7.4 General aims of the present experimental work
The aims of the present series of experiments 1‑9 are (i) to determine the effect of three‑dimensional versus two‑dimensional presentations of objects on recognition and naming latencies of normal adults (ii) to determine whether angle of view of rotated stimuli affects the latency of naming and recognition processes (iii) to ascertain whether there is any relation between the behaviour of recognizing the visual object and production of the name (iv) to determine if colour has any effect on the speed of recognition and naming processes.
The object of the present study, then, is to look for evidence of a clear distinction between the different mental representations of an object and the different psychological stages of recognition of an object and production of its name.
These aims were posed in the wider context of providing visual recognition and naming data which may contribute to a better understanding of disorders in visual object recognition and of naming disorders.
7.5 Ooutline of the research
A total of a thirteen experiments were carried out in the present thesis. These experiments were organized into stages as follows: a) nine experiments in the first stage, b) four experiments in the second stage of the present thesis.
In the first nine experiments in the present study a tachistoscopic method was employed to investigate the effects of axis rotation, and stereoscopic depth cues on object naming and object recognition in normal adult subjects.
With regard to the material used in Experiments 1‑9, for experiments 1‑2 the stimuli were photographs of three different commonly‑occurring objects; a screw‑ driver, a pair of scissors and a fork. Each object was photographed from an usual view, in that it depicts the object in an orientation which contains the important structural features necessary for recognition (Freeman, 1980. p. 346), against a white background.
For Experiments 3:6 the stimuli were photographs of the same objects of experiments 1 and 2 except with different depth rotated views for each object.
For Experiment 7, a new set of three photographs of objects were used and they were presented once with identical colour, and once in black and white. The same stimuli were used in Experiments 8 and 9, except they were presented in black and white and viewed only monocularly.
7.5.2. Construction of stimuli for binocular and monocular viewing condition
In order to provide vivid stereopsis, black and white photographs of these objects were taken so as produce two photographs of the object corresponding to the view which would be seen by each eye separately.
The slight differences between pairs of objects are caused by the slightly different viewpoints of the two cameras that were used to record them (equivalent to differences in left and right eye positions in the head). If the photographs are presented one to each eye simultaneously, the resulting impression is of a three‑dimensional image. The three different stimuli were of roughly equal size in terms of area of the screen covered, and of the same level of brightness, although this was not accurately determined.
For the binocular viewing condition, each pair of the photographs was produced by a eparation of 65 mm distance between the twophotographs of the same object.
7.5.3. Construction of the pair of photographs in depth
A pair of photographs of each object was produced. Depth in the pairs of photograph was constructed by rotating the object around an axis that was oblique with respect to the line of sight anti‑clockwise.
The objects were placed on a horizontal axis and rotated in depth from 0 degree to 90 degrees around vertical axis. The distance between the camera and axis of rotation was fixed, and the camera and the object were at the same elevation. The object, say a fork, was rotated around a vertical axis. One of its ends was on the rotation axis. The plane of rotation was horizontal, as shown in Figure 7.1.
Care was taken to ensure that all major parts or features were visible in the 90 degrees angle views of each object. For this view (conventional view)that did not foreshorten the major axis of elongation, the conventional view preserved the principal axis of information. For the view that foreshortened (unconventional view) the major axis, the objects were placed at 0 degree angle in depth, therefore, the salient features were invisible. This manipulation was done to allow examination of the influence of two plausible strategies on object constancy judgements, one relying on feature extraction, the other on main‑axes analysis (Humphreys & Riddoch, 1984; Warrington & James, 1986).
Figure 7.1. illustrates the rotation of an object in depth. Orientations are through three angles as follows: 0, 45, and 90 degrees.
7.6 Ooutline of the experimental part of the thesis
This thesis considers the effect of depth cues on recognition and naming common objects.
The first nine experiments in the present study they were replications of Warrington and James (1986) but with some added variable such as stereopsis. The stimuli were presented binocularly and monocularly to determine the effect of stereopsis on the naming and recognition latencies in normal adults.
In the second stage of the present thesis four Experiments (10: 13) were conducted to extend the findings of the first nine experiments to investigate the role of otherattributes of line drawings vs. words and photographs of common objects such as those edge‑ based cues in one hand and other variables such as imageability in the other hand.
The experiments reported reveals several suggestive proposals about the role of the perceptual information of the visual stimuli upon identification processes. The results of Experiments 1‑9 have shown the importance of stereoscopic information as depth cues upon naming common rotated objects. Stereopsis helped particularly location information in the display of the visual object. The dimensionality of the visual stimuli had strong effect on naming behaviour of normal subject.
