(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Attention is required for canonical brain signature of prediction error despite early encoding of the stimuli [1]

['Alie G. Male', 'Discipline Of Psychology', 'College Of Science', 'Health', 'Engineering', 'Education', 'Murdoch University', 'Perth', 'Department Of Psychiatry', 'Human Behavior']

Date: 2023-06

Abstract Prediction error is a basic component of predictive-coding theory of brain processing. According to the theory, each stage of brain processing of sensory information generates a model of the current sensory input; subsequent input is compared against the model and only if there is a mismatch, a prediction error, is further processing performed. Recently, Smout and colleagues found that a signature of prediction error, the visual (v) mismatch negativity (MMN), for a fundamental property of visual input—its orientation—was absent without endogenous attention on the stimuli. This is remarkable because the weight of evidence for MMNs from audition and vision is that they occur without endogenous attention. To resolve this discrepancy, we conducted an experiment addressing 2 alternative explanations for Smout and colleagues’ finding: that it was from a lack of reproducibility or that participants’ visual systems did not encode the stimuli when attention was on something else. We conducted a similar experiment to that of Smout and colleagues. We showed 21 participants sequences of identically oriented Gabor patches, standards, and, unpredictably, otherwise identical, Gabor patches differing in orientation by ±15°, ±30°, and ±60°, deviants. To test whether participants encoded the orientation of the standards, we varied the number of standards preceding a deviant, allowing us to search for a decrease in activity with the number of repetitions of standards—repetition suppression. We diverted participants’ attention from the oriented stimuli with a central, letter-detection task. We reproduced Smout and colleagues’ finding of no vMMN without endogenous attention, strengthening their finding. We found that our participants showed repetition suppression: They did encode the stimuli preattentively. We also found early processing of deviants. We discuss various explanations why the earlier processing did not extend into the vMMN time window, including low precision of prediction.

Citation: Male AG, O’Shea RP (2023) Attention is required for canonical brain signature of prediction error despite early encoding of the stimuli. PLoS Biol 21(6): e3001866. https://doi.org/10.1371/journal.pbio.3001866 Academic Editor: Ben Seymour, University of Cambridge, UNITED KINGDOM Received: September 24, 2022; Accepted: May 11, 2023; Published: June 20, 2023 Copyright: © 2023 Male, O’Shea. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are within the paper and its Supporting Information files. Funding: The authors received no specific funding for this work. Competing interests: The authors have declared that no competing interests exist. Abbreviations: EDRP, early deviant-related positivity; EEG, electroencephalogram; ERP, event-related potential; FIR, finite impulse response; ICA, independent component analysis; MMN, mismatch negativity; PCA, principal component analysis; PO, parieto-occipital; ROI, region of interest; vMMN, visual mismatch negativity

Our senses are flooded by information. For example, the 4 million cones of the human retina encode a staggering 101,200,000 bits of information at any instant if their responses to light were binary (which they are not) [1]. Still, we do not experience a flood—only a stream—of information to which we pay attention or of which we become conscious [2]. What we see does not require such an impossible burden of encoding because it is highly redundant, allowing prediction of the state of one photoreceptor at any instant from its state the instant before and from the state of its neighbors. This seminal idea developed into predictive-coding theory, first of the retina [3] and then of the brain’s hierarchical sensory systems [4] and ultimately to the whole brain [5–11].

Predictive-coding theory is a leading theory of how the brain deals with sensory input [12]. Friston [7,13–15] described the theory based on energy conservation and free-energy principles: It is that the brain uses past, bottom-up, sensory information, along with top-down information (such as prior probabilities, expectations, and attention) to generate predictive models of sensory input at various levels of the sensory pathway, linked by feedforward and feedback connections. To ensure good predictions, precision of predictions from previous bottom-up and top-down influences and from the success of previous predictions affect whether, or the degree to which, the model is updated [16–19]. For example, a prediction error for a small irregularity after high-precision predictions should prompt updating of the model, whereas the same irregularity after low-precision predictions may not [16].

