Pattern recognition in insects

A heuristic underlies the search for relief in Drosophila melanogaster

Humans rely on multiple types of sensory information to make decisions, and strategies that shorten decision‐making time by taking into account fewer but essential elements of information are preferred to strategies that require complex analyses. Such shortcuts to decision making are known as heuristics. The identification of heuristic principles in species phylogenetically distant to humans would shed light on the evolutionary origin of speed–accuracy trade‐offs and offer the possibility for investigating the brain representations of such trade‐offs, urgency and uncertainty. By performing experiments on spatial learning in the invertebrate Drosophila melanogaster, we show that the fly's search strategies conform to a spatial heuristic—the nearest neighbor rule—to avoid bitter taste (a negative stimulation). That is, Drosophila visits a salient location closest to its current position to stop the negative stimulation; only if this strategy proves unsuccessful does the fly use other learned associations to avoid bitter taste. Characterizing a heuristic in D. melanogaster supports the view that invertebrates can, when making choices, operate on economic principles, as well as the conclusion that heuristic decision making dates to at least 600 million years ago.

Vision in Drosophila: seeing the world through a model's eyes

The fruit fly, Drosophila melanogaster, has been used for decades as a genetic model for unraveling mechanisms of development and behavior. In order to efficiently assign gene functions to cellular and behavioral processes, early measures were often necessarily simple. Much of what is known of developmental pathways was based on disrupting highly regular structures, such as patterns of cells in the eye. Similarly, reliable visual behaviors such as phototaxis and motion responses provided a solid foundation for dissecting vision. Researchers have recently begun to examine how this model organism responds to more complex or naturalistic stimuli by designing novel paradigms that more closely mimic visual behavior in the wild. Alongside these advances, the development of brain-recording strategies allied with novel genetic tools has brought about a new era of Drosophila vision research where neuronal activity can be related to behavior in the natural world.

Visual artificial grammar learning: comparative research on humans, kea (Nestor notabilis) and pigeons (Columba livia)

Artificial grammar learning (AGL) provides a useful tool for exploring rule learning strategies linked to general purpose pattern perception. To be able to directly compare performance of humans with other species with different memory capacities, we developed an AGL task in the visual domain. Presenting entire visual patterns simultaneously instead of sequentially minimizes the amount of required working memory. This approach allowed us to evaluate performance levels of two bird species, kea (Nestor notabilis) and pigeons (Columba livia), in direct comparison to human participants. After being trained to discriminate between two types of visual patterns generated by rules at different levels of computational complexity and presented on a computer screen, birds and humans received further training with a series of novel stimuli that followed the same rules, but differed in various visual features from the training stimuli. Most avian and all human subjects continued to perform well above chance during this initial generalization phase, suggesting that they were able to generalize learned rules to novel stimuli. However, detailed testing with stimuli that violated the intended rules regarding the exact number of stimulus elements indicates that neither bird species was able to successfully acquire the intended pattern rule. Our data suggest that, in contrast to humans, these birds were unable to master a simple rule above the finite-state level, even with simultaneous item presentation and despite intensive training.

A comparative view of face perception

Face perception serves as the basis for much of human social exchange. Diverse information can be extracted about an individual from a single glance at their face, including their identity, emotional state, and direction of attention. Neuropsychological and functional magnetic resonance imaging (fMRI) experiments reveal a complex network of specialized areas in the human brain supporting these face-reading skills. Here we consider the evolutionary roots of human face perception by exploring the manner in which different animal species view and respond to faces. We focus on behavioral experiments collected from both primates and nonprimates, assessing the types of information that animals are able to extract from the faces of their conspecifics, human experimenters, and natural predators. These experiments reveal that faces are an important category of visual stimuli for animals in all major vertebrate taxa, possibly reflecting the early emergence of neural specialization for faces in vertebrate evolution. At the same time, some aspects of facial perception are only evident in primates and a few other social mammals, and may therefore have evolved to suit the needs of complex social communication. Because the human brain likely utilizes both primitive and recently evolved neural specializations for the processing of faces, comparative studies may hold the key to understanding how these parallel circuits emerged during human evolution.

