The present thesis
Chapter 2 describes Study 1, a multi-lab study from the Psychological Science Accelerator to which I contributed. This study revisited the object orientation effect (S.-C. Chen et al., 2018), which taps into sensorimotor simulation in conceptual processing. As I review in Chapter 2, replicating the object orientation effect is important due to its seminal influence and to the existence of mixed findings. Thanks to a worldwide network of laboratories, this study boasted a sample of over 3,000 participants. Furthermore, the study examined the role of individual differences in mental rotation as well as the role of language of testing. I contributed theoretical and methodological suggestions to this study from an early stage, recorded a demonstration video (https://osf.io/h36wr), contributed data collected from 50 participants at Lancaster University, and then contributed to the reporting. From now, I look forward to contributing to the latest revisions of this Stage II Registered Report, before the submission to Psychonomic Bulletin & Review, where this registered report was accepted in principle.
Study 1 did not replicate the object orientation effect. Furthermore, it revealed that neither mental rotation nor language interacted with the effect. Since the statistical power reached in this study is larger than that of most previous studies addressing the object orientation effect, it is especially important to reflect on the non-replication. It will be especially important to investigate the methodological characteristics that influence the robustness of effects. That is, in addition to the real effect size of each effect (which can only be estimated in research), and in addition to the statistical power of studies (which is highly dependent on the sample size), what does the operationalisation bear on the results? One possible influence is the categorical or continuous nature of independent variables.
Chapter 3 describes Study 2, which is formed of three studies (2.1, 2.2 and 2.3). These studies are thus grouped because they address the same overarching question (see Wingfield & Connell, 2022b for a similar format). This question is how the interplay between language and sensorimotor information manifests itself at the levels of individuals, words and tasks. Furthermore, the study addresses the importance of sample size by estimating how many participants would be necessary to reliably detect several effects of interest. Whereas in Study 1 the task used was specific to sensorimotor simulation, the tasks used in Study 2 are semantic priming, semantic decision and lexical decision, which allow the analysis of both language-based and vision-based information. It was important for us to incorporate language because research has suggested that language is complementary to sensorimotor information, and that it has a greater influence than the latter. At the level of individual differences, Study 2 examined vocabulary size and included general cognition covariates which are important to control for. In addition, the influence of participants’ gender was considered, as some research has suggested that language-based information could be more influential in females than in males. At the word level, text-based variables were used as indices of language-based information, whereas vision-based variables indexed sensorimotor information. As covariates, lexical information and word concreteness were included in the analyses. The covariates were included to attain a higher degree of certainty regarding the effects of interest, as the former and the latter were often correlated. Furthermore, these analyses included random effects, which were important to preserve the assumption of independence of observations, due to the presence of repeated measures.
The results revealed a prominent influence of language-based information and a smaller effect of vision-based information. Furthermore, the results revealed a set of interactions suggestive of a ‘relevance advantage’ in higher-vocabulary participants. That is, these participants were more sensitive to task-relevant features of the words, such as language-based information in lexical decision and word concreteness in semantic decision. Last, statistical power analyses revealed the need to dramatically increase sample sizes for studying vision-based information, as well as for studying the interaction of both language-based and vision-based information with vocabulary size, gender and presentation speed.
References
Thesis: https://doi.org/10.17635/lancaster/thesis/1795.
Online book created using the R package bookdown.