Alphabetic letter identification: effects of perceivability, similarity, and bias

Does stochastic resonance improve performance for individuals with higher autism-spectrum quotient?

While noise is generally believed to impair performance, the detection of weak stimuli can sometimes be enhanced by introducing optimum noise levels. This phenomenon is termed 'Stochastic Resonance' (SR). Past evidence suggests that autistic individuals exhibit higher neural noise than neurotypical individuals. It has been proposed that the enhanced performance in Autism Spectrum Disorder (ASD) on some tasks could be due to SR. Here we present a computational model, lab-based, and online visual identification experiments to find corroborating evidence for this hypothesis in individuals without a formal ASD diagnosis. Our modeling predicts that artificially increasing noise results in SR for individuals with low internal noise (e.g., neurotypical), however not for those with higher internal noise (e.g., autistic, or neurotypical individuals with higher autistic traits). It also predicts that at low stimulus noise, individuals with higher internal noise outperform those with lower internal noise. We tested these predictions using visual identification tasks among participants from the general population with autistic traits measured by the Autism-Spectrum Quotient (AQ). While all participants showed SR in the lab-based experiment, this did not support our model strongly. In the online experiment, significant SR was not found, however participants with higher AQ scores outperformed those with lower AQ scores at low stimulus noise levels, which is consistent with our modeling. In conclusion, our study is the first to investigate the link between SR and superior performance by those with ASD-related traits, and reports limited evidence to support the high neural noise/SR hypothesis.

Factors Influencing the Accurate Identification of Written Minimal Pairs with Graphemic Similarity: Evidence from Persian-Speaking Children and Adults

In this study, we compared children's and adults' ability to accurately identify target words in written minimal pairs (WMPs) with graphemically similar letters while accounting for factors such as gender, similarity of the middle letter in WMPs, mono- versus dimorphemic WMPs, number of syllable, homography, and imageability. Fifty children and fifty adults were exposed to a distractor stimulus as a pre-mask, followed by the target, and then a post-mask stimulus. Subsequently, the corresponding WMPs including the target word and its graphemically minimal contrast were presented to the participants to obtain their reaction time (RT) in accurately identifying the target word. Results demonstrated that children tend to slow down their reaction as a compensatory strategy to circumvent their less mature knowledge of graphophonic units/morphemes to achieve accuracy during word recognition. In addition, among all controlled factors, children's RT was significantly influenced by similarity of the middle letter in the WMPs. Adults' RT, however, was influenced by factors such as gender, similarity of the middle letter in WMPs, and homography.

General object-based features account for letter perception

After years of experience, humans become experts at perceiving letters. Is this visual capacity attained by learning specialized letter features, or by reusing general visual features previously learned in service of object categorization? To explore this question, we first measured the perceptual similarity of letters in two behavioral tasks, visual search and letter categorization. Then, we trained deep convolutional neural networks on either 26-way letter categorization or 1000-way object categorization, as a way to operationalize possible specialized letter features and general object-based features, respectively. We found that the general object-based features more robustly correlated with the perceptual similarity of letters. We then operationalized additional forms of experience-dependent letter specialization by altering object-trained networks with varied forms of letter training; however, none of these forms of letter specialization improved the match to human behavior. Thus, our findings reveal that it is not necessary to appeal to specialized letter representations to account for perceptual similarity of letters. Instead, we argue that it is more likely that the perception of letters depends on domain-general visual features.

For over a century, scientists have conducted behavioral experiments to investigate how the visual system recognizes letters, but it has proven difficult to propose a model of the feature space underlying this capacity. Here we leveraged recent advances in machine learning to model a wide variety of features ranging from specialized letter features to general object-based features. Across two large-scale behavioral experiments we find that general object-based features account well for letter perception, and that adding letter specialization did not improve the correspondence to human behavior. It is plausible that the ability to recognize letters largely relies on general visual features unaltered by letter learning.

