This article was downloaded by: [Universitat Politècnica de València] On: 26 October 2014, At: 01:33 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Neurocase: The Neural Basis of Cognition Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/nncs20 Naming Concepts: Evidence of Two Routes a Heidi L. Roth , Stephen E. Nadeau c Knight & Kenneth M. Heilman a b c d e c , Amber L. Hollingsworth , Ann Marie Cimino- b c e f Department of Neurology , University of North Carolina , Chapel Hill, USA b Department of Neurology , University of Florida College of Medicine , Gainesville, Florida, USA c The Brain Rehabilitation Research Center, Malcolm Randall DNA Medical Center , Gainesville, Florida, USA d The Rehabilitation Outcomes Research Center, Malcolm Randall DNA Medical Center , Gainesville, Florida, USA e The Geriatric Research, Education and Clinical Center, Malcolm Randall DNA Medical Center , Gainesville, Florida, USA f The Neurology Service, Malcolm Randall DNA Medical Center , Gainesville, Florida, USA Published online: 16 Feb 2007. To cite this article: Heidi L. Roth , Stephen E. Nadeau , Amber L. Hollingsworth , Ann Marie Cimino-Knight & Kenneth M. Heilman (2006) Naming Concepts: Evidence of Two Routes, Neurocase: The Neural Basis of Cognition, 12:1, 61-70, DOI: 10.1080/13554790500502892 To link to this article: http://dx.doi.org/10.1080/13554790500502892 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions Neurocase (2006) 12, 61–70 Copyright © Taylor & Francis Group, LLC ISSN: 1355-4795 print DOI: 10.1080/13554790500502892 Naming Concepts: Evidence of Two Routes NNCS HEIDI L. ROTH1, STEPHEN E. NADEAU2,3,4,5, AMBER L. HOLLINGSWORTH3, ANN MARIE CIMINO-KNIGHT3 and KENNETH M. HEILMAN2,3,5,6 Naming 1 Department of Neurology, University of North Carolina, Chapel Hill, USA Department of Neurology, University of Florida College of Medicine, Gainesville, Florida, USA 3 The Brain Rehabilitation Research Center, Malcolm Randall DNA Medical Center, Gainesville, Florida, USA 4 The Rehabilitation Outcomes Research Center, Malcolm Randall DNA Medical Center, Gainesville, Florida, USA 5 The Geriatric Research, Education and Clinical Center, Malcolm Randall DNA Medical Center, Gainesville, Florida 6 The Neurology Service, Malcolm Randall DNA Medical Center, Gainesville, Florida Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 2 This study examines a parallel distributed processing (PDP) model (partly based on the Wernicke-Lichtheim information processing model) that posits two routes for naming concepts, whole word and phonological. To test the two naming route hypothesis of this model, we performed confrontation naming tests that were either uncued, semantically cued, or phonologically cued in a patient with naming impairment due to Broca’s aphasia. In spoken language and in uncued naming to confrontation, word retrieval was severely impaired and marked by semantic but no phonemic paraphasic errors. With semantic cues, naming behavior was unchanged; however, with phonological cues, naming success was enhanced but frequent phonemic paraphasias were produced. These results suggest that the patient spontaneously engaged the whole word naming route, but when given phonological cues, he engaged an alternative phonological naming route that incorporated phonological sequence knowledge. Introduction The ability to attach names to concrete and abstract concepts, such as objects, actions, emotions and thoughts, constitutes one of the fundamental and critical functions of language. Disorders of naming are frequently associated with injury to the brain, and almost all types of aphasia are accompanied by naming deficits. Based on studies of normal subjects and patients with neurological injury, investigators have proposed models of how the brain mediates naming. The original Wernicke-Lichtheim (W-L) information processing model of language function has played a dominant role in understanding aphasic syndromes (Lichtheim, 1885). It provides a topographical map of the critical modules underlying language, relates this topography to brain anatomy, and is often used to explain naming deficits. The W-L model of language posits that there are three basic modules important for language function. First, there is the acoustic module that stores memories of word sounds (Wernicke’s area), located in the posterior portion of the dominant superior temporal gyrus. Second, there is the articulatory-motor area that stores the memories of movements needed to produce speech sounds, located in the frontal operculum, including pars opercularis and triangularis (Broca’s area). Finally, there is the conceptual module Received 26 March 2005; accepted 17 November 2005. We are most grateful to John Richardson for creation of the figures. Address correspondence to Stephen E. Nadeau, M.D., GRECC (182), Malcom Randall DVA Medical Center, 1601 SW Archer Road, Gainesville, FL 32608-1197. E-mail: snadeau@ufl.edu that stores the concepts or ideas that are expressed as language. This module is widely distributed throughout unimodal, polymodal, and supramodal association cortices of both hemispheres. Unfortunately, the W-L information processing model does not specify the characteristics of the representations and how they might be stored in the brain. It also does not address the means by which these modules might interact. We have proposed a parallel distributed processing (PDP) model that uses the same general topography as the W-L model (Nadeau, 2001), but also specifies how representations are generated in the modules and how knowledge is represented in the links between these modules (Figure 1). From this PDP model, it becomes apparent that there could be two alternative pathways supporting naming of concepts. The PDP modification of the W-L model posits that the acoustic module (Wernicke’s area) contains large numbers of units located in auditory association cortices that represent acoustic features of phonemes. The articulatory module (Broca’s area) contains units predominantly located in dominant frontal operculum that represent discrete articulatory features of speech (as opposed to continuously variable motor programs). The semantic or conceptual module contains an array of units distributed throughout unimodal, polymodal and supramodal association cortices that represent semantic features of concepts. Within any module, a representation corresponds to a specific pattern of activity of all the units, hence the term distributed representation. Each unit within each of these modules is connected via interposed hidden units to