Same-different Task Research Articles

Solving the same-different task with convolutional neural networks

Deep learning demonstrated major abilities in solving many kinds of different real-world problems in computer vision literature. However, they are still strained by simple reasoning tasks that humans consider easy to solve. In this work, we probe current state-of-the-art convolutional neural networks on a difficult set of tasks known as the same-different problems. All the problems require the same prerequisite to be solved correctly: understanding if two random shapes inside the same image are the same or not. With the experiments carried out in this work, we demonstrate that residual connections, and more generally the skip connections, seem to have only a marginal impact on the learning of the proposed problems. In particular, we experiment with DenseNets, and we examine the contribution of residual and recurrent connections in already tested architectures, ResNet-18, and CorNet-S respectively. Our experiments show that older feed-forward networks, AlexNet and VGG, are almost unable to learn the proposed problems, except in some specific scenarios. We show that recently introduced architectures can converge even in the cases where the important parts of their architecture are removed. We finally carry out some zero-shot generalization tests, and we discover that in these scenarios residual and recurrent connections can have a stronger impact on the overall test accuracy. On four difficult problems from the SVRT dataset, we can reach state-of-the-art results with respect to the previous approaches, obtaining super-human performances on three of the four problems.

Song Is More Memorable Than Speech Prosody: Discrete Pitches Aid Auditory Working Memory.

Vocal music and spoken language both have important roles in human communication, but it is unclear why these two different modes of vocal communication exist. Although similar, speech and song differ in certain design features. One interesting difference is in the pitch intonation contour, which consists of discrete tones in song, vs. gliding intonation contours in speech. Here, we investigated whether vocal phrases consisting of discrete pitches (song-like) or gliding pitches (speech-like) are remembered better, conducting three studies implementing auditory same-different tasks at three levels of difficulty. We tested two hypotheses: that discrete pitch contours aid auditory memory, independent of musical experience (“song memory advantage hypothesis”), or that the higher everyday experience perceiving and producing speech make speech intonation easier to remember (“experience advantage hypothesis”). We used closely matched stimuli, controlling for rhythm and timbre, and we included a stimulus intermediate between song-like and speech-like pitch contours (with partially gliding and partially discrete pitches). We also assessed participants' musicality to evaluate experience-dependent effects. We found that song-like vocal phrases are remembered better than speech-like vocal phrases, and that intermediate vocal phrases evoked a similar advantage to song-like vocal phrases. Participants with more musical experience were better in remembering all three types of vocal phrases. The precise roles of absolute and relative pitch perception and the influence of top-down vs. bottom-up processing should be clarified in future studies. However, our results suggest that one potential reason for the emergence of discrete pitch–a feature that characterises music across cultures–might be that it enhances auditory memory.

Sequential sampling models of same-different data and how they explain the fast-same effect.

There have been many empirical endeavors to explain the fact that participants correctly identify same stimuli faster than different stimuli. Despite a large body of literature, a conclusive and unanimous consensus to explain this effect is lacking. The present study offers a new perspective on the same-different task using best-fitting parametric values of 2 sequential sampling models (linear ballistic accumulator [LBA] and drift-diffusion model [DDM]) to evaluate the contribution of 3 theoretical concepts (starting point, accumulation rate, and base time) to the fast-same effect. Eighteen participants were recruited to perform 3 tasks in which they compared letter strings (a) in a normal setting, the control task; (b) with lower contrast, the contrast task; and (c) based on nominal identity, the case task. Both the LBA and the DDM suggests that participants have faster base time in the same condition compared with the different condition. They also suggest that the number of letters and the number of differences influence the accumulation rate. These results suggest that the cognitive processes do not accumulate information as efficiently when the workload is increased. Finally, the limited predictive power of the best-fitting models suggests that better theoretical grounds are needed to identify fundamental cognitive concepts underlying same-different judgments. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Masked form priming as a function of letter position: An evaluation of current orthographic coding models.

