The survival of animals depends on their ability to respond to different stimuli quickly and efficiently. From flies fluttering away when a swatter approaches, to deer running away at the sight of a lion to humans ducking to escape a looming punch, fast-paced reactions to harmful stimuli is what keep us (and other fauna) from getting injured or seriously maimed. This entire process is orchestrated by the nervous system, where cells called neurons carry signals from our senses to higher processing centres in the brain, allowing us to react appropriately.

However, this relay process from the sensory organs to the brain accumulates delays: it takes time for signals to be transmitted from cell to cell, and also for the brain to process these signals. This means that the information received by our brains is usually outdated, which could lead to delayed responses. Experiments done in cats and monkeys have shown that the brain can compensate for these delays by predicting how objects might move in the immediate future, essentially extrapolating the trajectories of objects moving in a predictable manner. This might explain why rabbits run in an impulsive zigzag manner when trying to escape a predator: if they change direction often enough, the predator may not be able to predict where they are going next.

Johnson et al. wanted to find out whether human brains can also compensate for delays in processing the movement of objects, and if so, at what point (early or late) in the processing pipeline the compensation occurs. To do this, they recorded the electrical activity of neurons using electroencephalography from volunteers who were presented with both static and moving stimuli. Electroencephalography or EEG records the average activity of neurons in a region of the brain over a period of time.

The data showed that the volunteers’ brains responded to moving stimuli significantly faster than to static stimuli in the same position on the screen, essentially being able to track the real-time position of the moving stimulus. Johnson et al. further analysed and compared the EEG recordings for moving versus static stimuli to demonstrate that compensation for processing delays occurred early on in the processing journey. Indeed, the compensation likely happens before the signal reaches a part of the brain called the visual cortex, which processes stimuli from sight. Any delays accrued beyond this point were not accommodated for.

Johnson et al. clearly demonstrate that the human brain can work around its own shortcomings to allow us to perceive moving objects in real time. These findings start to explain, for example, how sportspersons are able to catch fast-moving balls and hit serves coming to them at speeds of approximately 200 kilometres per hour. The results also lay the foundation for studying processing delays in other senses, such as hearing and touch.

Responding quickly to the world around us is a primary function of the central nervous system: catching prey, escaping predators, and avoiding moving objects (e.g. falling rocks) are all crucial to survival. This task is complicated by delays that accumulate during the neural transmission of information from the sensory organs to the brain. As a result, the brain only has access to outdated sensory information. Furthermore, additional delays accumulate during subsequent cortical processing of information. The world will inevitably change during this time, so how can the brain overcome this fundamental problem and keep up with an ever-changing world?

Several lines of evidence suggest that the brain can compensate for neural transmission delays through prediction: using information from the past to predict what is happening in the present (Nijhawan, 1994; Kiebel et al., 2008; Friston et al., 2010). Indeed, many researchers consider prediction to be a core objective of the central nervous system (Friston, 2010; Clark, 2013). This is relevant to all sensory processing, whether anticipating haptic input while manipulating an object or auditory input while listening to a melody. While experimental paradigms in cognitive neuroscience often remove the dimension of time by using artificially static displays (Millidge et al., 2022), delays are a fundamental obstacle in neural processing and should be accounted for in any comprehensive theory of brain function. This problem has received particular interest within the field of motion perception, as the future positions of a predictably moving object can be determined by the object’s current acceleration, velocity, and position. In this case, there is evidence that predictive processes help to compensate both for the neural delays incurred before visual input reaches the visual cortex and for the delays incurred during subsequent cortical processing (Orban et al., 1985; Berry et al., 1999; Jancke et al., 2004; Subramaniyan et al., 2018; Sundberg et al., 2006).

For example, neurophysiological recordings in animals reveal motion extrapolation mechanisms as early as the retina (Berry et al., 1999; Chen et al., 2013; Johnston and Lagnado, 2015; Liu et al., 2021; Souihel and Cessac, 2021). By responding to the leading edge of moving stimuli, retinal ganglion cells extrapolate the represented position of those stimuli, and are thought to transmit these extrapolated representations to visual cortex, thereby compensating for some of the lag that arises during transmission. These pre-cortical extrapolation mechanisms should effectively allow primary visual cortex to represent the position of a moving object with reduced latency, as observed in both cat and macaque V1 (Jancke et al., 2004; Subramaniyan et al., 2018). The existence of these extrapolation mechanisms opens the possibility that transmission delays on the way to visual cortex might be partially or fully compensated, allowing the early visual system to represent moving objects on predictable trajectories closer to their real-time locations.

It is unclear whether similar mechanisms operate along the cortical visual processing hierarchy to compensate for additional delays that accumulate as visual information is processed. On the one hand, there is suggestive evidence that position representations in areas V4 (Sundberg et al., 2006) and V5 (Maus et al., 2013a) are shifted for moving objects, potentially reflecting the effect of motion extrapolation in those areas. That interpretation is consistent with recent fMRI (Schneider et al., 2019; Harvey and Dumoulin, 2016), theoretical (Hogendoorn and Burkitt, 2019) and psychophysical (van Heusden et al., 2019) work suggesting that motion extrapolation mechanisms operate at multiple levels of the visual system. On the other hand, shifted position representations in higher areas might simply result from those areas inheriting extrapolated information from upstream areas such as V1. To our knowledge, no study to-date has investigated how the represented position of a smoothly moving object evolves over time as visual information about that object flows along the visual hierarchy.

Here, we address this question by using time-resolved EEG decoding to probe the position representations of smoothly moving objects across all levels of the human visual system in real time. We show that early position representations of moving objects are in close alignment with the veridical position of the object, providing the first direct evidence in humans that extrapolation processes allow the early visual system to localise moving objects in real time. We further show that during the course of cortical visual processing, position representations increasingly lag behind real-time stimulus position as information progresses through the visual hierarchy. This suggests that delay compensation is primarily achieved at very early stages of stimulus processing, and that subsequent cortical visual areas do not implement further compensation for neural delays. Nevertheless, this early compensation ensures that the represented position of a moving object throughout the entire visual hierarchy is far more up-to-date than could be expected on the basis of the latencies of neural responses to static objects. These findings demonstrate the existence of significant predictive processing during motion perception, but constrain any predictive mechanisms to acting relatively early in processing.

Twelve observers viewed sequences of black discs that were either flashed in one of 37 possible positions on a hexagonal grid (static trials) or smoothly moved through a series of positions on the grid along a straight trajectory (motion trials, Figure 1). Static trials were presented 252 times per position, and each of the 42 motion vectors was presented 108 times. EEG data were recorded over six testing sessions and analysed offline (see Methods). Multivariate pattern classifiers (linear discriminant analysis) were trained to discriminate stimulus position for all pairwise combinations of positions, using EEG activity evoked by static stimuli in those positions. These classifiers were subsequently tested on EEG data recorded during an independent subset of static trials, or during motion trials. Results of this classification analysis were combined to estimate the likelihood of the stimulus being present in each of the possible stimulus positions, $p_s(i)$ for $i\in \{1,2,\mathrm{\dots }37\}$, where $s$ is the presented position. From this, we traced the evolution over time of the estimated likelihood of the stimulus being present in the position in which it was actually presented (static trials) or moved through (motion trials), $p_s(s)$, hereafter referred to as the stimulus-position likelihood.

Figure 1

Download asset Open asset

This analysis was repeated for multiple combinations of training timepoint (i.e. time after onset of a static stimulus) and test timepoint (i.e. time after onset of a static stimulus or within a motion vector). Using different training timepoints allowed us to probe neural representations at different levels of the visual hierarchy, and testing at multiple timepoints allowed us to characterise how information flows through those levels over time during the epoch of interest (King and Dehaene, 2014). In this way, we were able to evaluate whether the neural position representation of a moving object flows through the visual hierarchy at the same latency as the position representation of a static flash. Additionally, this allowed us to evaluate how much the position representation of the moving object lags behind that object’s physical position.

Decoding position of static stimuli

First, we investigated the ability of classifiers to discriminate the presented position of static flashes based on the EEG signal. Figure 2a shows average classification accuracy across pairwise combinations of positions over time, grouped by distance between the two positions. Classifiers were trained and tested using subsets of data from the same timepoints. As expected, the performance of pairwise classifiers improved with increasing stimulus separation. This is due to the retinotopic organisation of visual cortex; stimuli elicit more distinct patterns of activity when they are further apart.

Figure 2 with 2 supplements see all

Download asset Open asset

Pairwise classification results were combined to calculate the stimulus-position likelihood. Stimulus-position likelihoods by eccentricity (distance from fixation) are plotted in Figure 2—figure supplement 1. We then averaged across all stimulus positions and participants. This likelihood was compared to a permuted null distribution to establish whether it was significantly above chance at each timepoint (Figure 2b, see Methods). The stimulus-position likelihood was above chance starting at 58 ms after stimulus onset, meaning that stimulus positions were decodable from this timepoint onward. An equivalent analysis was applied to eye position during static trials to ensure that EEG decoding results were not confounded by systematic microsaccades (Figure 2—figure supplement 1), either due to neural activity related to motor planning, movement-related EEG artefacts, or shifts in retinotopy (Mostert et al., 2018). For eye position, stimulus-position likelihood was not significantly above chance at any timepoint, which is in line with previous findings (Blom et al., 2020; Salti et al., 2015; Tse et al., 2002). Therefore, we can conclude that the classifiers trained on the EEG response to the static stimulus did not exploit eye movements to determine the stimulus position.

To assess whether position-related information was stable or variable across the time-course of the visual evoked response, the classification analysis was generalised across time (King and Dehaene, 2014): classifiers were trained and tested at all combinations of timepoints. Figure 2c shows the resulting temporal generalisation matrix (TGM), averaged across all stimulus positions and participants. The TGM was typical of position decoding plots seen in previous work (Hogendoorn and Burkitt, 2018). It appears that the stimulus representations before 150 ms training/test time did not generalise to other timepoints, but later reactivation of representations is apparent after 150 ms (off-diagonal red blobs). Finally, Figure 2d shows topographic maps of activation which contributed to classification of stimulus position (Haufe et al., 2014); these show that the relevant signal was mainly recorded from occipital electrodes, suggesting a source within the visual cortex, as expected.