Letter visual similarity of Japanese hiragana and katakana based on reaction times

Letter similarity (i.e., perceptual distance) is a critical measure to better understand letter perception and literacy development. Despite its importance, however, measurements of letter similarity for non-alphabetic scripts are limited, and the shortage of letter similarity for non-alphabetic script interferes with the identification of the universality and the uniqueness of letter perception systems across different scripts. In the present study, we provide a comprehensive matrix of letter similarity for Japanese kana letters (hiragana and katakana). We obtained the discrimination reaction times for simultaneously presented letter pairs and calculated the perceptual distance of 4,278 letter pairs by inversing the time. The matrix showed significant correlations with previously obtained letter similarity for hiragana and katakana. An additional experiment showed that letter pairs for the same sounds (え–エ) produced significantly slower responses compared with those for different sounds (え–コ). However, the differences in reaction times between the same and different sound conditions were smaller than the sequentially presented conditions, suggesting that the matrix was partially attributable to knowledge-based factors (e.g., letter-sound knowledge). This first comprehensive matrix of letter similarity (i.e., perceptual distance) for Japanese kana letters (hiragana and katakana) will be useful for researchers interested in letter perception and literacy development.

The effect of response biases on resolution thresholds of Sloan letters in central and paracentral vision

Sloan letters are one of the most commonly used optotypes in clinical practice. Sloan letters have different relative legibility which could be due to three factors: perceivability, response bias, and similarity. Similarities between Sloan letters are known to be the major source of errors in threshold determination. However, little is known about the effect of response biases on the resolution thresholds. The aim of the present study was to investigate the effect of response bias and similarity on resolution thresholds of Sloan letters in central and paracentral vision. Eight subjects with normal ocular health participated in this study. Using the method of constant stimuli, we measured resolution thresholds for the Sloan letters set at 0° (central) and ± 3° eccentricity along the vertical meridian of the visual field. We calculated thresholds from data pooled across the 10 Sloan letters (pooled threshold). For further analysis we also calculated thresholds for each of the 10 Sloan letters (individual threshold). Response biases and letter similarities were determined using Luce's choice model. Results showed statistically significant differences between the mean individual thresholds of Sloan letters at the central and the upper visual field, but not at the lower visual field. For equally-sized letters at pooled threshold, unlike letter similarity, response biases showed statistically significant correlations to the differences in individual thresholds at the central, upper and lower visual field locations. For equally legible letters at individual thresholds, response biases and similarities showed no significant correlations to the differences in individual thresholds at the central, the upper and the lower visual field locations. These results suggest that, for equally-sized letters at pooled threshold, the response biases may lead to an underestimation of the pooled threshold, i.e. an overestimation of visual acuity measurements when using Sloan letters.

Graphic complexity in writing systems

A writing system is a graphic code, i.e., a system of standardized pairings between symbols and meanings in which symbols take the form of images that can endure. The visual character of writing implies that written characters have to fit constraints of the human visual system. One aspect of this optimization lays in the graphic complexity of the characters used by scripts. Scripts are sets of graphic characters used for the written form of one language or more. Using computational methods over a large and diverse dataset (over 47,000 characters, from over 133 scripts), we answer three central questions about the visual complexity of written characters and the evolution of writing: (1) What determines character complexity? (2) Can we find traces of evolutionary change in character complexity? (3) Is complexity distributed in a way that makes character recognition easier? Our study suggests that (1) character complexity depends primarily on which linguistic unit the characters encode, and that (2) there is little evidence of evolutionary change in character complexity. Additionally (3) for an individual character, the half which is encountered first while reading tends to be more complex than that which is encountered last.

Matrices of the frequency and similarity of Arabic letters and allographs

Indicators of letter frequency and similarity have long been available for Indo-European languages. They have not only been pivotal in controlling the design of experimental psycholinguistic studies seeking to determine the factors that underlie reading ability and literacy acquisition, but have also been useful for studies examining the more general aspects of human cognition. Despite their importance, however, such indicators are still not available for Modern Standard Arabic (MSA), a language that, by virtue of its orthographic system, presents an invaluable environment for the experimental investigation of visual word processing. This paper presents for the first time the frequencies of Arabic letters and their allographs based on a 40-million-word corpus, along with their similarity/confusability indicators in three domains: (1) the visual domain, based on human ratings; (2) the auditory domain, based on an analysis of the phonetic features of letter sounds; and (3) the motoric domain, based on an analysis of the stroke features used to write letters and their allographs. Taken together, the frequency and similarity of Arabic letters and their allographs in the visual and motoric domains, as well as the similarities among the letter sounds, will be useful for researchers interested in the processes underpinning orthographic processing, visual word recognition, reading, and literacy acquisition.

