13.4. Contrasting angles effects in recognition and naming tasks of the present study
In considering the nature of the orientation function, overall, the results of the present study indicate that first, for the naming reaction times, in five of seven experiments 3:9 of the present study the overall RT differences between the three different angles was significant. The time taken to name and to match three different objects increased with increases in degree of rotation in depth.
Second, for the recognition tasks, in four of seven experiments of the present study the overall RT differences between the three different angles was significant.
For Experiment 3, the effect of angles the result on object naming was not significant. In contrast, the effect of angles on recognition task was found to be significant. This differential effect on both tasks in the same experiment was interpreted in terms of stimulus familiarity. That is, repeated exposure with an object rotated in depth in the two different tasks (recognition vs. naming) in the same experiment results in establishment of a broadly tuned memory representation that allows direct stimulus recognition over a relatively wide range of stimulus orientations.
Experiments 4 , 5, and 6 replicated the result of Experiment 3 by manipulating object naming and object recognition in a single task.
The negative results of Experiment 4 alone do not allow any firm conclusions concerning the familiarity hypothesis raised in Experiment 3. While the non-significant results of the different treatments of experiment 4 fits with the feature priming account, there are other possible explanations for the above findings. Thus, to supplement the results of Experiment 4, two further experiments of 5 and 6 were conducted.
Recognition without prior presentation of naming in Experiment 6, and naming without prior recognition in Experiment 5, the overall results indicated that the “familiarity” explanation to some extent is correct. Inference conclusion from replicated Experiments 4, 5 and 6, suggest that the different effects of objects rotated in depth causes mental transformation affects some early process prior to any identification or matching of objects. However, the contrasting results between experiments 4 to 6 in which the subjects required to deal with objects rotated in depth suggests that there is relatively slow and effortful mental transformation process, which uses whole-object description, and which is effected primarily in recognition tasks (see, Corballis, 1986). In addition, there is a relatively fast‑acting perhaps more automatic, compensatory process which is invoked in naming tasks.
A comparison between the overall RT differences of the three different angles for both recognition and naming tasks , indicated that the overall RT at angle 0 degree is longer than the overall RT at both angles 45 and 90 degrees except of Experiment 4 and the recognition task of Experiment 7, the differences between angles were negligible. Overall, unforeshortened views of objects (i.e., 90 degrees) were named and recognized more rapidly than foreshortened (i.e., 45 and 0 degrees) views. This result suggests that the perceptual tasks in which the normal redundancy (i.e., a stimulus description) present in the visual stimulus is reduced appear to be damaged (i.e., 0, and 45 degrees).
This result reveal the effects of the viewpoint of a particular orientation on mean RT for both object recognition and object naming.
There are many possible reasons for the fact that the appearance of object in depth rotations slow down the recognition and naming times. This is because when objects have been rotated around y-axis to achieve the rotation in depth, they have different visible parts that need to be reconciled as the corresponding, and parts of the objects may mentally “disappear” and “reappear” during the rotation process. For the presentation of the same objects at 90 degrees, thus, the same parts were visible in both objects, and did not becoming “hidden” as the objects were mentally rotated into congruence.
Taken together, the effects produced by objects and angles interactions in Experiments 3, 5, 7 and 8 indicated that the overall RT differences is caused by viewing a particular object at a particular orientation. This implies that the visual characteristics of objects with the canonical representation of orientation influence process RT. Namely the overall RT of the object “fork” in four experiments ( i.e., experiments 1, 2, 3 and 4) and the object “cup” in Experiment 7 was short at angle 90 degrees. However, the overall RT of the object “plug” of Experiment 8 was shorter at angle 90 degrees than at both angles 0 and 45 degrees respectively. This result suggests that object constancy is affected also by the internal representation determined by the characteristics of stimulus used in the present study.
It was claimed that for efficient recognition there was a “favored” view. Palmer, et al. (1981) showed that there is a canonical or favored perspective for viewing familiar objects. They suggested that these canonical views maximize the amount of information that is available for perceiving the object. Accordingly, the fast advantage in mean RT of viewed an object at 90 degrees over relatively slow in mean RTs of angles 45 and 0 degrees in the present study can be explained in terms of the way in which subjects set about the photographs of object task itself. If the object is in its usual view s/he accesses a mental representation for the “favored” view which in this case unforeshortened view by simply repeating it on the memory without referring back to the object itself. If the two other objects are different in angle of view, this may prompt the subject to look again after the first presentation.
Apparently the presence of the recognized view provided a context for judging the object seen from the other perspective. Perhaps the conventional view provided access to a categorical description based on the internal structure of the object which was then rotated to a viewing angle similar to the unconventional view to check for correspondence. An alternative suggestion would be that once the object was recognized from the conventional view (successful access of the category in memory), the subjects no longer dependent only on the aspect of the object rotation that they could see, but could use past experience with similar objects seen from similar angles to fill in details of the form which were obscured or distorted, Jolicoeur & Kosslyn (1983) offer evidence that both forms of representation are used for recognition.