many, if not most, of the units in the other modules. During language acquisition, the strengths of the connections Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 62 Fig. 1. Proposed parallel distributed processing model of language. Pathway 6–5 represents the phonological input lexicon. Pathways 1–2 and 4–3 represent the phonological output lexicon. See text for details. between the units are gradually adjusted so that a pattern of activity involving the units in one module elicits the correct pattern of activity in the units of the other modules. The entire set of connections between any two modules forms a pattern associator network. The hidden unit modules, in conjunction with nonlinear unit properties, enable the association of representations in two connected domains that are arbitrarily related to one another (e.g., word sound and word meaning). An important characteristic of PDP models is that knowledge is stored as patterns of connectivity not only within modules but also between modules. For example, understanding the meaning of a word that is heard is achieved through the connections between the module that contains the sound features of language and the module that contains concept features (the acoustic-concepts representations pattern associator, Figure 1, pathway 6–5). This pattern associator network corresponds to the cognitive neuropsychological concept of a phonological input lexicon (Ellis and Young, 1988). The knowledge that allows a person to translate heard sound sequences into articulatory sequences, and thereby mediates repetition of both real words and nonwords, is contained in the network that connects the acoustic module to the articulatory module (the acoustic-articulatory motor pattern associator, Figure 1, pathway 7–3). Because this network has acquired, through experience, knowledge of the systematic relationships between acoustic sequences and articulatory sequences, it has learned the sound sequence regularities of the language: the phonemic sequences of joint phonemes, rhymes, syllables, affixes, morphemes and words characteristic of the language (Nadeau, 2001). The knowledge that enables a person to translate a concept into a spoken word (the phonological output lexicon H. L. Roth et al. (Ellis and Young, 1988)) is contained in two different pattern associator networks that connect the concept representations module to the articulatory motor module (Figure 1, pathways 1–2 and 4–3). These two pattern associator networks support different forms of knowledge. The indirect concept representations-articulatory motor pathway (pathway 4–3, hereinafter referred to as the phonological route) provides a robust basis for knowledge of phonological sequences and sublexical entities because of the sequence knowledge stored in the acoustic-articulatory motor pattern associator. However, the direct concept representationsarticulatory motor pattern associator (pathway 1–2, hereinafter referred to as the whole word route) does not contain much knowledge of sequences and sublexical entities because it translates spatially distributed patterns of activity corresponding to concepts into temporally distributed sequences of motor activity corresponding to articulated words. This spatial-temporal translation precludes significant acquisition of sequence knowledge and makes this fundamentally a whole word pathway. The existence of this direct, whole word naming route finds support in studies of subjects with repetition conduction aphasia; some appear to have lost most phonological sequence knowledge (pathway 3/4–7)(resulting in a severe deficit in auditory verbal short term memory), but can speak quite well, producing few if any phonological paraphasic errors, can repeat real words (with evidence of influence by semantic attributes but little influence of word length), and are severely impaired in repeating nonwords and functors (Warrington and Shallice, 1969; Saffran and Marin, 1975; Caramazza et al., 1981; Friedrich et al., 1984). It also finds some support in reports of subjects with conduction aphasia who are able to repeat words better than nonwords (Saffran and Marin, 1975; Friedrich et al., 1984; McCarthy and Warrington, 1984; Caramazza et al., 1986), and who are able to repeat words better when they are given in a sentence context than when given as a single word (thereby increasing the likelihood of engaging concept representations) (McCarthy and Warrington, 1984). However, a model in which the only link from the concept representations module to the articulatory motor module is the direct one (pathway 1–2) cannot account for observations that normal subjects exhibit phonological slips-of-the-tongue, and aphasic subjects produce phonemic paraphasias in naming and spontaneous language quite comparable to those produced during repetition. To explain these observations, one must posit access from concept representations to phonological sequence knowledge, as indicated in pathway 4–3 of the model. Thus, this PDP model predicts two pathways enabling naming of concepts, which together instantiate the phonological output lexicon. Normal language production probably engages both output routes from the concept representations module, and language production might not be normal when either of these pattern associator networks is impaired. The case described here provides further evidence of the use of these two routes 63 Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 Naming Fig. 2. Axial T2-weighted MRI obtained 10 years after the stroke. The scan reveals infarction of nearly the entire territory of the left middle cerebral artery. All the perisylvian cortex, the basal ganglia, the insula, and most of the temporal, parietal and occipital cortices are severely damaged and largely replaced by cerebrospinal fluid. There is some sparing of the frontal pole and of posterior temporal cortex, as well as medial frontal and parietal cortex supplied by the anterior and posterior cerebral arteries. when producing spoken words. Our subject, who had Broca’s aphasia stemming from a very large left hemisphere infarct (Figure 2), produced only whole words, most often semantic paraphasias, when speaking spontaneously or naming objects to confrontation. In our initial evaluation of him, providing whole word cues to naming to confrontation neither improved his naming success nor changed his error pattern. Providing phonemic cues seemed to improve his naming success but also precipitated a profusion of phonological paraphasic errors. We posited that in the course of recovery from his stroke and the