Semantic naming errors 1 Mimicking aphasic semantic errors in normal speech production: Evidence from a novel experimental paradigm

Semantic errors are commonly found in semantic dementia (SD) and some forms of stroke aphasia and provide insights into semantic processing and speech production. Low error rates are found in standard picture naming tasks in normal controls. In order to increase error rates and thus provide an experimental model of aphasic performance, this study utilised a novel methodtempo picture naming. Experiment 1 showed that, compared to standard deadline naming tasks, participants made more errors on the tempo picture naming tasks. Further, RTs were longer and more errors were produced to living items than non-living items a pattern seen in both semantic dementia and semantically-impaired stroke aphasic patients. Experiment 2 showed that providing the initial phoneme as a cue enhanced performance whereas providing an incorrect phonemic cue further reduced performance. These results support the contention that the tempo picture naming paradigm reduces the time allowed for controlled semantic processing causing increased error rates. This experimental procedure would, therefore, appear to mimic the performance of aphasic patients with multi-modal semantic impairment that results from poor semantic control rather than the degradation of semantic representations observed in semantic dementia (Jefferies & Lambon Ralph, 2006). Further implications for theories of semantic cognition and models of speech processing are discussed. Semantic naming errors 3 Semantic memory is our store of meanings and factual knowledge. It allows the comprehension of our environment and underpins our ability to communicate effectively both verbally and nonverbally (Rogers, Lambon Ralph, Garrard, Bozeat, McClelland, Hodges & Patterson, 2004). Impairments to semantic cognition can be devastating and can result from neurodegenerative disease such as semantic dementia or aphasia due to cerebral vascular accident (CVA or stroke: Jefferies & Lambon Ralph, 2006). Semantic dementia (SD) is a neurodegenerative condition where all types of concepts slowly degrade whilst other cognitive and language functions remain relatively intact (Hodges, Patterson, Oxbury & Funnell, 1992; Snowden, Goulding & Neary, 1989). Semantic dementia is caused by progressive bilateral atrophy of the anterior and inferior temporal cortex (Lambon Ralph, McClelland, Patterson, Galton & Hodges, 2001; Mummery, Patterson, Price, Ashburner, Frackowiak & Hodges, 2000). Whilst there is a progressive degradation of semantic representations, this decline is not random and follows the same general pattern. For example, patients demonstrate better knowledge for general properties of objects than specific features in both receptive and expressive tasks (Hodges, Graham & Patterson, 1995; Warrington, 1975). This results in the production of category superordinates (e.g., “animal” for horse, elephant or whale). Furthermore patients frequently use more typical or familiar labels within a semantic category in place of less familiar/typical exemplars on naming tasks (e.g., "cat" for leopard, deer or goat; Hodges et al., 1995; Rogers et al., 2004). Alongside omissions, these two error types (semantic category superordinates and coordinates) dominate the naming performance of semantic dementia patients and can be explained in terms of the gradual deterioration of underlying semantic representations (Lambon Ralph, Graham, Ellis & Hodges, 1998; Lambon Ralph et al., 2001; Rogers et al., 2004). Semantic naming errors 4 A further feature of SD naming performance is differential error patterns to living (e.g., animals, plants) versus non-living (man-made artefacts) stimuli. In a study consisting of 15 patients, Rogers et al. (2004) compared the proportion of errors that were omissions, superordinate or semantic coordinates of the items to be named. For the living stimuli (birds, water creatures and land animals), there were more semantic coordinate and superordinate errors in relation to omissions. The non-living items (household objects, vehicles and musical instruments) showed the reverse pattern with increased omissions compared to the other error types. Using an implemented PDP model of conceptual knowledge, Rogers et al. were able to show that this difference in error type is due to the organisation of semantic memory: living things tend to be more tightly clustered in semantic space compared to non-living items. When these semantic representations degrade in SD, the tightly packed representations are more likely to be confused with each other, leading to the production of coordinate or superordinate semantic errors. In contrast, although the representations for non-living items also break down, their relative isolation within semantic space means that there is less opportunity for confusion with other concepts and so omission errors are the most likely outcome. This same difference in error patterns is seen in other patient groups, e.g., herpes simplex encephalitis (e.g., Barbarotto, Capitani & Laiacona, 1996; Borgo & Shallice, 2001; Warrington & Shallice, 1984) indicating that this is a general property of