1. Neuroscience
Download icon

Statistical learning attenuates visual activity only for attended stimuli

  1. David Richter  Is a corresponding author
  2. Floris P de Lange
  1. Radboud University Nijmegen, Netherlands
Research Article
Cite this article as: eLife 2019;8:e47869 doi: 10.7554/eLife.47869
6 figures, 1 data set and 2 additional files

Figures

Expectation suppression within the ventral visual stream depends on attention.

(A) Displayed are parameter estimates + /- within subject SE for responses to expected (blue) and unexpected (green) object stimuli during the objects attended task (attended) and objects unattended task (unattended). In all three ROIs, V1 (left), LOC (middle), and TOFC (right) BOLD responses were significantly suppressed in response to expected stimuli during the objects attended task. No difference was found between BOLD responses to expected and unexpected stimuli during the objects unattended task. The interaction effect between expectation and attention condition was significant in all three ROIs. (B) Expectation suppression in primary visual cortex is stimulus unspecific, and specific only in higher visual areas. Displayed is the average expectation suppression effect (BOLD responses, unexpected minus expected) split into stimulus-driven (light gray) and non-stimulus-driven (dark gray) gray matter voxels. Data are shown for the three ROIs, V1 (left bars), LOC (middle bars), and TOFC (right bars). Expectation suppression in LOC and TOFC was significantly larger for stimulus-driven than non-stimulus-driven voxels, while no such difference was evident in V1, indicating that expectation suppression in V1 was stimulus unspecific. Error bars indicate within-subject SE. Note, that the ROI masks in panel A and B differ, for details see: ROI definition and Stimulus specificity analysis in the Materials and methods section. *p<0.05. **p<0.01. ***p<0.001.

https://doi.org/10.7554/eLife.47869.002
Figure 1—source data 1

Expectation suppression within the ventral visual stream depends on attention.

The source data file contains a separate JASP file per ROI, containing BOLD data for expected and unexpected stimuli for both attention conditions (objects attended and unattended tasks; Figure 1A). Also contained is a JASP file showing expectation suppression per ROI split into stimulus driven and not stimulus driven voxel (Figure 1B).

https://doi.org/10.7554/eLife.47869.003
Expectation suppression across cortex for attended object stimuli only.

(A) Widespread expectation suppression across cortex in the objects attended condition. Displayed are parameter estimates for unexpected minus expected image pairs overlaid onto the MNI152 2 mm template. Color indicates unthresholded parameter estimates: red-yellow clusters represent expectation suppression. Opacity represents the z statistics of the contrasts. Black contours outline statistically significant clusters (GRF cluster corrected). Significant clusters included major parts of the ventral visual stream (early visual cortex, LOC, TOFC), anterior insula, and inferior frontal gyrus during the objects attended condition (upper row). No significant clusters were evident in the objects unattended condition (middle row). The interaction (attended >unattended; bottom row) showed significant clusters similar to those of the attended condition, albeit less extensive. (B) Expectation suppression across the ventral visual stream for attended objects, but with task-irrelevant predictions. Displayed are z statistics of the contrast unexpected minus expected of the conjunction: attended task-relevant predictionstask-irrelevant predictions; data of task-irrelevant predictions from Richter et al. (2018). Exclusively the ventral visual stream clusters showed significant expectation suppression in this conjunction, while all non-sensory area clusters were no longer significant. Thus, only the ventral visual stream clusters displayed a sensitivity to conditional probabilities, irrespective of whether predictions were task-relevant or task-irrelevant, as long as the predictable stimuli were attended.

https://doi.org/10.7554/eLife.47869.004
Figure 2—source data 1

Expectation suppression across cortex for attended object stimuli only.

