Elsevier

Neuropsychologia

Volume 44, Issue 7, 2006, Pages 1017-1028
Neuropsychologia

The role of semantic distance in category-specific impairments for living things: Evidence from a case of semantic dementia

https://doi.org/10.1016/j.neuropsychologia.2005.11.006Get rights and content

Abstract

In this paper, we describe a patient (LI) suffering from semantic dementia who showed a category-specific naming impairment for living things over and above the effects of several nonsemantic confounding variables. We investigated the characteristics of LI's impairment to address the following three issues raised in three different accounts of category-specific impairments: (i) the role of an imbalance in the loss of sensory compared to nonsensory features (assumed by the Sensory Functional Theory [Warrington, E. K., & Shallice, T. (1984). Category-specific semantic impairments. Brain, 107, 829–859]); (ii) the role of cross domain differences in Feature Correlation (assumed by the Conceptual Structure Account [Moss, H., Tyler, L. K., & Devlin, J. T. (2002). The emergence of category-specific deficits in a distributed semantic system. In: E. M. E. Forde & G. W. Humphreys (Eds.), Category Specificity in Brain and Mind (pp. 115–147). New York: Psychology Press]); (iii) the role of semantic distance (proposed by Cree and McRae [Cree, G. S., & McRae, K. (2003). Analyzing the factors underlying the structure and computation of the meaning of chipmunk, cherry, chisel, cheese, and cello (and many other such concrete nouns). Journal of Experimental Psychology: General, 132, 163–201]). We found that semantic distance was the only factor causally linked to LI's poorer performance on living things. In fact, her naming performance was less accurate on items that had many semantic neighbours, which is typical of living things. On the contrary, a feature listing task revealed that the features available to LI were not predicted by their level of correlation, as expected by the Conceptual Structure Account. Finally, at variance with the Sensory Functional Theory, although LI quoted sensory features less accurately than nonsensory ones, this did not give rise to a disproportionate loss of semantic features in the living domain.

Introduction

In the last 30 years, the literature has reported many cases of brain damaged patients with language impairments characterised by category-specific deficits (see Capitani, Laiacona, Mahon, & Caramazza, 2003 and Gainotti, 2000, for reviews). These patients performed disproportionately worse, on lexical-semantic tasks, with items belonging to particular semantic categories. In most cases, the functional locus of the impairment was at the level of semantic memory. Gainotti (2000) (see also Capitani et al., 2003) believes that this was true for three-quarters of the approximately sixty patients with category-specific impairments reported in the literature. The most commonly observed pattern of category-specific dissociation broadly follows the division between living things and man-made artefacts. In most individuals, categories within the living things domain, e.g., fruits, vegetables and animals, are more impaired than artefacts, e.g., vehicles, tools and furniture (Barbarotto, Capitani, Spinnler, & Trivelli, 1995; Caramazza & Shelton, 1998); however, in a few cases the reverse pattern is observed—poorer performances on artefacts than on living things (Sacchett & Humphreys, 1992; Warrington & McCarthy, 1987).

Studying individuals with category-specific deficits provides us with a unique opportunity to understand how semantic memory is organised. However, this is true only if the effect is genuine and not an artefact due to some extra semantic confounding variables. According to the so-called Artefact hypothesis, variables such as familiarity, word frequency, and visual complexity may be responsible for lower scores on lexical semantic tasks involving living items, because living things are on average less familiar, more visually complex, and designated by less frequent words (Funnell & Sheridan, 1992; Stewart, Parkin, & Hunkin, 1992). On the other hand, variables such as Age of Acquisition and imageability may act in the opposite direction, favouring living things (Lambon Ralph, Hovard, Nightingale, & Ellis, 1998; Silveri, Cappa, Mariotti, & Puopolo, 2002). Although the need to take confounding variables into account is now widely acknowledged, many authors claim that in most cases the Artefact hypothesis alone is not sufficient to explain the observed category-specific impairment (Caramazza & Shelton, 1998; Sartori, Job, & Zago, 2002).

