After reading or hearing the sentence context, ‘In the crib there is a sleeping …', we are easily able to predict the next word, ‘baby’. In other words, we are able to access a unique lexico-semantic representation of <baby> that is different from the lexico-semantic representation of any other word (e.g. <rose>), ahead of this information becoming available from the bottom-up input. In the present study, we used Magnetoencephalography (MEG), in combination with Representational Similarity Analysis (RSA), to show that the prediction of specific words is associated with distinct spatial and temporal patterns of neural activity before the predicted word is actually presented.

Prediction is hypothesized to be a core computational principle of brain function (Clark, 2013; Mumford, 1992). During language processing, probabilistic prediction at multiple levels of representation allows us to rapidly understand what we read or hear by giving processing a head start (see Kuperberg and Jaeger, 2016a, for a review). The strength of prediction, and the precise level of representation at which it occurs, is likely to depend on many factors (see Kuperberg and Jaeger, 2016a, section 3.4). However, there is now clear neural evidence that, at least in highly constraining sentence contexts, we are able to predict the semantic features of upcoming words.

This evidence comes from several sources. First, a large body of studies show that the N400 — an event related potential (ERP) that reflects semantic processing — is reduced in response to words whose semantic features match semantic predictions generated by highly predictable (versus less predictable) contexts. For example, the N400 elicited by ‘baby’ is smaller in the constraining context ‘In the crib, there is a sleeping …' than in the less constraining context, ‘Under the tree, there is a sleeping …' (Federmeier and Kutas, 1999; Kutas and Federmeier, 2011; Kuperberg, 2016b).

Second, several studies have reported differential modulation of brain activity following highly predictable versus less predictable sentence contexts, prior to the onset of predicted words. These include larger negative-going ERP effects (Freunberger and Roehm, 2017; Grisoni et al., 2017; León-Cabrera et al., 2017; Maess et al., 2016), increases in theta power (Dikker and Pylkkänen, 2013; Piai et al., 2016), and the suppression of alpha/beta power (Piai et al., 2014; Piai et al., 2015; Rommers et al., 2017; Wang et al., 2018). These anticipatory effects have been neuroanatomically localized to both neocortical (e.g. left frontal and temporal regions; Dikker and Pylkkänen, 2013; Piai et al., 2015; Wang et al., 2018) and subcortical (e.g. hippocampus and cerebellum; Bonhage et al., 2015; Lesage et al., 2017; Piai et al., 2016; Wang et al., 2018) regions. They have been attributed either to the process of generating predictions and/or access to the lexico-semantic representations that correspond to predicted words themselves. Importantly, however, previous studies have averaged across items that predict different upcoming words. It therefore remains unclear whether the brain produces unique patterns of neural activity that correspond to the prediction of item-specific lexico-semantic representations. For example, does the particular pattern of neural activity that is produced following the context, ‘In the crib there is a sleeping …' differ from the pattern of neural activity produced following the context, ‘On Valentine’s day, he sent his girlfriend a bouquet of red …'?

Multivariate Pattern Analysis (MVPA) provides one way of addressing this question (Kriegeskorte et al., 2008; Staudigl et al., 2015; Stokes et al., 2015a). Correlational approaches were first applied to fMRI data to identify spatial patterns of brain activity representing objects categories in the ventral stream (Haxby et al., 2001). They later evolved into Representational Similarity Analysis (RSA). The basic assumption of RSA is that similarities in patterns of brain activity reflect similarities between representationally similar items. Spatial RSA has been used to identify unique patterns of spatial activity during perception, cognition and action (Kriegeskorte et al., 2007; Kriegeskorte and Kievit, 2013). More recently, it has been applied to MEG and EEG data whose excellent temporal resolution can tell us exactly when such spatially-specific patterns of neural activity are activated in relation to the appearance of bottom-up input (Stokes et al., 2015a). Moreover, the precise temporal resolution of MEG/EEG also allows for the use of an analogous RSA approach that probes temporal rather than spatial patterns of neural similarity (Staudigl et al., 2015; Michelmann et al., 2016). Both spatial and temporal RSA approaches have been successfully used in combination with MEG and EEG to decode representationally specific visual information during the perception of bottom-up input (Cichy et al., 2014), as well as during its maintenance in working memory in the absence of bottom-up input (Wolff et al., 2017).

In the present study, we used MEG, together with both spatial and temporal RSA, to ask whether, under experimental conditions that are known to encourage specific lexico-semantic prediction, distinct words are associated with distinct spatial and temporal patterns of neural activity, prior to the appearance of the predicted input. Participants read 240 sentences, all with highly constraining contexts that predicted a specific word (Figure 1A). The sentences were visually presented at a slow rate of 1000 ms per word. This ensured the generation of specific lexico-semantic predictions and guaranteed sufficient time to detect any representationally specific neural activity before the onset of the predicted word. We constructed these sentences in pairs (120 pairs) such that each member of a pair predicted the same word, even though their contexts differed (e.g. ‘In the crib, there is a sleeping …' and ‘In the hospital, there is a newborn …'). During the experiment, sentences were presented in a pseudorandom order, with at least 30 other sentences (on average 88 sentences) in between each member of a given pair.

Figure 1

Download asset Open asset

Figure 1—source data 1

Chinese stimuli together with their English translations, as well as some measures for the context up until the sentence final word (SFW), the word before the sentence final word (SFW-1) and the sentence final words (SFW).