An important part of Marr theory concerns the ways in which the visual system uses canonical forms in a modular organization (that is, split into different parts). One possible canonical form which would be useful to achievement the 3-dimensional representation of a human being is the cylinder. Following the principle of the modular organization, an initial cylinder could be constructed to represent a person, provided the visual system could first decide upon the direction of the principal axis: in this case, from head to feet. This would be self-contained unit in the shape description. It is possible to enrich the description, using the same canonical form to represent the head, the torso, the arms and legs by smaller cylinder. Clearly, then, the internal representation of an external object captures its three‑dimensional structure, not as that structure exist in the object absolutely, but, only as it appears relative to a particular angle of regard. What I wish to argue is that the internal representation shares properties with both the three‑dimensional object and its two‑dimensional perspective projection, without being wholly isomorphic to either. It resembles the perspective view in its integration of information (not contained in the object itself) concerning the relation of the object to the viewer and it resembles the object, in the cognitive operations upon the representation which are more simply related to properties specifiable in three‑dimensional space than to properties peculiar to the particular two‑dimensional projection. What the representation really represents, then, is the appearance, from a particular point of view of the object in depth.
The overall result produced by objects and angles in the present study further suggest that subjects may use the canonical view to drive the correct description of an object presented in an unusual view.
For orientations between 90 degrees and 0 degree of rotation in depth there is a linear increase in naming time the further an object is rotated in depth. This pattern of effects suggests that successful recognition require a transformation of the input representation of rotated object to the 90 degrees for comparison with orientation-dependent representations stored in memory. One candidate process for accomplish this transformation is mental rotation, the analogue process assumed to underlie the orientation effect found for judgments of left-right reflection (Jolicoeur, 1985, 1988). This view would account for the reaction time increases with increases in degree of rotation in the present study.
That is for the finding that unusual views of an object at 0 and 45 degrees seen in the course of the experiment do not show a reduced in RTs when they are presented later for recognition or for naming tasks .
One further explanation rests on the assumption, first, that the subject is tested for a direct match between their internal representation and the external stimulus. Whenever they detect a mismatch, then, they require an additional fixed amount of time to switch to the other response (Clark & Chase, 1972).
The adequacy of this explanation rests on the variance hypothesis of shape perception which might account for the reaction time increases with increases in degree of rotation.
The accumulating evidence (Jolicoeur, 1985, 1988, 1990; Jolicoeur & Milliken, 1989; Maki, 1986; McMullen & Jolicoeur, 1992) in the literature seems to favour the significance effects of orientation on naming time are the result of the use of orientation-variant information hypothesis. For example Jolicoeur & Milliken (1989) found that the time to name disoriented line drawings of objects and animals increases systematically as a function of the amount of rotation of the object from the upright (for drawings rotated from 0 to 120 degrees). These results converge with evidence presented by Jolicoeur & Landau (1984). They reported orientation effects on the identification of alphanumeric characters. The individual characters were briefly presented and pattern masked, the proportion of correct identifications was taken as dependent variable. Correct identifications decreased linearly as a function of the angle of the characters from the upright (at least up to 180 degree rotations within‑ the plane). Such linear relationships have been taken as indicating the operation of an effortful mental transformation process in shape‑matching which is directly analogous to the external manipulation needed to transform one target object into another (Kosslyn, 1980, 1983; Shepard & Cooper, 1982).
One of the most interesting finding of Jolicoeur & Landau’s (1984) is that when performance is taken to limit, orientation effects are obtained even on the identification of highly familiar stimuli often presumed to contain Landmark features. There are some results consistent with the landmark feature account, (Hochberg & Gellman, 1977), suggesting that the landmark features could rapidly be identified irrespective of the shapes orientation.
Some previous research findings suggests that the changes in the angle of rotation of real common photographs of objects for naming and recognition can be diagnostic (Biederman & Ju, 1984) in evaluating models of pattern recognition. If the hypothesis of pattern recognition proceeds via ‘‘ extraction of features”, then we would expect no effect of orientation on recognition.
The feature model predicts that orientation could affect the pattern recognition process in a systematic way. In contrast, there are several recent studies which suggest that changes in retinal orientation have little or no effect on the time required to identify simple patterns (Corballis & Nagourney, 1978; Corballis, Zbrodoff, Shetzer & Buttler, 1978) Identification might then proceed via the extraction of the landmark feature whatever the viewpoint (often the feature of an object is visible). There is little evidence that these sorts of invariance actually occur at the level of feature extraction. It is well known that so called edge and line detectors in the visual systems of cats (Hubel & Wiesel, 1965) and monkeys (Hubel & Wiesel, 1968) are tuned to specific angular orientations and directions of movement; they do not respond if these orientations and directions are, say, reflected about the vertical.
