Concrete & Abstract Words in Alzheimer’s Disease: The influence of categories on processing

(1)

Concrete & Abstract Words in Alzheimer’s Disease:

The influence of categories on processing

Jet M.J. Vonk

jet_vonk@hotmail.com

University: University of Groningen Program: Research Master Linguistics

Course: Research Master's Thesis

First supervisor: Dr. Roel Jonkers Second supervisor: Dr. Loraine K. Obler

(2)

(3)

Abstract

(6)

ii

Acknowledgments

This thesis has occupied me for the last six months, but has definitely given me a great amount of satisfaction. When I first started my studies in Groningen in September 2006, I absolutely did not expect that it would end here. It is enjoyable to see that starting with a Bachelor's in Dutch Language and Culture and an interest in Journalism can end in an unlimited curiosity in the research of Neurolinguistics.

I owe an immensely great “thank you” to Dr. Roel Jonkers. He literally started and ended my six years of studentship, as I sharply remember my first class with him on a Monday morning in early September 2006. Through the years he taught me in several classes, supervised my Bachelor’s thesis, supervised my internship in New York and last, but definitely not least, supervised this Master’s thesis. His guidance gave me the ultimate opportunity to develop my skills and expertise. His approachability gave me the freedom to explore my own ideas about matters, which he critically reviewed every time and supported when they made sense, but rightly challenged in the cases they did not. I have been very fortunate with all his help and guidance over the years and hope to continue working with him in future research projects.

I would like to thank Dr. Loraine Obler very much for everything she has done for me in the past year. She gave me the opportunity to come to New York in September 2011 to do my research internship with her and taught me a lot in the six months I was there. Subsequently, as the second supervisor on my Master’s thesis, she gave helpful comments and great advice for which I am very grateful.

I am really grateful to Dr. Philip Scheltens, director at the Alzheimer Center VUmc in Amsterdam, whom I met at a conference in Miami and kindly allowed me to contact in case I ever needed participants for my study. Without him I do not think I would have succeeded in finding the patients with Alzheimer’s disease I needed for this thesis. I very much enjoyed attending his weekly consults and appreciate all the participants he referred to me. I would also like to thank Dr. Yolande Pijnenburg and Antoinette Keulen of the Alzheimer Center VUmc for their help. I found it very interesting to attend the weekly multidisciplinary meetings, which gave me a better insight into the aspects of dementia.

I owe a debt of gratitude to my friend Eve Higby for editing the manuscript on grammatical correctness of English. It means a lot to me that she altruistically read all the pages of this thesis very carefully and thoroughly, and provided comments not only on the English grammar, but also in some cases with respect to the content.

I would like to thank my parents, who have been supportive and encouraging during not only the process of this thesis, but during all the steps of my education. They have never doubted my capabilities and always given me the freedom to choose my own direction. Pushy as I am about my own aspirations, they knew when to support me and when to bring me down to earth again. They have always let me and my two brothers explore our own ideas and never put their own interest first; I am very grateful to them for everything they have done for us.

I am also thankful to all my friends for their encouragement, love, amusement, and playful nonsense that kept me going. They took me out at moments I needed to get away from my computer and ignored my calls and messages to go out in case they knew I needed to work on a deadline. They endlessly listened to my excitement about details they most probably had never heard of before and pretended to know or to be interested in them just to make me satisfied.

I would like to thank all the participants and their family members for their participation in this study. I also owe a big thank you to all the people who participated in the pilot studies and filled in the endless lists of words just to please me because I certainly know that could not have been fun.

(7)

iii

List of Tables

Table 1. The five semantic features of verbs and their neural substrates organized … 19

as a function of verb class as studied by Kemmerer et al. (2008)

Table 2. Demographical information of the participants … 20

Table 3. Test stimuli … 21

Table 4. Mean values, standard deviations (SD) and median values of … 22

(8)

iv

List of Figures

Figure 1. Mean accuracy in nouns and verbs in the control and AD group … 24

Figure 2. Mean reaction time in nouns and verbs in the control and AD group … 24

Figure 3. Mean accuracy in the different object categories in the control and AD group … 25

Figure 4. Mean reaction time in the different object categories in the control and … 25

AD group

Figure 5. Mean accuracy in the different action categories in the control and AD group … 25

Figure 6. Mean reaction time in the different action categories in the control and … 25

AD group

Figure 7. Mean accuracy in concrete words and abstract words in the control and … 26

AD group

Figure 8. Mean reaction time in concrete words and abstract words in the control … 26

and AD group

Figure 9. Mean accuracy in abstract nouns and verbs in the control and AD group … 26

Figure 10. Mean reaction time in abstract nouns and verbs in the control and AD group … 26

Figure 11. Mean accuracy in the different abstract noun categories in the control and … 27

AD group

Figure 12. Mean reaction time in the different abstract noun categories in the control … 27

and AD group

Figure 13. Mean accuracy in the different abstract verb categories in the control and … 28

AD group

Figure 14. Mean reaction time in the different abstract verb categories in the control … 28

(9)

1

1. Introduction

Words and the way in which words are processed have been the topic of hundreds of psycho- and neurolinguistic studies. An often studied aspect of this topic is the processing of nouns versus verbs, different semantic categories in words, and the concreteness of words. These topics are of particular interest because they give us insight into the cognitive and neural representation of words in the brain. However, the findings in different studies often do not align well with each other and are even contradictory in several cases. This is the case for both behavioral studies and studies that use brain imaging methods.

A valuable direction in the research on the cognitive and neural representation of words is the study of language-impaired populations. The effect of brain damage on language has provided us a basis on which anatomical-behavioral correlations can be drawn, connecting a repeatedly seen language deficit found in a specific group to their uniform neural damage. One of the populations that has recently received more research is individuals with Alzheimer’s disease (AD), although also in this population various results have been found.

In each of the above mentioned research topics (i.e., nouns versus verbs, different semantic categories in words, and the concreteness of words) multiple theories and explanations have been proposed to account for the (disparate) results. For each theory there are studies that either support or challenge that specific account, resulting in a lack of consensus on the appropriate theory that should be applied.

Although the research on nouns versus verbs, different semantic categories in words, and the concreteness of words have often been studied separately, or at best in combinations of two of these, there are few studies that unite them. In the present study, the focus will be on all three areas, putting forward specific expectations based on a thorough review of the findings so far. This literature review is presented in the first chapter of the thesis, followed by a description of the methodology of the current study in the second chapter. The third chapter presents the results of this study and in the final, fourth chapter, these results and related issues will be discussed.