We studied the visual mismatch negativity (vMMN) as our signature of prediction error [20]. The vMMN is an event-related potential (ERP) component from electroencephalography. Stefanics and colleagues [20] have reviewed its hundreds of studies: It occurs when a rare, unpredicted, deviant, visual stimulus occurs after a sequence of identical, standard, visual stimuli in a so-called oddball sequence. It is greater ERP negativity from parieto-occipital (PO) recording sites for deviants than for standards, occurring between 150 ms and 300 ms after the onset of the deviant. It is supposed to occur without endogenous attention on the deviant property of the stimuli.

There are 2 measures of the vMMN [20]. One is the classic vMMN, calculated as the difference between the ERP to deviants minus that to standards. The classic vMMN involves comparison of physically different stimuli and includes the greater neural adaptation of the frequently repeated standards than of the occasionally repeated deviants. The other is the genuine vMMN, calculated as the difference between the ERP to deviants minus that to some control (usually the equiprobable control [21,22]) stimuli, physically identical to the deviants and with equal neural adaptation, hence excluding it [23]. It has been used for hundreds of human ERP and animal single-cell experiments for various modalities. We focus on the genuine vMMN.

Recently, Smout and colleagues [24] found no ERP evidence of the genuine vMMN without endogenous attention to unexpected changes in orientation. In their attended condition, they showed each participant displays consisting of an annular Gabor patch (gratings of blurry dark and light bars) with a central gray patch containing a concentric black spot. They asked each participant to look only at the spot but to pay attention to the bars and to press a button on rare occasions when their spatial frequency increased. In their unattended condition, the same participant viewed the same displays but paid attention to the spot and pressed a button on rare occasions when its contrast decreased. The difficulty of the 2 tasks was equated for each participant in prior psychophysical testing.

Smout and colleagues’ stimuli of interest were the Gabor annuli. Participants viewed different-length sequences of displays of identical annuli—standards—followed by an otherwise identical annulus but with bars differing in orientation by ±20°, ±40°, ±60°, or ±80°—a deviant. On the next display, the deviant annulus was repeated, commencing a new sequence for which it now served as the standard—the so-called roving-standard paradigm [25]. To measure the genuine vMMN, Smout and colleagues used the equiprobable control.

Smout and colleagues found a genuine vMMN from PO electrodes to unexpected changes in the orientation of the bars in the attended condition (between about 170 ms to about 300 ms), but not in the unattended condition (their S1B Fig). They did find genuine, deviant-related responses in the unattended condition: a negativity from about 330 to 430 ms from a midline central electrode (Cz) and from about 300 to 470 ms from a midline frontal electrode (Fz). But these are too late to be considered vMMNs, or any sort of prediction error from such an early brain-processing feature of visual input as orientation [26].

Smout and colleagues confined their remaining analyses to their attended condition to answer questions about the timing and orientation tuning of prediction error. But the absence of a genuine vMMN without endogenous attention was surprising. Smout and colleagues did not discuss it except to cite one study for a similar finding in the auditory modality [27]. To resolve this discrepancy, we conducted an experiment addressing 2 alternative explanations for Smout and colleagues’ finding: that it was from a lack of reproducibility or that participants’ visual systems did not encode the stimuli when their endogenous attention was on something else.

We made some changes to improve our study’s ability to explore these alternatives. Smout and colleagues’ stimuli were Gabor annuli with an outer diameter of 11° diameter and an inner diameter of 0.83°. Following Smout and colleagues’ argument that bigger is better, we doubled the size of the outer diameter of our Gabor patches. Smout and colleagues’ inner gray patch interrupted the central bars over a region of about the size of the human foveola to which about 20% of the processing of the visual system is devoted [28]. To include as much as possible of that area, we eliminated any central gray patch, making our central bars continuous except for a very small area occupied by the lines of small task-relevant fixation letters.

We also omitted an attended condition, eliminating any possible confound in Smout and colleagues’ design from participants’ moving their eyes to the bars [29], thus bringing them onto central vision. We also analyzed clusters of electrodes around the single electrodes Smout and colleagues reported. This was to avoid issues related to variability in placement of single electrodes [30].

To look for evidence that participants encoded the orientation of the Gabor patches, we searched for repetition suppression: attenuated brain activity to repeated standard stimuli [31,32]. For Gabor patches similar to those used by Smout and colleagues and by us, this attenuation appears in various components of ERPs, as late as the P2 [32] and as early as the P1 at PO electrodes [33].

