Perceptual Classification Task Research Articles

Comparison of the prediction accuracy of machine learning algorithms in crosslinguistic vowel classification

Machine learning algorithms can be used for the prediction of nonnative sound classification based on crosslinguistic acoustic similarity. To date, very few linguistic studies have compared the classification accuracy of different algorithms. This study aims to assess how well machines align with human speech perception by assessing the ability of three machine learning algorithms, namely, linear discriminant analysis (LDA), decision tree (C5.0), and neural network (NNET), to predict the classification of second language (L2) sounds in terms of first language (L1) categories. The models were trained using the first three formants and duration of L1 vowels and fed with the same acoustic features of L2 vowels. To validate their accuracy, adult L2 speakers completed a perceptual classification task. The results indicated that NNET predicted with success the classification of all L2 vowels with the highest proportion in terms of L1 categories, while LDA and C5.0 missed only one vowel each. Furthermore, NNET exhibited superior accuracy in predicting the full range of above chance responses, followed closely by LDA. C5.0 did not meet the anticipated performance levels. The findings can hold significant implications for advancing both the theoretical and practical frameworks of speech acquisition.

Complementary deficits in perceptual classification in pure alexia and acquired prosopagnosia – New insights from two classic cases

Pure alexia and prosopagnosia traditionally have been seen as prime examples of dissociated, category-specific agnosias affecting reading and face recognition, respectively. More recent accounts have moved towards domain-independent explanations that postulate potential cross-links between different types of visual agnosia. According to one proposal, abnormal crowding, i.e. the impairment of recognition when features of adjacent objects are positioned too closely to each other, might provide a unified account for the perceptual deficits experienced by an agnosic patient. An alternative approach is based on the notion of complementary visual subsystems favouring the processing of abstract categories and specific exemplars, respectively. To test predictions of these two approaches with regard to pure alexia and prosopagnosia, we present previously unpublished data on digit recognition and visual crowding from two in the neuropsychological literature extensively studied patients, KD and MT (e.g., Campbell et al., 1986; Landis and Regard, 1988; Rentschler et al., 1994). Patient MT, diagnosed with pure alexia, showed pronounced abnormal foveal crowding, whereas KD, diagnosed with prosopagnosia, did not. These results form a distinct double dissociation with the performance of the two patients in other perceptual classification tasks involving Gabor micropatterns and textures, as well as Glass patterns, which revealed a significantly greater impairment in KD relative to MT. Based on an analysis of the specific task demands we argue that prosopagnosia and pure alexia may involve complementary deficits in instantiation and abstraction, respectively, during perceptual classification, beyond any category specificity. Such an explanation appears in line with previous distinctions between a predominantly left-hemispheric, abstract-category and a predominantly right-hemispheric, specific-exemplar subsystem underlying object recognition.

Integrating situational probability and kinematic information when anticipating disguised movements

ObjectivesThe current study sought to examine the relative contributions of kinematic and situational probability information to anticipation using different levels of disguised kinematics. More specifically, it was tested whether the weighting of the informational sources (kinematic vs. probabilistic) shifts relative to the certainty of the available kinematic information. Design and MethodHuman-like avatars were generated performing penalty throws and displayed in a virtual reality environment. The ambiguity of the kinematic information available from the avatars was systematically manipulated using linear morphing between genuine and disguised throws. In a perceptual classification task, trained novice observers (N = 23) were asked to classify as quickly and accurately as possible whether observed throws were either genuine or disguised. In addition, information about the performer’s action preferences was also systematically manipulated by explicitly informing participants about the performer’s AP to disguise their throw (25%, 50%, and 75%). ResultsParticipants’ response behavior showed that observers relied more heavily on the probabilistic information when the kinematics were ambiguous. For the AP 25% condition, observers were more likely to report that ambiguous throws were genuine (p < 0.001), whereas they classified the ambiguous throws as being disguised in the AP 75% condition (p < 0.001). ConclusionFindings suggest that observers rely more strongly on non-kinematic (situational probability) information when the reliability of the observable movement kinematics becomes less certain.