Detection of patches of coloured discs by bees

To find out how grouping of flowers into patches improves their detectability by hymenopteran pollinators, we trained honeybees and bumblebees to detect groups of three spatially separated disks and compared results with the detection limit for single disks. When the discs presented contrast to the long-wavelength-sensitive (L) receptor, grouping of disks improved the detectability. The disks were optically resolvable for the honeybee eye. The improvement of detectability was stronger for bumblebees than for honeybees. When disks did not present contrast to the L-receptor, the grouping did not improve the detectability, i.e. the detection limit was set by the size of a single disk. We conclude that in bees the neural mechanisms that improve detectability of grouped elements require input from the L-receptor. Our results indicate that grouping of flowers into sparse patches can improve their detectability by bees, even when individual flowers can be optically resolved by the eyes of bees, as long as flowers can be detected by the long-wavelength-sensitive receptor.

Cracking neural circuits in a tiny brain: new approaches for understanding the neural circuitry of Drosophila

Genetic screens in Drosophila have identified many genes involved in neural development and function. However, until recently, it has been impossible to monitor neural signals in Drosophila central neurons, and it has been difficult to make specific perturbations to central neural circuits. This has changed in the past few years with the development of new tools for measuring and manipulating neural activity in the fly. Here we review how these new tools enable novel conceptual approaches to 'cracking circuits' in this important model organism. We discuss recent studies aimed at defining the cognitive demands on the fly brain, identifying the cellular components of specific neural circuits, mapping functional connectivity in those circuits and defining causal relationships between neural activity and behavior.

Different parameters support generalization and discrimination learning in Drosophila at the flight simulator

We have used a genetically tractable model system, the fruit fly Drosophila melanogaster to study the interdependence between sensory processing and associative processing on learning performance. We investigated the influence of variations in the physical and predictive properties of color stimuli in several different operant-conditioning procedures on the subsequent learning performance. These procedures included context and stimulus generalization as well as color, compound, and conditional discrimination (colors and patterns). A surprisingly complex dependence of the learning performance on the colors' physical and predictive properties emerged, which was clarified by taking into account the fly-subjective perception of the color stimuli. Based on estimates of the stimuli's color and brightness values, we propose that the different tasks are supported by different parameters of the color stimuli; generalization occurs only if the chromaticity is sufficiently similar, whereas discrimination learning relies on brightness differences.

Insect vision: remembering the shape of things

How does the nervous system store a newly experienced visual pattern, and how is that pattern subsequently made available for recognition? Recent work in Drosophila suggests that specific pattern features are stored separately in the nervous system.

Recognition of a familiar place by the honeybee (Apis mellifera)

Recent work shows that at any one place bees detect a limited variety of simple cues in parallel. At each choice point, they recognize a few cues in the range of positions where the cues occurred during the learning process. There is no need to postulate that they re-assemble the surrounding panorama in memory; only that they retain memories of the coincidences of cues in the expected retinotopic directions. The cues could be stimuli that excite groups of peripheral visual neurons. All the experimentally known cues are described, including modulation of the receptors, the locations of areas of black or colour, the nearness, size, averaged edge orientation, and radial and tangential edges. Cues of each type are separately summed within large fields, the size of which varies with the cue. Local orientation cues from edges at right angles cancel each other within each field, which also suggests that the discrimination of shape and texture is limited. Resolution depends on lateral interactions and the number of ommatidia required for each cue. To identify a new place, a few sparse cues, together with their directions, are learned in orientation flights. When the bee returns, the cues in the panorama are progressively matched as they coincide with the cues in memory. The limited number of cues, though economical for memory, may restrict the foraging behaviour and lead to flower constancy. This kind of a visual system is a candidate model for other animals or machines with economical processing systems.

Visual pattern recognition in Drosophila is invariant for retinal position

Vision relies on constancy mechanisms. Yet, these are little understood, because they are difficult to investigate in freely moving organisms. One such mechanism, translation invariance, enables organisms to recognize visual patterns independent of the region of their visual field where they had originally seen them. Tethered flies (Drosophila melanogaster) in a flight simulator can recognize visual patterns. Because their eyes are fixed in space and patterns can be displayed in defined parts of their visual field, they can be tested for translation invariance. Here, we show that flies recognize patterns at retinal positions where the patterns had not been presented before.

