Identification of everyday objects on the basis of fragmented outline versions

Although Attneave (1954 Psychological Review 61 183 193) and Biederman (1987 Psychological Review 94 115-147) have argued that curved contour segments are most important in shape perception, Kennedy and Domander (1985 Perception 14 367-370) showed that fragmented object contours are better identifiable when straight segments are shown. We used the set of line drawings published by Snodgrass and Vanderwart (1980 Journal of Experimental Psychology: Human Learning and Memory 6 174-215), to make outline versions that could be used to investigate this issue with a larger and more heterogeneous stimulus set. Fragments were placed either around the 'salient' points or around the midpoints (points midway between two salient points), creating curved versus relatively straight fragments when the original outline was fragmented (experiment 1), or angular and straight fragments when straight-line versions were fragmented (experiment 2). We manipulated fragment length in each experiment except the last one, in which we presented only selected points (experiment 3). While fragmented versions were on average more identifiable when straight fragments were shown, certain objects were more identifiable when the curved segments or the angles were shown. A tentative explanation of these results is presented in terms of an advantage for straight segments during grouping processes for outlines with high part salience, and an advantage for curved segments during matching processes for outlines with low part salience.     

Discrimination of contour-deleted images by pigeons.

Three experiments attempted to determine which properties of pictorial representations of objects control their discrimination by pigeons. A particular focus was whether the representation mediating such discriminations could be described by the simple viewpoint-invariant primitive volumes of Biederman's (1987) recognition-by-components theory of object recognition or by Cerella's (1990) particulate features. In all 3 experiments, pigeons were first trained to discriminate drawings of 4 stimulus objects with half of the contour deleted but with the component geons postulated by Biederman's theory recoverable. Discrimination accuracy was then compared for test items containing the original particulate features, affording the retrieval of the original component geons, or having neither of these properties of the training stimuli. Although response accuracy was significantly greater when the component geons of the original objects were retrievable, measurable control over recognition by the particulate features of the objects and by their specific locations was also found. The results are consistent with the idea of component geon recognition as one of the important factors in object discrimination.     

Contour deletion as a method for identifying the weights of features underlying object recognition

A prerequisite for comparative work on object recognition is a method for identifying the features actually extracted from the form. The method introduced here with pigeons is discrimination training between two simple line drawings, followed by a generalization test in which contour is deleted from the reinforced drawing. In Condition 1, the line drawings were a square (S+) versus a triangle (S-); for Condition 2, the line drawings were planar projections of a cube (S+) versus a truncated pyramid (S-). The generalization decrement between responses to S+ and responses to test stimuli provides a quantitative index of the weight assigned to each feature. Contour deletion at either vertices or midsegments produced a decrement in the rate of responding, showing that each contour was represented as a feature. The generalization decrement to forms containing vertices with midsegments deleted was larger than the generalization decrement to forms containing midsegments with vertices deleted. Therefore, it appears that midsegments are weighted more strongly as features than vertices. Contour deletion provides a direct method for identifying the visual features underlying object recognition and lays a foundation for the development of comparative theories of object recognition.     

Incomplete figure perception and invisible masking

The Gollin test (measuring recognition thresholds for fragmented line drawings of everyday objects and animals) has traditionally been regarded as a test of incomplete figure perception or 'closure', though there is a debate about how such closure is achieved. Here, figural incompleteness is considered to be the result of masking, such that absence of contour elements of a fragmented figure is the result of the influence of an 'invisible' mask. It is as though the figure is partly obscured by a mask having parameters identical to those of the background. This mask is 'invisible' only consciously, but for the early stages of visual processing it is real and has properties of multiplicative noise. Incomplete Gollin figures were modeled as the figure covered by the mask with randomly distributed transparent and opaque patches. We adjusted the statistical characteristics of the contour image and empty noise patches and processed those using spatial and spatial-frequency measures. Across 73 figures, despite inter-subject variability, mean recognition threshold was always approximately 15% of total contour in naive observers. Recognition worsened with increasing spectral similarity between the figure and the 'invisible' mask. Near threshold, the spectrum of the fragmented image was equally similar to that of the 'invisible' mask and complete image. The correlation between spectral parameters of figures at threshold and complete figures was greatest for figures that were most easily recognised. Across test sessions, thresholds reduced when either figure or mask parameters were familiar. We argue that recognition thresholds for Gollin stimuli in part reflect the extraction of signal from noise.     

Contour-based object identification and segmentation: stimuli, norms and data, and software tools.

