A model of character recognition and legibility

This article presents a model of character recognition and the experiments used to develop and test it. The model applies to foveal viewing of blurred or unblurred characters and to tactile sensing of raised characters using the fingerpad. The primary goal of the model is to account for variations in legibility across character sets; a secondary goal is to account for variations in the cell entries of the confusion matrix for a given character set. The model consists of two distinct processing stages. The first involves transformation of each stimulus into an internal representation; this transformation consists of linear low-pass spatial filtering followed by nonlinear compression of stimulus intensity. The second stage involves both template matching of the transformed test stimulus with each of the stored internal representations of the characters within the set and response selection, which is assumed to conform to the unbiased choice model of Luce (1963). Though purely stimulus driven, the model accounts quite well for differences in the legibility of character sets differing in character type, size of character, and number of characters within the set; it is somewhat less successful in accounting for the details of each confusion matrix.     

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.     

The effect of orientation on tactual braille recognition: optimal touching positions.

Subjects in five experiments matched tangible braille against a visible matching code. In Experiment 1, braille recognition suffered when entire lines of braille characters were tilted in varying amounts from the upright. Experiment 2 showed that tilt lowered performance for tangible, large embossed letters, as well as for braille. However, recognition was better for print letters than it was for braille. In Experiment 3, subjects attempted to match the upright array against embossed braille that was left/right reversed, inverted up/down, or rotated +180 degrees. Performance was close to that for normal braille in the left/right reversal condition, and very low for the +180 degrees rotation group. These results on braille tilt in the "picture plane" may reflect difficulty in manipulating the tangible "image." Braille recognition performance was not lowered when the visible matching array was tilted -45 degrees or -90 degrees from the upright but the tangible stimuli were upright. In Experiment 4, recognition of left/right reversed braille that was physically horizontal (on the bottom of a shelf) was compared with that of braille left/right reversed due to its location on the back of a panel, in the vertical plane. Braille recognition accuracy was higher with braille located vertically. An additional experiment showed the beneficial effect of locating braille in the vertical, frontoparallel plane, obtained with +90 degree rotated braille. It is proposed that optimal tactual performance with tangible arrays might depend on touching position, and on the physical position of stimuli in space. Just as there are good and poor viewing positions, there may be optimal touching positions.(ABSTRACT TRUNCATED AT 250 WORDS)     

The effect of orientation on tactual braille recognition: Optimal touching positions

Subjects in five experiments matched tangible braille against a visible matching code. In Experiment 1, braille recognition suffered when entire lines of braille characters were tilted in varying amounts from the upright. Experiment 2 showed that tilt lowered performance for tangible, large embossed letters, as well as for braille. However, recognition was better for print letters than it was for braille. In Experiment 3, subjects attempted to match the upright array against embossed braille that was left/right reversed, inverted up/down, or rotated +180°. Performance was close to that for normal braille in the left/right reversal condition, and very low for the +180° rotation group. These results on braille tilt in the “picture plane” may reflect difficulty in manipulating the tangible “image.” Braille recognition performance was not lowered whenthe visible matching array was tilted ?45° or ?90° from the upright but the tangible stimuli were upright. In Experiment 4, recognition of left/right reversed braille that was physically horizontal (on the bottom of a shelf) was compared with that of braille left/right reversed due to its location on the back of a panel, in the vertical plane. Braille recognition accuracy was higher with braille located vertically. An additional experiment showed the beneficial effect of locating braille in the vertical, frontoparallel plane, obtained with +90° degree rotated braille. It is proposed that optimal tactual performance with tangible arrays might depend on touching position, and on the physical position of stimuli in space. Just as there are good and poor viewing positions, there may be optimal touching positions. The effects of tilt on braille identification were diminished for blind subjects, suggesting the importance of tactile experience and skill.     

Teaching identity matching of braille characters to beginning braille readers

We taught three children with visual impairments to make tactile discriminations of the braille alphabet within a matching‐to‐sample format. That is, we presented participants with a braille character as a sample stimulus, and they selected the matching stimulus from a three‐comparison array. In order to minimize participant errors, we initially arranged braille characters into training sets in which there was a maximum difference in the number of dots comprising the target and nontarget comparison stimuli. As participants mastered these discriminations, we increased the similarity between target and nontarget comparisons (i.e., an approximation of stimulus fading). All three participants’ accuracy systematically increased following the introduction of this identity‐matching procedure.     