We found no genuine vMMN to unattended orientation deviants even though participants encoded the stimuli.

Discussion The current study has 3 important findings: When unexpected, unattended changes to only the orientation of visual input occur, the typical brain signature of prediction error, the genuine vMMN, does not occur. This confirms the similar finding by Smout and colleagues, suggesting that theirs is not due to any reproducibility error.

Participants encoded the stimuli by showing repetition suppression. The encoding extends into the time range of the vMMN, as others have reported [32,33,35,36].

There is a sign of prediction error about 90 ms after onset of the stimuli: the EDRP. Below, we consider why unattended, unexpected changes of orientation do not seem to produce a well-accepted neural signature of prediction error, the vMMN, and why they seem to produce an EDRP. We also consider the implications of our findings for existing models of attentional control and stimulus selection. Why no vMMN but an EDRP? Absence of a genuine vMMN to unattended, unexpected changes in orientation is consistent with 5 experiments including ours [24,29,37,38], but inconsistent with 4 [22,39–41]. We consider 5 possible explanations: There really is no genuine vMMN to unexpected, unattended orientation changes, and the 4 out of 9 studies that reported it are examples of publication bias or methodological errors. There is a genuine vMMN to unexpected, unattended orientation changes, and the 5 studies that failed to find it suffered from some yet-to-be-determined methodological differences or problems. Because the orientation changes were not attended, they were processed early, but not for long enough to engage the vMMN mechanism. Endogenous attention is necessary for the vMMN. The vMMN is affected by the EDRP. 1. No genuine vMMN to unexpected orientation changes? A 44% hit rate across studies for finding a genuine vMMN to unexpected, unattended orientation changes is something to take notice of. But it is possible there are studies reporting no vMMN that never made it into print, which would reduce that rate. The vMMN literature does indeed show signs of publication bias. For example, of the 22 orientation-vMMN published studies, including those only on the classic vMMN, we are aware of only 4 reporting no vMMN for a hit rate of 82%. All but one [29] also reported a different condition or experiment in which a vMMN was found, suggesting that experiments in which all conditions failed to show a vMMN are either not submitted for publication or submitted and rejected. Such a bias would make it difficult to identify the experimental parameters required to resolve whether the genuine vMMN to orientation changes occur without attention. 2. Methodological differences or problems? We considered whether our experimental parameters were responsible for our failure to find a genuine vMMN. We can rule out that our: Orientation differences (15° was our smallest) were too small, because observers can discriminate orientation differences between gratings of about 0.4° [42]. Moreover, the experiments that did find a genuine vMMN tested an orientation difference of 30° to 36° [22,39,43–45], whereas we and Smout and colleagues fruitlessly tested some much larger orientation differences.

Presentation time (80 ms) was too short, because others have shown genuine vMMNs for other unattended, more complex deviants with presentation times as short as 17 ms [46].

Our interstimulus interval (280 ms) was too short because others have shown genuine vMMNs with intervals as short as 210 ms [47].