We summarize five studies of our large-scale research program, in which we examined aspects of contour-based object identification and segmentation, and we report on the stimuli we used, the norms and data we collected, and the software tools we developed. The stimuli were outlines derived from the standard set of line drawings of everyday objects by Snodgrass and Vanderwart (1980). We used contour curvature as a major variable in all the studies. The total number of 1,500 participants produced very solid, normative identification rates of silhouettes and contours, straight-line versions, and fragmented versions, and quite reliable benchmark data about saliency of points and object segmentation into parts. We also developed several software tools to generate stimuli and to analyze the data in nonstandard ways. Our stimuli, norms and data, and software tools have great potential for further exploration of factors influencing contour-based object identification, and are also useful for researchers in many different disciplines (including computer vision) on a wide variety of research topics (e.g., priming, agnosia, perceptual organization, and picture naming). The full set of norms, data, and stimuli may be downloaded from www.psychonomic.org/archive/.     

Contour-based object identification and segmentation: Stimuli,norms and data,and software tools

We summarize five studies of our large-scale research program, in which we examined aspects of contour-based object identification and segmentation, and we report on the stimuli we used, the norms and data we collected, and the software tools we developed. The stimuli were outlines derived from the standard set of line drawings of everyday objects by Snodgrass and Vanderwart (1980). We used contour curvature as a major variable in all the studies. The total number of 1,500 participants produced very solid, normative identification rates of silhouettes and contours, straight-line versions, and fragmented versions, and quite reliable benchmark data about saliency of points and object segmentation into parts. We also developed several software tools to generate stimuli and to analyze the data in nonstandard ways. Our stimuli, norms and data, and software tools have great potential for further exploration of factors influencing contour-based object identification, and are also useful for researchers in many different disciplines (including computer vision) on a wide variety of research topics (e.g., priming, agnosia, perceptual organization, and picture naming). The full set of norms, data, and stimuli may be downloaded fromwww.psychonomic.org/archive/.     

Spatiotemporal integration and contour interpolation revealed by a dot localization task with serial presentation paradigm

The visual system seems to integrate information that is presented over time in a spatially fragmented fashion, with the result that observers are able to report the whole shape of objects. This research considers relations in space and time that allow the integrated percepts of complete objects. Specifically, temporal characteristics for spatiotemporal integration of illusory contour and spatial characteristics of interpolated contour are examined. A serial presentation paradigm and a dot localization task were used in two experiments; observers localized a probe dot relative to a perceived contour of an illusory object. Each of four inducing figures was briefly presented in a serial order to observers and the total time of the series was manipulated. In Experiment 1 short time ranges varied up to 180 ms, whereas longer times were examined in Experiment 2. Overall, the results demonstrate that a short time allows spatiotemporal integration, and that the perceived location of contour consistently shifts with time range. These experiments suggest that the mechanism of spatiotemporal integration operates on spatial integration as a limiting case.     

Haptic object recognition: how important are depth cues and plane orientation?

Raised-line drawings of familiar objects are very difficult to identify with active touch only. In contrast, haptically explored real 3-D objects are usually recognised efficiently, albeit slower and less accurately than with vision. Real 3-D objects have more depth information than outline drawings, but also extra information about identity (eg texture, hardness, temperature). Previous studies have not manipulated the availability of depth information in haptic object recognition whilst controlling for other information sources, so the importance of depth cues has not been assessed. In the present experiments, people named plastic small-scale models of familiar objects. Five versions of bilaterally symmetrical objects were produced. Versions varied only in the amount of depth information: minimal for cookie-cutter and filled-in outlines, partial for squashed and half objects, and full for 3-D models. Recognition was faster and much more accurate when more depth information was available, whether exploration was with both hands or just one finger. Novices found it almost impossible to recognise objects explored with two hand-held probes whereas experts succeeded using probes regardless of the amount of depth information. Surprisingly, plane misorientation did not impair recognition. Unlike with vision, depth information, but not object orientation, is extremely important for haptic object recognition.     

Detection and Identification of Change in Naturalistic Scenes

Research using change detection paradigms has demonstrated that only limited scene information remains available for conscious report following initial inspection of a scene. Previous researchers have found higher change identification rates for deletions of parts of objects in line drawings of scenes than additions. Other researchers, however, have found an asymmetry in the opposite direction for addition/deletion of whole objects in line drawings of scenes. Experiment 1 investigated subjects' accuracy in detecting and identifying changes made to successive views of high quality photographs of naturalistic scenes that involved the addition and deletion of objects, colour changes to objects, and changes to the spatial location of objects. Identification accuracy for deletions from scenes was highest, with lower identification rates for object additions and colour changes, and the lowest rates for identification of location changes. Data further suggested that change identification rates for the presence/absence of objects were a function of the number of identical items present in the scene. Experiment 2 examined this possibility further, and also investigated whether the higher identification rates for deletions found in Experiment 1 were found for changes involving whole objects or parts of objects. Results showed higher identification rates for deletions, but only where a unique object was deleted from a scene. The presence of an identical object in the scene abolished this deletion identification advantage. Results further showed that the deletion/addition asymmetry occurs both when the objects are parts of a larger object and when they are entire objects in the scene.     