How reading braille is both like and unlike reading print

The letter detection errors of blind readers of braille were compared with those of sighted readers of print in order to determine whether braille can be read in units larger than the individual braille character. Sighted readers missed a disproportionate number ofhs in the wordhe whether the prose was intact or scrambled, supporting Healy’s (1976) conclusion that high-frequency words are read in units larger than the individual letter. By contrast, blind readers missed only a disproportionate number ofhs inhe when reading prose, suggesting that their detection failures were based on redundancy rather than unitization. Although the unit of perception for braille appears to differ from that for print, the factors underlying braille comprehension ability do not. Braille comprehension was correlated with listening comprehension and working memory capacity, a finding consistent with the visual reading literature which has shown that it is the higher level linguistic processes that tax working memory capacity and not the lower level visual word encoding processes that chiefly underlie individual differences in comprehending print.     

Acquisition and generative responding following print‐to‐braille construction response training with sighted learners

Previous researchers have taught sighted adults to match braille sample stimuli to print comparisons in a matching‐to‐sample (MTS) format and assessed the emergence of other braille repertoires, such as transcribing and reading braille following this training. Although participants learned to match to sample with braille, they displayed limited emergence of other braille repertoires. Lack of generative responding may have resulted from participants' over‐selective attending to components of compound braille characters during instruction. We taught undergraduates to construct braille characters given a print sample, which required attending to each individual braille symbol, and assessed generative braille responding. Participants met mastery of 378 braille construction responses and demonstrated modestly improved responding compared with previous research.     

Familiarity does not affect the unilateral field advantage for repetition detection

We have previously reported evidence that repetitions of letters, colors, sizes, and common motion paths are more rapidly detected when they are presented unilaterally (i.e., both in the same visual field) versus bilaterally (one element in each visual field; Butcher and Cavanagh (Attention, Perception, & Psychophysics 70:714?C724, 2008). Here, we report evidence that this unilateral field advantage (UFA) for repetition detection does not depend on prior experience with the elements that comprise the repetition. In Experiment 1, native English, Persian, and Japanese speakers were tested on a repetition detection task involving characters from Western, Arabic, and Japanese character sets. The character sets were tested in blocks, in each of which subjects were presented with four characters for 16 ms and asked to report whether any two of the characters were identical. The subjects were faster detecting repetitions that were presented unilaterally rather than bilaterally, and there was no interaction with stimulus familiarity. A second experiment replicated this finding with native English speakers only, using a longer stimulus duration (150 ms). We had previously proposed that the UFA arises because the low-level processes that group physically identical items operate more efficiently within than across hemifields. Our data now indicate that this grouping process is insensitive to item familiarity, supporting the claim that the process is low-level.     

Fuzzy picture processing: effects of size reduction and blurring on emotional processing

Previous studies have suggested that picture size reduction affects emotional reactions, possibly because scenes subtending a small visual angle are perceived as being more distant and less relevant compared to larger stimuli. However, pictures that subtend a small visual angle also contain few fine-grained details, which may determine less vivid representations and responses compared to larger and more detailed images. Critically, the present study compared two different types of manipulations, namely size reduction and low-pass spatial filtering, which determined similar detail loss but affected visual angles differently. Affective modulation was assessed using an evaluative task and a behavioral interference task. Results showed that the availability of fine-grained details, independently of visual angle, modulated emotional evaluation. Moreover, interference in an unrelated task was unaffected by either size reduction or low-pass spatial filtering. These findings suggest that high spatial frequencies affect subjective emotional response whereas attentional capture by affective stimuli seems to rely on information that is sufficient to allow a categorization of picture content.     

Two types of word superiority effects in a speeded matching task

Two experiments examined unitization in word and nonword four-letter sequences and word superiority effects (WSE) usingsame-different reaction time tasks. Stimuli consisted of letters presented directly above a subset of letters within four-letter sequences (e.g., \(\mathop {ABCD}\limits^{ABC} \) ). Probes either matched the sequence letters or differed in a single letter, and all possible probe configurations were used as stimuli. When complex probe configurations were matched as rapidly as single-letter probes, it was taken as evidence suggestive of possible multiletter unitization. Results indicated that a WSE occurred simultaneously with apparent unitization differences between word and nonword contexts only when entire four-letter sequences were matched. However, other WSEs were found, suggesting that familiarity affected processing efficiency of both postperceptual inference and perceptual analysis within the same task.     