Dissociation of visual associative and motor learning in Drosophila at the flight simulator

Ever since operant conditioning was studied experimentally, the relationship between associative learning and possible motor learning has become controversial. Although motor learning and its underlying neural substrates have been extensively studied in mammals, it is still poorly understood in invertebrates. The visual discriminative avoidance paradigm of Drosophila at the flight simulator has been widely used to study the flies' visual associative learning and related functions, but it has not been used to study the motor learning process. In this study, newly-designed data analysis was employed to examine the flies' solitary behavioural variable that was recorded at the flight simulator-yaw torque. Analysis was conducted to explore torque distributions of both wild-type and mutant flies in conditioning, with the following results: (1) Wild-type Canton-S flies had motor learning performance in conditioning, which was proved by modifications of the animal's behavioural mode in conditioning. (2) Repetition of training improved the motor learning performance of wild-type Canton-S flies. (3) Although mutant dunce(1) flies were defective in visual associative learning, they showed essentially normal motor learning performance in terms of yaw torque distribution in conditioning. Finally, we tentatively proposed that both visual associative learning and motor learning were involved in the visual operant conditioning of Drosophila at the flight simulator, that the two learning forms could be dissociated and they might have different neural bases.

Catchment areas of panoramic snapshots in outdoor scenes

We took panoramic snapshots in outdoor scenes at regular intervals in two- or three-dimensional grids covering 1 m2 or 1 m3 and determined how the root mean square pixel differences between each of the images and a reference image acquired at one of the locations in the grid develop over distance from the reference position. We then asked whether the reference position can be pinpointed from a random starting position by moving the panoramic imaging device in such a way that the image differences relative to the reference image are minimized. We find that on time scales of minutes to hours, outdoor locations are accurately defined by a clear, sharp minimum in a smooth three-dimensional (3D) volume of image differences (the 3D difference function). 3D difference functions depend on the spatial-frequency content of natural scenes and on the spatial layout of objects therein. They become steeper in the vicinity of dominant objects. Their shape and smoothness, however, are affected by changes in illumination and shadows. The difference functions generated by rotation are similar in shape to those generated by translation, but their plateau values are higher. Rotational difference functions change little with distance from the reference location. Simple gradient descent methods are surprisingly successful in recovering a goal location, even if faced with transient changes in illumination. Our results show that view-based homing with panoramic images is in principle feasible in natural environments and does not require the identification of individual landmarks. We discuss the relevance of our findings to the study of robot and insect homing.

Visual signals of individual identity in the wasp Polistes fuscatus

Individual recognition is an essential component of interactions in many social systems, but insects are often thought incapable of the sophistication necessary to recognize individuals. If this were true, it would impose limits on the societies that insects could form. For example, queens and workers of the paper wasp Polistes fuscatus form a linear dominance hierarchy that determines how food, work and reproduction are divided within the colony. Such a stable hierarchy would be facilitated if individuals of different ranks have some degree of recognition. P. fuscatus wasps have, to our knowledge, previously undocumented variability in their yellow facial and abdominal markings that are intriguing candidates for signals of individual identity. Here, I describe these highly variable markings and experimentally test whether P. fuscatus queens and workers use these markings to identify individual nest-mates visually. I demonstrate that individuals whose yellow markings are experimentally altered with paint receive more aggression than control wasps who are painted in a way that does not alter their markings. Further, aggression declines towards wasps with experimentally altered markings as these novel markings become familiar to their nestmates. This evidence for individual recognition in P. fuscatus indicates that interactions between insects may be even more complex than previously anticipated.

Insects could exploit UV-green contrast for Landmark navigation

Illumination-invariant detection of landmark features is a prerequisite for landmark navigation in insects. It is suggested that a contrast mechanism involving the UV and green receptors of insect eyes could guarantee a robust separation between natural objects as foreground and sky as background. Using a sensor with a UV and a green channel that in their spectral characteristics are close to the corresponding insect photoreceptors, data of natural objects and sky were collected. The data show that the two classes can be separated by a fixed threshold in the UV-green color space, offering an advantage over a purely UV-based separation that would require a dynamic threshold. Based on a numerical method, UV-green antagonism is shown to guarantee a more reliable discrimination than UV-blue antagonism.

Template-matching describes visual pattern-recognition tasks in the weakly electric fishGnathonemus petersii

