Article Figures & Data
Figures
- Figure 1.
θ oscillations during Go/No-Go task in frontal and visual cortices. A, Schematic illustration of experimental setup. Mice performed Go/No-Go task in custom-built head restrainer. B, Experimental procedures. Task required subject to discriminate direction of random-dot motion (rightward: S+, downward: S–). Go response (i.e., licking) to S+ and S– motion resulted in reward and punishment, respectively. There was no particular reinforcing stimulus following No-Go response (i.e., no licking). The variation of the performance across days and within session was analyzed (Extended Data Fig. 1-1). C, Example EEG data of two trials. Upper, For S+ motion, licking response was counted as “hit,” and water reward was followed instantaneously. Motion stimulus was ceased at first lick (RT). Lower, For S– motion, No-Go response was counted as “CR.” Scale bars on bottom right indicate 500 ms and 300 µV. D, Summary of behavioral results. From all subjects (five mice), and all sessions (10 d), data of 8390 trials were collected. In this study, only the trials from CR (n = 2411) were analyzed. E, Time-frequency representation of grand-averaged event-related spectral perturbation (ERSP) of each cortical area. In both the frontal and visual cortices, prominent θ (4–12 Hz) oscillations were observed during 4 s of stimulus presentation (No-Go period). The stimulus-locked and response-locked spectrograms in all trials are summarized in Extended Data Figure 1-2. F, Temporal average (0.5–4 s) of θ oscillations in each cortical area. Numbers denote peak frequency of each channel. FA: false alarm, Fr: frontal cortex, Vis: visual cortex, Amptd: amplitude, Freq: frequency.
- Figure 2.
Task performance and θ amplitude. A, Distribution of pc (proportion correct) in all CR trials (n = 2411). pc was calculated across 10 previous trials. With a threshold of chance level (pc = 0.5), each trial was classified as either attentive (pc > 0.5) or inattentive (pc ≤ 0.5) state. B, Change of mean RT as function of pc. The RT of hit trials decreased as pc increased (Pearson’s r = –0.051, p < 0.05), suggesting the change of attentional level within the hit trial epochs. C, Scatterplot of frontal/visual θ amplitude with color coding of pc. Each gray dot represents the observed θ amplitude value of a single CR trial. Gray solid lines indicate the probability density function of each channel. Along with the changes of pc, the mean amplitudes of frontal and visual θ showed a negative relationship. D, Mean amplitude of frontal/visual θ as function of pc. The linear correlation between θ amplitude and pc was positive in the frontal (r = 0.076, p < 0.001) and negative in the visual (r = –0.099, p < 0.001), suggesting their distinct behavioral correlates. E, Time-frequency representation of oscillatory amplitudes highlighting state-dependent and region-dependent change of θ amplitudes. Black dashed lines indicate the start/end of the visual stimuli. Error bars represent ±1 SD (C) and SE (B, D) of the means. pc: probability correct. The changes in power spectral density and its peak frequency were calculated for different values of pc (Extended Data Fig. 2-1).
- Figure 3.
Task performance and θ-γ phase-amplitude coupling. A, Color-coded fast-frequency amplitudes as function of θ phase in frontal (top) and visual (bottom) and as function of pc (left to right). B, C, MI for local θ-γ coupling as function of pc (left: frontal, right: visual). In the frontal, the effect of attention on CFC was frequency-specific (increased CFC with low γ, decreased CFC with high γ). The attentional modulation of the visual coupling was not significant. The dependency of CFC on θ power or attentional state or their interaction was further analyzed using a general linear model, showing both θ power and its interaction with attentional state are significant independent parameters for CFC values (Extended Data Fig. 3-1). D, Grand-averaged amplitudes of fast oscillations as function of θ phase, showing noticeable regional difference of preferred θ phase (left: frontal, right: visual). E, Trial histogram of preferred θ phase of 20- to 40-Hz oscillations (left). Hexagons and dotted lines denote the peak and mean of each polar histogram, respectively. Mean preferred angle of each area (right). The distances between preferred angles over sites were both significant. Error bars represent ±1 SEM; *p < 0.05, ***p < 0.001. INA: inattentive, ATT: attentive.
- Figure 4.
Synchrony and lags of θ oscillations. A, Two-dimensional representation of joint phase histogram of frontal and visual θ oscillations as function of pc. The white color of each histogram indicates the higher probability of colocalization of the phase of two oscillations, suggesting a higher level of synchrony between two oscillations. The inset text indicates the maximum value of colocalization probability in each pc bin. B, Mean fronto-visual phase synchrony calculated by PLV as function of pc. The phase synchrony was higher for the attentive state than the inattentive state. Extended Data Figure 4-1A illustrates the phase-locking behaviors of two rhythms appeared as a transient locking. The phase synchrony is its maximal at the moment of frequency locking of two θ rhythms (Extended Data Fig. 4-1B), which is also dependent on the attentional state (Extended Data Fig. 4-1C). C, Cross-correlation analysis of example EEG epoch. Top, To calculate temporal lag (τ), raw time traces of EEG data from two channels were bandpass filtered in the θ band. Bottom, After filtering, the CCF was obtained and τ was defined as a temporal lag where the correlation coefficient reaches its maximum. Positive and negative values of τ indicate a lead of the visual and frontal θ, respectively. D, Trial histograms of τ between frontal and visual θ. The dots denote the peak of each histogram. Color-coded texts indicate the peak position from the lowest pc (green) and to the highest pc (red) bin. E, Mean τ as function of pc. The τ was smaller in the attentive state than in the inattentive state. Error bars represent ± 1 SEM; *p < 0.05, ***p < 0.001. INA: inattentive, ATT: attentive.