Detecting distortions of peripherally presented letter stimuli under crowded conditions

When visual features in the periphery are close together they become difficult to recognize: something is present but it is unclear what. This is called “crowding”. Here we investigated sensitivity to features in highly familiar shapes (letters) by applying spatial distortions. In Experiment 1, observers detected which of four peripherally presented (8 deg of retinal eccentricity) target letters was distorted (spatial 4AFC). The letters were presented either isolated or surrounded by four undistorted flanking letters, and distorted with one of two types of distortion at a range of distortion frequencies and amplitudes. The bandpass noise distortion (“BPN”) technique causes spatial distortions in Cartesian space, whereas radial frequency distortion (“RF”) causes shifts in polar coordinates. Detecting distortions in target letters was more difficult in the presence of flanking letters, consistent with the effect of crowding. The BPN distortion type showed evidence of tuning, with sensitivity to distortions peaking at approximately 6.5 c/deg for unflanked letters. The presence of flanking letters causes this peak to rise to approximately 8.5 c/deg. In contrast to the tuning observed for BPN distortions, RF distortion sensitivity increased as the radial frequency of distortion increased. In a series of follow-up experiments, we found that sensitivity to distortions is reduced when flanking letters were also distorted, that this held when observers were required to report which target letter was undistorted, and that this held when flanker distortions were always detectable. The perception of geometric distortions in letter stimuli is impaired by visual crowding.     

Asymmetric binding in serial memory for verbal and spatial information

As the number of studies showing that items can be retained as bound representations in memory increases, researchers are beginning to investigate how the different features are bound together. In the present study, we examined the relative importances of the verbal and spatial features in serial memory for visual stimuli. Participants were asked to memorize the order of series of letters presented visually in different locations on the computer screen. The results showed that manipulating the phonological similarity of the letters affected recall of their spatial locations, but that increasing the complexity of the spatial pattern had no effect on recall of the letters. This finding was observed in both order reconstruction (Exps. 1 and 2) and probe serial recall (Exps. 3 and 4), suggesting that verbal–spatial binding in serial memory for visual information is asymmetric.     

Evidence for scanning with unilateral visual presentation of letters

When letters and words are presented tachistoscopically, material from the right visual field (RVF) can be reported more accurately than that from the left visual field (LVF). The RVF superiority may reflect either left hemispheric dominance for language or directional scanning. Previous studies have deliberately focused on the cerebral asymmetry factor while "controlling" scanning and, thus, have cast some doubt on the potency of the scanning factor. Two experiments were conducted to show that scanning can induce a RVF superiority comparable to that often associated with cerebral asymmetry. The first experiment required bilingual subjects to report six English or six Hebrew letters, shown briefly in either the LVF or RVF, with order of report controlled. A RVF superiority found with English characters was matched by an equal but opposite LVF effect with Hebrew. In a second experiment, five English characters were shown briefly in either the LVF or RVF, and subjects had to identify a single character indicated by a post exposural cue. Using a spatial cue to by pass scanning, there were no field differences; with an ordinal position cue--a procedure thought to force scanning--there was a strong RVF superiority. The results show clearly that scanning can induce visual field differences.     

Memory and learning for visual signals in time and space

Vision is often characterized as a spatial sense, but what does that characterization imply about the relative ease of processing visual information distributed over time rather than over space? Three experiments addressed this question, using stimuli comprising random luminances. For some stimuli, individual items were presented sequentially, at 8 Hz; for other stimuli, individual items were presented simultaneously, as horizontal spatial arrays. For temporal sequences, subjects judged whether each of the last four luminances matched the corresponding luminance in the first four; for spatial arrays, they judged whether each of the right-hand four luminances matched the corresponding left-hand luminance. Overall, performance was far better with spatial presentations, even when the entire spatial array was presented for just tens of milliseconds. Experiment 2 demonstrated that there was no gain in performance from combining spatial and temporal information within a single stimulus. In a final experiment, particular spatial arrays or temporal sequences were made to recur intermittently, interspersed among, non-recurring stimuli. Performance improved steadily as particular stimulus exemplars recurred, with spatial and temporal stimuli being learned at equivalent rates. Logistic regression identified several shortcut strategies that subjects may have exploited while performing our task.     

