Using spatial frequency adaptation to study word recognition

The study of spatial frequency is being used increasingly often to investigate processes underlying visual word recognition. However, research in this area has adopted techniques that require the physical deformation of word targets used in experiments (e.g., filtered images of words, words embedded in visual noise), and this approach may limit the inferences that can be made about the role of spatial frequencies in normal word recognition. Spatial frequency adaptation is described in this article as an additional technique for studying the role of spatial frequency information in word recognition. The advantage of this technique is that it alters participants' sensitivity to particular spatial frequencies and so allows the study of spatial frequency involvement in word recognition using normal images of word stimuli. The application of the adaptation technique to studies of word recognition is explained in detail and its potential is then demonstrated by an example word recognition experiment in which spatial frequency adaptation was used.     

Learning influences the encoding of static and dynamic faces and their recognition across different spatial frequencies

Studies on face recognition have shown that observers are faster and more accurate at recognizing faces learned from dynamic sequences than those learned from static snapshots. Here, we investigated whether different learning procedures mediate the advantage for dynamic faces across different spatial frequencies. Observers learned two faces—one dynamic and one static—either in depth (Experiment 1) or using a more superficial learning procedure (Experiment 2). They had to search for the target faces in a subsequent visual search task. We used high-spatial frequency (HSF) and low-spatial frequency (LSF) filtered static faces during visual search to investigate whether the behavioural difference is based on encoding of different visual information for dynamically and statically learned faces. Such encoding differences may mediate the recognition of target faces in different spatial frequencies, as HSF may mediate featural face processing whereas LSF mediates configural processing. Our results show that the nature of the learning procedure alters how observers encode dynamic and static faces, and how they recognize those learned faces across different spatial frequencies. That is, these results point to a flexible usage of spatial frequencies tuned to the recognition task.     

Effects of image background on spatial-frequency thresholds for face recognition

A great deal of work has been devoted to the question of which spatial frequencies, if any, are optimal for various visual tasks, such as face and object recognition. However, to date these studies have all been carried out with stimuli set against a uniform background. It is possible that this type of stimulus does not produce ecologically valid results. The natural world in which visual tasks normally take place involves a great deal of luminance variation and distracting visual structure, which may alter the spatial frequencies necessary for a task. We conducted two experiments that examined the effects of image background on the spatial-frequency thresholds (50% maximum of a low-pass or high-pass Butterworth filter) for face recognition by the psychophysical methods of adjustment and constant stimuli. In both experiments we found no significant difference in spatial-frequency thresholds between uniform-grey backgrounds and natural-scene backgrounds, and only minor differences between uniform-grey backgrounds and fractal noise backgrounds. This suggests that results obtained with uniform backgrounds are ecologically valid and that background effects, if they exist, are small.     

Face perception: An integrative review of the role of spatial frequencies

The aim of this article is to reinterpret the results obtained from the research analyzing the role played by spatial frequencies in face perception. Two main working lines have been explored in this body of research: the critical bandwidth of spatial frequencies that allows face recognition to take place (the masking approach), and the role played by different spatial frequencies while the visual percept is being developed (the microgenetic approach). However, results obtained to date are not satisfactory in that no single explanation accounts for all the data obtained from each of the approaches. We propose that the main factor for understanding the role of spatial frequencies in face perception depends on the interaction between the demands of the task and the information in the image (the diagnostic recognition approach). Using this new framework, we review the most significant research carried out since the early 1970s to provide a reinterpretation of the data obtained.     

Hemispheric asymmetry in the processing of high and low spatial frequencies: a facial recognition task

The research investigated the relationship between spatial frequency and visual field in a facial recognition task. Faces of neutral affect (Ekman, 1979) were tachistoscopically presented to the right or left visual field. The faces were presented alone, or masked with square wave gratings of 1, 24, or 48 cycles/degree, for a duration of 10 msec. Accuracy in recognizing each target face from a group of five served as the dependent measure. Subjects were 15 males and 15 females. ANOVA results included a frequency x visual field interaction effect (p less than .001). As was hypothesized, LVF errors were highest in the absence of low spatial frequencies, while RVF errors were highest when a higher range of spatial frequencies was removed. These results confirm that the hemispheres show a differential efficiency in processing high and low spatial frequency information in faces. They also offer empirical evidence to support the clinical findings that both hemispheres contribute to facial recognition.     