The results of the naming tasks of Experiments 5, 7, and 8‑9 show that the time required to identify rotated common objects is dependent on the orientation of particular object. Thus the familiarity with a particular object lead to a reduction in the effect of orientation upon object recognition. This suggests the interesting possibility that object perception and matching can be influenced by the visual attributes of the object.
In every experiment in both recognition and naming tasks, however, response time was greater for objects rotated in depth (i.e. 0 degree) than for the same objects in usual view (i.e. 45 and 90 degrees). This result be interpreted to mean that subjects normalize an object with respect to usual view as well as compute the rotational transformations used to determine object identity.
The result also suggest that in some stage object recognition processing depend upon the mode of presentation of the visual stimuli and task demand.
Experiments 1 and 2 dealt with the role of stereopsis on the visual appearance of the stimuli. Experiments 1 and 2 different in that for experiment 1 object factor in the recognition task was between subjects design whereas for experiment 2 it was within subjects design. Both in experiment 1 and 2 there was reliable effect of binocular over monocular viewing condition in the naming task. Overall the patterns of results between experiments 1 and 2 are consistent, with the exception of the marginal interactions was found; a) the effect of stereopsis was found significant in the naming task; b) the mean reaction time difference between “yes” and “no” was significant. The time to respond “yes” is less “no”; c) the reaction time was long when different objects contained a few attributes than when they contained many attributes. The results suggest that both the 2‑Dimensional and 3‑Dimensional views as depth cues play a differential role in two different tasks (recognition vs. naming) of everyday life common objects.
Experiments 3, 4, 5, and 6 replicated the findings of Experiments 1 and 2 using the same objects but with different subject by investigating the effect of rotated objects in depth on the time of naming and recognition and how this might be differentially affected by binocular and monocular viewing.
Depth in the photograph of object pair constructed by rotating the object around the vertical axis, so that different parts were visible that need to be reconciled as corresponding, and parts of the objects may mentally “disappear” and “reappear” during the rotation process.
Corroboration of the conclusions of Experiments 1 and 2 with an entirely new set of subjects and designs, Experiments 3 to 6 has succeeded in replicating the three principal findings of Experiments 1 and 2; viz; that reaction time a) increases with monocular view in particular with naming task, b) the effective role of surfaces detail of a particular object on naming and recognition processes, c) the fast‑same response over different response. In addition, the effect of rotated objects in depth on naming time.
The result of experiment 3, effect of orientation as a main effect was not significant. This is could due to that subjects learned to minimize the effects of orientation with repeated exposure to the same view in the two tasks.
Experiment 4 was replicated the findings of experiment 3, using a group of 8 subjects with different combinations of presentations order and with the same objects but with naming only.
In experiment 5 the design of experiment 4 was modified. The order of presentation was balanced between subjects.
Experiment 6 was designed to reduce the reliance on increasing familiarity with repeated exposure to the same object by investigating the recognition only without prior object naming.
Experiment 7 replicated the findings of experiments 3 : 6, using different subjects, objects and colours. This experiment was designed to answer two questions; a) how the visual system processes combinations of colour and form? b) does processing occur at different levels of the visual system, thus independently, or the two factors share a common processing level. The difference in the mean reaction time between colour and black‑white did not approach significance.
In experiment 8 and 9 I sought to investigate further the effects of rotated objects in depth without colour and through the introduction of monocular viewing only, as a control experiments, where stereoscopic depth information would not be available.
Experiments 8 and 9 differed in that for experiment 8 objects were the same black‑white objects of experiment 7, whereas for experiment 9 they were only two objects.
In sum, the results of the series of experiments of the recognition task regard the significance or non‑significance differences between “yes” and “no” as a main effect have proved to be useful in interpreting the different processes in visually matching visually presented objects.
The consistent result except the result of Experiment 7 has been that the overall mean time to respond “no” that three views of the same objects rotated in depth in three different angles is longer than the mean time to respond “yes”.
Overall, the results of experiments 1‑9 demonstrated the importance of distance and orientation information for object constancy. That is, as objects are rotated further in depth (increasing their distance from the viewer), subjects tend to identify the object in terms of its projected, retinally based description. Presumably such effects occur because depth cues become degraded in such views. Similarly, shape constancy is considerably better with binocular than monocular viewing (Epstein & Hatfield, 1978).