A word's exterior letters, particularly its initial letter, appear to have a special status when reading. Therefore, most orthographic coding models incorporate assumptions giving initial letters and, in some cases, final letters, enhanced importance during the orthographic coding process. In the present article, 3 masked priming experiments were carried out, using the conventional lexical-decision task, the sandwich priming lexical-decision task and the masked priming same-different task, in an attempt to examine a number of those models with a specific focus on the implications of the models' assumptions concerning the different letter positions. The related primes and targets were 6-letter strings that differed in 2 letter positions, initial (e.g., jnckey-HOCKEY), middle (e.g., hojney-HOCKEY), or final (hockjn-HOCKEY), with the middle-letters different primes being the primes that maintained both end letters. To the extent possible, the predictions of the models were derived by using easyNet, the simulation program recently developed by Adelman, Gubian, and Davis (2017). In all experiments, the final-letters different primes were the most effective primes with there being no clear distinction between the other 2 prime types, a pattern that none of the models predicted. The lack of an advantage for the middle-letters different primes suggests that the orthographic code driving masked priming is not one that places a special emphasis on the identities of the exterior letters. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

The role of the written script in shaping mirror-image discrimination: Evidence from illiterate, Tamil literate, and Tamil-Latin-alphabet bi-literate adults

Learning a script with mirrored graphs (e.g., d ≠ b) requires overcoming the evolutionary-old perceptual tendency to process mirror images as equivalent. Thus, breaking mirror invariance offers an important tool for understanding cultural re-shaping of evolutionarily ancient cognitive mechanisms. Here we investigated the role of script (i.e., presence vs. absence of mirrored graphs: Latin alphabet vs. Tamil) by revisiting mirror-image processing by illiterate, Tamil monoliterate, and Tamil-Latin-alphabet bi-literate adults. Participants performed two same-different tasks (one orientation-based, another shape-based) on Latin-alphabet letters. Tamil monoliterate were significantly better than illiterate and showed good explicit mirror-image discrimination. However, only bi-literate adults fully broke mirror invariance: slower shape-based judgments for mirrored than identical pairs and reduced disadvantage in orientation-based over shape-based judgments of mirrored pairs. These findings suggest learning a script with mirrored graphs is the strongest force for breaking mirror invariance.

Is zjudge a better prime for JUDGE than zudge is?: A new evaluation of current orthographic coding models.

Three masked priming paradigms, the conventional masked priming lexical-decision task (Forster & Davis, 1984), the sandwich priming task (Lupker & Davis, 2009), and the masked priming same-different task (Norris & Kinoshita, 2008), were used to investigate priming for a given target (e.g., JUDGE) from primes created by either adding a letter to the beginning of the target (e.g., zjudge) or replacing the target's initial letter (e.g., zudge). Virtually all models of orthographic coding that allow calculation of orthographic similarity measures predict that zjudge should be the better prime because zjudge contains all the letters in JUDGE in their correct order whereas zudge does not. Nonetheless, Adelman et al.'s (2014) megastudy data indicated no difference in the effectiveness of these two prime types. The present experiments provide additional support for the conclusion of no difference between these two prime types with the only observed difference being a small zudge prime advantage in Experiment 1b (sandwich priming). These results suggest that models of orthographic coding/word recognition may be well served by allowing inconsistent information (e.g., the "z" in both zjudge and zudge indicates that the presented prime is not JUDGE) to be given considerable weight during the orthographic coding/word recognition process. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

How the forest interacts with the trees: Multiscale shape integration explains global and local processing.

