How to constrain and maintain a lexicon for the treatment of progressive semantic naming deficits: Principles of item selection for formal semantic therapy

Abstract

The progressive degradation of semantic memory is a common feature of many forms of dementia, including Alzheimer's disease and the semantic variant of primary progressive aphasia (svPPA). One of the most functionally debilitating effects of this semantic impairment is the inability to name common people and objects (i.e., anomia). Clinical management of a progressive, semantically based anomia presents extraordinary challenges for neurorehabilitation. Techniques such as errorless learning and spaced-retrieval training show promise for retraining forgotten words. However, we lack complementary detail about what to train (i.e., item selection) and how to flexibly adapt the training to a declining cognitive system. This position paper weighs the relative merits of several treatment rationales (e.g., restore vs. compensate) and advocates for maintenance of known words over reacquisition of forgotten knowledge in the context of semantic treatment paradigms. I propose a system for generating an item pool and outline a set of core principles for training and sustaining a micro-lexicon consisting of approximately 100 words. These principles are informed by lessons learned over the course of a Phase I treatment study targeting language maintenance over a 5-year span in Alzheimer's disease and SvPPA. Finally, I propose a semantic training approach that capitalises on lexical frequency and repeated training on conceptual structure to offset the loss of key vocabulary as disease severity worsens.

Keywords:

• Dementia
• Primary progressive aphasia
• Language treatment
• Language disorder
• Anomia
• Item selection

Notes

¹We enrolled all patients prior to more contemporary diagnostic criteria for AD (McKhann et al., 2011) and primary progressive aphasia (Gorno-Tempini et al., 2011) were published.

²One might look to an extensive literature on category-specific language impairments to glean insight into global word classes that are differentially vulnerable to the effects of semantic impairment. For example, dissociations of natural kinds vs manufactured artifacts (Gonnerman, Andersen, Devlin, Kempler, & Seidenberg, 1997), nouns vs. verbs (Hillis, Tuffiash, & Caramazza, 2002) and abstract vs. concrete words (Breedin, Saffran, & Coslett, 1994) have all been reported. Most of these dissociations remain controversial, and individual differences add another source of variability. Although global trends might inform treatment (e.g., verbs and abstract words are often difficult for patients), explicit prediction of the sequence of forgetting individual words remains impossible.

³The “dyadic” component involving direct interaction with another sentient being may be especially critical. The data in support of transfer benefits from computer-based brain training paradigms remain limited (Owen et al., 2010).

⁴Of note, we did not conduct a parallel longitudinal study using random picture presentation. Thus, this semantic clustering hypothesis remains tentative and is an open empirical question.

⁵Item lists are freely available for download and use at www.reilly-coglab.com/train-lex.