Fuzzy Logical Model Of Perception Research Articles

Bayes factor of model selection validates FLMP.

The fuzzy logical model of perception (FLMP; Massaro, 1998) has been extremely successful at describing performance across a wide range of ecological domains as well as for a broad spectrum of individuals. An important issue is whether this descriptive ability is theoretically informative or whether it simply reflects the model's ability to describe a wider range of possible outcomes. Previous tests and contrasts of this model with others have been adjudicated on the basis of both a root mean square deviation (RMSD) for goodness-of-fit and an observed RMSD relative to a benchmark RMSD if the model was indeed correct. We extend the model evaluation by another technique called Bayes factor (Kass & Raftery, 1995; Myung & Pitt, 1997). The FLMP maintains its significant descriptive advantage with this new criterion. In a series of simulations, the RMSD also accurately recovers the correct model under actual experimental conditions. When additional variability was added to the results, the models continued to be recoverable. In addition to its descriptive accuracy, RMSD should not be ignored in model testing because it can be justified theoretically and provides a direct and meaningful index of goodness-of-fit. We also make the case for the necessity of free parameters in model testing. Finally, using Newton's law of universal gravitation as an analogy, we argue that it might not be valid to expect a model's fit to be invariant across the whole range of possible parameter values for the model. We advocate that model selection should be analogous to perceptual judgment, which is characterized by the optimal use of multiple sources of information (e.g., the FLMP). Conclusions about models should be based on several selection criteria.

Tests of auditory-visual integration efficiency within the framework of the fuzzy logical model of perception.

The fuzzy logical model of perception [FLMP, Massaro, Perceiving Talking Faces: From Speech Perception to a Behavioral Principle (MIT Press, Cambridge, MA, 1998)] has been extremely successful at describing performance across a wide range of ecological domains as well as for a broad spectrum of individuals. Because the model predicts optimal or maximally efficient integration, an important issue is whether this is the case for most individuals. Three databases are evaluated to determine to what extent a significant quantitative improvement in predictive ability can be obtained if integration is assumed to be somewhat inefficient. For the most part, there were no significant signs of inefficient integration. The previous differences found by Grant and Seitz [J. Acoust. Soc. Am. 104, 2438-2450 (1998)] must be due to their measures of efficiency, which appear to be invalid and/or conflate information with integration efficiency. Finally, the descriptive ability of the FLMP is shown to be theoretically informative and not simply the model's ability to describe any possible outcome.

Fuzzy logical model of bimodal emotion perception: Comment on “The perception of emotions by ear and by eye” by de Gelder and Vroomen

De Gelder and Vroomen (this issue) studied how emotions are perceived from information given by the face and the voice. Emotions in the face and the voice were presented under both unimodal and bimodal conditions. When participants identified the affect, they were influenced by information from both modalities. This result occurred even when they were instructed to base their judgement on just one of the modalities. These experiments and results provide an independent replication of similar studies published in Massaro (1998). Given this opportunity for a new set of tests, the fuzzy logical model of perception (FLMP) was fit to the new results provided by de Gelder and Vroomen. Of central interest is the nature of the bimodal performance as a function of the unimodal performance. The FLMP gave a good fit of performance. The description reveals that, although information differences exist across different instruction conditions, the information processing involved in pattern recognition appears to be the same and well-described by the FLMP.

Accuracy, Scope, and Flexibility of Models

Traditionally, models are compared on the basis of their accuracy, their scope, and their simplicity. Simplicity is often represented by parameter counts; the fewer the parameters, the simpler the model. Arguments are presented here suggesting that simplicity has little place in discussions of modeling; instead, the concept of flexibility should be substituted. When comparing two models one should be wary of the possibility of their differential flexibility. Several methods for assessing relative flexibility are possible, as represented in this special issue of the Journal of Mathematical Psychology. Here, the method of cross-validation is applied in the comparison of two models, a linear integration model (LIM) and the fuzzy-logical model of perception (FLMP), in the fitting of 44 data sets concerning the perception of layout seen among three panels with the presence or absence of four sources of information for depth. Prior to cross-validation the two models performed about equally well; after cross-validation LIM was statistically superior to FLMP, but the overall pattern of fits remained nearly the same for both models.

