The role of lexical competition and acoustic-phonetic structure in lexical processing: evidence from normal subjects and aphasic patients

A tale of two lexica: Investigating computational pressures on word representation with neural networks

Introduction

The notion of a single localized store of word representations has become increasingly less plausible as evidence has accumulated for the widely distributed neural representation of wordform grounded in motor, perceptual, and conceptual processes. Here, we attempt to combine machine learning methods and neurobiological frameworks to propose a computational model of brain systems potentially responsible for wordform representation. We tested the hypothesis that the functional specialization of word representation in the brain is driven partly by computational optimization. This hypothesis directly addresses the unique problem of mapping sound and articulation vs. mapping sound and meaning.

Results

We found that artificial neural networks trained on the mapping between sound and articulation performed poorly in recognizing the mapping between sound and meaning and vice versa. Moreover, a network trained on both tasks simultaneously could not discover the features required for efficient mapping between sound and higher-level cognitive states compared to the other two models. Furthermore, these networks developed internal representations reflecting specialized task-optimized functions without explicit training.

Discussion

Together, these findings demonstrate that different task-directed representations lead to more focused responses and better performance of a machine or algorithm and, hypothetically, the brain. Thus, we imply that the functional specialization of word representation mirrors a computational optimization strategy given the nature of the tasks that the human brain faces.

Audiological Assessment of Word Recognition Skills in Persons With Aphasia

PURPOSE

The purpose of this study was to evaluate the ability of persons with aphasia, with and without hearing loss, to complete a commonly used open-set word recognition test that requires a verbal response. Furthermore, phonotactic probabilities and neighborhood densities of word recognition errors were assessed to explore potential underlying linguistic complexities that might differentially influence performance among groups.

METHOD

Four groups of adult participants were tested: participants with no brain injury with normal hearing, participants with no brain injury with hearing loss, participants with brain injury with aphasia and normal hearing, and participants with brain injury with aphasia and hearing loss. The Northwestern University Auditory Test No. 6 (NU-6; Tillman & Carhart, 1966) was administered. Those participants who were unable to respond orally (repeating words as heard) were assessed with the Picture Identification Task (Wilson & Antablin, 1980), permitting a picture-pointing response instead. Error patterns from the NU-6 were assessed to determine whether phonotactic probability influenced performance.

RESULTS

All participants with no brain injury and 72.7% of the participants with aphasia (24 out of 33) completed the NU-6. Furthermore, all participants who were unable to complete the NU-6 were able to complete the Picture Identification Task. There were significant group differences on NU-6 performance. The 2 groups with normal hearing had significantly higher scores than the 2 groups with hearing loss, but the 2 groups with normal hearing and the 2 groups with hearing loss did not differ from one another, implying that their performance was largely determined by hearing loss rather than by brain injury or aphasia. The neighborhood density, but not phonotactic probabilities, of the participants' errors differed across groups with and without aphasia.

CONCLUSIONS

Because the vast majority of the participants with aphasia examined could be tested readily using an instrument such as the NU-6, clinicians should not be reticent to use this test if patients are able to repeat single words, but routine use of alternative tests is encouraged for populations of people with brain injuries.

Inferior Frontal Cortex Contributions to the Recognition of Spoken Words and Their Constituent Speech Sounds

Speech perception and comprehension are often challenged by the need to recognize speech sounds that are degraded or ambiguous. Here, we explore the cognitive and neural mechanisms involved in resolving ambiguity in the identity of speech sounds using syllables that contain ambiguous phonetic segments (e.g., intermediate sounds between /b/ and /g/ as in "blade" and "glade"). We used an audio-morphing procedure to create a large set of natural sounding minimal pairs that contain phonetically ambiguous onset or offset consonants (differing in place, manner, or voicing). These ambiguous segments occurred in different lexical contexts (i.e., in words or pseudowords, such as blade-glade or blem-glem) and in different phonological environments (i.e., with neighboring syllables that differed in lexical status, such as blouse-glouse). These stimuli allowed us to explore the impact of phonetic ambiguity on the speed and accuracy of lexical decision responses (Experiment 1), semantic categorization responses (Experiment 2), and the magnitude of BOLD fMRI responses during attentive comprehension (Experiment 3). For both behavioral and neural measures, observed effects of phonetic ambiguity were influenced by lexical context leading to slower responses and increased activity in the left inferior frontal gyrus for high-ambiguity syllables that distinguish pairs of words, but not for equivalent pseudowords. These findings suggest lexical involvement in the resolution of phonetic ambiguity. Implications for speech perception and the role of inferior frontal regions are discussed.

