Information theory, novelty and hippocampal responses: unpredicted or unpredictable?

doi:10.1016/j.neunet.2004.12.004

Neural Networks

Volume 18, Issue 3, April 2005, Pages 225-230

https://doi.org/10.1016/j.neunet.2004.12.004 Get rights and content

Abstract

Shannon's information theory provides a principled framework for the quantitative analysis of brain responses during the encoding and representation of event streams. In particular, entropy measures the expected uncertainty of events in a given context. This contextual uncertainty or unpredictability may, itself, be important for balancing [bottom-up] sensory information and [top-down] prior expectations during perceptual synthesis. Using event-related functional magnetic resonance imaging (fMRI), we found that the anterior hippocampus is sensitive to the entropy of a visual stimulus stream. In contrast, activity in an extensive bilateral cortico-thalamic network was dictated by the surprise or information associated with each particular stimulus. In short, we show that the probabilistic structure or context in which events occur is an important predictor of hippocampal activity.

Section snippets

Subjects

Informed consent was obtained from 12 right-handed subjects (7 male; age range 19–39 years; mean age 27.2). Ethics approval was obtained from the local Ethics Committee.

Bayesian learning

For the four outcomes, where j=1–4, we denote the total number of occurrences of outcome j up to observation i by $n_{j}^{i}$ . Our estimate of the probability of outcome j in the next observation is then $p (x_{i}) = \frac{n_{j}^{i} + 1}{\sum_{k} n_{k}^{i} + 1}$ where k is for all four outcomes, and hence the lower term of the equation is equivalent to the number of completed

Results

Consistent with previous observations (Hick, 1952, Hyman, 1953), entropy and surprise both modulated reaction times significantly (Fig. 1c). Fig. 2 demonstrates that activation in left anterior hippocampus was modulated by the entropy of the sampling distribution. Greater activation was observed for unpredictable stimulus streams, i.e. blocks with more even sampling distributions, evident in the plot of response estimates. The parameter estimates also demonstrate that the anterior hippocampus

Discussion

Our study addresses the role of the hippocampus from a particular (information theoretic) perspective that is closely related to novelty detection. Functional neuroimaging studies have consistently demonstrated novelty-dependent activation of anterior hippocampus (Tulving et al., 1996, Dolan and Fletcher, 1997, Strange et al., 1999). More recent evidence, however, suggests that the hippocampal role in novelty detection reflects a more general role in detecting mismatches between expectation and

Acknowledgements

BAS is supported by the Mary Kinross Trust. AD, WP, RJD and KJF are supported by the Wellcome Trust. We thank R. Frackowiak and N. Burgess for internal review of this manuscript.

References (35)

M.W. Brown et al.
Recognition memory: Neuronal substrates of the judgement of prior occurrence
Progress in Neurobiology
(1998)
S.M. Courtney et al.
What has fMRI taught us about human vision?
Current Opinion in Neurobiology
(1997)
H. Eichenbaum et al.
The hippocampus, memory, and place cells: Is it spatial memory or a memory space?
Neuron
(1999)
K.J. Friston
Functional integration and inference in the brain
Progress in Neurobiology
(2002)
G. Rees et al.
What can functional imaging reveal about the role of attention in visual awareness?
Neuropsychologia
(2001)
A.J. Yu et al.
Acetylcholine in cortical inference
Neural Networks
(2002)
J.M. Bernardo et al.
Bayesian theory
(1994)
A. Bischoff-Grethe et al.
Conscious and unconscious processing of nonverbal predictability in Wernicke's area
Journal of Neuroscience
(2000)
R.J. Dolan et al.
Dissociating prefrontal and hippocampal function in episodic memory encoding
Nature
(1997)
J. Downar et al.
A multimodal cortical network for detecting changes in the sensory environment
Nature Neuroscience
(2000)

K.J. Friston et al.

Spatial registration and normalisation of images

Human Brain Mapping

(1995)

K.J. Friston et al.