Hierarchical stimuli have been widely used to study global and local processing. Two classic phenomena have been observed using these stimuli: the global advantage effect (we identify the global shape faster) and an interference effect (we identify shape slower when the global and local shapes are different). Because these phenomena have been observed during shape categorization tasks, it is unclear whether they reflect the categorical judgment or the underlying shape representation. Understanding the underlying shape representation is also critical because both global and local processing are modulated by stimulus properties.We performed two experiments to investigate these issues. In Experiment 1, we show that these phenomena can be observed in a same-different task, and that participants show systematic variation in response times across image pairs. We show that the response times to any pair of images can be accurately predicted using two factors: their dissimilarity and their distinctiveness relative to other images. In Experiment 2, we show that these phenomena can also be observed in a visual search task where participant did not have to make any categorical shape judgments. Here too, participants showed highly systematic variations in response time that could be explained as a linear sum of shape comparisons across global and local scales. Finally, the dissimilarity and distinctiveness factors estimated from the same-different task were systematically related to the search dissimilarities observed during visual search.In sum, our results show that global and local processing phenomena are properties of a systematic shape representation governed by simple rules.

On the noisy spatiotopic encoding of word positions during reading: Evidence from the change-detection task

The present study builds on our prior work showing evidence for noisy word-position coding in an immediate same-different matching task. In that research, participants found it harder to judge that two successive brief presentations of five-word sequences were different when the difference was caused by transposing two adjacent words compared with different word replacements – a transposition effect. Here we used the change-detection task with a 1-s delay introduced between sequences – a task thought to tap into visual short-term memory. Concurrent articulation was used to limit the contribution of active rehearsal. We used standard response-time (RT) and error-rate analyses plus signal detection theory (SDT) measures of discriminability (d’) and bias (c). We compared the transposition effects for ungrammatical word sequences and nonword sequences observed with these different measures. Although there was some evidence for transposition effects with nonwords, the effects were much larger with word sequences. These findings provide further support for the hypothesized noisy assignment of word identities to spatiotopic locations along a line of text during reading.

The fast-same effect of an exclusive-OR task.

Participants are faster to decide that two stimuli are identical than to decide that they are different. Opposing theories suggested that this fast-same effect is either due (a) to a response bias toward similarity or (b) to facilitation caused by the repetition of the stimuli attributes. Although both theories predict the fast-same effect in a conventional same-different task, they make distinct predictions for tasks in which response bias is removed. In such tasks, the bias theory predicts that the fast-same would disappear whereas the facilitation theory predicts that the fast-same would remain. We tested those hypotheses using a same-different task in which participants had to indicate if all the attributes of the stimuli were matching or all were mismatching by pressing one response key, or if some attributes were matching and some were mismatching, by pressing another response key. We call this an exclusive-OR same-different task. Results show that participants were much faster in the "all-matching" condition compared with the "all-mismatching" condition, therefore supporting the facilitation theory. A fit of the linear ballistic accumulator model to the observed data provide additional supports that the fast-same effect is not caused by bias, but by a faster accumulation rate of evidence in the "all-matching" condition. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Phonological priming effects with same-script primes and targets in the masked priming same-different task.

Norris, Kinoshita and colleagues (Kinoshita & Norris, Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(1), 1-18, 2009; Journal of Experimental Psychology: General, 137(3), 434-455, 2010; Norris & Kinoshita, Quarterly Journal of Experimental Psychology, 63(1), 194-204, 2008) have suggested that the masked priming same-different task (SDT) is an excellent tool for studying the orthographic coding process because, in most circumstances, performance in that task is driven entirely by orthographic codes. More specifically, although evidence of phonological influences (i.e., phonological priming effects in the SDT) have been reported, Kinoshita, Gayed, and Norris (Journal of Experimental Psychology: Human Perception and Performance, 44(11), 1661-1671, 2018) have claimed that phonological priming does not arise in the SDT when the prime and target are written in the same script and the targets are words, the most typical experimental situation. Indeed, it does appear that no-one has yet reported phonological priming effects in such situations. The question of whether it is possible to observe phonological priming in such situations was more fully examined in the present experiments. Experiment 1 involved a masked priming SDT using Japanese Kanji script in which the primes and targets were homophonic but shared no characters. Experiment 2 was a parallel experiment using Chinese stimuli. In both experiments, phonological priming effects were observed for both one- and two-character words. These experiments indicate that, although the priming effects in masked priming SDTs undoubtedly have a strong orthographic basis, phonological codes also play a role even when the prime and (word) target are written in the same script.