Our use of the roving-standard paradigm was responsible because others [24,48] did find a genuine orientation vMMN using it. Moreover, that paradigm has been used successfully to show MMNs to various deviants [9,49–51]. In a review of the literature, we noted that researchers generally failed to isolate individual low-level features, such as orientation, from other, confounding factors including eye movements that are capable of changing simple orientation changes into more complex visual changes [29]. Even Smout and colleagues’ study suffered from this, with very careful control of fixation in their unattended condition (the participants’ task was to detect contrast changes in a small central spot, requiring them to look at it) but not in their attended condition (the participants’ task was to detect spatial-frequency changes in the bars of the oriented stimuli, 0.42° away from the fixation spot). Male and colleagues [29] have shown evidence that participants’ eyes do stray away from central fixation when the task-relevant information is elsewhere, despite instructions to keep fixation central. A potential limitation of our study is that we cannot discount the possibility that participants found the letter task so easy that endogenous attention, or even overt attention, drifted away from fixation. If so, we would anticipate some evidence of a vMMN given that Smout and colleagues found a mismatch signal when participants attended to the stimulus of interest. We did not find any such evidence. Additionally, if attention or distraction were different for roving-standard versus equiprobable-control sequences, we would anticipate differences in the ERP waveforms for deviant and control stimuli, if not in task performance; however, we found both to be markedly similar. We take this as a good indicator that our control condition, like Smout and colleagues’, was suitably designed for exploring unattended irregularities in low-level properties of visual input. 3. Unattended orientation changes are processed early, but not in the vMMN range? Potentially, unattended orientation changes are processed early but not for long enough or for far enough in the visual pathways to engage the vMMN mechanism. This could explain our finding of the EDRP, our findings on repetition suppression, and our (and Smout and colleagues’) failure to find a vMMN to unattended stimuli. It would also be consistent with the theory that although unattended sensory inputs are fully preserved in the early, feedforward pathways for about 200 ms, only attended inputs are maintained for further processing in later, large-scale, recurrent networks for consciousness [52,53]. However, our and other researchers’ experimental data show that the repetition suppression we observed overlaps with the time range of the vMMN as does the processing of unattended inputs. There is indeed experimental evidence that processing of unattended visual stimuli diminishes with time [54–56]. For example, Moerel and colleagues [56] showed participants superimposed rectangular gratings of different orientations and colors (blue and orange) for 100 ms; participants pressed a key when a cued grating (e.g., 90°, blue) appeared in a particular sequence of trials including all possible combinations of orientations and colors. They analyzed EEG data only from trials well before or after trials containing a cued or partially cued grating (e.g., matched the orientation but not the color). They trained pattern classifiers to discriminate 2 presented orientations 90° apart. Moerel and colleagues found that classifier accuracy was above chance for uncued orientations from about 80 ms after onset but diminished with time, whereas accuracy for cued orientations diminished less with time. But the times for which unattended stimuli yield significant activity are all more than 400 ms after onset [54–56]. This is well into, and past, the vMMN range, from 150 ms to 300 ms. Even our P2, repetition-suppression result, with peak activity at 268 ms, is well beyond the notional time at which feedforward inputs are theoretically meant to give way to recurrent processing. We conclude that the processing of our unattended stimuli, both standards and deviants, endured well into the vMMN range. 4. Endogenous attention is necessary for the vMMN? One could argue that we and Smout and colleagues did not find a vMMN because endogenous attention is necessary for it. Auksztulewicz and Friston [27] have made this proposition for auditory input. Doing so seems to us to undermine one of the most frequently cited evolutionary purposes of prediction error: to monitor one’s immense sensory input preattentively and to select only those that elicit prediction errors. We have no dispute with the notion that endogenous attention serves to increase the precision of prediction errors, as Smout and colleagues and others have argued [16,17,27], thereby engaging the vMMN mechanism. The neural mechanism is via regulation of the postsynaptic gain of error units undertaking the top down-sensory comparison and production of prediction errors. 5. The vMMN is affected by the EDRP? We found evidence of an early prediction error, the genuine EDRP, at the P1, about 90 ms after onset of the Gabor patch. Of course, we need to be cautious about the EDRP’s reliability. Although we have found a similar early genuine, deviant-related positivity for orientation and for other low-level stimulus features [29] and others have reported early positivities for deviants either with a vMMN [57–60] and similar [61] or without [38], Smout and colleagues showed no sign of an EDRP. Theories on the early positivity range from memory-based change-detection, to changes in stimulus eccentricity [61], to hemi-field mediated deviance detection [60]. Nevertheless, we can think of 3 reasons why presence or absence of the vMMN might be affected by the EDRP: Energy efficiency,

