NEUROBIOLOGY OF LANGUAGE Edited by GREGORY HICKOK Department of Cognitive Sciences, University of California, Irvine, CA, USA STEVEN L. SMALL Department of Neurology, University of California, Irvine, CA, USA AMSTERDAM • BOSTON • HEIDELBERG • LONDON NEW YORK • OXFORD • PARIS • SAN DIEGO SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO Academic Press is an imprint of Elsevier SECTION J SPEAKING CHAPTER 56 Word Production from the Perspective of Speech Errors in Aphasia Myrna F. Schwartz1 and Gary S. Dell2 1Moss Rehabilitation Research Institute, Elkins Park, PA, USA; 2University of Illinois, Urbana-Champaign, Beckman Institute, Urbana, IL, USA A common strategy for understanding the inner 1995) and clinical profiles (Blumstein, 1973; Jefferies & workings of a complex system is to study the ways in Lambon Ralph, 2006; Romani & Galluzzi, 2005). which it breaks down. Research in naturally occurring Furthermore, computational models of word produc- speech errors (“slips of the tongue”) and research in tion have been used to simulate lesions in aphasia and aphasia both take advantage of this strategy. Language test competing theories of how errors arise. We exem- scientists study speech errors for insight into the work- plify this strategy with our own computational studies ings of the cognitive system that enables the skilled of naming and repetition. production of words and sentences. Neuroscientists study language deficits in aphasia to learn how the mechanisms of language processing are represented in 56.1 SPEECH ERRORS IN APHASIA: THE the brain. This chapter presents research that falls at NEUROLOGICAL TRADITION the intersection of these two historical traditions. The studies we highlight use computational models to sim- The long history of neurological research in aphasia ulate the cognitive mechanisms responsible for speech has been dominated by the classical syndromes or sub- errors in aphasia and large-scale lesion-symptom map- types. These include, but are not limited to, Broca’s ping to link those cognitive mechanisms with brain aphasia (BA), Wernicke’s aphasia (WA), Conduction regions. aphasia (CA), and Anomic aphasia (AA). Patients are Naturally occurring speech errors are the product of assigned to subtypes based primarily on how they per- momentary, functional disruption within a normal lan- form on clinical examination of expressive language guage system. Aphasia causes structural damage that (fluency, grammar, naming), receptive language (com- has persistent functional consequences. The most prehension of spoken words and sentences), and repe- important of these, for the present purposes, is exacer- tition (of words and sentence). Within this framework, bation of the normal tendency to occasionally select BA is considered primarily a disorder of expressive the wrong word or the wrong phonological segments speech, featuring dysfluency, altered and inconsistent when speaking (Dell, Schwartz, Martin, Saffran, & articulation (“apraxia of speech”), and simplified Gagnon, 1997; Ellis, 1985; Freud, 1953). Aphasia creates grammar. WA is primarily a disorder of speech com- vulnerability to speech errors in spontaneous speech prehension, CA is primarily a disorder of repetition, and in single word production tasks, such as spoken and AA is primarily a disorder of word retrieval. object naming and auditory word repetition. The high Speech error tendencies also enter into the diagnosis degree of experimental control afforded by naming of WA and CA. Patients with WA characteristically and word repetition tasks has made them the vehicle produce speech that is fluent but replete with errors of choice for studying how patients’ error types and (historically, “paraphasias”) that particularly distort error frequencies vary in relation to target properties the semantic and also phonological content, sometimes (Cuetos, Aguado, Izura, & Ellis, 2002; Kittredge, Dell, to a degree that obscures the intended meaning. CA Verkuilen, & Schwartz, 2008; Nickels & Howard, 1994, patients are less compromised in their functional Neurobiology of Language. DOI: http://dx.doi.org/10.1016/B978-0-12-407794-2.00056-0 701 © 2016 Elsevier Inc. All rights reserved. 