Department of Neurology, Northwestern University, USAAccepted 6 July 2006IntroductionThe complexity account of treatment efficacy (CATE; Thompson,Shapiro, Kiran, & Sobecks, 2003) predicts that training complex languagestructures in agrammatism will result in generalize to less complex struc-tures, but only when they share similar derivations. For instance, trainingcomplex sentences derived by ‘wh’ movement, such as object clefts (e.g., Itwas the coach who the skater chased) results in generalization to less com-plex ‘wh’ movement structures such as object-extracted wh questions (Whodid the skater chase?), but not to ‘NP’ movement structures (e.g. passivevoice: The coach was chased by the skater). To date, CATE has been testedwith noncanonical sentence structures, which are particularly difficult foragrammatic speakers. In this paper, we examined the generalization pat-terns that result from training functional morphology in basic canonicalsentences. Specifically, we examined patterns of acquisition and general-ization of:(1) Complementizers: They wonder if the cat is following the mouse.(2) Past tense inflections: Yesterday the cat followed the mouse.(3) Presentagreement inflections: Nowadaysthe cat follows the mouse.While all three morphemes are considered functional category mem-bers, complementizers and tense/agreement are not equivalent with respectto their structural position in the syntactic tree. Complementizers are ele-ments generated in the complementizer phrase (CP); while tense and agree-ment inflections are both elements licensed by the inflection phrase (IP).Therefore, based on CATE, we predicted generalization between tense(2) and agreement (3), but NOT between complementizers (1) and tense(2) or between complementizers (1) and agreement (3), since they arelicensed by distinct functional projections.In contrast to CATE,the Tree pruninghypothesis (TPH; FriedmannG Thompson, 2005), and spontaneouslanguage patterns. Participants were between 36–68 years old, and hadbetween 12–20 years of education.Design and analysisA single subject multiple baseline design across behaviors and partici-pants was used. Pre- and post-testing measures included the Westernaphasia battery (WAB; Kertesz, 1982), the Verb Inflection Test (VIT;Bastiaanse & Thompson, 2005), and two neuroimaging (fMRI) tasksdesigned to examine neural patterns associated with production and com-prehension of the target morphemes. Baseline probes were also adminis-tered to assess production of all target morpheme using picture stimuli(line drawings).During each treatment session participants were asked to produce 15sentences containing the target morpheme. Three types of sentences weretrained: those with complementizers (1), past tense (2), and present tense(3). Training included: (a) thematic role training, (b) placement of writtensentence constituent cards in their surface structure position, (c) readingtarget sentences, (d) reassembly of scrambled written sentence constituentcards, and (d) re-reading sentences. Prior to each session probes, identicalto baseline, were administered to evaluate learning and generalizationpatterns.doi:10.1016/j.bandl.2006.06.049