Neighbourhood density effects in auditory nonword processing in aphasia

Abstract

Utrecht institute of Linguistics OTS, Utrecht University, The NetherlandsAccepted 8 July 2005Available online 29 August 2005IntroductionThe question how prior knowledge of words affects processing ofnonwords is important with respect to the study of lexical access andlexical competition. According to the Neighbourhood Activationmodel (Luce & Pisoni, 1998), the auditory presentation of a word ornonword yields automatic activation of all words in the mental lexiconthat sound similar. Experimental results with unimpaired listeners ona lexical decision task showed that nonwords with dense lexicalneighbourhoods led to slower responses and lower accuracy rates thannonwords with sparse neighbourhoods: the nonword must be com-pared against existing entries, which takes more time when there aremore neighbours.The present study examined these effects of neighbourhood densityon auditory lexical decision performance for nonwords with threegroups of listeners: Broca’s aphasic patients, Wernicke’s aphasic pa-tients, and age-matched nonbrain-damaged participants.Broca’s and Wernicke’s aphasic patients have been shown to bedifferentially impaired in lexical access: whereas a number of auditorylexical access studies have shown that lexical activation levels arereduced in Broca’s aphasic patients (Aydelott Utman, Blumstein, S Milberg, Blumstein, & Dworetzky, 1988), these levelsare increased in Wernicke’s aphasic patients (Milberg et al., 1988)who may thus be impaired in selecting one winning word candidateamong the activated candidates. In unimpaired speech processing,words are activated relative to the overall goodness of fit with theauditory input. In Wernicke’s aphasia, less appropriate acousticforms seem to be able to activate lexical items almost to the sameextent as exact-match items. Broca’s aphasics, on the other hand, didnot activate words such as coat when presented with an acousticallydistorted form (VOT-manipulated c*oat) if there was a lexical com-petitor (such as goat: cf. Aydelott Utman et al., 2001). If overalllexical activation is reduced for Broca’s aphasic patients, the differ-ence between nonwords from dense neighbourhoods and from sparseneighbourhoods may be reduced as well. For Wernicke’s aphasicpatients with increased lexical activation levels, however, neighbour-hood density may exert a bigger influence.Thus, based on the assumption that the magnitude of the neigh-bourhood density effect is influenced by the overall level of lexicalactivation, smaller neighbourhood density effects on lexical decisionperformance were expected for Broca’s aphasic patients, whereas en-larged density effects were expected for Wernicke’s aphasic patients,compared to age-matched control participants.MethodEighty monosyllabic Dutch CVC nonwords were constructed: 40with a high number of real-word neighbours (16 or more neighbours)and 40 with a low number of real-word neighbours (10 or less neigh-bours). In addition, 80 Dutch monosyllabic real words were chosen tobalance for lexical status. The words and nonwords were spoken inisolation by a male native speaker of Dutch. An experimental softwareprogramme randomised the order of presentation of the words andnonwords for each participant. Participants were presented with thematerials via headphones. They were asked to provide a lexical deci-sion response after each auditory presentationby pressing either of twobuttons (labelled YES and NO) on a response box as fast and accu-rately as possible.Participants included 8 patients classified as Broca’s aphasic pa-tients (on the basis of the Dutch version of the Aachen Aphasia Test),five patients classified as Wernicke’s aphasic patients, and 11 age-matched nonbrain-damaged control participants.ResultsLexical decision times were initially measured from target onset.However, because Low Density nonwords were generally longer thanHigh Density nonwords, each item’s duration was subtracted from thelexical decision time. Subsequently, the mean item duration of 606 mswas added again to avoid negative response times. The obtained re-sponse times (collapsed over correct NO responses) and accuracypercentages are presented in Fig. 1.Analyses of accuracy rates showed a main effect of Density:accuracy scores were significantly lower for the High Densitynonwords than for the Low Density nonwords (F (1, 21) = 6.30,Brain and Language 95 (2005) 24–25www.elsevier.com/locate/b&lE-mail address: esther.janse@let.uu.nl.0093-934X/$ - see front matter 2005 Elsevier Inc. All rights reserved.doi:10.1016/j.bandl.2005.07.027