The source data file contains nifti images for the whole brain contrast unexpected >expected (expectation suppression). Separate files are included for each attention condition, as well as their interaction (attended >unattended), both in terms of unthresholded parameter estimate, z and thresholded z-maps (Figure 2A). The thresholded z map of the conjunction analysis (Figure 2B) is also included.

https://doi.org/10.7554/eLife.47869.005
Figure 3 with 3 supplements
Larger pupil dilations in response to unexpected compared to expected stimuli during the objects attended task.

Displayed are pupil diameter traces over time, relative to trailing image onset. Pupil diameter data for expected (blue) and unexpected (green) image pairs are shown for the objects attended task (left) and objects unattended task (right). The black line on the abscissa denotes statistically significant differences in pupil dilations between expected and unexpected images (cluster permutation test, p<0.05). In the objects attended condition significantly larger pupil dilations in response to unexpected images are evident between 1.52 to 2.88 s after trailing image onset (left). No significant difference is found in the objects unattended condition (right), nor in the interaction between conditions. The first vertical dashed line indicates leading image onset, the second vertical line trailing image onset. Shaded areas denote within-subject SE. Timepoints from −1.0 to −0.5 s served as baseline period.

https://doi.org/10.7554/eLife.47869.006
Figure 3—source data 1

Larger pupil dilations in response to unexpected compared to expected stimuli during the objects attended task.

The source data file contains the preprocessed pupil diameter traces (participants by timepoints) for each of the four experimental conditions separately (two attention by two expectation conditions).

https://doi.org/10.7554/eLife.47869.010
Figure 3—figure supplement 1
Pupil dilation influences BOLD responses in V1.

Displayed are the parameter estimates of the influence of pupil size on BOLD responses in V1. BOLD responses increase with larger pupil dilations regardless of whether stimuli were attended and expected. However, pupil dilation influenced BOLD responses more when objects were unattended than attended. Whether stimuli were expected or unexpected did not change the association between BOLD and pupil dilation. Error bars indicate within-subject SEM.

*p<0.05, **p<0.01, ***p<0.001.

https://doi.org/10.7554/eLife.47869.007
Figure 3—figure supplement 2
Pupil dilation influences BOLD responses more in non-stimulus-driven than stimulus-driven V1 voxels.

Displayed are the parameter estimates of the influence of pupil size on BOLD responses in V1. BOLD responses increase with larger pupil dilation. This association was stronger in non-stimulus-driven (left) than stimulus-driven (right) V1 gray matter voxels. Error bars indicate within-subject SEM.

*p<0.05, **p<0.01, ***p<0.001.

https://doi.org/10.7554/eLife.47869.008
Figure 3—figure supplement 3
No difference in baseline pupil size between attention tasks, nor expectation conditions.

Displayed are mean pupil sizes during the baseline period in raw units for expected and unexpected trials during the objects attended and unattended task. Pupil sizes during baseline were similar for trials with expected and unexpected object stimuli, as well as during both tasks. Error bars indicate within-subject SEM.

https://doi.org/10.7554/eLife.47869.009
Expectation suppression is associated with pupil dilation differences and behavioral benefits of expectations.

(A) Correlation of expectation suppression magnitude and pupil dilation differences due to expectation. When predictable objects are attended, trailing images that induce larger pupil dilation differences are also showing larger expectation suppression magnitudes in V1. No such association is evident when objects are unattended. (B) Correlation of expectation suppression magnitude and RT benefits due to expectation. When predictable objects are attended, larger RT benefits are associated with larger expectation suppression effects in V1 and TOFC. This association is absent when objects are unattended. Error bars indicate within-subject SEM. *p<0.05.

https://doi.org/10.7554/eLife.47869.011
Figure 4—source data 1

Neural effects of expectations are associated with pupil dilation differences and reaction time benefits.

The source data file contains two JASP files of the analyses conducted on the correlation coefficients (Fisher z-transformed Rho), correlating expectation suppression (neural metric) with (A) pupil dilation differences due to expectations and (B) RT benefits due to expectations (behavioral measure). Correlation coefficients for data from the three ROIs (V1, LOC, TOFC) and both attention tasks are provided.

https://doi.org/10.7554/eLife.47869.012
Behavioral results demonstrate statistical learning.

