Advances in Neurolinguistics: Semantics of Word Production in Aphasia
Arpita Bose, Ph.D., & Lori Buchanan, Ph.D.
Department of Psychology
University of Windsor, Canada
Gary Libben, Ph.D.
Department of Linguistics
University of Alberta, Canada
Introduction:
Neurolinguistics is focused on understanding neural mechanisms underlying comprehension, production and abstract knowledge of language, be it spoken, signed or written. Historically, neurolinguistics has been most closely associated with aphasiology, the study of linguistic deficits resulting from brain damage. However, in recent years, the field of neurolinguistics has broadened. Researchers are investigating various types of speech-language disorders (e.g., stuttering, dyslexia) as well as normal language processing to better understand the neurological and cognitive bases of language processing.
Neurolinguistics is an interdisciplinary endeavor involving linguistics, psychology, neurobiology, speech-language pathology and computer science, to name a few. Researchers are drawn from a variety of backgrounds and bring a wide range of experimental techniques and differing theoretical perspectives. Researchers in neurolinguistics employ traditional and advanced neurophysiological and brain imaging techniques. Traditional methodologies may include; behavioral data, reaction time and error rate measures to tasks such as picture naming or word comprehension. Advanced technologies may include brain imaging techniques (e.g., Positron Emission Tomography [PET], functional Magnetic Resonance Imaging [fMRI]) and gross electrophysiological techniques (e.g., Electroencephalography [EEG], Event Related Potentials [ERP]).
This article will focus on word processing and lexical access with respect to aphasia, with an emphasis on semantics of word processing. We chose to focus on word processing and/or lexical access because difficulty in naming (e.g., anomia) is virtually universal in aphasia. Errors in naming are generically referred to as paraphasia. Collectively, this term is applied to any unintended error of word or sound choice. Paraphasias include phonemic paraphasia (production of unintended sounds or syllables in the utterance of partially recognizable word, e.g., ‘paker’ for ‘paper’), semantic paraphasia (production of word similar in concept or meaning to the correct production, e.g., ‘butter’ for ‘bread’) and neologism (the production of nonsense word or words, usually without recognition of errors, e.g., ‘table’ becomes ‘tilto’).
The most studied paraphasia is the semantic error. Semantic errors have long been of interest to aphasiologists for what they reveal about the organization of semantic knowledge and how this explains speech planning (Dell, Schwartz, Martin, Saffran, & Gagnon, 1997). The origin and locus of semantic errors in aphasia have been explained from different perspectives depending on the theoretical background of the researcher (Dell et al., 1997; Levelt, Roelofs, & Meyer, 1999).
Three Major Levels of Processing:
In general, current models of language formulation in adults assume the existence of three major levels of processing (Bock, 1982; Dell et al., 1997; Levelt et al., 1999): conceptualization, formulation (grammatical encoding and phonological encoding) and articulation.
The conceptualization stage involves generation of so-called preverbal messages which specify concepts to be verbally expressed. In the formulation stage, preverbal messages are mapped onto a linguistic form. During the formulation stage, the process can be broken down into a number of sub-processes, including grammatical encoding (the selection of semantically appropriate lexical items and the generation of a syntactic frame or surface form) and phonological encoding. Phonological encoding involves retrieval of segmental and supra-segmental information and the generation of a syllabified phonological word, and the computation of the phonetic form of the intended utterance, referred to as phonetic encoding. The articulation stage involves initiation and execution of the speech motor plan generated in the phonetic encoding stage, which results in overt speech as a physical consequence. Various models differ with regard to the specifics of the operations that occur at a given level and the relative importance given to any particular level.
It is a core assumption of cognitive neuropsychology that brain damage can selectively impair specific components of a processing model. Damage to one component may have an effect on other processes within that model. For example, in an interactive activation model, Dell et al., (1997) noted a deficit at one level did have widespread effects throughout the model. However, we will restrict our discussion for the sources of semantic errors by focusing on specific levels of impairments in a word production model.
Semantic problems could occur at the level of degraded representation at the conceptual knowledge in the word production model. Usually, patients with progressive neurological disorders such as semantic dementia, appear to show deficits in semantic knowledge and presentation. Individuals with aphasia may show difficulty in semantic representation and knowledge as evidenced by strong correlations between the incidence of semantic errors in word comprehension and semantic errors in picture naming (e.g., Nickels & Howard, 1994); but the semantic deficits in severe aphasia and degenerative neurological disorders may have very different properties (Warrington & Cipolotti, 1996).
Semantic errors in aphasic speech primarily results from deficit at the level of lexical representation (also referred as lemma in some models). Access to incorrect lemma would result in semantic errors (e.g., dog for cat). According to Dell’s model each lemma (e.g., cat) is considered as a node and is connected with related lemma (e.g., dog, furry) in the form of networks. Additionally, there is a bi-directional connection with phonological level. The existence of feedback in the network means that during lemma access, the word nodes and it semantic neighbors (e.g., dog) are activated (Dell et al., 1997). During word production, if the target is cat, then cat will be activated along with other semantic neighbors. Once cat is selected the activation to other semantic neighbors is reduced. In aphasia, activation to the competing neighbor is not attenuated properly or activation to the appropriate target is attenuated too quickly. This results in selection of wrong targets, such as dog for cat. This is a very simple explanation of semantic errors in aphasia.