Speech perception in perceivers with hearing loss: synergy of multiple modalities.

Although we would expect that the role of visible speech in multimodal speech perception would have the greatest relevance to individuals with hearing impairment, relatively few analytic studies have been done with these participants. Their adequate understanding of spoken language usually requires information from several modalities or other sources of information. The framework of the fuzzy logical model of perception (FLMP) is used to assess how individuals with hearing impairment evaluate and integrate multiple sources of information. Given this framework, a distinction can be made between information and information processing. Within this framework, we can ask what information differences and information-processing differences exist among individuals with normal hearing and those with hearing impairment. Four experimental studies from the literature are analyzed to address these questions. Test items are presented under both unimodal and bimodal conditions. Of central interest is the nature of the bimodal performance as a function of the unimodal performance. The findings show that, although information differences obviously exist across different populations, their information processing involved in pattern recognition appears to be the same and is well described by the FLMP.

Perceiving talking faces: Insights into attention

Perceivers naturally integrate auditory and visual information in the perception of speech and emotion. Although both modalities contribute to perception, instructions and intention appear to modulate the impact of these modalities. Although there is some influence of the to-be-ignored modality, perceivers can attenuate its influence so that some degree of control is possible. The fuzzy logical model of perception provides a good account of performance under the different instruction conditions. For the estimation of the free parameters corresponding to the parameters that correspond to the evaluation of the auditory and visual sources, the FLMP assumed different parameter values in the different instruction conditions. Different parameter values were sufficient to capture the results. Intention can thus be accounted for in terms of information with no qualitative changes in information processing. Within the framework of the FLMP, the fundamental pattern recognition algorithm does not change. Bearing out this assumption, intention can enhance or attenuate the information contribution of sources that are available, but intention does not seem capable of enforcing another type of pattern recognition algorithm. Future research using expanded factorial designs should be able to provide a more definitive understanding of the role of intention and attentional set in pattern recognition. [Work supported by NIDCD.]

Perception of visible speech: influence of spatial quantization.

Visible speech reading was studied to determine which features are functional and to test several models of pattern recognition. Nine test syllables differing in their initial consonant were presented in intact form or under various levels of spatial quantization. Performance decreased in increasing quantization but remained relatively good at moderate levels of degradation. Different models were tested against the confusion matrices. Six features were identified as functional in distinguishing among the nine consonant--vowel syllables. These features were used as sources of information in a fuzzy-logical model of perception and an additive model. The fuzzy-logical model provided a significantly better description of the confusion matrices, showing that speech reading is analogous to other domains of pattern recognition such as face recognition and facial-affect perception.

Speechreading talking faces

It is now well established that visible speech is an important source of information in face-to-face communication. Given the valuable role of audible speech synthesis in experimental, theoretical, and applied arenas, visible speech synthesis has been developed. Research has shown that the talking head closely resembles real heads in the quality of its speech and its realism (when texture mapping is used). The talking head can be heard, communicates paralinguistic as well as linguistic information, and is controlled by a text-to-speech system. Several sources of evidence are presented which show that visible speech perception (speechreading) is fairly robust across various forms of degradation. Speechreading remains fairly accurate even when the mouth is viewed in noncentral vision; eliminating and distorting high-spatial frequency information does not completely disrupt speechreading; and speechreading is possible when additional visual information is simultaneously being used to recognize the speech input. The results are consistent with the fuzzy logical model of perception in which multiple sources of information are used to recognize patterns. Various visible feature sets are tested within the framework of the model to determine which visible features are functional in speechreading. Demonstrations of the talking head and various psychological phenomena will be provided. [Work supported by NIDCD.]

Information integration in cross-modal pattern recognition: An argument for acquired modularity

Three experiments addressed the issue of whether modularized operations (Fodor, 1983) in pattern recognition could be acquired. In Experiment 1, blind and sighted individuals were compared in the speech-tactile domain of pattern recognition. The results were inconsistent with the assumptions of the fuzzy logical model of perception (FLMP) (Oden and Massaro, 1978) and with the assumptions of modularized pattern recognition processes, but were consistent with the assumptions of ‘penetrability’ into non-modularized operations. Experiments 2 and 3 compared sighted individuals across speech-visual and speech-tactile domains, revealing fundamentally different patterns of performance that we argue demonstrate modularized and non-modularized FLMP operations in pattern recognition, respectively.

