-tone image such as a "Mooney" image is seen for the first time, it is unrecognizable in the initial seconds. The recognition of such an image is facilitated by giving prior information on the object, which is known as top-down facilitation and has been intensively studied. Even in the absence of any prior information, however, we experience sudden perception of the emergence of a salient object after continued observation of the image, whose processes remain poorly understood. This emergent recognition is characterized by a comparatively long reaction time ranging from seconds to tens of seconds. In this study, to explore this time-consuming process of emergent recognition, we investigated the properties of the reaction times for recognition of degraded images of various objects. The results show that the time-consuming component of the reaction times follows a specific exponential function related to levels of image degradation and subject's capability. Because generally an exponential time is required for multiple stochastic events to co-occur, we constructed a descriptive mathematical model inspired by the neurophysiological idea of combination coding of visual objects. Our model assumed that the coincidence of stochastic events complement the information loss of a degraded image leading to the recognition of its hidden object, which could successfully explain the experimental results. Furthermore, to see whether the present results are specific to the task of emergent recognition, we also conducted a comparison experiment with the task of perceptual decision making of degraded images, which is well known to be modeled by the stochastic diffusion process. The results indicate that the exponential dependence on the level of image degradation is specific to emergent recognition. The present study suggests that emergent recognition is caused by the underlying stochastic process which is based on the coincidence of multiple stochastic events.

Language comprehension dependent on emotional context: a magnetoencephalography study.

In communication, language can be interpreted differently depending upon the emotional context. To clarify the effect of emotional context on language processing, we performed experiments using a cross-modal priming paradigm with an auditorily presented prime and a visually presented target. The primes were the names of people that were spoken with a happy, sad, or neutral intonation; the targets were interrogative one-word sentences with emotionally neutral content. Using magnetoencephalography, we measured neural activities during silent reading of the targets presented in a happy, sad, or neutral context. We identified two conditional differences: the happy and sad conditions produced less activity than the neutral condition in the right posterior inferior and middle frontal cortices in the latency window from 300 to 400 ms; the happy and neutral conditions produced greater activity than the sad condition in the left posterior inferior frontal cortex in the latency window from 400 to 500 ms. These results suggest that the use of emotional context stored in the right frontal cortex starts at ∼300 ms, that integration of linguistic information with emotional context starts at ∼400 ms in the left frontal cortex, and that language comprehension dependent on emotional context is achieved by ∼500 ms.

Phase-interpolated averaging for analyzing electroencephalography and magnetoencephalography epochs.

Stimulus-locked averages of electroencephalography (EEG) and magnetoencephalography (MEG) epochs reveal characteristic waveforms. EEG/MEG generation models to have reconstruct such waveforms have been recently proposed. These models assume that evoked, phase-modulated, and amplitude-modulated activities occur solely or simultaneously. We propose a two-stage stimulus-locked averaging method, called phase-interpolated averaging, to investigate the EEG/MEG generation process. First, virtual EEG/MEG epochs, which would be obtained as if instantaneous phases for each time sampling point were on a phase-grid, are interpolated from actually measured EEG/MEG epochs. Then, the virtual EEG/MEG epochs are discrete Fourier transformed. A simulation revealed that the zeroth Fourier term revealed the evoked activity, the first Fourier term revealed the amplitude-modulated activity, and the condition number of the interpolation reflected the phase-modulated activity. On the basis of these facts, a preliminary EEG analysis implied that the evoked activity is much smaller than what would be expected by using conventional averaging, the evoked and phase-modulated activities simultaneously occur, and the amplitude-modulated activity occasionally associates with the evoked and phase-modulated activities. To the best of our knowledge, this is the first time that these three activities have been shown to coexist by actually separating them.

Inference in alpha rhythm phase and amplitude modeled on Markov random field using belief propagation from electroencephalograms.

Alpha rhythm is a major component of spontaneous electroencephalographic (EEG) data. We develop a novel method that can be used to estimate the instantaneous phases and amplitudes of the alpha rhythm with high accuracy by modeling the alpha rhythm phase and amplitude as Markov random field (MRF) models. By using a belief propagation technique, we construct an exact-inference algorithm that can be used to estimate instantaneous phases and amplitudes and calculate the marginal likelihood. Maximizing the marginal likelihood enables us to estimate the hyperparameters on the basis of type-II maximum likelihood estimation. We prove that the instantaneous phase and amplitude estimation by our method is consistent with that by the Hilbert transform, which has been commonly used to estimate instantaneous phases and amplitudes, of a signal filtered from observed data in the limited case that the observed data consist of only one frequency signal whose amplitude is constant and a Gaussian noise. Comparison of the performances of observation noise reduction by our method and by a Gaussian MRF model of alpha rhythm signal indicates that our method reduces observation noise more efficiently. Moreover, the instantaneous phase and amplitude estimates obtained using our method are more accurate than those obtained by the Hilbert transform. Application of our method to experimental EEG data also demonstrates that the relationship between the alpha rhythm phase and the reaction time emerges more clearly by using our method than the Hilbert transform. This indicates our method's practical usefulness. Therefore, applying our method to experimental EEG data will enable us to estimate the instantaneous phases and amplitudes of the alpha rhythm more precisely.