How “mere” is the mere ownership effect in memory? Evidence for semantic organization processes

Memory is better for items arbitrarily assigned to the self than for items assigned to another person (mere ownership effect, MOE). In a series of six experiments, we investigated the role of semantic processes for the MOE. Following successful replication, we investigated whether the MOE was contingent upon semantic processing: For meaningless stimuli, there was no MOE. Testing for a potential role of semantic elaboration using meaningful stimuli in an encoding task without verbal labels, we found evidence of spontaneous semantic processing irrespective of self- or other-assignment. When semantic organization was manipulated, the MOE vanished if a semantic classification task was added to the self/other assignment but persisted for a perceptual classification task. Furthermore, we found greater clustering of self-assigned than of other-assigned items in free recall. Taken together, these results suggest that the MOE could be based on the organizational principle of a "me" versus "not-me" categorization.

Showing a model's eye movements in examples does not improve learning of problem-solving tasks

Eye movement modeling examples (EMME) are demonstrations of a computer-based task by a human model (e.g., a teacher), with the model's eye movements superimposed on the task to guide learners' attention. EMME have been shown to enhance learning of perceptual classification tasks; however, it is an open question whether EMME would also improve learning of procedural problem-solving tasks. We investigated this question in two experiments. In Experiment 1 (72 university students, Mage = 19.94), the effectiveness of EMME for learning simple geometry problems was addressed, in which the eye movements cued the underlying principle for calculating an angle. The only significant difference between the EMME and a no eye movement control condition was that participants in the EMME condition required less time for solving the transfer test problems. In Experiment 2 (68 university students, Mage = 21.12), we investigated the effectiveness of EMME for more complex geometry problems. Again, we found no significant effects on performance except for time spent on transfer test problems, although it was now in the opposite direction: participants who had studied EMME took longer to solve those items. These findings suggest that EMME may not be more effective than regular video examples for teaching procedural problem-solving skills.

Intrinsically motivated learning systems based on biologically-inspired novelty detection

Intrinsic motivations play an important role in human learning, particularly in the early stages of childhood development, and ideas from this research field have influenced robotic learning and adaptability. In this paper we investigate one specific type of intrinsic motivation, that of novelty detection and we discuss the reasons that make it a powerful facility for continuous learning. We formulate and present one original type of biologically inspired novelty detection architecture and implement it on a robotic system engaged in a perceptual classification task. The results of real-world robot experiments we conducted show how this original architecture conforms to behavioural observations and demonstrate its effectiveness in terms of focusing the system’s attention in areas that are potential for effective learning.

Priming and stimulus–response learning in perceptual classification tasks

Participants often respond more quickly and more accurately to a repeated stimulus compared to a non-repeated one, a phenomenon known as repetition priming. In semantic classification tasks priming appears to be largely attributable to the learning of stimulus–decision and stimulus–response associations, which allow participants to bypass many of the processes engaged during initial stimulus analysis. The current study tested whether stimulus–response learning plays a similarly dominant role in priming that occurs in perceptual classification tasks. Unfamiliar objects were used as stimuli to reduce the influence of semantic processes on priming and the task switched for all test trials to eliminate stimulus–decision learning. The results showed across-task priming as measured by reaction time facilitation and improved accuracy when the response remained the same during the encoding and test phases. When the response switched, similar levels of reaction time facilitation were observed, but priming as measured by accuracy was significantly reduced and no longer significant. These findings indicate that stimulus–response learning contributes to priming in perceptual classification tasks, but does not play a dominant role. Significant stimulus-level learning that is independent of the task and response also occurs and likely indexes facilitated perceptual processing of the objects.

