Mental maze solving

Peripheral vision and crowding in mental maze-solving

Solving a maze effectively relies on both perception and cognition. Studying maze-solving behavior contributes to our knowledge about these important processes. Through psychophysical experiments and modeling simulations, we examine the role of peripheral vision, specifically visual crowding in the periphery, in mental maze-solving. Experiment 1 measured gaze patterns while varying maze complexity, revealing a direct relationship between visual complexity and maze-solving efficiency. Simulations of the maze-solving task using a peripheral vision model confirmed the observed crowding effects while making an intriguing prediction that saccades provide a conservative measure of how far ahead observers can perceive the path. Experiment 2 confirms that observers can judge whether a point lies on the path at considerably greater distances than their average saccade. Taken together, our findings demonstrate that peripheral vision plays a key role in mental maze-solving.

End-to-End One-Shot Path-Planning Algorithm for an Autonomous Vehicle Based on a Convolutional Neural Network Considering Traversability Cost

Path planning plays an important role in navigation and motion planning for robotics and automated driving applications. Most existing methods use iterative frameworks to calculate and plan the optimal path from the starting point to the endpoint. Iterative planning algorithms can be slow on large maps or long paths. This work introduces an end-to-end path-planning algorithm based on a fully convolutional neural network (FCNN) for grid maps with the concept of the traversability cost, and this trains a general path-planning model for 10 × 10 to 80 × 80 square and rectangular maps. The algorithm outputs the lowest-cost path while considering the cost and the shortest path without considering the cost. The FCNN model analyzes the grid map information and outputs two probability maps, which show the probability of each point in the lowest-cost path and the shortest path. Based on the probability maps, the actual optimal path is reconstructed by using the highest probability method. The proposed method has superior speed advantages over traditional algorithms. On test maps of different sizes and shapes, for the lowest-cost path and the shortest path, the average optimal rates were 72.7% and 78.2%, the average success rates were 95.1% and 92.5%, and the average length rates were 1.04 and 1.03, respectively.

Eye movements reveal spatiotemporal dynamics of visually-informed planning in navigation

Goal-oriented navigation is widely understood to depend upon internal maps. Although this may be the case in many settings, humans tend to rely on vision in complex, unfamiliar environments. To study the nature of gaze during visually-guided navigation, we tasked humans to navigate to transiently visible goals in virtual mazes of varying levels of difficulty, observing that they took near-optimal trajectories in all arenas. By analyzing participants’ eye movements, we gained insights into how they performed visually-informed planning. The spatial distribution of gaze revealed that environmental complexity mediated a striking trade-off in the extent to which attention was directed towards two complimentary aspects of the world model: the reward location and task-relevant transitions. The temporal evolution of gaze revealed rapid, sequential prospection of the future path, evocative of neural replay. These findings suggest that the spatiotemporal characteristics of gaze during navigation are significantly shaped by the unique cognitive computations underlying real-world, sequential decision making.

One-Shot Multi-Path Planning Using Fully Convolutional Networks in a Comparison to Other Algorithms

Path planning plays a crucial role in many applications in robotics for example for planning an arm movement or for navigation. Most of the existing approaches to solve this problem are iterative, where a path is generated by prediction of the next state from the current state. Moreover, in case of multi-agent systems, paths are usually planned for each agent separately (decentralized approach). In case of centralized approaches, paths are computed for each agent simultaneously by solving a complex optimization problem, which does not scale well when the number of agents increases. In contrast to this, we propose a novel method, using a homogeneous, convolutional neural network, which allows generation of complete paths, even for more than one agent, in one-shot, i.e., with a single prediction step. First we consider single path planning in 2D and 3D mazes. Here, we show that our method is able to successfully generate optimal or close to optimal (in most of the cases <10% longer) paths in more than 99.5% of the cases. Next we analyze multi-paths either from a single source to multiple end-points or vice versa. Although the model has never been trained on multiple paths, it is also able to generate optimal or near-optimal (<22% longer) paths in 96.4 and 83.9% of the cases when generating two and three paths, respectively. Performance is then also compared to several state of the art algorithms.

