Family
size, frequency and word length effects across language borders: Dutch
at the conceptual intersection of Finnish and Hebrew
Harald Baayen [1], Avital Deutsch [2],
Ram Frost [2], Raymond Bertram [3], Tuomo Häikiö
[3], Nivja de Jong [1], Rob Schreuder [1] and Fermín
Moscoso del Prado Martin [1].
(1) Interfaculty Research Unit for Language and Speech, University of
Nijmegen,and Max Planck Institute for Psycholinguistics, Nijmegen, The
Netherlands; (2)Hebrew University of Jerusalem, Israel; (3) University
of Turku, Finland.
Recently, a new frequency component in human memory, family size, has
beendiscovered to affect lexical processing in the Germanic languages
Dutch(Schreuder & Baayen, 1997), English (De Jong, Feldman, Schreuder,
Pastizzo & Baayen, 2001), and German (Lüdeling & De Jong,
2002). Family size is the typecount of the number of derived and compound
words in which a certain stemoccurs. Whereas effects of word form frequency
are often assumed to ariseduring the early stages of word identification,
the effect of family size hasbeen shown to arise at the semantic level
of lexical processing. This paper reports three visual lexical decision
experiments, using translationequivalents in Hebrew, Finnish and Dutch.
The family size effect for Dutch(Germanic, family sizes in the medium
range 0-550) is replicated. Moreimportantly, family size also affects
response latencies in Hebrew (Semitic,non-concatenative morphology,
family sizes in the small range 0-25) and Finnish(Finno-Ugric, rich
agglutinative morphology with substantial morphophonologicalaltenations,
family sizes in the huge range 0-7000). The focus of this paper is on
the surprising degree to which reaction times inone language can be
predicted from lexical statistical properties of theirtranslation equivalents
in genetically unrelated languages. Frequency in anyof the three languages
under investigation predicts reaction times in any ofthem even after
partialling out the effects of word length and family size.Similarly,
word length is an excellent cross-linguistic predictor of responselatencies.
Finally, family size emerges as a bi-directional predictor forFinnish
and Dutch as well as for Dutch and Hebrew, but not for Finnish andHebrew
in any direction. Since the translation equivalents in Hebrew, Finnish,
and Dutch share onlymeaning but no form, these results strongly suggest
that not only family size,but also word frequency and word length capture
aspects of the conceptualorganization in the mental lexicon.The asymmetry
in the cross-linguistic predictivty of family size is probablydue to
typological distance: The productivity of Finnish morpholology exceedsthat
of Hebrew morphology by several orders of magnitude, as witnessed by
thedifferences in the ranges of family size counts. In terms of morphological
productivity, Dutch occupies the middle ground between Hebrew and Finnish.