Bottom-up Information Flow Research Articles

Spatially dynamic recurrent information flow across long-range dorsal motor network encodes selective motor goals.

Performing voluntary movements involves many regions of the brain, but it is unknown how they work together to plan and execute specific movements. We recorded high-resolution ultra-high-field blood-oxygen-level-dependent signal during a cued ankle-dorsiflexion task. The spatiotemporal dynamics and the patterns of task-relevant information flow across the dorsal motor network were investigated. We show that task-relevant information appears and decays earlier in the higher order areas of the dorsal motor network then in the primary motor cortex. Furthermore, the results show that task-relevant information is encoded in general initially, and then selective goals are subsequently encoded in specifics subregions across the network. Importantly, the patterns of recurrent information flow across the network vary across different subregions depending on the goal. Recurrent information flow was observed across all higher order areas of the dorsal motor network in the subregions encoding for the current goal. In contrast, only the top-down information flow from the supplementary motor cortex to the frontoparietal regions, with weakened recurrent information flow between the frontoparietal regions and bottom-up information flow from the frontoparietal regions to the supplementary cortex were observed in the subregions encoding for the opposing goal. We conclude that selective motor goal encoding and execution rely on goal-dependent differences in subregional recurrent information flow patterns across the long-range dorsal motor network areas that exhibit graded functional specialization.

The neurobiology of emotion regulation in posttraumatic stress disorder: Amygdala downregulation via real-time fMRI neurofeedback.

Amygdala dysregulation has been shown to be central to the pathophysiology of posttraumatic stress disorder (PTSD) representing a critical treatment target. Here, amygdala downregulation was targeted using real-time fMRI neurofeedback (rt-fMRI-nf) in patients with PTSD, allowing us to examine further the regulation of emotional states during symptom provocation. Patients (n = 10) completed three sessions of rt-fMRI-nf with the instruction to downregulate activation in the amygdala, while viewing personalized trauma words. Amygdala downregulation was assessed by contrasting (a) regulate trials, with (b) viewing trauma words and not attempting to regulate. Training was followed by one transfer run not involving neurofeedback. Generalized psychophysiological interaction (gPPI) and dynamic causal modeling (DCM) analyses were also computed to explore task-based functional connectivity and causal structure, respectively. It was found that PTSD patients were able to successfully downregulate both right and left amygdala activation, showing sustained effects within the transfer run. Increased activation in the dorsolateral and ventrolateral prefrontal cortex (PFC), regions related to emotion regulation, was observed during regulate as compared with view conditions. Importantly, activation in the PFC, rostral anterior cingulate cortex, and the insula, were negatively correlated to PTSD dissociative symptoms in the transfer run. Increased functional connectivity between the amygdala- and both the dorsolateral and dorsomedial PFC was found during regulate, as compared with view conditions during neurofeedback training. Finally, our DCM analysis exploring directional structure suggested that amygdala downregulation involves both top-down and bottom-up information flow with regard to observed PFC-amygdala connectivity. This is the first demonstration of successful downregulation of the amygdala using rt-fMRI-nf in PTSD, which was critically sustained in a subsequent transfer run without neurofeedback, and corresponded to increased connectivity with prefrontal regions involved in emotion regulation during the intervention. Hum Brain Mapp 38:541-560, 2017. © 2016 Wiley Periodicals, Inc.

Gut vagal afferents differentially modulate innate anxiety and learned fear.

Vagal afferents are an important neuronal component of the gut-brain axis allowing bottom-up information flow from the viscera to the CNS. In addition to its role in ingestive behavior, vagal afferent signaling has been implicated modulating mood and affect, including distinct forms of anxiety and fear. Here, we used a rat model of subdiaphragmatic vagal deafferentation (SDA), the most complete and selective vagal deafferentation method existing to date, to study the consequences of complete disconnection of abdominal vagal afferents on innate anxiety, conditioned fear, and neurochemical parameters in the limbic system. We found that compared with Sham controls, SDA rats consistently displayed reduced innate anxiety-like behavior in three procedures commonly used in preclinical rodent models of anxiety, namely the elevated plus maze test, open field test, and food neophobia test. On the other hand, SDA rats exhibited increased expression of auditory-cued fear conditioning, which specifically emerged as attenuated extinction of conditioned fear during the tone re-exposure test. The behavioral manifestations in SDA rats were associated with region-dependent changes in noradrenaline and GABA levels in key areas of the limbic system, but not with functional alterations in the hypothalamus-pituitary-adrenal grand stress. Our study demonstrates that innate anxiety and learned fear are both subjected to visceral modulation through abdominal vagal afferents, possibly via changing limbic neurotransmitter systems. These data add further weight to theories emphasizing an important role of afferent visceral signals in the regulation of emotional behavior.