The first tab shows the full list of the stimuli, the second tab shows a subset of sentence pairs for the control analysis.

https://doi.org/10.7554/eLife.39061.003

Download elife-39061-fig1-data1-v2.xlsx

There is some evidence that the various semantic features and properties associated with words and concepts are represented in the brain across spatially distributed multimodal networks (Damasio, 1989; Price, 2000; Martin and Chao, 2001), which can be detected using spatial RSA (e.g. Devereux et al., 2013). For example, the particular set of semantic features and properties associated with the concept, <baby> (e.g. <human>, <small>, <cries>), might be represented by a particular spatially distributed pattern of neural activity, whereas the semantic features and properties associated with the concept, <rose> (e.g. <plant>, <scalloped petals>, <fragrant smell>) might be represented by a different spatially distributed pattern of neural activity. If, following a constraining context (e.g. In the crib, there is a sleeping …'), the prediction of a unique lexico-semantic item (<baby>) is represented by a unique spatial pattern of brain activity, then this spatial pattern should be more similar following another context that predicts the same word, that is within-pair (e.g. ‘In the hospital, there is a newborn …') than following another context that predicts a different word, that is between-pair (e.g. On Valentine’s day, he sent his girlfriend a bouquet of red …'). This should be just as true if we average across all within-pair sentences and compare them with all between-pair sentences. Importantly, this effect should be evident prior to the onset of the predicted word.

To test this hypothesis, we correlated the spatial pattern of MEG data across all sensors, between all possible pairs of sentences, at all time points over the last three words of the sentences. We were particularly interested in activity following the first word at which specific lexico-semantic predictions of upcoming words could be generated: the word before the SFW. We asked whether, at this point, the resulting spatial similarity values were larger following sentence contexts that constrained for the same word (within-pairs) than those that constrained for a different word (between-pairs), see Figure 1B.

A classic hypothesis of how spatially distributed semantic information becomes bound together to represent specific concepts in the brain is through a process of ‘temporal synchrony’ (Damasio, 1989). If, following a highly constraining context, the prediction of a unique lexico-semantic item is instantiated through a unique temporal pattern of brain activity, then the temporal patterns of neural activity should be more similar following pairs of sentence contexts that constrain for the same word (within-pairs) than following pairs that constrain for a different word (between-pairs). To test this hypothesis, we correlated the temporal pattern of MEG data evoked within the prediction period (before the onset of predicted word) between all possible pairs of sentences at each MEG sensor, and we asked whether there were any sensors in which the resulting temporal similarity values were larger for within-pair sentences than between-pair sentences.

Damasio (1989) also hypothesized that temporal binding occurred within so-called ‘convergence zones’ of the brain. Although it is still a matter of debate whether multiple convergence zones exist, parts of the temporal lobe, including anterior, ventral and medial regions, have been identified as ‘semantic hubs’ that bring spatially distributed semantic information together to form single concepts (Patterson et al., 2007; Ralph et al., 2017). If these regions play a functional role in instantiating the prediction of specific lexico-semantic items through temporal binding, then unique temporal patterns of prediction (Figure 1C) should localize to these regions. To test this hypothesis, we used source localization techniques to determine the neuroanatomical source of any increased temporal similarity following sentence contexts that predicted the same word (within-pairs) versus a different word (between-pairs).

Twenty-six participants read 240 sentences, presented at a rate of one word per second, while MEG data were acquired. The sentences were constructed in pairs (120 pairs) that strongly predicted the same sentence-final word (SFW), although, during presentation, members of the same pair were separated by at least 30 (on average 88) other sentences. As an example (Figure 1A), sentences S1 (e.g. ‘In the crib there is a sleeping…') and S2 (‘In the hospital, there is a newborn…') both predicted the word ‘baby’. To avoid repetition of the predicted word across sentence pairs, one member of each pair ended with the predicted word (e.g. in S1, ‘baby’) while the other member ended with an unpredicted but plausible word (e.g. in S2, ‘child’). Participants were asked to read each sentence carefully and to answer yes/no comprehension questions following 1/6th of the sentences. Comprehension accuracy was high (98% ± 2.0%). We compared both spatial and temporal similarity patterns of sentence pairs that predicted the same SFWs (within-pairs) to those that predicted different SFWs (between-pairs), before the SFW actually appeared. All trials were included in the analysis.

Spatial RSA: The spatial pattern of neural activity was more similar in sentence pairs that predicted the same versus different words, and this effect began before the onset of the predicted word

In each participant, we quantified the degree of spatial similarity of MEG activity (30 Hz low-pass filter) produced by pairs of sentences that predicted either the same SFW (i.e. within-pairs, for example S1-A vs. S2-A’) or a different SFW (i.e. between-pairs, for example S1-A vs. S3-B) by correlating the spatial pattern of signal across sensors at each sampling point from −2000 ms until 1000 ms, relative to the onset of the SFW. We then averaged the resulting time series of spatial correlations (R-values), first within each participant and then across participants (Figure 2B). Both the within- and between-pair group-averaged time series of spatial correlation values showed a sharp increase at ~100 ms after the onset of the penultimate word (SFW-1; at −1000 ms) that lasted ~400 ms before sharply decreasing again (Figure 2A). The same pattern was observed around the onset of the previous word (SFW-2) and around the onset of the SFW itself. We attribute this general increase in spatial similarity to the visual onset and offset of each word. This general increase in spatial correlation was largest between −880 – −485 ms (R > 0.04) before the onset of the SFW, and between −897 – −507 ms (R > 0.04) before the onset of SFW-1.

Figure 2 with 4 supplements see all

Download asset Open asset

Figure 2—source data 1

Data used for plotting Figure 2 as well as its supplementary Figures.

https://doi.org/10.7554/eLife.39061.009

Download elife-39061-fig2-data1-v2.zip

Averaged across the −880 – −485 ms interval before the onset of the SFW (corresponding to 120–515 ms after the onset of SFW-1), we found that the spatial pattern of neural activity was more similar in sentence pairs that predicted the same SFW (within-pairs: R = 0.074 + /- 0.02) than in pairs that predicted different SFWs (between-pairs: R = 0.067 + /- 0.02): t₍₂₅₎ = 3.751, p < 0.001, see Figure 2B. Figure 2C shows a scatter plot of the averaged R-values per participant within this interval. Eighteen out of 26 subjects had R-values below the diagonal, that is larger values for the within-pair than the between-pair spatial correlations. In contrast, there was no difference between the within-pair (R = 0.066 + /- 0.02) and between-pair (R = 0.068 + /- 0.02) spatial correlation values averaged across the −897 – −507 ms interval before the SFW-1 (corresponding to 103–493 ms after the onset of SFW-2): t₍₂₅₎ = −0.937, p = 0.358.