Object-recognition tasks: comparing paper versions to computerized laboratory methods

This study attempts to generalize Biederman's 1987 findings regarding Recognition by Components theory, which were obtained using a computer administered object-recognition task, to an analog or paper task that is consistent with typical assessment or testing procedures. Three versions of an object-recognition task were developed after the Structure of Intellect-Learning Abilities Test by Meeker, Meeker, and Roid. One version contained randomly fragmented objects, one contained objects with vertices present, and the third contained objects with midsegments. 30 participants were administered each of the three versions in a counterbalanced order. The results are consistent with those of Biederman. Objects with missing vertices were more difficult to recognize than objects with missing midsegments. There was no difference between randomly fragmented objects and those with vertices present. Implications for object-recognition research and test-item development are discussed. In particular, it is suggested that perceptual theories should be used in developing test items to gain greater control in creating items of appropriate difficulty and to increase the validity of the overall instrument.     

Temporal variations in visual completion: a reflection of spatial limits?

The completion of partly occluded objects appears instantaneous and effortless, but empirically takes measurable time. The current study investigates how amount of occlusion affects the time course and mechanisms of visual completion. Experiment 1 used a primed-matching paradigm to determine completion times for objects occluded by various amounts. Experiments 2 and 3 used a dot-localization paradigm to probe completed contour representations for a qualitative shift above some spatial limit. The results demonstrate that time to completion rises with amount of occlusion. Nonetheless, the visual system can complete highly occluded objects, even when the occlusion renders visible contours nonrelatable. Furthermore, prolonged completion times for highly occluded objects do not result from a breakdown of low-level interpolation processes: The same contour completion mechanism operates on objects occluded by different spatial extents.     

Semantic consistency versus perceptual salience in visual scenes: Findings from change detection

In a one-shot change detection task, we investigated the relationship between semantic properties (high consistency, i.e., diagnosticity, versus inconsistency with regard to gist) and perceptual properties (high versus low salience) of objects in guiding attention in visual scenes and in constructing scene representations. To produce the change an object was added or deleted in either the right or left half of coloured drawings of daily-life events. Diagnostic object deletions were more accurately detected than inconsistent ones, indicating rapid inclusion into early scene representation for the most predictable objects. Detection was faster and more accurate for high salience than for low salience changes. An advantage was found for diagnostic object changes in the high salience condition, although it was limited to additions when considering response speed. For inconsistent objects of high salience, deletions were detected faster than additions. These findings may indicate that objects are primarily selected on a perceptual basis with subsequent and supplementary effect of semantic consistency, in the sense of facilitation due to object diagnosticity or lengthening of processing time due to inconsistency.     

Segmentation of object outlines into parts: a large-scale integrative study

In this study, a large number of observers (N=201) were asked to segment a collection of outlines derived from line drawings of everyday objects (N=88). This data set was then used as a benchmark to evaluate current models of object segmentation. All of the previously proposed rules of segmentation were found supported in our results. For example, minima of curvature (i.e. locations along the contour where negative curvature takes an extreme value) were often used as segmentation points. The second point of a pair connected by a segmentation line often depended on more global shape characteristics such as proximity, collinearity, symmetry, and elongation. Based on these results, a framework is presented in which all of the previously proposed (and now empirically validated) segmentation rules or rules for part formation are integrated.     

Category effects on the processing of plane-rotated objects

We examined the effects of plane rotations on the identification of exemplars of three semantic categories. In the first two experiments line drawings belonging to three categories (animals, inanimate objects, and vegetables) were presented at four orientations (0 degree, 60 degrees, 120 degrees, and 180 degrees of clockwise rotation). The response time was found to depend on stimulus category. In particular, whereas rotation effects were shown for animals, no effect at all was found for vegetables and only partial effects were found for inanimate objects. The unclear pattern found for inanimate objects was further examined in experiment 3 where the orientation effects on the identification of two subsets of the inanimate category were studied. The hypothesis of view-observation frequency was confirmed. In experiment 4, line drawings of objects at different orientations were presented in physically degraded versions. The minimum amount of visual information necessary to identify rotated stimuli was found to vary as a function of stimulus category as well. Results are discussed, combining current research on both viewpoint-dependence/independence and neural systems involved in category processing.     