Spatial frequency content of visual imagery

Three experiments employing the McCollough paradigm were conducted to determine the spatial-frequency content of visual imagery. In Experiment 1, large and reliable pattern-contingent color aftereffects were obtained after adaptation to visual imagery. The direction of the aftereffects indicated that subjects were adapting to higher spatial frequencies in their imagery. These results contrast with the data of Experiment 2, which demonstrate that color aftereffects obtained with adaptation to physically present stimuli are mediated by the fundamental spatial frequency components. The magnitude of the imagery-induced aftereffects in Experiment 1 equaled the magnitude of the externally induced aftereffects obtained in Experiment 2 with the same subjects. By blurring the to-be-imaged patterns (Experiment 3), the fundamental Fourier components became the salient perceptual features of the stimuli, and the direction of the imagery-induced aftereffects was reversed from that of Experiment 1, indicating that the spatial frequency content of the imagery had changed from higher to lower frequencies. Under normal viewing conditions, subjects use the higher spatial frequencies associated with the perceptually salient edges of stimuli to construct their images. The results of Experiments 1 and 3 are discussed in light of a current controversy over the nature of information representation in imagery, and it is concluded that support has been obtained for the analog model of visual imagery.     

The effect of combinations of image degradations in a discrimination task

This paper explores the ways in which combinations of image degradations affect discrimination. Nine experiments are described that examine the discriminability of visual images that are degraded with three types of information reducing transformations: random punctate visual interference, lowpass spatial frequency filtering, and local area (i.e., block) averaging. The results of these experiments characterize a powerful visual ability to discriminate highly degraded stimuli unless that ability is severely challenged by relatively high levels of random visual interference. Discriminative commutativity of the orders in which the other two degradations are imposed is demonstrated. That is, the order in which the degradations are applied does not affect the final discriminative outcome. This result is in contrast to predictions from relevant mathematics and direct examination of the images produced by both orders of degradation. The commutativity is attributed to the particularly strong effect of the low-pass spatial frequency filtering degradation on the discrimination process. This study also demonstrates that combinations of degradations in a discrimination task always result in a reduction in performance, and never in the improvement that has been reported for recognition. This difference is attributed to the fact that form discrimination is mediated mainly by local features and high-frequency spatial components, whereas recognition is mediated mainly by global features and low-frequency spatial components.     

Newborns' face recognition is based on spatial frequencies below 0.5 cycles per degree

A critical question in Cognitive Science concerns how knowledge of specific domains emerges during development. Here we examined how limitations of the visual system during the first days of life may shape subsequent development of face processing abilities. By manipulating the bands of spatial frequencies of face images, we investigated what is the nature of the visual information that newborn infants rely on to perform face recognition. Newborns were able to extract from a face the visual information lying from 0 to 1 cpd (Experiment 1), but only a narrower 0-0.5 cpd spatial frequency range was successful to accomplish face recognition (Experiment 2). These results provide the first empirical support of a low spatial frequency advantage in individual face recognition at birth and suggest that early in life low-level, non-specific perceptual constraints affect the development of the face processing system.     

上下文预期在快速场景识别中的作用

本研究首先考察上下文预期是否能够影响快速场景识别,进而通过比较不同空间频率的场景信息来探究上下文预期在快速场景识别不同阶段的作用。共包括3个实验:实验1采用双眼竞争范式,考察对快速呈现场景刺激的主观选择是否受到上下文预期的影响;实验2采用词汇分类与快速场景识别的双任务范式,比较预期与非预期条件下快速场景识别绩效的差异;实验3分别以低频信息(实验3a)和高频信息(实验3b)作为实验材料,探讨预期效应的作用阶段。结果发现:上下文预期可以影响快速场景识别过程中的主观选择与反应绩效;对不同空间频率的场景信息,场景刺激与预期相一致时均可以提高辨别力,但预期对反应偏好的影响只发生在对高空间频率场景信息的加工阶段。因此,上下文预期在快速场景识别不同加工阶段的作用不同,快速场景识别需要结合不同空间频率信息加工的结果。     

Spatial-frequency thresholds for object categorisation at basic and subordinate levels

In an attempt to understand how low-level visual information contributes to object categorisation, previous studies have examined the effects of spatially filtering images on object recognition at different levels of abstraction. Here, the quantitative thresholds for object categorisation at the basic and subordinate levels are determined by using a combination of the method of adjustment and a match-to-sample method. Participants were asked to adjust the cut-off of either a low-pass or high-pass filter applied to a target image until they reached the threshold at which they could match the target image to one of six simultaneously presented category names. This allowed more quantitative analysis of the spatial frequencies necessary for recognition than previous studies. Results indicate that a more central range of low spatial frequencies is necessary for subordinate categorisation than basic, though the difference is small, at about 0.25 octaves. Conversely, there was no effect of categorisation level on high-pass thresholds.     