Stimulus complexity, and

Precision of prediction. Energy efficiency. One feature of predictive-coding theory is that it is designed for energy efficiency [7,14,15]. It is inefficient to respond to a prediction error at one level of the visual system if that error had been responded to at an earlier level. To decide whether this is reasonable, we had recourse to the better-studied auditory system. When there are unexpected, unattended changes to auditory input, an MMN occurs from about 150 to 250 ms after onset [62–64]. More recently, ERP evidence has been found for processing of changes in physical properties of auditory input as early as 40 ms: the middle-latency response [65–70]. It might be akin to our EDRP. However, one important characteristic of the middle-latency response is that it is always followed by the MMN, possibly because model updating is accomplished only by the MMN. This is consistent with predictive coding theory’s harmony with evolutionary reasons. A prediction error requires a listener to assess whether a preattentive stimulus represents a threat or a mating opportunity, thereby converting the stimulus into one demanding exogenous attention, yielding later ERP activity, in this case the MMN. The visual EDRP does not show the same properties as the auditory middle-latency response: We found no hint of a vMMN after the EDRP. It will remain unknown whether the EDRP performs model updating until the EDRP is replicated and scrutinized for evidence of model updating. Although our study is not designed for revealing model updating, we hope that our findings will encourage studies better suited in pursuing the answer to this question. Stimulus complexity. If the EDRP for simple features of visual input involves both detection of a prediction error and updating of the predictive model, it is possible that more complex aspects of visual input require the vMMN. Our Gabor patches changed only one low-level property—orientation—without any other low-level changes, such as of contrast, luminance, spatial frequency, or retinal location. Other studies with spatially complex figures, such as windmill patterns, providing more global, higher-level properties of the stimuli, have generally found a genuine vMMN [37,71–73]. Precision of prediction. The most likely tool in predictive-coding theory’s box is precision of prediction. This is the notion that variability of environment determines prediction precision, which, in turn, determines the prediction error [16,74]. Attention modulates precision [13,75]. Smout and colleagues found supporting evidence, admittedly only from their endogenously attended condition, by showing that the selectivity of a mismatch signal was greater when an unexpected orientation change followed more identical presentations of an unchanging orientation. Implications for models of attentional control and stimulus selection There are various neural models of attentional control and stimulus selection [76–78]. They share being brain-wide neural pathways, focusing on separate but interacting pathways from the sensory cortices to higher levels: the dorsal attention network and the ventral attention network [79,80]. The dorsal network, involving goal-directed, endogenous attention, includes sensory cortices for whatever modality is required to accomplish a goal, such as the visual cortex when searching for red fruit in green foliage. It has feedforward and feedback connections to numerous brain areas mainly of the dorsal pathways including dorsal frontoparietal regions involved with attentional control. The dorsal network is active for as long as the searching continues, maintaining goal-directed behavior including finding and picking the fruit, enhancing processing in the visual cortex for features specific to red fruit, and suppressing the ventral network. The ventral network also begins with the visual cortex when an unattended, unexpected change in visual input happens, such as the appearance of a venomous snake in the fruit tree. The visual cortex has other feedforward and feedback connections to numerous brain areas mainly of the ventral pathways including middle and inferior prefrontal regions. Then, both networks become transiently activated to accomplish the new goal-directed behavior: to escape from the snake. Ideally, the activity of the 2 networks correspond to the components typically found from ERP studies of changes in visual input. When an unexpected change is unattended, the ventral network is supposed to mediate the vMMN [81]; when the unexpected change is attended (task relevant), then the dorsal network is supposed to mediate the later, P300 component [82]. But the coarse resolution of source localization from ERP studies casts some doubt on whether this ideal is true [79]. The speculative implication of our finding of the EDRP is that sometimes an unattended, unexpected orientation change can fail to activate the supposed ventral network’s vMMN. To distinguish this notion from some explanation via low precision of prediction will require studies beyond the scope of ours, perhaps with advanced source resolution or using dynamic causal modeling.

Conclusion We hope we have convinced you it is possible that unattended, unexpected changes in orientation do not yield the vMMN (consistent with the findings of Smout and colleagues) but might yield the EDRP. If so, the EDRP is a useful lead to understanding how the absence of a vMMN without attention could be reconciled with predictive-coding theory and models of attentional control and stimulus selection.