1.1 Nouns and verbs

The topic of nouns and verbs and their organization in both cognitive and neuroanatomical ways has been of interest for decades. Researchers have studied healthy and impaired populations, used behavioral and objective measures, and proposed a widespread range of theories concerning this topic throughout the years. However, no consensus has yet been reached on this topic.

Joined with the question of the organizing principle of nouns and verbs is the question of what the exact neural substrates of nouns and verbs are. In behavioral studies this question has been looked at by correlating specific areas of brain damage with category-specific language impairment. In neuroimaging studies, tasks have been developed to show cortical activation in healthy populations. But even with these tools, also in this matter there is no general agreement.

In this chapter an overview will be given of the behavioral studies on impaired populations and the anatomical-behavioral correlations that have been made, the neuroimaging and electrophysiological studies that have been done, and the theories and explanations researchers have developed about the organization of lexical information and its cortical distribution. Furthermore, some methodological concerns will be discussed that have come up regarding this topic.

1.1.1 Behavioral studies

(10)

2 deficits (Mätzig, Druks, Masterson, & Vigliocco, 2009). In general, these studies have been taken as strong evidence that several linguistic components are separately organized in the brain. However, the answer to the question about how this organization actually operates and functions is still heavily discussed. This is because many studies show conflicting results, on the one hand because populations with comparable brain damage show differently impaired linguistic components and on the other hand because different sorts of brain damage show identical impaired linguistic components.

1.1.1.1 Impaired populations

A great deal of research on impairment of nouns and verbs has been done on aphasic speakers. This language disorder, mostly caused by a stroke or trauma, has a wide variety of types and symptoms. Often studied on their noun and verb processing are patients with Broca’s aphasia (also often referred to as agrammatic or non-fluent aphasia), Wernicke’s syndrome (also often referred to as fluent aphasia), and anomic aphasia. Initially, specific deficits observed in aphasia were seen in verbs, and taken together with the belief that even for healthy speakers verbs, in general, are more difficult to process than nouns (Berndt, Mitchum, Haendiges, & Sandson, 1997), this led to the hypothesis that verbs are simply more complex (e.g., Saffran, Schwartz, & Marin, 1980a, 1980b). This hypothesis had to be abandoned, though, when studies were published showing several aphasic patients with a noun-specific deficit (e.g, Miceli, Silveri, Villa, & Caramazza, 1984).

One of the first studies to describe a relationship between the type of aphasia and the specific deficit in nouns or verbs was Miceli et al. (1984). They found that agrammatic speakers named verbs significantly worse than nouns, while anomic speakers showed the opposite pattern. This exact finding was also reported in the study by Zingeser and Berndt (1990). The studies of McCarthy and Warrington (1985) and Chen and Bates (1998) showed a verb deficit in agrammatic speakers as well. In addition, Chen and Bates (1998) showed a noun deficit for Wernicke’s aphasia. The finding by Miceli et al. (1984) on anomic speakers was supported by Miozzo, Soardi and Cappa (1994), who found that action naming was spared in a case of pure anomia. Although these findings strongly suggest a correlation between type of aphasia and a specific deficit in either nouns or verbs, other studies have shown contradictory results. Williams and Canter (1987) as well as Berndt, Mitchum, Haendiges, and Sandson (1997) documented patients with Wernicke’s aphasia who had greater difficulty with verbs than nouns, and Breedin and Martin (1996) described an anomic patient who was severely impaired on verbs.

In addition to the population of aphasic speakers, other brain-damaged populations have also been studied on their ability to process nouns and verbs. Among these are several syndromes of dementia, like frontotemporal dementia (FTD), Alzheimer’s disease (AD) and primary progressive aphasia (PPA). For example, Caño et al. (2010) described the performance of a patient with non-fluent PPA, who had worse verb than noun naming. Hillis, Oh and Ken (2004) also reported that patients with non-fluent PPA are more impaired on verbs than nouns, while patients with fluent PPA show the opposite pattern. Cappa, Binetti, Pezzini, Padovani, Rozzini, and Trabucchi (1998) reported on FTD and AD patients, showing that both groups were impaired on nouns. Contrasting herewith is the study by Rhee, Antiquena, and Grossman (2001) that reported FTD patients to be significantly less accurate and slower on verbs than nouns.

1.1.1.2 Anatomical-behavioral correlations

(11)

3 the involvement of distinct brain mechanisms in the processing of these two categories (Shapiro et al., 2005).

Damasio and Tranel (1993) stated more precisely that the left anterior midtemporal and inferior temporal regions are responsible for the retrieval of concrete nouns, based on two patients. A third patient, who had a lesion in the left premotor cortex (the posterior segment of the left inferior frontal gyrus and the anterior segment of the left precentral gyrus), was thought to provide evidence that retrieval for verbs is processed in the left frontal region. Therefore, Damasio and Tranel (1993) put forward the fronto-temporal dichotomy hypothesis (FTDH). In a later study, Tranel, Adolphs, Damasio and Damasio (2001) found that 75 subjects with action naming deficits overlapped maximally in lesions in the left frontal operculum, in the underlying white matter, and in the anterior insula. In addition, lesions of the left anterior temporal and inferior temporal regions resulted in impaired naming of concrete entities, but not in impaired action naming.

However, even in the anatomical-behavioral correlative point of view, multiple exceptions for the lesion deficit pattern have been found. De Renzi and di Pellegrino (1995) reported a patient with a large left prefrontal lesion who showed no verb impairment with a normal production of verbs both in picture naming and sentence completion tasks. Silveri and Di Betta (1997), and Silveri, Perri, and Cappa (2003) report patients with verb impairment who have lesions that lie outside of the left frontal regions. The findings of Aggujaro, Crepaldi, Pistarini, Taricco, and Luzzatti (2006) suggest that the left posterior temporal lobe and the inferior parietal lobule also play a crucial role in verb retrieval. Based on these exceptions, it can be concluded that the FTDH is not comprehensive enough to explain the neural distribution of nouns and verbs in the brain.

Nevertheless, independently of the anatomical-behavioral correlations made, all these results do indicate the existence of a double dissociation, which by these studies is assumed to be between nouns and verbs. This would argue for representations of nouns and verbs that are separate and functionally independent (Crepaldi, Berlingeri, Paulesu, & Luzzatti, 2011).