Decoding position of moving stimuli

To decode the position of moving stimuli, we again trained classifiers on pairwise combinations of static stimuli, then applied these classifiers to EEG data recorded during motion trials. An illustration of each step in the analysis of motion trials is shown in Figure 3. As before, the stimulus-position likelihood was calculated, this time at each timepoint during motion epochs.

Figure 3 with 1 supplement see all

Download asset Open asset

We considered EEG epochs from 500 ms before to 500 ms after the timepoint at which the moving stimulus was exactly in each possible static stimulus location. This time-window was chosen to be broad enough to capture stimulus-evoked activity as the stimulus approached and receded from each position (moving from one position to the next took 400 ms). We then averaged the time-course of stimulus-position likelihoods across all six motion directions and 37 stimulus positions. The first position along each trajectory was excluded due to observed strong EEG responses to the initial onset of the stimulus.

The TGM derived from classifiers trained on static trials and tested on motion trials (Figure 3, step 3) revealed that classifiers trained on timepoints from around 100 ms were able to decode the position of moving objects. To identify timepoints at which classification was significantly above chance, we considered the performance of classifiers trained and tested on matching timepoints (diagonal of the TGM). Permutation testing revealed that decoding was significantly above chance for timepoints between 102 and 180 ms (Figure 3, step 4). Note that because we are investigating possible latency differences between the neural response to static and moving stimuli, maximal decoding is likely achieved off-diagonal, making this a conservative analysis choice.

Although the average stimulus-position likelihood was smaller in magnitude for moving stimuli compared to static stimuli, we observed that the location-specific neural response to motion over time was characterised by a gradual increase of the stimulus-position likelihood as the stimulus approached the centre of the position, then a decrease as the stimulus moved away on the other side. This is illustrated in Figure 3, step 5 (see Figure 3—figure supplement 1 for more training timepoints), and is similar to the pattern of activity found in response to a moving bar with direct recordings from cat V1 (Jancke et al., 2004).

Establishing the latency of position representations of moving stimuli

To investigate the latency at which neural position representations are activated for moving objects, we calculated the timepoint at which the peak stimulus-position likelihood was reached during motion sequences. Again, this was repeated for different training times as a proxy for different stages of neural processing. The time to peak likelihood in the test data for each training timepoint reflects the time at which the location-specific activity is most similar in the train and test set. This is assumed to be the time that the brain is representing the moving stimulus at the centre of a flash position. We use peak likelihood, as opposed to onset or a peak percentage, as the latency measure due to possible variations in receptive field (RF) size over the course of visual processing. As time elapses during stimulus processing, visual information reaches visual areas further up the processing hierarchy, which contain stimulus-selective neurons with larger RFs (Johnson et al., 2021; Harvey and Dumoulin, 2011). This means that a moving stimulus would enter the RF earlier in these later-activated brain regions. Looking at the peak neural response avoids this problem, because peak response would be expected when the stimulus is at the centre of the RF, irrespective of RF size.

To establish the latency with which the position of a moving object is represented at different stages of visual processing, we identified the timepoint at which our classification analysis yielded maximum stimulus-position likelihood. This was achieved by fitting a Gaussian curve to the observed time-course of the calculated likelihood averaged across participants, separately for each training time (Figure 3, step 5). Adjusted R² of these fits can be found in Figure 3, step 6. For training timepoints later than ∼80 ms, the Gaussian curves provided a very good fit to the evolution of stimulus-position likelihood over time, with R² values over 0.5. Although the window of significant cross-classification of static stimuli to moving stimuli is restricted (Figure 3, step 4), the sustained high adjusted R² values indicate that even for training times at which the stimulus-position likelihood was close to chance level, the likelihood increased and decreased as the stimulus traversed each flash location.

Figure 4a shows the time to peak likelihood for motion across all training timepoints at which adjusted R² exceeded a minimum value of 0.5. The choice of R² cutoff is essentially arbitrary, but the pattern of points in Figure 3, step 6 shows that this selection is relatively robust to changes in the cutoff value. Up to ∼150 ms training time, the time to peak likelihood increases with increasing training time. This follows the same pattern as the static trials (see Figure 4b), where earlier representations of the stimulus (i.e. early training times) were activated at a shorter latency in the testing epoch than later representations of the stimulus. This sequential pattern is consistent with the first feedforward sweep of stimulus-driven activation. As information flows through the visual processing hierarchy, representations of the stimulus will gradually change over time. The order of these changes appeared to be consistent between static and motion trials.

Figure 4

Download asset Open asset

This pattern subsequently reverses between 150 and 200 ms, indicating that hierarchically later representations were activated at a shorter latency. Finally, from a training time of ∼250 ms the time to peak likelihood was stable at approximately 50 ms. The non-monotonic relationship between training time and time to peak likelihood could emerge because there was variable compensation for neural delays at different training times. Perhaps more likely, this pattern could reflect feedforward and feedback sweeps of activity in the visual cortex evoked by the static stimulus: the feedforward sweep activates sequential representations, and information flowing backward along the hierarchy reactivates the same activity patterns in reverse order. The timescale of this wave of activity was in line with previous findings from Dijkstra et al., 2020, who showed approximately 10 Hz oscillations evoked by face/house stimuli. Additionally, previous TMS, MEG, and fMRI results suggest that 150 ms is a reasonable estimate for the time it takes for visual information to reach hierarchically later visual processing areas, such as V5/MT+ (Sack et al., 2006; Ahlfors et al., 1999; Hotson et al., 1994; Yoshor et al., 2007; Mohsenzadeh et al., 2018; Lamme and Roelfsema, 2000). This would imply that the activity evoked by static compared to moving stimuli is different after about 150 ms, due to a divergence in further stimulus processing. Therefore, cross-generalisation of position decoding across stimulus types is possible after this timepoint because a portion of the flash-evoked signal reflects reactivation of earlier activity patterns, which are common to static and moving stimulus processing. If this later activity (>150 ms training time) does indeed reflect feedback processing of the static stimulus, then, for these later timepoints, the latency measure we have calculated might not be informative about the time necessary to first represent the moving object, because the initial activations and the reactivations are indistinguishable.

In order to confirm which timepoints predominantly correspond to feedforward processes rather than feedback, we applied piecewise regression, as implemented in the Shape Language Modeling toolbox (D’Errico, 2022). In this approach, several polynomials are smoothly joined together at ‘knots’. Placement of knots, at the start and end of each segment, is optimised by reducing root mean squared error. We fit straight lines, and varied the number of knots between four and seven, in order to identify the optimal number. The best piecewise regression fit, with six knots, is shown in Figure 4a. Corroborating our observation, the first internal knot was placed at 158 ms. We took this inflection point as the end of the initial feedforward sweep of information through the visual cortex. The piecewise regression revealed further positive and negative slopes, suggesting that feedforward and feedback sweeps of activity continue during later stimulus processing (Dijkstra et al., 2020).

In this study we investigated how the visual system compensates for neural transmission delays when tracking the positions of moving objects. We investigated the latency of neural position representations of moving stimuli compared to unpredictably presented static stimuli, as well as the real-time position of the stimulus. By computing the timepoint at which each position representation of the moving stimulus was most similar to a static stimulus in the same location, we tracked the represented position of the stimulus over time across the visual hierarchy.

We demonstrate that classifiers trained to locate static stimuli based on the stimulus-evoked EEG signal could also localise moving stimuli. This is the first study, to our knowledge, to demonstrate cross-classification between stationary and smoothly moving stimuli with EEG, showing that population neural codes for location are (at least partially) shared between the two stimulus types. We, therefore, were able to investigate the timing of neural responses to motion in humans with a fine temporal resolution. We subsequently showed that, during the first feedforward sweep of activity, the neural response encoding the position of a moving stimulus was shifted approximately 75 ms earlier than the response to a static stimulus. The early decoded representations of the position of a moving stimulus aligned with the real-time position of the object, rather than the position corresponding to afferent retinal input (subject to transmission and processing delays), which would instead signal outdated position information in visual cortex. Finally, we showed that delay compensation was primarily achieved before information reached visual cortex, as the recorded cortical processing of static and motion stimuli followed a similar time-course. Overall, this study shows direct neural evidence of motion extrapolation enabling accurate real-time representation of moving objects in humans.

These results are consistent with findings of RF shifts across the visual cortex in response to motion. Many earlier fMRI studies showing RF shifts against the direction of motion Liu et al., 2006; Whitney et al., 2003; Raemaekers et al., 2009 have been dismissed because of the ‘aperture-inward’ bias, in which the trailing edge of a motion stimulus evokes larger responses than the leading edge (Wang et al., 2014). This is not an issue for the present study, as we can determine the timing of neural responses at a fine temporal scale, rather than looking at aggregate responses over whole motion trajectories. Neural recordings from animals and more recent fMRI studies in humans have reliably shown RF shifts throughout the visual cortex in response to motion (Harvey and Dumoulin, 2016), and that these displacements are against the direction of motion (Schneider et al., 2019; Sundberg et al., 2006; Fu et al., 2004). However, several differences remain between the previous fMRI results and the present study. Harvey and Dumoulin, 2016, found that RF shifts in response to motion scale with the size of the RF across the visual hierarchy. This implies that visual areas higher up the processing hierarchy that are activated later in time would shift their RFs more than lower visual areas which are activated earlier. In contrast, our results suggest that later visual areas show RF shifts of the same magnitude as earlier visual areas. However, it is not clear whether it is possible to map time elapsing after stimulus onset in EEG to processing in different visual areas as recorded by fMRI. The longer timescale of the fMRI signal means that it could be indexing later activity than we are recording with EEG, or include signals that emerge after integration of many feedforward and feedback sweeps of activity. While further research is needed to understand how extrapolation operates at different spatial scales, this converging evidence of RF shifts against the direction of motion suggests that the positions of moving objects are predictively encoded during processing. Furthermore, we provide evidence that RF shifts occur during the initial feedforward sweep of the visual response.