A multistream model of visual word recognition

Four experiments are reported that test a multistream model of visual word recognition, which associates letter-level and word-level processing channels with three known visual processing streams isolated in macaque monkeys: the magno-dominated (MD) stream, the interblob-dominated (ID) stream, and the blob-dominated (BD) stream (Van Essen & Anderson, 1995). We show that mixing the color of adjacent letters of words does not result in facilitation of response times or error rates when the spatial-frequency pattern of a whole word is familiar. However, facilitation does occur when the spatial-frequency pattern of a whole word is not familiar. This pattern of results is not due to different luminance levels across the different-colored stimuli and the background because isoluminant displays were used. Also, the mixed-case, mixed-hue facilitation occurred when different display distances were used (Experiments 2 and 3), so this suggests that image normalization can adjust independently of object size differences. Finally, we show that this effect persists in both spaced and unspaced conditions (Experiment 4)—suggesting that inappropriate letter grouping by hue cannot account for these results. These data support a model of visual word recognition in which lower spatial frequencies are processed first in the more rapid MD stream. The slower ID and BD streams may process some lower spatial frequency information in addition to processing higher spatial frequency information, but these channels tend to lose the processing race to recognition unless the letter string is unfamiliar to the MD stream—as with mixed-case presentation.     

Analysis of tactile and visual confusion matrices

Confusion matrices were compiled for uppercase letters and for braille characters presented to observers in two ways: as raised touch stimuli and as visual stimuli that had been optically filtered of their higher spatial frequencies. These and other existing matrices were subjected to a number of analyses, including the choice model and hierarchical clustering. The strong similarity of the visual and tactile matrices from this study lends additional support to the claim that visual recognition of low-pass filtered characters, to a first approximation, can be taken as a model of tactile recognition of small two-dimensional raised patterns. Besides this, the analysis questions the widely held assumption that response bias contributes significantly to the stimulus-response contingencies in a character-recognition task.     

Hemispheric asymmetry in the processing of absolute versus relative spatial frequency.

Observers indicated whether a stimulus presented to one visual field or the other consisted of two sine-wave gratings (the baseline stimulus) or those same two gratings with the addition of a 2 cycle per degree (cpd) component. When the absolute spatial frequencies of the baseline stimulus were low (0.5 and 1.0 cpd), there was a left visual field-right hemisphere (LVF-RH) advantage in reaction time (RT) to respond to the baseline stimulus which disappeared when the 2 cpd component was added (i.e., the stimulus consisted of 0.5, 1.0, and 2.0 cpd components). When the absolute spatial frequencies of the baseline stimulus were moderate to high (4.0 and 8.0 cpd), a right visual field-left hemisphere advantage in RT to respond to the baseline stimulus approached significance and shifted to a significant LVF-RH advantage when the 2 cpd component was added (i.e., the stimulus consisted of 2.0, 4.0, and 8.0 cpd components. That is, adding the same 2 cpd component caused opposite shifts in visual laterality depending on whether 2 cpd was a relatively high or relatively low frequency compared to the baseline.     

Subordinate-level categorization relies on high spatial frequencies to a greater degree than basic-level categorization

In two experiments, category verification of images of common objects at subordinate, basic, and superordinate levels was performed after low-pass spatial filtering, high-pass spatial filtering, 50% phase randomization, or no image manipulation. Both experiments demonstrated the same pattern of results: Low-pass filtering selectively impaired subordinate-level category verification, while having little to no effect on basic-level category verification. Subordinate categorization consequently relies to a greater degree on high spatial frequencies of images. This vulnerability of subordinate-level processing was specific to a lack of high spatial frequency information, as opposed to other visual information, since neither high-pass filtering nor the addition of phase noise produced a comparable reduction in performance. These results are consistent with the notion that object recognition at basic levels relies on the general shapes of objects, whereas recognition at subordinate levels relies on finer visual details.     

Reduced Reliance on Optimal Facial Information for Identity Recognition in Autism Spectrum Disorder

Previous research into face processing in autism spectrum disorder (ASD) has revealed atypical biases toward particular facial information during identity recognition. Specifically, a focus on features (or high spatial frequencies [HSFs]) has been reported for both face and nonface processing in ASD. The current study investigated the development of spatial frequency biases in face recognition in children and adolescents with and without ASD, using nonverbal mental age to assess changes in biases over developmental time. Using this measure, the control group showed a gradual specialization over time toward middle spatial frequencies (MSFs), which are thought to provide the optimal information for face recognition in adults. By contrast, individuals with ASD did not show a bias to one spatial frequency band at any stage of development. These data suggest that the “midband bias” emerges through increasing face-specific experience and that atypical face recognition performance may be related to reduced specialization toward optimal spatial frequencies in ASD.     

