Modeling a theory without a model theory, or, computational modeling “after Feyerabend”

Abstract

According to the theorv of lexical uccess preseuted l» Le clt. Roeluf, &. ~ll'H·r. processiug of sr-mantic-svntactic intormutiou (i.e.. Icmmu informarion ' and phonologtcal information (i.e. lexeme information r proceedin il strictlv discrcte. seriul manner, \\'(' \\;1I evaluute this daim inlight of recent evidence l'rom the literuture and unpuhlished flndillg~ From our laborutorv. ln their target article. Levelt, Roelofs & ~le~er present a theorv or lexical <leee.sin. peech production that is unique in that it tries to cover niost of the subprocesses involved in lexical Ht:Cl'SSin peaking. ln our commentarv. \\'Cc' will locus on one central aspect of the theorv, namely. the assumption of strictlv discrete serial ,tage. in It'\ical uccess. Accordinz to this assumption. semannc -svntactic (i.e .. lenuun) processing and phonologtcal (i.e .. lexeme l processing proceed ill hm nonoverlapplng stages with phonological activation hl?ing contingent on the selection of the corresponding lemma. \\'hile semantic competitors Illight well become acnvated at the lemma level earlv in lexicahzatiou. their word forms remain inactive. Phouological coactivation. that is. the simultuneous activation of word [orms ether than the target word form, is explicltlv excluded. This i. at least our reading: of the strict principle proposed bv Le\'elt et al. 11991<1\. ln the pre ent \'ersion of tht' tht'Ol:, thi prindpl!' Il<l~bt'en somewhat fehtecl. It nO\\' allo\\'s for one exception: If hm le:--iC'al items are contextuall~' C:'(luall~appropriate to express a gi\en concept, both lemmas might be sdected and each.ofthem will consequenthumlergo phOllOlogic,u acti\ation (herea!ter, Cclllie.t1l1nl appmpr{atelll'Ss aecol/llt). Tlus relaxation of the prindple is <1reaction to recent e:--pt'rilllC'ntal \\'ork delllonstmting that nt'ar~110n~111011' nontarg;et t:ompetitors (e,g., couch-sofa) hecollle phonologkalI~ adh'e. \\'herea' the findinus for nons~llon~'mous semantic l'ompetitors are less eondllsi\'e (Cutting & Ft'ITeim. in press: Jescheniak &. Schriefers 199i; 19913: Le\·elt et al, 1991a: Peterson & Sa\'O\' 19915). The collte:--tual appropriateness account accommodate' tht'se e:--1Jt'liml'ntal f1.ndillg yet maintains the spirit of thC'original plinciple, Ho\\'e\'er, an aIte11lati\'e (':--1)lanationexists, namely that semantie competition must be suffit:iently strong in order to yield measurable phonological (-oactinltion eflet:ts (hereafter strl'lI{!,th of COIlIj)(1iti(1I! aeCcll/lltl. ln studies ill\'e ·tigating nons~11on~,nous semantk competitors, this compt'tition might ha\'e been too sll1<ul (e,g .. Dell & O'Seaghdha 1992; Harle~' 1993; O'Seaghdha & BEHAVIORAL AND BRAIN SCIENCES (1999) 22: 1 47