Serial grouping of 2D-image regions with object-based attention in humans

After an initial stage of local analysis within the retina and early visual pathways, the human visual system creates a structured representation of the visual scene by co-selecting image elements that are part of behaviorally relevant objects. The mechanisms underlying this perceptual organization process are only partially understood. We here investigate the time-course of perceptual grouping of two-dimensional image-regions by measuring the reaction times of human participants and report that it is associated with the gradual spread of object-based attention. Attention spreads fastest over large and homogeneous areas and is slowed down at locations that require small-scale processing. We find that the time-course of the object-based selection process is well explained by a 'growth-cone' model, which selects surface elements in an incremental, scale-dependent manner. We discuss how the visual cortical hierarchy can implement this scale-dependent spread of object-based attention, leveraging the different receptive field sizes in distinct cortical areas.

DOI:http://dx.doi.org/10.7554/eLife.14320.001

When we look at an object, we perceive it as a whole. However, this is not how the brain processes objects. Instead, cells at early stages of the visual system respond selectively to single features of the object, such as edges. Moreover, each cell responds to its target feature in only a small region of space known as its receptive field. At higher levels of the visual system, cells respond to more complex features: angles rather than edges, for example. The receptive fields of the cells are also larger. For us to see an object, the brain must therefore 'stitch' together diverse features into a unified impression.

This process is termed perceptual grouping. But how does it work? Jeurissen et al. hypothesized that this process depends on the visual system’s attention spreading over a region in the image occupied by an object, and that the speed of the process will depend on the size of the receptive fields involved. If an image region is narrow, the visual system must recruit cells with small receptive fields to process the individual features. Grouping will therefore be slow. By contrast, if the object consists of large uniform areas lacking in detail, grouping should be fast. These assumptions give rise to a model called the “growth-conemodel”, which makes a number of specific predictions about reaction times during perceptual grouping.

Jeurissen et al. tested the growth-cone model’s predictions by measuring the speed of perceptual grouping in 160 human volunteers. These volunteers looked at an image made up of two simple shapes, and reported whether two dots fell on the same or different shapes. The results supported the growth-cone model. People were able to group large and uniform areas quickly, but were slower for narrow areas. Grouping also took more time when the distance between the dots increased. Hence, perceptual grouping of everyday objects calls on a step-by-step process that resembles solving a small maze.

The results also revealed that perceptual grouping of simple shapes relies on the spreading of visual attention over the relevant object. Furthermore, the data support the hypothesis that perceptual grouping makes use of the different sizes of receptive fields at various levels of the visual system. Further research will be needed to translate these findings to the more complex natural scenes we encounter in our daily lives.

DOI:http://dx.doi.org/10.7554/eLife.14320.002

Neural mechanisms underlying the exploration of small city maps using magnetoencephalography

The neural mechanisms underlying spatial cognition in the context of exploring realistic city maps are unknown. We conducted a novel brain imaging study to address the question of whether and how features of special importance for map exploration are encoded in the brain to make a spatial decision. Subjects explored by eyes small city maps exemplifying five different street network types in order to locate a hypothetical City Hall, while neural activity was recorded continuously by 248 magnetoencephalography (MEG) sensors at high temporal resolution. Monitoring subjects' eye positions, we locally characterized the maps by computing three spatial parameters of the areas that were explored. We computed the number of street intersections, the total street length, and the regularity index in the circular areas of 6 degrees of visual angle radius centered on instantaneous eye positions. We tested the hypothesis that neural activity during exploration is associated with the spatial parameters and modulated by street network type. All time series were rendered stationary and nonautocorrelated by applying an autoregressive integrated moving average model and taking the residuals. We then assessed the associations between the prewhitened time-varying MEG time series from 248 sensors and the prewhitened spatial parameters time series, for each street network type, using multiple linear regression analyses. In accord with our hypothesis, ongoing neural activity was strongly associated with the spatial parameters through localized and distributed networks, and neural processing of these parameters depended on the type of street network. Overall, processing of the spatial parameters seems to predominantly involve right frontal and prefrontal areas, but not for all street network layouts. These results are in line with findings from a series of previous studies showing that frontal and prefrontal areas are involved in the processing of spatial information and decision making. Modulation of neural processing of the spatial parameters by street network type suggests that some street network layouts may contain other types of spatial information that subjects use to explore maps and make spatial decisions.