All data files have been deposited on Dryad at https://doi.org/10.5061/dryad.vx0k6djw0. Code is available on the Open Science Framework at https://osf.io/jbw9m/.

1. 1. Johnson PA
   2. Blom T
   3. van Gaal S
   4. Feuerriegel D
   5. Bode S
   6. Hogendoorn H

   (2022) Dryad Digital Repository

   EEG and eyetracking response to static and moving stimuli.

   https://doi.org/10.5061/dryad.vx0k6djw0
2. 1. Johnson P

   (2020) Open Science Framework

   ID jbw9m. Position representations of moving objects align with real-time position in the early visual response.

   https://osf.io/jbw9m/

1. 1. Ahlfors SP
   2. Simpson GV
   3. Dale AM
   4. Belliveau JW
   5. Liu AK
   6. Korvenoja A
   7. Virtanen J
   8. Huotilainen M
   9. Tootell RB
   10. Aronen HJ
   11. Ilmoniemi RJ

   (1999) Spatiotemporal activity of a cortical network for processing visual motion revealed by MEG and fmri

   Journal of Neurophysiology 82:2545–2555.

   https://doi.org/10.1152/jn.1999.82.5.2545

   • PubMed
   • Google Scholar
2. 1. Alilović J
   2. Timmermans B
   3. Reteig LC
   4. van Gaal S
   5. Slagter HA

   (2019) No evidence that predictions and attention modulate the first feedforward sweep of cortical information processing

   Cerebral Cortex 29:2261–2278.

   https://doi.org/10.1093/cercor/bhz038

   • PubMed
   • Google Scholar
3. Preprint

   1. Benvenuti G
   2. Chemla S
   3. Boonman A
   4. Perrinet L
   5. Masson GS
   6. Chavane F

   (2020) Anticipatory Responses along Motion Trajectories in Awake Monkey Area V1

   bioRxiv.

   https://doi.org/10.1101/2020.03.26.010017

   • Google Scholar
4. 1. Berens P

   (2009) CircStat: A MATLAB toolbox for circular statistics

   Journal of Statistical Software 31:1–21.

   https://doi.org/10.18637/jss.v031.i10

   • Google Scholar
5. 1. Berry MJ
   2. Brivanlou IH
   3. Jordan TA
   4. Meister M

   (1999) Anticipation of moving stimuli by the retina

   Nature 398:334–338.

   https://doi.org/10.1038/18678

   • PubMed
   • Google Scholar
6. Conference

   1. Bigdely-Shamlo N
   2. Kreutz-Delgado K
   3. Kothe C
   4. Makeig S

   (2013) EyeCatch: data-mining over half a million EEG independent components to construct a fully-automated eye-component detector

   Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference. pp. 5845–5848.

   https://doi.org/10.1109/EMBC.2013.6610881

   • Google Scholar
7. 1. Blom T
   2. Feuerriegel D
   3. Johnson P
   4. Bode S
   5. Hogendoorn H

   (2020) Predictions drive neural representations of visual events ahead of incoming sensory information

   PNAS 117:7510–7515.

   https://doi.org/10.1073/pnas.1917777117

   • PubMed
   • Google Scholar
8. 1. Blom T
   2. Bode S
   3. Hogendoorn H

   (2021) The time-course of prediction formation and revision in human visual motion processing

   Cortex; a Journal Devoted to the Study of the Nervous System and Behavior 138:191–202.

   https://doi.org/10.1016/j.cortex.2021.02.008

   • PubMed
   • Google Scholar
9. 1. Bode S
   2. Feuerriegel D
   3. Bennett D
   4. Alday PM

   (2019) The decision decoding toolbox (DDTBOX) - a multivariate pattern analysis toolbox for event-related potentials

   Neuroinformatics 17:27–42.

   https://doi.org/10.1007/s12021-018-9375-z

   • PubMed
   • Google Scholar
10. 1. Brainard DH

    (1997) The psychophysics toolbox

    Spatial Vision 10:433–436.

    https://doi.org/10.1163/156856897X00357

    • PubMed
    • Google Scholar
11. 1. Bullmore ET
    2. Suckling J
    3. Overmeyer S
    4. Rabe-Hesketh S
    5. Taylor E
    6. Brammer MJ

    (1999) Global, voxel, and cluster tests, by theory and permutation, for a difference between two groups of structural MR images of the brain

    IEEE Transactions on Medical Imaging 18:32–42.

    https://doi.org/10.1109/42.750253

    • PubMed
    • Google Scholar
12. 1. Chemla S
    2. Reynaud A
    3. di Volo M
    4. Zerlaut Y
    5. Perrinet L
    6. Destexhe A
    7. Chavane F

    (2019) Suppressive traveling waves shape representations of illusory motion in primary visual cortex of awake primate

    The Journal of Neuroscience 39:4282–4298.

    https://doi.org/10.1523/JNEUROSCI.2792-18.2019

    • PubMed
    • Google Scholar
13. 1. Chen EY
    2. Marre O
    3. Fisher C
    4. Schwartz G
    5. Levy J
    6. da Silveira RA
    7. da Silviera RA
    8. Berry MJ

    (2013) Alert response to motion onset in the retina

    The Journal of Neuroscience 33:120–132.

    https://doi.org/10.1523/JNEUROSCI.3749-12.2013

    • PubMed
    • Google Scholar
14. 1. Clark A

    (2013) Whatever next? predictive brains, situated agents, and the future of cognitive science

    The Behavioral and Brain Sciences 36:181–204.

    https://doi.org/10.1017/S0140525X12000477

    • PubMed
    • Google Scholar
15. 1. Delorme A
    2. Makeig S

    (2004) EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis

    Journal of Neuroscience Methods 134:9–21.

    https://doi.org/10.1016/j.jneumeth.2003.10.009

    • PubMed
    • Google Scholar
16. Website

    1. D’Errico J

    (2022) SLM -shape language modeling

    MATLAB Central File Exchange. Accessed March 11, 2022.

    https://www.mathworks.com/matlabcentral/fileexchange/24443-slm-shape-language-modeling
17. 1. Dijkstra N
    2. Ambrogioni L
    3. Vidaurre D
    4. van Gerven M

    (2020) Neural dynamics of perceptual inference and its reversal during imagery

    eLife 9:e53588.

    https://doi.org/10.7554/eLife.53588

    • PubMed
    • Google Scholar
18. 1. Di Russo F
    2. Martínez A
    3. Sereno MI
    4. Pitzalis S
    5. Hillyard SA

    (2002) Cortical sources of the early components of the visual evoked potential

    Human Brain Mapping 15:95–111.

    https://doi.org/10.1002/hbm.10010

    • PubMed
    • Google Scholar
19. 1. Fahrenfort JJ
    2. Scholte HS
    3. Lamme VAF

    (2007) Masking disrupts reentrant processing in human visual cortex

    Journal of Cognitive Neuroscience 19:1488–1497.

    https://doi.org/10.1162/jocn.2007.19.9.1488

    • PubMed
    • Google Scholar
20. 1. ffytche DH
    2. Guy CN
    3. Zeki S

    (1995) The parallel visual motion inputs into areas V1 and V5 of human cerebral cortex

    Brain 118 (Pt 6):1375–1394.

    https://doi.org/10.1093/brain/118.6.1375

    • PubMed
    • Google Scholar
21. 1. Friston K

    (2010) The free-energy principle: a unified brain theory?

    Nature Reviews. Neuroscience 11:127–138.

    https://doi.org/10.1038/nrn2787

    • PubMed
    • Google Scholar
22. 1. Friston KJ
    2. Daunizeau J
    3. Kilner J
    4. Kiebel SJ

    (2010) Action and behavior: a free-energy formulation

    Biol Cybern 102:227–260.

    https://doi.org/10.1007/s00422-010-0364-z

    • PubMed
    • Google Scholar
23. 1. Fu YX
    2. Shen Y
    3. Gao H
    4. Dan Y

    (2004) Asymmetry in visual cortical circuits underlying motion-induced perceptual mislocalization

    The Journal of Neuroscience 24:2165–2171.

    https://doi.org/10.1523/JNEUROSCI.5145-03.2004

    • PubMed
    • Google Scholar
24. 1. Guo K
    2. Robertson RG
    3. Pulgarin M
    4. Nevado A
    5. Panzeri S
    6. Thiele A
    7. Young MP

    (2007) Spatio-temporal prediction and inference by V1 neurons

    The European Journal of Neuroscience 26:1045–1054.

    https://doi.org/10.1111/j.1460-9568.2007.05712.x

    • PubMed
    • Google Scholar
25. 1. Harrison LM
    2. Stephan KE
    3. Rees G
    4. Friston KJ

    (2007) Extra-classical receptive field effects measured in striate cortex with fmri

    NeuroImage 34:1199–1208.

    https://doi.org/10.1016/j.neuroimage.2006.10.017

    • PubMed
    • Google Scholar
26. 1. Harvey BM
    2. Dumoulin SO

    (2011) The relationship between cortical magnification factor and population receptive field size in human visual cortex: constancies in cortical architecture

    The Journal of Neuroscience 31:13604–13612.

    https://doi.org/10.1523/JNEUROSCI.2572-11.2011

    • PubMed
    • Google Scholar
27. 1. Harvey BM
    2. Dumoulin SO

    (2016) Visual motion transforms visual space representations similarly throughout the human visual hierarchy

    NeuroImage 127:173–185.

    https://doi.org/10.1016/j.neuroimage.2015.11.070

    • PubMed
    • Google Scholar
28. 1. Haufe S
    2. Meinecke F
    3. Görgen K
    4. Dähne S
    5. Haynes JD
    6. Blankertz B
    7. Bießmann F