This difference in spatial similarity prior to the SFW cannot be explained by differences in the number of within-pair and between-pair sentences used to compute these mean spatial correlation values. This is because the number of trials per condition can affect the variance of the estimated mean value, but not the value of the estimated mean itself. Given that we carried out statistical analyses on the estimated mean values, the different number of within-pairs and between-pairs should not affect statistical inference at the participant level (Groppe et al., 2011 and Thomas et al., 2004). Nonetheless, to convince skeptics, we repeated the analysis using a randomly selected subset of between-pair correlations that matched the number of within-pair correlations. This analysis confirmed that the within-pair spatial correlation values remained significantly greater than the between-pair correlations (t₍₂₅₎ = 2.393, p = 0.025; Figure 2—figure supplement1).

The difference in spatial similarity prior to the SFW cannot be explained by differences in lexical processing of the word before the SFW (SFW-1): this word always differed within sentence pairs, and any differences in its lexical properties (visual complexity, word frequency and syntactic class) between members of a pair were matched between pairs that constrained for the same SFW (within-pairs) and pairs that constrained for a different SFW (between-pairs). The spatial similarity effect also cannot be explained by differences in the predictability of the SFW-1: the cloze probability of these words was fairly low (11% on average) and any difference in cloze probability between members of a pair was matched between pairs that constrained for the same SFW (within-pairs) and pairs that constrained for a different SFW (between-pairs).

However, to fully exclude the possibility that the spatial similarity effect was driven by lexical processing of the SFW-1 rather than anticipatory processing of the SFW itself, we carried out an additional control analysis. First, we selected a subset of 31 pairs of sentences that contained exactly the same SFW-1, but nonetheless predicted a different SFW (43 unique SFWs; this between-pairs subset can be found in Figure 1—source data 1). Then we selected sentence pairs that constrained for these same SFWs (within-pairs), but which differed in the SFW-1. These constituted 43 within-pair sentences (also shown in Figure 1—source data 1). Various global contextual properties (such as the number of words, number of clauses, and syntactic complexity) as well as the cloze probability of the SFW-1 were matched between this within-pair and between-pair subset (all ps > 0.05). We then compared the spatial similarity between these two subsets of sentence pairs. If the increased spatial similarity associated with the within-pairs versus between-pairs was due to the lexical processing of the SFW-1, then the spatial similarity should be greater in sentence pairs containing exactly the same SFW-1 (i.e. in the subset of between-pairs) than in sentence pairs that predicted the same SFW (i.e. in the subset of within-pairs). We found no evidence for this. Instead, we found that the averaged spatial similarity across the −880 – −485 ms interval before the onset of the SFW (corresponding to 120–515 ms after the onset of SFW-1) remained larger for the within-pair sentences (R = 0.072 + /- 0.02) than the between-pair sentences (R = 0.063 + /- 0.03): t₍₂₅₎ = 1.81, p = 0.08 (Figure 2—figure supplement 2), although in this subset analysis, the difference only approached significance due to the limited statistical power (on average there were only 40 within-pairs and 29 between-pairs after artifact rejection). Interestingly, the spatial correlation values, averaged across the −897 – −507 ms interval before the SFW-1 (corresponding to 103–493 ms after the onset of SFW-2), was larger for the between-pairs (R = 0.069 + /- 0.03) than the within-pairs (R = 0.058 + /- 0.03): t₍₂₅₎ = −2.295, p = 0.03. It is possible that this difference was driven by the prediction of the same SFW-1 in the between-pairs. However, this interpretation is speculative.

During sentence presentation, we avoided the repetition of the SFW (e.g. ‘baby’) within pairs by replacing the predicted SFW of one member of a pair with an unpredicted but plausible word in the other member of the pair (e.g. ‘child’). However, one might argue that, after encountering the predicted word (‘baby’), participants retained this item within memory, and that the increased spatial similarity of brain activity when reading the other member of the pair was due to anticipatory retrieval of this item that was facilitated by its previous presentation as a SFW. To address this concern, we divided the sentence pairs into two subsets according to whether the sentences with expected or unexpected SFWs were presented first. We then applied the spatial similarity analysis to both subsets (Figure 2—figure supplement 3) and compared their spatial similarity values. A repeated measures ANOVA with the factors Order (Expected SFW first, Unexpected SFW first) and Pairs (Within-pair, Between-pair) showed no main effect of Order (F_(1,25) = 0.747, p = 0.396, η² = 0.029), nor an interaction between Order and Pairs (F_(1,25) = 1.804, p = 0.191, η² = 0.067). We conclude that previously encountering a sentence ending with the expected SFW did not inflate the spatial similarity between sentence pairs that predicted the same SFW.

We then asked whether the increased spatial similarity associated with the within-pair versus between-pair sentences reflected the prediction of semantic features over and above the prediction of a general syntactic category (it is known that nouns and verbs are associated with distinct spatial patterns of activity; Vigliocco et al., 2011). To do this, we calculated within-category spatial similarity values by averaging the spatial similarity between all pairs of sentences that predicted the same syntactic category of words (i.e. nouns or verbs) and compared these values to the within-pair spatial similarity values. We found that the spatial similarity associated with pairs that predicted the same specific words (within-pair spatial similarity: R = 0.074 + /- 0.02) was significantly larger than the spatial similarity associated with pairs that predicted the same category (within-category spatial similarity: R = 0.068 + /- 0.02), (t₍₂₅₎ = −3.559, p = 0.002; Figure 2—figure supplement 4). This suggests that the greater within-pair versus between-pair spatial similarity effect was not simply reducible to the prediction of general syntactic category.