1.1.2 Brain imaging

The conflicting results from behavioral studies of brain damage and its relationship to noun-specific or verb-specific impairments initiated the discussion on the cortical areas involved in the processing of these two categories. It is questioned whether nouns and verbs are represented in different neural substrates or if both categories are processed in the same anatomical area, but with functional differences between them (Crepaldi et al., 2011). To provide an answer to this question, multiple neuroimaging studies with Positron Emission Tomography (PET) and functional magnetic resonance imaging (fMRI) have been performed. In addition, electrophysiological studies using electroencephalography (EEG), Magnetoencephalography (MEG), and Transcranial Magnetic Stimulation (TMS) have tried to provide a different approach to identifying the neural correlates for nouns and verbs.

(12)

4 Other studies showed deviating dissociations than those of the anatomical-behavioral correlative studies. Although Shapiro, Moo and Caramazza (2006) found selective activation in the left prefrontal cortex for verbs and in the left inferior temporal lobe for nouns with fMRI, they also found activation in the left superior parietal lobule for verbs. Dehaene (1995) found left temporal-parietal negativity for verbs in his EEG study. Similarly, the PET data of Warburton et al. (1996) showed that the left inferolateral temporal cortex and the posterior part of the inferior parietal lobe were activated particularly for the retrieval of verbs. Also, in an fMRI study Bedny and Thompson-Schill (2006) showed a role for the temporal lobe for verbs by finding greater activity in the left superior temporal gyrus for verbs than nouns.

There are also studies that report no dissociation between nouns and verbs at all. Tyler, Russell, Fadili, and Moss (2001) had healthy subjects perform a lexical decision task and a semantic categorization task, but found in both situations no differences in activation between the word classes. Vigliocco, Warren, Siri, Arcuili, Scott, and Wise (2006) reported, in contrast to studies described above, to have found no effects of grammatical class in the left inferior frontal gyrus. Also the MEG study by Sörös, Cornelissen, Laine, and Salmelin (2003) showed no different patterns of activation for nouns and verbs in healthy subjects.

1.1.3 Theories and explanations

Both neuroimaging and electrophysiological studies, as well as behavioral studies show many conflicting and inconsistent results. To explain the findings in these studies in a bigger framework, researchers have proposed several theories and explanations.

One explanation researchers have argued for is that the double dissociation provides evidence for the lexical forms (i.e., grammatical class), as an organizational principle for knowledge of language in the brain (e.g., Caramazza & Hillis, 1991; Hillis & Caramazza, 1995; Silveri & Di Betta, 1997) and that there are separate neural substrates for both grammatical classes (Daniele et al., 1994). This would mean that the difference between nouns and verbs is not only based on concrete nouns (objects) and concrete verbs (actions), but also on their abstract counterparts. Shapiro et al. (2005) supported the grammatical class hypothesis by showing patterns of left frontal and temporal cortical activation not only for verbs and nouns, respectively, but also for pseudowords used as nouns and verbs, suggesting that no semantics are involved in the dissociation between nouns and verbs. However, Collina, Marangolo, and Tabossi (2001) show results that conflict with the grammatical class hypothesis, by testing the production of argumental nouns and verbs in patients with Broca’s aphasia. In initial assessment tests, the patients showed a clear dissociation between nouns and verbs, but when their production of argumental nouns and verbs was assessed, no substantial difference was found.

A second theory revolves around a morphological deficit in either nouns or verbs (Shapiro, Shelton, & Caramazza, 2000; Shapiro & Caramazza, 2003). In this hypothesis the grammatical class deficit surfaces not as early as the lexical form, but only later, at the level of morphology. Shapiro et al. (2000) describe a fluent aphasic patient who is able to produce the third-person singular forms of verbs, but has difficulty producing the plural forms of phonologically identical nouns. This was the case not only for existing words, but also for pseudowords. A comparable result of a deficit for verbs but not nouns on homonyms for existing words and pseudowords was found in the study by Shapiro and Caramazza (2003). Tsapkini, Jarema, and Kehayia (2002) report data of a Greek non-fluent aphasic patient that suggest that morphological processes in relation to grammatical class play an important role in lexical processing, although the impairment of the patient was only observed in production and not in comprehension.

(13)

5 words referring to objects. They show that nouns referring to actions are more similar to verbs referring to actions than to nouns referring to objects, thereby providing evidence against the grammatical class hypothesis.

A fourth hypothesis put forward is that of syntactic damage (e.g., Friedmann & Grodzinsky, 1997; Friedmann, 2000, 2002, 2006). In the so called ‘tree pruning hypothesis’ (Friedmann & Grodzinsky, 1997) it is stated that there is damage on a level in the syntactic tree through which verbs cannot be moved up in the tree anymore and therefore cannot be inflected. This hypothesis is, however, not widely accepted due to the fact that it cannot account for the impairment of nouns and cannot explain why non-agrammatic patients may suffer from verb damage too (Crepaldi et al., 2006).

Although all of the above theories and explanations have been debated in the literature, none of them has been commonly accepted among researchers. Crepaldi et al. (2011) recently reviewed a large number of studies of impaired populations, anatomical-behavioral correlations, neuroimaging, and electrophysiological measurement on nouns and verbs. They concluded that neither different techniques of investigation nor the use of different cognitive tasks can fully explain the variability in the results so far. Regarding the neural substrates of nouns and verbs, Crepaldi et al. (2011) suggest that the trouble in finding neural correlates which are agreed upon is due to either no existence of spatial segregation or the lack of a detailed enough spatial resolution in the current techniques available. Another important notion of several researchers is that multiple factors and aspects could play a very influential role in the processing of nouns and verbs, like, for instance, psycholinguistic characteristics (e.g., Bird, Howard, & Franklin, 2003), argument structure (e.g., Jonkers & Bastiaanse, 1997) and more detailed semantic features (e.g., Fung et al., 2001).

1.1.4 Methodological concerns

Several methodological issues need to be taken into account in discussing the processing of nouns and verbs. For instance, many studies have based their conclusions on one task that tests one output modality. Some studies use multiple tasks and test multiple output modalities, although few studies have tested all output modalities. However, there are multiple examples in the literature in which a patient shows a selective impairment that is restricted to a certain modality. A substantial example of this is the description of two cases by Hillis, Wityk, Barker, and Caramazza (2002). One of these two patients had an oral deficit in the production of verbs, which was not present in writing, while the other showed the exact opposite pattern. In addition, both patients had spared oral and written naming of nouns. This shows that deficits can be modality specific, which means that an absent result in a patient or patient group does not also have to mean that the deficit does not surface when tested in another modality.

Another methodological concern is the task used to test the processing of word categories (Vigliocco, Vinson, Druks, Barber, & Cappa, 2011). Some studies revolve, for instance, around the retrieval of lexical information from memory, while others investigate lexical integration into sentences. Comparing the results of different kinds of tasks could have an influence on how the results are interpreted because of the different cognitive processes involved and linguistic levels addressed in each task.