Does orthographic processing emerge rapidly after learning a new script?

Orthographic processing is characterized by location-invariant and location-specific processing (Grainger, 2018): (1) strings of letters are more vulnerable to transposition effects than the strings of symbols in same-different tasks (location-invariant processing); and (2) strings of letters, but not strings of symbols, show an initial position advantage in target-in-string identification tasks (location-specific processing). To examine the emergence of these two markers of orthographic processing, we conducted a same-different task and a target-in-string identification task with two unfamiliar scripts (pre-training experiments). Across six training sessions, participants learned to fluently read and write one of these scripts. The post-training experiments were parallel to the pre-training experiments. Results showed that the magnitude of the transposed-letter effect in the same-different task and the serial function in the target-in-string identification tasks were remarkably similar for the trained and untrained scripts. Thus, location-invariant and location-specific processing does not emerge rapidly after learning a new script; instead, they may require thorough experience with specific orthographic structures.

A reexamination of consonant-vowel differences in masked transposed letter priming effects in the lexical decision task.

Most orthographic coding models are based on the assumption that the orthographic code does not distinguish between vowels and consonants and, therefore, those models predict no difference between vowel (cisano-CASINO) and consonant (caniso-CASINO) transposed-letter (TL) effects. The available data, however, do provide some evidence for a consonant-vowel distinction at the level of the orthographic code. Most centrally, masked priming lexical decision tasks, mainly carried out in Spanish, have shown priming from consonant TL primes (e.g., caniso) but not from vowel TL primes (e.g., cisano). The present experiments were an investigation of this pattern. Experiment 1, based on Schubert, Kinoshita, and Norris' (2018) stimuli which showed no consonant-vowel differences in an unprimed same-different task, also showed no consonant-vowel differences in masked TL priming effects in lexical decision showing, for the first time, a vowel TL priming effect in that task. Experiment 2, using Lupker, Perea, and Davis' (2008) Experiment 1a stimuli, also showed a small but significant vowel TL priming effect (a nonreplication of that experiment), while replicating the consonant TL priming effect that those authors originally reported. In Experiment 3, TL priming was again essentially unaffected by the consonant-vowel status of the letters involved as well as by target frequency, a variable on which the Experiment 1 and 2 stimuli differed. These results, supported by evoked response potential (ERP) results from other labs, suggest that consonant-vowel TL differences, when they do emerge in English, are likely are not due to the nature of the orthographic code. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

The EZ Diffusion Model: An overview with derivation, software, and an application to the Same-Different task

The EZ Diffusion Model: An overview with derivation, software, and an application to the Same-Different task

Performing the hand laterality judgement task does not necessarily require motor imagery

When people judge the laterality of rotated hand images, that is they perform the laterality judgement task (LJT), they are thought to use motor imagery. However, recent studies have suggested that its completion does not necessarily require the use of motor imagery. In this study, we investigated whether and how many people preferentially use motor imagery to perform the LJT in 37 healthy adults. We assessed the presence of behavioural features associated with motor imagery at the individual level, namely, the linear angle–response time (RT) relationship and the biomechanical constraints effect in the LJT and in the same-different judgement task (SDJT), in which people are not thought to use motor imagery. We found that at most 50% of participants showed both behavioural features in the palmar view condition of the LJT. Moreover, this proportion did not differ from that in the dorsal view condition of the LJT or that in both view conditions of the SDJT. These results demonstrate that a motor imagery–based strategy is not universally and specifically used to perform the LJT. Therefore, previous results of the LJT, in particular, regarding the biomechanical constraints effect, should be reinterpreted in light of our findings.

Cognitive architecture and capacity of the cognitive system responsible for Same - Different judgments.