Thinking in spatial terms: decoupling spatial representation from sensorimotor control in monkey posterior parietal areas 7a and LIP

Perhaps the simplest and most complete description of the cerebral cortex is that it is a sensorimotor controller whose primary purpose is to represent stimuli and movements, and adaptively control the mapping between them. However, in order to think, the cerebral cortex has to generate patterns of neuronal activity that encode abstract, generalized information independently of ongoing sensorimotor events. A critical question confronting cognitive systems neuroscience at present therefore is how neural signals encoding abstract information emerge within the sensorimotor control networks of the brain. In this review, we approach that question in the context of the neural representation of space in posterior parietal cortex of non-human primates. We describe evidence indicating that parietal cortex generates a hierarchy of spatial representations with three basic levels: including (1) sensorimotor signals that are tightly coupled to stimuli or movements, (2) sensorimotor signals modified in strength or timing to mediate cognition (examples include attention, working memory, and decision-processing), as well as (3) signals that encode frankly abstract spatial information (such as spatial relationships or categories) generalizing across a wide diversity of specific stimulus conditions. Here we summarize the evidence for this hierarchy, and consider data showing that signals at higher levels derive from signals at lower levels. That in turn could help characterize neural mechanisms that derive a capacity for abstraction from sensorimotor experience.

Exploring small city maps

The exploration of city maps has exploded recently due to the wide availability, increasing use of, and reliance on small positioning and navigational devices for personal use. In this study, subjects explored small, 3-mile diameter circular maps exemplifying five different types of street networks common in the United States, in order to locate a hypothetical city hall. Chosen locations indicated that subjects are able to identify more accessible sites. Monitoring eye position revealed that women explored maps faster, using more widely dispersed but more narrowly focused gaze clusters than men. The type of street network influenced the time spent by the eyes in a locale and differentially affected the size of gaze clusters between women and men, underscoring the complex interactions of gender-specific strategies with street network types.

Do left and right asymmetries of hemispheric preference interact with attention to predict local and global performance in applied tasks?

Many cognitive neuroscience studies show that the ability to attend to and identify global or local information is lateralised between the two hemispheres in the human brain; the left hemisphere is biased towards the local level, whereas the right hemisphere is biased towards the global level. Results of two studies show attention-focused people with a right ear preference (biased towards the left hemisphere) are better at local tasks, whereas people with a left ear preference (biased towards the right hemisphere) are better at more global tasks. In a third study we determined if right hemisphere-biased followers who attend to global stimuli are likely to have a stronger relationship between attention and globally based supervisor ratings of performance. Results provide evidence in support of this hypothesis. Our research supports our model and suggests that the interaction between attention and lateral preference is an important and novel predictor of work-related outcomes.

Solving mazes with memristors: a massively parallel approach

Solving mazes is not just a fun pastime: They are prototype models in several areas of science and technology. However, when maze complexity increases, their solution becomes cumbersome and very time consuming. Here, we show that a network of memristors--resistors with memory--can solve such a nontrivial problem quite easily. In particular, maze solving by the network of memristors occurs in a massively parallel fashion since all memristors in the network participate simultaneously in the calculation. The result of the calculation is then recorded into the memristors' states and can be used and/or recovered at a later time. Furthermore, the network of memristors finds all possible solutions in multiple-solution mazes and sorts out the solution paths according to their length. Our results demonstrate not only the application of memristive networks to the field of massively parallel computing, but also an algorithm to solve mazes, which could find applications in different fields.

Eye movements reveal solution knowledge prior to insight

In two experiments, participants solved anagram problems while their eye movements were monitored. Each problem consisted of a circular array of five letters: a scrambled four-letter solution word containing three consonants and one vowel, and an additional randomly-placed distractor consonant. Viewing times on the distractor consonant compared to the solution consonants provided an online measure of knowledge of the solution. Viewing times on the distractor consonant and the solution consonants were indistinguishable early in the trial. In contrast, several seconds prior to the response, viewing times on the distractor consonant decreased in a gradual manner compared to viewing times on the solution consonants. Importantly, this pattern was obtained across both trials in which participants reported the subjective experience of insight and trials in which they did not. These findings are consistent with the availability of partial knowledge of the solution prior to such information being accessible to subjective phenomenal awareness.