Not only could the task influence the results, but also the types of materials that are used. Fung et al. (2001) argue that in picture-naming tasks, animations should be used when testing action words, rather than still pictures. In their study they report results which confirm that the apparent deficit in action naming found in earlier studies might be due to the nature of the stimuli used. Moreover, Rodríguez-Ferreiro, Davies, González-Nosti, Barbón, and Cuetos (2009) underline the importance of psycholinguistic characteristics of stimuli (e.g., frequency, age of acquisition, and imageability) that need to be controlled for when conducting a study.

(14)

6

frequency items. Furthermore, they argued that static or dynamic depictions of actions in a naming task cannot be responsible for a lower score on verbs because they were able to replicate their initial findings when videotaped stimuli were used instead of drawings. On the other hand, Druks and Shallice (2000) reported a patient who showed a great advantage when actions were acted out in front of him or when he carried them out himself instead of seeing them depicted as drawings. Luzzatti et al. (2002) showed that the selective impairment of patients on either nouns or verbs disappeared when the data were not controlled for word frequency and imageability.

The above mentioned methodological concerns are important to consider, but are, unfortunately, not applicable and/or reasonable in all cases to take into account. Moreover, the possible influence and extent of the influence of several variables, tasks and modalities on contradictory findings are still under discussion. However, as much as possible, the control of certain variables and a methodologically premeditated task in future studies would be helpful in determining the position of results that are up for discussion.

1.2 Semantic categories

Although the research on lexical retrieval has mainly focused on nouns versus verbs as general word class categories, the importance of more specific semantic categories has been taken into account in some studies as well. For several years now, the interest in category-specific deficits in impaired populations has grown. As suggested by the name, a category-specific deficit is characterized by a disproportionately worse performance on items of a specific semantic category compared to other semantic categories (of the same grammatical class).

Category-specific deficits have mostly been interpreted as reflecting a deficit in the semantic memory and the knowledge stored with it (Capitani, Laiacona, Mahon, & Caramazza, 2003). Semantic memory and episodic memory together form the declarative memory system, which exclusively stores explicit information (i.e., information that can be verbalized). In episodic memory personal events are saved, while semantic memory is meant for saving general information. Examples of information stored in semantic memory are facts, concepts, and the meanings of words. In contrast to episodic memory, which contains personal information, the information stored in semantic memory largely overlaps between individuals (Yi, Moore, & Grossman, 2007).

By performing neuropsychological studies in impaired populations, and neuroimaging and electrophysiological studies in healthy individuals, researchers have gathered important information about the structure of the semantic memory system. However, this information has not given us a clear-cut insight into the contents, organization and neuroanatomical bases of the semantic memory yet. In this section, an overview will be given of the category-specific deficits found in neurologically impaired populations and the anatomical-behavioral correlations that have been made between brain damage and category-specific deficits. Furthermore, the neuroimaging and electrophysiological studies that have considered semantic categories and the theories and explanations that have been developed on semantic memory will be described. Finally, some methodological concerns related to this topic will be discussed.

1.2.1 Behavioral studies

(15)

7

In the literature, several specific categories that can be impaired are described. The most common finding is a category-specific deficit for natural items (e.g., animals, fruits, and vegetables) compared to manufactured artifacts (e.g., vehicles, tools, and furniture). According to a review by Capitani et al. (2003) this pattern is seen in 75% of the described cases. For example, Sartori and Job (1988) described a brain-damaged patient with a selective impairment for the categories of animals, fruits, and vegetables, which surfaced in multiple modalities. Basso, Capitani, and Laiacona (1988) reported a patient with an auditory comprehension deficit that was the most severe for animals, fruits, and

vegetables. The study by Silveri and Gainotti (1988) described a patient with a partial recovery from

herpes simplex encephalitis, who was selectively impaired for living things and food items. Barbarotto, Capitani, Spinnler, and Trivelli (1995) reported a case study of a patient with semantic dementia with a lexical-semantic deficit for living things while knowledge about tools, vehicles, and

furniture was completely spared. Another case showing impairment for living things is described in

the study by Forde, Francis, Riddoch, Rumiati, and Humphreys (1997). The patient described was known to be impaired in naming living things compared to non-living things, and further testing showed that also the retrieval of perceptual attributes of living things was impaired. A sixth example is a case series described by Humpreys and Riddoch (2003), who report seven brain-damaged patients who all showed an apparent category-specific deficit affecting living things.

However, the opposite pattern, namely, greater impairment for manufactured objects than objects of the natural kind, is also reported in multiple cases, although it is comparatively rare. An example of this pattern is the study by Warrington and McCarthy (1983). They described the case of a global dysphasic patient who demonstrated selective preservation of foods, animals, and flowers compared to other objects. Another example is the study by Lambon Ralph, Howard, Nightingale and Ellis (1998), who reported two patients, one of which showed a deficit for living things and the other for non-living things. That both patterns occur is an important observation because it suggests that it is not the case that in general one semantic category is inherently more difficult than the other (Reilly, Rodriguez, Peelle, & Grossman, 2011).

Multiple studies have focused on more narrowly defined domains within the broad object categories of natural kinds and artifacts. Several subcategories have been distinguished, like animals,

fruits/vegetables, furniture, tools/instruments, and body parts. In the study by Hart and Gordon

(1992) a patient is described who was impaired in naming animals, but not in other living things and objects. Crutch and Warrington (2003) revealed a specific deficit, in a single patient, for fruits and

vegetables at the level of comprehension. The patient reported by Cappa, Frugoni, Pasquali, Perani,

and Zorat (1998) was severely impaired in furniture and tools. And, a specific deficit for body parts, while all other categories were comprehended well, was present in the case described by Suzuki, Yamadori and Fujii (1997).

(16)

8

Category-specific deficits surface in rather different kinds of brain damage, ranging from focal lesions to diffuse neural damage (for a detailed overview see Capitani et al., 2003). Multiple studies on category-specific deficits in impaired populations have linked the impaired object category to the brain damage of a patient. Gainotti (2000) states that the right hemisphere plays an important role in the processing of living things because a deficit to this category mostly involves bilateral damage to the anteromesial and inferior parts of the temporal lobes, while a deficit of manufactured artifacts mostly involves only left hemispheric dorsolateral damage. Barbarotto et al. (1995) described the category-specific deficit for living things and in addition, gave an analysis of the MRI scan, which showed severe atrophy of the temporal lobes, most severely on the right side, and which also included the basal neocortex, hippocampus and parahippocampal gyri. According to Reilly et al. (2011) the most commonly found lesion sites in the case of a deficit for manufactured artifacts are the left posterior middle temporal gyrus and the left ventral premotor cortex.