Participants tend to match identical pairs of stimuli faster than different pairs. Despite many endeavours to explain this fast-same effect, there is still no theoretical consensus. A potential reason for the lack of consensus is that the cognitive architecture and capacity underlying such phenomenon is assumed and not formally tested. For example, the dual-process approach suggests that Same responses arise from a parallel treatment, whereas Different responses arise from a serial treatment. It also suggests that in both conditions, the capacity of the process is unaffected by workload (unlimited capacity). Alternative approaches argue that the fast-same effect can be explained by parallel or coactive architectures with channels working in either limited or super capacity. In this study, we formally assess the architecture (three possibilities: serial, parallel and coactive) and the capacity (three possibilities: unlimited, limited and super-capacity) of the cognitive system in a Same-Different task using Systems Factorial Technology (SFT). We recruited twenty participants to perform a double-factorial task lasting four sessions. Because of the lack of effectiveness of the blurring manipulation, we cannot draw a strong conclusion about the cognitive architecture. As for the capacity, the results show that it is mostly limited for the majority of participants. However, between 300 and 500 ms, participants tend to have a much stronger processing capacity in the Same condition compared to the Different condition. This short but strong burst of activity for identical stimuli might explain the fast-same effect.

La variación en maya yucateco: un estudio descriptivo desde la dialectología perceptual

Desde la perspectiva de la dialectología perceptiva y mediante un estudio exploratorio y descriptivo, nos propusimos examinar la habilidad de quince bilingües de maya-español para identificar posibles variedades de lengua maya y relacionarlas con una región geográfica, determinar los rasgos más salientes que usan para distinguir una variedad de otra e identificar algunos factores que pueden afectar esta percepción dialectal. Algunas tareas de la metodología usada por Preston (1999), Mase (1999) y Clopper y Pisoni (2004) fueron adaptadas en el presente estudio, como la tarea de categorización controlada con mapa y la tarea de iguales o diferentes. Los resultados muestran que la habilidad de categorización de variedades entre los participantes fue pobre o deficiente, afectada principalmente por la escasa exposición y experiencia con mayahablantes de otras comunidades. Los límites geográficos, la distinción entre habla urbana / rural y entre la jach maya (la maya pura, antigua, verdadera) y el xe’ek’ (la maya mezclada) también se identificaron como una guía de reconocimiento de variedades o zonas dialectales. Los elementos suprasegmentales (entonación, melodía, ritmo, velocidad de la elocución) y el léxico, en general, resultaron los elementos lingüísticos salientes que guiaron la percepción dialectal de los participantes.

The Security-Utility Trade-off for Iris Authentication and Eye Animation for Social Virtual Avatars.

The gaze behavior of virtual avatars is critical to social presence and perceived eye contact during social interactions in Virtual Reality. Virtual Reality headsets are being designed with integrated eye tracking to enable compelling virtual social interactions. This paper shows that the near infra-red cameras used in eye tracking capture eye images that contain iris patterns of the user. Because iris patterns are a gold standard biometric, the current technology places the user's biometric identity at risk. Our first contribution is an optical defocus based hardware solution to remove the iris biometric from the stream of eye tracking images. We characterize the performance of this solution with different internal parameters. Our second contribution is a psychophysical experiment with a same-different task that investigates the sensitivity of users to a virtual avatar's eye movements when this solution is applied. By deriving detection threshold values, our findings provide a range of defocus parameters where the change in eye movements would go unnoticed in a conversational setting. Our third contribution is a perceptual study to determine the impact of defocus parameters on the perceived eye contact, attentiveness, naturalness, and truthfulness of the avatar. Thus, if a user wishes to protect their iris biometric, our approach provides a solution that balances biometric protection while preventing their conversation partner from perceiving a difference in the user's virtual avatar. This work is the first to develop secure eye tracking configurations for VR/AR/XR applications and motivates future work in the area.