702 56. WORD PRODUCTION FROM THE PERSPECTIVE OF SPEECH ERRORS IN APHASIA communication. Their comprehension is quite good normal speech depends on the integrity of the and their speech is more semantically coherent. Their auditory-articulatory mapping served by the dorsal primary deficits are inability to repeat spoken speech, route (for discussion and evidence, see Buchsbaum reduced auditory-verbal short-term memory span, and et al., 2011). vulnerability to phonemic errors in all types of produc- Apart from the Dual Stream theory, the legacy of tion tasks. Whereas WA patients are generally the classical aphasia model is seen in modern empiri- unaware of their errors, errorful productions in CA are cal studies of the relationship between aphasia syn- often accompanied by successive repair attempts, dromes and speech error patterns. This issue was referred to as “conduit d’approache.” featured in Blumstein’s (1973) seminal monograph, “A The following speech samples exemplify paraphasic phonological investigation of aphasic speech.” The speech production in WA (Ex. 1) and CA (Ex. 2). The monograph’s linguistic analysis of phonemic parapha- patients were asked to describe the Cookie Theft picture, sias in the speech of those with Broca’s, Wernicke’s which shows a woman washing dishes at an overflow- and Conduction aphasia revealed surprising unifor- ing sink while a girl and boy, the latter perched on a mity in the patterns of phonological breakdown. For tipping stool, grab cookies from a cabinet (Goodglass example, in all three groups, phonemes were substi- & Kaplan, 1983). tuted more often than they were added or deleted, substitutions were strongly constrained by similarity (1) “So the two boys work together an one is sneakin’ in acoustic-articulatory features, and errors tended to around here, making his...work an’ his further funnas replace marked structures (i.e., more complex, less fre- his time he had.” quent, and later acquired) with less marked ones. (2) “Well this um...somebody’s...ah mather is takin Blumstein (1973) recognized that the mechanism for the...washin’ the dayshes [material deleted] and paraphasia might well be different in the three aphasic there’s a...then the girl...not the girl...the boy who’s groups, notwithstanding their linguistic similarities. In getting the cooking is on this ah...strool and’ startin’ to line with this, researchers who study phonemic errors fall off.” in WA often start from the assumption that these The aphasia syndrome classification is rooted in the patients suffer from a central lexical deficit that 19th century BrocaÀWernickeÀLichtheim model of impacts the retrieval of phonological and semantic language and brain. The well-known claims of the information in language production, as it does in com- model are that motor and auditory speech engrams prehension. In contrast, researchers who study phone- localize to the left inferior frontal gyrus (IFG) and left mic errors in CA tend to emphasize that such errors posterior superior temporal gyrus (pSTG), respectively; occur even in tasks with an optional or obligatory these sensory and motor engrams are interconnected “direct” route that bypasses the lexicon (e.g., repeating through the arcuate fascisculus. IFG lesions give rise to and reading words). On this basis (and others), it has the expressive deficit in BA, pSTG lesions to the com- been argued that phonemic paraphasia in at least some prehension deficit in WA, and arcuate lesions to the CA patients arises subsequent to lexical processing repetition deficit in CA. The paraphasic speech pat- (i.e., at a postlexical phonological or phonetic stage of terns of WA and CA patients arise because the intact word production) (Caplan, Vanier, & Baker, 1986; motor speech engrams are deprived of governing Garrett, 1984; Goldrick & Rapp, 2007; Pate, Saffran, & input from the speech comprehension center due to Martin, 1987). direct damage (WA) or disconnection (CA) The notion that different mechanisms underlie the (Compston, 2006; Geschwind, 1965). phonemic paraphasias of CA, WA, and BA patients is The aphasia syndromes and classical model have an difficult to reconcile with the homogenizing impact of enduring legacy. The contemporary Dual Stream the- partial recovery. Experience suggests that some ory of auditory language processing (Hickok & patients who are diagnosed in the acute phase with Poeppel, 2004, 2007) bases its characterization of the WA or BA in time evolve either to the CA profile, with dorsal stream on neuroimaging data and evidence con- its characteristic pattern of phonemic paraphasia and cerning CA. The theory assigns to the dorsal stream conduit d’approache, or to AA. This reminds us that the function of mapping between auditory and articu- what looks to be a difference in kind (e.g., phonemic latory representations in speech. It identifies an area in errors, which resemble a plausible target, vs. neolo- the posterior Sylvian fissure at the temporoparietal gisms, which do not) may actually be one of degree. boundary (area Spt) that plays a central role in this Moreover, a particular diagnostic feature may become dorsal stream specialization and postulates that lesions salient only on resolution of another symptom that ini- here give rise to the CA symptom complex. Phonemic tially masked or distorted its expression, as when the paraphasia