Finally, to further characterize the time course of brain activity reflecting unique lexico-semantic predictions, we correlated the spatial pattern of activity (across sensors) between each sentence (e.g. S1-A) at each time sample (e.g. t₁) with that of its paired sentence (e.g. S2-A’) at all time samples (e.g. from t₁ to t_n) in each participant (see also King and Dehaene, 2014 and Stokes et al., 2015a), yielding a cross-temporal within-pair similarity matrix. We also calculated between-pair cross-temporal similarity matrices and averaged these within each participant and then across participants (Figure 2D). As expected, both the within- and between-pair group-averaged cross-temporal spatial similarity matrices showed that the spatial similarity was strongest around the diagonal in the first 500 ms after the onset of SFW-1. This was also the case for the difference between the within-pair and between-pair matrices (cluster-based permutation test: p = 0.002). This effect along the diagonal is consistent with the spatial similarity difference reported in Figure 2B and C. Moreover, the absence of an effect off the diagonal suggests that the spatial patterns associated with prediction changed over time.

Temporal RSA: The temporal pattern of neural activity was more similar in sentence pairs that predicted the same versus different words, and this effect localized to left inferior temporal regions

As described above, across the −880 – −485 ms interval prior to the onset of the SFW, we observed a general increase in spatial similarity between all pairs of sentences, regardless of whether they constrained for the same SFW (within-pairs) or a different SFW (between-pairs). We next asked whether, within this time window, there were any brain regions in which the temporal pattern of neural activity was more similar for within-pairs than between-pairs. Note that this temporal RSA approach is fairly conservative in that it was limited to the time window that showed a spatial similarity effect, and so it may not have captured more extended temporal similarity effects that were not accompanied by a spatial similarity effect. The reason we took this approach is that we were interested, a priori, in any functional relationship between these measures, that is whether the spatial similarity effect reflected brain activity associated with the prediction of spatially distributed semantic representations, and whether the temporal similarity effect reflected brain activity associated with temporal binding of these spatially distributed representations. However, in order to fully exploit the spatiotemporal pattern of the data, future studies could examine the spatial and temporal patterns simultaneously using a spatiotemporal searchlight approach (Nili et al., 2014; Su et al., 2012; Su et al., 2014).

In each participant, we quantified the degree of temporal similarity of MEG activity produced by sentence pairs that predicted the same versus a different SFW by correlating the temporal pattern of signal produced within this time window at each sensor, yielding spatial topographic maps of temporal correlations. We then averaged these spatial temporal similarity maps within each participant and then across participants (Figure 1C). The group-averaged temporal similarity maps revealed a general increase in temporal similarity over bilateral temporal and posterior sensors, regardless of whether sentences predicted the same or a different SFW. When comparing the within- and between-pair temporal similarity topographic maps (Figure 3A), the temporal pattern of neural activity was more similar in pairs that predicted the same versus different words over central and posterior sensors (cluster-based randomization test: p = 0.002; Figure 3A: right panel). The comparison of a randomly selected subset of between-pair correlations that matched the number of within-pair correlations showed a similar, albeit slightly reduced, effect (marginally significant cluster: p = 0.0679; a cluster-randomization approach controlling for multiple comparisons over sensors; see Figure 3—figure supplement 1A for the group-averaged temporal similarity maps).

Figure 3 with 2 supplements see all

Download asset Open asset

Figure 3—source data 1

Data used for plotting Figure 3 as well as its supplementary Figures.

https://doi.org/10.7554/eLife.39061.013

Download elife-39061-fig3-data1-v2.zip

In order to estimate the underlying neuroanatomical source of the increased temporal similarity associated with the within-pair versus between-pair sentences, we repeated this analysis in source space. We first discretized the full brain volume using a grid. At each grid point, we constructed spatial filters using a ‘beamforming approach’ (a linearly constrained minimum variance technique; Van Veen et al., 1997) and applied it to the MEG data. Then, we performed the temporal similarity analysis on the time series from the spatial filters. The differences in the temporal similarity R-values were mapped on to the grid in each participant. These difference values were then morphed to the MNI brain and averaged. The source localization of the difference (corresponding to the difference of the topographic distribution, see Figure 3A: right panel) is shown in Figure 3B. It shows that the temporal pattern of neural activity was more similar in sentence pairs that predicted the same versus different words within a cluster over the left hemisphere (cluster-randomization controlling for multiple comparisons: p = 0.006). The strongest effect was found in the left inferior temporal lobe, as suggested by the 85% maximum difference of the temporal correlation values between the within-pair and between-pair sentences within the statistically significant cluster (see Figure 3—figure supplement 2A). The source extended medially into the left fusiform cortex, parahippocampus and hippocampus, as well as posteriorly into the left cerebellum. The comparison of a randomly selected subset of between-pair correlations that matched the number of within-pair correlations confirmed this finding (cluster-randomization controlling for multiple comparisons: p = 0.034, see Figure 3—figure supplement 1B for the statistically significant cluster and Figure 3—figure supplement 2B for the 85% maximum difference).

We asked whether the prediction of distinct words in highly constraining contexts is associated with distinct spatial and temporal patterns of neural activity before the appearance of new bottom-up input. To this end, we used MEG in conjunction with an RSA approach to index brain activity as participants read sentences in which the final word was highly predictable from the context. Based on a spatial correlation measure, we were able to provide evidence that the prediction of specific individual words produced unique spatial patterns of brain activity. This activity was evident between 120 and 515 ms following the word prior to the predicted sentence-final word (SFW-1). Moreover, within this time window, using a temporal correlation measure, we show that the prediction of specific individual words produced distinct temporal patterns of neural activity, which localized to the left inferior temporal region and neighboring areas. To the best of our knowledge, this is the first study to show that unique spatial and temporal patterns of neural activity are associated with the prediction of distinct words during language processing.

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 2 and 3.

