Brain-imaging evidence for compression of binary sound sequences in human memory
- Cognitive Neuroimaging Unit, Université Paris-Saclay, INSERM, CEA, CNRS, NeuroSpin center, France
- Institute of Neuroscience, Key Laboratory of Primate Neurobiology, CAS Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, China
- Collège de France, Université Paris Sciences Lettres (PSL), France
Figures
Figure 1
Experimental design.
(A) List of the different 16-item sequences used in the magneto-encephalography (MEG) and fMRI experiments, with associated language of thought (LoT) complexity, and categorized according to the …
Figure 2 with 1 supplement
Behavioral data.
(A) Group-averaged sensitivity (d’) and response times for each sequence in the deviant detection task, plotted against the language of thought (LoT) complexity. A significant linear relationship …
Figure 2—figure supplement 1
Task performance: average sensitivity (d’), for each position and each sequence.
Error bars represent SEM.
Figure 3 with 2 supplements
Sequence complexity in the proposed language of thought (LoT) modulates fMRI responses.
(A) Brain areas showing an increase (hot colors) or a decrease (cold colors) in activation with sequence LoT complexity during habituation (voxel-wise p<0.001, uncorrected; cluster-wise p<0.05, FDR …
Figure 3—figure supplement 1
Positive (hot colors) and negative (cold colors) effects of language of thought (LoT) complexity effects on standard trials (voxel-wise p<0.001, uncorrected; cluster-wise p<0.05, FDR corrected).
Figure 3—figure supplement 2
Time course of group-averaged BOLD signals for each sequence in nine regions of interest (ROIs) where a language of thought (LoT) complexity effect was found.
Each mini-session lasted 160 s and was composed of five blocks (two habituation and three tests) interspersed with short rest periods of variable duration (depicted in light gray). The full time …
Figure 4
Brain responses to deviants decrease with language of thought (LoT) complexity.
Colors indicate the brain areas whose activation on deviant trials decreased significantly with complexity, in two distinct general linear models (GLMs): one in which all deviant stimuli were …
Figure 5
Sequence complexity effects in mathematics and language networks.
(A) Overlap between the brain areas showing an increase of activation with sequence language of thought (LoT) complexity during habituation in the main experiment (in red) and the brain areas …
Figure 6 with 4 supplements
Sequence complexity in the proposed language of thought (LoT) modulates magneto-encephalography (MEG) signals to habituation, standard, and deviant trials.
(A) Global field power computed for each sequence (see color legend) from the evoked potentials of the habituation, standard, and deviant trials. 0 ms indicates sound onset. Note that the time …
Figure 6—figure supplement 1
Sequence complexity modulates the contrast of deviant / matched standard trials.
(A) Global field power computed for the deviant / matched standard contrast for each sequence (see color legend). 0 ms indicates sound onset. Significant correlation with sequence complexity is …
Figure 6—figure supplement 2
Unconfounding the effects of statistical surprise and sequence complexity on magneto-encephalography (MEG) signals.
Left: amplitude of the regression coefficients β of the complexity regressor for each MEG sensor, in a general linear model where transition-based surprise, repetition, and alternation were also …
Figure 6—figure supplement 3
Spatiotemporal clusters for the complexity regressor in sensor space, shown separately for the three trial types (habituation, standard, deviant) and three general linear models of magneto-encephalography (MEG) signals: with complexity alone (left column); with complexity, transition-based surprise and repeat/alternate (middle column); and with complexity after regressing out transition-based surprise and repeat/alternate signals.
The clusters are very similar in all three cases, suggesting a robust effect of complexity irrespectively of transition statistics.
Figure 6—figure supplement 4
Amplitude of the regression coefficient β for each magneto-encephalography (MEG) sensor for the four regressors of transition statistics: repetition/alternation for item n (presented at t=0 ms), repetition/alternation for item n+1 (presented at t=250 ms), transition-based surprise for item n, and transition-based surprise for item n+1.
The transition-based surprise predictor is computed using an ideal observer estimating surprise over 100 past observations. The projection on the source space at the time of its maximal amplitude is …
Figure 7
Multivariate decoding of deviant trials from magneto-encephalography (MEG) signals, and its variation with sequence complexity.
(A) A decoder was trained to classify standard from deviant trials from MEG signals at a given time point. We here show the difference in the projection on the decision vector for standard and devian…
Figure 8
Time course of the deviancy decoder across the different types of sequences and deviant positions.