Concurrent processing of spatial relations and objects in two cerebral hemispheres

This study examined hemispheric asymmetry for concurrent processing of object and spatial information. Participants viewed two successive stimuli, each of which consisted of two digits and two pictures that were randomly located and judged them as identical or different. A sample stimulus was presented in a central visual field, followed by a matching stimulus presented briefly in a left or right visual field. The matching stimuli were different from the sample stimuli with respect to the object (digit or picture) or spatial (locations or distances of items) aspect. No visual field asymmetry was found in the detection of object change. However, a left visual field advantage was found in the detection of spatial change. This result can be explained by the double filtering by frequency theory of Ivry and Robertson, who asserted that the left hemisphere has a bias for processing information contained in relatively high spatial frequencies whereas the right hemisphere has a bias for processing information contained in relatively low spatial frequencies. Based upon this evidence, the importance of interhemispheric integration for visual scene perception is discussed.     

A test of the spatial-frequency explanation of the Müller-Lyer illusion

Previous investigations have shown that the response of spatial-frequency-specific channels in the human visual system is differentially affected by adaptation to gratings of distinct spatial frequencies and/or orientations. A study is reported of the effects of adaptation to vertical or horizontal gratings of a high or a low spatial frequency on the extent of the Brentano form of the Müller-Lyer illusion in human observers. It is shown that the illusion decreases after adaptation to vertical gratings of low spatial frequency, but seems unaffected otherwise. These results are consistent with the notion of visual channels that are spatial-frequency and orientation specific, and support the argument that the Müller-Lyer illusion may be due primarily to lower-spatial-frequency components in the Fourier spectra of the image.     

Internal representation of visual texture as the basis for the judgment of similarity

The relationship between Fourier spectra of visual textures (represented both by the actual frequency components and by the response of four hypothetical channels selectively sensitive to spatial frequency) and the perceptual appearance of the textures was investigated. Thirty textures were synthesized by combining seven spatial frequencies whose amplitudes were randomly chosen and then scaled to give an overall contrast of .9. Similarity judgments were collected using both the method of triadic comparison (two subjects, 4,060 trials each) and the method of paired comparison (six subjects, 435 trials each). The similarity judgments were subjected to MDSCAL and INDSCAL dimensions were found to be optimally oriented in terms of spatial frequency information without rotation. The seven spatial frequency components accounted for 90.6% of the variance in the 3-D INDSCAL space, while the four channels accounted for 91.8% of the variance in the first two dimensions. The data further suggest that the four channels may interact in an opponent process manner. The results support the idea that the visual internal representation of stimuli is based on spatial frequency analysis rather than feature extraction.     

Spatial frequencies in short-term memory for faces: A test of three frequency-dependent hypotheses

Spatial frequency manipulations have been shown to have significant utility in ascertaining the various types of information that might be important in object identification and recognition tasks. This utility suggests that, given some mapping between ranges of spatial frequencies and different types of psychological information, it should be possible to examine the roles of these different types of psychological information by way of spatial frequency manipulations. One potential problem, however, is that there is no well-specified, unambiguous mapping between the distinctions in the frequency domain and the distinctions in the informational domain. Three experiments provide tests of three general hypotheses regarding the ways in which different spatial frequencies might map to different information in facial perception and memory tasks: (1) the low-frequency dominance hypothesis, which proposes that low-frequency information should be superior (to high-frequency information) as a cue to perception and memory; (2) the distinct informational roles hypothesis, which holds that high spatial frequencies should carry featural information while low spatial frequencies should carry information about the configuration of those features; and (3) the task-dependent information hypothesis, which suggests that high-frequency information should be best suited to discrimination tasks while low-frequency information should be best suited for recognition tasks. Results generally contradict the first two of these hypotheses, while providing support for the third. Implications with regard to the various issues related to the mapping between spatial frequencies and the informational content of faces, as well as the need to consider important interactions among perceptual and memory processes, are discussed.     

Facelikeness matters: A parametric multipart object set to understand the role of spatial configuration in visual recognition

There is a view that faces and objects are processed by different brain mechanisms. Different factors may modulate the extent to which face mechanisms are used for objects. To distinguish these factors, we present a new parametric multipart three-dimensional object set that provides researchers with a rich degree of control of important features for visual recognition such as individual parts and the spatial configuration of those parts. All other properties being equal, we demonstrate that perceived facelikeness in terms of spatial configuration facilitated performance at matching individual exemplars of the new object set across viewpoint changes (Experiment 1). Importantly, facelikeness did not affect perceptual discriminability (Experiment 2) or similarity (Experiment 3). Our findings suggest that perceptual resemblance to faces based on spatial configuration of parts is important for visual recognition even after equating physical and perceptual similarity. Furthermore, the large parametrically controlled object set and the standardized procedures to generate additional exemplars will provide the research community with invaluable tools to further understand visual recognition and visual learning.