240 likes | 411 Views
Affixal Salience and Lexical Processing The Role of Suffix Allomorphy Juhani Järvikivi 1 , Raymond Bertram 2 & Jussi Niemi 1 University of Joensuu 1 University of Turku 2 Finland. The present study investigates:
E N D
Affixal Salience and Lexical ProcessingThe Role of Suffix AllomorphyJuhani Järvikivi1, Raymond Bertram2 &Jussi Niemi1University of Joensuu1 University of Turku2Finland Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
The present study investigates: • The role of morphological constituents in the processing of Finnish derived words with several allomorphic stems vs. derived words with an invariant stem. Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Evidence from Finnish derivations so far: • There is not much in terms of morphological computation (e.g., Niemi et al., 1994, Bertram et al., 2000) • Cross-linguistically, more computation for derived words found in morphologically less productive languages like English (Vannest, Bertram, Järvikivi & Niemi, 2002) Question: Why is this? • Perhaps employed Finnish affixes not salient enough. Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Affixal Salience(Laudanna & Burani 1995): The likelihood of an affix to occur as a processing unit in a given language Affixal Salience is influenced by • quantitative and/or distributional factors such as: • affix length (phonological/orthographic) • affix frequency • confusability: the ratio of a phonological string serving as a real or pseudo affix, e.g., -er in ’sister – worker’ Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Affixal Salience might also be influenced by linguistic properties such as: • Productivity • Semantic coherence • Phonological transparency • Perhaps the salience of Finnish affixes was compromized by one or more of the above mentioned factors • Question: Is the lack of computation as found for Finnish derivations due to the use of affixes of low salience? Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Vannest et al. (2002) found no computation for three Finnish derivational suffixes forming adjectives from noun bases: • -isA: e.g., sotaisa’warlike, bellicose’ of fairly low productivity, low frequency and also semantically unpredictable • -kAs: e.g., tehokas’efficient, effective’ moderately frequent but of quite low productivity • -tOn: e.g., peloton ’fearless’ productive but only moderately frequent Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Thus: Maybe computation for Finnish derivations was not observed because of low salience of the suffixes employed • What we need are more salient suffixes Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Increasing affix salience -Us: forms nouns from verbs (cf. English –ing) e.g., kloonaus ’cloning’ -productive and – different from all other suffixes used so far – highly frequent -(U)Us: the most productive means of forming nouns from adjectives in Finnish (cf. English –ness) e.g., juoppous ’drunkenness’ - very productive and highly frequent (in both respects more salient than any other suffix used so far) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Method • Frequency manipulation (Taft 1979, Vannest et al. 2002) • first set: Surface Frequency varied, Base Frequency kept constant • second set: Base Frequency varied, Surface Frequency kept constant • Visual lexical decision Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Interpretations of Base and Surface Frequency effects: • Traditionally: • Base Frequency Effect: morpheme-based access • No Base Frequency Effect: no morpheme-based access • Surface Frequency Effect: whole word access • No Surface Frequency Effect: no whole word access Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Traditional Interpretations under fire: • Most notably Taft (Nijmegen, 2001) argued that • no base frequency effect does not necessarily imply that morpheme-based access has not taken place • Indeed it is possible, that a high-base frequency word like ’moons’ is accessed rapidly via the base ’moon’, but that in a later stage of combining stem and suffix information, processing is slowed down, because of the unusual combination of ’moon + –s’. • For a low-base frequency word like ’cliffs’, it could be the other way around, slow access of the base, relatively quick combination of stem and suffix information. • Taken together, this means that in effect the base frequency effect could have been wiped out. Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
In any case, however: we can claim that • a positive Base Frequency effect means that the base is accessed • no base frequency effect means that the decomposition route is not effective • Frequency manipulation is an adequate experimental tool for testing the processing of derivational morphology • Experiment with the productive and highly frequent deverbal suffix -Us Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Results XP1: Deverbal -Us Reliable effect of Surface Frequency No effect of Base Frequency (t1, p > .05, t2 < 1) Hi Base:688 (5.2 %) Lo Base:704 (6.8 %) Hi Sur: 650 (1.8 %) Lo Sur:705 (6.8 %) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Then – although highly frequent – perhaps the deverbal –Us is not productive enough after all • Experiment 2 with the highly productive deadjectival suffix -(U)Us (cf., Eng. –ness) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Results XP2:-(U)Us Reliable effect of Surface Frequency Apparent effect of Base Frequency in the wrong direction • non-significant in item analysis (t2, p > .1) Hi Base: 664 (2.9 %) Lo Base: 641 (1.8 %) Hi Sur: 616 (1.8 %) Lo Sur: 641 (2.0 %) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Perhaps Affixal Salience can be compromised in other ways, for instance, when there is non one-to-one relation between form and meaning: • affixal homonymy(Bertram et al., 2000) e.g., Dutch suffix -er: • forms agentive nouns and comparatives FORM SEMANTICS -er ROWER[agent] SLOWER [comparative] Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Also in this study there is lack of one-to-one correspondence between form and semantics manifesting itself differently, namely the other way around affixal allomorphy: for example, Dutch diminutive suffix (Schreuder & Baayen, 1995) FORM SEMANTICS -pje -tje DIMINUTIVE -etje -je Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Suffix allomorphy in Finnish: Examples Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
General Question: • Is affixal salience compromized by suffix allomorphy? More specifically: • Is the lack of computation for highly productive, monofunctional suffixes in low surface frequency ranges, such as –Us (-Uks, -Ukse), due to suffix allomorphy? Affixes that areproductive, monofunctional and phonologically transparent Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Highly productive suffix • -hkO: forms moderative adjectives korkeahko [HIGH + mod] ’quite high, highish’ NO allomorphy e.g., korkeahko, korkeahkon, korkeahkoa, korkeahkoja, korkeahkona etc. • XP with frequency manipulation Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Results XP3:-hkO A reliable Surface Frequency effect In addition: a significant Base Frequency effect Hi Base: 635 (4.5 %) Lo Base: 667 (6.9 %) Hi Sur: 619 (3.4 %) Lo Sur: 643 (6.1 %) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
However, -hkO: corpus-based index of productivity even higher than for –(U)Us Possibility that the results were due to extreme productivity of –hkO alone Experiment with another invariant suffix: • -stO: collective nouns kirjasto [BOOK + coll] ’library’ kirjasto, kirjaston, kirjastoa, kirjastoja, etc. - only moderately productive and of comparatively low frequency Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
Results XP4:-stO A reliable Surface Frequency effect Again: a significant Base Frequency effect Hi Base:687 (3.2 %) Lo Base:738 (6.6 %) Hi Sur: 677 (4.4 %) Lo Sur: 709 (9.4 %) Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002
In conclusion • Affixal Salience is compromized by suffix allomorphy • In case of many allomorphemic variants (as with both –Us suffixes) affixal salience is decreased and the decomposition process is troubled. • In case of formal invariance (as with the phonologically transparent suffixes –stO and –hkO), suffixes are salient enough to allow effective decomposition Third International Conference on the Mental Lexicon, Banff, Alberta, Canada, Oct. 6-8, 2002