Reconciling phonological neighborhood effects in speech production through single trial analysis

Abstract

A crucial step for understanding how lexical knowledge is represented is to describe the relative similarity of lexical items, and how it influences language processing. Previous studies of the effects of form similarity on word production have reported conflicting results, notably within and across languages. The aim of the present study was to clarify this empirical issue to provide specific constraints for theoretical models of language production. We investigated the role of phonological neighborhood density in a large-scale picture naming experiment using finegrained statistical models. The results showed that increasing phonological neighborhood density has a detrimental effect on naming latencies, and re-analyses of independently obtained data sets provide supplementary evidence for this effect. Finally, we reviewed a large body of evidence concerning phonological neighborhood density effects in word production, and discussed the occurrence of facilitatory and inhibitory effects in accuracy measures. The overall pattern shows that phonological neighborhood generates two opposite forces, one facilitatory and one inhibitory. In cases where speech production is disrupted (e.g. certain aphasic symptoms), the facilitatory component may emerge, but inhibitory processes dominate in efficient naming by healthy speakers. These findings are difficult to accommodate in terms of monitoring processes, but can be explained within interactive activation accounts combining phonological facilitation and lexical competition. 2Reconciling phonological neighborhood effects in speech production through single trial analysis Native speakers of a language know a myriad of different words. This so-called mental lexicon is often described as an interconnected network in which representation distance may depend on meaning or form similarities among the words. One crucial step for understanding this network is to describe which kind of similarity influences language processing and how it modulates performance. In this context, the current research is concerned with the role of phonological similarity in word retrieval and speech production. This issue has been addressed in earlier theoretical work (e.g. Chen & Mirman, 2012;Dell & Gordon, 2003), but the empirical evidence on which this work is grounded remains controversial. Our goal in this article is to clarify the empirical facts regarding form similarity effects in speech production, and to integrate them within a single account. Clearing the empirical constraints will allow further refinement of theoretical models to advance our understanding of the cognitive processes underlying speech production.An approximation of how similar or interconnected a word is within the lexical network can be obtained by computing its phonological neighborhood density (PhND). PhND refers to the number of words that can be formed from a given word by substituting, adding or deleting one phoneme (Luce, 1986). For example, the word "bat" sounds similar to many other words (e.g., "cat", "fat",...

Activation cascading in sign production.

“…Iconicity also makes the composition of activated signs different in signing and speaking, where we found no effects of iconicity (see also Baus & Costa, in press). Research on speaking has made clear that the composition of co-activated words is a major factor in determining lexical selection -both correct and erroneous (Gordon, 2002;Sadat et al, 2014;Ziegler, Muneaux, & Grainger, 2003). Having established that iconicity determines which signs are activated, aspects of lexical selection affected by coactivation can be better characterized in sign processing.…”

“…Their activation arises from activation spreading to related concepts within the semantic systems and subsequently cascading to word phonology. Other co-activated words are phonologically related -e.g., sinister or mister for the target word sister -as a result of activation spreading across words part of the same 'phonological neighborhood' (Sadat, Martin, Costa, & Alario, 2014;Vitevitch, 2002). Were there cascading of activation in sign production, co-activated signs would not be similar in terms of sounds, sharing instead features of hand movement and configuration.…”

Typing pictures: Linguistic processing cascades into finger movements

“…Given the high level of correlation among some of the variables considered as predictors, the first step was to restrict the number of predictors using a principled approach. To this end, we used random forest analyses, an analytical tool that provides a measure of variable importance for each predictor, using a data-driven approach based on a collection of classification trees (e.g., Breiman, 2001; for applications in psycholinguistics, see Tagliamonte & Baayen, 2012;Sadat, Martin, Costa, & Alario, 2014). In a second step we used mixed effects regression models to estimate the effect of the selected predictors on the dependent variables (response latency and mean interkeystroke intervals).…”

“…Following the rationale outlined in the introduction and previous work in the field of word production in both speech (e.g., Alario et al, 2004;Barry et al, 1997;Bates et al, 2003;Sadat et al, 2014) and writing (Bonin, Chalard, Méot, & Fayol, 2002;Bonin, Méot, Lagarrigue, & Roux, 2015), we started our analysis with 11 potential predictor variables, encompassing lexical-semantic, lexical-orthographic/phonological and sublexical effects. Each variable is described below.…”

“…In cases in which two predictors were very strongly correlated (r > .8), we retained only the variables emerging as more important in accounting for the results using datadriven methods for assessing variable importance based on random forest analysis (Hothorn, Buehlmann, Dudoit, Molinaro, & Van Der Laan, 2006;Strobl, Boulesteix, Kneib, Augustin, & Zeileis, 2008;Strobl, Boulesteix, Zeileis, & Hothorn, 2007). This method has been indeed suggested as a black-box method to identify, amongst a larger set of variables, a smaller sample of potentially relevant predictors (Strobl, Malley, & Tutz, 2009), to be later tested with classic regression approach (Sadat et al, 2014). Random forest analysis provides a trial and error method for establishing whether a given variable is a useful predictor (Tagliamonte & Baayen, 2012).…”