A compositional neural code in high-level visual cortex can explain jumbled word reading

We read jubmled wrods effortlessly, but the neural correlates of this remarkable ability remain poorly understood. We hypothesized that viewing a jumbled word activates a visual representation that is compared to known words. To test this hypothesis, we devised a purely visual model in which neurons tuned to letter shape respond to longer strings in a compositional manner by linearly summing letter responses. We found that dissimilarities between letter strings in this model can explain human performance on visual search, and responses to jumbled words in word reading tasks. Brain imaging revealed that viewing a string activates this letter-based code in the lateral occipital (LO) region and that subsequent comparisons to stored words are consistent with activations of the visual word form area (VWFA). Thus, a compositional neural code potentially contributes to efficient reading.

Spatially Specific Attention Mechanisms Are Sensitive to Competition during Visual Search

Extensive studies have focused on selection mechanisms during visual search. One important influence on these mechanisms is the perceptual characteristics of the stimuli. We investigated the impact of perceptual similarity between targets and nontargets (T-N similarity) in a visual search task using EEG. Participants searched for a predefined target letter among five nontargets. The T-N similarity was manipulated with three levels: high, middle, and low. We tested for the influences of T-N similarity on an ERP (e.g., N2pc) and alpha oscillations. We observed a significant N2pc effect across all levels of similarity. The N2pc amplitude was reduced and occurred later for high similarity relative to low and middle similarities. We also showed that the N2pc amplitude was inversely correlated with the RTs across all similarities. Importantly, we found a significant alpha phase adjustment about the same time as the N2pc for high similarity; by contrast, no such effect was observed for middle and low similarities. Finally, we showed a positive correlation between the phase-locking value and the N2pc—the stronger the alpha phase-locking value, the larger the N2pc, when the T-N similarity was high. In conclusion, our results provide novel evidence for multiple competitive mechanisms during visual search.

Working memory training and perceptual discrimination training impact overlapping and distinct neurocognitive processes: Evidence from event-related potentials and transfer of training gains

There is emerging evidence that working memory (WM) can potentially be enhanced via targeted training protocols. However, the differential effects of targeted training of WM vs. training of general attentional processes on distinct neurocognitive mechanisms is not well understood. In the present study, we compared adaptive n-back WM training to an adaptive visual search training task that targeted perceptual discrimination, in the absence of demands on WM. The search task was closely matched to the n-back task on difficulty and participant engagement. The training duration for both protocols was 20 sessions over approximately 4 weeks. Before and after training, young adult participants were tested on a battery of cognitive tasks to examine transfer of training gains to untrained tests of WM, processing speed, cognitive control, and fluid intelligence. Event-related brain potential (ERP) measures obtained during a Letter 3-Back task and a Search task were examined to determine the neural processes that were affected by each training protocol. Both groups improved on measures of cognitive control and fluid intelligence at post- compared to pretest. However, n-back training resulted in more pronounced transfer effects to tasks involving WM compared to search training. With respect to ERPs, both groups exhibited enhancement of P3 amplitude following training, but distinct changes in neural responses were also observed for the two training protocols. The search training group exhibited earlier ERP latencies at post- compared to pretest on the Search task, indicating generalized improvement in processing speed. The n-back group exhibited a pronounced enhancement and earlier latency of the N2 ERP component on the Letter 3-back task, following training. Given the theoretical underpinnings of the N2, this finding was interpreted as an enhancement of conflict monitoring and sequential mismatch identification. The findings provide evidence that n-back training enhances distinct neural processes underlying executive aspects of WM.