An fMRI examination of the effects of acoustic-phonetic and lexical competition on access to the lexical-semantic network

The current study explored how factors of acoustic-phonetic and lexical competition affect access to the lexical-semantic network during spoken word recognition. An auditory semantic priming lexical decision task was presented to subjects while in the MR scanner. Prime-target pairs consisted of prime words with the initial voiceless stop consonants /p/, /t/, and /k/ followed by word and nonword targets. To examine the neural consequences of lexical and sound structure competition, primes either had voiced minimal pair competitors or they did not, and they were either acoustically modified to be poorer exemplars of the voiceless phonetic category or not. Neural activation associated with semantic priming (Unrelated-Related conditions) revealed a bilateral fronto-temporo-parietal network. Within this network, clusters in the left insula/inferior frontal gyrus (IFG), left superior temporal gyrus (STG), and left posterior middle temporal gyrus (pMTG) showed sensitivity to lexical competition. The pMTG also demonstrated sensitivity to acoustic modification, and the insula/IFG showed an interaction between lexical competition and acoustic modification. These findings suggest the posterior lexical-semantic network is modulated by both acoustic-phonetic and lexical structure, and that the resolution of these two sources of competition recruits frontal structures.

The cortical organization of lexical knowledge: a dual lexicon model of spoken language processing

Current accounts of spoken language assume the existence of a lexicon where wordforms are stored and interact during spoken language perception, understanding and production. Despite the theoretical importance of the wordform lexicon, the exact localization and function of the lexicon in the broader context of language use is not well understood. This review draws on evidence from aphasia, functional imaging, neuroanatomy, laboratory phonology and behavioral results to argue for the existence of parallel lexica that facilitate different processes in the dorsal and ventral speech pathways. The dorsal lexicon, localized in the inferior parietal region including the supramarginal gyrus, serves as an interface between phonetic and articulatory representations. The ventral lexicon, localized in the posterior superior temporal sulcus and middle temporal gyrus, serves as an interface between phonetic and semantic representations. In addition to their interface roles, the two lexica contribute to the robustness of speech processing.

Theories of spoken word recognition deficits in aphasia: evidence from eye-tracking and computational modeling

We used eye-tracking to investigate lexical processing in aphasic participants by examining the fixation time course for rhyme (e.g., carrot-parrot) and cohort (e.g., beaker-beetle) competitors. Broca's aphasic participants exhibited larger rhyme competition effects than age-matched controls. A re-analysis of previously reported data (Yee, Blumstein, & Sedivy, 2008) confirmed that Wernicke's aphasic participants exhibited larger cohort competition effects. Individual-level analyses revealed a negative correlation between rhyme and cohort competition effect size across both groups of aphasic participants. Computational model simulations were performed to examine which of several accounts of lexical processing deficits in aphasia might account for the observed effects. Simulation results revealed that slower deactivation of lexical competitors could account for increased cohort competition in Wernicke's aphasic participants; auditory perceptual impairment could account for increased rhyme competition in Broca's aphasic participants; and a perturbation of a parameter controlling selection among competing alternatives could account for both patterns, as well as the correlation between the effects. In light of these simulation results, we discuss theoretical accounts that have the potential to explain the dynamics of spoken word recognition in aphasia and the possible roles of anterior and posterior brain regions in lexical processing and cognitive control.

The role of Broca's area in regular past-tense morphology: an event-related potential study