    (2014) On the interpretation of weight vectors of linear models in multivariate neuroimaging

    NeuroImage 87:96–110.

    https://doi.org/10.1016/j.neuroimage.2013.10.067

    • PubMed
    • Google Scholar
29. 1. Hogendoorn H
    2. Burkitt AN

    (2018) Predictive coding of visual object position ahead of moving objects revealed by time-resolved EEG decoding

    NeuroImage 171:55–61.

    https://doi.org/10.1016/j.neuroimage.2017.12.063

    • PubMed
    • Google Scholar
30. 1. Hogendoorn H
    2. Burkitt AN

    (2019) Predictive coding with neural transmission delays: a real-time temporal alignment hypothesis

    ENeuro 6:ENEURO.0412-18.2019.

    https://doi.org/10.1523/ENEURO.0412-18.2019

    • PubMed
    • Google Scholar
31. 1. Hogendoorn H

    (2020) Motion extrapolation in visual processing: lessons from 25 years of flash-lag debate

    The Journal of Neuroscience 40:5698–5705.

    https://doi.org/10.1523/JNEUROSCI.0275-20.2020

    • PubMed
    • Google Scholar
32. 1. Hotson J
    2. Braun D
    3. Herzberg W
    4. Boman D

    (1994) Transcranial magnetic stimulation of extrastriate cortex degrades human motion direction discrimination

    Vision Research 34:2115–2123.

    https://doi.org/10.1016/0042-6989(94)90321-2

    • PubMed
    • Google Scholar
33. 1. Jancke D
    2. Erlhagen W
    3. Schöner G
    4. Dinse HR

    (2004) Shorter latencies for motion trajectories than for flashes in population responses of cat primary visual cortex

    The Journal of Physiology 556:971–982.

    https://doi.org/10.1113/jphysiol.2003.058941

    • PubMed
    • Google Scholar
34. 1. Jatoi MA
    2. Kamel N
    3. Malik AS
    4. Faye I
    5. Begum T

    (2014) A survey of methods used for source localization using EEG signals

    Biomedical Signal Processing and Control 11:42–52.

    https://doi.org/10.1016/j.bspc.2014.01.009

    • Google Scholar
35. Conference

    1. Johnson P
    2. Grootswagers T
    3. Moran C
    4. Hogendoorn H

    (2021) Temporal dynamics of visual population receptive fields

    In: 43rd European Conference on Visual Perception (ECVP) 2021 Online, vol. pp. 1–244.

    https://doi.org/10.1177/03010066211059887

    • Google Scholar
36. Data

    1. Johnson PA
    2. Blom T
    3. Van Gaal S
    4. Feuerriegel D
    5. Bode S
    6. Hogendoorn H

    (authors) (2022) EEG and eyetracking response to static and moving stimuli

    Dryad.

    https://doi.org/10.5061/dryad.vx0k6djw0
37. 1. Johnston J
    2. Lagnado L

    (2015) General features of the retinal connectome determine the computation of motion anticipation

    eLife 4:e06250.

    https://doi.org/10.7554/eLife.06250

    • PubMed
    • Google Scholar
38. 1. Kiebel SJ
    2. Daunizeau J
    3. Friston KJ

    (2008) A hierarchy of time-scales and the brain

    PLOS Computational Biology 4:e1000209.

    https://doi.org/10.1371/journal.pcbi.1000209

    • PubMed
    • Google Scholar
39. 1. King JR
    2. Dehaene S

    (2014) Characterizing the dynamics of mental representations: the temporal generalization method

    Trends in Cognitive Sciences 18:203–210.

    https://doi.org/10.1016/j.tics.2014.01.002

    • PubMed
    • Google Scholar
40. 1. King J-R
    2. Wyart V

    (2021) The human brain encodes a chronicle of visual events at each instant of time through the multiplexing of traveling waves

    The Journal of Neuroscience 41:7224–7233.

    https://doi.org/10.1523/JNEUROSCI.2098-20.2021

    • PubMed
    • Google Scholar
41. 1. Kleiner M
    2. Brainard D
    3. Pelli D

    (2007)

    What’s new in psychtoolbox-3

    Perception 36:1–16.

    • Google Scholar
42. Preprint

    1. Kovesi P

    (2015) Good Colour Maps: How to Design Them

    arXiv.

    https://arxiv.org/abs/1509.03700

    • Google Scholar
43. 1. Krekelberg B
    2. Lappe M
    3. Whitney D
    4. Cavanagh P
    5. Eagleman DM
    6. Sejnowski TJ

    (2000) The position of moving objects

    Science 289:1107a.

    https://doi.org/10.1126/science.289.5482.1107a

    • PubMed
    • Google Scholar
44. 1. Lamme V
    2. Roelfsema PR

    (2000) The distinct modes of vision offered by feedforward and recurrent processing

    Trends in Neurosciences 23:571–579.

    https://doi.org/10.1016/s0166-2236(00)01657-x

    • PubMed
    • Google Scholar
45. 1. Liu JV
    2. Ashida H
    3. Smith AT
    4. Wandell BA

    (2006) Assessment of stimulus-induced changes in human V1 visual field maps

    Journal of Neurophysiology 96:3398–3408.

    https://doi.org/10.1152/jn.00556.2006

    • PubMed
    • Google Scholar
46. 1. Liu B
    2. Hong A
    3. Rieke F
    4. Manookin MB

    (2021) Predictive encoding of motion begins in the primate retina

    Nature Neuroscience 24:1280–1291.

    https://doi.org/10.1038/s41593-021-00899-1

    • PubMed
    • Google Scholar
47. 1. Manyakov NV
    2. Van Hulle MM

    (2010) Decoding grating orientation from microelectrode array recordings in monkey cortical area V4

    International Journal of Neural Systems 20:95–108.

    https://doi.org/10.1142/S0129065710002280

    • PubMed
    • Google Scholar
48. 1. Maris E
    2. Oostenveld R

    (2007) Nonparametric statistical testing of EEG- and MEG-data

    Journal of Neuroscience Methods 164:177–190.

    https://doi.org/10.1016/j.jneumeth.2007.03.024

    • PubMed
    • Google Scholar
49. 1. Maus GW
    2. Fischer J
    3. Whitney D

    (2013a) Motion-dependent representation of space in area MT+

    Neuron 78:554–562.

    https://doi.org/10.1016/j.neuron.2013.03.010

    • PubMed
    • Google Scholar
50. 1. Maus GW
    2. Ward J
    3. Nijhawan R
    4. Whitney D

    (2013b) The perceived position of moving objects: transcranial magnetic stimulation of area MT+ reduces the flash-lag effect

    Cerebral Cortex 23:241–247.

    https://doi.org/10.1093/cercor/bhs021

    • PubMed
    • Google Scholar
51. 1. McKeefry DJ
    2. Watson JDG
    3. Frackowiak RSJ
    4. Fong K
    5. Zeki S

    (1997) The activity in human areas V1/V2, V3, and V5 during the perception of coherent and incoherent motion

    NeuroImage 5:1–12.

    https://doi.org/10.1006/nimg.1996.0246

    • PubMed
    • Google Scholar
52. Preprint

    1. Millidge B
    2. Seth A
    3. Buckley CL

    (2022) Predictive Coding: A Theoretical and Experimental Review

    arXiv.

    https://arxiv.org/abs/2107.12979

    • Google Scholar
53. 1. Mohsenzadeh Y
    2. Qin S
    3. Cichy RM
    4. Pantazis D

    (2018) Ultra-rapid serial visual presentation reveals dynamics of feedforward and feedback processes in the ventral visual pathway

    eLife 7:e36329.

    https://doi.org/10.7554/eLife.36329

    • PubMed
    • Google Scholar
54. 1. Mostert P
    2. Kok P
    3. de Lange FP

    (2015) Dissociating sensory from decision processes in human perceptual decision making

    Scientific Reports 5:18253.

    https://doi.org/10.1038/srep18253

    • PubMed
    • Google Scholar
55. 1. Mostert P
    2. Albers AM
    3. Brinkman L
    4. Todorova L
    5. Kok P
    6. de Lange FP

    (2018) Eye movement-related confounds in neural decoding of visual working memory representations

    ENeuro 5:ENEURO.0401-17.2018.

    https://doi.org/10.1523/ENEURO.0401-17.2018

    • PubMed
    • Google Scholar
56. 1. Nijhawan R

    (1994) Motion extrapolation in Catching

    Nature 370:256–257.

    https://doi.org/10.1038/370256b0

    • PubMed
    • Google Scholar
57. 1. Nijhawan R
    2. Kirschfeld K

    (2003) Analogous mechanisms compensate for neural delays in the sensory and the motor pathways: evidence from motor flash-lag

    Current Biology 13:749–753.

    https://doi.org/10.1016/s0960-9822(03)00248-3

    • PubMed
    • Google Scholar
58. 1. Noda H
    2. Freeman RB
    3. Gies B
    4. Creutzfeldt OD

    (1971) Neuronal responses in the visual cortex of awake cats to stationary and moving targets

    Experimental Brain Research 12:389–405.

    https://doi.org/10.1007/BF00234494

    • Google Scholar
59. 1. Orban GA
    2. Hoffmann KP
    3. Duysens J

    (1985) Velocity selectivity in the cat visual system. I. responses of LGN cells to moving bar stimuli: a comparison with cortical areas 17 and 18

    Journal of Neurophysiology 54:1026–1049.

    https://doi.org/10.1152/jn.1985.54.4.1026

    • PubMed
    • Google Scholar
60. 1. Pelli DG

    (1997)

    The videotoolbox software for visual psychophysics: transforming numbers into movies

    Spatial Vision 10:437–442.