A Hierarchical Model of Inhibitory Control

Inhibitory control describes the suppression of goal-irrelevant stimuli and behavioral responses. Current developmental taxonomies distinguish between Response Inhibition – the ability to suppress a prepotent motor response, and Attentional Inhibition – the ability to resist interference from distracting stimuli. Response Inhibition and Attentional Inhibition have exhibited moderately strong positive correlations in previous studies, suggesting they are closely related cognitive abilities. These results may reflect the use of cognitive tasks combining Stimulus–Stimulus- and Stimulus–Response-conflict as indicators of both constructs, which may have conflated their empirical association. Additionally, previous statistical modeling studies have not controlled for individual differences in Working Memory Capacity, which may account for some of the empirical overlap between Response Inhibition and Attentional Inhibition. The aim of the current study was to test a hierarchical model of inhibitory control that specifies Working Memory Capacity as a higher-order cognitive construct. Response Inhibition and Attentional Inhibition were conceptualized as lower-order cognitive mechanisms that should be empirically independent constructs apart from their shared reliance on Working Memory Capacity for active maintenance of goal-relevant representations. Measures of performance on modified stimulus–response compatibility tasks, complex memory span, and non-selective stopping tasks were obtained from 136 preadolescent children (M = 11 years, 10 months, SD = 8 months). Consistent with hypotheses, results from Structural Equation Modeling demonstrated that the Response Inhibition and Attentional Inhibition factors were empirically independent constructs that exhibited partial statistical dependence on the Working Memory Capacity factor. These findings have important implications for current theories and models of inhibitory control during development.

GraphCom: A multidimensional measure of graphic complexity applied to 131 written languages

We report a new multidimensional measure of visual complexity (GraphCom) that captures variability in the complexity of graphs within and across writing systems. We applied the measure to 131 written languages, allowing comparisons of complexity and providing a basis for empirical testing of GraphCom. The measure includes four dimensions whose value in capturing the different visual properties of graphs had been demonstrated in prior reading research-(1) perimetric complexity, sensitive to the ratio of a written form to its surrounding white space (Pelli, Burns, Farell, & Moore-Page, 2006); (2) number of disconnected components, sensitive to discontinuity (Gibson, 1969); (3) number of connected points, sensitive to continuity (Lanthier, Risko, Stolz, & Besner, 2009); and (4) number of simple features, sensitive to the strokes that compose graphs (Wu, Zhou, & Shu, 1999). In our analysis of the complexity of 21,550 graphs, we (a) determined the complexity variation across writing systems along each dimension, (b) examined the relationships among complexity patterns within and across writing systems, and (c) compared the dimensions in their abilities to differentiate the graphs from different writing systems, in order to predict human perceptual judgments (n = 180) of graphs with varying complexity. The results from the computational and experimental comparisons showed that GraphCom provides a measure of graphic complexity that exceeds previous measures in its empirical validation. The measure can be universally applied across writing systems, providing a research tool for studies of reading and writing.

Is nevtral NEUTRAL? Visual similarity effects in the early phases of written-word recognition

For simplicity, contemporary models of written-word recognition and reading have unspecified feature/letter levels-they predict that the visually similar substituted-letter nonword PEQPLE is as effective at activating the word PEOPLE as the visually dissimilar substituted-letter nonword PEYPLE. Previous empirical evidence on the effects of visual similarly across letters during written-word recognition is scarce and nonconclusive. To examine whether visual similarity across letters plays a role early in word processing, we conducted two masked priming lexical decision experiments (stimulus-onset asynchrony = 50 ms). The substituted-letter primes were visually very similar to the target letters (u/v in Experiment 1 and i/j in Experiment 2; e.g., nevtral-NEUTRAL). For comparison purposes, we included an identity prime condition (neutral-NEUTRAL) and a dissimilar-letter prime condition (neztral-NEUTRAL). Results showed that the similar-letter prime condition produced faster word identification times than the dissimilar-letter prime condition. We discuss how models of written-word recognition should be amended to capture visual similarity effects across letters.

The role of allograph representations in font-invariant letter identification

The literate brain must contend with countless font variants for any given letter. How does the visual system handle such variability? One proposed solution posits stored structural descriptions of basic letter shapes that are abstract enough to deal with the many possible font variations of each letter. These font-invariant representations, referred to as allographs in this paper, while frequently posited, have seldom been empirically evaluated. The research reported here helps to address this gap with 2 experiments that examine the possible influence of allograph representations on visual letter processing. In these experiments, participants respond to pairs of letters presented in an atypical font in 2 tasks-visual similarity judgments (Experiment 1) and same/different decisions (Experiment 2). By using representational similarity analysis (RSA) in conjunction with linear mixed effect models (LMEM; RSA-LMEM) we show that the similarity structure of the responses to the atypical font is influenced by the predicted similarity structure of allograph representations even after accounting for font-specific visual shape similarity. Similarity due to symbolic (abstract) identity, name, and motor representations of letters are also taken into account providing compelling evidence for the unique influence of allograph representations in these tasks. These results provide support for the role of allograph representations in achieving font-invariant letter identification. (PsycINFO Database Record