The theories of feature extraction suggest that the recognition may be based on those features that would not be affected by rotation: a capital letter A would remain the sharp point and P the closed loop after frontal plane rotation. If a set of such orientation‑free features is stored for each form and used for its recognition, form recognition will be released from the orientation problem.
The “object‑centred coordinate system” proposed by Marr & Nishihara (1978; 1982) was designed to structure separate orientation‑free features into an integrated form.
Marr & Nishihara (1978) proposed that the shape of a three‑dimensional objects is represented with respect to a frame of reference that is aligned with the direction of maximum elongation of the object. The resulting representation is independent of the orientation of the object because the frame of reference used to encode the object remains invariant as the object is rotated in space. That is, the principal axis serves as a basis to construct an entire coordinate system. All the subsidiary axes are located with reference to this coordinate system. Such a system has an important characteristic which was the very purpose of its development: the structural description of the object remains the same irrespective of the orientation of its principal axis, namely, the orientation of the whole object. This is because the structure is described without making reference to any external framework. It follows that the two identical objects have the same description even when they are placed in totally different orientations. A direct comparison of the perspective descriptions will suffice to determine whether the two objects are the same or different; no mental rotation is needed.
However, rotating the object may alter the ease with which the viewer can assign the intrinsic or object‑centred frame of reference to the object ( Jolicoeur & Kosslyn, 1983; Marr & Nishihara, 1978; Warrington & Taylor, 1973). Thus, while the representation of the object is orientation invariant, the processes that produce that representation may not be completely insensitive to orientation.
In contrast, to feature extraction theories, a number of studies demonstrate great effects of orientation on identification of pattern ( Jolicoeur & Landau, 1984; Kolers & perkins, 1969a, 1969b).
An alternative interpretation, which makes no reference to feature detection per se, is that subjects generally adopt a mental‑rotation strategy when the task requires them to recognize that the visual display is same or different from the target object in the memory. Although, several mental rotation studies emphasis that when subjects identify or classify disoriented patterns, generally, mental rotation would not be involved, whether they are familiar alphanumeric character (Corballis & Nagourency, 1978; Corballis, Zbrodoff, Shetzer & Butler, 1978; White, 1980) or relatively unfamiliar letter-like‑forms (Eley, 1982). There may be exceptions, however. For instance Shwartz (1981) found that subjects apparently did adopt a mental‑rotation strategy in identifying pictures of disoriented real‑world objects, such as animals, inanimate objects (e.g., airplane, gun), and well‑known faces. Similarly, presumably, subjects did adopt a mental‑rotation strategy in recognize pictures of photographs of disoriented common objects ( e.g., plug, stamp and screw‑driver). However, it must be supposed in the present study that the subjects recognize the objects in at least an initial manner prior to mental rotation, for otherwise they could not have known what position to mentally rotate the objects to. Therefore, mental rotation may have served as a check on some prior recognition.
The overall difference between usual view of object at angle 90 degrees in the present study is equivalent to the “canonical representation” and rotated object in depth at angle 0 degree is equivalent to the “foreshortened representation” reaction times suggests further that a match is generally tested for first. If this match fails, then extra time is evidently needed to switch to the “yes” response. This initial comparison of the transformed internal representation with a memory image of usual view “canonical” version is consistent with the “congruence” principles discussed by Clark & Chase (1972) and Trabasso, Rollins, & Shaughnessy (1971). The linear relationship between reaction time and angular difference in rotated object, obtained in the present study, is one form of evidence that the internal process underlying the observed reaction times is passing through an ordered series of stages. That is, the reaction‑time increases with increases in degree of rotation indicates that the time needed to compare between two objects one in the memory and the other on the visual display presented at orientations 0 and 90 degrees is an additive combination of the time needed to compare those objects presented at orientations 45 and 90 degrees. This finding is indirect evidence for the claim that the internal process underlying comparison of the objects presented in orientations A and C passes through an intermediate state corresponding to orientation B (Cooper, 1975).
The theory of mental rotation seems to furnish one of the possibilities to account for the principal findings of the experiments reported here: namely, (a) that the time required to complete the naming process and possibly also the recognition process increases linearly with the angle of rotation, (b) the effects produced by a particular object at a particular orientation.
The tendency to normalize the orientation of a pattern which is tilted from the vertical in a frontal plane has been claimed to be a diagnostic feature of certain neurological and psychiatric disorders ( Royer & Holland, 1975). For example, Elderd (1973) asked children to draw a copy of a standard pattern when the standard pattern is in one of several orientations.