the semantic system. Impaired semantic cognition due to stroke results from lesions in the temporoparietal and/or prefrontal regions and is often associated with Wernicke’s, transcortical sensory and global aphasic subtypes (Berthier, 2001; Chertkow, Bub, Deaudon & Whitehead, 1997; Jefferies & Lambon Ralph, 2006). Whilst stroke aphasics with semantic impairment (herein termed semantic aphasics) produce category superordinates and coordinates they also produce additional errors including semantic associative errors Semantic naming errors 5 (e.g., ‘bone’ for the target dog) which are absent in SD patients (Jefferies & Lambon Ralph, 2006). Although semantic dementia and stroke aphasia can both affect semantic cognition, until recently little research has been carried out that directly compares the deficits associated with the two aetiologies. Jefferies and Lambon Ralph (2006) directly compared the two groups across a variety of semantic tasks including picture naming. In order to compare SD to the closest possible stroke aphasia model, their patients were selected on the basis of showing multi-modal semantic impairments in the context of aphasia. Half had a transcortical sensory aphasia classification (TSA – i.e., able to repeat but not understand). The others had a classification indicating primarily semantic impairments. This selection criterion meant that, in practice, none conformed to classical Wernicke’s aphasia (we refer to their aphasia group as semantic aphasia for short throughout this paper). Both groups, SD and stroke aphasia, produced the same proportion of correct responses (.41) and omissions (.37 and .32, SD patients and stroke aphasics respectively). The SD patients produced more semantic errors overall (.45 compared to .33) with a higher proportion of these as category coordinates and superordinates (.99 compared to .73 of the total semantic errors) and very few associate errors (.01) compared to the stroke aphasic group (.27). In addition to the error analyses, the two groups were compared on several further measures. Although the stroke aphasic and SD patients failed the same semantic tests and obtained relatively equivalent scores, there were clear qualitative differences between them (see Table 2, Jefferies & Lambon Ralph, 2006). The SD patients showed high correlations between scores on different semantic tasks and strong item consistency across different tests. This group were also highly sensitive to item familiarity/frequency. The stroke aphasic patients showed a different pattern. They were insensitive to the effects of familiarity/frequency. They only showed significant item consistency/correlations across Semantic naming errors 6 tasks requiring the same type of semantic judgement (e.g., word vs. picture semantic association tests.) Unlike SD patients, the semantic aphasia patients’ scores across different types of semantic task (e.g., semantic association vs. picture naming) did not correlate. A deregulation of semantic cognition (i.e., less precise executive control of semantic processing) rather than a degradation of core amodal semantic knowledge (as observed in SD) would seem to explain the behavioural profile of the semantic aphasic patients. Their deficit is multimodal (i.e., affects all verbal and nonverbal modalities of input and output) because all tasks, irrespective of which sensory/verbal modalities are involved, require at least some degree of semantic control. They demonstrate similar levels of semantic performance across different versions of the same semantic task (picture version of the Camel and Cactus test versus all word version of the task) because the semantic control requirements are held constant. However, this consistency drops away when comparing across different semantic tasks because the semantic control requirements change; although the aphasic patients may be able to regulate the activation of information appropriate for one task (e.g., naming), they may be unable to reshape the information required for another test/situation (e.g., word-picture matching) even though the same concept is being tapped (Jefferies & Lambon Ralph, 2006). The implications of “dimmed” or degraded (SD) vs. “deregulated” (semantic aphasia) semantic cognition were explored further using phonemic cueing and miscuing in picture naming (Jefferies, Patterson, Hopper, Corbett, Baker & Lambon Ralph, submitted). As expected, SD patients exhibited minimal effects of cueing and miscuing. In contrast, the aphasic patients (the patients tested in this study were the same as those reported in Jefferies and Lambon Ralph (2006) demonstrated considerable effects of both cueing and 1 A helpful reviewer noted that ‘reshaping’ information could be interpreted as conscious processing, we were not making claims as whether the ability to reshape the information is conscious or not. It is possible that it may be done without conscious awareness, but also a patient could consciously try to reject distractors during a word-to-picture matching task. In essence, this does not alter our contention that semantic control or controlled semantic processing is impaired in the semantic aphasia patients. Semantic naming errors 7 miscuing. Phonemic miscues (e.g., table + /ch/) increased the activation of specific, semantically-related distractor words, making these competitors more likely to be selected instead of the target word. When provided with such phonemic miscues, the stroke aphasics generated more semantic errors that were consistent with the miscue (e.g., table + /ch/ “chair”). Such results suggest that the presence or absence of cueing-miscuing effects can be used to distinguish between these different kinds of deficits of semantic cognition (see below). While considerable knowledge can be gained by studying patient populations, it is necessary to assess non-brain damaged participants as well. In addition to careful explorations of normal performance, it can be revealing to create experimental models of patient-like symptoms in normal participants by manipulating experimental tasks. Using both methods, one can better ascertain the relationship between normal and pathological performance. The first aim of this paper was to induce semantic naming errors in controls. If one looks at naming errors from a general aphasic population then various different types are observed (Lambon Ralph, Moriarty & Sage, 2002; Schwartz, Dell, Martin, Gahl & Sobel, 2006). We concentrated, however, on the semantic errors made by the subtype of patients studied by Jefferies and Lambon Ralph (2006) – that is semantic dementia patients and aphasic patients with multi-modal semantic impairment. We selected, therefore, experimental paradigms that were most suited to eliciting semantic errors predominantly and excluded other methods that tend to produce phonological and other speech errors. Specifically, we explored the novel application of a picture version of the tempo naming technique developed by Kello and Plaut (2000). Kello and Plaut introduced a new technique, known as tempo word naming to investigate subjective control and speed of responses. During tempo word naming, participants are presented with a series of evenly 2 It should be noted that in their original paper Kello & Plaut used the term tempo naming not tempo word naming as we have adopted here. We adopted tempo word naming in relation to their work as to avoid confusion with the picture-based tasks used in this study. Semantic naming errors 8 spaced beeps (forming a steady rhythm) along with a decreasing visual cue. The letter string to-be-read is presented on the final beep and the task is to pronounce the letter string in time with the beginning of the next beep (which does not actually occur). Using this method, participants’ naming times can be experimentally manipulated by slowing or quickening the tempo. In a series of experiments, Kello and Plaut (2000; Kello, 2004) compared the tempo word naming paradigm to a standard word naming paradigm under different conditions. Across all three experiments, word, nonword and articulatory errors increased as tempo increased (with a relative increase in lexicalisation errors). Kello and Plaut (2000) suggested that tempo word naming is a form of the more standard deadline naming method, which is widely used both in word naming and object naming studies. The obvious difference between the two paradigms is that in tempo naming there is an explicit and precise cue for when to initiate each response, rather than a single beep (or some other signal), often used in naming-to-deadline paradigms, which participants are asked to “beat”. Kello and Plaut (2003) used a PDP model to simulate the effects of the tempo word naming task. As well as an increased error rate, Kello and Plaut found that response duration also reduced as the tempo increased. They simulated all three effects (reduced response time and duration with an increased error rate) by forcing an increased processing speed in the network (increasing the unit input gain function (Kello & Plaut, 2003)). Importantly for the present study, these simulations indicated that a forced increase of processing speed comes at the expense of less controlled processing resulting in increased error rates. This follows from the increase in the gain function to all units: a response reaches threshold more rapidly (reducing response times) because the inputs to the relevant units are amplified but, in doing so, the model is less precise in activating the correct pattern, leading to errors. Semantic naming errors 9 The current study addressed four main questions: 1. Does the tempo procedure induce semantic errors, like those seen in stroke aphasic and semantic dementia (Jefferies & Lambon Ralph, 2006), when applied to picture naming in normal controls? 2. Is the tempo naming technique better than other methods of inducing errors in normal controls such as the standard naming-to-deadline paradigm (Vitkovitch & Humphreys, 1991; Vitkovitch, Humphreys & Lloyd Jones, 1993)? 3. Is the pattern of errors across domains (living/non-living) the same as that found in the patient groups? 4. If the tempo naming paradigm reduces semantic control then naming performance should become sensitive to the effects of cueing and miscuing as observed in semantic aphasia (Jefferies et al., submitted). Questions 1-3 are addressed in Experiment 1 and the final question is addressed in Experiment 2.