Intact word processing in developmental prosopagnosia

A wealth of evidence from behavioural, neuropsychological and neuroimaging research supports the view that face recognition is reliant upon a domain-specific network that does not process words. In contrast, the recent many-to-many model of visual recognition posits that brain areas involved in word and face recognition are functionally integrated. Developmental prosopagnosia (DP) is characterised by severe deficits in the recognition of faces, which the many-to-many model predicts should negatively affect word recognition. Alternatively, domain-specific accounts suggest that impairments in face and word processing need not go hand in hand. To test these possibilities, we ran a battery of 7 tasks examining word processing in a group of DP cases and controls. One of our prosopagnosia cases exhibited a severe reading impairment with delayed response times during reading aloud tasks, but not lexical decision tasks. Overall, however, we found no evidence of global word processing deficits in DP, consistent with a dissociation account for face and word processing.

A New Font, Specifically Designed for Peripheral Vision, Improves Peripheral Letter and Word Recognition, but Not Eye-Mediated Reading Performance

Reading speed is dramatically reduced when readers cannot use their central vision. This is because low visual acuity and crowding negatively impact letter recognition in the periphery. In this study, we designed a new font (referred to as the Eido font) in order to reduce inter-letter similarity and consequently to increase peripheral letter recognition performance. We tested this font by running five experiments that compared the Eido font with the standard Courier font. Letter spacing and x-height were identical for the two monospaced fonts. Six normally-sighted subjects used exclusively their peripheral vision to run two aloud reading tasks (with eye movements), a letter recognition task (without eye movements), a word recognition task (without eye movements) and a lexical decision task. Results show that reading speed was not significantly different between the Eido and the Courier font when subjects had to read single sentences with a round simulated gaze-contingent central scotoma (10° diameter). In contrast, Eido significantly decreased perceptual errors in peripheral crowded letter recognition (-30% errors on average for letters briefly presented at 6° eccentricity) and in peripheral word recognition (-32% errors on average for words briefly presented at 6° eccentricity).

The effects of alphabet and expertise on letter perception

Long-standing questions in human perception concern the nature of the visual features that underlie letter recognition and the extent to which the visual processing of letters is affected by differences in alphabets and levels of viewer expertise. We examined these issues in a novel approach using a same-different judgment task on pairs of letters from the Arabic alphabet with 2 participant groups: 1 with no prior exposure to Arabic and 1 with reading proficiency. Hierarchical clustering and linear mixed-effects modeling of reaction times and accuracy provide evidence that both the specific characteristics of the alphabet and observers' previous experience with it affect how letters are perceived and visually processed. The findings of this research further our understanding of the multiple factors that affect letter perception and support the view of a visual system that dynamically adjusts its weighting of visual features as expert readers come to more efficiently and effectively discriminate the letters of the specific alphabet they are viewing. (PsycINFO Database Record

Residents' numeric inputting error in computerized physician order entry prescription

BACKGROUND

Computerized physician order entry (CPOE) system with embedded clinical decision support (CDS) can significantly reduce certain types of prescription error. However, prescription errors still occur. Various factors such as the numeric inputting methods in human computer interaction (HCI) produce different error rates and types, but has received relatively little attention.

OBJECTIVE

This study aimed to examine the effects of numeric inputting methods and urgency levels on numeric inputting errors of prescription, as well as categorize the types of errors.

METHODS

Thirty residents participated in four prescribing tasks in which two factors were manipulated: numeric inputting methods (numeric row in the main keyboard vs. numeric keypad) and urgency levels (urgent situation vs. non-urgent situation). Multiple aspects of participants' prescribing behavior were measured in sober prescribing situations.

RESULTS

The results revealed that in urgent situations, participants were prone to make mistakes when using the numeric row in the main keyboard. With control of performance in the sober prescribing situation, the effects of the input methods disappeared, and urgency was found to play a significant role in the generalized linear model. Most errors were either omission or substitution types, but the proportion of transposition and intrusion error types were significantly higher than that of the previous research. Among numbers 3, 8, and 9, which were the less common digits used in prescription, the error rate was higher, which was a great risk to patient safety.