extensive speech-language therapy he received, he learned to exclusively employ the whole word naming route, thereby avoiding undesirable phonological paraphasic errors. However, the simultaneous presentation of a picture, which generated a concept representation and top down activation of the phonological route, and a phonemic cue, which provided bottom-up activation of the phonological route, sufficiently activated the phonological route that its use was unavoidable. Because in this situation, he was presumably engaging both whole word and phonological routes, his naming ability should improve. However, because he was also engaging a damaged or underdeveloped phonological route, he was producing abundant phonological paraphasic errors. Our initial observations and the hypotheses we posited to account for them led to a systematic study of the subject’s baseline and cued naming to confrontation. Published data on phonological processing provide converging evidence of the existence of two routes by which concepts might be translated into spoken words, but they do not prove this hypothesis. The present case, in which a single subject alternatively engages one route or the other, depending upon the circumstances, therefore provides important new evidence of the existence of two naming routes. Methods Case Description This 41-year-old left-handed man was shot in the neck at age 30. The left carotid artery was penetrated at the bifurcation and was repaired with a saphenous vein graft. The immediate sequelae were right hemiparesis and aphasia. He underwent extensive rehabilitation therapy for several years following the injury. He was first seen at our institution 10 years later. On examination, except for a right hemiparesis, his deficits were exclusively cognitive. He was fully oriented and was able to recall three objects immediately and after several minutes. He had right-left confusion, mild ideomotor apraxia of the left hand, normal performance on a go-no-go test, a contrasting programs test, and a test of drawing ability. Spontaneous language was nonfluent, slightly effortful, and agrammatic. He had word finding difficulty and made frequent semantic paraphasic errors but he produced no phonological paraphasic errors. Verbal expression was usually limited to one or two word utterances comprised almost exclusively of nouns and verbs. He employed exaggerated prosody, animated facial expressions, and elaborate hand gestures in his efforts to communicate. He named 23 of 60 items on the Boston Naming Test (Kaplan et al., 1983) and made frequent semantic errors. His reading and writing were agrammatic. He could sing well. He scored 58 on the Western Aphasia Battery and his profile on this test was consistent with a non-fluent or Broca’s aphasia (Kertesz, 1982). He named 11 farms animals in one minute and produced a total of two words in three minutes on the Controlled Oral Word Association test (Benton and Hamsher, 1989) employing the letters F, A, and S. He was able to correctly answer yes/no 64 H. L. Roth et al. Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 questions and follow complex multi-step commands. He could repeat single words and two-word phrases but failed with phrases longer than two words. For phrases longer than two words, he would correctly repeat one or two of the words, but would fail to complete the entire phrase. When pressed to repeat a sequence of unrelated words, he made semantic, but not phonemic paraphasic errors. For example, when asked to repeat “brown, honesty, tulip” he repeated “brown, dishonesty, flower.” Results of brain imaging are illustrated in Figure 2. The following is an example of his internally generated language in response to a question about the events surrounding his stroke: stroke … Sunday night… . Navy, Army, Airforce … Sunday night … pool (gestures motion for shooting pool) … . pay phone… . “I got your girlfriend”… . I don’t know … . memories … memories … shot twice (gestures trigger pull with hand) … pints of blood (holding his neck area) … bar … helpless … helpless … rescue … helicopter … coma … helicopter … passed out … helicopter.. memories … drugged up … . drill, drill (points to head) … drugged up … . tragedy, tragedy, tragedy … lucky … talk … today … hard … hard … hard … speech … hard. During confrontation naming, he usually produced a stream of semantic paraphasic errors that only occasionally culminated in the correct response but usually made the meaning apparent—in effect an attempt at a semantic conduite d’approche. In an unsuccessful attempt to name a lobster, he responded “The [Florida] Keys … alligator … octopussy … .octopus … . hot … steaming … . raw … the Keys … big … . money … . money-money-money … . delicious … .food … . expensive … .fish … . hot … steamin … dinner … . stopped up … . tight … hand-cuff … hot.” In an ultimately successful attempt to name a sword, he responded “ war … long time … knife … .scissors … razor … .disappeared … knife … razor … . big … . saw … . knife … . razors … . sword.” Only rarely did he make phonemic paraphasic errors during spontaneous speech or confrontation naming. However, when he was given a phonemic cue, his naming was characterized by large numbers of phonemic paraphasias produced in a usually unsuccessful attempt at a phonological conduite d’approche. For example, when asked to name a faucet, he responded “ dishes … . chairs … .dishwasher … shut … . water … ready to go … water … shut … hot … cold … sink … water … heavy … water … heavy … washer … tub” When he was given the phonological cue “faus” and tried to name the picture of the faucet again, he said: “fauwash … fau … fau … fauswah … thafaush … fallshine … . fallsha … fallshvine … fallswash … fallsh.” These patterns of response to bedside testing suggested that the patient normally used a whole word route to confrontation naming (as well as in internally generated language) (pathway 1–2, Figure 1), but that by providing him with a phonemic cue, we could induce him to employ a phonological route—a route that engaged sublexical representations implicit in his stores of phonological sequence knowledge (pathway 4–3, Figure 1). To test this hypothesis, we conducted more systematic cued naming studies. Procedures Confrontation naming was further assessed with a set of 254 black and white line drawings of objects on 4x6 inch cards, most of which were taken from the Snodgrass and Vanderwart collection (Snodgrass and Vanderwart, 1980). Naming was tested under three conditions: no cue, semantic cue, and phonological cue. The patient was given one