Methods We showed 21 [Smout and colleagues had 24] participants sequences of identically oriented Gabor patches (Fig 1A), standards, and, unpredictably, otherwise identical Gabor patches differing in orientation by ±15°, ±30°, and ±60° [Smout and colleagues used ±20°, ±40°, ±60°, and ±80°; we continue to give their values in brackets after ours], deviants (80 ms duration [100 ms], 280 ms [500 ms] interstimulus interval, 634 patches [415] per block; see Fig 1A). Participants looked at a central, desynchronized stream of letters and pressed a key whenever an X appeared (Fig 1B). Standard sequences ranged in length randomly from 3 to more than 11. For an equiprobable control, we presented the same stimuli, but in a random order (see Fig 1D). Ethics statement The Human Research Ethics Committee at Murdoch University approved the experiment (permit 2015 208). All protocols adhered to the principles of the Declaration of Helsinki. All participants provided their written informed consent and were free to withdraw from the experiment at any time. Participants Twenty-one self-declared healthy adults (8 males, 18 right-handed) with normal or corrected-to-normal vision participated in the experiment (statistical power = .88, based on Kimura and Takeda’s [39] reported effect size for genuine vMMN at PO8: Cohen’s d = −0.73). The mean age was 33 years with a range from 18 to 60 years. Most of the participants were undergraduate psychology students at Murdoch University. Participants received course credit or entry into a draw to win a $50 gift card in return for participation. Apparatus Participants sat in a light-attenuated chamber viewing a calibrated monitor (17-inch, color, cathode ray tube display; Sony Trinitron Multiscan E230) from 57 cm. The monitor showed 1,280 × 1,024 pixels (75 Hz refresh rate) and was the only source of light. A chin rest stabilized each participant’s head. Participants gave their responses by pressing a key on a 4-key EGI response box with the index finger of their dominant hand. A PC running Linux (v4.13.0), GNU Ubuntu (v16.04.4), Octave (v4.0.0) [83], and Psychophysics Toolbox (v3.0.14) [84–86] delivered the visual stimuli and recorded behavioral responses. An iMac running EGI’s NetStation 5.2 recorded EEG data. Stimuli We used achromatic Gabor patches (mean RGB values of 128, 127, 128) on an average gray background (RGB values of 128, 127, 128). A Gabor patch comprises a grating of a particular spatial frequency and orientation whose contrast reduces with distance from the center of the grating according to a Gaussian function [87,88]. These stimuli are like those used by Smout and colleagues. Gabor patches had a contrast of .999 [Smout and colleagues used 1.00], a phase of 0 radians—the black-to-white crossing was in the center of the patch [not specified], a spatial frequency of 2.4 cycles per degree of visual angle [2.73], and a standard deviation of the Gaussian of 3.84° of visual angle [not specified]. The visible parts of each Gabor patch had a diameter of approximately 22° of visual angle [11°, with a 0.83° central blank area]. There were 12 possible orientations: 8°, 23°, 38°, 53°, 68°, 83°, 98°, 113°, 128°, 143°, 158°, and 173° clockwise from vertical (0°) [9 possible orientations from 0° to 160°, in 20° steps] (further details in S1 Table). For the participants’ primary task, capitalized letters were superimposed on the center of the Gabor patch in cyan (RGB values of 0, 255, 255) in 30-point Courier font [a black fixation spot]. Letters occupied 0.5° (width) × 0.6° (height) with a line width of 0.06° [visual angle of 0.3° diameter]. We used all letters of the English alphabet except for “I”. We give an illustration in Fig 1. Procedure The participants’ task was to fixate on, and attend to, the continually changing random sequence of letters in the center of the monitor and to press a key with the dominant hand whenever an X appeared. Each letter occurred for 600 ms. None of the letters was immediately repeated. If the participant responded between 0.15 and 1.2 seconds after target onset, the response was correct. There were 383 letter changes during a block. On average, there were 15 targets (ranging from 7 to 25) per block. Desynchronized from the letters, we presented our stimuli of interest—concentric Gabor patches—for 80 ms separated by a 280-ms interstimulus interval. We had 2 sorts of blocks. Roving-standard blocks The experimental, roving-standard blocks contained 634 trials [Smout and colleagues: 415] showing Gabor patches of which 114 (17.98%) [59 (14.29%)] were deviants. Standard trials showed orientations from among twelve 15° steps from 8° clockwise from vertical (0°) [nine 20° steps from 0°]. Deviant Gabor patches differed in orientation from the preceding standard stimuli by ±15°, ±30°, or ±60° [±20°, ±40°, ±60°, or ±80°]. We used a hazard distribution to equalize the probability of the appearance of a deviant on any trial. Consequently, the distribution of numbers of standard trials against serial position had a gamma shape, with fewest trials for the largest serial positions [89,90]. We give the mean percentages of the numbers of standard-trial runs across all participants in S1 Table. On average, 5.53 standards preceded each deviant. Orientation for each standard was approximately 8.3% each. We give the mean numbers over participants in S1 Table. Equiprobable-control blocks For each participant, we built equiprobable-control blocks from that participant’s roving-standard blocks. We replaced each block’s standard stimuli—except for the ones immediately preceding a deviant and the deviant itself—with a Gabor patch whose orientation was randomly chosen among the 12 possible orientations. The stimulus preceding the deviant and the deviant itself were identical in the 2 sorts of blocks. Control trials were the trials in the equiprobable-control block that were identical to the roving-standard block’s deviant trials. These 2 identical trials in both sorts of blocks allow us to discount the possibility that a difference in pairing of trials contributed to any deviant-related differences. There were 6 roving-standard blocks and 6 equiprobable-control blocks. We randomized block order afresh for each participant. Each block took less than 4 minutes to complete. Participants were free to take breaks between blocks. EEG recording and analysis We recorded the electroencephalogram (EEG) using an EGI, 129-channel, dense-array HydroCel geodesic sensor net, and Net Amps 300 signal amplifier. We recorded EEG at a 500-Hz sampling rate. Impedances were below 50 kΩ, per recommendation [91] and existing standard (e.g., [72,92–95]) for high-input impedance amplifiers. All electrodes were referenced to Cz. We processed the EEG data offline using MATLAB 2015b (MathWorks, USA), EEGLAB 14.1.1 [96], and ERPLAB 6.1.4 [97]. We re-referenced the signal of all electrodes to the common average and filtered the EEG with a low-pass 40-Hz Kaiser-windowed (beta 5.65) sinc finite impulse response (FIR) filter (order 184) followed by a high-pass 0.1-Hz Kaiser-windowed (beta 5.65) sinc FIR filter (order 9056). Epochs were 400 ms long, featuring a 50-ms prestimulus period for baseline correction to accommodate the short 360 ms stimulus-onset-asynchrony. We excluded epochs with amplitude changes exceeding 800 μV at any electrode [98,99]. We identified electrodes with unusually high deviations in EEG activity relative to the average standard deviation pooled from all electrodes using the method described by Bigdely-Shamlo and colleagues [100]. A robust z-score was calculated for each electrode by replacing the mean by the median and the standard deviation by the robust standard deviation (0.7413 times the interquartile range). We removed any electrode with a robust z-score exceeding 2.0 provided at least 4 others surrounded it for later interpolation. We created vertical and horizontal EOG channels by bipolarizing data from electrodes above and below the right eye (electrodes 8 and 126) and outer canthi of both eyes (electrodes 1 and 32), respectively (as in [72]). To ensure that trials where participants moved their eyes or blinked were not included in the final analysis of the data, we identified epochs containing amplitude changes exceeding ±60 μV at these EOG channels for rejection prior to ICA correction. We did not exclude these epochs at this stage, and ICA artifact correction had not yet occurred. We performed independent component analysis (ICA) with AMICA [101]. To improve the decomposition, we performed the analysis on raw data (excluding bad electrodes) that was first filtered using a 1-Hz high-pass (Kaiser-windowed sinc FIR filter, order 804, beta 5.65) and 40-Hz low-pass filter as suggested by Winkler and colleagues [98]. Thereafter, we segmented into epochs, but without baseline correction [102]. We simultaneously reduced the data to 32 components. We applied the demixing matrix to the 0.1- to 40-Hz filtered data and used SASICA [103] to identify which components exhibited low autocorrelation, low focal electrode or trial activity, high correlation with vertical or horizontal EOG, or met ADJUST criteria [104]. We assessed the remaining components using criteria described by Chaumon and colleagues [105], classifying components based on consistent activity time-locked to stimulus onset across all trials, on topography, or on power spectrum. We removed components identified as unrelated to brain activity. We then removed epochs that were previously identified for rejection. This was followed by a final artifact