Ultra-high field parallel imaging of the superior parietal lobule during mental maze solving

We used ultra-high field (7 T) fMRI and parallel imaging to scan the superior parietal lobule (SPL) of human subjects as they mentally traversed a maze path in one of four directions (up, down, left, right). A counterbalanced design for maze presentation and a quasi-isotropic voxel (1.46 x 1.46 x 2 mm thick) collection were implemented. Fifty-one percent of single voxels in the SPL were tuned to the direction of the maze path. Tuned voxels were distributed throughout the SPL, bilaterally. A nearest neighbor analysis revealed a "honeycomb" arrangement such that voxels tuned to a particular direction tended to occur in clusters. Three-dimensional (3D) directional clusters were identified in SPL as oriented centroids traversing the cortical depth. There were 13 same-direction clusters per hemisphere containing 22 voxels per cluster, on the average; the mean nearest-neighbor, same-direction intercluster distance was 9.4 mm. These results provide a much finer detail of the directional tuning in SPL, as compared to those obtained previously at 4 T (Gourtzelidis et al. Exp Brain Res 165:273-282, 2005). The more accurate estimates of quantitative clustering parameters in 3D brain space in this study were made possible by the higher signal-to-noise and contrast-to-noise ratios afforded by the higher magnetic field of 7 T as well as the quasi-isotropic design of voxel data collection.

Brain activation during mental maze solving

BACKGROUND

So-called Porteus mazes are used to investigate prefrontal cortex (PFC) functioning in normal subjects and patients with different neuropsychiatric disorders. Here we present data confirming the involvement of the PFC for the first time by means of functional magnetic resonance imaging (fMRI). To minimize motor-related activation, mental mazes were used.

METHODS

Mazes as well as pseudo-mazes without any bifurcations were presented to 49 healthy participants during fMRI scans.

RESULTS

Both, mazes as well as pseudo-mazes, activated a large network from visual to parietal regions, reflecting the dorsal stream of visual information processing. Mazes but not pseudo-mazes also activated bilateral areas of the PFC indicating their special role in decision processes. In addition, although no motor response was required during maze performance, both tasks activated subcortical and cortical motor areas.

CONCLUSIONS

These tasks are suitable for investigating and specifying PFC functioning and its impairment in psychiatric disorders such as schizophrenia. In addition, mental mazes might be a suitable task for the investigation of patients with motor disturbances.

Dynamics of Parietal Neural Activity during Spatial Cognitive Processing

Dynamic neural processing unrelated to changes in sensory input or motor output is likely to be a hallmark of cognitive operations. Here we show that neural representations of space in parietal cortex are dynamic while monkeys perform a spatial cognitive operation on a static visual stimulus. We recorded neural activity in area 7a during a visual maze task in which monkeys mentally followed a path without moving their eyes. We found that the direction of the followed path could be recovered from neuronal population activity. When the monkeys covertly processed a path that turned, the population representation of path direction shifted in the direction of the turn. This neural population dynamic took place during a period of unchanging visual input and showed characteristics of both serial and parallel processing. The data suggest that the dynamic evolution of parietal neuronal activity is associated with the progression of spatial cognitive operations.

Mental maze solving: directional fMRI tuning and population coding in the superior parietal lobule

The superior parietal lobule (SPL) of six human subjects was imaged at 4 T during mental traversing of a directed maze path. Here we demonstrate the orderly involvement of the SPL in this function, as follows. Forty-two percent of the voxels were tuned with respect to the direction of the maze path. This suggests a coherent tuning of local neuronal populations contributing to the change of the single-voxel BOLD signal. Preferred directions ranged throughout the directional continuum of 360 degrees. Voxels with similar preferred directions tended to cluster together: on average there were seven same-direction clusters per slice, with an average cluster membership of five voxels/cluster and an average nearest-neighbor same-direction intercluster distance of 13.1 mm. On the other hand, the average nearest-neighbor intercluster distance between a given direction and all other directions was 3.1 mm. This suggests a patchy arrangement such that patches of directionally tuned voxels, containing voxels with different preferred directions, alternate with patches of non-tuned voxels. Finally, the population vector predicted accurately the direction of the maze path (with an error of 12.7 degrees), and provided good estimates (with an error of 29 degrees) when calculated within parts of the SPL. Altogether, these findings document a new, orderly functional organization of the SPL with respect to mental tracing.