In general, these findings would support the suggestion made in the review by Gainotti, Silveri, Daniele, and Guistolisi (1995) that there is a correlation between deficits for living things and damage to the anteromedial parts of both temporal lobes, and between deficits for non-living things and damage involving the left frontal parietal lobe. However, other regions have also been implicated as being important in the processing of specific subcategories. Tranel, Damasio, and Damasio (1997) found that a deficit in the category animals was correlated with damage centered in the right mesial and inferior occipital-temporal region or in the left mesial occipital region. Furthermore, they found that a deficit in tools and utensils was correlated with damage centered in the left posterior temporal-occipital and parietal-temporal-occipital regions. The study by Capitani et al. (2009) also focuses on the subcategories within objects and puts certain anatomical-behavioral correlations forward. They found that a category-specific deficit in fruits and vegetables might be due to left fusiform gyrus damage and a deficit to animals might be due to damage in the anterior temporal lobe. Gainotti (2010) supports these latter suggestions through a review of single-case studies, although he suggests that the gender of the patient plays an important role in the neuroanatomical distribution of categories.

1.2.2 Brain imaging

Although anatomical-behavioral correlative studies in general suggest that damage in different neural sites causes different category-specific deficits, the exact correlation between specific brain damage and specific category remains unclear. To obtain more insight into the exact neural distribution of specific categories of nouns and verbs, several neuroimaging studies have been conducted on this subject.

Overall, there is very little consistency regarding anatomical specialization of categories across studies. In a short review of PET and fMRI studies on the anatomical basis of categories by Devlin et al. (2002) this becomes apparent in, for example, the association of the left lingual gyrus in some PET studies with natural kinds (e.g., Perani et al., 1995), while in other PET studies with artifacts (e.g., Moore & Price, 1999). In their own study, consisting of two PET experiments and one fMRI experiment, Devlin et al. (2002) found no consistent results regarding more narrow subcategories, but did identify different neural systems for natural kinds versus artifacts. Studies with fMRI have also suggested certain areas that would be activated in categories, for example the activation of the right inferior frontal, middle temporal, and fusiform gyrus for living objects (Leube, Erb, Grodd, Bartels, & Kircher, 2001).

(17)

9 in accordance with the statement by Van Schie, Wijers, Mars, Benjamins, and Stowe (2005) that neural activation for animals is typically present in occipital and fusiform areas, while activation for

tools is mostly present in motor function areas. Grabowski, Damasio, and Damasio (1998) found

activation for tools in the anterior bank of the precentral gyrus along the inferior and middle frontal gyri.

Studies with fMRI also found category-specific findings for more narrow categories, for example the left middle, right middle, and inferior frontal areas, plus the superior and middle temporal areas animals and furniture (Spitzer et al., 1998) and in the left ventromedial occipital cortex for animals (Grossman et al., 2002). In a recent study by Lin, Lu, Fang, Han, and Bi (2011) activation for animals was found in the posterior superior temporal sulcus and activation for tools in the posterior middle temporal gyrus.

Contrary to the statement by Devlin et al. (2002) about the absence of consistency across studies on the subcategories within objects, with the exception of tools, Martin (2007) concluded in his review that the literature is generally in agreement on the neural sites of animals, as well. Animals tend to activate two regions in the posterior temporal cortex, namely the lateral portion of the fusiform gyrus and the posterior superior temporal sulcus. Furthermore, evidence was reviewed that the amygdala also plays a role in the processing of animals. Regarding tools, Martin (2007) states that across studies the left ventral premotor cortex, intraparietal sulcus, and posterior middle temporal gyrus show activation for this subcategory.

Even though anatomical-behavioral correlative studies on categories within verbs are rather rare, there are several neuroimaging studies on types of verbs and their assumed neural distribution. Grossman et al. (2002) studied the activation patterns of verbs of motion and verbs of cognition. They found that cognition verbs showed activation of the left posterior lateral temporal cortex, while motion verbs had more activation in the caudate, left ventral temporal occipital cortex, and the bilateral prefrontal cortex. However, Wallentin et al. (2011) found a robust activation of motion verbs in the left posterior middle temporal gyrus. Lin et al. (2011) went into further detail with motion verbs and found that biological-motion verbs activate the posterior superior temporal sulcus, while mechanical-motion verbs show more activation in the posterior middle temporal gyrus. Earlier, a study of Bedny, Caramazza, Grossman, Pascual-Leone, and Saxe (2008) instead generalized the recruitment of the posterior lateral temporal cortex to be high not only for action verbs, but also for mental verbs. Kemmerer, Gonzalez Castillo, Talavage, Patterson, and Wiley (2008) studied five different semantic components of verbs, namely action, motion, contact, change of state, and tool

use. They confirmed that motion verbs activate the posterior lateral temporal cortex. Furthermore,

they found that the action component activates the primary motor and premotor cortices, the

contact component the intraparietal sulcus and inferior parietal lobule, the change of state

component the ventral temporal cortex, and the tool use component a distributed network of temporal, parietal, and frontal regions. Their finding regarding the action component is consistent with the finding by Hauk, Johnsrude, and Pulvermüller (2004) that action words attributed to arm, leg, or face actions activate areas in the motor cortex.

The studies described above give a broad overview of the neuroimaging studies done so far on the categories within nouns and verbs. Even though in general no consensus on the exact neural distribution of semantic subcategories has been reached, these studies taken together provide convincing evidence that there may be domain-specific lexical regions in the cortex. The object subcategories of animals and tools have been studied the most and are the ones most consistently linked to certain neural sites. Also, most of the verb subcategories seem to show some overlapping regions across several studies.

1.2.3 Theories and explanations

(18)

10 given of some well-known theories developed over the years and then further detail will be given on the more recently proposed theory of embodied cognition and its counterpart, the abstract model of representation.

One of the first theories put forward for the organization of semantic memory was the hierarchical network model by Collins and Quillian (1969). Following this model, information would be organized in categories that are logically related to each other on the basis of hierarchy. For instance, a canary is a bird, a bird is an animal, and an animal is a living creature. In 1975, the spreading activation theory of semantic processing was introduced, which states that when one word is activated, other related words will automatically be activated as well through a network of links (Collins & Loftus, 1975). Both of these theories have been of great influence on studies of semantic memory and form the basis for later proposed hypotheses.