rejection—removing epochs containing amplitude changes exceeding ±60 μV at any electrode. Finally, using spherical splines [106], we interpolated data for removed electrodes. For exploring deviant-related responses, we averaged ERPs for the standard, deviant, and control trials, excluding epochs immediately following a deviant or control trial. We produced difference waves by subtracting ERPs to standards and ERPs to controls from ERPs to deviants. For exploring repetition suppression, we averaged PCA components for standards from roving-standard blocks and for randomly oriented stimuli from equiprobable-control blocks at varying sequential positions, excluding deviant and control trials. We retained standard trials immediately following deviants; we retained randomly oriented trials immediately following controls (S1 Table for number of epochs in each ERP). We defined anterior, central, and PO regions of interest (ROIs). The signal from all electrodes within each ROI was averaged over time. We visually inspected ERPs for electrodes within each ROI and then when averaged together. There was middle (M) frontal (M F) central (M C), parieto-occipital (M PO), left (L) frontal (L F), central (M C), parieto-occipital (L PO), and right frontal (R F), central (R C), and parieto-occipital (R PO). We display the results of 6 ROIs (M F, M C, M PO, L PO, R PO; see S2 Table for electrodes within each cluster) and 1 PZ electrode for comparison with Smout and colleagues. Using the EP Toolkit (v2.64; [34]), we conducted temporal PCA on the individual average ERP data from all 129 electrodes, thereby allowing for deviant-related responses isolated to any one electrode to emerge. For exploring deviant-related responses, we included deviant and control trials. For exploring repetition suppression, we included standard and random trials. We used Promax orthogonal rotation (κ = 3) with a covariance relationship matrix and Kaiser weighting as recommended by Dien and colleagues [107]. PCA reduced the data to those components explaining most of the observed data. The component that explained the greatest amount of signal is component 1, and every component thereafter explained less of the data than the component before it. We show how much of the signal a component is responsible for by plotting the component’s loading (scaled by SD) over time [108]. Using PCA, one can identify separate components in the ERP waveform and extract an alternative measure of ERP component amplitudes for inferential testing [107,109,110]. Each PCA component has a peak latency, a site of maximum positivity on the scalp (i.e., a component’s positive pole), and a site of maximum negativity on the scalp (i.e., a component’s negative pole). Plotting a topographical map of the microvolt scaled PCA data of a single component at the time of its peak latency shows the component’s positive and negative pole. A vMMN component would emerge as a component that is largest (most negative) between 150 and 300 ms, based on previous vMMN studies using orientation deviants (e.g., 162 to 170 ms in [43]; 200 to 250 ms in [22]; 190 to 220 ms in [39]). A genuine vMMN component should also yield scores that are more negative for deviants compared to controls at the component’s negative pole. In addition to our frequentist paired t tests, we conducted Bayesian paired t tests to determine the likelihood of obtaining the data. We used a medium prior (with a Cauchy prior whose width was set to 0.707) for all Bayesian analyses. For interpreting our findings, a model with the largest Bayes Factor (BF 10 ) is the model that best explains the data; this is the favored model. All main effects and interactions in the favored model are, therefore, important for explaining the data. Evidence against the null is considered weak if a BF 10 is between 1 and 3. It is positive for a BF 10 between 3 and 20, strong for a BF 10 between 20 and 150, and very strong given a BF 10 greater than 150 [111]. We also performed Bayes Factor replication (BF r0 ) tests [112] on mean amplitudes between 197 and 207 ms at each PO ROI given a prior reflecting the effect size (Cohen’s d = −0.73) for the genuine vMMN at the PO8 in Kimura and Takeda’s [39] oddball condition for 32.7° orientation deviants. For all frequentist ANOVAs, we applied the Greenhouse–Geisser correction where necessary (ε < .750). Eta squared (η2) denotes the estimated effect size.

Acknowledgments We are grateful to Urte Roeber, who contributed to conception, methodology, review, and supervision; to Sabine Grimm, Cooper Smout, and Philip Smith for helpful discussion; and to the students who assisted in collecting the data for this experiment.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001866

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.