Into the Mind’s Eye: Exploring the Fast-Same Effect in the Same-Different Task

Abstract The fast-same effect is the observation that “same” responses are much faster than “different” responses in the same-different task. Moreover, identical stimuli are responded to faster than stimuli that are the same in name only (e.g., B and b). We examine Bamber’s (1969) identity reporter model (a two-stage model predicting load effects), Proctor’s (1981) facilitation framework, and Krueger and Shapiro’s (1981) priming framework, proposed to account for these effects. Facilitation and priming are strong for identical and repeated stimuli, for phonological associates, and, in principle, for any form of association. We thus manipulated two types of associations: nominal (changing the letter case, preserving phonology) and learned (matching arbitrary symbols to letters) associations and used extended training to see variations in load effects. We found that overall performance benefits from phonological information and from training, although training did not change load effects. Additionally, the results from a transfer phase show capacity limitation for learned associates, which severely constrains facilitation. This finding is inconsistent with a priming framework. The results are discussed using an expanded version of Bamber’s identity reporter model, which is also compatible with Proctor’s facilitation framework.

Developmental Letter Position Dyslexia in Turkish, a Morphologically Rich and Orthographically Transparent Language.

We present the first report of a specific type of developmental dyslexia in Turkish, letter position dyslexia (LPD). LPD affects the encoding of letter positions, leading to letter migrations within words. In a multiple case study of 24 Turkish-speaking children with developmental LPD, we examined in detail the characteristics of this dyslexia and explored its manifestation in Turkish. We used migratable words, in which a migration creates another existing word (e.g., signer-singer), which exposed the migration errors of the participants. In sharp contrast with the common assumption that dyslexics in transparent languages, including Turkish, do not make reading errors, we have shown that right stimuli can detect even up to 30% reading errors. The participants made migrations in reading aloud, comprehension, lexical decision, and same-different tasks, in both words and non-words. This indicates that their deficit is in the orthographic-visual analysis stage, a stage that precedes the orthographic input lexicon and is shared by the lexical and non-lexical routes. Their repetition of non-words and migratable words was normal, indicating that their phonological output stages are intact. They did not make digit migrations in reading numbers, indicating that the orthographic-visual analyzer deficit is orthographic-specific. The properties of Turkish allowed us to examine two issues that bear on the cognitive model of reading: consonant-consonant transpositions were far more frequent than consonant-vowel and vowel-vowel migrations. This indicates that the orthographic-visual analyzer already classifies letters into consonants and vowels, before or together with letter position encoding. Furthermore, Turkish is very rich morphologically, which has allowed us to examine the effect of the morphological structure of the target word on migrations. We found that there was no morphological effect on migrations: morphologically complex words did not yield more (nor fewer) migrations than monomorphemic ones, migrations crossed morpheme boundaries and did not preserve the morphological structure of the target word. This suggests that morphological analysis follows the letter-position encoding stage.

Self-Construal Priming Affects Holistic Face Processing and Race Categorization, but Not Face Recognition.

Self-construal priming can affect an individual’s cognitive processing. Participants who were primed with interdependent self-construal showed more holistic process bias than those who were primed with independent self-construal. The holistic processing of a face also differs across cultures. As such, the purpose of the present study was to explore whether the cultural differences in holistic face processing can be interpreted from the perspective of self-construal, as well as to investigate the relationship between self-construal and holistic face processing/face recognition/race categorization. In Experiment 1, participants were primed with control, interdependent, or independent self-construal, respectively, and then they completed a feature-space same-different task (Experiment 1A) or a composite face effect task (Experiment 1B). Results showed no priming effect in Experiment 1A, whereas independent self-construal priming resulted in less holistic processing in Experiment 1B. In Experiment 2, participants were primed with control, collective/interdependent, relational, or independent self-construal, respectively, and then they completed a Vanderbilt Holistic Face Processing Test and Cambridge Face Memory Test. Participants who were primed as independent showed greater congruency effect than the relational group. Self-construal priming had no effect on face recognition. In Experiment 3, we manipulated self-construal in the same way as that in Experiment 2 and monitored the eye movement of Chinese participants while they learned, recognized, and categorized their own-/other-race faces. Self-construal priming had no effect on face recognition. Compared with other groups, collective-/interdependent-self priming increased the fixation time of eyes and decreased the fixation time of nose in the race categorization task. These results indicated that the cultural differences in self-construal could not mirror the cultural differences in face processing in a simple way.