    • PubMed
    • Google Scholar
61. 1. Pernet CR
    2. Chauveau N
    3. Gaspar C
    4. Rousselet GA

    (2011) LIMO EEG: a toolbox for hierarchical linear modeling of electroencephalographic data

    Computational Intelligence and Neuroscience 2011:831409.

    https://doi.org/10.1155/2011/831409

    • PubMed
    • Google Scholar
62. Conference

    1. Price D
    2. Knerr S
    3. Personnaz L
    4. Dreyfus G

    (1995)

    Pairwise neural network classifiers with probabilistic outputs

    Neural Information Processing Systems. pp. 1109–1116.

    • Google Scholar
63. 1. Raemaekers M
    2. Lankheet MJM
    3. Moorman S
    4. Kourtzi Z
    5. van Wezel RJA

    (2009) Directional anisotropy of motion responses in retinotopic cortex

    Human Brain Mapping 30:3970–3980.

    https://doi.org/10.1002/hbm.20822

    • PubMed
    • Google Scholar
64. 1. Rousselet GA
    2. Husk JS
    3. Bennett PJ
    4. Sekuler AB

    (2005) Spatial scaling factors explain eccentricity effects on face erps

    Journal of Vision 5:755–763.

    https://doi.org/10.1167/5.10.1

    • PubMed
    • Google Scholar
65. 1. Sack AT
    2. Kohler A
    3. Linden DEJ
    4. Goebel R
    5. Muckli L

    (2006) The temporal characteristics of motion processing in hmt/V5+: combining fmri and neuronavigated TMS

    NeuroImage 29:1326–1335.

    https://doi.org/10.1016/j.neuroimage.2005.08.027

    • PubMed
    • Google Scholar
66. 1. Salti M
    2. Monto S
    3. Charles L
    4. King JR
    5. Parkkonen L
    6. Dehaene S

    (2015) Distinct cortical codes and temporal dynamics for conscious and unconscious percepts

    eLife 4:e05652.

    https://doi.org/10.7554/eLife.05652

    • PubMed
    • Google Scholar
67. 1. Schellekens W
    2. van Wezel RJA
    3. Petridou N
    4. Ramsey NF
    5. Raemaekers M

    (2016) Predictive coding for motion stimuli in human early visual cortex

    Brain Structure & Function 221:879–890.

    https://doi.org/10.1007/s00429-014-0942-2

    • PubMed
    • Google Scholar
68. 1. Schellekens W
    2. Ramsey NF
    3. van Wezel RJA
    4. Raemaekers M

    (2017) Changes in fMRI BOLD dynamics reflect anticipation to moving objects

    NeuroImage 161:188–195.

    https://doi.org/10.1016/j.neuroimage.2016.09.017

    • PubMed
    • Google Scholar
69. Preprint

    1. Schneider M
    2. Marquardt I
    3. Sengupta S
    4. De Martino F
    5. Goebel R
    6. Martino FD
    7. Goebel R

    (2019) Motion Displaces Population Receptive Fields in the Direction Opposite to Motion

    bioRxiv.

    https://doi.org/10.1101/759183

    • Google Scholar
70. 1. Souihel S
    2. Cessac B

    (2021) On the potential role of lateral connectivity in retinal anticipation

    Journal of Mathematical Neuroscience 11:3.

    https://doi.org/10.1186/s13408-020-00101-z

    • PubMed
    • Google Scholar
71. 1. Subramaniyan M
    2. Ecker AS
    3. Patel SS
    4. Cotton RJ
    5. Bethge M
    6. Pitkow X
    7. Berens P
    8. Tolias AS

    (2018) Faster processing of moving compared with flashed bars in awake macaque V1 provides a neural correlate of the flash lag illusion

    Journal of Neurophysiology 120:2430–2452.

    https://doi.org/10.1152/jn.00792.2017

    • PubMed
    • Google Scholar
72. 1. Sundberg KA
    2. Fallah M
    3. Reynolds JH

    (2006) A motion-dependent distortion of retinotopy in area V4

    Neuron 49:447–457.

    https://doi.org/10.1016/j.neuron.2005.12.023

    • PubMed
    • Google Scholar
73. 1. Thaler L
    2. Schütz AC
    3. Goodale MA
    4. Gegenfurtner KR

    (2013) What is the best fixation target? the effect of target shape on stability of fixational eye movements

    Vision Research 76:31–42.

    https://doi.org/10.1016/j.visres.2012.10.012

    • PubMed
    • Google Scholar
74. 1. Tse PU
    2. Sheinberg DL
    3. Logothetis NK

    (2002) Fixational eye movements are not affected by abrupt onsets that capture attention

    Vision Research 42:1663–1669.

    https://doi.org/10.1016/s0042-6989(02)00076-7

    • PubMed
    • Google Scholar
75. Preprint

    1. van Driel J
    2. Olivers CNL
    3. Fahrenfort JJ

    (2019) High-Pass Filtering Artifacts in Multivariate Classification of Neural Time Series Data

    bioRxiv.

    https://doi.org/10.1101/530220

    • Google Scholar
76. 1. van Heusden E
    2. Harris AM
    3. Garrido MI
    4. Hogendoorn H

    (2019) Predictive coding of visual motion in both monocular and binocular human visual processing

    Journal of Vision 19:3.

    https://doi.org/10.1167/19.1.3

    • PubMed
    • Google Scholar
77. 1. Vanni S
    2. Dojat M
    3. Warnking J
    4. Delon-Martin C
    5. Segebarth C
    6. Bullier J

    (2004) Timing of interactions across the visual field in the human cortex

    NeuroImage 21:818–828.

    https://doi.org/10.1016/j.neuroimage.2003.10.035

    • PubMed
    • Google Scholar
78. 1. Vidaurre D
    2. Myers NE
    3. Stokes M
    4. Nobre AC
    5. Woolrich MW

    (2019) Temporally unconstrained decoding reveals consistent but time-varying stages of stimulus processing

    Cerebral Cortex 29:863–874.

    https://doi.org/10.1093/cercor/bhy290

    • PubMed
    • Google Scholar
79. 1. Wang HX
    2. Merriam EP
    3. Freeman J
    4. Heeger DJ

    (2014) Motion direction biases and decoding in human visual cortex

    The Journal of Neuroscience 34:12601–12615.

    https://doi.org/10.1523/JNEUROSCI.1034-14.2014

    • PubMed
    • Google Scholar
80. 1. Whitney D
    2. Murakami I

    (1998) Latency difference, not spatial extrapolation

    Nature Neuroscience 1:39–50.

    https://doi.org/10.1038/3659

    • Google Scholar
81. 1. Whitney D
    2. Cavanagh P

    (2000) Motion distorts visual space: shifting the perceived position of remote stationary objects

    Nature Neuroscience 3:954–959.

    https://doi.org/10.1038/78878

    • PubMed
    • Google Scholar
82. 1. Whitney D
    2. Murakami I
    3. Cavanagh P

    (2000) Illusory spatial offset of a flash relative to a moving stimulus is caused by differential latencies for moving and flashed stimuli

    Vision Research 40:137–149.

    https://doi.org/10.1016/s0042-6989(99)00166-2

    • PubMed
    • Google Scholar
83. 1. Whitney D
    2. Goltz HC
    3. Thomas CG
    4. Gati JS
    5. Menon RS
    6. Goodale MA

    (2003) Flexible retinotopy: motion-dependent position coding in the visual cortex

    Science 302:878–881.

    https://doi.org/10.1126/science.1087839

    • PubMed
    • Google Scholar
84. 1. Wibral M
    2. Bledowski C
    3. Kohler A
    4. Singer W
    5. Muckli L

    (2009) The timing of feedback to early visual cortex in the perception of long-range apparent motion

    Cerebral Cortex 19:1567–1582.

    https://doi.org/10.1093/cercor/bhn192

    • PubMed
    • Google Scholar
85. 1. Wojtach WT
    2. Sung K
    3. Truong S
    4. Purves D

    (2008) An empirical explanation of the flash-lag effect

    PNAS 105:16338–16343.

    https://doi.org/10.1073/pnas.0808916105

    • PubMed
    • Google Scholar
86. 1. Yan C
    2. Chang X
    3. Luo M
    4. Zheng Q
    5. Zhang X
    6. Li Z
    7. Nie F

    (2021) Self-weighted robust LDA for multiclass classification with edge classes

    ACM Transactions on Intelligent Systems and Technology 12:1–19.

    https://doi.org/10.1145/3418284

    • Google Scholar
87. 1. Yoshor D
    2. Bosking WH
    3. Ghose GM
    4. Maunsell J

    (2007) Receptive fields in human visual cortex mapped with surface electrodes

    Cerebral Cortex 17:2293–2302.

    https://doi.org/10.1093/cercor/bhl138

    • PubMed
    • Google Scholar
88. 1. Yuen KK

    (1974) The two-sample trimmed T for unequal population variances

    Biometrika 61:165–170.

    https://doi.org/10.1093/biomet/61.1.165

    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Position representations of moving objects align with real-time position in the early visual response" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Clare Press as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Tirin Moore as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Samson Chota (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Summary of Revisions:

Overall assessment:

The reviewers find the patterns in the data interesting and believe this view would be shared by other visual neuroscientists as well as neuroscientists from other fields – where neural delays are likely to prove a universal sticking point to many theories. One of its major strengths lies in the successful attempt to generalize neural representations of static objects to moving objects. They also found the manuscript to be well-written – well-tuned, clear, and interesting, with well-presented figures.

Introduction:

1. A little more work could be done to sell this to readers who do not study motion perception in vision. Compensating for neural processing delays will be a problem that most neuroscientific theories need to incorporate; hence why we believe this is especially suitable for publication in eLife. We would suggest adding some literature considering an application to other sensory (e..g, auditory) processing, as well as potentially higher-level cognitive domains.

Results:

2. We found the Results section to be well-written, like the rest of the manuscript, but due to the complexity of the analysis we think the reader could be given a little more help. E.g., in the final paragraph of the Results section can some clear flags be inserted to show which aspects of the regression result support which of the conclusions? We think this would help more non-specialised readers grasp the reasons for the conclusion.

