Skip to content
This repository contains all experimental data, including every respondent's survey, the final data set in Excel or CSV format, and the analysis code in R.
Jupyter Notebook R
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
Code with output.docx
Norms excerpt.png
Sound symbolism, Dutch words.png
Unimodal, bimodal, trimodal.png
dashboard gif.gif
file1, Object Stimulus validation.xlsx
file2, Object Validation of stimulus.xlsx
file3, Object Validation Stimulus.xlsx
file4, Property Stimulus validation.xlsx
file5, Property Validation of stimulus.xlsx
file6, Property Validation Stimulus.xlsx
markdown launcher.R
sound symbolism regression and barplots.xlsx

Modality exclusivity norms for 747 Dutch properties and concepts in Dutch: a replication of English

This repository contains all experimental data, including every respondent's survey, the final data set in Excel or CSV format, and the analysis code in R (norms.R).

The norms, which are ratings of linguistic stimuli, served a twofold purpose: first, the creation of linguistic stimuli (see also Speed & Majid, 2017), and second, a conceptual replication of Lynott and Connell's (2009, 2013) analyses. In the collection of the ratings, forty-two respondents completed surveys for the properties or the concepts separately. Each word was rated by eight participants on average (see data set), with a minimum of five (e.g., for bevriezend) and a maximum of ten ratings per word (e.g., for donzig). The instructions to participants were similar to those used by Lynott and Connell (2009, 2013), except that we elicited three modalities (auditory, haptic, visual) instead of five.

'This is a stimulus validation for a future experiment. The task is to rate how much you experience everyday' [properties/concepts] 'using three different perceptual senses: feeling by touch, hearing and seeing. Please rate every word on each of the three senses, from 0 (not experienced at all with that sense) to 5 (experienced greatly with that sense). If you do not know the meaning of a word, leave it blank.'

These norms were validated in an experiment showing that shifts across trials with different dominant modalities incurred semantic processing costs (Bernabeu, Willems, & Louwerse, 2017). All data for that study are available, including a dashboard (in case of downtime of the dashboard site, please see this alternative).

The properties and the concepts were analysed separately. Properties were more strongly perceptual than concepts. Distinct relationships also emerged among the modalities, with the visual and haptic modalities being closely related, and the auditory modality being relatively independent (cf. Lynott & Connell's data for English. This ties in with findings that, in conceptual processing, modalities can be collated based on language statistics (Louwerse & Connell, 2011).

The norms also served to investigate sound symbolism, which is the relation between the form of words and their meaning. The form of words rests on their sound more than on their visual or tactile properties (at least in spoken language). Therefore, auditory ratings should more reliably predict the lexical properties of words (length, frequency, distinctiveness) than haptic or visual ratings would. Lynott and Connell's (2013) findings were replicated, as auditory ratings were either the best predictor of lexical properties, or yielded an effect that was opposite in polarity to the effects of haptic and visual ratings.

The present analyses and further ones will be reported in a forthcoming paper. All data and analysis code for the norms are available for re-use under a CC-BY licence, provided acknowledgment of the following publication:

Bernabeu, P., Willems, R. M., & Louwerse, M. M. (2017). Modality switch effects emerge early and increase throughout conceptual processing: Evidence from ERPs. In G. Gunzelmann, A. Howes, T. Tenbrink, & E. J. Davelaar (Eds.), Proceedings of the 39th Annual Conference of the Cognitive Science Society (pp. 1629-1634). Austin, TX: Cognitive Science Society.


This study is a cross-linguistic, conceptual replication of Lynott and Connell’s (2009, 2013) modality exclusivity norms. Their English properties and concepts were translated into Dutch, then independently tested as follows. Forty-two respondents rated the auditory, haptic, and visual strength of those words. Mean scores were then computed, with a high interrater reliability and interitem consistency. Based on the three modalities, each word also features a specific modality exclusivity, and a dominant modality. The norms also include external measures of word frequency, length, distinctiveness, age of acquisition, and known percentage. Starting with the results, unimodal, bimodal, and tri-modal words appear. Visual and haptic experience are quite related, leaving a more independent auditory experience. These different relations are important because they may correlate with different levels of detail in word comprehension (Louwerse & Connell, 2011). Auditory and visual words tend towards unimodality, whereas haptic words tend towards multimodality. Likewise, properties are more unimodal than concepts. Last, the 'sound symbolism' hypothesis was tested by means of a regression: Auditory strength predicts lexical properties of the words (frequency, distinctiveness...) better than the other modalities do, or else with a different polarity.

Further links:

Dashboard presenting the data and analyses (in case of downtime, please see alternative site)

Summarised report

The relation among auditory, haptic, and visual modalities in property and concept words, in English and Dutch.

Definitions (as in Lynott & Connell, 2009, 2013)

  • Dominant modality: Highest-rated modality;

  • Modality exclusivity: Range of the three modality ratings divided by the sum;

  • Perceptual strength: Highest rating across modalities.

Corpus measures used

  • Concreteness and age of acquisition: Brysbaert et al. (2014);

  • Phonological and orthographic neighbours: Marian et al.'s (2012) DutchPOND;

  • Word frequency and contextual diversity: Keuleers et al.'s (2010) SUBTLEX-NL;

  • Lemma frequency: Baayen et al.'s (1993) CELEX.


Baayen, R. H., Piepenbrock, R., & van Rijn, H. (1993). The CELEX Lexical Database [CD-ROM]. Philadelphia: Linguistic Data Consortium, University of Pennsylvania.

Field, A. P., Miles, J., & Field, Z. (2012). Discovering Statistics Using R. London, UK: Sage.

Keuleers, E., Brysbaert, M. & New, B. (2010). SUBTLEX-NL: A new frequency measure for Dutch words based on film subtitles. Behavior Research Methods, 42, 3, 643-650.

Louwerse, M., & Connell, L. (2011). A taste of words: linguistic context and perceptual simulation predict the modality of words. Cognitive Science, 35, 2, 381-98.

Lynott, D., & Connell, L. (2009). Modality exclusivity norms for 423 object concepts. Behavior Research Methods, 41, 2, 558-564.

Lynott, D., & Connell, L. (2013). Modality exclusivity norms for 400 nouns: The relationship between perceptual experience and surface word form. Behavior Research Methods, 45, 2, 516-526.

Marian, V., Bartolotti, J., Chabal, S., & Shook, A. (2012). CLEARPOND: Cross-Linguistic Easy-Access Resource for Phonological and Orthographic Neighborhood Densities. PLoS ONE, 7, 8: e43230.

Speed, L. J., & Majid, A. (2017). Dutch modality exclusivity norms: Simulating perceptual modality in space. Behavior Research Methods, 49, 6, 2204-2218.


Pablo Bernabeu. Email:


You can’t perform that action at this time.