Reply to Greyson et al.: Experimental evidence lays a foundation for a rational understanding of near-death experiences

Greyson et al. (1) state that it is misleading to describe the rat brain as hyperaroused because the EEG power at near death was only a small fraction of that of the waking state in our study (2). However, they neglect to note that we showed that the power of EEG signals associated with consciousness increased in every rat we tested [see figure 2A of Borjigin et al. (2)]. Greyson et al. posit that “the pertinent question here is not whether there is any brain electrical activity at all after cardiac arrest, but whether there is activity of the type currently thought to be necessary for conscious experience.” In complete accordance with their opinion, our entire study (2) is devoted to demonstrating the presence of electrical fingerprints of consciousness in the near-death brain. We report increased power and global synchrony in the gamma bandwidth, two neurophysiologic features associated with conscious processing. Moreover, this gamma band exhibits an eightfold increase in top-down information processing (thought to be a key element of consciousness) and fivefold increase in bottom-up information flow (thought to represent sensory information processing) at near death. In addition, we found tight coupling of gamma bands with both theta and alpha bands, yet another indicator of conscious information processing in the postarrest brain. Thus, our work (2) directly answers the concern of Greyson et al.

Hierarchical spatiotemporal feature extraction using recurrent online clustering

Deep machine learning offers a comprehensive framework for extracting meaningful features from complex observations in an unsupervised manner. The majority of deep learning architectures described in the literature primarily focus on extracting spatial features. However, in real-world settings, capturing temporal dependencies in observations is critical for accurate inference. This paper introduces an enhancement to DeSTIN – a compositional deep learning architecture in which each layer consists of multiple instantiations of a common node – that learns to represent spatiotemporal patterns in data based on a novel recurrent clustering algorithm. Contrary to mainstream deep architectures, such as deep belief networks where layer-by-layer training is assumed, each of the nodes in the proposed architecture is trained independently and in parallel. Moreover, top-down and bottom-up information flows facilitate rich feature formation. A semi-supervised setting is demonstrated achieving state-of-the-art results on the MNIST classification benchmarks. A GPU implementation is discussed further accentuating the scalability properties of the proposed framework.

”BOTTOM-UP” AND ”TOP-DOWN” APPROACH FOR DEFINING ROAD SAFETY STRATEGY - CASE STUDY: CITY OF BELGRADE

The forming process of road safety strategy is quite complex because it implies a system approach that consists of several more or less complex steps. The analysis of the existing situation, defining the problem and creating a strategic document are the basic steps in forming road safety strategy. Each one of the mentioned steps involves several activities that must be carried out. Implementation, monitoring and revision are the next steps in a life cycle of the strategy, which provide continuous monitoring and adjustment of actions and programs for road safety improvement. In this study, on the example of the strategic document for the city of Belgrade, the concept of defining strategy and key areas of road safety strategy, which uses a top-down and bottom-up information flow approach, was shown. The specificity of strategy creation lays in the fact that the city of Belgrade consists of 10 urban and 7 rural municipalities, therefore it was necessary to find a model of strategic document that would include all the problems in the city but in also in all 17 municipalities. The model presented in this study provides the creation of a so-called overarching strategic document and 17 action plans, for each of the municipalities, and the overarching document is obtained by the so-called ''overlap'' 17 action plans.

Are phonological influences on lexical (mis)selection the result of a monitoring bias?