Masked immediate-repetition-priming effect on the early lexical process in the bilateral anterior temporal areas assessed by neuromagnetic responses.

A masked priming paradigm has been used to measure unconscious and automatic context effects on the processing of words. However, its spatiotemporal neural basis has not yet been clarified. To test the hypothesis that masked repetition priming causes enhancement of neural activation, we conducted a magnetoencephalography experiment in which a prime was visually presented for a short duration (50 ms), preceded by a mask pattern, and followed by a target word that was represented by a Japanese katakana syllabogram. The prime, which was identical to the target, was represented by another hiragana syllabogram in the "Repeated" condition, whereas it was a string of unreadable pseudocharacters in the "Unrepeated" condition. Subjects executed a categorical decision task on the target. Activation was significantly larger for the Repeated condition than for the Unrepeated condition at a time window of 150-250 ms in the right occipital area, 200-250 ms in the bilateral ventral occipitotemporal areas, and 200-250 ms and 200-300 ms in the left and right anterior temporal areas, respectively. These areas have been reported to be related to processing of visual-form/orthography and lexico-semantics, and the enhanced activation supports the hypothesis. However, the absence of the priming effect in the areas related to phonological processing implies that automatic phonological priming effect depends on task requirements.

Phase-compensated averaging for analyzing electroencephalography and magnetoencephalography epochs.

Stimulus-locked averaging for electroencephalography and/or megnetoencephalography (EEG/MEG) epochs cancels out ongoing spontaneous activities by treating them as noise. However, such spontaneous activities are the object of interest for EEG/MEG researchers who study phase-related phenomena, e.g., long-distance synchronization, phase-reset, and event-related synchronization/desynchronization (ERD/ERS). We propose a complex-weighted averaging method, called phase-compensated averaging, to investigate phase-related phenomena. In this method, any EEG/MEG channel is used as a trigger for averaging by setting the instantaneous phases at the trigger timings to 0 so that cross-channel averages are obtained. First, we evaluated the fundamental characteristics of this method by performing simulations. The results showed that this method could selectively average ongoing spontaneous activity phase-locked in each channel; that is, it evaluates the directional phase-synchronizing relationship between channels. We then analyzed flash evoked potentials. This method clarified the directional phase-synchronizing relationship from the frontal to occipital channels and recovered another piece of information, perhaps regarding the sequence of experiments, which is lost when using only conventional averaging. This method can also be used to reconstruct EEG/MEG time series to visualize long-distance synchronization and phase-reset directly, and on the basis of the potentials, ERS/ERD can be explained as a side effect of phase-reset.

Neuroimaging study on brain asymmetries in situs inversus totalis.

Situs inversus totalis (SI) is a rare condition in which all visceral organs are arranged as mirror images of the usual pattern. The objective of this study was to determine whether SI individuals have reversed brain asymmetries. We performed a neuroimaging study on 3 SI subjects and 11 control individuals with normally arranged visceral organs. The language-dominant hemisphere was determined by magnetoencephalography. Left-hemispheric dominance was observed in 1 SI subject and all controls, whereas right-hemispheric dominance was observed in the remaining 2 SI subjects. Statistical analysis revealed that language dominance patterns in SI subjects were different from those in the controls, suggesting that the developmental mechanisms underlying visceral organ asymmetries are related to those underlying functional brain asymmetry. Anatomical brain asymmetries were determined by magnetic resonance imaging. SI subjects had the same planum temporale (PT) asymmetry pattern as the controls, but a reversed petalia asymmetry pattern. The inferior frontal gyrus (IFG) asymmetry pattern varied within both groups, indicating a relationship between the rightward IFG and right-hemispheric language dominance. These results suggest that the developmental mechanisms underlying visceral organ asymmetries are related to those underlying petalia asymmetry but not to those underlying PT and IFG asymmetries, and that brain asymmetries might develop via multiple region-dependent mechanisms.

