1

Semantic Memory: which side are you on?

Karalyn Patterson1,2, Michael D. Kopelman3, Anna M. Woollams4,

Sonia L.E. Brownsett5, Fatemeh Geranmayeh5 and Richard J.S. Wise5*

1. Department of Clinical Neurosciences, University of Cambridge

2. MRC Cognition & Brain Sciences, Cambridge

3. Institute of Psychiatry, Kings College London

4. School of Psychological Sciences, University of Manchester

5. Computational, Cognitive, and Clinical Neuroimaging Laboratory, Imperial College London

*Corresponding author:

Professor Richard J S Wise

Computational, Cognitive, and Clinical Neuroimaging Laboratory

Imperial College London,

Hammersmith Hospital, London, W12 0NN.

richard.wise@imperial.ac.uk

Abstract

Patients with the neurodegenerative condition known as semantic dementia (SD) have provided behavioural and brain-imaging data of considerable importance in understanding the organisation and functional anatomy of semantic memory. The majority of SD patients present with more prominent atrophy of the left rostroventral temporal lobe (rvTL), but they also have a degree of atrophy in the mirror region on the right that progresses. SD therefore provides no clear evidence about the specific role of the left rvTL. We had the opportunity to study two unusual patients. One had had a single lobar stroke with an unusual distribution, largely destroying the whole of the left temporal lobe ventral to the superior temporal sulcus. The other patient had had herpes simplex encephalitis with destruction that, based on imaging data, was entirely confined to the left cerebral hemisphere. The distribution of this damage was very similar to that of the stroke patient, mainly affecting the temporal lobe, though also involving the left inferior frontal gyrus. Cognitive outcomes in the two patients were compared with published results from patients with SD and patients with left fronto-parietal strokes resulting in so-called semantic aphasia. The two patients showed a highly consistent cognitive profile, similar in many respects to that observed in patients with mild-moderate SD, with impaired performance on non-verbal as well as verbal tasks. Certain key features of the SD profile, howeverincluding striking sensitivity to the familiarity and typicality of the stimulus materials in all expressive and receptive tasksappeared only in tasks requiring verbal output in these two patients with unilateral left temporal lesions. Results in these cases address the debate about whether the left temporal lobe processes and represents verbal semantics while the right is specialised for non-verbal semantics, or whether there is a more amodal distribution of semantic knowledge between both temporal lobes, but with the left having direct access to verbal output. The evidence presented here strongly favours the second hypothesis. Comment by PA to Prof Kopelman: [MDK1] I know where you are coning from, but it seems to me a bit unusual to begin an Abstract about a condition (SD) for which we are not actually presenting new data. My inclination would be to begin the abstract along the lines: In this paper, we describe two patients with severe anomia but only mild/minimal semantic impairments, resulting from extensive left temporal lobe damage of differing aetiologies. One had had a single lobar stroke ... ... We will compare these findings with those obtained in previous studies of semantic dementia (SD) patients, who have prominent atrophy of the left rostroventral temporal lobe (rvTL). ...]

Introduction

Semantic memory became the subject of serious experimental study about 40 years ago, following the proposal that episodic and semantic knowledge are distinct forms of long-term memory (Tulving and Donaldson, 1972). In clinical settings, it was common for patients to present with a predominant impairment in episodic memory, with difficulty encoding specific life events, such as where they went and what they did on their last holiday. In contrast, patients with a predominant impairment of semantic memory were not recognised until the clinical neuropsychological study of Warrington (Warrington, 1975). Such patients, with what is now known as the semantic variant of fronto-temporal dementia, or semantic dementia (SD), demonstrated a progressive loss of conceptual knowledge while most other cognitive abilities remained largely unaffected for some years. Comment by PA to Prof Kopelman: {MDK2] I have inserted a paragraph break here so that the opening paragraph appears more general see my comments on the Abstract. Again, this paper is not so much about SD per se as to demonstrate what these two patients tell us about the hemispheric underpinning of semantics and semantic memory.