The most straightforward explanation of the category effect is that it arises as a consequence of being semantic memory directly organised into categories. According to the Domain-specific hypothesis (Caramazza & Shelton, 1998; Santos & Caramazza, 2002), specialised neurological subsystems may have developed under evolutionary pressure to categorise efficiently items referring to animals, plant life, and artefacts. Since these substrates are thought to be localised in discrete brain regions, a category-specific impairment for either living things or artefacts may arise as the consequence of a focal brain damage selectively involving the corresponding substrate.

Warrington and co-workers (Warrington and McCarthy, 1983, Warrington and McCarthy, 1987; Warrington & Shallice, 1984) challenged the assumption of a truly categorical organisation of semantic memory and proposed a different explanation of category-specific deficits. The Sensory Functional Theory (SFT) claims that the two major semantic domains (living things and artefacts) differ with regard to the type of knowledge that distinguishes one exemplar from another within the same semantic category. To identify living things, we rely mostly on sensory information (e.g., it is critical to know that zebras 〈have stripes〉 to distinguish them from horses). On the contrary, to operate correctly within category distinctions in the artefacts domain, functional/associative knowledge is necessary (e.g., the most important difference between a vase and a bottle is not what the two items look like but what they are designed to hold). Since sensory and functional/associative knowledge are likely both subserved by discrete neurological substrates, this account explains both patterns of category-specific impairments, because it assumes the presence of selective brain damage involving only one of the two critical substrates.

The SFT receives some empirical support from single case studies that show an association between disproportionate loss of sensory knowledge and a category-specific effect that penalises living things (De Renzi & Lucchelli, 1994; Sartori & Job, 1988; Silveri & Gainotti, 1988). However, other findings undermine the explanatory power of this account. In particular, two lines of research that investigate issues relevant to the present study report data at variance with the SFT. First, recent feature listing studies provide little support for the assumption that living things rely more heavily on sensory information than nonliving things. Garrard, Lambon Ralph, Hodges, & Patterson (2001) demonstrated that among the more distinctive features (i.e., the features shared by few category members, which are thought to be critical for distinguishing some concepts from others) the amount of sensory and nonsensory features does not significantly differ across domains. In a recent study (Zannino, Perri, Pasqualetti, Caltagirone, & Carlesimo, in press), we found a similar pattern. In the same study, we found a significantly larger mean number of sensory features for living concepts (when distinctive and shared features were collapsed), because the living concepts’ representations had more features on average than the nonliving ones; however, the percentage of sensory features across living and nonliving representations was virtually the same. Lambon Ralph and co-workers proposed the second line of research, which undermines the assumption of a causal link between a category-specific impairment for living things and a disproportionate loss of sensory features. These authors carried out a number of studies on patients suffering from semantic dementia (SD) (Garrard, Lambon Ralph, & Hodges 2002; Lambon Ralph, Hovard, et al., 1998; Lambon Ralph, Patterson, Garrard, & Hodges, 2003), a degenerative disease that selectively affects semantic memory and spares other language processing levels (such as phonology and syntax) and nonverbal cognitive skills (Hodges, Patterson, Oxbury, & Funnell, 1992; Snowden, Goulding, & Neary, 1989). In fact, they demonstrated that a disproportionate loss of sensory information is a hallmark of SD, while, at variance with the expectations of the SFT, category-specific impairments are only exceptionally observed in this subject population.

Our data also address another hypothesis—the so-called Conceptual Structure Account (CSA; Moss, Tyler, & Devlin, 2002). According to this hypothesis, “systematic differences in the internal structure of concepts in different categories and domains” (p. 117) play a major role in the genesis of the category effect. These differences are due to the typical values taken on by the Feature Distinctiveness and Feature Correlation variables according to the different categories and domains. More distinctive features take part in the semantic representation of only one or a few members of a semantic category (e.g., 〈has a trunk〉, which is true only for elephant, is a very distinctive feature in the animals category); on the opposite end of this continuum, we find features shared by all or by most category members (e.g., 〈has legs〉 or 〈breathes〉 for animals). As for Feature Correlation, two features are said to be correlated if the presence of a pair member in the semantic representation of a given concept significantly increases the probability that the other pair member will occur in the same representation (e.g., if the semantic representation of concept X contains the property 〈can fly〉 it will probably also contain the property 〈has wings〉). In this model distinctiveness is a measure of the usefulness of a feature for distinguishing among category members. As such, it is also a measure of its usefulness in performing many lexical-semantic tasks requiring the operation of such distinctions (Tyler, Moss, Durrant-Peatfield, & Levy, 2000). For example, if we are requested to find the picture of an elephant among the pictures of various animals, it is more important to know that an elephant 〈has a trunk〉 than to know that an elephant 〈has legs〉. Correlation, on the contrary, is not related to feature helpfulness but to feature resilience against neurological damage, because more correlated features are thought to be more resistant than less correlated ones (Devlin, Gonnerman, Andersen, & Seidenberg, 1998; but see Caramazza & Mahon, 2003, for counter arguments).