Decision making without bounds? Evidence from humans and monkeys

Event Abstract Back to Event Decision making without bounds? Evidence from humans and monkeys Decision making in perceptual classification tasks is generally assumed to depend on accumulation of noisy information. In the reaction time paradigm, behavioral and neurophysiological data support the notion that responses are initiated when a threshold or boundary is reached. However, the termination rule is less conclusive in time-controlled paradigms, where the experimenter controls the timing of the response signal. Two approaches have been used in such cases: 1) At the response cue, a decision is made by comparing the levels of evidence favoring the two choices (Bogacz et. al. 2006, Ratcliff 1978); 2) Alternatively, evidence accumulation may continue only until a bound is reached, and this may occur before response cue onset. If the bound is reached, neural activity and behavioral responses reflect the state of a binary decision variable (Kiani et. al. 2008). We looked at two relevant experiments, one with human participants and the other with monkeys. The human participants saw a rectangle offset to the left or right of fixation and were required to indicate the direction of shift immediately after a response signal that could occur at a delay of 0 to 2000 msec relative to stimulus onset. In the monkey experiment, the stimulus in each trial was coherently moving dots lasting 500 msec. Monkeys were trained to discriminate the motion direction and indicate their decisions by an eye movement after a signal occurring 300-550 msec after the motion termination. In both experiments, a reward cue was presented prior to stimulus onset indicating which response would, if correct, receive a larger reward. For each stimulus condition, we sorted trials into two groups according to the behavioral response latency following the response cue (shorter or longer than the median), and analyzed choices separately for each group. Surprisingly, we found that choices differed systematically between these groups, even for long trial durations in which any hypothesized bound should already have been reached. Human choices were simply less accurate for trials with longer response latencies (5 subjects), whereas monkeys were more biased toward the larger reward option at longer latencies (2 subjects). These effects were observed for trial durations of up to 2.7 seconds in humans, and up to 1.3 seconds in monkeys. These data indicate that decisions in these tasks are not determined solely by a decision variable that crosses a hard bound before receipt of the response cue. Our observations can, however, be accommodated within the framework of the leaky competing accumulator model (Usher & McClelland, 2001) in which the representation of the evidence for each of the alternatives remains unbounded and continuous even after the response signal, with the continuous values influencing both the outcome and the time taken to respond. Conference: Computational and systems neuroscience 2009, Salt Lake City, UT, United States, 26 Feb - 3 Mar, 2009. Presentation Type: Poster Presentation Topic: Poster Presentations Citation: (2009). Decision making without bounds? Evidence from humans and monkeys. Front. Syst. Neurosci. Conference Abstract: Computational and systems neuroscience 2009. doi: 10.3389/conf.neuro.06.2009.03.189 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 03 Feb 2009; Published Online: 03 Feb 2009. Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Google Google Scholar PubMed Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Chess players' intake of task-relevant cues.

Patterns of chess pieces are important for chess players in storing information in memory-recall tasks. It also may be assumed that these patterns play a central role in the perceptual coding of task-relevant cues. Two perceptual classification experiments were done to test this assumption. Chess-players' intake of task-relevant cues was studied by measuring the reaction times of subjects in counting the minor pieces on the chessboard or searching for checks in the game and in random positions. Surprisingly, there was no interaction between the level of skill and the type of position. In Experiment 3, both a perceptual classification task and a memory-recall task using the same positions were presented to a group of chess players at the same session. In the perceptual classification task, there was no interaction between the level of skill and the type of position, but in the recall task there was strong interaction. The result suggests that the information intake of chess players, in its lowest levels, is not dependent on the learned patterns. The role of the patterns becomes important in defining the information to be looked for, but the patterns are not important in the early-stage perceptual structuring of a position.

The perceptual classification of speech and nonspeech sounds

It is well known that the formant transition cues which signal the presence of [b] and [d] vary greatly according to vowel context [Liberman et al., Psychol. Rev. 74, 431–461 (1967)]. Yet, listeners nonetheless classify these different physical cues as belonging to the same phonetic segment. The present study focusses on the way in which adults are able to utilize formant transition information in perceptual classification tasks. Is there information in the formant transition that cues [b]'s and [d]'s in various vowel contexts which adults can use to form consistent groups in perceptual classification tasks? Do the perceptual groupings obtained conform to phonetic categories? Is formant transition information presented in a nonspeech context (as either two or three formant chirp patterns) grouped in the same way as when it occurs in the context of speech syllables? Our results suggest that (1) though classification according to phonetic categories is not subjects' preferred mode of grouping they can be instructed to classify in this way; (2) different groupings occur for the speech and nonspeech contexts; (3) there is sufficient information in three‐formant chirps to form groups based on phonetic categories. [Work supported by N.S.E.R.C. and NSF.]

Verbal interference with encoding in a perceptual classification task.

Verbal interference with encoding in a perceptual classification task.