Perceiving affect from the voice and the face

This experiment examines how emotion is perceived by using facial and vocal cues of a speaker. Three levels of facial affect were presented using a computer-generated face. Three levels of vocal affect were obtained by recording the voice of a male amateur actor who spoke a semantically neutral word in different simulated emotional states. These two independent variables were presented to subjects in all possible permutations-visual cues alone, vocal cues alone, and visual and vocal cues together-which gave a total set of 15 stimuli. The subjects were asked to judge the emotion of the stimuli in a two-alternative forced choice task (either HAPPY or ANGRY). The results indicate that subjects evaluate and integrate information from both modalities to perceive emotion. The influence of one modality was greater to the extent that the other was ambiguous (neutral). The fuzzy logical model of perception (FLMP) fit the judgments significantly better than an additive model, which weakens theories based on an additive combination of modalities, categorical perception, and influence from only a single modality.

Integration of orthographic, conceptual, and episodic information on implicit and explicit tests.

An experiment was conducted to determine how orthographic and conceptual information are integrated during incidental and intentional retrieval. Subjects studied word lists with either a shallow (counting vowels) or deep (rating pleasantness) processing task, then received either an implicit or explicit word fragment completion (WFC) test. At test, word fragments contained 0, 1, 2, or 4 letters, and were accompanied by 0, 1, 2, or 3 semantically related words. On both the implicit and explicit tests, performance improved with increases in the numbers of letters and words. When semantic cues were presented with the word fragments, the implicit test became more conceptually drive. Still, conceptual processing had a larger effect in intentional than in incidental retrieval. The Fuzzy Logical Model of Perception (FLMP) provided a good description of how orthographic, semantic, and episodic information were combined during retrieval.

Perceiving speech from inverted faces.

We examined whether the orientation of the face influences speech perception in face-to-face communication. Participants identified auditory syllables, visible syllables, and bimodal syllables presented in an expanded factorial design. The syllables were /ba/, /va/, /oa/, or /da/. The auditory syllables were taken from natural speech whereas the visible syllables were produced by computer animation of a realistic talking face. The animated face was presented either as viewed in normal upright orientation or inverted orientation (180 degrees frontal rotation). The central intent of the study was to determine if an inverted view of the face would change the nature of processing bimodal speech or simply influence the information available in visible speech. The results with both the upright and inverted face views were adequately described by the fuzzy logical model of perception (FLMP). The observed differences in the FLMP's parameter values corresponding to the visual information indicate that inverting the view of the face influences the amount of visible information but does not change the nature of the information processing in bimodal speech perception.

