Learning to recognize objects

Abstract - A theory of object recognition requires a theory of shape. Despite considerable empirical and theoretical research, however, a definition of object shape has proved elusive. Two experiments provide new insights by showing that children's object recognition changes dramatically during the period between 17 and 25 months. During this time, children develop the ability to recognize stylized three-dimensional caricatures of known and novel objects. This ability is linked to the number of object names in children's vocabularies, suggesting that category learning may be a driving force behind the developmental changes.     

Learning to see faces and objects

Visual recognition of objects is an impressively difficult problem that biological systems solve effortlessly. We consider two aspects of this ability. First, is the recognition of all objects accomplished by either a single system or multiple, domain-specific systems? Behavioral, neuropsychological and neuroimaging data indicate that a single system is sufficient for the recognition of all objects at all levels. Second, how does such a system 'tune' itself to the constraints imposed by recognition at different levels of specificity? Evidence indicates that the task demands and learning that arise from different forms of feedback determine which computational routines are recruited automatically in object recognition.     

Learning of lists of properties related to familiar objects

Four groups of Ss simultaneously learn six-word lists that indicate common objects' properties. Words of each list may either all correspond to the same object (highest homogeneity level), or, inversely, correspond to six different objects (highest heterogeneity level), with two intermediate situations in between.The homogeneity of lists has a main effect upon learning. It is hypothesized that the object corresponding to the words of a list is used, when it is detected, as a federator for this words, and thus facilitates their recall. It is concluded there is a certain similarity between word lists corresponding to several properties of an object and taxonomic lists.     

Rotating objects to recognize them: A case study on the role of viewpoint dependency in the recognition of three-dimensional objects

Successful object recognition is essential for finding food, identifying kin, and avoiding danger, as well as many other adaptive behaviors. To accomplish this feat, the visual system must reconstruct 3-D interpretations from 2-D “snapshots” falling on the retina. Theories of recognition address this process by focusing on the question of how object representations are encoded with respect to viewpoint. Although empirical evidence has been equivocal on this question, a growing body of surprising results, including those obtained in the experiments presented in this case study, indicates that recognition is often viewpoint dependent. Such findings reveal a prominent role for viewpointdependent mechanisms and provide support for themultiple-views approach, in which objects are encoded as a set of view-specific representations that are matched to percepts using normalization procedures.     

Learning to recognize talkers from natural,sinewave, and reversed speech samples

In 5 experiments, the authors investigated how listeners learn to recognize unfamiliar talkers and how experience with specific utterances generalizes to novel instances. Listeners were trained over several days to identify 10 talkers from natural, sinewave, or reversed speech sentences. The sinewave signals preserved phonetic and some suprasegmental properties while eliminating natural vocal quality. In contrast, the reversed speech signals preserved vocal quality while distorting temporally based phonetic properties. The training results indicate that listeners learned to identify talkers even from acoustic signals lacking natural vocal quality. Generalization performance varied across the different signals and depended on the salience of phonetic information. The results suggest similarities in the phonetic attributes underlying talker recognition and phonetic perception.     

The ability to recognize objects from bottlenose dolphin (Tursiops truncatus) echoes generalizes across multiple orientations in humans and neural networks

Object constancy, the ability to recognize objects despite changes in orientation, has not been well studied in the auditory modality. Dolphins use echolocation for object recognition, and objects ensonified by dolphins produce echoes that can vary significantly as a function of orientation. In this experiment, human listeners had to classify echoes from objects varying in material, shape, and size that were ensonified with dolphin signals. Participants were trained to discriminate among the objects using an 18-echo stimulus from a 10° range of aspect angles, then tested with novel aspect angles across a 60° range. Participants were typically successful recognizing the objects at all angles (M = 78 %). Artificial neural networks were trained and tested with the same stimuli with the purpose of identifying acoustic cues that enable object recognition. A multilayer perceptron performed similarly to the humans and revealed that recognition was enabled by both the amplitude and frequency of echoes, as well as the temporal dynamics of these features over the course of echo trains. These results provide insight into representational processes underlying echoic recognition in dolphins and suggest that object constancy perceived through the auditory modality is likely to parallel what has been found in the visual domain in studies with both humans and animals.     