A monitoring bias account is often used to explain speech error patterns that seem to be the result of an interactive language production system, like phonological influences on lexical selection errors. A biased monitor is suggested to detect and covertly correct certain errors more often than others. For instance, this account predicts that errors that are phonologically similar to intended words are harder to detect than those that are phonologically dissimilar. To test this, we tried to elicit phonological errors under the same conditions as those that show other kinds of lexical selection errors. In five experiments, we presented participants with high cloze probability sentence fragments followed by a picture that was semantically related, a homophone of a semantically related word, or phonologically related to the (implicit) last word of the sentence. All experiments elicited semantic completions or homophones of semantic completions, but none elicited phonological completions. This finding is hard to reconcile with a monitoring bias account and is better explained with an interactive production system. Additionally, this finding constrains the amount of bottom-up information flow in interactive models.

Knowing a word affects the fundamental perception of the sounds within it.

Understanding spoken language is an exceptional computational achievement of the human cognitive apparatus. Theories of how humans recognize spoken words fall into two categories: Some theories assume a fully bottom-up flow of information, in which successively more abstract representations are computed. Other theories, in contrast, assert that activation of a more abstract representation (e.g., a word) can affect the activation of smaller units (e.g., phonemes or syllables). The two experimental conditions reported here demonstrate the top-down influence of word representations on the activation of smaller perceptual units. The results show that perceptual processes are not strictly bottom-up: Computations at logically lower levels of processing are affected by computations at logically more abstract levels. These results constrain and inform theories of the architecture of human perceptual processing of speech.

A new approach to linking strategy formulation and strategy implementation: an example from the UK banking sector

Broady-Preston and Hayward (1998) make the point that ‘the attempt by organisations to move away from purely top-down strategic formulation is in its infancy and different organisations favour different structures and models to ease the bottom-up flow of information’ (p. 285). They further suggested that the balanced scorecard (BSC) model might be one tool to help the bottom-up flow of information. This paper uses the experience of one of the major UK banks to illustrate how the BSC might be used as part of a strategy process which can capture bottom-up information. The conclusion to be derived from this experience is that the implementation role of BSCs needs to be supported by a defined strategy formulation process. The solution which has emerged is the linking, for the first time, of the strategic architecture formulation process of Hamel and Prahalad (1996a) with the BSC strategy implementation tool of Kaplan and Norton (1996a), the balance scorecard. Boston MA: Harvard Business School Press).

Lexical and Sublexical Feedback in Auditory Word Recognition

Currently, there are two qualitatively different model classes in the field of spoken language understanding. Autonomous models allow only bottom-up information flow, whereas interactive models allow higher level representations (e.g., lexical) to affect processing at lower levels (e.g., phonemic). Part 1 of the present study included a test of a prediction that differentiates the two model classes: Is phoneme monitoring faster for targets in real words than in pseudowords, even before the word could in principle be recognized? The results indicate that this lexical advantage does occur, in accord with the predictions of interactive models. In Part 2, speech compression and expansion were used to assess the sufficiency or necessity of bottom-up evidence and of processing time in accomplishing lexical access. The results of Parts 1 and 2 suggested that in addition to the lexical effects posited by current models, sublexical activation may also play an important role. Data are presented in Part 3 that support this interpretation Collectively, the results in the current study support interactive models of lexical processing, but require additional sublexical processes as well.

The problems of language control: Editing, monitoring, and feedback

The incorporation of editors and monitors in models of language production has become a common practice over the past ten years or so. However, there is much confusion in the psycholinguistics literature about these two control mechanisms. A monitor is redefined as a television viewer who can spot problems on the screen but cannot do anything about it. An editor, in contrast, is powerful enough to change the programme so as to produce fewer and less anomalous errors. Although there is no doubt that our speech is internally monitored, it is far from clear that this is achieved via editing. Not only can editing models be shown to suffer from serious defects, but also the data which motivated the introduction of editors in the first place lend themselves more plausibly to a reanalysis in terms of evidencing properties of the regular programming process. It is argued that McClelland and Rumelhart's interactive activation model accommodates the monitoring function quite nicely without any additional theoretical apparatus. Monitoring is carried out by a mechanism called non-local feedback which results automatically from bottom-up information flow in a parallel processing system. All the speaker has to do is to check the activation levels of the nodes in his internal network. He discovers problems whenever the levels of activation are too low with respect to the idea he wishes to communicate.