3. We would recommend the authors include an illustration of the compensation achieved at different cortical processing stages based on their findings. Perhaps it also makes sense to include approximate neural delays at different stages for a direct comparison. In addition, it might also help to illustrate how this compensation is presumably achieved in space by shifting encoding locations. We believe that these concepts can sometimes be tricky to communicate and the paper would benefit from those visual aids.

4. Have the authors tested for potential differences in how well static positions and movements can be decoded at different locations in the visual field? We can imagine that many readers would be curious if (at least static) decoding is uniform across the visual field. If certain positions/movements are decoded better then the authors could also consider the following analysis: Was there a group of subjects (or selection of best stimulus locations for decoding) for which the early static representations (<100 ms) reliably generalized to motion? This might help to distinguish between poor signal-to-noise and entirely different codes (or direct connections from LGN/pulvinar) between early processing of static and motion stimuli.

Discussion:

5. The findings reported are derived from a specific case of motion perception which may not reflect the general mechanisms optimized for motion perception. The limitations related to task designs and the neural readouts should be discussed as they affect the way that the reported results will be interpreted.

6. This study is clearly different from the authors' earlier work, e.g., Blom et al. However, we think it would be useful to add some discussion to the Discussion section to outline what these Blom et al. studies showed and clearly, therefore, flag the novelty of this study/analysis.

7. The approach of using different epochs to train to approximate different stages of neural processing is clever. However, it seems that later epochs should reflect processing in later hierarchical points while still also reflecting processing in early points in the hierarchy. The authors should discuss the implications of this for their conclusions.

8. Could there be an impact on these findings of a large neural response when a stimulus enters the edge of a neuron's RF, followed by some adaptation? This impact should be discussed.

9. The findings displayed in Figure 4 (reversal of time to peak likelihood) are very intriguing. We believe the manuscript would benefit from discussing these particular results in a bit more detail in the discussion, especially in the light of feedback signals during motion processing. If the observed 'reversed' patterns indeed reflect feedback then by the time the feedback arrives back in the early visual cortex the original stimulus will have moved on and might be replaced by activity from subsequent (mismatching) bottom-up stimulus processing. In addition, referring to Hogendoorn and Burkitt (2019) the authors make a point that delayed compensation should occur in feedback activity, parts of which they do seem to capture with their method (Figure 4). Can the authors elaborate a bit on the potential functional relevance of the observed feedback signals and how they might differ from motion extrapolation-specific feedback?

Reviewer #1 (Recommendations for the authors):

1. I think a little more work could be done to sell this to readers who do not study motion perception in vision. Compensating for neural processing delays will be a problem that most neuroscientific theories need to incorporate; hence why I believe this is especially suitable for publication in eLife. I would suggest adding some literature considering an application to other sensory (e.g, auditory) processing, as well as potentially higher-level cognitive domains.

2. This study is clearly different from the authors' earlier work, e.g., Blom et al. However, I think it would be useful to add some discussion to the Discussion section to outline what these Blom et al. studies showed and clearly, therefore, flag the novelty of this study/analysis.

3. I found the Results section to be well-written, like the rest of the manuscript, but due to the complexity of analysis, I think the reader could be given a little more help. E.g., in the final paragraph of the Results section can some clear flags be inserted to show which aspects of the regression result support which of the conclusions? I think this would help more non-specialised readers grasp the reasons for the conclusion.

4. Could there be an impact on these findings of a large neural response when a stimulus enters the edge of a neuron's RF, followed by some adaptation?

5. The approach of using different epochs to train to approximate different stages of neural processing is clever. However, it seems that later epochs should reflect processing in later hierarchical points while still also reflecting processing in early points. The authors should discuss the implications of this for their conclusions.

Reviewer #2 (Recommendations for the authors):

I honestly have very few remarks on this manuscript all of which concern minor details on form and interpretation. It's fantastic to work, and I am happy to recommend this paper for publication in eLife.

Remarks

1. I would recommend the authors include an illustration of the compensation achieved at different cortical processing stages based on their findings. Perhaps it also makes sense to include approximate neural delays at different stages for a direct comparison. In addition, it might also help to illustrate how this compensation is presumably achieved in space by shifting encoding locations. I believe that these concepts can sometimes be tricky to communicate and the paper would benefit from those visual aids.

2. Have the authors tested for potential differences in how well static positions and movements can be decoded at different locations in the visual field? I'm happy for the authors that the analysis works when collapsing across all positions, but I can imagine that many readers would be curious if (at least static) decoding is uniform across the visual field. If certain positions/movements are decoded better then the authors could consider the analysis in point 4 (at their own discretion).

3. The findings displayed in Figure 4 (reversal of time to peak likelihood) are very intriguing. I was hoping the authors could discuss these particular results in a bit more detail in the discussion, especially in the light of feedback signals during motion processing. If the observed 'reversed' patterns indeed reflect feedback then by the time the feedback arrives back in the early visual cortex the original stimulus will have moved on and might be replaced by activity from subsequent (mismatching) bottom-up stimulus processing. This seems to be somewhat problematic for their interpretation. In addition, referring to Hogendoorn and Burkitt (2019) the authors make a point that delay compensation should occur in feedback activity, parts of which they do seem to capture with their method (Figure 4). Can the authors elaborate a bit on the potential functional relevance of the observed feedback signals and how they might differ from motion extrapolation specific feedback?

4. Was there a group of subjects (or selection of best stimulus locations for decoding) for which the early static representations (<100 ms) reliably generalized to motion? This might help to distinguish between poor signal-to-noise and entirely different codes (or direct connections from LGN/pulvinar) between early processing of static and motion stimuli.

Reviewer #3 (Recommendations for the authors):

I found the question under investigation to be interesting and I appreciate the authors' choice of recording modality and analytical methods. The paper is also well-organized and the results are presented with clarity. However, I believe the scope of this work is rather narrow and is better suited for a more specialized journal with a focus on human cognition/perception.

Summary of Revisions (for the authors):

Overall assessment:

The reviewers find the patterns in the data interesting and believe this view would be shared by other visual neuroscientists as well as neuroscientists from other fields – where neural delays are likely to prove a universal sticking point to many theories. One of its major strengths lies in the successful attempt to generalize neural representations of static objects to moving objects. They also found the manuscript to be well-written – well-tuned, clear, and interesting, with well-presented figures.

We thank the reviewers for their thoughtful, positive and constructive comments on the manuscript. Please find responses to the outstanding questions below. We believe that the edits made while addressing these questions have substantively improved the manuscript.

Introduction:

1. A little more work could be done to sell this to readers who do not study motion perception in vision. Compensating for neural processing delays will be a problem that most neuroscientific theories need to incorporate; hence why we believe this is especially suitable for publication in eLife. We would suggest adding some literature considering an application to other sensory (e.g, auditory) processing, as well as potentially higher-level cognitive domains.

Thank you for pointing out the broader application of our work. We certainly agree that the problem of overcoming neural delays is not restricted to motion perception. We have therefore broadened the scope of the introductory paragraphs:

Lines 23-43: “Responding quickly to the world around us is a primary function of the central nervous system: catching prey, escaping predators, and avoiding moving objects (e.g., falling rocks) are all crucial to survival. […] In this case, there is evidence that predictive processes help to compensate both for the neural delays incurred before visual input reaches the visual cortex and for the delays incurred during subsequent cortical processing (Berry et al., 1999; Jancke et al., 2004; Orban et al., 1985; Subramaniyan et al., 2018; Sundberg et al., 2006).”

Results:

2. We found the Results section to be well-written, like the rest of the manuscript, but due to the complexity of the analysis we think the reader could be given a little more help. E.g., in the final paragraph of the Results section can some clear flags be inserted to show which aspects of the regression result support which of the conclusions? We think this would help more non-specialised readers grasp the reasons for the conclusion.

Thank you for this comment. We have added some pointers to the specific equations that support our conclusions in the final section of the Results (see below), as well as rephrasing some sections throughout the Results, which we hope will improve clarity. We have also added an additional statistical test that enables us to directly compare the gradient and intercept of the relationship between training time and peak times for static and motion stimuli, which we hope makes it clearer from where our conclusions are coming. Additionally, please see the response to Question 3 for a new figure, which we hope helps illustrate our findings.

Results, Lines 260-290: “Furthermore, the difference between the static and motion gradients was not significantly different from zero (t = -1.58, p = 0.13) but the difference between the two intercepts was (t = 5.32, p = 3.32x10^-5).[…] Based on the training time, these early stimulus representations likely originated in early visual cortex (V1-3), meaning that the early visual system was able to almost completely compensate for neural delays accumulated during processing up to that point and represent moving objects close to their real-time position.”

3. We would recommend the authors include an illustration of the compensation achieved at different cortical processing stages based on their findings. Perhaps it also makes sense to include approximate neural delays at different stages for a direct comparison. In addition, it might also help to illustrate how this compensation is presumably achieved in space by shifting encoding locations. We believe that these concepts can sometimes be tricky to communicate and the paper would benefit from those visual aids.

Thank you for this suggestion. An additional panel has been added to Figure 4, as well as text to explain it in the final section of the Results. We hope this illustration helps to communicate the delays that are compensated at different levels of processing. We chose not to illustrate the spatial lag, but have added additional text to explain how the temporal shift corresponds to a spatial shift.

Lines 266-268: In other words, delays that accumulate during cortical processing did not appear to be compensated when processing motion. This is illustrated in Figure 4c, where the relative delay between neural representations is preserved regardless of whether the stimulus is moving or not.

Lines 280-287: In turn, this means that these neural representations were activated at the time that the moving object was physically centred on the corresponding position. This can also be seen in Figure 4c: the representations of the moving stimulus emerge 73ms earlier than the same representations of the flashed stimulus. The earliest representation of the moving stimulus peaks when t = 0ms, the time at which the moving stimulus is in the same position of the flash. This corresponds to a spatial shift in the encoding of the moving stimulus, as, by the time the representations of the flash emerge, the moving stimulus will be represented further ahead on its path.

