Figure - available from: Journal of Cognitive Neuroscience
This content is subject to copyright.
Visualization of the grand average ERPs, separately for people and places as well as personally familiar and famous entities. We report changes in voltage averaged across all electrodes (y axis) for all time points (x axis). We also run statistical tests, comparing ERP voltage for people versus places and for personally familiar versus famous entities. Statistically significant differences (p ≤.05 after TFCE correction) in both cases are indicated by solid color dots. Although some differences can be visually detected already from 250 msec on, the only time window where differences between different ERPs are statistically significant is the LPC/SFE range.
Source publication
Proper names are linguistic expressions referring to unique entities, such as individual people or places. This sets them apart from other words like common nouns, which refer to generic concepts. And yet, despite both being individual entities, one's closest friend and one's favorite city are intuitively associated with very different pieces of kn...
Similar publications
Graph-based classification techniques applied to neuroimaging data have played a pivotal role in advancing our comprehension of brain functionalities. Existing graph-based classification methods primarily focus on undirected networks. However, networks derived from neuroimaging data often exhibit a directed nature. In this study, we propose an inno...
Citations
... The parts of the data that could be published, together with the code, the ratings and all results and plots, are publicly available online, on the Open Science Foundation website (https://osf.io/sjtmn/). Also, notice that the same dataset has been used for a different study, focused on decoding semantic categories [60]. ...
... We controlled for name length for personally familiar people and places, since they could not be controlled a priori during stimulus selection, as was instead the case for famous names. To this aim, we used a cross-validated confound regression procedure that was previously validated [70] and used with this dataset in a category decoding setting [60]. For each train-test split, we fitted a linear regression model from the confound variable to the brain data within the train set. ...
... First of all, it is possible that the questionnaire could not capture place-specific types of information-for instance, sensory and modality-specific features-that are crucial when it comes to cognitive processing of familiar places. Secondly, as discussed in [2,60], the identity of personally familiar places may be harder to process than that of personally familiar people-for various reasons that could be evolutionary [133], social [134], or related to the availability of semantic features during retrieval [135]. This would then make it harder to correctly distinguish place-specific, as opposed to person-specific, signatures in brain activity-a result that converges with the overall lower encoding scores for places reported here ( Fig 5) and that has previously been found in the literature [60,135,136]. ...
Knowledge about personally familiar people and places is extremely rich and varied, involving pieces of semantic information connected in unpredictable ways through past autobiographical memories. In this work, we investigate whether we can capture brain processing of personally familiar people and places using subject-specific memories, after transforming them into vectorial semantic representations using language models. First, we asked participants to provide us with the names of the closest people and places in their lives. Then we collected open-ended answers to a questionnaire, aimed at capturing various facets of declarative knowledge. We collected EEG data from the same participants while they were reading the names and subsequently mentally visualizing their referents. As a control set of stimuli, we also recorded evoked responses to a matched set of famous people and places. We then created original semantic representations for the individual entities using language models. For personally familiar entities, we used the text of the answers to the questionnaire. For famous entities, we employed their Wikipedia page, which reflects shared declarative knowledge about them. Through whole-scalp time-resolved and searchlight encoding analyses, we found that we could capture how the brain processes one’s closest people and places using person-specific answers to questionnaires, as well as famous entities. Overall encoding performance was significant in a large time window (200-800ms). Using spatio-temporal EEG searchlight, we found that we could predict brain responses significantly better than chance earlier (200-500ms) in bilateral temporo-parietal electrodes and later (500-700ms) in frontal and posterior central electrodes. We also found that XLM, a contextualized (or large) language model, provided superior encoding scores when compared with a simpler static language model as word2vec. Overall, these results indicate that language models can capture subject-specific semantic representations as they are processed in the human brain, by exploiting small-scale distributional lexical data.
... 44 Secondly, even if one were able to sidestep this issue, a more pervasive one would 45 emerge: namely, that each person has highly idiosyncratic and subjective ways of 46 perceiving and describing personally familiar people and places. This makes it hard to 47 capture semantic representations from recollections of autobiographic memories 48 expressed in natural language, which constitute an exceptionally diverse and reduced 49 linguistic dataset [26][27][28]. 50 Such limitations posed by language models have had an impact on studies employing 51 them as models to capture semantic representations in the brain. ...
... The parts of the data that could be published, together with the 185 code, are publicly available online, on the Open Science Foundation website 186 (https://osf.io/sjtmn/?view_only=49dcdbf7aa2649fa9e376f07c26ee417). The 187 same dataset was used for a different study, currently under review [48]. ...
Knowledge about personally familiar people and places is extremely rich and varied, involving pieces of semantic information connected in unpredictable ways through past autobiographical memories. In this work we investigate whether we can capture brain processing of personally familiar people and places using subject-specific memories, after transforming them into vectorial semantic representations using language models. First we asked participants to provide us with the names of the closest people and places in their lives. Then we collected open-ended answers to a questionnaire, aimed at capturing various facets of declarative knowledge. We collected EEG data from the same participants while they were reading the names and subsequently mentally visualizing their referents. As a control set of stimuli, we also recorded evoked responses to a matched set of famous people and places. We then created original semantic representations for the individual entities using language models. For personally familiar entities, we used the text of the answers to the questionnaire. For famous entities, we employed their Wikipedia page, which reflects shared declarative knowledge about them. Through whole-scalp time-resolved and searchlight encoding analyses we found that we could capture how the brain processes one's closest people and places using person-specific answers to questionnaires, as well as famous entities. Encoding performance was significant in a large time window (200-800ms). In terms of spatio-temporal clusters, two main axes where encoding scores are significant emerged, in bilateral temporo-parietal electrodes first (200-500ms) and frontal and posterior central electrodes later (500-700ms). We also found that XLM, a contextualized language model or large language model, provided superior encoding scores when compared with a simpler static language model as word2vec. Overall, these results indicate that language models can capture subject-specific semantic representations as they are processed in the human brain, by exploiting small-scale distributional lexical data.