Elderd reported that children tend to normalize their drawings to an upright orientation. However, in this study, as in most other studies of this kind, the standard pattern and the copy were lying flat on a table top, so that normalization was certainly not to the gravitational vertical. Furthermore, the model was presented on rectangular pieces of paper, so that normalization may not even have been to the normally vertical meridian of the eye, but rather to the edges of the paper.
Gombrich (1969) has argue that the child simplifies an internal representation by first normalizing the drawing with respect to the sides of the paper, or with respect to his own body if the paper is circular, and then draws the description which he has compiled and ignores the standard pattern. I propose that if the child can decide whether or not two objects have the same orientation when they do not intersect or when they are presented one after the other. In these cases, the task entails transferring the attention from one object to the other and comparing the iconic image of one with the current image of the other. In other words, the child forms an internal abstract representation, or description, of one and compares it with a description of the other.
It can be argued that in the present study, the subject in the recognition task is ‘distinguishing’ between a several visual object presentations and by forms an internal abstract representation which makes it easier by distinguish each object from the other.
With respect to depth rotation, views which foreshorten objects seem particularly difficult to recognize, as suggested both from everyday experience, and from studies of neuropsychological patients.
The results of the present experiments demonstrated that the identification of an object depicted from a view that foreshortens a major axis of elongation is difficult for normal subjects, which is similar to what reported for the right‑hemisphere‑damaged patients in the neuro‑psychological studies. Such a result might suggest that normal subjects would show fast reaction time in identifying or recognizing usual view over foreshortened views. The subjects may use the canonical view to drive the correct description of an object presented in an unusual view.
It was shown that the right‑hemisphere lesion group were impaired on a same‑ or different matching task in which one prototypical view are paired with a non prototypical‑view object photograph (Humphreys & Riddoch; 1984). These findings were interpreted in terms of a failure of perceptual categorization, a mechanism whereby two or more stimulus inputs are allocated to the same class. It appeared that the detection of similarity in different views of the same object had broken down and categorization of percepts by physical identity was impaired. Warrington & Taylor (1978) argued that this perceptual categorization mechanism was a post‑sensory but presemantic stage of object recognition.
13.5. RESPONSES SELECTION
In general, the results of the recognition task in Experiments 1‑3; 6, 8 and 9 reveals that the
recognition time as reflected in “yes” and “no” response type increased with increases of the degree of rotation between objects. This result reveals several suggestive differences regarding the issues of matching process.
First, the issue regarding the disparity between physical match and name match and second, the issue regarding the effect of mode of presentation upon recognition of visual stimuli were explored in the present series of experiments.
It is assumed that normalization operations take an amount of time that is proportional to the difference between the input and memory representation along the dimension to be normalized ( Dixon & Just, 1978; Neisser, 1967). Thus, the normalization of orientation should show recognition disparity between input pattern and the stored representation. Similarly, the input in this study was viewing the same objects from different angle and the subject stored the canonical visual image in his mind’s eye.
The subject in his mind’s eye merely observe it. S/he has seen a shape in this particular orientation before, and s/he could attach a yes or no response to a set of orientations that are equivalent with respect to having been seen before. The general feature of an identification is that each member of at least one proper identity or equivalence subset of the stimulus (i.e., 0, 45 and 90 degree of the same stimulus) domain must evoke the same response and each member of at least non‑target must evoke a “No” response.
” an identification task is one that requires the subject to pair all members of at least one proper subset of defined stimulus domain with the same identifying response, and members of another, non‑overlapping proper subset with a different identifying response.”
In other words the subject was asked to identify the oriented target, by making some distinctive response, “Yes/No”. Thus, s/he is not required to give a structural description. It has been argued that subjects do not behave this way and explicit structural descriptions are normally involved in human perception ( Pylyshn, 1973 ).
For the object recognition task, the subject was given the name of an object before the actual visual display [memory item to held in memory] and subsequently the subject was asked to judge whether or not a singly presented item (target object) was the same or different. Thus the matching paradigm in this case is said to be memory‑dependent in which stimuli are presented successively, [memory‑task].
Since the stimuli in the present study differed in a number of attributes such as viewing an object binocularly and monocularly, orientation, and colour versus black‑white, the visual objects in this case possess two important features: the target defining attribute, the physical object, and the response attribute itself (i.e. yes and no response). Therefore, the major concern of the present study was to explore the mechanisms by which these two characteristics of the visual object are analyzed. The question arises; how the response type as reflected in recognition latency in this situation can be constructed.
In sum, the results of the different treatments of Experiment 4 were negative, except, the effects produced by subjects, viewing, object and angle suggest that the interaction between the order of presentation (as each subject of eight assigned to a particular order of presentation formed by combination of the different variables) is one of the possible reason induced to detect the different treatments effects.