minute to respond to each item. If he produced the correct response in less than one minute, the next stimulus was presented immediately. All procedures were videotaped using a Panasonic Ag-187 VHS recorder. Responses were scored independently from the videotapes by three different clinicians experienced in the evaluation of aphasic errors. Errors were first scored according to lexical status: word or nonword. Errors were then scored as conceptually related to the target, phonologically related to the target, unrelated to the target, or phonologically related to a semantic relative of the target. Criteria for determination of phonological relatedness were based on established methodology (Gagnon et al., 1997). Informed consent for the study was obtained under a protocol approved by the University of Florida Health Science Center Institutional Review Board. During the baseline, uncued naming task, the drawings of objects were presented to the patient; 164 of the 254 items were named in three or fewer attempts and were classified as “easy to name.” The remaining 90 pictures were named after more than three utterances or were not named successfully within one minute, and were classified as “difficult to name.” Only responses to the difficult to name items were scored, and only these 90 items were tested in the cued naming experiments. For the cued conditions, the 90 items were divided into two lists of 45 items each, balanced for word frequency (Francis and Kucera, 1982) and age of acquisition (Carroll and White, 1973; Gilhooly and Logie, 1980). Cues were presented immediately before the presentation of the picture. During one testing session, one set of 45 items was semantically cued and the other phonologically cued. During a second testing session, the cues for each set of 45 items were reversed. Responses for the 90 naming trials were classified according to the categories depicted in Tables 1 and 2. Results Clinician raters categorized 2,527 error responses that were produced in the 270 naming trials in the baseline (90), semantic cueing (90) and phonological cueing (90) conditions. The Kappa statistic measuring inter-rater reliability for the classification of naming responses was 0.93. Overall agreement on classifications was 95 percent. Results are depicted in Tables 65 Naming Table 1. Summary of performance under different cue conditions Baseline (no cue) Semantic cue Phonological cue 974 28.9 948 92 923 25.5 900 94.2 722 47.8 679 75 Total responses Percent of 90 targets named correctly Total errors Percent of errors that were words Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 Table 2. Percentage of error types under different cue conditions Baseline (no cue) Semantic cue Phonological cue General error types (%) Phonologically related to target Semantically related to target Unrelated to target Phonologically related to semantic relative 9 75.2 16 5.4 8.2 77.9 15.1 5.0 33.9 54.2 11.9 5.2 Specific error types Word, phonologically related to target only Word, semantically related to target only Word, phonologically and semantically related to target Word, phonologically related to semantic relative Word, unrelated to target Nonword, phonologically related to target Nonword, phonologically related to semantic relative Nonword, unrelated to target 1.7 69.5 5.7 1.8 13 1.6 3.6 2.8 1.1 71.7 5.9 2.3 12.8 0.9 2.7 2.2 12.8 49 5.2 0.4 7.5 15.9 4.7 4.4 1 and 2. Naming success was quite similar in the baseline and semantic conditions (26/90 and 23/90 correct, respectively) but the subject was more successful in the phonological cueing condition (43/90 correct)(p = 0.0032, Fisher’s exact test, phonological versus semantic cueing). The percentages of error types were similar for the baseline and semantic cueing conditions, but the phonological cueing condition resulted in more phonological errors (33.9% vs. 9% in baseline and 8.2% in semantically cued conditions)(p < 0.0001) and nonword errors (25% vs. 8% in baseline and 5.8 % in semantically cued conditions)(p < 0.0001)(Fisher’s exact test, phonological versus semantic cueing). Discussion These results provide support for the dual naming route model presented here. When spontaneously speaking or naming to confrontation, with or without auditory whole word prompts, our patient employs a whole word process. In these conditions, his word retrieval success is poor and marked by frequent semantic paraphasic errors. To produce a target word, he appears to use a lexical-semantic conduite d’approche, but this strategy usually is futile. However, using this whole word process, he does avoid phonological errors and manages to communicate by “boxing in” the concept he is trying to convey. In contrast, when naming to confrontation after auditory phonological prompts, he frequently employs a phonological process. Using this process, he produces words that are phonologically closer to the target word and he achieves greater naming success than when he employs a whole word process. However, by doing so he makes many phonological paraphasic errors. In healthy individuals, the whole word and phonological naming routes are most likely used simultaneously, constraining each other in ways yet to be determined. However, when the brain is injured, as in this patient, the ability to use one of these routes may be lost or one route may be used more than the other. We posit that in the course of recovery from his stroke and the extensive speech-language therapy our subject received, he learned to exclusively employ the whole word naming route, thereby avoiding undesirable phonological paraphasic errors. However, the simultaneous presentation of a picture, which generated a concept representation and top down activation of the phonological route, and a phonemic cue, which provided bottom-up activation of the phonological route, sufficiently activated the phonological route that its use became unavoidable. Because in this situation, he was presumably engaging both whole word and phonological routes, his naming ability improved. However, because he was also engaging a damaged or underdeveloped phonological route, he produced abundant phonological paraphasic errors. Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 66 It is possible that our patient uses a completely different mechanism when he names to visual confrontation than when he produces internally generated spoken language. When using a whole word naming approach, he might be eliciting articulatory motor representations directly from visual representations (as seems likely in patients with nonoptic aphasia (Shuren and Heilman, 1993)). However, he makes frequent semantic paraphasic errors during word naming, just as in internally generated spoken language, suggesting that he is engaging concept representations when naming to confrontation. Since our patient is left-handed and left-handed people more often have bi-hemispheric mediation of speech and language, we cannot exclude the possibility that the systems mediating his naming and spoken language are organized fundamentally differently from those mediating these processes in most right-handed people. Information Processing Models Information processing models, now commonly invoked by cognitive neuropsychologists to account for aphasic phenomenology, posit a single naming route from semantic representations (Patterson and Shewell, 1987). In these models, both semantic and phonological cueing should increase naming success by virtue of increased activation of lexical representations. It is not clear, however, how this single naming route mechanism could account for the dissociation we observed in our subject, where semantic cueing had no influence, but phonological cueing had a dramatic influence. According to these cognitive information-processing models, spoken word semantic cueing would also be expected to activate phonological representations, and since these representations are degraded, it is not clear why such cueing would not also lead to the production of phonological paraphasic errors through the activation of competing phonemes. In contrast, the dual route connectionist model we propose provides a ready explanation for all of the phenomena observed in our subject. Other Support for the Dual Route Model The model that was tested in this investigation was developed on the basis of insights into the nature of the neural representation of various types of knowledge that have been provided by connectionist research, coupled with the need to account for two consistent observations: 1) patients with conduction aphasia frequently make phonemic paraphasic errors in internally generated spoken language quite comparable to those made during repetition; and 2) normal subjects make phonemic paraphasic errors in internally generated spoken language (Nadeau, 2001). This model can account, within the limits so far probed by experimentation, for an extensive body of findings from research on phonological function in normal and brain damaged individuals (Nadeau, 2001). H. L. Roth et al. In the introduction we noted the evidence supporting the existence of a direct, whole word naming route, largely derived from studies of subjects with conduction aphasia. Extensive additional support is provided for the existence of the phonological route by evidence of ongoing semantic and phonological interactions in word production, for example, in malapropisms (formal or verbal paraphasias) (Dell and Reich, 1981; Harley, 1984; Martin et al., 1989; Harley, 1990; Brédart and Valentine, 1992; Gagnon et al., 1995; Best, 1996; Laine and Martin, 1996; Dell et al., 1997; Gagnon et al., 1997; Goodglass et al., 1997), semantic paraphasias and word blends (MacKay, 1973; Motley and Baars, 1976; Dell, 1980; Dell et al., 1993); by evidence that valid phonological information is available to subjects, even when word production is not successful, as in neologisms (Ellis et al., 1983; Miller and Ellis, 1987; Valdois et al., 1989; Wilshire, 1998) and the tip-of-the-tongue phenomenon (Brown and McNeill, 1966; Barton, 1971; Yarmey, 1973; Koriat and Lieblich, 1974; Rubin, 1975; Goodglass et al., 1976; Marin et al., 1976; Feinberg et al., 1986; Kohn et al., 1987; Bub et al., 1988; Brown, 1991; Burke et al., 1991; Laine and Martin, 1996); and by evidence that lexical effects in real word repetition may paradoxically have an adverse effect on repetition success, as damaged connections from the concept representations to acoustic-articulatory motor pathway (engaged only when real words are repeated) serve as an additional source of phonemic paraphasias (Dubois et al., 1973; Caplan et al., 1986; Martin et al., 1999). The dual route model presented here is not simply a heuristic model. Rather, it represents a hypothesis about neural network structure. Computational models incorporating connectionist principles are now widely accepted as a valid representation of neural network structure and a key to understanding neural network function (Buonomano and Merzenich, 1998; Rolls and Treves, 1998; Rolls and Deco, 2002). In the domain of cognitive function, connectionist models have provided the means for understanding complex behavior in terms of the behavior of neurons, and computer simulations involving connectionist models have proven remarkably capable of recapitulating both normal human behavior and the behavior of brain-damaged individuals (see e.g., Cohen et al., 2000). Phonological Paraphasic Errors as Reflection of Underdeveloped Networks Phonological paraphasic errors are generally thought to reflect damage to dominant hemisphere networks supporting phonological processing. Nonpropositional spoken language, which is often supported by the nondominant hemisphere (Speedie et al., 1993), does reflect sequence knowledge but discrete phrasal, lexical or sublexical elements of this knowledge cannot be selected at will as they can in propositional language processing. However, given the extensive damage to this patient’s left hemisphere, his performance when Naming 67 naming after phonemic cueing suggests that he was using discretely accessible phonological sequence knowledge represented in his right hemisphere. This suggests in turn that deficient development of networks instantiating this knowledge (as contrasted to damage to fully developed networks) can provide the basis for phonological paraphasic errors. Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 Factors Influencing the Pattern of Errors in Internally Generated Language In the two route naming model we have introduced, three factors may influence the pattern of errors observed in internally generated spoken language in perisylvian aphasias (and whether this pattern is marked exclusively by impaired word retrieval, or additionally, by phonemic paraphasic errors). First, given the likely anatomic representation of the network shown in Figure 1, most dominant perisylvian lesions probably damage both the whole word and the phonological output routes (see Figure 3) and the pattern of spoken output may reflect the relative degree to which these two pattern associator networks are affected. This would explain why patients with Wernicke’s or conduction aphasia apparently do not have the option of relying entirely on the whole word route. Second, these two output pattern associator networks are likely to be differentially represented in the two hemispheres, the phonological pathway being more frequently better developed in the dominant hemisphere and the whole word pathway more equally developed in the two hemispheres. Subjects with dominant hemisphere lesions almost invariably demonstrate impaired if not completely absent phonologic sequence knowledge but often exhibit partial sparing of lexical semantic knowledge. Studies of patients with callosal disconnection demonstrate that the disconnected right hemisphere has a phonological input lexicon and conceptual semantic knowledge but impoverished phonological processing (Zaidel et al., 2003). Third, there may be individual variability in the degree to which connectivity is developed in these two output routes and this individual variability may vary as a function of hemisphere. Thus, in order to fully understand language production following a left-hemisphere lesion, a bi-hemispheric language model that incorporates both developmental attributes and the impact of the lesion must be considered (Figure 4). As our case suggests, in the presence of an extensive left hemisphere perisylvian lesion, deficient development of connectivity in nondominant hemisphere concept representations may lead to the generation of anomia and semantic paraphasic errors, and deficient development of connectivity in the nondominant hemisphere phonological route may lead to the generation of phonemic paraphasic errors. The deficient development of both systems, particularly the phonological one, may be a general characteristic of the right hemisphere, but deficient development of the phonological route may also occur to one degree or another in the left hemisphere as a result of normal variability in phonological network ontogenesis (Goodglass, 1993; Plaut et al., 1996). Fig. 3. Cartoon depicting the network of Figure 1 mapped onto the brain. Color coding as in Figure 1. Concept representations are assumed to be widely distributed across association cortices throughout the brain. In this cartoon, only the region of presumed interface between concept representations and the remainder of the model is depicted. Given the paucity of information about the anatomic organization of the human perisylvian region, the mapping depicted here is, at best, approximate; our goal is primarily to demonstrate the feasibility of mapping a connectionist architecture of phonologic processing to cortical anatomy. However, recent MRI diffusion tensor imaging tractographic studies of deep white matter pathways are shedding some light on the anatomic details. Catani et al. (2005) have delineated two dominant perisylvian pathways linking Wernicke’s and Broca’s areas, a direct one, corresponding to the arcuate fasciculus (possibly corresponding to pathway 3 in our model), and an indirect one, projecting from Wernicke’s area to inferior parietal cortex (Brodmann’s areas 39 and 40), with apparent relay to Broca’s area via what is likely component III of the superior longitudinal fasciculus (Makris et al., 2005) (possibly corresponding to pathway 1–2 in our model). They concluded that the broad extent of origin and termination of these pathways and their large cross sections favored a connectionist account. In summary, with any given left hemisphere lesion, the actual pattern of internally generated spoken language may reflect three factors: effects of the lesion on the two output routes (phonological and whole word), the degree to which one or both of these routes is represented in the right hemisphere, and the degree of development of connectivity in each of the two routes in the two hemispheres. If both naming routes are involved and the combined effect of the three factors differentially impacts the phonological routes, the net result will be output marked predominantly by word retrieval deficits (with or without semantic paraphasic errors) as whole word routes become the predominant means of language production (as in repetition conduction aphasia). If whole word routes are differentially impacted, the net result will be output marked by word retrieval deficits and a substantial incidence of phonemic paraphasic errors as damaged or inadequately developed phonological routes become the predominant means of language production. Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 68 Fig. 4. Cartoon depicting mapping of a bi-hemispheric model to the brain. This provides the basis for a fuller explanation of the results of left hemisphere lesions on language output in terms of bihemispheric contributions, the effect of the lesion, and the degree to which various networks are developed in each hemisphere. Color coding as in Figures 1 and 3. HU = hidden units. Patients with reproduction conduction or Wernicke’s aphasia predominantly use damaged or inadequately developed phonological pathways. Wernicke’s aphasia may reflect more severe impairment, hence greater difficulty with word retrieval and more profuse phonemic paraphasic errors than with conduction aphasia. Patients with Wernicke’s aphasia may also have damage to acoustic representations or the acoustic representations-concept representations pathways that enable verbal comprehension. Naming difficulty may arise through mechanisms discussed in the preceding paragraph or in two additional ways. First, damage to Brodmann’s areas 37 and 39 in the dominant hemisphere may be associated with word retrieval deficits (Hart and Gordon, 1990; Chertkow et al., 1997; Raymer et al., 1997; Whatmough et al., 2002). These areas may constitute the interface between association cortices throughout the brain supporting concept representations and the phonological apparatus (proximal to the point at which the phonological route provides differential access to sequence knowledge). Second, dysfunction of conceptual (semantic) networks because of left hemisphere damage and deficient development of right hemisphere networks, as in our patient, would be expected to yield naming difficulty with production of semantic paraphasic errors. Processes Occurring during Recovery The dual route naming model, as elaborated in the foregoing, can also account for the patterns of evolution of internally H. L. Roth et al. generated spoken language observed during recovery. First, recovery of connectivity in the dominant hemisphere phonological route, or enhancement of deficient connectivity in phonological routes in either