(A) Average projection of magneto-encephalography (MEG) signals onto the decoding axis of the standard/deviant decoder. For each sequence, the time course of the projection was computed separately …
Author response image 1
Author response image 2
Author response image 3
Author response image 4
Author response image 5
Author response image 6
Tables
Table 1
Coordinates of brain areas modulated by language of thought (LoT) complexity during habituation.
| Positive LoT complexity effect in habituation trials | ||||||||
|---|---|---|---|---|---|---|---|---|
| Region | H | k | p(unc.) | p(FWE-corr) | T | x | y | z |
| Supplementary motor area, precentral gyrus, superior frontal gyrus (dorsolateral), middle frontal gyrus | L/R | 8991 | <0.0001 | <0.0001 | 6.62 | 1 | 5 | 65 |
| <0.0001 | <0.001 | 5.82 | 8 | 12 | 49 | |||
| <0.0001 | <0.05 | 5.59 | 27 | 5 | 52 | |||
| Lobule VIII of cerebellar hemisphere | L | 1411 | <0.0001 | <0.0001 | 6.19 | 22 | 68 | 53 |
| Lobule VI and Crus I of cerebellar hemisphere | L | 939 | <0.0001 | <0.001 | 5.97 | 29 | 56 | 28 |
| Superior temporal gyrus, middle temporal gyrus | L | 2022 | <0.0001 | <0.05 | 5.56 | 68 | 23 | 5 |
| <0.0001 | <0.05 | 4.80 | 59 | 35 | 12 | |||
| <0.0001 | 0.213 | 4.25 | 55 | 42 | 23 | |||
| Lobule VI of cerebellar hemisphere | R | 1216 | <0.0001 | <0.05 | 5.45 | 27 | 58 | 27 |
| Lobule VIII of cerebellar hemisphere | L | 1549 | <0.0001 | <0.05 | 5.04 | 22 | 67 | 53 |
| <0.0001 | 0.118 | 4.44 | 33 | 54 | 55 | |||
| Superior temporal gyrus | R | 1039 | <0.0001 | <0.05 | 4.93 | 48 | 30 | 3 |
| <0.0001 | <0.05 | 4.79 | 67 | 44 | 17 | |||
| <0.001 | 0.880 | 3.55 | 69 | 23 | 3 | |||
| Postcentral gyrus, Inferior parietal gyrus | R | 1478 | <0.0001 | <0.05 | 4.79 | 36 | 46 | 56 |
| <0.0001 | 0.061 | 4.63 | 46 | 35 | 61 | |||
| <0.0001 | 0.170 | 4.33 | 46 | 32 | 47 | |||
| Superior parietal gyrus, Precuneus | R | 547 | <0.0001 | 0.085 | 4.54 | 17 | 67 | 58 |
| <0.001 | 0.792 | 3.65 | 24 | 60 | 42 | |||
| Inferior parietal gyrus, Postcentral gyrus | L | 1570 | <0.0001 | 0.106 | 4.47 | 31 | 42 | 44 |
| <0.0001 | 0.149 | 4.37 | 45 | 35 | 38 | |||
| <0.0001 | 0.530 | 3.90 | 40 | 42 | 61 | |||
| Negative LoT complexity effect in habituation trials | ||||||||
| Region | H | k | p(unc.) | p(FWE-corr) | T | x | y | z |
| Superior frontal gyrus (dorsolateral, medial, medial orbital), middle frontal gyrus | L/R | 12366 | <0.0001 | <0.001 | 5.86 | 19 | 67 | 12 |
| <0.0001 | <0.05 | 5.42 | 29 | 25 | 47 | |||
| <0.0001 | <0.05 | 5.33 | 6 | 44 | 58 | |||
| Middle cingulate and paracingulate gyri, precuneus | L | 1444 | <0.0001 | <0.05 | 5.26 | 1 | 33 | 51 |
| Angular gyrus | L | 1530 | <0.0001 | 0.060 | 4.63 | 43 | 65 | 37 |
| <0.001 | 0.816 | 3.63 | 33 | 54 | 24 | |||
| <0.001 | 0.938 | 3.45 | 27 | 82 | 44 | |||
| IFG pars orbitalis | L | 522 | <0.0001 | 0.354 | 4.07 | 52 | 35 | 14 |
| <0.0001 | 0.473 | 3.95 | 34 | 40 | 7 | |||
| <0.0001 | 0.645 | 3.80 | 27 | 33 | 16 |
Additional files
-
Supplementary file 1
Selected sequences.
The first column indicates the list of the different 16-item sequences used in the magneto-encephalography (MEG) and fMRI experiments. *Sequences used only in the fMRI experiment. The second column provides the sequence description obtained from the LoT. The third column is its verbal description, meant to ease the understanding of the formal expression.
- https://cdn.elifesciences.org/articles/84376/elife-84376-supp1-v1.docx
-
Supplementary file 2
fMRI complexity effect on standard trials (voxel-wise p<0.001, uncorrected; cluster-wise p<0.05, FDR corrected).
- https://cdn.elifesciences.org/articles/84376/elife-84376-supp2-v1.docx
-
Supplementary file 3
fMRI complexity effect on deviant trials (voxel-wise p<0.001, uncorrected; cluster-wise p<0.05, FDR corrected).
- https://cdn.elifesciences.org/articles/84376/elife-84376-supp3-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/84376/elife-84376-mdarchecklist1-v1.docx