A main assumption of the CSA is that “random, global damage throughout the system will affect concepts in different ways, as a function of their internal structure” (Moss et al., 2002, p. 117). Correlated features are thought to be more resilient against neurological damage so that in cross domains “different patterns of correlation among properties within a concept will lead to different patterns of loss and preservation of information, given the same degree of overall damage” (Moss et al., 2002, p. 118). For example, in the case of relative impairment for living things, concepts referring to the impaired domain should show a disproportionate loss of distinctive features (the critical ones for distinguishing between category members) because they are less correlated in this domain than in the artefacts domain1 (Moss, Tyler, Durrant-Peatfield, & Bunn, 1998).

Recently, Cree and McRae (2003) identified the semantic distance between category members as another factor able to account for category specificity. A number of feature listing studies, including ours (Cree & McRae, 2003; Garrard et al., 2001; McRae, de Sa, & Seidenberg, 1997; Zannino et al., in press), found that living things concepts on average share more features with other category members than nonliving concepts. In other words, semantic representations in the living domain show more overlapping. Cree and McRae (2003) believe that lower semantic distance between living things could be one of several structural differences supporting the emergence of category-specific impairments (penalising this domain) in brain damaged people. We also believe that this may be the reason for poorer performances on living things, because concepts that are semantically very similar tend to be particularly confounding. A major verifiable prediction of this account is that a living things impairment may occur without a disproportionate loss of semantic knowledge in the affected domain provided there is a cross domain imbalance for semantic distance in the testing material. This expectation is at variance with all of the above-reviewed accounts (SFT, Domain-specific hypothesis and CSA), which hold that patients perform worse in the category that has suffered the greatest loss of semantic information, for whatever reason. Therefore, the CSA posits that semantic knowledge is more severely hit in the affected category due to a different degradation rate; the Domain-specific hypothesis that the affected semantic category suffers a major reduction of knowledge because of the loss of the neural substrate subserving category-specific semantic knowledge; finally, the SFT posits that an asymmetry in the availability of semantic information across domains is expected because of the loss of the neural substrate subserving a feature type, which is more represented in the affected domain.

Recently, we (Zannino et al., in press) made a feature listing study of 64 concepts (32 living—16 fruits and 16 mammals and 32 nonliving—16 vehicles and 16 furniture and household items) to obtain normative data on a number of indices for items to be used in lexical-semantic tasks. The collected database allowed us to quantify the following indices: (i) the number of features listed by the subjects for each concept included in the corpus (e.g., donkey), according to different types of knowledge (sensory, nonsensory, superordinate); (ii) the distinctiveness of each feature, i.e., the number of concepts within a given semantic category (fruits, mammals, vehicles, furniture and household items) sharing that feature; e.g., 〈has engine〉 was listed for 9 out of the 16 vehicles in our corpus; (iii) the correlation level of each feature according to the particular concept in which it occurs (for further details on this measure, see Experiment 3); and finally, (iv) the semantic distance between concepts. Details about the computation of this index are reported in Experiments 1b and 2.