hemisphere, would lead to increased naming success and a reduction in phonemic paraphasic errors. Second, because articulatory motor representations normally depend on input from both the whole word and phonological routes for activation, damage to either route could lead to anomia as representations fail to reach threshold for activation (a diaschisis effect). Changes may occur in the connectivity within articulatory representations or in surviving connectivity within the two naming routes such that this diaschisis effect diminishes with time. Third, in normal brain development, neural network connectivity in the two hemispheres presumably evolves such that transcallosal pathways facilitate coordinated bi-hemispheric activity and do not impede independent uni-hemispheric activity. A lesion may disrupt this precise inter-hemispheric coordination (another form of diaschisis), potentially rendering transcallosal input dysfunctional until appropriate readjustment of connectivity within and between hemispheres occurs in the course of recovery. With such readjustment, word retrieval may improve as the nondominant whole word route achieves a progressively greater transcallosal impact on dominant hemisphere spoken language output. Finally, because phonemic paraphasias have such an adverse effect on communication, patients like ours, in order to make themselves more readily understood, may be able to learn a strategy of using only the whole word route and semantically “boxing in” the concept they are trying to convey. Language Rehabilitation The presence of dual routes for naming may have implications for language rehabilitation. Currently, treatment of word retrieval deficits in aphasia consists overwhelmingly of practicing naming to confrontation, at times in conjunction with semantic or phonological cueing (Wilshire and Coslett, 2000; Nadeau and Rothi, 2004). Knowledge acquired in this way is limited to the information provided during training, it tends to generalize minimally (to semantically related words), and it is unlikely to expand over time. However, to the extent that word retrieval deficits are related to loss of phonological sequence knowledge, retraining of this knowledge may improve these deficits. Furthermore, unlike knowledge acquired in therapy that focuses on whole word production, knowledge of a repertoire of phonological sequences should generalize to all words that employ these sequences, including words that the patient might have the opportunity to relearn after completion of therapy. In this way, phonological sequence therapy for anomia might recapitulate the normal language learning process as it occurs in young children (Gathercole, 1995; Nadeau, 2001). It is likely that for such therapy to work, there must be some residual phonological sequence knowledge to build upon, as in our patient. Naming Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 References Barton MI. Recall of generic properties of words in aphasic patients. Cortex 1971; 7: 73–82. Benton AL, Hamsher KdeS. Multilingual Aphasia Examination. Iowa City: AJA Associates, 1989. Best W. When racquets are baskets but baskets are biscuits, where do the words come from? A single case study of formal paraphasic errors in aphasia. Cogn Neuropsychol 1996; 13: 443–480. Brown AS. A review of the tip-of-the-tongue experience. Psychol Bul 1991; 109: 204–223. Brown R, McNeill D. The “tip of the tongue” phenomenon. J Verb Learning Verb Behav 1966; 5: 325–337. Brédart S, Valentine T. From Monroe to Moreau: An analysis of face naming errors. Cognition 1992; 45: 187–223. Bub DN, Black S, Hamson E, Kertesz A. Semantic encoding of pictures and words. Some neuropsychological observations. Cogn Neuropsychol 1988; 5: 27–66. Buonomano DV, Merzenich MM. Cortical plasticity: From synapses to maps. Annu Rev Neurosci 1998; 21: 149–186. Burke DM, MacKay DG, Worthley JS, Wade E. On the tip of the tongue: What causes word finding failures in young and older adults? J Mem Lang 1991; 30: 542–579. Caplan D, Vanier M, Baker C. A case study of reproduction conduction aphasia I: Word production. Cogn Neuropsychol 1986; 3: 99–128. Caramazza A, Basili AG, Koller JJ, Berndt RS. An investigation of repetition and language processing in a case of conduction aphasia. Brain and Lang 1981; 14: 235–271. Caramazza A, Miceli G, Villa G. The role of the (output) phonological buffer in reading, writing, and repetition. Cogn Neuropsychol 1986; 3: 37–76. Carroll JB, White MN. Age-of-acquisition norms for 220 picturable nouns. J Verb Learn Verb Behav 1973; 12: 563–576. Catani M, Jones DK, Fytche DH. Perisylvian language networks of the human brain. Ann Neurol 2005; 57: 8–16. Chertkow H, Bub D, Deaudon C, Whitehead V. On the status of object concepts in aphasia. Brain & Lang 1997; 58: 203–232. Cohen G, Johnston RA, Plunkett K, editors. Exploring cognition: Damaged brains and neural networks. Philadelphia: Taylor and Francis Group, 2000. Dell GS. Phonological and lexical encoding in speech production: An analysis of naturally occurring and experimentally elicited speech errors. Unpublished doctoral dissertation, University of Toronto, 1980. Dell GS, Juliano C, Govindjee A. Structure and content in language production: A theory of frame constraints in phonological speech errors. Cogn Science 1993; 17: 149–195. Dell GS, Reich PA. Stages in sentence production: An analysis of speech error data. J Verb Learning Verb Behav 1981; 20: 611–629. Dell GS, Schwartz MF, Martin N, Saffran EM, Gagnon DA. Lexical access in normal and aphasic speakers. Psychol Rev 1997; 104: 801–838. Dubois J, Hécaen H, Angelergues R, Maufras de Chatelier A, Marcie P. Neurolinguistic study of conduction aphasia. In: H Goodglass, S Blumstein, editors. Psycholinguistics and aphasia. Baltimore: Johns Hopkins University Press, 1973; 283–300. Ellis AW, Miller D, Sin G. Wernicke’s aphasia and normal language processing: A case study in cognitive neuropsychology. Cognition 1983; 15: 111–144. Ellis AW, Young AW. Human cognitive neuropsychology. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 1988. Feinberg TE, Gonzalez Rothi LJ, Heilman KM. ‘Inner speech’ in conduction aphasia. Arch Neurol 1986; 43: 591–593. Francis WN, Kucera H. Frequency analysis of English usage: lexicon and grammar. Boston: Houghton Mifflin, 1982. Friedrich FJ, Glenn CG, Marin OSM. Interruption of phonological coding in conduction aphasia. Brain Lang 1984; 22: 266–291. 