Warrington and McCarthy (1987) proposed the sensory/functional hypothesis, which states that sensory (i.e., visual) features and non-sensory (i.e., functional) features have distinct semantic systems. Objects are assumed to differ in their ratio of sensory and functional features, with natural kinds having primarily sensory features that are represented in the temporal limbic lobe, and artifacts primarily having functional features that are represented in the frontal parietal cortex. Caramazza, Hillis, Rapp, and Romani (1990) proposed another theory, the organized unitary content hypothesis (OUCH). In this view, a semantic representation consists of conceptual features, and features that often occur are therefore strongly linked together and clustered together, while less often co-occurring features are more restricted in their strength of association. Later, Caramazza and Shelton (1998) proposed the domain-specific account, which in addition assumes that by natural selection different neural circuits have been developed to organize categories, for example animals being a potential predator and fruits and vegetables being food or medicine. Another hypothesis that has been considered is the artifact account (Funell & Sheridan, 1992). This account proposes that, in general, names of living things are more difficult to process than names of non-living things, based on features like frequency, imageability, visual complexity, and age of acquisition. Although the importance of taking psycholinguistic characteristics into account is widely acknowledged, the influence of these characteristics alone cannot fully explain the observed category-specific deficits. Therefore, this hypothesis is not sufficient to explain the performance (e.g., Caramazza & Shelton, 1998).

Humphreys and Forde (2001) propose the hierarchical interactive theory, which states that the problem lies at a visual processing level because living things have more similar perceptual structures than non-living things and therefore living things would be harder to process. However, Moss et al. (2002) think that category-specific deficits are at a cognitive level and reflect a problem with the interactive system of semantic features that are shared among concepts. They call this the conceptual structure account and take findings that show no anatomical segregation of categories as evidence for this theory. In 2004, Sartori and Lombardi (2004) introduced the semantic relevance theory, based on the assumption that there is a difference in processing demands between semantic categories. This difference is caused by the subset of semantic features that form the core meaning of a concept. The core meaning consists of features that make it possible to differentiate a concept from other similar concepts (Sartori & Lombardi, 2004). Zannino, Perri, Pasqualetti, Caltagirone, and Carlesimo (2006) built on the semantic relevance theory by proposing that living concepts have a higher degree of semantic similarity and therefore cause higher processing demands.

1.2.3.1 The embodied view of cognition

(19)

11 between and even within disciplines, the framework has several variations and claims concerning its scope.

In the context of language, the theory of embodied cognition is mainly interpreted as semantic knowledge being not purely amodal, but grounded in sensorimotor systems. For object knowledge, there is a strong belief that sensory and motor attributes and features acquired during experience shape our object concepts. Therefore, it has been suggested that objects are represented through distributed networks of sensory, motor and/or more abstract functional information (e.g., Warrington & McCarthy, 1987; Damasio, 1990; Farah & McClelland, 1991). However, even in linguistics several variations of the embodied view of cognition are present. The interpretation of the embodied view of cognition ranges from all to only some concepts being grounded in sensorimotor systems.

Multiple neuroimaging studies have provided evidence that supports the embodied view of cognition. Kemmerer et al. (2008) showed that semantic components of verbs are processed by different parts of the brain and that the components of action and tool use are partly grounded in sensorimotor systems. However, Hickok (2010) is not completely convinced by the conclusion made by Kemmerer et al. (2008) in the ongoing discussion about the somatotopic organization of action verbs. He postulates that it might not be the semantic components of verbs that activate the sensorimotor systems, but the association between the verb meanings and the actions they refer to. To test which of these two explanations would be applicable, Hickok proposes testing the prediction on neurologically impaired individuals. If individuals with impairment in a specific area are not able to properly perform categories associated with that area, this would support Kemmerer’s point.

Evidence for the embodied view of cognition has also been provided for nouns, mostly based on concrete noun categories, such as animals and tools. Sabsevitz, Medler, Seidenberg, and Binder (2005) found that concrete concepts of nouns activate perceptually based representations in the brain. Martin, Haxby, Lalonde, Wiggs, and Ungerleider (1995) showed that generating the name of an action that is typically associated with an object activated the posterior region of the left middle temporal gyrus. Martin (2001) reviewed over 20 findings showing greater activity in the left posterior middle temporal gyrus for generating action words and stated that these are consistent with the idea that the neural systems that are active during perception also store information about object-specific features. Martin and Chao (2001) stated that areas close to regions that mediate perception of object motion, i.e., the posterior region of the lateral temporal lobe, are activated during naming and identifying objects with motion-related attributes. In another review, Martin (2007) indicated that object properties are stored throughout the brain. The specific sensory and motor-based information of objects would be stored in their corresponding sensorimotor systems and this results in certain object-category-related neural circuits.

1.2.3.2 The abstract model of representation

There have also been studies supporting a ‘disembodied’ view of cognition, also called the abstract or symbolic model of representation. This theory states that conceptual representations are symbolic and abstract and therefore qualitatively distinct and separate from sensorimotor information. The ideas of Landauer and Dumas (1997) are an example of this view, proposing that the meaning of concepts is derived from the statistical co-occurrence of linguistic forms together. Another hypothesis of abstract representation is that meaning is drawn from syntactic relations among abstract language symbols (Kintsch, 2008). In this view, the meaning of symbols cannot be reduced to perception and action.

(20)

12 more abstract ideas. The notion of domains is at the basis of this hierarchy. According to this idea, each piece of information is sorted into a class or group (domain) and not by its relation to functionality or sensory perceptions. Buxbaum, Veramonti, and Schwartz (2000) documented that participants with apraxia showed a deficit in their manipulation knowledge, but that they still had functional knowledge about tools. In other words, there was a lack of sensorimotor knowledge, while conceptual knowledge was still present. Mahon and Caramazza (2009) interpret this finding as a dissociation between sensory, motor, and conceptual knowledge and present it as an argument against the strict embodied view of cognition. In their critical review Mahon and Caramazza (2008) argue that the evidence that is used in favor of the embodied view of cognition in earlier studies can also be explained by their proposal called ‘grounding by interaction.’ This account considers a middle course between the hypotheses of an embodied and disembodied view of cognition. In this view, concepts are abstract and symbolic, but they can be closely connected through experiences with perceptions and actions (Harrison & Trafton, 2010).

