Maybe "Diable", But Not "Carrotable": How Pseudo-words Can Trick Colombian Spanish Speakers' Brains?
It has been shown that morphologically complex words are decomposed when processed. According to Schreuder, R., & Baayen, R. H. (1995) there are two separate sub-processes after decomposition, namely licensing (checking stem category), and composition (checking semantic compatibility of stem+affix combinations). This has been evidenced by Manouilidou, C., & Stockall, L. (2014) showing that pseudowords (PWs) containing argument violations (*morible “diable”) are more likely to be accepted than those with category violations (*zanahoriable “carrotable”), and they are rejected at a slower rate. We investigate whether the two sub-processes are distinguishable for Colombian Spanish native speakers when suffix attachment rules of Spanish are violated. Contrary to previous studies, here suffix productivity is manipulated. Additionally, we compare verb-attaching to noun-attaching affixes, while previous research has mainly focused on verb-attaching affixes. The experiment involved a 2 x 2 factorial design with two factors: suffix productivity (low, high) and base category (nouns, verbs). For noun bases, we used: 1. suffix -udo (=low productivity) which attaches to nouns that refer to body parts, so we created semantic violations (SemViol) PWs by attaching it to object nouns; 2. suffix -ano (=high productivity) that attaches to proper nouns, so concrete nouns were used for SemViol PWs. In both cases, we created category violations (CatViol) PWs by attaching the suffixes to verbs. For verb bases we used:1. -ble (=high productivity) and 2. -dizo (low productivity) that attach to various verbs, but they seem not to attach to verbs of growing, change of state, or existence, which we used as bases for SemViol PWs. CatViol PWs were created by attaching -ble and -dizo to nouns. We also added Grammatical existing words for the four suffixes, 40 fillers, and 40 nonwords, and created a lexical decision task with 310 items in total.
Preliminary Linear mixed-effects models analysis [accuracy/RTs~Condition (CatViol, SemViol, Gramm) x Suffix (udo,ble,diz,ano)+by-item and by-subject random intercepts] show the expected pattern for BLE and DIZO, with SemViols eliciting higher acceptability, higher RTs (BLE z=-2.31, p= 0.09, DIZO z= -2.90, p=.02) and more errors (BLE z=9.09, p<.0001, DIZO z=4.14 p=.0002). The effect was not found for ANO and UDO. The results show that the pattern does not seem to depend on affix productivity, but they suggest dependence on base category: with noun-attaching affixes, violations of category and violations of semantics seem to be equally bad, fast, and accurate. Further research is needed to show if this pattern is triggered by differences in processing noun vs. verb-attaching affixes, or specific semantic restrictions.
 C. Manouilidou and L. Stockall, "Teasing apart syntactic category vs. argument structure information in deverbal word formation: A comparative psycholinguistic study," Italian Journal of Linguistics, vol. 26, no. 2, pp. 71-98, 2014.
 R. Schreuder and R. H. Baayen, "Modeling morphological processing," in L. B. Feldman (Ed.), Morphological aspects of language processing. Lawrence Erlbaum Associates, 1995, pp. 131–154