Category-specific verb-semantic deficits in Alzheimer’s disease: Evidence from static and dynamic action naming

ABSTRACT

We investigated the representation and breakdown of verb knowledge employing different syntactic and semantic classes of verbs in a group of individuals with probable Alzheimer’s Disease (pAD). In an action naming task with coloured photographs (Fiez & Tranel, 1997. Standardized stimuli and procedures for investigating the retrieval of lexical and conceptual knowledge for action. Memory and Cognition, 25(4), 543–569. https://doi.org/10.3758/BF03201129), pAD individuals were impaired for naming actions compared to objects. Verb tense was also affected, with simple-past (e.g., chopped) being more difficult to name than the gerundial form (e.g., chopping). Employing action-naming with short movies depicting events and states, we contrasted three verb classes based on their hypothetical structural and semantic/conceptual properties: argument structure, thematic structure, and conceptual templates. The three classes were: causatives (peel), verbs of perception (hear), and verbs of motion (run) Overall, results suggest that individuals with pAD are selectively impaired for verb tense and thematic assignment, but not conceptual-template complexity. Methodologically, we also show that dynamic scenes are more ecologically valid than static scenes to probe verb knowledge in AD.

KEYWORDS:

• Verb-semantic deficits
• Alzheimer’s disease
• category-specific deficits
• action naming
• dynamic scenes
• verb meaning
• thematic roles
• semantic templates
• thematic hierarchy
• concepts

Acknowledgements

We are thankful to two anonymous reviewers and especially to Brad Mahon for insightful comments to a previous version of this paper. Our colleague George Schwartz passed away before the completion of the present article. We dedicate it to his fond memory.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Although the literature on category-specific deficits usually takes the term “semantic” to refer to categories of knowledge affected in semantic memory, “semantic” may be used refer to properties of verb meanings that are “linguistically active”, to use Grimshaw’s (2005) expression – i.e., those that affect linguistic structure. The term “conceptual” or “concepts” is usually taken to refer to the units of higher cognitive representation and processing, independent of their linguistic implementation. We will make this distinction, where context makes it necessary. But, for the most part, we will use “semantic” and “conceptual” interchangeably, with both referring to word meanings as well as to the units of semantic memory.

2 Some authors (e.g., Levin & Rappaport Hovav, 2005) take these predicates to specify subevents in the whole event denoted by the verb and its carrier sentence. For instance, (1) implies two subevents – that the agent x did something, and that the object y changed its state. We will avoid a detailed discussion of this theory, but address the nature of predicate decomposition in the General Discussion.

3 It is possible that this hierarchy implies a form of redundant representation: that, say, GO is represented redundantly at possibly all three levels – it is a feature, in Kim and Thompson’s (2004) account. Thus, losing GO means losing its associated superordinate verbs (go, perhaps), but not other levels of representation which incorporate or inherit the feature/predicate GO (e.g., run). This “redundancy” hypothesis, however, raises numerous questions about the nature of the representation of concepts that is postulated and how concepts might be affected selectively. It implies, for instance, that GO is affected at the top level but continues to contribute content to lower-level verbs. If that is not the case, it remains to be explained how a concept can have a feature such as GO without other concepts also incorporating its content – including the meaning of the verb go itself.

4 This analysis is restricted to this example. Although Grossman and White-Devine (1998) used 60 causatives, it is not clear how many of those are morphological causatives (that is, overtly marked by morphemes such as -en and -ify) rather than lexical causatives.

5 Henceforth, we will refer to this class as perception, though recognizing that many verbs within this class can be taken as agentive and that the psychological class can sometimes be analyzed as causative (e.g., Pesetsky, 1995). In the present investigation, only Subject-Experiencer verbs were employed (see Manouilidou & de Almeida, 2009, for discussion).

6 Notice that the causative verbs we used can also be inchoatives (one argument), thus they could, in principle, be easier than perception/psychological verbs. However, our stimuli and task are setup in a way that requires an agentive interpretation of causatives, thus yielding a transitive interpretation.

7 We recognize that the use of the NINCDS-ADRDA is a potential limitation of our study, given newer diagnosis criteria (Dubois et al., 2007; McKhann et al., 2011). However, we should note that our pAD sample was selected based on a battery of clinical and neurocognitive assessments, in addition to the NINCDS-ADRDA.

8 It should be noted that, as shown by several studies (see Capitani et al., 2003, for review), musical instruments usually pattern with living things, and body parts, with nonliving things. Although it is not the main scope of the present study to investigate subcategories of living and nonliving categories, our results show a similar pattern, with body parts being significantly better than other living subcategories, and musical instruments being significantly worse than other nonliving subcategories. Removing these items did not change the nature of the dissociation we obtained.

Additional information

Funding

This work was supported in part by a grant from the Social Sciences and Humanities Research Council of Canada (SSHRC) to R. G. de Almeida, by a SSHRC Major Collaborative Research Initiative grant to G Libben, G. Jarema, E. Kehayia, B. Derwing, L. Buchanan, and R. G. de Almeida (co-investigators), by a SSHRC fellowship to C. Antal, and by funds from the Centre for Interdisciplinary Research in Rehabilitation to E. Kehayia.