Extended Data
Extended Data Figure 1-1
Performances changing across- and within-session. A, Performance varies across sessions. Across the 10 d of sessions, overall behavioral performance (d’, M = 0.475, SE = 0.08) was significantly larger than chance level, Z = 5.845, p < 0.001. Learning effect, calculated by the difference of mean d’ between first 5 d (M = 0.399, SE = 0.07) and last 5 d (M = 0.551, SE = 0.21), was not significant, Z = 0.405, p = 0.686. Error bars represent ± 1 SEM. B, Performance varies within a session. Splitting trials of single session into three different groups, initial (first 1/3), middle (second 1/3), and final (last 1/3), change of behavioral performance within session was analyzed. In general, proportion of Go response trials (i.e., hit and FA) was higher at initial phase, and gradually decreased, suggesting the change of animals’ motivational level to water reward. Interestingly, despite the decrease of probability Go response, p(Go), overall probability of correct (i.e., hit and CR), p(correct), showed relatively small variation within session. C, Behavioral state changes across- and within-session. Changes of proportion of “attentive”-labeled trials (pc > 0.5) as a function of experimental session (left) and as a function of within-session progress (right) were analyzed. To calculate the effect size, one-way ANOVA (Kruskal–Wallis test) was performed. We found experimental session explained 17% of the variance, while the within-session progress explained 35% of the variance, suggesting the larger effect size of motivation-related factors relative to the learning-related factors. Error bars represent ± 1 SEM. D, Duration and transitions of behavioral state. Behavioral state showed transitions (i.e., transition from inattentive to attentive, and vice versa) and stationary phase with certain duration (inattentive: 6.60 trials in average, attentive: 7.81 trials in average; left). Using Markov-chain model with two-state process, this state transition was summarized in a state diagram (right). Download Figure 1-1, TIF file.
Extended Data Figure 1-2
Grand-averaged EEG responses during Go/No-Go task. A, Stimulus-locked grand-averaged amplitude spectrogram of four behavioral outcomes. Black dashed lines indicate the onset of visual stimulus. Hit and FA trials exhibited relatively large changes in oscillatory amplitudes, compared to those of Miss and CR trials. These large activities were considered as reinforcer-related activities (i.e., water reward, air puff punishment, motor movement, etc.) as they were locked to the moment of reaction (see C). B, Averaged (t = 0.5–4 s after motion onset) amplitude spectrum of spectrograms in A. C, Same as A, but response-locked. For the trials which do not contain behavioral response (CR, Miss), mean RTs of the session from hit and FA trials were used to align the epoch. D, Same as B, but response locked. Download Figure 1-2, TIF file.
Extended Data Figure 2-1
Baseline-uncorrected amplitude- and cross-spectral densities. A, Mean amplitude- or cross-spectrum of CR trials (n = 2,411, t = 0.5–4 s) calculated from the frontal. Error bars represent ± 1 SEM. B, Same as A, from the visual. C, Same as A, from cross-spectrum of the frontal and visual. D, Mean frontal θ (4–12 Hz) amplitude. Error bars represent ± 1 SEM. E, Same as D, from the visual. F, Same as D, from cross-spectrum of the frontal and visual. G, Mean peak frequency of the frontal θ-amplitude spectrum. Peak values were calculated in trial-by-trial basis. Colored numbers illustrate the duration of one cycle of each frequency. Error bars represent ± 1 SEM. H, Same as G, from the visual. I, Same as G, from cross-spectrum of the frontal and visual. Download Figure 2-1, TIF file.
Extended Data Figure 3-1
GLM analysis of CFC. To dissect the confounding effect of θ and attention on MI, MIs were further analyzed through GLM with three predictors: θ amplitude, attentive state, and two-way interaction of θ amplitude and attentive state. As a result, we found statistically significant main effect of θ amplitude on θ-low-γ coupling in both regions (frontal: t = 14.529, p < 0.001, visual: t = 9.902, p < 0.001) and on θ-high-γ coupling in the visual (t = 2.648, p < 0.01). The effect of θ amplitude on frontal θ-high-γ coupling did not reach statistical significance (t = 1.747, p = 0.081). Also, the effect of attentional state on MI was statistically significant only for visual θ-low-γ coupling (t = –2.864, p < 0.01) with non-significant effect on the other couplings (|t|s < 1.688, ps > .091). Most importantly, the two-way interaction between θ amplitude and attentional state was statistically significant for θ-low-γ coupling in both regions (ts > 2.294, ps < 0.05), but not for θ-high-γ coupling (|t|s < 1.301, ps > .193). Such two-way interaction with positive slope value (ß = 1.0 × 10–6 for frontal, ß = 1.3 × 10–6 for visual) suggests the effect of θ amplitude on θ-γ coupling becomes stronger with increased level of attention, especially for θ-low-γ coupling in both regions. In short, the change of CFC observed in this study can be best explained by a synergetic effect of θ amplitude and attentional state, rather than by single factor. Download Figure 3-1, PDF file.
Extended Data Figure 4-1
Transient frequency matching during fronto-visual θ synchrony. A, Example single-trial data showing transient frequency matching and increased phase synchrony. To calculate frequency difference and phase synchrony, raw data were converted into Hilbert instantaneous angle after band-pass filtering. Frequency difference was obtained by taking derivate of unwrapped angle difference between two channels’ data. Fr: frontal, Vis: visual, Ang: angle, Diff: difference, Freq: frequency. B, Two-dimensional histogram of instantaneous frequency difference and phase synchrony during the No-Go period of all trials (irrespective of attentional state). High synchrony was observed when frequency difference is low. Note that synchrony value was calculated with 1-s width and 10-ms sliding temporal window, and instantaneous frequency difference was time averaged with same size sliding window.
: instantaneous frequency of frontal θ,
In this issue