Early Understanding and Use of Symbols: The Model Model

3. D.W. Massaro, Broadening the domain of the fuzzy logical model of perception, in Cognition: Con ceptual and Methodological Issues, H.L. Pick, Jr., P. Van den Broek, and D.C. Knill, Eds. (American Psy chological Association, Washington, DC, 1992). 4. M.M. Cohen and D.W. Massaro, Synthesis of visible speech, Behavioral Research Methods, In strumentation, & Computers, 22, 260-263 (1990); M.M. Cohen and D.W. Massaro, Modeling coartic ulation in synthetic visual speech, in Models and Techniques in Computer Animation, N.M. Thal mann and D. Thalmann, Eds. (Springer-Verlag, To kyo, 1993); M.M. Cohen and D.W. Massaro, De velopment and experimentation with synthetic visible speech, Behavioral Research Methods, In strumentation, & Computers, 26, 260-265 (1994); F.l. Parke, A model for human faces that allows speech synchronized animation, Computers and Graphics Journal, 7(1), 1-4(1975). 5. B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. McNeilage, and ). Morton, Eds., Devel opmental Neurocognition: Speech and Face Pro cessing in the First Year of Life (Kluwer, Dordrecht, The Netherlands, 1993); C.A. Fowler and D.J. Dekle, Listening with eye and hand: Cross-modal contributions to speech perception, Journal of Ex perimental Psychology: Human Perception and Per formance, 17,8} 6-828 (1991 ); Q. Summerfield, Vi sual perception of phonetic gestures, in Modularity and the Motor Theory of Speech Perception, I.G. Mattingly and M. Studdert-Kennedy, Eds. (Erlbaum, Hillsdale, NJ, 1991). 6. A. Liberman and I.C. Mattingly, The motor theory of speech perception revised, Cognition, 21, 1-33 (1985); I.G. Mattingly and M. Studdert Kennedy, Eds., Modularity and the Motor Theory of Speech Perception (Erlbaum, Hillsdale, NJ, 1991). 7. J.W. Ellison and D.W. Massaro, Evaluating and integrating features in the identification of facial affect, unpublished manuscript, University of Cali fornia, Santa Cruz (1994). 8. S.C. Levine, M.T. Banich, and M.P. Koch Weser, Face recognition: A general or specific right hemisphere capacity? Brain and Cognition, 8, 303 325(1988). 9. B. Dodd and R. Campbell, Eds., Hearing by Eye: The Psychology of Lip-Reading (Erlbaum, Hills dale, NJ, 1987).

A Comparison of Learning Models

We investigate learning in a probabilistic task, called "medical diagnosis." On each trial, a subject is presented with a stimulus configuration indicating the value of four medical symptoms. The subject responds by guessing which of two diseases is present and is then given feedback about which disease was actually present. The feedback is determined according to fixed conditional probabilities unknown to the subject. We test a normative Bayesian model as well as simple variants of well-known psychological models including the Fuzzy Logical Model of Perception, an Exemplar model, a two-layer Connectionist model and an ALCOVE model. Both the asymptotic predictions of these models (i.e., predictions regarding behavior after it has stabilized and learning is complete) and predictions of trial-by-trial changes in behavior are tested. The models are tested against existing data from Estes et al. (1989, Journal of Experimental Psychology: Learning, Memory, & Cognition,15, 556-571) and new data from medical diagnosis tasks that include not only asymmetric but also symmetric base rates. Learning was observed in all cases in that subjects tended to match the objective probabilities of the symptom configurations more closely in later trials. All of the descriptive models give a more accurate account of performance than the normative Bayesian model. Relative to a benchmark measure, however, none of these models does an especially good job of characterizing asymptotic performance or the learning process. We suggest that future experiments should address individual performance, rather than group learning curves.

Perceiving talking faces

Speech perception has been studied extensively in the last decades. It has been learned that people use many sources of information in perceiving and understanding speech. This talk focuses on the important contribution of visible information given in the talker’s face in face-to-face communication. This visible speech is particularly helpful when the auditory speech is degraded due to noise, bandwidth filtering, or hearing impairment. Although the influence of visible speech is substantial when auditory speech is degraded, visible speech also contributes to performance even when paired with intelligible speech sounds. The importance of visible speech is most directly observed when conflicting visible speech is presented with intelligible auditory speech. These studies use a synthetic talking face to achieve control over the visible speech and to study those visible aspects that are informative. The talking head can be heard, communicates paralinguistic as well as linguistic information, and is controlled by a text-to-speech system. A fuzzy logical model of perception (FLMP) has been shown to provide an accurate description of how people evaluate and integrate audible and visible speech in a broad range of experimental studies. Demonstrations of the talking head and various psychological phenomena will be provided. [Work supported by NIDCD.]

Independence of lexical context and phonological information in speech perception.

M. A. Pitt (1995a) studied the joint influence of phonological information and lexical context in W. F. Ganong's (1980) task. Pitt improved on earlier studies by collecting enough observations to make possible the quantitative analyses of an individual's data. The present article shows that the results of such analyses demonstrate that the integration of phonological information and lexical context is very well accounted for by the fuzzy logical model of perception (FLMP). Although Pitt concluded that the results of his research argued against the FLMP in favor of an interactive feedback system, his conclusion was based on an analysis of transformed results. It is argued that this use of a response transformation led to incorrect conclusions and that ultimately, models must be tested directly against observed behavior.