69 Gagnon DA, Schwartz MF, Martin N, Dell GS, Saffran EM. The origins of form-related word and nonword errors in aphasic naming. Brain Cog 1995; 28: 192. Gagnon DA, Schwartz MF, Martin N, Dell GS, Saffran EM. The origin of formal paraphasias in aphasic’s picture naming. Brain Lang 1997; 59: 450–472. Gathercole SE. Is nonword repetition a test of phonological memory or long-term knowledge? It all depends on the nonwords. Memory and Cognition 1995; 23: 83–94. Gilhooly KJ, Logie RH. Age-of-acquisition, imagery, concreteness, familiarity and ambiguity measures of 1944 words. Behav Res Methods Instrumentation 1980; 12: 395–427. Goodglass H. Understanding aphasia. San Diego: Academic Press, 1993. Goodglass H, Kaplan E, Weintraub S, Ackerman N. The “tip of the tongue” phenomenon in aphasia. Cortex 1976; 12: 145–153. Goodglass H, Wingfield A, Hyde MR, Gleason JB, Bowles NL, Gallagher RE. The importance of word initial phonology: Error patterns in prolonged naming efforts by aphasic patients. JINS 1997; 3: 128–138. Harley TA. A critique of top-down independent levels models of speech production: Evidence from non-plan-internal speech errors. Cogn Science 1984; 8: 191–219. Harley TA. Phonologic activation of semantic competitors during lexical access in speech production. Lang Cogn Process 1990; 8: 291–309. Hart J, Gordon B. Delineation of single-word semantic comprehension deficits in aphasia, with anatomical correlation. Ann Neurol 1990; 27: 226–231. Kaplan E, Goodglass H, Weintraub S, Segal O. Boston Naming Test. Philadelphia: Lea and Febiger, 1983. Kertesz A. Western Aphasia Battery. New York: Grune and Stratton, 1982. Kohn SE, Wingfield A, Menn L, Goodglass H, Gleason JB, Hyde M. Lexical retrieval: The tip-of-the-tongue phenomenon. Applied Psycholinguistics 1987; 8: 245–266. Koriat A, Lieblich I. What does a person in a “TOT” state know that a person in a “don’t know” state doesn’t know. Memory and Cognition 1974; 2: 647–655. Laine M, Martin N. Lexical retrieval deficit in picture naming: Implications for word production models. Brain Lang 1996; 53: 283–314. Lichtheim L. On aphasia. Brain 1885; 7: 433–484. MacKay DG. Complexity in output systems: evidence from behavioral hybrids. Am J Psychol 1973; 86: 785–806. Makris N, Kennedy DN, McInerney S, Sorensen AG, Wang R, Caviness VS, et al. Segmentation of subcomponents within the superior longitudinal fascicle in humans: A quantitative, in vivo, DT-MRI study. Cereb Cortex 2005; 15: 854–869. Marin OM, Saffran EM, Schwartz MF. Dissociation of language in aphasia: Implications for normal function. In: HB Stekles, SR Harnard, J Lancaster, editors. Origins and evolution of language and speech. New York: New York Academy of Sciences, 1976. Martin N, Saffran EM. A computational account of deep dysphasia: Evidence from a single case study. Brain Lang 1992; 43: 240–274. Martin N, Saffran EM, Dell GS. Recovery in deep dysphasia: evidence for a relation between auditory-verbal STM capacity and lexical errors in repetition. Brain Lang 1996; 52: 83–113. Martin N, Weisberg RW, Saffran EM. Variables influencing the occurrence of naming errors: Implications for a model of lexical retrieval. J Mem Lang 1989; 28: 462–485. Martin RC, Lesch MF, Bartha MC. Independence of input and output phonology in word processing and short-term memory. J Mem Lang 1999; 41: 3–29. McCarthy R, Warrington EK. A two-route model of speech production. Evidence from aphasia. Brain 1984; 107: 463–485. Miller D, Ellis AW. Speech and writing errors in “neologistic jargonaphasia”: A lexical activation hypothesis. In: M Coltheart, G Sartori, R Job, editors. The cognitive neuropsychology of language. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1987; 253–271. Motley MT, Baars BJ. Semantic bias effects on the outcomes of verbal slips. Cognition 1976; 4: 177–187. Downloaded by [Universitat Politècnica de València] at 01:33 26 October 2014 70 Nadeau SE. Phonology: A review and proposals from a connectionist perspective. Brain Lang 2001; 79: 511–579. Nadeau SE, Rothi LJG. Rehabilitation of language disorders. In: J Ponsford, editor. Cognitive and behavioral rehabilitation: From neurobiology to clinical practice. New York: Guilford, 2004; 129–174. Patterson K, Shewell C. Speak and spell: dissociations and word-class effects. In: M Coltheart, G Sartori, R Job, editors. The cognitive neuropsychology of language. London: Lawrence Erlbaum Associates, 1987; 273–294. Plaut DC, McClelland JL, Seidenberg MS, Patterson K. Understanding normal and impaired word reading: Computational principles in quasiregular domains. Psychol Rev 1996; 103: 56–115. Raymer AM, Foundas AL, Maher LM, Greenwald ML, Morris M, Rothi LJ, et al. Cognitive neuropsychological analysis and neuroanatomic correlates in a case of acute anomia. Brain Lang 1997; 58: 137–156. Rolls ET, Deco G. Computational neuroscience of vision. Oxford: Oxford University Press, 2002. Rolls ET, Treves A. Neural networks and brain function. New York: Oxford University Press, 1998. Rubin DC. Within word structure in the tip-of-the-tongue phenomenon. J Verb Learn Verb Behav 1975; 14: 392–397. Saffran EM, Marin OSM. Immediate memory for word lists in a patient with deficient auditory short-term memory. Brain Lang 1975; 2: 420–433. Shuren J, Heilman KM. Non-optic aphasia. Neurology 1993; 43: 1900–1907. H. L. Roth et al. Snodgrass JG, Vanderwart M. A standardized set of 260 pictures: Norms for naming agreement, familiarity, and visual complexity. J Exp Psychol: Hum Learning & Memory 1980; 6: 174–215. Speedie LJ, Wertman E, Ta’ir J, Heilman KM. Disruption of automatic speech following a right basal ganglia lesion. Neurology 1993; 43: 1768–1774. Valdois S, Joanette Y, Nespoulous J-L. Intrinsic organization of sequences of phonemic approximations: A preliminary study. Aphasiology 1989; 3: 55–73. Warrington EK, Shallice T. The selective impairment of auditory verbal short-term memory. Brain 1969; 92: 885–896. Whatmough C, Chertkow H, Murtha S, Hanratty K. Dissociable brain regions process object meaning and object structure during picture naming. Neuropsychologia 2002; 40: 174–186. Wilshire CE. Three “abnormal” features of aphasic phonologic errors. Brain & Lang 1998; 65: 219–222. Wilshire CE, Coslett HB. Disorders of word retrieval in aphasia: Theories and potential applications. In: SE Nadeau, LJG Rothi, B Crosson, editors. Aphasia and language. Theory to practice. New York: Guilford Press, 2000; 82–107. Yarmey AD. I recognize your face but I can’t remember your name: Further evidence on the tip-of-the-tongue phenomenon. Memory and Cognition 1973; 1: 287–290. Zaidel E, Iacoboni M, Zaidel DW, Bogen J. The callosal syndromes. In: KM Heilman, E Valenstein, editors. Clinical neuropsychology. New York: Oxford University Press, 2003; 347–403.