A computer-based program to teach braille reading to sighted individuals

Instructors of the visually impaired need efficient braille‐training methods. This study conducted a preliminary evaluation of a computer‐based program intended to teach the relation between braille characters and English letters using a matching‐to‐sample format with 4 sighted college students. Each participant mastered matching visual depictions of the braille alphabet to their printed‐word counterparts. Further, each participant increased the number of words they read in a braille passage following this training. These gains were maintained at variable levels on a maintenance probe conducted 2 to 4 weeks after training.     

Flying to Neverland: How readers tacitly judge norms during comprehension

As readers gain experience with specific narrative worlds, they accumulate information that allows them to experience events as normal or unusual within those worlds. In this article, we contrast two accounts for how readers access information about specific narrative worlds to make tacit judgments of normalcy. We conducted two experiments. In Experiment 1, participants read stories about an ordinary character (e.g., a police officer in Boston) or a familiar fantastic character (e.g., Superman). Each story described a realistic event (e.g., the character being killed by bullets) or a fantastic event (e.g., bullets bouncing off the character’s chest). Participants were faster to read events that were consistent with their prior knowledge about the story world. In Experiments 2a and 2b, participants read stories about familiar fantastic characters, unfamiliar fantastic characters (e.g., a Kryptonian named Dev-em), and unfamiliar ordinary characters. In Experiment 2a, participants were equally fast to read about the familiar and unfamiliar fantastic characters experiencing fantastic events, both of which were read faster than the unfamiliar ordinary characters sentences. In Experiment 2b, participants were fastest to read about unfamiliar ordinary characters experiencing realistic events and were equally slow for familiar and unfamiliar fantastic characters. Our experiments provide evidence that readers routinely use inductive reasoning to go beyond their prior knowledge when reading fictional narratives, affecting whether they experience events as normal or unusual.     

Children's left parietal brain activation during mental rotation is reliable as well as specific

Some recent evidence suggests that mental rotation of characters in children aged 7 or 8 years might be lateralized to the left parietal hemisphere. An alternative statement exists, however, the finding might be completely unspecific for mental rotation but either be simply a function of task difficulty or a consequence of the use of characters as stimuli. To test these alternatives, ERPs of 24 second graders were measured twice: (a) during mental rotation with characters as stimuli and orientations of 30°, 90°, or 150° and (b) during memory scanning with characters as stimuli and set sizes of 1, 2, or 3 letters. In both cases, an amplitude modulation was found. The effect of mental rotation as a function of character orientation turned out to be lateralized to the left parietal hemisphere. The effect of memory scanning as a function of set size, however, turned out to be completely non-lateralized. Thus, children's left hemisphere activation during mental rotation is reliable as well as specific.     

The robustness of communication of emotion via facial expression: Emotion recognition from photographs with deteriorated pictorial quality

This study

1 The studies reported here were conducted within the TELEMED project which is funded by the European Community within the RACE program.

examines whether the recognition of emotion from facial expressions is impaired by deterioration of spatial resolution, contrast resolution, and picture size. Eighty judges rated 65 stimuli under 11 conditions: Undistorted, reduced spatial resolution (three steps), reduced contrast resolution (three steps), reduced picture size (three steps), and a very ‘hard’ condition combining the severest spatial and contrast resolution. Variation in picture quality was achieved by using a digital video recorder. Recognition rate and intensity ratings were not significantly affected by variations in contrast resolution or picture size. The only significant reduction of recognition rate and intensity ratings resulted from reduction in spatial resolution, but only with the largest deterioration in such resolution. Results are discussed with respect to the fundamental importance of facial expressions in interaction and communication, and with respect to applications, such as tele-conferencing systems.     

Stimulus averaging and the context for judgment

Loudness judgments of stimulus sets composed of four bursts of noise were analyzed for two types of contextual effects: between-set and within-set. Experiment I demonstrated between-set effects for these four-component stimuli; they were shown to be similar to those found in previous work with single stimuli. Experiment 2 tested an averaging model for within-set contextual effects. The results were inconsistent with the model. One interpretation is that there are within-set effects and that these are caused by shifts in the effective range of stimuli. Alternative interpretations attribute the apparent contextual effects either to an averaging of physical values or to an inappropriate scale of judgment.