Learning to use novel objects: a training study on the acquisition of novel action representations

Many studies have suggested that the motor system is organized in a hierarchical fashion, around the prototypical end location associated with using objects. However, most studies supporting the hierarchical view have used well-known actions and objects that are highly over-learned. Accordingly, at present it is unclear if the hierarchical principle applies to learning the use of novel objects as well. In the present study we found that when learning to use a novel object subjects acquired an action representation of the end location associated with using the object, as evidenced by slower responses in an action observation task, when the object was presented at an incorrect end location. By showing the importance of knowledge about end locations when learning to use a novel object, the present study suggests that end locations are a fundamental organizing feature of the human motor system.     

Why people fail to recognize their own incompetence

Successful negotiation of everyday life would seem to require people to possess insight about deficiencies in their intellectual and social skills. However, people tend to be blissfully unaware of their incompetence. This lack of awareness arises because poor performers are doubly cursed: Their lack of skill deprives them not only of the ability to produce correct responses, but also of the expertise necessary to surmise that they are not producing them. People base their perceptions of performance, in part, on their preconceived notions about their skills. Because these notions often do not correlate with objective performance, they can lead people to make judgments about their performance that have little to do with actual accomplishment.     

Great apes’ capacities to recognize relational similarity

Recognizing relational similarity relies on the ability to understand that defining object properties might not lie in the objects individually, but in the relations of the properties of various object to each other. This aptitude is highly relevant for many important human skills such as language, reasoning, categorization and understanding analogy and metaphor. In the current study, we investigated the ability to recognize relational similarities by testing five species of great apes, including human children in a spatial task. We found that all species performed better if related elements are connected by logico-causal as opposed to non-causal relations. Further, we find that only children above 4 years of age, bonobos and chimpanzees, unlike younger children, gorillas and orangutans display some mastery of reasoning by non-causal relational similarity. We conclude that recognizing relational similarity is not in its entirety unique to the human species. The lack of a capability for language does not prohibit recognition of simple relational similarities. The data are discussed in the light of the phylogenetic tree of relatedness of the great apes.     

Learning new names for new objects: cortical effects as measured by magnetoencephalography

We tracked the evolvement of naming-related cortical dynamics with magnetoencephalography when five normal adults successfully learned names and/or meanings of unfamiliar objects. In all subjects, the learning of new names was associated with pronounced cortical effects. The learning effect was of long latency and emerged as a change of activation in the same cortical network that was active during naming of familiar items. In four out of five subjects, the cortical learning effect occurred in the inferior parietal lobe. In three of these subjects, the cortical effect was left-sided. These results suggest that the inferior parietal lobe plays an important role in the acquisition of novel words, presumably as a part of working memory systems.     

Attentional prioritization to contextually new objects

Using a hybrid paradigm of a contextual cuing task and a probe detection task, we tested whether or not contextually new objects can be prioritized in visual search. After several hundred visual search practice trials with repeated and nonrepeated layouts, an additional distractor that was contextually new was presented in the repeated layouts. The results showed that detection of probe dots appearing at the location of a search target was faster in the repeated than in the nonrepeated layouts. More importantly, detection of probe dots at the new object locations was as fast as that of probe dots at the target locations in repeated layouts, suggesting that the attentional system implicitly prioritizes the processing of a location where a change in contextual information has occurred.     