1.2.4 Methodological concerns

There are also some methodological concerns regarding research on semantic categories. In principle, the methodological issues that were raised concerning research on nouns and verbs are also applicable to the research on different semantic categories (see section 1.1.4). These are the consideration of different output modalities, the task that is used, the materials that are used, and the psycholinguistic features that should be taken into account. In addition, the different semantic categories raise one more important methodological concern. The division of concepts into categories makes these more sensitive to gender-specificity. For example, tools are often more familiar to men, while kitchen utensils are stereotypically more familiar to women (Barbarotto, Laiacona, Macchi, & Capitani, 2002). Laiacona, Barbarotto, and Capitani (2006) review several cases of category-specific deficits and find a strong interaction with gender. Therefore, they argue that materials used to detect differences between semantic categories should be carefully composed regarding gender specificity.

1.3 Concrete and abstract words

The research on nouns versus verbs and different semantic categories within nouns and verbs has mainly focused on concrete words, and actually little is known about the representation of abstract words in semantic memory. This could be considered odd since abstract words provide an extensive part of our vocabulary and are very frequent in our daily use of language. Although studies have been done on concrete versus abstract words, the discussion continues concerning what the cognitive representation is, which neural substrates are involved, and which of the theories is applicable.

In this section an overview will be given of the psycholinguistic features of frequency, concreteness, and imageability, the concreteness effect, the studies on impaired populations regarding concrete versus abstract words, and the anatomical-behavioral correlations that have been drawn from those. Furthermore, neuroimaging and electrophysiological studies, the theories and explanations that have been developed, and some methodological concerns that have come up regarding research on concrete versus abstract words will be reported.

1.3.1 Imageability, concreteness, and frequency

(21)

13 these were rated high on imageability, but low on concreteness. Apart from these exceptions, imageability and concreteness are usually highly correlated (Paivio et al., 1968), and therefore, from now on, high imageable words will be referred to as concrete and low imageable words as abstract. The influence of imageability is, among others, shown in the study by Paivio, Walsh, and Bons (1994), who found that words with a high imageability rating were remembered better than low-imageability rated words.

Frequency is a psycholinguistic lexical feature that indicates how often a word occurs in the language. This is standardly measured by counting the occurrence of a word in a structurally defined unit (e.g., occurrence per 10,000 words). Frequency can be defined for written language use or spoken language use. The influence of frequency is, among other things, reflected in the study by Forster and Chambers (1973), who showed that naming times for words with high frequency were shorter than for words with low frequency.

1.3.2 Concreteness effect

The concreteness of words is determined by the degree to which the meaning of the concept that the word refers to is tangible and sensorily perceptible. It is relatively easy to form a mental image of concrete words. On the contrary, abstract words do not refer to physical objects, but to ideas or moods, and are therefore difficult to form a mental image of (Jefferies, Patterson, Jones, & Lambon Ralph, 2009). Examples of concrete words are sweater and chair, while abstract words are things like

love and hope. Concrete and abstract words have different semantic natures, and this plays a role in

language production and language comprehension. It is frequently demonstrated in neurologically unimpaired individuals that concrete words are easier to process than abstract words in all linguistic aspects (e.g., James, 1975). Examples are slower processing of abstract than concrete words in word association (De Groot, 1989), lexical decision (Kroll & Merves, 1986), naming (Swanenflugel, Harnishfeger, & Stowe, 1988), and free recall (Paivio, 1986). This has not only been found in monolingual individuals, but also in multilingual language situations. For instance, concrete words are translated quicker and more accurately than abstract words (e.g., De Groot, 1992). This phenomenon of better concrete than abstract word processing is called the concreteness effect and is one of the most robust phenomena in psycholinguistic research (Reilly, Grossman, & McCawley, 2006).

1.3.3 Behavioral studies

Not only is the concreteness effect highly consistent in healthy participants, but it also has often been observed in brain-damaged individuals. However, there are also some cases in which the concreteness effect is reversed. In these cases patients have, in general, overlapping lesions, which could give us a clue about the anatomical-behavioral correlation between concrete versus abstract words and their neuroanatomical distribution. First, the cases of impaired populations and their deficits on either concrete or abstract words will be reported, and then the focus will be on the anatomical-behavioral correlations that have been proposed.

(22)

14 (2005) showed that their patient with global aphasia made more errors on abstract than concrete words.

The unusual pattern of concrete words being more impaired than abstract words has also been reported multiple times. This reversed concreteness effect has mostly been reported in cases of semantic dementia (e.g., Breedin, Saffran, & Coslett, 1994; Papagno, Capasso, Zerboni, & Miceli, 2007; Yi et al., 2007; Bonner et al., 2009). The effect was found for both abstract and concrete nouns (e.g., Reilly et al., 2006) as well as for abstract and concrete verbs (Reilly, Cross, Troiani, & Grossman, 2007). Besides semantic dementia, the reversed concreteness effect has also been documented in cases of herpes simplex encephalitis (Sirigu, Duhamel, & Poncet, 1991; Mattioli, 2008).

Although Grossman and Ash (2004) have proposed that a reversal of the concreteness effect is a characteristic feature of semantic dementia, the phenomenon has not always been found in this population. For example, Jefferies et al. (2009) found no reversed concreteness effect in the eleven patients they examined, with all of them having higher success rates on more imageable words. One explanation for these contradictory results could be the stage of the semantic dementia the patient has progressed to (Reilly et al., 2006).

The double dissociation in the performance on concrete and abstract words of brain-damaged individuals suggests that the neuroanatomical organization of concrete and abstract words is at least partially distinct. Especially the reversed concreteness effect in semantic dementia has led to suggestions of neural areas that are responsible for concrete words. A greater impairment for concrete over abstract words has been documented in patients with lesions in the anterior and inferior portions of the temporal lobes (e.g., Macoir, 2009; Papagno, Capasso, & Miceli, 2009). Also, the additional findings of a reversed concreteness effect in patients with herpes simplex encephalitis strengthen this assumption because the similarity between semantic dementia and herpes simplex encephalitis is a deficit in the anterior temporal area. Papagno et al. (2007) present a case that showed a reversed concreteness effect with selective atrophy of the left inferiolateral temporal cortex and Yi et al. (2007) associate the atrophy of the left inferior temporal cortex of their patient with the relative difficulty shown in concrete words. Loiselle et al. (2012) compared two groups with different brain damage, namely patients with a selective unilateral anterior temporal resection and patients with a selective unilateral amygdalohippocampectomy. They found that both groups were impaired on concrete and abstract nouns, but that the former were significantly more impaired on concrete words than abstract words, while there was no difference between concrete and abstract words in the latter group. Therefore, they conclude that the anterior temporal lobe may play a critical role in the processing of concrete concepts.