It has been suggested that damage to anterior regions of the left hemisphere results in a dissociation in the perception and lexical activation of past-tense forms. Specifically, in a lexical-decision task in which past-tense primes immediately precede present-tense targets, such patients demonstrate significant priming for irregular verbs (spoke-speak), but, unlike control participants, fail to do so for regular verbs (looked-look). Here, this behavioral dissociation was first confirmed in a group of eleven patients with damage to the pars opercularis (BA 44) and pars triangularis (BA 45) of the left inferior frontal gyrus (i.e., Broca's area). Two conditions containing word-onset orthographic-phonological overlap (bead-bee, barge-bar) demonstrated that the disrupted regular-verb priming was accompanied by, and covaried with, disrupted ortho-phonological priming, regardless of whether prime stimuli contained the regular inflectional rhyme pattern. Further, the dissociation between impaired regular-verb and preserved irregular-verb priming was shown to be continuous rather than categorical; priming for weak-irregular verbs (spent-spend) was intermediate in size between that of regular verbs and strong verbs. Such continuous dissociations grounded in ortho-phonological relationships between present- and past-tense forms are predicted by single-system, connectionist approaches to inflectional morphology and not predicted by current dual-system, rule-based models. Event-related potential data demonstrated that N400 priming effects were intact for both regular and irregular verbs, suggesting that the absence of significant regular-verb priming in the response time data did not result from a disruption of lexical access, and may have stemmed instead from post-lexical events such as covert articulation, segmentation strategies, and/or cognitive control.

A PET study of word generation in Huntington's disease: effects of lexical competition and verb/noun category

Huntington's disease (HD) patients show language production deficits that have been conceptualized as a consequence of executive disorders, e.g. selection deficit between candidate words or switching between word categories. More recently, a deficit of word generation specific to verbs has been reported, which might relate to impaired action representations in HD. We studied the brain correlates of language impairment in HD using H(2)O(15) positron emission tomography (PET). The activation task consisted of generation of semantically appropriate nouns and verbs in dominant (low lexical selection) and selective conditions (high lexical selection). Reaction times were longer and number of errors was higher in 12 non-demented HD than in 17 age-matched controls in all conditions. In both groups, the selective condition yielded longer reaction time and a greater number of errors than the dominant one. PET data revealed that, in control subjects, the left inferior temporal gyrus was involved in the selective condition whereas it was not in HD. Moreover, activity in the anterior cingulate and the inferior frontal gyri was correlated with behavioral performance in control subjects only. In HD, the lack of implication of these regions, already shown to be crucial in lexical selection, might have been partly compensated by the activation in the left supramarginal gyrus (phonological loop activity) and the right inferior frontal gyrus (effortful retrieval processes), which might support accessory language strategies allowing patients to achieve word generation.

Auditory word recognition: evidence from aphasia and functional neuroimaging

This review examines the neural systems underlying auditory word recognition processes using both lesion and functional neuroimaging studies. Focus is on the influence of the sound properties of language (its phonetic as well as its phonological properties) in the service of identifying a particular word or the conceptual/meaning associated with that word. Results indicate that auditory word recognition recruits a neural system in which information is passed through the network in what appears to be functionally distinct stages - acoustic-phonetic analysis in temporal areas, mapping of sound structure to the lexicon, accessing a lexical candidate and its associated lexical-semantic network in temporo-parietal areas, and lexical selection in frontal areas. Information cascades throughout the network as shown by the influence of 'goodness' of fit and phonological/lexical competition on modulation of activation in both posterior areas including the superior temporal gyrus and supramarginal gyrus and in frontal areas including the inferior frontal gyrus.

Neighbourhood density effects in auditory non-word processing in aphasic listeners

This study investigates neighbourhood density effects on lexical decision performance (both accuracy and response times) of aphasic patients. Given earlier results on lexical activation and deactivation in Broca's and Wernicke's aphasia, the prediction was that smaller neighbourhood density effects would be found for Broca's aphasic patients, compared to age-matched non-brain-damaged control participants, whereas enlarged density effects were expected for Wernicke's aphasic patients. The results showed density effects for all three groups of listeners, and overall differences in performance between groups, but no significant interaction between neighbourhood density and listener group. Several factors are discussed to account for the present results.

Drinking for negative reinforcement: the semantic priming of alcohol concepts

Cognitive models of alcohol abuse posit that the context typically associated with alcohol use, such as negative affect, implicitly activates alcohol use cognitions, which in turn leads to alcohol consumption. We selected 40 undergraduate women based upon their alcohol use and reported anxiety sensitivity, and proposed that drinking for the purpose of negative reinforcement would predict increased semantic priming between anxiety and alcohol concepts. A lexical decision task compared the response latencies of alcohol targets preceded by anxiety words to those same targets preceded by neutral words (anxiety-alcohol priming). Level of anxiety sensitivity did not relate to anxiety-alcohol priming, but drinking following social conflict was associated with increased anxiety-alcohol priming. This study specifically suggests that the contextual antecedents to drinking behavior relate to the organization of semantic information about alcohol, and more generally supports cognitive models of substance abuse.