1. 1. Bell AJ
   2. Sejnowski TJ

   (1997) The "independent components" of natural scenes are edge filters

   Vision Research 37:3327–3338.

   https://doi.org/10.1016/S0042-6989(97)00121-1

   • PubMed
   • Google Scholar
2. 1. Bonhage CE
   2. Mueller JL
   3. Friederici AD
   4. Fiebach CJ

   (2015) Combined eye tracking and fMRI reveals neural basis of linguistic predictions during sentence comprehension

   Cortex 68:33–47.

   https://doi.org/10.1016/j.cortex.2015.04.011

   • PubMed
   • Google Scholar
3. 1. Caramazza A

   (1996) Neuropsychology. The brain's dictionary

   Nature 380:485–486.

   https://doi.org/10.1038/380485a0

   • PubMed
   • Google Scholar
4. 1. Chen J
   2. Olsen RK
   3. Preston AR
   4. Glover GH
   5. Wagner AD

   (2011) Associative retrieval processes in the human medial temporal lobe: hippocampal retrieval success and CA1 mismatch detection

   Learning & Memory 18:523–528.

   https://doi.org/10.1101/lm.2135211

   • PubMed
   • Google Scholar
5. 1. Chen Y
   2. Shimotake A
   3. Matsumoto R
   4. Kunieda T
   5. Kikuchi T
   6. Miyamoto S
   7. Fukuyama H
   8. Takahashi R
   9. Ikeda A
   10. Lambon Ralph MA

   (2016) The 'when' and 'where' of semantic coding in the anterior temporal lobe: Temporal representational similarity analysis of electrocorticogram data

   Cortex 79:1–13.

   https://doi.org/10.1016/j.cortex.2016.02.015

   • PubMed
   • Google Scholar
6. 1. Cichy RM
   2. Pantazis D
   3. Oliva A

   (2014) Resolving human object recognition in space and time

   Nature Neuroscience 17:455–462.

   https://doi.org/10.1038/nn.3635

   • PubMed
   • Google Scholar
7. 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
8. Conference

   1. Da J

   (2004)

   A corpus-based study of character and bigram frequencies in Chinese e-texts and its implications for Chinese language instruction

   The Studies on the Theory and Methodology of the Digitalized Chinese Teaching to Foreigners: Proceedings of the Fourth International Conference on New Technologies in Teachingand Learning Chinese. pp. 501–511.

   • Google Scholar
9. 1. Damasio AR

   (1989) The brain binds entities and events by multiregional activation from convergence zones

   Neural Computation 1:123–132.

   https://doi.org/10.1162/neco.1989.1.1.123

   • Google Scholar
10. 1. Damasio H
    2. Grabowski TJ
    3. Tranel D
    4. Hichwa RD
    5. Damasio AR

    (1996) A neural basis for lexical retrieval

    Nature 380:499–505.

    https://doi.org/10.1038/380499a0

    • PubMed
    • Google Scholar
11. 1. Davachi L
    2. DuBrow S

    (2015) How the hippocampus preserves order: the role of prediction and context

    Trends in Cognitive Sciences 19:92–99.

    https://doi.org/10.1016/j.tics.2014.12.004

    • PubMed
    • Google Scholar
12. 1. DeLong KA
    2. Urbach TP
    3. Kutas M

    (2005) Probabilistic word pre-activation during language comprehension inferred from electrical brain activity

    Nature Neuroscience 8:1117–1121.

    https://doi.org/10.1038/nn1504

    • PubMed
    • Google Scholar
13. 1. Devereux BJ
    2. Clarke A
    3. Marouchos A
    4. Tyler LK

    (2013) Representational similarity analysis reveals commonalities and differences in the semantic processing of words and objects

    Journal of Neuroscience 33:18906–18916.

    https://doi.org/10.1523/JNEUROSCI.3809-13.2013

    • PubMed
    • Google Scholar
14. 1. Dikker S
    2. Pylkkänen L

    (2013) Predicting language: MEG evidence for lexical preactivation

    Brain and Language 127:55–64.

    https://doi.org/10.1016/j.bandl.2012.08.004

    • PubMed
    • Google Scholar
15. 1. Federmeier KD
    2. Kutas M

    (1999) A rose by any other name: long-term memory structure and sentence processing

    Journal of Memory and Language 41:469–495.

    https://doi.org/10.1006/jmla.1999.2660

    • Google Scholar
16. 1. Freunberger D
    2. Roehm D

    (2017) The costs of being certain: brain potential evidence for linguistic preactivation in sentence processing

    Psychophysiology 54:824–832.

    https://doi.org/10.1111/psyp.12848

    • PubMed
    • Google Scholar
17. 1. Grisoni L
    2. Miller TM
    3. Pulvermüller F

    (2017) Neural correlates of semantic prediction and resolution in sentence processing

    The Journal of Neuroscience 37:4848–4858.

    https://doi.org/10.1523/JNEUROSCI.2800-16.2017

    • PubMed
    • Google Scholar
18. 1. Groppe DM
    2. Urbach TP
    3. Kutas M

    (2011) Mass univariate analysis of event-related brain potentials/fields II: Simulation studies

    Psychophysiology 48:1726–1737.

    https://doi.org/10.1111/j.1469-8986.2011.01272.x

    • PubMed
    • Google Scholar
19. 1. Gupta AS
    2. van der Meer MA
    3. Touretzky DS
    4. Redish AD

    (2012) Segmentation of spatial experience by hippocampal θ sequences

    Nature Neuroscience 15:1032–1039.

    https://doi.org/10.1038/nn.3138

    • PubMed
    • Google Scholar
20. 1. Harrison LM
    2. Duggins A
    3. Friston KJ

    (2006) Encoding uncertainty in the hippocampus

    Neural Networks 19:535–546.

    https://doi.org/10.1016/j.neunet.2005.11.002

    • PubMed
    • Google Scholar
21. 1. Haxby JV
    2. Gobbini MI
    3. Furey ML
    4. Ishai A
    5. Schouten JL
    6. Pietrini P

    (2001) Distributed and overlapping representations of faces and objects in ventral temporal cortex

    Science 293:2425–2430.

    https://doi.org/10.1126/science.1063736

    • PubMed
    • Google Scholar
22. 1. Hindy NC
    2. Ng FY
    3. Turk-Browne NB

    (2016) Linking pattern completion in the hippocampus to predictive coding in visual cortex