Statistical parametric maps in functional imaging: A general linear approach

Human Brain Mapping

(1995)

E. Halgren et al.

Endogenous potentials generated in the human hippocampal formation and amygdala by infrequent events

Science

(1980)

M.E. Hasselmo et al.

Encoding and retrieval of episodic memories: Role of cholinergic and GABAergic modulation in the hippocampus

Hippocampus

(1996)

W.E. Hick

On the rate of gain of information

Quartenary Journal of Experimental Psychology

(1952)

G.E. Hinton et al.

The “wake-sleep” algorithm for unsupervised neural networks

Science

(1995)

R. Hyman

Stimulus information as a determinant of reaction time

Journal of Experimental Psychology

(1953)

Cited by (208)

A characterization of the neural representation of confidence during probabilistic learning
2023, NeuroImage
Learning in a stochastic and changing environment is a difficult task. Models of learning typically postulate that observations that deviate from the learned predictions are surprising and used to update those predictions. Bayesian accounts further posit the existence of a confidence-weighting mechanism: learning should be modulated by the confidence level that accompanies those predictions. However, the neural bases of this confidence are much less known than the ones of surprise. Here, we used a dynamic probability learning task and high-field MRI to identify putative cortical regions involved in the representation of confidence about predictions during human learning. We devised a stringent test based on the conjunction of four criteria. We localized several regions in parietal and frontal cortices whose activity is sensitive to the confidence of an ideal observer, specifically so with respect to potential confounds (surprise and predictability), and in a way that is invariant to which item is predicted. We also tested for functionality in two ways. First, we localized regions whose activity patterns at the subject level showed an effect of both confidence and surprise in qualitative agreement with the confidence-weighting principle. Second, we found neural representations of ideal confidence that also accounted for subjective confidence. Taken together, those results identify a set of cortical regions potentially implicated in the confidence-weighting of learning.
Interindividual differences in interoception modulate behavior and brain responses in emotional inference
2022, NeuroImage
Emotional experiences are proposed to arise from contextualized perception of bodily responses, also referred to as interoceptive inferences. The recognition of emotions benefits from adequate access to one's own interoceptive information. However, direct empirical evidence of interoceptive inferences and their neural basis is still lacking. In the present fMRI study healthy volunteers performed a probabilistic emotion classification task with videotaped dynamically unfolding facial expressions. In a first step, we aimed to determine functional areas involved in the processing of dynamically unfolding emotional expressions. We then tested whether individuals with higher interoceptive accuracy (IAcc), as assessed by the Heartbeat detection task (HDT), or higher interoceptive sensitivity (IS), as assessed by the Multidimensional Assessment of Interoceptive Awareness, Version 2 (MAIA-2), benefit more from the contextually given likelihood of emotional valence and whether brain regions reflecting individual IAcc and/or IS play a role in this. Individuals with higher IS benefitted more from the biased probability of emotional valence. Brain responses to more predictable emotions elicited a bilateral activity pattern comprising the inferior frontal gyrus and the posterior insula. Importantly, individual IAcc scores positively covaried with brain responses to more surprising and less predictable emotional expressions in the insula and caudate nucleus. We show for the first time that IAcc score is associated with enhanced processing of interoceptive prediction errors, particularly in the anterior insula. A higher IS score seems more likely to be associated with a stronger weighting of attention to interoceptive changes processed by the posterior insula and ventral prefrontal cortex.
Surprise and recency in novelty detection in the primate brain
2022, Current Biology
Primates and other animals must detect novel objects. However, the neuronal mechanisms of novelty detection remain unclear. Prominent theories propose that visual object novelty is either derived from the computation of recency (how long ago a stimulus was experienced) or is a form of sensory surprise (stimulus unpredictability). Here, we use high-channel electrophysiology in primates to show that in many primate prefrontal, temporal, and subcortical brain areas, object novelty detection is intertwined with the computations of recency and sensory surprise. Also, distinct circuits could be engaged by expected versus unexpected sensory surprise. Finally, we studied neuronal novelty-to-familiarity transformations during learning across many days. We found a diversity of timescales in neurons’ learning rates and between-session forgetting rates, both within and across brain areas, that are well suited to support flexible behavior and learning in response to novelty. Our findings show that novelty sensitivity arises on multiple timescales across single neurons due to diverse but related computations of sensory surprise and recency and shed light on the computational underpinnings of novelty detection in the primate brain.