Object appearance,disappearance, and attention prioritization in real-world scenes

We examined the prioritization of abruptly appearing and disappearing objects in real-world scenes. These scene changes occurred either during a fixation (transient appearance/disappearance) or during a saccade (nontransient appearance/disappearance). Prioritization was measured by the eyes’ propensity to be directed to the region of the scene change. Object additions and deletions were fixated at rates greater than chance, suggesting that both types of scene change are cues used by the visual system to guide attention during scene exploration, although appearances were fixated twice as often as disappearances, indicating that new objects are more salient than deleted objects. New and deleted objects were prioritized sooner and more frequently if they occurred during a fixation, as compared with during a saccade, indicating an important role of the transient signal that often accompanies sudden changes in scenes. New objects were prioritized regardless of whether they appeared during a fixation or a saccade, whereas prioritization of a deleted object occurred only if (1) a transient signal was present or (2) the removal of the object revealed previously occluded objects.     

Intuition in the context of object perception: intuitive gestalt judgments rest on the unconscious activation of semantic representations

Intuition denotes the ability to judge stimulus properties on the basis of information that is activated in memory, but not consciously retrieved. In three experiments we show that participants discriminated better than chance fragmented line drawings depicting meaningful objects (coherent fragments) from fragments consisting of randomly displaced line segments (incoherent fragments) or from fragments which were rotated by 180 degrees (inverted fragments), even if participants did not consciously recognize the objects. Unrecognized coherent, but not incoherent or inverted fragments produced reliable priming of correct object names in a lexical decision task, indicating that coherent fragments activated an unconscious semantic object representation. Priming effects were larger for coherent fragments judged as coherent compared to coherent fragments judged as incoherent. We conclude that intuitive gestalt judgments for coherent fragments rested on the activation of semantic object representations, which biased participants' intuitive impression of "gestalt-ness" even when the underlying object representations remained unconscious.     

Outline shape is a mediator of object recognition that is particularly important for living things

We assess the importance of outline shape in mediating the recognition of living and nonliving things. Natural objects were presented as shaded line drawings or silhouettes, and were living and nonliving things. For object decision (deciding whether an object may be encountered in real life) there were longer response times to nonliving than to living things. Importantly, this category difference was greater for silhouettes than for shaded line drawings. For naming, similar category and stimulus differences were evident, but were not as pronounced. We also examined effects of prior naming on subsequent object decision performance. Repetition priming was equivalent for nonliving and living things. However, prior presentation of silhouettes (but not shaded line drawings) reduced the longer RT to nonliving things relative to living things in silhouette object decision. We propose that outline contour benefits recognition of living things more than nonliving things: For nonliving things, there may be greater 2-D/3-D interpretational ambiguity, and/or they may possess fewer salient features.     

Orientation-specific effects in picture matching and naming

In two experiments, subjects made timed decisions about the second of two sequentially presented rotated drawings of objects. When the two objects were physically identical, response times to decide whether the two drawings depicted the same object varied as a function of the shortest distance between the orientation of the second drawing and either the orientation of the previous drawing or the upright. This was found for both short (250-msec) and long (2-sec) interstimulus-intervals. The result was also obtained when subjects named the second drawing after deciding whether the first drawing faced left or right. Following repeated experience with the drawings in the left/right task over four blocks of trials, time to name the second drawing in the same-object sequences was independent of orientation. These results suggest that, initially, object- and orientation-specific representations can be formed following a single presentation of a rotated object and subsequently used to identify drawings of the same object at either the same or different orientations. Alignment of the second drawing with either the canonical representation or the new representation at the previous orientation is achieved by normalization through the shortest path. Following experience with the objects, orientation-invariant representations are formed.     

Another reason why adults find it hard to draw accurately

Misperception of objects is a major cause of inaccuracies in adults' drawing. It has previously been established that participants' drawings are biased by their knowledge of the drawn object. We hypothesised that additional inaccuracy arises because drawings are biased towards participants' idiosyncratic canonical representations of the object. We report that participants' free drawings of a cylinder are correlated with their observational drawings of the same shape, providing evidence that people's observational drawings are distorted by their individual schematic representations of the objects in question. It is unclear whether this reflects a perceptual distortion or a bias in drawing production; in either case, this result provides a further explanation why people are poor at drawing from observation.     

Color makes a difference: two-dimensional object naming in literate and illiterate subjects

Previous work has shown that illiterate subjects are better at naming two-dimensional representations of real objects when presented as colored photos as compared to black and white drawings. This raises the question if color or textural details selectively improve object recognition and naming in illiterate compared to literate subjects. In this study, we investigated whether the surface texture and/or color of objects is used to access stored object knowledge in illiterate subjects. A group of illiterate subjects and a matched literate control group were compared on an immediate object naming task with four conditions: color and black and white (i.e., grey-scaled) photos, as well as color and black and white (i.e., grey-scaled) drawings of common everyday objects. The results show that illiterate subjects perform significantly better when the stimuli are colored and this effect is independent of the photographic detail. In addition, there were significant differences between the literacy groups in the black and white condition for both drawings and photos. These results suggest that color object information contributes to object recognition. This effect was particularly prominent in the illiterate group.