Inhibition of return to occluded objects

Since many visual objects are vulnerable to occlusion, an active process that tracks objects behind occluders confers considerable ecological validity to the visual system. We studied this possibility by testing whether inhibition of return can be observed with occluded objects. In our experiments, two moving objects disappeared or reappeared behind occluders while a cue and a probe were presented. Contrary to the results of a previous study (Tipper, Weaver, Jerreat, & Burak, 1994), responses were consistently delayed for the cued object that was occluded when it was cued (Experiment 1), when it was probed (Experiment 2), or both (Experiment 3). These results suggest that attention can select occluded objects that are out of view. Our findings are in line with prior studies that have demonstrated similar perceptual/attentional effects for occluded objects, as well as for visible objects.     

Rigid objects that appear to bend

Simple rigid objects are presented that appear to bend when viewed from certain angles. These illusions illustrate that perspective information is used by the stereo system, that projective distortions can override rigidity constraints in motion perception, and that touch only corrects the illusion for a local region.     

Sensitivity to viewpoint in children's drawings of objects and relations between objects.

Previous literature suggests that young children are relatively insensitive to viewpoint, only showing their view when the task is manipulated to provoke it. In contrast, older children appear to become more sensitive to viewpoint and it has been claimed that there is a developmental progression toward use of linear perspective as a means of drawing a view-specific scene. This study investigates sensitivity to viewpoint by manipulating it directly. Children between the ages of 6 and 14 years were asked to draw an L-shaped array of three cubes from one of three views: frontal eye level, frontal looking down, and corner looking down. At every age children showed sensitivity to their view in the sense that there were consistent differences between the drawings produced in the three viewing conditions. In the case of younger children this did not lead to an accurate portrayal of either their view or the array relations. Older children portrayed their view and the array relations more accurately, and viewpoint had a strong effect on the choice of projection system both within and between objects. There was no evidence of a general progression toward use of linear perspective.     

Young children's ability to recognize advertisements in web page designs

Identifying what is, and what is not an advertisement is the first step in realizing that an advertisement is a marketing message. Children can distinguish television advertisements from programmes by about 5 years of age. Although previous researchers have investigated television advertising, little attention has been given to advertisements in other media, even though other media, especially the Internet, have become important channels of marketing to children. We showed children printed copies of invented web pages that included advertisements, half of which had price information, and asked the children to point to whatever they thought was an advertisement. In two experiments we tested a total of 401 children, aged 6, 8, 10 and 12 years of age, from the United Kingdom and Indonesia. Six‐year‐olds recognized a quarter of the advertisements, 8‐year‐olds recognized half the advertisements, and the 10‐ and 12‐year‐olds recognized about three‐quarters. Only the 10‐ and 12‐year‐olds were more likely to identify an advertisement when it included a price. We contrast our findings with previous results about the identification of television advertising, and discuss why children were poorer at recognizing web page advertisements. The performance of the children has implications for theories about how children develop an understanding of advertising.     

Privileged access to action for objects relative to words

We compared action (pour or twist?) and contextual/semantic (found in kitchen?) decisions made to pictures of objects, nonobjects, and words. Although there was no advantage for objects over words in contextual/semantic decisions, there was an advantage for objects over words and nonobjects in action decisions. For objects, both action and contextual/semantic decisions were faster than naming; for words, the opposite occurred. These results extend the early results of Potter and Faulconer (1975) that there is privileged access to semantic memory for objects relative to that for words and privileged access to phonology for words. Our data suggest that, for objects, there is privileged access to action knowledge rather than to all forms of semantic knowledge and that this is contingent on learned associations between objects and actions.     

Infants' ability to use luminance information to individuate objects

Recent research indicates that infants first use form and then surface features as the basis for individuating objects. However, very little is known about the underlying basis for infants' differential sensitivity to form than surface features. The present research assessed infants' sensitivity to luminance differences. Like other surface properties, luminance information typically reveals little about an object. Unlike other surface properties (e.g. pattern, color), the visual system can detect luminance differences at birth. The outcome of two experiments indicated that 11.5-month-olds, but not 7.5-month-olds, used luminance differences to individuate objects. These results suggest that it is not the age at which infants can detect a feature, but the nature of the information carried by the feature, that determines infants' capacity to individuate objects.