Visual information transfer across eye movements in the monkey

During normal viewing, the eyes move from one location to another in order to sample the visual environment. Information acquired before the eye movement facilitates post-saccadic processing. This "preview effect" indicates that some information is maintained in transsaccadic memory and combined with information acquired at the next fixation. However, the nature of transsaccadic memory remains a subject of debate. Here, we investigate preview effects in monkeys that carry out a contour-grouping (curve-tracing) task, by manipulating the consistency between pre- and post-saccadic information. The results show that consistent information causes a preview benefit, whereas inconsistent information causes a preview cost. These preview effects are relatively independent of the pre-saccadic viewing duration, and they occur even when the stimulus is exposed for only approximately 10 ms. The results further demonstrate that an entire relevant curve is stored in transsaccadic memory, instead of just the items at the saccade goal. They suggest that preview effects are caused by a mechanism that stores attended sensory information to make it available at the next fixation. The results are discussed within a theoretical framework that establishes an intimate relationship between attention, short-term memory and transsaccadic memory.

Correlates of transsaccadic integration in the primary visual cortex of the monkey

We make several eye movements per second when we explore a visual scene. Each eye movement sweeps the scene's projection across the retina and changes its representation in retinotopic areas of the visual cortex, but we nevertheless perceive a stable world. Here we investigate the neuronal correlates of visual stability in the primary visual cortex. Monkeys were trained to make two saccades along a single curve and to ignore another, distracting curve. Attention enhanced neuronal responses to the entire relevant curve before the first saccade. This response enhancement was rapidly reestablished after the saccade, although the image was shifted across the primary visual cortex. We argue that this fast postsaccadic restoration of the attentional response enhancement contributes to the stability of vision across eye movements, and reduces the impact of saccades on visual cognition.

Participation of primary motor cortical neurons in a distributed network during maze solution: representation of spatial parameters and time-course comparison with parietal area 7a

Traditionally, primary motor cortex (M1) has been thought to be involved solely in planning and generating movements. Recent evidence suggests that the arm area of M1 plays a role in other functions, such as the representation of serial order (Pellizzer et al. 1995, Science 269:702-705; Carpenter et al. 1999, Science 283:1752-1757) and spatial processing (Georgopoulos et al. 1989, Science 243:234-236). Previous studies of such cognitive processes have used tasks in which a directed arm movement was required, raising a question as to whether this brain area is involved in cognitive processing per se, or whether such cognitive signals may be gated into the arm area of M1 only when arm movements are required. To study this question, we developed a task that required a spatial analysis of a complex visual stimulus, but required no arm movement as a response. In this task, monkeys were shown an octagonal maze. After an imposed delay of 2 to 2.5 s, they indicated whether a path that emanated from the center of the maze exited at the perimeter (exit maze) or terminated within the maze (no-exit maze) by pressing a pedal with their left or right foot, respectively. We recorded from 785 cells from the arm area of M1 from two monkeys during the delay period of the maze task. We found that cell activity was influenced by both the exit status and the direction of the path, beginning soon after the maze was displayed. This activity was not related to the activation of arm muscles, suggesting that the directional signals observed represented abstract spatial aspects of maze processing. Finally, we compared maze-related activity of M1 neurons with those recorded from posterior parietal area 7a, reported previously (Crowe et al. 2004). Interestingly, cells from each area exhibited similar properties. Both the exit status and path direction were encoded by cells in M1 and 7a, although to different extents. An analysis of the time-course of the neural representation of these factors revealed that area 7a and M1 begin to encode these factors at the same time, suggesting these brain areas are part of a distributed system performing the spatial computations involved in maze solution.

Attention! V1 neurons lining up for inspection

In this issue of Neuron, Roelfsema and Spekreijse report that macaque V1 neuron responses are correlated with target choice in a task requiring monkeys to attentively trace a line to plan a saccade. These results provide evidence that V1 is actively involved in the interpretation of visual stimuli.