4. Have the authors tested for potential differences in how well static positions and movements can be decoded at different locations in the visual field? We can imagine that many readers would be curious if (at least static) decoding is uniform across the visual field. If certain positions/movements are decoded better then the authors could also consider the following analysis: Was there a group of subjects (or selection of best stimulus locations for decoding) for which the early static representations (<100 ms) reliably generalized to motion? This might help to distinguish between poor signal-to-noise and entirely different codes (or direct connections from LGN/pulvinar) between early processing of static and motion stimuli.

We had not tested this, but we agree that these aspects of the decoding analysis are worth exploring further.

Firstly, we have investigated how stimulus-position likelihood varies over the visual field for the static stimulus presentations. We have now included the graph as a supplement to Figure 2. It shows the difference from chance likelihood as a function of time, with different lines corresponding to stimuli at different eccentricities. We decided to average by eccentricity, as the graph is quite difficult to read with 37 lines (i.e. one for each stimulus position; this can be seen in Figure II, top left panel).

It can be observed that decoding performance is highest for stimuli at fixation, and that performance generally decreases with increasing eccentricity. However, such differences are not clearly visible for the three furthest eccentricities relative to fixation. We believe that this is because, although the most eccentric stimuli probably elicit the smallest EEG signal, these likelihood calculations include comparisons between stimuli that are very far apart. For example, some pairwise comparisons included in the likelihood calculation for the stimulus on the far left are up to 24dva apart, but at fixation the maximum distance between the stimulus and the comparison positions is half the grid, 12dva. In Figure 2A in the manuscript, it can be seen that classification performance is highest for pairwise comparisons between stimuli further apart. Additionally, it has been shown that foveally presented stimuli elicit a much larger event-related potential (ERP) than peripheral stimuli (Rousselet et al., 2005), which would likely result in larger-magnitude ERPs for static stimuli at fixation. This could explain why the likelihood is dramatically higher for stimuli presented at fixation. We have noted these issues in the Figure 2 supplement caption.

Results, Lines 112-114: Stimulus-position likelihoods by eccentricity (distance from fixation) are plotted in Figure 2 —figure supplement 1.

Secondly, we explored whether selecting certain participants/positions with better position decoding would enable us to investigate a wider time window of motion processing. To this end, we selected good participants or good positions on the basis of the strength of decoding of flashes (to avoid cherry-picking within the motion decoding). More specifically, we selected either participants (Author response image 1) or positions (Author response image 2) for which decoding of the position of the flash was above 2*chance likelihood (i.e. 2*1/37). Through this approach we selected three participants and six positions. With regards to position, the best decoding of stimulus location was possible when the stimulus was horizontally aligned with, or just below, fixation. This is consistent with known anisotropies in visual field processing and representation (e.g. Carrasco et al., 2001; Van Essen et al., 1984). This selection did not enable us to meaningfully extend our results to earlier timepoints and, due to the exclusion of significant amounts of data, we believe our signal-to-noise ratio was lowered. In the case of looking at the best positions, we can now fit a Gaussian with R² greater than 0.5 at training timepoints before 50ms training time, when it is impossible that stimulus information is present in the brain. This indicates that the selection of a subset of participants and positions may be biasing our results toward finding decodable information at times when this would not be available to participants in the flashed condition.

One point of interest is that with both of these selection methods, the time of peak likelihood for training times 80-100ms is well before 0ms. This implies that these early representations of the stimulus might be over-extrapolated and that the brain may be representing the stimulus in a position it has not yet reached. However, we are reluctant to overinterpret analyses that lack power and as such we would prefer not to include these additional analyses in the main manuscript.

Author response image 1

Download asset Open asset

Author response image 2

Download asset Open asset

Discussion:

5. The findings reported are derived from a specific case of motion perception which may not reflect the general mechanisms optimized for motion perception. The limitations related to task designs and the neural readouts should be discussed as they affect the way that the reported results will be interpreted.

We agree that these findings reflect just one specific case of motion, in particular because the stimulus moved at a constant (rather slow) speed, and along a predictable path. We have added the following sections to the Discussion to ensure this limitation is explicit. Some limitations of the neural readout approach are already discussed, namely the inability to pinpoint the signal to a particular neural source, and the low signal-to-noise ratio. To flag this section more clearly, and to better organise the Discussion in general, we have added subheadings to the Discussion. This section is now titled ‘Limitations’. We would welcome any further suggestions of limitations that we should include in this section. We have also compared the smooth motion used in our study to apparent motion, and have highlighted important differences between these, in our response to comment 6.

Lines 364 to 369: “If information about object motion is used to compensate for delays via feedback connections, it could be expected that the particular speed of the stimulus would be exploited to maximise the accuracy of the position representation. In the present study, we only tested one stimulus speed, so further research including a range of speeds, as found in natural scenes (Wojtach et al., 2008), is necessary to examine whether compensatory mechanisms work in this way.”

Lines 415 to 417: “Future research should extend the present findings by investigating the neural consequences of an unexpected change of direction, and generally unpredictable motion, in a smooth motion trajectory.”

6. This study is clearly different from the authors' earlier work, e.g., Blom et al. However, we think it would be useful to add some discussion to the Discussion section to outline what these Blom et al. studies showed and clearly, therefore, flag the novelty of this study/analysis.

Thank you for pointing this out. We have added a paragraph to the discussion comparing our results with Hogendorrn and Burkitt's (2018) and Blom et al.’s (2020; 2021) previous findings regarding apparent motion. Importantly, this prior work differs from ours in that we investigated smooth (rather than apparent) motion in the manuscript. The primary difference between the previous findings and ours is the extent of the compensation found. We have outlined these discrepancies and the possible reasons for differences in findings below:

Lines 304 – 420: “Comparison between smooth and apparent motion

Previous EEG research has investigated whether neural delays are compensated when viewing apparent motion. […] Delays seem to be compensated to a greater extent in this case of smooth motion, so the visual system would have to employ a larger correction for erroneous position encoding if the stimulus changed direction.”

7. The approach of using different epochs to train to approximate different stages of neural processing is clever. However, it seems that later epochs should reflect processing in later hierarchical points while still also reflecting processing in early points in the hierarchy. The authors should discuss the implications of this for their conclusions.

We agree that later timepoints in the epoch could capture ongoing processing at early points in the visual hierarchy, as well as processing in later hierarchical stages. Previous research suggests that the early visual response is dominated by a feedforward cascade, due to the timing of neural responses and delays between activity in different cortical areas (Lamme and Roelfsema, 2000). Additionally, the TGM in Figure 2c shows that there is not substantial cross-generalisation across time before 150ms (i.e. the red blob is quite restricted to the diagonal). Because our regression analyses were restricted to multivariate patterns occurring less than 150 post stimulus onset, we therefore believe this is not a significant issue in the present research. We have now added a sentence to the Results about this issue. However, even if there is lingering activity from earlier visual areas, we would argue that this does not alter the main conclusions of the paper (processing happens at the same rate regardless of whether the stimulus is moving or not; there is a lot of compensation for delays very early in processing). This is now mentioned in the Limitations section of the Discussion.

Results, Lines 130-132: “It appears that the stimulus representations before 150ms training/test time did not generalise to other timepoints, but later reactivation of representations is apparent after 150ms (off-diagonal red blobs).”

Discussion, Lines 435-440: “While the initial response is likely to be dominated by a feedforward cascade through the visual hierarchy (Lamme and Roelfsema, 2000), it is possible that activity recorded at later timepoints in the training epoch reflects ongoing processing in hierarchically early visual areas, as well as additional processing in hierarchically later visual areas. However, this issue does not alter our conclusions concerning the relative timing of activity when viewing static or moving stimuli.”

Relatedly, we agree with the suggestion that later timepoints within the training epoch could reflect reactivation of earlier stimulus representations, as well as further processing of the stimulus. We believe this is the reason that the observed oscillatory pattern emerges. For more discussion of this point, please refer to our response to Question 9.

8. Could there be an impact on these findings of a large neural response when a stimulus enters the edge of a neuron's RF, followed by some adaptation? This impact should be discussed.

Thank you for this suggestion. We agree that adaptation in response to the leading edge of the stimulus is a possible mechanism for the observed effect. This idea is similar to theories about how delays are compensated in the retina when viewing motion (e.g., Berry et al., 1999). Previous studies have suggested that similar computation could indeed occur in the visual cortex, as the organisation of neurons could bring about the balance of excitation and inhibition necessary to compensate for delays in this way (Johnston and Lagnado, 2015). It takes the stimulus 400ms to move from one flash position to the next, so the leading edge of the stimulus is closer to a particular flash position than the previous at -200ms. We see that the likelihood of the stimulus being present in that position starts ramping up much earlier that -200ms, so we think that this compensation is unlikely to be completely driven by a large response to the leading edge, followed by adaptation. We have added words to this effect on Lines 267-273, as shown below:

Lines 347 – 356: “In particular, the retinal compensation mechanisms only appear to act up to speeds of about 5dva/s (Berry et al., 1999; Jancke et al., 2004). This is thought to be achieved by a large response to the leading edge of the stimulus, followed by gain control mechanisms reducing neuronal firing rates. In the present study, it takes 400ms for the stimulus to travel from one static stimulus position to the next, so the leading edge of the moving stimulus is closest to the stimulus position it is approaching 200ms before it reaches that position. In general, we see a ramping in the likelihood earlier than -200ms (see Figure 3 – Supplementary Figure 1), so we believe that this adaptation is unlikely to be the sole mechanism responsible for the observed shift in encoded location.”