The present paper describes a detailed investigation of semantic knowledge in a patient (LI) suffering from SD who showed a category-specific naming deficit for living things that was still reliable even after the role of several confounding variables was taken into account. The major issues addressed by this study were the following: First, we attempted to evaluate if LI's category effect was accompanied by a disproportionate loss of semantic information in the affected domain, as predicted by all of the above-reviewed accounts (Domain-specific hypothesis, SFT and CSA), except for the one based on the role of semantic distance. Second, we wanted to ascertain if LI's category effect was accompanied by a disproportionate loss of sensory information, and if a causal link could be established between category-specific and modality-specific deficits, as held by the SFT. Third, we wanted to evaluate if (as held by the CSA) the Feature Correlation factor was able to predict the availability of semantic information in LI. Finally, we wanted to investigate the role of an imbalance of semantic distance across domains in the emergence of LI's poorer performance on living things. With this aim, we submitted LI to a battery of semantic tests devised using the items from our feature listing study.

Section snippets

Case report

LI is a right-handed woman who was born in 1940 and has a degree in literature; she was referred to our neuropsychology service in October 2001. At that time LI reported a 2-year history of word-finding difficulties and reading problems; her spontaneous speech was fluent with no evidence of phonemic paraphasias or deficits at the grammatical or syntactic level. Her nonverbal abilities also seemed fully spared upon clinical observation. Periodically, LI still undergoes neuropsychological

Material and methods

LI was asked to perform the picture-naming task specifically developed by Laiacona et al. (1993) to partial out the role of confounding variables in the genesis of category effects. The task is comprised of 60 items selected from Snodgrass and Vanderwart's set (1980). Six semantic categories are included: three in the living domain (vegetables, animals and fruits) and three in the nonliving domain (tools, vehicles and furniture). The following indices are available for each item:

  • Frequency of

Discussion

In this paper we describe the semantic memory status of a patient who met Hodges et al.'s (1992) neuropsychological criteria for the diagnosis of SD. LI suffered from a selective deficit of semantic memory, while her other verbal and nonverbal skills appeared to be fully preserved. LI's semantic deficit was present when she was tested using visual stimuli and verbal tasks that did not rely on picture processing. Her deficit was accompanied by a profound naming deficit resulting in semantic

References (56)

  • U. Bortolini et al.

    Lessico di frequenza della lingua italiana contemporanea

    (1972)
  • R. Capasso et al.

    Esame neuropsicologico per l’afasia E.N.P.A.

    (2001)
  • E. Capitani et al.

    What are the facts of semantic category-specific deficits? A critical review of the clinical evidence

    Cognitive Neuropsychology

    (2003)
  • A. Caramazza et al.

    Domain-specific knowledge systems in the brain: The animate-inanimate distinction.

    Journal of Cognitive Neuroscience

    (1998)
  • G.A. Carlesimo et al.

    Standardizzazione di due test di memoria per uso clinico: Breve Racconto e Figura di Rey

    Nuova Rivista di Neurologia

    (2002)
  • G.A. Carlesimo et al.

    The mental deterioration battery: Normative data, diagnostic reliability and qualitative analyses of cognitive impairment. The Group for the Standardization of the Mental Deterioration Battery

    European Neurology

    (1996)
  • J.R. Crawford et al.

    Payne and Jones revisited: Estimating the abnormality of test score differences using a modified paired samples t-test

    Journal of Clinical and Experimental Neuropsychology

    (1998)
  • G.S. Cree et al.

    Analyzing the factors underlying the structure and computation of the meaning of chipmunk, cherry, chisel, cheese, and cello (and many other such concrete nouns)

    Journal of Experimental Psychology: General

    (2003)
  • R. Dell’Acqua et al.

    Naming times and standardized norms for the Italian PD/DPSS set of 266 pictures: Direct comparisons with American, English, French, and Spanish published databases.

    Behavior Research Methods, Instruments, & Computers

    (2000)
  • J.T. Devlin et al.

    Category-specific semantic deficits in focal and widespread brain damage: A computational account

    Journal of Cognitive Neuroscience

    (1998)
  • E. Funnell et al.

    Categories of knowledge: Unfamiliar aspects of living and nonliving things

    Cognitive Neuropsychology

    (1992)
  • P. Garrard et al.

    Semantic dementia: Implications for the neural basis of language and meaning

    Aphasiology

    (1999)
  • P. Garrard et al.