Data fitting and detection theory: reply to Massaro and Oden (1995).

D. W. Massaro and G. C. Oden (1995) claimed that M. A. Pitt's (1995) data provide strong evidence in favor of independence, not interactivity, as argued by Pitt. Massaro and Oden's arguments rested on an evaluation of the fit of the fuzzy logical model of perception (FLMP) to the identification data and on criticisms of the detection theory analyses. In this reply, Pitt shows that the latter criticisms were unfounded and that the data-fitting demonstrations raised questions about FLMP's ability to capture the phenomenon of interest (i.e., lexical context effects).

Cross-linguistic comparisons in the integration of visual and auditory speech

We examined how speakers of different languages perceive speech in face-to-face communication. These speakers identified synthetic unimodal and bimodal speech syllables made from synthetic auditory and visual five-step /ba/-/da/ continua. In the first experiment, Dutch speakers identified the test syllables as either /ba/ or /da/. To explore the robustness of the results, Dutch and English speakers were given a completely open-ended response task. Tasks in previous studies had always specified a set of alternatives. Similar results were found in the two-alternative and open-ended task. Identification of the speech segments was influenced by both the auditory and the visual sources of information. The results falsified an auditory dominance model (ADM) which assumes that the contribution of visible speech is dependent on poor-quality audible speech. The results also falsified an additive model of perception (AMP) in which the auditory and visual sources are linearly combined. The fuzzy logical model of perception (FLMP) provided a good description of performance, supporting the claim that multiple sources of continuous information are evaluated and integrated in speech perception. These results replicate previous results found with English, Spanish, and Japanese speakers. Although there were significant performance differences, the model analyses indicated no differences in the nature of information processing across language groups. The performance differences across languages were caused by information differences due to different phonologies in Dutch and English. These results suggest that the underlying mechanisms for speech perception are similar across languages.

A pattern recognition account of decision making.

In the domain of pattern recognition, experiments have shown that perceivers integrate multiple sources of information in an optimal manner. In contrast, other research has been interpreted to mean that decision making is nonoptimal. As an example, Tversky and Kahneman (1983) have shown that subjects commit a conjunction fallacy because they judge it more likely that a fictitious person named Linda is a bank teller and a feminist than just a bank teller. This judgment supposedly violates probability theory, because the probability of two events can never be greater than the probability of either event alone. The present research tests the hypothesis that subjects interpret this judgment task as a pattern recognition task. If this hypothesis is correct, subjects' judgments should be described accurately by the fuzzy logical model of perception (FLMP)--a successful model of pattern recognition. In the first experiment, the Linda task was extended to an expanded factorial design with five vocations and five avocations. The probability ratings were described well by the FLMP and described poorly by a simple probability model. The second experiment included (1) two fictitious people, Linda and Joan, as response alternatives and (2) both ratings and categorization judgments. Although the ratings were accurately described by both the FLMP and an averaging of the sources of information, the categorization judgments were described better by the FLMP. These results reveal important similarities in recognizing patterns and in decision making. Given that the FLMP is an optimal method for combining multiple sources of information, the probability judgments appear to be optimal in the same manner as pattern-recognition judgments.

Orthographic redundancy in letter recognition: Orthographic neighbourhood or orthographic context?

Abstract In explaining the word-superiority effect (i.e. the better detection of a letter in a word than in a nonword), the Interactive Activation Model (IAM) of McClelland and Rumelhart (1981) and the Fuzzy Logical Model of Perception (FLMP) of Massaro (1979) emphasise the importance of orthographic redundancy (i.e. the regularities of letters within words) in different ways. In the IAM, orthographic redundancy is defined by the number of “friends”; that is, words sharing the same letters except one with the word containing the target letter. Such friends constitute the orthographic “neighbourhood”. FLMP stresses the orthographic “context”; that is, the similarity of the word with a representation in the lexicon. The orthographic neighbourhood and context are manipulated independently in Experiment 1, and the findings are better understood in terms of the orthographic neighbourhood. By increasing the number of friends in nonwords, better letter detection is also obtained in nonwords as compared with lett...