The sets of objects in Experiments 1‑9 were familiar. Because they are familiar, it is possible the advantage of viewing an object at 90 degrees could not be attributed solely to perceptual processes. Instead, the advantage of usual position of object at 90 degrees could be due to the fact these objects are seen in their usual positions more frequently than in other positions, so their memory representations reflect this fact. Further, because the objects had been rotated in depth, they may have different sizes and in this way they could account for an increase in reaction time.
With respect to the possibility of other different mechanisms underlying the effect of rotated object in depth on encoding time in particular, the advantage for viewing an object at 90 degrees over 0 and 45 degrees which could be attributed to the presence of surface, edge‑based, and depth cues that available for recognizing real objects but not for line drawings. Thus, the nature of object depicted in the photograph play an important factor in recognition and naming processes. This is the focus of Experiments 10 and 11 and also to search on the relationship between visual imagery and visual perception. To do so experiment 10 was designed to investigate the effect of imagery values and size of selected line drawings on naming time. The result did not yielded a significant effect between high imagery values and low imagery values and the differences is due to size and the objects per se.
Experiment 11 replicated the findings of Experiment 10 using different subject and different stimuli to investigate the extent to which the same line drawings as those used in experiment 10 could be presented in different mode of presentation, that is, “words”.
A strong effect of imagery on naming high versus low words was obtained. Overall the results of experiments 10 and 11 may be interpreted in terms of the dual‑code theories of semantic memory. The different entry level could possibly cause performance differences in semantic decisions between naming line drawings and corresponding words (Jolicoeur et al, 1984).
However, any direct quantitative comparison between the two experiments is subject to some uncertainty owing to differences between the populations from which the two sets of stimuli and subjects were sampled. In addition, in all the first nine experiments of the present thesis the inter stimulus interval (ISI) was kept very brief as that 100 milliseconds (ms) to prevent the subjects moving their eyes. Thus the fast‑same effect for viewing an object at 90 degrees over rotated objects in 0 and 45 degrees might be due to the short interval between successive presentation of stimuli.
Experiments 12 and 13 were designed to test the selective effects of variables such as the inter‑stimulus interval, and exposure duration on subsequent recognition.
Experiment 12 and 13 different in that for Experiment 12 objects were a set of 5 line drawings with high imagery values and a set of 5 line drawings with low imagery values taken from Experiment 10 whereas for Experiment 13 the stimuli were the same as those of Experiment 12 except they were photographs of real objects.
In both experiments 12 and 13, subjects had to match successively presented photographed stimuli half with high imagery values and the other half with low imagery values (line drawings for experiment 12 and real objects in experiment 13), if these had the same‑different name at a three different ISI (100 ms, 200 ms and 2500 ms. “Same” trials could result from two identical photographs, from two photographs of mirror image object, or from two photographs of the objects which share the same name (e.g. saws).
In summary, imageability has been found as an important influence on the overall of matching time for the four different conditions (i.e. identical, mirror image, and different photographs share name). Overall, identical photographs was matched faster than mirror image and same name conditions. At the three different ISIs (100 ms vs. 200 ms vs. 2500 ms) the advantage of identical over mirror image and same name persisted. This results could be explained in terms of the availability of three different types of code, a “pictorial” or literal description of a particular photographs of object, an “object” level code, and a non‑visual semantic level code, at which two different objects sharing the same name would be equivalent. Although the results of these two separate experiments provide evidence for the independence of viewer‑centred from object‑centred visual codes. However, the results of experiments 12 and 13 were failed to show that both viewer‑centred and object‑centred affected by ISI factor.
However, the present study provide quite strong evidence that visual object recognition and object naming is not a unitary process because of the subprocess such as transformation, reduction and interrelating different mental codes, as suggested by Marr, (1982); Warrington (1988).
7.7 Summary of experiments 1 ‑ 9
In the first nine experiments, 118 subjects (aged 23‑35) were shown stereoscopic pictures of three rotated common objects, presented tachistoscopically, for the purpose of examining the effects of stereopsis, angle of view and colour on reaction time for recognition and naming of those objects.
The results of these studies show that naming speed is related to viewing condition and angle of view. Angle of view is related to the efficiency of recognition. In addition, the results indicate that under these experimental conditions, stereopsis significantly affects naming and recognition but for different reasons. These results provide clues about the interpretation, by the visual system, of shape from stereopsis cues and the relationship of shape from stereopsis to other depth cues in determining the perceived rotated object in depth and surfaces detail such as colour of objects. The results were discussed within the framework of the 2.5‑dimensional sketch, hypothesized as representing the orientations and distances of visible surfaces relative to the viewer (Marr 1982).
- Google patents new object recognition technology, likely has plans to use with YouTube (androidauthority.com)