9. The findings displayed in Figure 4 (reversal of time to peak likelihood) are very intriguing. We believe the manuscript would benefit from discussing these particular results in a bit more detail in the discussion, especially in the light of feedback signals during motion processing. If the observed 'reversed' patterns indeed reflect feedback then by the time the feedback arrives back in the early visual cortex the original stimulus will have moved on and might be replaced by activity from subsequent (mismatching) bottom-up stimulus processing. In addition, referring to Hogendoorn and Burkitt (2019) the authors make a point that delayed compensation should occur in feedback activity, parts of which they do seem to capture with their method (Figure 4). Can the authors elaborate a bit on the potential functional relevance of the observed feedback signals and how they might differ from motion extrapolation-specific feedback?

Thank you for this comment; we agree that feedback activity likely plays an important role in compensation for delays. However, while we do find evidence of continuing feedforward and feedback sweeps of activity, we do not think that this is particularly informative to the research question. This is because, rather than corresponding to further processing of the moving stimulus, we believe the oscillatory pattern emerges because part of the flash-evoked signal reflects reactivation of earlier stimulus representations. To explain further, as time passes after the onset of a static stimulus, stimulus-evoked activity progresses through different visual areas. However, there is also an increase in recurrent processing, whereby earlier representations of the stimulus are reactivated. We believe that after ~150ms, the further processing of the static and moving stimuli diverges. However, because part of the flash-evoked signal reflects recurrent processing in earlier visual areas, it is still possible to decode the position of the motion on the basis of these reactivated representations. We therefore think this oscillatory pattern (which is similar to that seen in Dijkstra et al. (2020)) is more informative about processing of static stimuli than motion. We have added a few sentences to this effect in the Discussion section about feedback. We additionally realised that this line of reasoning was not clear in the Results, so we have tried to clarify this point there as well.

Results, Lines 199-219: “This pattern subsequently reverses between 150 and 200ms, indicating that hierarchically later representations were activated at a shorter latency. Finally, from a training time of ~250ms the time to peak likelihood was stable at approximately 50ms. […] If this later activity (>150ms training time) does indeed reflect feedback processing of the static stimulus, then, for these later timepoints, the latency measure we have calculated might not be informative about the time necessary to first represent the moving object, because the initial activations and the reactivations are indistinguishable.”

Discussion, Lines 383 – 391: “Although this analysis captures feedforward and feedback sweeps of activity through the visual cortex, we believe that these oscillations are present in the processing of the flash, not the motion. This would mean that the portion of the later flash-evoked signal that cross-generalises to motion reflects reactivation of earlier stimulus representations rather than further stimulus processing. However, there remains the possibility that these oscillations are evoked by motion. This would imply that, during feedback activity, the position of the stimulus is not updated as the stimulus moves, and that the stimulus is likely represented in multiple locations concurrently in the visual system (King and Wyart, 2021). Further research is needed to tease apart these options.”

References

Benvenuti, G., Chemla, S., Boonman, A., Perrinet, L., Masson, G. S., and Chavane, F. (2020). Anticipatory responses along motion trajectories in awake monkey area V1. BioRXiv. https://doi.org/10.1101/2020.03.26.010017

Berry, M. J., Brivanlou, I. H., Jordan, T. A., and Meister, M. (1999). Anticipation of moving stimuli by the retina. Nature, 398(6725), 334–338. https://doi.org/10.1038/18678

Blom, T., Bode, S., and Hogendoorn, H. (2021). The time-course of prediction formation and revision in human visual motion processing. Cortex, 138, 191–202. https://doi.org/10.1016/j.cortex.2021.02.008

Blom, T., Feuerriegel, D., Johnson, P., Bode, S., and Hogendoorn, H. (2020). Predictions drive neural representations of visual events ahead of incoming sensory information. Proceedings of the National Academy of Sciences of the United States of America, 117(13). https://doi.org/10.1073/pnas.1917777117

Carrasco, M., P.Talgar, C., and Cameron, E. L. (2001). Characterizing visual performance fields: Effects of transient covert attention, spatial frequency, eccentricity, task and set size. Spatial Vision, 15(1), 61–75. https://doi.org/10.1163/15685680152692015

Chemla, S., Reynaud, X. A., Di Volo, M., Yann Zerlaut, X., Perrinet, X. L., Destexhe, A., and Chavane, X. F. (2019). Suppressive traveling waves shape representations of illusory motion in primary visual cortex of awake primate. Journal of Neuroscience, 39(22). https://doi.org/10.1523/JNEUROSCI.2792-18.2019

Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204. https://doi.org/10.1017/S0140525X12000477

Dijkstra, N., Ambrogioni, L., Vidaurre, D., and van Gerven, M. (2020). Neural dynamics of perceptual inference and its reversal during imagery. ELife, 9. https://doi.org/10.7554/eLife.53588

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787

Friston, K. J., Daunizeau, J., Kilner, J., and Kiebel, S. J. (2010). Action and behavior: A free-energy formulation. Biological Cybernetics, 102(3), 227–260. https://doi.org/10.1007/s00422-010-0364-z

Guo, K., Robertson, R. G., Pulgarin, M., Nevado, A., Panzeri, S., Thiele, A., and Young, M. P. (2007). Spatio-temporal prediction and inference by V1 neurons. European Journal of Neuroscience, 26(4), 1045–1054. https://doi.org/10.1111/j.1460-9568.2007.05712.x

Harvey, B. M., and Dumoulin, S. O. (2011). The relationship between cortical magnification factor and population receptive field size in human visual cortex: Constancies in cortical architecture. Journal of Neuroscience, 31(38), 13604–13612. https://doi.org/10.1523/JNEUROSCI.2572-11.2011

Hogendoorn, H., and Burkitt, A. N. (2018). Predictive coding of visual object position ahead of moving objects revealed by time-resolved EEG decoding. NeuroImage, 171, 55–61. https://doi.org/10.1016/j.neuroimage.2017.12.063

Jancke, D., Erlhagen, W., Schöner, G., and Dinse, H. R. (2004). Shorter latencies for motion trajectories than for flashes in population responses of cat primary visual cortex. J Physiol, 556, 971–982. https://doi.org/10.1113/jphysiol.2003.058941

Johnston, J., and Lagnado, L. (2015). General features of the retinal connectome determine the computation of motion anticipation. ELife, 2015(4). https://doi.org/10.7554/eLife.06250

Kiebel, S. J., Daunizeau, J., and Friston, K. J. (2008). A Hierarchy of Time-Scales and the Brain. PLoS Computational Biology, 4(11), e1000209. https://doi.org/10.1371/journal.pcbi.1000209

King, J.-R., and Wyart, V. (2021). The Human Brain Encodes a Chronicle of Visual Events at each Instant of Time thanks to the Multiplexing of Traveling Waves. The Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.2098-20.2021

Lamme, V. A. F. F., and Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11), 571–579. https://doi.org/10.1016/S0166-2236(00)01657-X

Millidge, B., Seth, A., and Buckley, C. L. (2022). Predictive Coding: A Theoretical and Experimental Review. ArXiv. https://doi.org/10.48550/arXiv.2107.12979

Nijhawan, R. (1994). Motion extrapolation in catching. Nature, 370(6487), 256–257. https://doi.org/10.1038/370256b0

Orban, G. A., Hoffmann, K. P., and Duysens, J. (1985). Velocity selectivity in the cat visual system. I. Responses of LGN cells to moving bar stimuli: A comparison with cortical areas 17 and 18. Journal of Neurophysiology, 54(4), 1026–1049. https://doi.org/10.1152/jn.1985.54.4.1026

Rousselet, G. A., Husk, J. S., Bennett, P. J., and Sekuler, A. B. (2005). Spatial scaling factors explain eccentricity effects on face ERPs. Journal of Vision, 5(10), 755–763. https://doi.org/10.1167/5.10.1

Subramaniyan, M., Ecker, A. S., Patel, S. S., Cotton, R. J., Bethge, M., Pitkow, X., Berens, P., and Tolias, A. S. (2018). Faster processing of moving compared with flashed bars in awake macaque V1 provides a neural correlate of the flash lag illusion. Journal of Neurophysiology, 120(5), 2430–2452. https://doi.org/10.1152/jn.00792.2017

Sundberg, K. A., Fallah, M., and Reynolds, J. H. (2006). A motion-dependent distortion of retinotopy in area V4. Neuron, 49(3), 447–457. https://doi.org/10.1016/j.neuron.2005.12.023

Van Essen, D. C., Newsome, W. T., and Maunsell, J. H. R. (1984). The visual field representation in striate cortex of the macaque monkey: Asymmetries, anisotropies, and individual variability. Vision Research, 24(5), 429–448. https://doi.org/10.1016/0042-6989(84)90041-5

Wojtach, W. T., Sung, K., Truong, S., and Purves, D. (2008). An empirical explanation of the flash-lag effect. Proceedings of the National Academy of Sciences of the United States of America, 105(42), 16338–16343. https://doi.org/10.1073/pnas.0808916105

Author details

1. Philippa Anne Johnson

   University of Melbourne, Melbourne, Australia

   Present address

   Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Australia

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   pajohnson@student.unimelb.edu.au

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6125-3138

2. Tessel Blom

   University of Melbourne, Melbourne, Australia

   Present address

   Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Australia

   Contribution

   Formal analysis, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

3. Simon van Gaal

   University of Amsterdam, Amsterdam, Netherlands

   Present address

   ABC Amsterdam Brain and Cognition, University of Amsterdam, Amsterdam, Netherlands

   Contribution

   Supervision, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6628-4534

4. Daniel Feuerriegel

   University of Melbourne, Melbourne, Australia

   Present address

   Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Australia

   Contribution

   Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

5. Stefan Bode

   University of Melbourne, Melbourne, Australia

   Present address

   Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Australia

   Contribution

   Resources, Supervision, Writing – review and editing

   Competing interests

   No competing interests declared

6. Hinze Hogendoorn

   University of Melbourne, Melbourne, Australia

   Present address

   Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Australia

   Contribution

   Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

• 828

  Page views

• 105

  Downloads

• 4

  Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.