    Nature Neuroscience 19:665–667.

    https://doi.org/10.1038/nn.4284

    • PubMed
    • Google Scholar
23. 1. Jafarpour A
    2. Piai V
    3. Lin JJ
    4. Knight RT

    (2017) Human hippocampal pre-activation predicts behavior

    Scientific Reports 7:5959.

    https://doi.org/10.1038/s41598-017-06477-5

    • PubMed
    • Google Scholar
24. 1. Jung TP
    2. Makeig S
    3. Westerfield M
    4. Townsend J
    5. Courchesne E
    6. Sejnowski TJ

    (2000) Removal of eye activity artifacts from visual event-related potentials in normal and clinical subjects

    Clinical Neurophysiology 111:1745–1758.

    https://doi.org/10.1016/S1388-2457(00)00386-2

    • PubMed
    • Google Scholar
25. 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
26. 1. Kriegeskorte N
    2. Formisano E
    3. Sorger B
    4. Goebel R

    (2007) Individual faces elicit distinct response patterns in human anterior temporal cortex

    PNAS 104:20600–20605.

    https://doi.org/10.1073/pnas.0705654104

    • PubMed
    • Google Scholar
27. 1. Kriegeskorte N
    2. Mur M
    3. Bandettini P

    (2008) Representational similarity analysis - connecting the branches of systems neuroscience

    Frontiers in Systems Neuroscience 2:4.

    https://doi.org/10.3389/neuro.06.004.2008

    • PubMed
    • Google Scholar
28. 1. Kriegeskorte N
    2. Kievit RA

    (2013) Representational geometry: integrating cognition, computation, and the brain

    Trends in Cognitive Sciences 17:401–412.

    https://doi.org/10.1016/j.tics.2013.06.007

    • PubMed
    • Google Scholar
29. 1. Kuperberg GR
    2. Jaeger TF

    (2016a) What do we mean by prediction in language comprehension?

    Language, Cognition and Neuroscience 31:32–59.

    https://doi.org/10.1080/23273798.2015.1102299

    • PubMed
    • Google Scholar
30. 1. Kuperberg GR

    (2016b) Separate streams or probabilistic inference? What the N400 can tell us about the comprehension of events

    Language, Cognition and Neuroscience 31:602–616.

    https://doi.org/10.1080/23273798.2015.1130233

    • PubMed
    • Google Scholar
31. 1. Kutas M
    2. Federmeier KD

    (2011) Thirty years and counting: finding meaning in the N400 component of the event-related brain potential (ERP)

    Annual Review of Psychology 62:621–647.

    https://doi.org/10.1146/annurev.psych.093008.131123

    • PubMed
    • Google Scholar
32. 1. León-Cabrera P
    2. Rodríguez-Fornells A
    3. Morís J

    (2017) Electrophysiological correlates of semantic anticipation during speech comprehension

    Neuropsychologia 99:326–334.

    https://doi.org/10.1016/j.neuropsychologia.2017.02.026

    • PubMed
    • Google Scholar
33. 1. Lesage E
    2. Hansen PC
    3. Miall RC

    (2017) Right lateral cerebellum represents linguistic predictability

    The Journal of Neuroscience 37:6231–6241.

    https://doi.org/10.1523/JNEUROSCI.3203-16.2017

    • PubMed
    • Google Scholar
34. 1. Lisman J
    2. Redish AD

    (2009) Prediction, sequences and the hippocampus

    Philosophical Transactions of the Royal Society B: Biological Sciences 364:1193–1201.

    https://doi.org/10.1098/rstb.2008.0316

    • PubMed
    • Google Scholar
35. 1. Lüders H
    2. Lesser RP
    3. Hahn J
    4. Dinner DS
    5. Morris H
    6. Resor S
    7. Harrison M

    (1986) Basal temporal language area demonstrated by electrical stimulation

    Neurology 36:505–510.

    https://doi.org/10.1212/WNL.36.4.505

    • PubMed
    • Google Scholar
36. 1. Lüders H
    2. Lesser RP
    3. Hahn J
    4. Dinner DS
    5. Morris HH
    6. Wyllie E
    7. Godoy J

    (1991) Basal temporal language area

    Brain 114:743–754.

    https://doi.org/10.1093/brain/114.2.743

    • PubMed
    • Google Scholar
37. 1. Maess B
    2. Mamashli F
    3. Obleser J
    4. Helle L
    5. Friederici AD

    (2016) Prediction Signatures in the Brain: Semantic Pre-Activation during Language Comprehension

    Frontiers in Human Neuroscience 10:.

    https://doi.org/10.3389/fnhum.2016.00591

    • PubMed
    • Google Scholar
38. 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
39. 1. Martin A
    2. Chao LL

    (2001) Semantic memory and the brain: structure and processes

    Current Opinion in Neurobiology 11:194–201.

    https://doi.org/10.1016/S0959-4388(00)00196-3

    • PubMed
    • Google Scholar
40. 1. McCarthy G
    2. Nobre AC
    3. Bentin S
    4. Spencer DD

    (1995) Language-related field potentials in the anterior-medial temporal lobe: I. Intracranial distribution and neural generators

    The Journal of Neuroscience 15:1080–1089.

    https://doi.org/10.1523/JNEUROSCI.15-02-01080.1995

    • PubMed
    • Google Scholar
41. 1. Michelmann S
    2. Bowman H
    3. Hanslmayr S

    (2016) The Temporal signature of memories: identification of a general mechanism for dynamic memory replay in humans

    PLOS Biology 14:e1002528.

    https://doi.org/10.1371/journal.pbio.1002528

    • PubMed
    • Google Scholar
42. 1. Mumford D

    (1992) On the computational architecture of the neocortex

    Biological Cybernetics 66:241–251.

    https://doi.org/10.1007/BF00198477

    • Google Scholar
43. 1. Mummery CJ
    2. Patterson K
    3. Wise RJ
    4. Vandenberghe R
    5. Vandenbergh R
    6. Price CJ
    7. Hodges JR