Common manifestations of SD include failure to recognise the meanings of words, to name or draw objects, to recall the properties and uses of man-made artefacts, and to recall widely known facts about famous people or places. Once the clinical descriptions of SD began to appear in the literature (Hodges et al., 1992; Snowden et al., 1989), cases were increasingly identified by clinicians and referred for specialist study. Imaging data and, to a lesser extent, post mortem studies in these patients have been analysed to identify the neuroanatomical correlates of semantic memory. Clinical studies have been supplemented by functional neuroimaging studies on healthy participants as they performed semantic tasks. Focal regions of maximum atrophy were presumed to constitute an important component of the network for conceptual knowledge in the healthy brain; and although the extent and relative importance of the parts of this network are not yet a matter of complete consensus, the consistent lesion location in SD is the rostroventral temporal lobes (rvTL) (Acosta-Cabronero et al., 2011; Agosta et al., 2010; Binney et al., 2010; Mion et al., 2010; Snowden et al., 1996). Brain atrophy in this region in SD is dramatic, with 50-80% grey-matter loss even in mild-moderate stages of the condition (Hodges and Patterson, 2007; Mesulam et al., 2013). Comment by Fatemeh Geranmayeh: There are two snowden 1996 papers. is this the correct one?

Like most neurodegenerative conditions, SD is a bilateral disease, though typically asymmetrical at least until the later stages. In the majority of cases, the left temporal lobe shows the most atrophy (SD-L), while in a smaller number the atrophy is either greater on the right (SD-R) or approximately symmetrical. In a cohort of 15 SD patients (Snowden et al., 2004), the numbers of cases in these three sub-groups were ten, three and two, respectively. Unsurprisingly, researchers have attempted to determine whether SD-L and SD-R subgroups differ reliably in their cognitive profile. To some extent they do, and in a manner that accords with what is known about specialisation in the two hemispheres. In the same study, for example, when asked to supply information about famous people such as Tony Blair or David Beckham, SD-L patients performed more poorly when probed with the peoples names whereas SD-R cases had more trouble when the stimuli were faces. Other researchers have likewise demonstrated that performance on tasks requiring verbal output, such as naming objects or generating the names of members of a category, is more impaired in SD-L than SD-R (Lambon Ralph et al., 2001; Mion et al., 2010; Thompson et al., 2003). In these and other studies, however, both groups were substantially impaired relative to controls on all tests requiring semantic memory, whatever the modality of input or output. Therefore, although some researchers have placed considerable weight on the cognitive differences (e.g. (Gainotti, 2012)), others consider that SD results in a consistent pattern of semantic deficit, in which SD-L and SD-R cases are more similar than different.

As atrophy in SD is bilateral, if often asymmetrical, such cases do not provide clear evidence on the contribution, jointly or separately, of the left and right temporal lobes to the various representations and processes of semantic memory. Two studies by Lambon Ralph and colleagues (Lambon Ralph et al., 2010; 2012) each identified 20 patients with unilateral temporal-lobe lesions: the former with a variety of aetiologies but mostly surgical resections for tumours or Temporal Lobe Epilepsy, the latter all with resections for TLE. When assessed with the tests commonly administered in research on SD, the patients in these studies did exhibit some semantic deficits relative to matched controls, especially on more demanding tests or (as in (Lambon Ralph et al., 2012)) on reaction-time measures. In neither study, however, did the performance of the patients with either unilateral left or right lesions resemble SD, suggesting that the deterioration and ultimately profound impairment of semantic memory characteristic of SD is the result of bilateral temporal lobe pathology. Nevertheless, it can be argued that the growth of tumours and the chronicity of repeated focal epileptic discharges allow time for reorganisation of function. Furthermore, the areas resected in TLE are more discrete than the regions of atrophy and hypometabolism in SD.

The most robust evidence regarding the contribution of the rostroventral temporal lobe in one hemisphere would come from patients with an acute or sub-acute onset of focal damage in this region; but such patients are very rare. Owing to the distribution of the arterial supply that is vulnerable to thromboembolism, lesions consequent on stroke are almost never confined to this region. Herpes simplex virus encephalitis (HSVE) almost invariably affects the ventromedial temporal lobes, but is rarely a unilateral disease (for example, see Figure 9 in (Kopelman et al. (, 1999); and Figure 2 in (Reed et al. (, 20