    Semantic dementia: Clinical, radiological and pathological perspectives

    Journal of Neurology

    (2000)
  • P. Garrard et al.

    Semantic dementia: A category-specific paradox

  • P. Garrard et al.

    Prototypicality, distinctiveness, and intercorrelation: Analyses of the semantic attributes of living and nonliving concepts

    Cognitive Neuropsychology

    (2001)
  • J.R. Hodges et al.

    Semantic dementia: Progressive fluent aphasia with temporal lobe atrophy

    Brain

    (1992)
  • M. Laiacona et al.

    Dissociazioni semantiche intercategoriali: Descrizione di una batteria standardizzata e dati normativi

    Archivio di Psicologia, Neurologia e Psichiatria

    (1993)
  • M.A. Lambon Ralph et al.

    Naming in semantic dementia—What matters?

    Neuropsychologia

    (1998)
  • Cited by (29)

    • Neural pattern similarity underlies the mnemonic advantages for living words

      2016, Cortex
      Citation Excerpt :

      For example, nonliving things showed greater within-category (structural) similarity than did living things when pictures were used (Gale & Laws, 2006; Laws & Gale, 2002; Laws & Neve, 1999; Marques, Raposo, & Almeida, 2013) or when particular subcategories of living (i.e., body parts) and nonliving things (i.e., musical instruments) were used (Laws, Gale, Frank, & Davey, 2002). Nevertheless, these studies converged with the other studies to show that higher similarity (either semantic or structural similarity) is associated with more processing errors (i.e., naming) but better memory (Laws & Gale, 2002; Zannino et al., 2006). These results provide further evidence to support the global matching hypothesis, which would predict that the overlap of representations during encoding would lead to better memory.

    • Association Splitting: A randomized controlled trial of a new method to reduce craving among inpatients with alcohol dependence

      2016, Psychiatry Research
      Citation Excerpt :

      Once a node in a semantic network is activated, associations with other nodes are (automatically) activated and these associations are strengthened with increasing co-occurrence of cognitions. Significant evidence exists for these semantic networks among individuals with semantic aphasia (e.g., Lambon et al., 2007; Zannino et al., 2006). Imaging studies have also demonstrated that semantically-related words activate neighboring cortical regions (e.g., Sachs et al., 2008).

    • The contribution of neurodegenerative diseases to the modelling of semantic memory: A new proposal and a review of the literature

      2015, Neuropsychologia
      Citation Excerpt :

      As for group performance, as a rule SD patients do not exhibit any category effect (Carroll and Garrard, 2005; Lambon Ralph et al., 2003), but occasionally group performances can show a marginal nonliving things advantage (Libon et al., 2013). On the other hand, single patients have been reported with a living things disadvantage that either persisted after controlling for several confounding variables (Barbarotto et al., 1995; Lambon Ralph et al., 2003) or disappeared after taking some of these factors into account (Parkin, 1993; Zannino et al., 2006c). In a few single cases (Lambon Ralph et al., 1998; Silveri et al. 1997), the reverse pattern, i.e. better performance on living- than nonliving things, has been reported.

    • Abstract and concrete categories? Evidences from neurodegenerative diseases

      2014, Neuropsychologia
      Citation Excerpt :

      Within the concrete domain, the living–non living dissociation in AD is generally characterized by a better performance on non living items (Chertkow and Bub, 1990; Garrard et al., 1998; Gonnerman et al., 1997; Silveri et al., 1991; Zannino et al., 2002), although there are also studies reporting the opposite dissociation (Gonnerman et al., 1997) or no significant differences (Tippett et al., 1996). Only a limited number of case studies showed a dissociation between living and non-living in sv-PPA (i.e., Barbarotto et al., 1995; Lambon Ralph et al., 2003; Lambon Ralph et al., 1998; Zannino et al., 2006a), as sv-PPA patients generally show a poor performance for both categories. These studies, however, are difficult to compare, due to differences in the characteristics of stimuli and in the tasks.

    View all citing articles on Scopus

    This study was carried out at the IRCCS S. Lucia, Roma, Italy.

    View full text