La sílaba en la producción del habla de individuos con afasia de Broca

OBJETIVOS: analizar los sonidos [+ líquidos] en dos tipos de estructura silábica en la producción de habla de 35 hablantes de español con afasia de Broca con el fin de determinar si dichas estructuras se conservan o se pierden. MÉTODOS: se utilizó una muestra de habla obtenida a partir de conversaciones grabadas con 35 pacientes, 12 hombres y 23 mujeres, con afasia de Broca. A partir de esa muestra se seleccionaron las producciones silábicas de estructura CCV y de estructura CV para determinar las frecuencias respectivas y los posibles errores o sustituciones de fonemas. RESULTADOS: del total de estructuras silábicas analizadas ninguna mostró una estructura diferente a la de la sílaba española. Tanto la estructura CCV como la CV se han mantenido en la muestra de habla afásica analizada. Ninguno de los errores observados contiene secuencias de sonidos mal formadas. CONCLUSIONES: independientemente de los procesos de reducción, asimilación, sustitución, elisión de segmentos, se ha podido determinar que en la producción de habla de los sujetos afásicos hay un gran apego a la estructura silábica. Se sugiere, además, que la sílaba es una unidad funcional que se codifica en el nivel de programación fonética.

Effect of schooling in auditory lexical decision

The task of lexical decision demands the functioning of the phonological loop to identify and discriminate strings of sounds and lexical knowledge to identify if this string can be taken as a real word or pseudo-word.

Lexical-semantic activation in Broca's and Wernicke's aphasia: evidence from eye movements

Lexical processing requires both activating stored representations and selecting among active candidates. The current work uses an eye-tracking paradigm to conduct a detailed temporal investigation of lexical processing. Patients with Broca's and Wernicke's aphasia are studied to shed light on the roles of anterior and posterior brain regions in lexical processing as well as the effects of lexical competition on such processing. Experiment 1 investigates whether objects semantically related to an uttered word are preferentially fixated, for example, given the auditory target "hammer," do participants fixate a picture of a nail? Results show that, like normal controls, both groups of patients are more likely to fixate on an object semantically related to the target than an unrelated object. Experiment 2 explores whether Broca's and Wernicke's aphasics show competition effects when words share onsets with the uttered word, for instance, given the auditory target "hammer," do participants fixate a picture of a hammock? Experiment 3 investigates whether these patients activate words semantically related to onset competitors of the uttered word, for example, given the auditory target "hammock," do participants fixate a nail due to partial activation of the onset competitor hammer? Results of Experiments 2 and 3 show pathological patterns of performance for both Broca's and Wernicke's aphasics under conditions of lexical onset competition. However, the patterns of deficit differed, suggesting different functional and computational roles for anterior and posterior areas in lexical processing. Implications of the findings for the functional architecture of the lexical processing system and its potential neural substrates are considered.

Lexical competition effects in aphasia: deactivation of lexical candidates in spoken word processing

Research has shown that Broca's and Wernicke's aphasic patients show different impairments in auditory lexical processing. The results of an experiment with form-overlapping primes showed an inhibitory effect of form-overlap for control adults and a weak inhibition trend for Broca's aphasic patients, but a facilitatory effect of form-overlap was found for Wernicke's aphasic participants. This suggests that Wernicke's aphasic patients are mainly impaired in suppression of once-activated word candidates and selection of one winning candidate, which may be related to their problems in auditory language comprehension.

Effects of nonlinguistic auditory variations on lexical processing in Broca's aphasics

In a series of experiments, the effect of white noise distortion and talker variation on lexical access in normal and Broca's aphasic participants was examined using an auditory lexical decision paradigm. Masking the prime stimulus in white noise resulted in reduced semantic priming for both groups, indicating that lexical access is degraded by nonlinguistic white noise distortion. However, talker variation within a prime-target pair had no effect upon the performance of either the normal or aphasic individuals. The absence of a talker variation effect suggests that voice-specific information is not encoded in the lexical representations of words. The normal performance of Broca's aphasics under conditions of both white noise and talker variation supports the view that these patients have a lexical processing impairment rather than a more generalized language deficit characterized by limited computational resources.