    (1999) Disrupted temporal lobe connections in semantic dementia

    Brain 122:61–73.

    https://doi.org/10.1093/brain/122.1.61

    • PubMed
    • Google Scholar
44. 1. Nieuwland MS
    2. Politzer-Ahles S
    3. Heyselaar E
    4. Segaert K
    5. Darley E
    6. Kazanina N
    7. Von Grebmer Zu Wolfsthurn S
    8. Bartolozzi F
    9. Kogan V
    10. Ito A
    11. Mézière D
    12. Barr DJ
    13. Rousselet GA
    14. Ferguson HJ
    15. Busch-Moreno S
    16. Fu X
    17. Tuomainen J
    18. Kulakova E
    19. Husband EM
    20. Donaldson DI
    21. Kohút Z
    22. Rueschemeyer SA
    23. Huettig F

    (2018) Large-scale replication study reveals a limit on probabilistic prediction in language comprehension

    eLife 7:e33468.

    https://doi.org/10.7554/eLife.33468

    • PubMed
    • Google Scholar
45. 1. Nili H
    2. Wingfield C
    3. Walther A
    4. Su L
    5. Marslen-Wilson W
    6. Kriegeskorte N

    (2014) A toolbox for representational similarity analysis

    PLOS Computational Biology 10:e1003553.

    https://doi.org/10.1371/journal.pcbi.1003553

    • PubMed
    • Google Scholar
46. 1. Nobre AC
    2. Allison T
    3. McCarthy G

    (1994) Word recognition in the human inferior temporal lobe

    Nature 372:260–263.

    https://doi.org/10.1038/372260a0

    • PubMed
    • Google Scholar
47. 1. Nobre AC
    2. McCarthy G

    (1995) Language-related field potentials in the anterior-medial temporal lobe: II. Effects of word type and semantic priming

    The Journal of Neuroscience 15:1090–1098.

    https://doi.org/10.1523/JNEUROSCI.15-02-01090.1995

    • PubMed
    • Google Scholar
48. 1. Nolte G

    (2003) The magnetic lead field theorem in the quasi-static approximation and its use for magnetoencephalography forward calculation in realistic volume conductors

    Physics in Medicine and Biology 48:3637–3652.

    https://doi.org/10.1088/0031-9155/48/22/002

    • PubMed
    • Google Scholar
49. 1. Oostenveld R
    2. Fries P
    3. Maris E
    4. Schoffelen JM

    (2011) FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data

    Computational Intelligence and Neuroscience 2011:1–9.

    https://doi.org/10.1155/2011/156869

    • PubMed
    • Google Scholar
50. 1. Patterson K
    2. Nestor PJ
    3. Rogers TT

    (2007) Where do you know what you know? The representation of semantic knowledge in the human brain

    Nature Reviews Neuroscience 8:976–987.

    https://doi.org/10.1038/nrn2277

    • PubMed
    • Google Scholar
51. 1. Piai V
    2. Roelofs A
    3. Maris E

    (2014) Oscillatory brain responses in spoken word production reflect lexical frequency and sentential constraint

    Neuropsychologia 53:146–156.

    https://doi.org/10.1016/j.neuropsychologia.2013.11.014

    • PubMed
    • Google Scholar
52. 1. Piai V
    2. Roelofs A
    3. Rommers J
    4. Maris E

    (2015) Beta oscillations reflect memory and motor aspects of spoken word production

    Human Brain Mapping 36:2767–2780.

    https://doi.org/10.1002/hbm.22806

    • PubMed
    • Google Scholar
53. 1. Piai V
    2. Anderson KL
    3. Lin JJ
    4. Dewar C
    5. Parvizi J
    6. Dronkers NF
    7. Knight RT

    (2016) Direct brain recordings reveal hippocampal rhythm underpinnings of language processing

    PNAS 113:11366–11371.

    https://doi.org/10.1073/pnas.1603312113

    • PubMed
    • Google Scholar
54. 1. Price CJ

    (2000) The anatomy of language: contributions from functional neuroimaging

    Journal of Anatomy 197:335–359.

    https://doi.org/10.1046/j.1469-7580.2000.19730335.x

    • PubMed
    • Google Scholar
55. 1. Rabovsky M
    2. Hansen SS
    3. McClelland JL

    (2018) Modelling the N400 brain potential as change in a probabilistic representation of meaning

    Nature Human Behaviour 2:693–705.

    https://doi.org/10.1038/s41562-018-0406-4

    • Google Scholar
56. 1. Ralph MA
    2. Jefferies E
    3. Patterson K
    4. Rogers TT

    (2017) The neural and computational bases of semantic cognition

    Nature Reviews Neuroscience 18:42–55.

    https://doi.org/10.1038/nrn.2016.150

    • PubMed
    • Google Scholar
57. 1. Reddy L
    2. Kanwisher N

    (2006) Coding of visual objects in the ventral stream

    Current Opinion in Neurobiology 16:408–414.

    https://doi.org/10.1016/j.conb.2006.06.004

    • PubMed
    • Google Scholar
58. 1. Rommers J
    2. Dickson DS
    3. Norton JJS
    4. Wlotko EW
    5. Federmeier KD

    (2017) Alpha and theta band dynamics related to sentential constraint and word expectancy

    Language, Cognition and Neuroscience 32:576–589.

    https://doi.org/10.1080/23273798.2016.1183799

    • Google Scholar
59. 1. Rommers J
    2. Federmeier KD

    (2018) Lingering expectations: A pseudo-repetition effect for words previously expected but not presented

    NeuroImage 183:263–272.

    https://doi.org/10.1016/j.neuroimage.2018.08.023

    • Google Scholar
60. 1. Schiffer AM
    2. Ahlheim C
    3. Wurm MF
    4. Schubotz RI

    (2012) Surprised at all the entropy: hippocampal, caudate and midbrain contributions to learning from prediction errors

    PLOS ONE 7:e36445.