(A) Behavioral benefits of expectations demonstrate statistical learning. Displayed are mean accuracy (left) and mean reaction time (right) + /- within subject SE. Responses to expected stimuli are significantly more accurate and faster, an effect exclusively observed during the objects attended condition. Thus, object identity expectations benefit behavioral performance during object classification and do not impact letter classification. (B) Pairs of both the objects attended image set and the objects unattended image set were classified significantly above chance, indicating a learning of the pairs for both conditions. Displayed are mean accuracy (left) and mean reaction time (right) during the post-scanning pair recognition task, + /- within subject SE. The dashed line indicates chance level. During the pair recognition task, no differences in either classification accuracy (left) or response speed (right) were observed between pairs previously belonging to the objects attended task compared to the objects unattended task. *p<0.05. ***p<0.001.

https://doi.org/10.7554/eLife.47869.013
Figure 5—source data 1

Behavioral results demonstrate statistical learning.

The source data file contains separate JASP files containing the behavioral performance data and conducted analyses for the object and letter classification tasks (both in terms of RTs and response accuracy; Figure 4A), as well as data from the post-scanning object recognition task (Figure 4B).

https://doi.org/10.7554/eLife.47869.014
Experimental paradigm.

(A) A single trial is displayed, starting with a 500 ms presentation of the leading object and the leading character superimposed at fixation. Next, without ISI, the trailing object and trailing character are shown for 500 ms. Each trial ends with a 4000–6000 ms ITI (MRI session; 1000–2000 ms ITI learning session), showing only a fixation dot. (B) Statistical regularities depicted as image transition matrix with object pairs and trial numbers during MRI scanning. L1 to L12 represent leading objects, while T1 to T12 represent the trailing objects. Leading and trailing objects were randomly selected per participant from a larger pool of images - that is, leading images of one participant may occur as trailing objects of another participant, in a different task, or not at all. Blue cells denote expected object pairs of the objects attended (object categorization) task, while green indicates unexpected object pairs of the objects attended task. Red denotes expected objects of the objects unattended (character categorization) task, and orange indicates unexpected objects of the objects unattended task. Each participant was also assigned 12 leading and 12 trailing characters (six letters, six symbols each). Unlike the object images, there was no association between leading and trailing characters – that is, the identity of the leading and trailing character was unpredictable. White numbers represent the total number of trials of that cell during MRI scanning. In total 120 trials of each of the four conditions were shown during MRI scanning per participant. In the behavioral learning session, participants performed an orthogonal oddball detection task, during which only expected pairs were shown (i.e., only the diagonal of the matrix), for a total of 80 trials per expected pair (960 trials total).

https://doi.org/10.7554/eLife.47869.015

Data availability

All data and code necessary to replicate the reported results are available via the following URL: http://hdl.handle.net/11633/aacg3rkw.

The following data sets were generated
  1. 1
    Donders Institute
    1. D Richter
    2. Lange FP de
    (2019)
    Attentional modulation of perceptual predictions.

Additional files

Supplementary file 1

Overview of expectation suppression across cortex.

Brain areas showing significant expectation suppression (GRF cluster corrected). Listed are significant clusters with their respective area label, MNI coordinate of the peak z value, the number of voxels in the cluster, as well as the p value of the cluster and its max z statistic. For large clusters (n voxels >700) additional local z maxima (z > 3.72; that is, p<0.0001, one-sided) are also shown with area label, MNI coordinates and max z statistic. Unexp. = unexpected image pairs; Exp. = expected image pairs; Att. = objects attended task; Unatt. = objects unattended (characters attended) task.

https://doi.org/10.7554/eLife.47869.016
Transparent reporting form
https://doi.org/10.7554/eLife.47869.017

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)