An alternative account for semantic errors has been provided by Buchanan and colleagues (Buchanan, McEwen, Westbury & Libben, 2003; Colangelo & Buchanan, in press; Libben, Buchanan, & Colangelo, 2003). According to this Failure of Inhibition (FIT) view, there is a lack of deactivation (i.e., inhibitory processes) in lexical processing in individuals with aphasia and deep dyslexia which results in semantic errors. Buchanan et al. argue that deficits shown by these patients can be traced to a failure to accomplish the task of lexical deactivation, which is an integral part of lexical processing. To illustrate this point, we report (below) data from an individual (RS) with aphasia who demonstrated semantic processing deficits, as reported in Libben (1998).
Patient RS showed a general semantic processing deficit following a cerebral hemorrhage at age 26. As a consequence, she had difficulty supplying semantically superordinate terms such as “furniture” when presented with words such as table, chair and couch, and asked “These are all types of?” RS also showed an unusual tendency to supply interpretations of words that would normally not be considered in a particular discourse context. For example, when asked which of the words in the set [oak, tulip, carnation, rose] does not belong, she unexpectedly chose “carnation.” When asked why she made that choice, she reported carnation does not belong because “it comes in a can”. While it is true that as the brand name for a type of condensed milk, “Carnation” would not belong, one would expect that prior to the onset of her aphasia, this secondary interpretation of the word would have been overpowered by the activation of the “flower” meaning, leading her to have chosen the word “oak” (a type of tree rather than a type of flower) as the item that did not belong in the set.Particularly relevant is RS’s interpretation of semantically opaque compounds, for which she showed evidence of both whole-word and constituent activation. Or, as we interpret this scenario, failure to deactivate inappropriate representations.
When asked, what the word butterfly means, she responded “a pretty yellow fly”. Note she was blending together the meaning of the whole word and the meaning of the constituents. The word “yellow” is an associate of the constituent butter. However, the descriptor “pretty” indicates a semantic influence of the whole-word meaning—butterflies are perhaps the only flies, from a human perspective, that are likely to be perceived as pretty.
Other paraphrases produced by RS showed a similar pattern of blending meanings of whole-words and constituent meanings. For summersault she said “you roll on the grass in the summer” again blending together whole word and constituent meaning. Finally, when asked the meaning of the word dumbbell, she responded “stupid weights… Arnold”. In this case she showed activation of the whole word meaning of dumbbell, which is a type of exercise weight, the constituent dumb, and the association to Arnold Schwarzenegger.
Discussion:
Our interpretation is that RS’s processing of semantically opaque compounds can be accounted for within a deactivation framework. She achieves both the activation of the whole compound words and also successfully parses them into their two morphological constituents, activating each. However, what is not accomplished is the subsequent and necessary deactivation of the constituent meanings that are irrelevant to the meanings of the whole words. The failure of this deactivation, results in miscomprehension because, in the normal case, the meaning of the whole word must take precedence over the meanings of the constituents.
This lack of deactivation is the core of Buchanan et al.’s “Failure of Inhibition” model (2003; in press) which supposes that brain damage associated with aphasia allows lexical entries to be activated but does not support subsequent pruning of activated, but irrelevant, entries. The theory predicts that patients with deep dyslexia will demonstrate increased semantic errors when asked to read a number of semantically related words.
This prediction was further supported by deep dyslexic patient JO (in Colangelo & Buchanan, in press) who made 10 times as many semantic errors to words when they were presented in category-related lists (e.g., mug, tea, spoon, coffee, sugar) than for the same words presented randomly.
Thus, the FIT view of semantic error production appears a tenable explanation and current work is determining the extent to which FIT can accommodate the production of the other types of errors common in aphasic responses. The hypothesis is also being examined with respect to the contributions of right and left hemispheres in terms of both activation and inhibition. This approach, based on the initial observations from patient RS, has prompted a theoretically rigorous and promising method for tying behavioral data with neurophysiological data to develop a neuroanatomically and theoretically sound model of language impairment subsequent to stroke.
References:
Bock, K. J. (1982). Towards a cognitive psychology of syntax: Information processing contributions to sentence formulation. Psychological Review, 89, 1-47.
Buchanan, L., McEwen, S., Westbury, C., & Libben, G. (2003). Semantics and semantic errors: Implicit access to semantic information from words and nonwords in deep dyslexia, Brain and Language, 84, 65-83.
Colangelo, A. & Buchanan, L. (in press). Implicit and Explicit Processing in Deep Dyslexia: Semantic Blocking as a Test for Failure of Inhibition in the Phonological Output Lexicon . Brain and Language.
Dell, G. S., Schwartz, M.F., Martin, N., Saffran, E., & Gagnon, D. A. (1997). Lexical access in aphasic and nonaphasic speakers. Psychological Review, 104, 801-838.
Levelt, W.J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22, 1-38.
Libben, G. (1998). Semantic transparency in the processing of compounds: Consequences for representation, processing, and impairment. Brain and Language, 61, 30-44.
Libben, G., Buchanan, L. & Colangelo, A. (2003), Morphology, Semantics and Aphasia: The failure of inhibition hypothesis. Logos and Language, IV, 45-53.
Nickels, L.A., & Howard, D. (1994). A frequent occurrence? Factors affecting the production of semantic errors in aphasic naming. Cognitive Neuropsychology, 11, 289-320.
Warrington, E. K., & Cipolotti, L. (1996). Word comprehension: The distinction between refractory and storage impairments, Brain, 119, 611-625.
Contact Author:
Arpita Bose, Ph.D.
Post Doctoral Research Fellow,
Computational and Cognitive Neuroscience Lab,
Department of Psychology, University of Windsor
Email: bosea@uwindsor.ca