    https://doi.org/10.1371/journal.pone.0036445

    • PubMed
    • Google Scholar
61. 1. Shimotake A
    2. Matsumoto R
    3. Ueno T
    4. Kunieda T
    5. Saito S
    6. Hoffman P
    7. Kikuchi T
    8. Fukuyama H
    9. Miyamoto S
    10. Takahashi R
    11. Ikeda A
    12. Lambon Ralph MA

    (2015) Direct exploration of the role of the ventral anterior temporal lobe in semantic memory: cortical stimulation and local field potential evidence from subdural grid electrodes

    Cerebral Cortex 25:3802–3817.

    https://doi.org/10.1093/cercor/bhu262

    • PubMed
    • Google Scholar
62. 1. St. John MF
    2. McClelland JL

    (1990) Learning and applying contextual constraints in sentence comprehension

    Artificial Intelligence 46:217–257.

    https://doi.org/10.1016/0004-3702(90)90008-N

    • Google Scholar
63. 1. Staudigl T
    2. Vollmar C
    3. Noachtar S
    4. Hanslmayr S

    (2015) Temporal-pattern similarity analysis reveals the beneficial and detrimental effects of context reinstatement on human memory

    Journal of Neuroscience 35:5373–5384.

    https://doi.org/10.1523/JNEUROSCI.4198-14.2015

    • PubMed
    • Google Scholar
64. 1. Stokes MG
    2. Wolff MJ
    3. Spaak E

    (2015a) Decoding rich spatial information with high temporal resolution

    Trends in Cognitive Sciences 19:636–638.

    https://doi.org/10.1016/j.tics.2015.08.016

    • PubMed
    • Google Scholar
65. 1. Stokes MG

    (2015b) 'Activity-silent' working memory in prefrontal cortex: a dynamic coding framework

    Trends in Cognitive Sciences 19:394–405.

    https://doi.org/10.1016/j.tics.2015.05.004

    • PubMed
    • Google Scholar
66. Conference

    1. Su L
    2. Fonteneau E
    3. Marslen-Wilson W
    4. Kriegeskorte N

    (2012)

    Spatiotemporal searchlight representational similarity analysis in EMEG source space

    Pattern Recognition in Neuroimaging (Prni), 2012 International Workshop On.

    • Google Scholar
67. 1. Su L
    2. Zulfiqar I
    3. Jamshed F
    4. Fonteneau E
    5. Marslen-Wilson W

    (2014) Mapping tonotopic organization in human temporal cortex: representational similarity analysis in EMEG source space

    Frontiers in Neuroscience 8:.

    https://doi.org/10.3389/fnins.2014.00368

    • PubMed
    • Google Scholar
68. Data

    1. Sun M

    (authors) (2003) Chinese lexicon

    973 Project . ID G1998030501A-03.

    http://www.chineseldc.org/doc/CLDC-LAC-2003-001/
69. 1. Thomas DG
    2. Grice JW
    3. Najm-Briscoe RG
    4. Miller JW

    (2004) The influence of unequal numbers of trials on comparisons of average event-related potentials

    Developmental Neuropsychology 26:753–774.

    https://doi.org/10.1207/s15326942dn2603_6

    • PubMed
    • Google Scholar
70. 1. Van Veen BD
    2. van Drongelen W
    3. Yuchtman M
    4. Suzuki A

    (1997) Localization of brain electrical activity via linearly constrained minimum variance spatial filtering

    IEEE Transactions on Biomedical Engineering 44:867–880.

    https://doi.org/10.1109/10.623056

    • PubMed
    • Google Scholar
71. 1. Vigliocco G
    2. Vinson DP
    3. Druks J
    4. Barber H
    5. Cappa SF

    (2011) Nouns and verbs in the brain: a review of behavioural, electrophysiological, neuropsychological and imaging studies

    Neuroscience & Biobehavioral Reviews 35:407–426.

    https://doi.org/10.1016/j.neubiorev.2010.04.007

    • PubMed
    • Google Scholar
72. 1. Visser M
    2. Jefferies E
    3. Lambon Ralph MA

    (2010) Semantic processing in the anterior temporal lobes: a meta-analysis of the functional neuroimaging literature

    Journal of Cognitive Neuroscience 22:1083–1094.

    https://doi.org/10.1162/jocn.2009.21309

    • PubMed
    • Google Scholar
73. 1. Wang L
    2. Hagoort P
    3. Jensen O

    (2018) Language prediction is reflected by coupling between frontal gamma and posterior alpha oscillations

    Journal of Cognitive Neuroscience 30:432–447.

    https://doi.org/10.1162/jocn_a_01190

    • PubMed
    • Google Scholar
74. 1. Wolff MJ
    2. Jochim J
    3. Akyürek EG
    4. Stokes MG

    (2017) Dynamic hidden states underlying working-memory-guided behavior

    Nature Neuroscience 20:864–871.

    https://doi.org/10.1038/nn.4546

    • PubMed
    • Google Scholar
75. Preprint

    1. Yan S
    2. Kuperberg GR
    3. Jaeger TF

    (2017) Prediction (or not) during language processing. a commentary on Nieuwland, et al (2017) And Delong, et al. (2005)

    bioRxiv.

    https://doi.org/10.1101/143750

    • Google Scholar

Author details

1. Lin Wang

   1. Department of Psychiatry, Harvard Medical School, Boston, United States
   2. Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, United States
   3. Department of Psychology, Tufts University, Medford, United States

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   For correspondence

   wanglinsisi@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6911-0660

2. Gina Kuperberg

   1. Department of Psychiatry, Harvard Medical School, Boston, United States
   2. Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, United States
   3. Department of Psychology, Tufts University, Medford, United States

   Contribution

   Conceptualization, Supervision, Funding acquisition, Visualization, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6093-7872

3. Ole Jensen

   School of Psychology, Centre for Human Brain Health, University of Birmingham, Birmingham, United Kingdom

   Contribution

   Conceptualization, Software, Supervision, Funding acquisition, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8193-8348

• 2,406

  views

• 369

  downloads

• 43

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.