Investigating the influence of semantic factors on word retrieval: Reservations, results and recommendations

ABSTRACT

There is consensus that word retrieval starts with activation of semantic representations. However, in adults without language impairment, relatively little attention has been paid to the effects of the semantic attributes of to-be-retrieved words. This paper, therefore, addresses the question of which item-inherent semantic factors influence word retrieval. Specifically, it reviews the literature on a selection of these factors: imageability, concreteness, number of semantic features, typicality, intercorrelational density, featural distinctiveness, concept distinctiveness, animacy, semantic neighbourhood density, semantic similarity, operativity, valence, and arousal. It highlights several methodological challenges in this field, and has a focus on the insights from studies with people with aphasia where the effects of these variables are more prevalent. The paper concludes that further research simultaneously examining the effects of different semantic factors that are likely to affect lexical co-activation, and the interaction of these variables, would be fruitful, as would suitably scaled computational modelling of these effects in unimpaired language processing and in language impairment. Such research would enable the refinement of theories of semantic processing and word production, and potentially have implications for diagnosis and treatment of semantic and lexical impairments.

KEYWORDS:

• Semantic processing
• lexical access
• picture naming
• aphasia

Acknowledgements

We would like to thank all those people with aphasia who have so willingly supported our research over the years, and all our colleagues that have inspired and supported us through ongoing discussion and debate. This paper began its life as an Academy of Aphasia Keynote presentation in Hong Kong in 2019 by the first author. This keynote was sponsored by Cognitive Neuropsychology. We would like to thank Professor Brenda Rapp for inviting this paper, and for her careful and interesting comments during the review process. We also thank the three anonymous reviewers for their helpful comments. The manuscript has been very much improved as a result of this rigorous peer review process.

Disclosure statement

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

Notes

1 The decision not to provide a formal systematic review was motivated by several factors. First, that there were other papers (cited in the sections on individual variables) that already provided relatively systematic reviews of the literature on each variable. Second that systematic reviews may lead to the tendency for potentially important differences in methodology across studies to be obscured, and this is particularly problematic if a metanalysis is performed. Finally, a systematic review of all the relevant variables would have been too large in scope for a single article, not to mention the difficulty finding the appropriate articles with the relevant search terms without risking becoming either over, or under inclusive (e.g., ‘semantic’ and ‘naming’ as search terms in Scopus results in @14,000 hits).

2 We do not mean to imply that it is not important to understand how tasks work, merely to highlight that the simpler the task, the less research effort needs to be expended in working out the processes involved in that task, and the more research effort that can be spent using the task to develop theory.

3 Although this task could, accurately, be described as continuous picture naming, this term is often associated with the cumulative semantic interference literature where the number of previously presented items from a target’s category is manipulated.

4 Nozari and Pinet (2020) in their review of co-activation of representations describe this as the use of ‘indices’ to examine the effects of this co-activation.

5 The effects were stronger on latencies in the speeded naming task than in the standard naming task for name agreement (facilitatory in both tasks), age of acquisition (inhibitory only in speeded naming), image agreement (facilitatory in both tasks), frequency (facilitatory only in speeded naming), familiarity (facilitatory in both tasks), trial order (facilitatory only in speeded naming), and distinctiveness (inhibitory in both tasks). Although five more semantic variables (number of semantic features, intercorrelational density, number of near semantic neighbours, semantic similarity, typicality) did not show a difference between tasks, they also did not show significant simple effects on either task.

6 Perret and Bonin (2019) do also locate conceptual familiarity at this level of processing. Note too that in their analysis they combine effects of imageability and image variability as a single variable (see section on imageability effects for further discussion).

7 Much work on semantic representations and semantic processing conducted outside the area of word production (e.g., Cree et al., 1999; Farah & McClelland, 1991; Masson, 1995; Plaut & Shallice, 1993) has not (yet) been incorporated in theories of word production, perhaps because most experimental findings in word production research can be explained in the context of highly simplified semantic representations (Vinson et al., 2013).

8 Note that Rogers et al. (2015) argue that in Semantic Dementia the cooccurrence of features provides a typicality advantage, due to the hypothesised loss of distinguishing features. In contrast, in ‘semantic aphasia’ typical items create more competition leading to poorer performance as greater ‘semantic control’ is required to distinguish between these items and ‘semantic control’ is hypothesised to be impaired in this population.

9 A related yet different measure is intercorrelational strength (e.g., McRae et al., 1997, 1999), which captures the degree to which a specific feature (e.g., has fur) is correlated with the other features of a concept (e.g., ‘cat’). Intercorrelational strength determines the level of activation of that particular feature but also of the other features of the concept (e.g., has four legs, has whiskers, has a tail, etc.) as features in a cluster of intercorrelated features boost each other’s activity. This can affect processing, with more strongly correlated features speeding up activation and thus decreasing processing times. For example, strength of the correlation between a feature and a concept has been shown to predict response times in feature verification tasks where participants were asked to verify features as true or false of a concept (McRae et al., 1997, 1999). For instance, is hunted is more strongly correlated with the other features of ‘deer’ than of ‘duck’ and was therefore verified more quickly than for the concept ‘deer’ than the concept ‘duck’ (for similar findings see also e.g., Garrard et al., 2005; Randall et al., 2004; Taylor et al., 2004).

10 Humphreys et al. (1988) and Riddoch and Humphreys (1987) did investigate a related concept—the extent to which a category was rated as having shared features between category members (structural similarity of items within a category). Looking at the same single case, both papers reported greater naming accuracy for items from categories rated as having fewer shared features (i.e., higher distinctiveness). However, effects of distinctiveness were not assessed at the item level (only the category level).

11 There was also no interaction between conceptual processing (Pyramids and Palm Trees score) and effect of feature-based neighbourhood in the main analysis.

12 This is also broadly related to the literature on visual ‘affordances’ (shapes allow for certain types of manipulations and actions independent of specific knowledge of their identity, Cosentino, 2019; Gibson, 1979) which have been suggested to be intact even in cases of severe conceptual impairment in Semantic Dementia (Hodges et al., 2000).

13 We report imageabiity and concreteness together here, as they are so highly intercorrelated and few studies have attempted to dissociate the two.

Additional information

Funding

Solene Hameau was supported by an Australian Research Council Discovery Project grant (DP190101490) and Catherine Mason by a Macquarie University Research Excellence Scholarship (MQRES; scholarship number: 20191031).