Neural activations correlated with reading speed during reading novels.

Functional magnetic resonance imaging was used to measure neural activations in subjects instructed to silently read novels at ordinary and rapid speeds. Among the 19 subjects, 8 were experts in a rapid reading technique. Subjects pressed a button to turn pages during reading, and the interval between turning pages was recorded to evaluate the reading speed. For each subject, we evaluated activations in 14 areas and at 2 instructed reading speeds. Neural activations decreased with increasing reading speed in the left middle and posterior superior temporal area, left inferior frontal area, left precentral area, and the anterior temporal areas of both hemispheres, which have been reported to be active for linguistic processes, while neural activation increased with increasing reading speed in the right intraparietal sulcus, which is considered to reflect visuo-spatial processes. Despite the considerable reading speed differences, correlation analysis showed no significant difference in activation dependence on reading speed with respect to the subject groups and instructed reading speeds. The activation reduction with speed increase in language-related areas was opposite to the previous reports for low reading speeds. The present results suggest that subjects reduced linguistic processes with reading speed increase from ordinary to rapid speed.

Early neural activation for lexico-semantic access in the left anterior temporal area analyzed by an fMRI-assisted MEG multidipole method.

To determine the time and location of lexico-semantic access, we measured neural activations by magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) and estimated the neural sources by fMRI-assisted MEG multidipole analysis. Since the activations for phonological processing and lexico-semantic access were reported to overlap in many brain areas, we compared the activations in lexical and phonological decision tasks. The former task required visual form processing, phonological processing, and lexico-semantic access, while the latter task required only visual form and phonological processing, with similar phonological task demands for both tasks. The activation areas observed among 9 or 10 subjects out of 10 were the superior temporal and inferior parietal areas, anterior temporal area, and inferior frontal area of both hemispheres, and the left ventral occipitotemporal area. The activations showed a significant difference between the 2 tasks in the left anterior temporal area in all 50-ms time windows between 200-400 ms from the onset of visual stimulus presentation. Previous studies on semantic dementia and neuroimaging studies on normal subjects have shown that this area plays a key role in accessing semantic knowledge. The difference between the tasks appeared in common to all areas in the time windows of 100-150 ms and 400-450 ms, suggesting early differences in visual form processing and late differences in the decision process, respectively. The present results demonstrate that the activations for lexico-semantic access in the left anterior temporal area start in the time window of 200-250 ms, after early visual form processing.

Phonological influences on lexicosemantic processing of kanji words.

To investigate the phonological influences on the lexicosemantic process with a strong orthographic constraint, we used kanji (morphogram) homophone words and measured, using magnetoencephalography, the neural activities during the silent reading of prime-target pairs. The primes were phonologically the same as or different from the targets or pseudocharacters. The neural activities in the left posterior temporal and inferior parietal areas became weaker with phonological repetition. Furthermore, stronger activities for the different condition in the left anterior temporal area and for the same condition in the left inferior frontal cortex, respectively, suggest the roles of these areas of the brain in the semantic processing of words and in the selection of appropriate meanings. We conclude that phonological information affects the lexicosemantic process even with a strong orthographic constraint.

Brownian motion, fluctuation and life.

The measurements of dynamic behaviors of biomolecules in relation to their functions have been allowed using single molecule measurements. Thermal Brownian motion causes random step motion of motor proteins and structural fluctuation of protein molecules between multiple states. In hierarchic structure of life, the fluctuation is modulated. Random fluctuation is biased to directional motion and reactions as a result of interaction of proteins. The fluctuation of kinetic state of signaling proteins results in polarization and localization of cells. A recognition process in brain is also explained by the equation analogous to biochemical reaction at the molecular level. Thus dynamic processes originated from thermal motion may play an important role in activation processes in life.

Discrete stochastic process underlying perceptual rivalry.

In perceptual rivalry such as ambiguous figure perception and binocular rivalry, the conscious percept spontaneously alternates between two stable interpretations of an unchanging stimulus. It is well known that the time intervals of the perceptual alternation follow a gamma distribution (GD), but its implication for the alternation mechanism has not been clarified. We examined quantitatively GDs fitted to alternation intervals, and found that the shape-determining parameter alpha of the GDs took natural numbers. Because a GD determined by a natural number alpha is mathematically obtained from a discrete stochastic process (Poisson process), our result indicates that such a stochastic process underlies perceptual rivalry and that the alpha-time accumulation of the discrete events causes a perceptual alternation.