Welcome to the Noun Associations for German database

Alissa Melinger and Andrea Weber

Saarland University

Copyright © 2006 Melinger & Weber

Go directly to search

I. The database

With this website we make a large database of free associations for German nouns available to interested researchers and scholars. 409 German nouns referring to picturable objects were chosen as target words. Target words represented a variety of semantic classes including animals, plants, professions, food, furniture, vehicles, and tools. Native speakers of German, all students at Saarland University, were asked to write the first three words (continued free association) that came to their mind when presented with a stimulus word. A critical component of the current study was the presentation of stimuli in two forms: lexical stimuli consisted of the printed name of target objects; pictorial stimuli consisted of the printed names accompanied by black and white line drawings of the referred-to objects (see Figures 1a and b). Drawings were based on Snodgrass and Vandervart (1980) and the picture database from the Max Planck Institute for Psycholinguistics in the Netherlands. We started collecting these norms because we were interested in differences in associate responses when stimulus words were presented in printed form compared to when the printed form was accompanied by a pictorial representation; for instance, images might increase the salience of physical attributes of objects, or when word forms have different shades of meaning, responses evoked by lexical stimuli might index any of the words´ meanings while responses evoked by pictorial representations might be more biased towards the depicted sense. But of course, the NAG database will be of interest to any researcher who is in more general terms interested in German semantic associations (for a restricted set of German associations see Schlaghecken & Bölte, 1998).





Figure 1: Example stimulus in lexical (a) and pictorial (b) presentation mode.

The 409 target stimuli were divided randomly into three separate questionnaires consisting of approximately 136 nouns each. Two presentation modes were prepared for the questionnaires; one with printed words only and one in which the printed words were accompanied by pictures. Each version of the questionnaire was filled out by 50 participants, resulting in a maximum of 300 data points for any given target stimulus (50 participants X 2 presentation modes X 3 responses). Target stimuli and collected associate responses were entered into a database with the following additional information:

Target stimuli: Associate responses: Participants:

II. How to use the database

If you want to start a search in the NAG database, you are first asked to select the information you want to keep in your search results. Whatever box you tick will be information that is displayed in the output; you can further constrain your search by choosing from the drop down menus or filling in the search boxes (As the default, all items of the drop down menus are selected; you can select specific items by mouseclicking them or return to selecting all items by clicking on "select all" under the box). For example, if you type Apfel "apple" in the stimulus search box, you will be shown all 300 responses that were given to the stimulus Apfel; if you neither constrain your stimulus or response search box, all responses to all stimuli will be shown. You can also use * as a wildcard for your search. The default settings for the output presentation contain the following information: stimulus name, response name, response part of speech, relation type between response and stimulus. If you would additionally like to see, for example whether a response was given as an answer to a lexical or pictoral stimulus, you have to tick the presentation box. If you are only interested in the first response provided to a stimulus, then you have to tick the rank box and select rank 1.

The default output format will display every answer in a separate row. If you would like identical responses to be collapsed in a single row with a count of occurrences, you can tick the collapsing box further down the page. Note that all displayed cells in a row must be identical for collapsing to take place. For example, if the response tree was given once to the lexical stimulus apple and once to the pictorial stimulus apple, this information will not be collapsed if you have chosen to display presentation mode; it will be collapsed, however, if presentation mode is not displayed.

Collapsing gives you the option either to treat all responses with the same weight, despite their rank, or to choose weights. For example, you could choose to weigh a first response as three times more important than a third response. If you want to treat all responses with the same weight, you have to give the same value to all three ranks. Collapsing also introduces two new search criteria: token count is the number of responses that have been collapsed; weighted count is the token count multiplied by the weights that you had assigned.

After you have selected the information you want to keep in your search results, you hit continue which will bring you to the output table. On this page, you have the possibility to refine your search. All the search criteria you had selected are again available at the top of the page. You can sort your search results on any one column within the displayed table. If you want to sort on more than one column, please go to advanced sorting (right top corner). If you want to save your search results, the tables can be downloaded as unsorted CSV tables. Links to the line drawings we used as stimuli, are included in the results table.

III. Some notes on data input

How to cite these associate norms in your references:

Melinger, A. & Weber, A. (2006). Database of Noun Associations for German. http://www.coli.uni-saarland.de/projects/nag/.

Other references:

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

Melinger, A., Schulte im Walde, S., & Weber, A. (2006). Characterizing response types and revealing noun ambiguity in German association norms. Proceedings of the 11th Conference of the European Chapter of the Association for Computational Linguistics.

Schlaghecken, F., & Bölte J. (1998). Assoziationsklassen bei Basis- und Oberbegriffen. Sprache & Kognition, 17, 21-40.

Snodgrass, J.G., & Vanderwart, M. (1980). A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6, 174-215.

Acknowledgements: This research was supported by a research grant "Anschubfinanzierung" from Saarland University.


  1. selection criteria
  2. collapsing equivalent rows
  3. lines displayed per page
1. Select the database fields to be displayed in the search result. Whatever box you tick will be information that is displayed in the output; you can also constrain your search by choosing from the drop down menus or filling in the search boxes.

Lex. freq.
Part of Speech
Relation Type
Same Cat.
Secondary Meaning
Lex. freq.
Category Second Meaning
2. Select if you want to collapse equivalent rows
2.1 If you want to collapse:

Display Weighted Count

Select the weights you want to map to if you collapse
(Weights can be mapped to other values and are added in the collapsing)
Weight for rank 1:
Weight for rank 2:
Weight for rank 3:
3. Select how many lines you would like to see at once?