The hypothesis by Reilly et al. (2006) that the presence or absence of a reversed concreteness effect in semantic dementia is dependent on the stage of the disease is based on the amount of atrophy seen in the demented patients. Reilly et al. (2006) studied three cases of semantic dementia, which were all in different stages of the disease. They hypothesized that in an early stage of semantic dementia, a reversed concreteness effect is shown due to focal lesion of the left inferior temporal lobe. In a later stage, this lesion would expand to both temporal lobes during which both concrete and abstract words would be impaired. The results support their hypothesis, with the least impaired patient showing a reversed concreteness effect, while there was no difference between concrete and abstract words in the two more impaired patients. The longitudinal study by Macoir (2009) supports this hypothesis by initially finding a reversed concreteness effect in a patient with semantic dementia that disappeared over the course of the disease.

1.3.4 Brain imaging

(23)

15 been performed to discover the neural correlates of concrete and abstract words. For instance, Sabsevitz et al. (2005) performed an fMRI study and found greater activation for concrete than abstract words in a bilateral network, including the ventral and medial temporal, posterior inferior parietal, dorsal prefrontal, and posterior cingulate cortex. Abstract words, however, showed more activity almost exclusively in the left superior temporal and inferior frontal cortex. This was also concluded in the short review by Bedny and Thompson-Schill (2006) on this topic, with most studies reporting that left inferior temporal and parietal cortices are more activated with concrete words, while the left lateral temporal and inferior frontal cortices show higher activation with abstract words. Papagno, Fogliata, Catricala, and Miniussi (2009) concluded in their TMS study that abstract words are stored in the left posterior temporal superior gyrus and possibly in the left frontal inferior gyrus, while concrete words are also partly stored in right temporal cortex. Although a number of studies generally overlap in their results on the neural distribution of concrete and abstract words, the exact detailed correlates still differ among many studies.

Apart from the general overlap in results on the neural distribution of concrete and abstract words, there are also deviant results found. Contradictory to the studies described above, Giesbrecht, Camblin, and Swaab (2004) report more activity in the left inferior frontal and middle temporal gyri for concrete over abstract words. Whatmough, Verret, Fung, and Chertkow (2004) found in their PET study greater activation for concrete words in the left medial fusiform gyrus, while the right medial fusiform gyrus produced more activation for abstract words. This shows that not only do the exact, detailed correlates of concrete and abstract words need to be figured out more precisely, but that there are also indications that other areas could be involved in concrete and abstract words too.

1.3.5 Theories and explanations

Whereas the concreteness effect is a robust and widely acknowledged phenomenon, the source of the effect is still a matter of discussion. Two competing theories that have been proposed to explain the concreteness effect have been of great influence. There is, on the one hand, a single-code model, called the context availability theory (Schwanenflugel & Shoben, 1983), and on the other hand the dual-coding theory (Paivio, 1971). According to the context availability theory, all concepts are represented in an amodal semantic system in which concrete nouns are processed more efficiently than abstract nouns. This would be because they have stronger and more extensive links to contextual information, which makes recognition of concrete words easier. The consequence of this theory is that the advantage of concrete words compared to abstract words should disappear when abstract words are presented in a context with sufficient supporting verbal information. This prediction was confirmed in the study of Schwanenflugel and Stowe (1989). On the other hand, the dual-coding theory proposes that concrete concepts are bilaterally encoded into memory in both verbal and visuoperceptual representations, while abstract nouns are only encoded in the form of verbal representations in the left hemisphere. This would make it easier to recognize and recall concrete words than abstract words. Support is found in studies that show that individuals who use imaginable mediators to remember words recall them better than individuals who use verbal mediators (Paivio, Yuille, & Smythe, 1966; Paivio, Smythe, & Yuille, 1968).

(24)

16

1.3.6 Methodological concerns

One of the most undervalued methodological points of interest in the research on concrete and abstract words is grammatical class. Grammatical class is generally overlooked since most of the studies focus on nouns. Nevertheless, concreteness effects are frequently reported as if they apply to words in general. Bedny and Thompson-Schill (2006) rightly notice that there are almost no studies that control for and/or manipulate both imageability and grammatical class in the same experiment. Bird et al. (2003) showed that the dissociation between nouns and verbs dissolves when the feature of imageability is taken into account, thereby providing strong argumentation for the need to control both imageability and grammatical class in the methodology of an experiment.

In addition, in the case of concrete and abstract words it is also important to take into account the effect of different tasks, different modalities, and psycholinguistic features of stimuli like frequency and imageability. The influence of imageability is inherent in studies on concrete and abstract words. Most studies assume low imageability is equal to abstractness and high imageability is equal to concreteness. However, the measure of imageability is a continuum. Therefore, it is worth considering taking the aspect of different sorts of abstractness into account (Crutch & Warrington, 2005).

1.4 Alzheimer’s disease

Over the past decades, an increase in research on language and aging is noticeable, as a result of the growing population of elderly people. Research in this discipline has practical applications because knowledge of normal language behavior in the elderly is essential for the diagnosis and therapy of neurologically impaired individuals. In addition, the theoretical application is of importance as well. Knowledge of language in aging, including similarities and differences with language use in younger speakers, adds to our knowledge about the human linguistic capacity and its interaction with other cognitive abilities (Obler & Albert, 1980).

It is well-known that certain aspects of language, i.e., lexical retrieval (naming) and sentence processing (comprehension), decline with age (Obler & Albert, 1984). What is less known is if and how this influences neural representations. For numerical cognition, Ansari and Dhital (2006) show in their fMRI study that even the most basic aspects are subject to age-related changes in neural representations. This gives reason to believe that in linguistic processing the role of age should be considered as well in neuroimaging studies and that it could play an important factor in the organization of certain neural correlates.

A consequence of the growth of the older population is an automatic increase in age-related neurodegenerative diseases like dementia (Kinsella & Phillips, 2005). An interesting issue regarding aging and language in general is the question to what extent symptoms of language decline are part of the normal aging process, and when features are great enough to be seen as symptoms of cognitive impairment, like dementia (e.g., Hyltenstam & Obler, 1989; Druks, Masterson, Kopelman, Clare, Rose, & Rai, 2006).

This section will focus on one of the syndromes of dementia, namely Alzheimer’s disease. First, general information of this disease population including its occurrence, symptoms and diagnosis will be described in this section. Then further detail on the course of language in this clinical picture will be included and an overview of the studies on nouns and verbs, semantic categories, and concrete and abstract words in Alzheimer’s disease will be given.

Concrete & Abstract Words in Alzheimer’s Disease: The influence of categories on processing