Several insects use template-matching systems to recognize objects or environmental landmarks by comparing actual and stored retinal images. Such systems are not viewpoint-invariant and are useful only when the locations in which the images have been stored and where they are later retrieved coincide. Here, we describe that a vertebrate, the weakly electric fish Gnathonemus petersii, appears to use template-matching to recognize visual patterns that it had previously viewed from a fixed vantage point. This fish is nocturnal and uses its electrical sense to find its way in the dark, yet it has functional vision that appears to be well adapted to dim light conditions. We were able to train three fish in a two-alternative forced-choice procedure to discriminate a rewarded from an unrewarded visual pattern. From its daytime shelter, each fish viewed two visual patterns placed at a set distance behind a transparent Plexiglas screen that closed the shelter. When the screen was lifted, the fish swam towards one of the patterns to receive a food reward or to be directed back into its shelter. Successful pattern discrimination was limited to low ambient light intensities of approximately 10 lx and to pattern sizes subtending a visual angle greater than 3°. To analyze the characteristics used by the fish to discriminate the visual training patterns, we performed transfer tests in which the training patterns were replaced by other patterns. The results of all such transfer tests can best be explained by a template-matching mechanism in which the fish stores the view of the rewarded training pattern and chooses from two other patterns the one whose retinal appearance best matches the stored view.

Discrimination of closed shapes by two species of bee, Apis mellifera and Megachile rotundata

In the present study, the performance of two bee species, the honeybee Apis mellifera and the leaf-cutter bee Megachile rotundata, in discriminating among various closed (convex) shapes was examined systematically for the first time. Bees were trained to each of five different shapes, a disc, a square, a diamond and two different triangles, all of the same area, using fresh bees in each experiment. In subsequent tests, the trained bees were given a choice between the learned shape and each of the other four shapes. Two sets of experiments were conducted with both species. In the first, solid black shapes were presented against a white background, thus providing a high luminance contrast. In the second, the shapes carried a random black-and-white pattern and were presented 5 cm in front of a similar pattern, thus producing motion contrast, rather than luminance contrast, against the background.

The results obtained with the solid shapes reveal that both bee species accomplish the discrimination, although the performance of the honeybee is significantly better than that of the leaf-cutter bee. Furthermore, the effectiveness of the various shapes differs between the two species. However, in neither species is the discrimination performance correlated with the amount of overlap of the black areas contained in the various pairs of shapes, suggesting that, in our experiments, shape discrimination is not based on a template-matching process. We propose that it is based on the use of local parameters situated at the outline of the shape, such as the position of angles or acute points and, in particular, the position and orientation of edges. This conclusion is supported by the finding that bees of both species accomplish the discrimination even with the patterned shapes. These shapes are visible only because of the discontinuity of the speed of image motion perceived at the edge between the shape and the background.

Do insects use templates or parameters for landmark navigation?

As an explanation for the landmark navigation abilities of insects, the template hypothesis is currently clearly dominating the parameter hypothesis. In a series of computer simulations, this paper investigates the conspicuous behavior of desert ants in experimental setups with expanded landmark arrays. It is shown that a parameter-based model accounts much better for the data than different template models.

Evidence against a retinotopic-template matching in honeybees' pattern recognition

Currently two hypotheses exist as to how insects process visual images, as photograph-like 'retinotopic-templates', or as a set of features extracted by the visual system. Several results obtained in honeybees cannot be reconciled with a retinotopic-template matching. (i) Bees discriminated between two patterns that should not be distinguished according to the template hypothesis. (ii) Bees preferred patterns that showed no overlap with the assumed template to patterns that had such an overlap. (iii) Bees showed a generalization of properties of the rewarded pattern to other patterns. Thus, in our paradigm, the bees must have used additional mechanisms and cues for the processing and classification of patterns.

Orientation discrimination independent of retinal matching by blowflies

Blowflies, Phaenicia sericata, can be trained to discriminate in a learning paradigm in which one of the two visual cues is positively rewarded. Retinotopic matching of a learned visual image to the same retinal location from viewing to viewing has been hypothesized to underlie visual pattern learning and memory in insects. To address the theory of retinotopic matching, a detailed analysis was made of the flies' body orientations during learned discriminations between +45 degrees and −45 degrees gratings. Initial approaches to the positive rewarded visual cue did not originate from the same spatial location within the behavioral arena with respect to the visual cues; thus, individual flies approached the positive cue from a different vantage point from trial to trial. During initial approaches to the rewarded visual cue, the distributions of body angles with respect to the cue were different from trial to trial for each individual. These data suggest that Phaenicia sericata can learn a visual pattern with one eye region and later recognize the same pattern with another eye region. Thus, retinotopic matching is not necessary for the recognition of pattern orientation in the experimental paradigm used here. The average amount of head turning in the yaw plane was too small to compensate for the changes in body orientation exhibited by the flies. Flies view the visual patterns with distinct retinal regions from trial to trial during orientation discrimination.