Similar network compositions, but distinct neural dynamics underlying belief updating in environments with and without explicit outcomes
2022, NeuroImage
Citation Excerpt :
Despite this opacity, we can still form and update beliefs about how likely our chosen actions are to deliver what we need or want, based on other sources of information (Ma and Jazayeri, 2014; Pouget et al., 2013). For instance, previous studies have reported that, in probabilistic environments without explicit outcomes, the anterior hippocampus monitors the entropy in sequences of visual or auditory events in order to generate expectations regarding future stimuli and guide behavior accordingly (Harrison et al., 2006; Krug et al., 2014; Strange et al., 2005; Tobia et al., 2012). However, it is yet to be determined whether action-outcome belief updating, under different conditions of access to immediate feedback, relies on different neural patterns or network dynamics.
Classic decision theories typically assume the presence of explicit value-based outcomes after action selections to update beliefs about action-outcome contingencies. However, ecological environments are often opaque, and it remains unclear whether the neural dynamics underlying belief updating vary under conditions characterized by the presence or absence of such explicit value-based information, after each choice selection. We investigated this question in healthy humans (n = 28) using Bayesian inference and two multi-option fMRI tasks: a multi-armed bandit task, and a probabilistic perceptual task, respectively with and without explicit value-based feedback after choice selections. Model-based fMRI analysis revealed a network encoding belief updating which did not change depending on the task. More precisely, we found a confidence-building network that included anterior hippocampus, amygdala, and medial prefrontal cortex (mPFC), which became more active as beliefs about action-outcome probabilities were confirmed by newly acquired information. Despite these consistent responses across tasks, dynamic causal modeling estimated that the network dynamics changed depending on the presence or absence of trial-by-trial value-based outcomes. In the task deprived of immediate feedback, the hippocampus increased its influence towards both amygdala and mPFC, in association with increased strength in the confidence signal. However, the opposite causal relations were found (i.e., from both mPFC and amygdala towards the hippocampus), in presence of immediate outcomes. This finding revealed an asymmetric relationship between decision confidence computations, which were based on similar computational models across tasks, and neural implementation, which varied depending on the availability of outcomes after choice selections.
Molecular basis underlying changes of brain entropy and functional connectivity in major depressive disorders after electroconvulsive therapy
2024, CNS Neuroscience and Therapeutics
Recent statistics shift object representations in parahippocampal cortex
2024, bioRxiv

View all citing articles on Scopus

View full text

Information theory, novelty and hippocampal responses: unpredicted or unpredictable?

Abstract

Section snippets

Subjects

Bayesian learning

Results

Discussion

Acknowledgements

Progress in Neurobiology

Current Opinion in Neurobiology

Neuron

Progress in Neurobiology

Neuropsychologia

Neural Networks

Bayesian theory

Conscious and unconscious processing of nonverbal predictability in Wernicke's area

Journal of Neuroscience

Dissociating prefrontal and hippocampal function in episodic memory encoding

Nature

A multimodal cortical network for detecting changes in the sensory environment

Nature Neuroscience

Spatial registration and normalisation of images

Human Brain Mapping

Statistical parametric maps in functional imaging: A general linear approach

Human Brain Mapping

Endogenous potentials generated in the human hippocampal formation and amygdala by infrequent events

Science

Encoding and retrieval of episodic memories: Role of cholinergic and GABAergic modulation in the hippocampus

Hippocampus

On the rate of gain of information

Quartenary Journal of Experimental Psychology

The “wake-sleep” algorithm for unsupervised neural networks

Science

Stimulus information as a determinant of reaction time

Journal of Experimental Psychology