CONCLUSIONS

Urgency played a more important role in CPOE numeric typing error-making than typing skills and typing habits. It was recommended that inputting with the numeric keypad had lower error rates in urgent situation. An alternative design could consider increasing the sensitivity of the keys with lower frequency of occurrence and decimals. To improve the usability of CPOE, numeric keyboard design and error detection could benefit from spatial incidence of errors found in this study.

Context-dependent similarity effects in letter recognition

In visual word recognition tasks, digit primes that are visually similar to letter string targets (e.g., 4/A, 8/B) are known to facilitate letter identification relative to visually dissimilar digits (e.g., 6/A, 7/B); in contrast, with letter primes, visual similarity effects have been elusive. In the present study we show that the visual similarity effect with letter primes can be made to come and go, depending on whether it is necessary to discriminate between visually similar letters. The results support a Bayesian view which regards letter recognition not as a passive activation process driven by the fixed stimulus properties, but as a dynamic evidence accumulation process for a decision that is guided by the task context.

Using 10AFC to further improve the efficiency of the quick CSF method

The contrast sensitivity function (CSF) provides a fundamental characterization of spatial vision, important for basic and clinical applications, but its long testing times have prevented easy, widespread assessment. The original quick CSF method was developed using a two-alternative forced choice (2AFC) grating orientation identification task (Lesmes, Lu, Baek, & Albright, 2010), and obtained precise CSF assessments while reducing the testing burden to only 50 trials. In this study, we attempt to further improve the efficiency of the quick CSF method by exploiting the properties of psychometric functions in multiple-alternative forced choice (m-AFC) tasks. A simulation study evaluated the effect of the number of alternatives m on the efficiency of the sensitivity measurement by the quick CSF method, and a psychophysical study validated the quick CS method in a 10AFC task. We found that increasing the number of alternatives of the forced-choice task greatly improved the efficiency of CSF assessment in both simulation and psychophysical studies. The quick CSF method based on a 10-letter identification task can assess the CSF with an averaged standard deviation of 0.10 decimal log unit in less than 2 minutes.

Visual similarity effects on masked priming

We investigated the role of the visual similarity of masked primes to targets in a lexical decision experiment. In the primes, some letters in the target (e.g., A in ABANDON) had either visually similar letters (e.g., H), dissimilar letters (D), visually similar digits (4), or dissimilar digits (6) substituted for them. The similarities of the digits and letters to the base letter were equated and verified in a two-alternative forced choice (2AFC) perceptual identification task. Using targets presented in lowercase (e.g., abandon) and primes presented in uppercase, visually similar digit primes (e.g., 484NDON) produced more priming than did visually dissimilar digit primes (676NDON), but little difference was found between the visually similar and dissimilar letter primes (HRHNDON vs. DWDNDON). These results were explained in terms of task-driven competition between the target letter and the visually similar letter.

The influence of reading expertise in mirror-letter perception: Evidence from beginning and expert readers

The visual word recognition system recruits neuronal systems originally developed for object perception which are characterized by orientation insensitivity to mirror reversals. It has been proposed that during reading acquisition beginning readers have to "unlearn" this natural tolerance to mirror reversals in order to efficiently discriminate letters and words. Therefore, it is supposed that this unlearning process takes place in a gradual way and that reading expertise modulates mirror-letter discrimination. However, to date no supporting evidence for this has been obtained. We present data from an eye-movement study that investigated the degree of sensitivity to mirror-letters in a group of beginning readers and a group of expert readers. Participants had to decide which of the two strings presented on a screen corresponded to an auditorily presented word. Visual displays always included the correct target word and one distractor word. Results showed that those distractors that were the same as the target word except for the mirror lateralization of two internal letters attracted participants' attention more than distractors created by replacement of two internal letters. Interestingly, the time course of the effects was found to be different for the two groups, with beginning readers showing a greater tolerance (decreased sensitivity) to mirror-letters than expert readers. Implications of these findings are discussed within the framework of preceding evidence showing how reading expertise modulates letter identification.