Flexibility in a Single Behavioral Variable of Drosophila

The flexibility of behavior is so rich, and its components are so exquisitely interwoven, that one may be well advised to turn to an isolated behavioral module for study. Gill withdrawal inAplysia, the proboscis extension reflex in the honeybee, and lid closure in mammals are such examples. We have chosen yawing, a single component of flight orientation in Drosophila melanogaster, for this approach. A specialty of this preparation is that the behavioral output can be reduced beyond the single module by one further step. It can be studied in tethered animals in which all turns are blocked while the differentially beating wings still provide the momentum. These intended yaw turns are measured by a torque meter to which the fly is hooked. The fly is held horizontally as if cruising at high speed. The head is glued to the thorax. It can bend its abdomen, extend its proboscis, and move its legs but cannot shift its direction of gaze or its orientation in space. Evidently, a fly hardly ever encounters this bizarre situation in the wild. We describe here the flexibility in this single behavioral variable. It provides insights into the relation between classical and operant conditioning, the processing of and interactions between the conditioned visual stimuli, early visual memory, visual pattern recognition, selective attention, and several other experience-dependent properties of visual orientation behavior. We start with a brief summary of visual flight control at the torque meter.

The operant and the classical in conditioned orientation of Drosophila melanogaster at the flight simulator

Ever since learning and memory have been studied experimentally, the relationship between operant and classical conditioning has been controversial. Operant conditioning is any form of conditioning that essentially depends on the animal's behavior. It relies on operant behavior. A motor output is called operant if it controls a sensory variable. The Drosophila flight simulator, in which the relevant behavior is a single motor variable (yaw torque), fully separates the operant and classical components of a complex conditioning task. In this paradigm a tethered fly learns, operantly or classically, to prefer and avoid certain flight orientations in relation to the surrounding panorama. Yaw torque is recorded and, in the operant mode, controls the panorama. Using a yoked control, we show that classical pattern learning necessitates more extensive training than operant pattern learning. We compare in detail the microstructure of yaw torque after classical and operant training but find no evidence for acquired behavioral traits after operant conditioning that might explain this difference. We therefore conclude that the operant behavior has a facilitating effect on the classical training. In addition, we show that an operantly learned stimulus is successfully transferred from the behavior of the training to a different behavior. This result unequivocally demonstrates that during operant conditioning classical associations can be formed.

The memory template in Drosophila pattern vision at the flight simulator

Pattern recognition is studied in flight orientation of fixed flying Drosophila melanogaster controlling the horizontal rotations of an arena. Earlier experiments had suggested a simple mechanism of pattern recognition in which a memory template and the actual image are retinotopically matched. In contrast, we now show that Drosophila extracts at least two and probably four pattern parameters: size, vertical position of the center of gravity and, presumably horizontal/vertical extent as well as vertical separatedness of pattern elements. Moreover, the fly treats isolated pattern elements as a compound figure. Retinal transfer is possible between training and test if the centers of gravity of the compound figures are retained.

Limited translation invariance of human visual pattern recognition

Visual object recognition is considered to be largely translation invariant. An earlier study (Foster & Kahn, 1985), however, has indicated that recognition of complex novel stimuli is partially specific to location in the visual field: It is significantly easier to determine the identity of two briefly displayed random patterns if both stimuli are presented at the same, rather than at different, locations. In a series of same/different discrimination tasks, we characterize the processes underlying this "displacement effect": Horizontal and vertical translations are equally effective in reducing performance. Making the task more difficult by increasing pattern similarity leads to even higher positional specificity. The displacement effect disappears after rotation or contrast reversal of the patterns, indicating that positional specificity depends on relatively low levels of processing. Control experiments rule out explanations that are independent of visual pattern memory, such as spatial attention, eye movements, or retinal afterimages. Positional specificity of recognition is found only for same trials. Our results demonstrate that position invariance, a widely acknowledged property of the human visual system, is limited to specific experimental conditions. Normalization models involving mental shifts of an early visual representation or of a window of attention cannot easily account for these findings.

Insect visual perception: complex abilities of simple nervous systems

Despite their relatively simple nervous systems, insects display a rich behavioural repertoire, in which vision plays a major role. In the past two years, much knowledge has been gained about how insects are capable of a variety of flexible, visually guided tasks that involve a high level of complexity. From long-range navigation to median-range orientation and close-up recognition, insects apply different strategies that complement each other, that are used sequentially during their approach flight towards their goals, and that may replace each other, depending on the salience of, and the attention towards, particular visual cues.