Frequency in lexical processing

ABSTRACT

Background: Frequency of occurrence is a strong predictor of lexical processing across modalities and experimental paradigms. However, frequency is part of a large set of collinear predictors including not only frequencies collected from different registers, but also a wide range of other lexical properties such as length, neighbourhood density, measures of valence, arousal, and dominance, semantic diversity, dispersion, age of acquisition, and measures grounded in discrimination learning.

Aims: The aim of this study is to provide a critical examination of these variables, the sources on which they are based, the way they are calculated and evaluated, and their potential causal relations.

Main Contribution: We show that age of acquisition ratings and subtitle frequencies constitute (reconstructed) genres that favour frequent use for very different subsets of words. As a consequence of the very different ways in which collinear variables profile as a function of genre, the fit between these variables and measures of lexical processing depends on both genre and task. A graphical model suggests that neither frequency nor age of acquisition are primary causal factors, but rather semantic and emotion measures as well as measures derived from discriminative learning.

Conclusions: The methodological implication of these results is that when evaluating effects of lexical predictors on processing it is advisable to carefully consider what genres were used to obtain these predictors, and to consider the system of predictors and potential conditional independencies using graphical modelling.

KEYWORDS:

• Lexical processing
• frequency
• age of acquisition
• naive discrimination learning
• activation
• prior availability
• subtitles

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. A further disadvantage of this strategy is that it comes with the danger of circularity: frequency counts collected to predict lexical processing are themselves based on decisions about data preprocessing that are informed by how well candidate counts predict lexical processing.

2. We backed off from zero by adding 1 to the corpus frequency before taking the (natural) logarithm.

3. A thin plate regression spline approximates a wiggly curve as a weighted sum of mathematically regular curves (named basis functions), with a penalty on wiggliness. The estimation algorithms make sure a good balance is found between fidelity to the data and model simplicity.

4. Data sets and analyses reported in this manuscript are available in the Mind Research Repository at http:////openscience.uni-leipzig.de/index.php/mr2.

5. A tensor product smooth uses higher-dimensional simple basis functions (restricted cubic splines) to set up a mesh for a wiggly surface. Again, penalties on wiggliness guarantee an optimal balance between over- and undersmoothing.

6. For the RTs, we decomposed the interaction of frequency and PC1 into two main effects and the remaining joint effect. For the ratings, a similar decomposition resulted in an inferior fit, which is why a non-decompositional tensor product smooth is reported for this response variable.

7. For details on a model for auditory comprehension, see Baayen et al. (2015).

Additional information

Funding

This work was supported by the Alexander von Humboldt Foundation [grant number 1141527].