We examine the syntactic structures exhibited by persons with dementia in conversation. Traditionally, research on the linguistic abilities of persons with dementia is either observational (reports kept by caregivers) or experimental (e.g. based on comprehension tasks), and the focus is not on the syntactic competence of the speaker. We combine insights from corpus-based syntactic analysis with methods from interactional approaches. Our close syntactic analysis of the talk of 20 persons diagnosed with dementia and their non-impaired co-participants document both grammatical accuracy and the range of linguistic complexity produced as compared to patterns reported in studies of non-impaired speakers. We note the selection of complex grammatical structures (i.e. verbal arguments) is largely a consequence of the interaction rather than evidence of linguistic decrement. We report on the most pertinent of these interactional influences. Our findings support strategies that promote complex linguistic and interactional talk with persons diagnosed with dementia. We believe that these strategies will help increase the interactional opportunities that are lacking as reported by and about persons residing in an assisted living facility (Brooker 2011). INTRODUCTION The increasing population of people over 65 has led to a new area of research in linguistics, sometimes called ‘gerontolinguistics’, the study of the language behavior of older people (Sugatani 2008). While many people suffer no significant loss of linguistic abilities as they age, others experience such losses as the result of age-related neurocognitive impediment. These changes are primarily brought on by dementia, a syndrome caused by any number of progressive disorders that affect memory, thinking, communication, and the ability to perform everyday activities (American Psychological Association 2013). Changes in language use are one of the predominant indicators of dementia progression. One of the first behavioral symptoms of dementia is difficulty with lexical retrieval. Besides noticeable and frequent word searches, a decline in short-term memory processing affects both productive and receptive language processing. As the disease progresses, these symptoms manifest as disjunctions in semantic and pragmatic language use. Late-stage dementia results in greater deficits in linguistic processing that can lead to reliance on formulaic expressions, the loss of affective responses, and, in many cases, ultimate loss of all verbal communication, that is mutism (Bourgeois and Hickey 2009: 54). Decrement in cognitive and linguistic abilities of persons with any form of dementia has been shown to have negative consequence on the types and frequency of their interactions. Most relevant to our study is that persons diagnosed with dementia syndrome or a particular type of dementia who reside in assisted living facilities are estimated to be unengaged with others between 65 per cent (Cohen-Mansfield et al. 1992) and 85 per cent (Burgio et al. 1994; Logsdon et al. 2007) of their day; in other words, persons with dementia have few daily meaningful social interactions (Brooker 2011). As a result of cognitive decrement and the consequential losses in engagement both with others and with their immediate surroundings, persons with dementia report increased frustration, tension, and depression (Nawate et al. 2008). A better understanding of the linguistic and interactional competence of persons living with dementia may lead to communication strategies resulting in improved interactions, for both those with dementia and the people interacting with them. Additionally, a better understanding of the specific areas of linguistic decrement and resilience experienced by persons with dementia during actual conversation may help identify early symptoms and signs of progression. Our investigation is a step toward that understanding. Transitivity patterns and dementia For many types of dementia, disruptions in communication are the result of breaches in semantic memory. These are breaks in the link between real-world knowledge and the retrieval of the linguistic representation of that knowledge. Persons diagnosed with dementia are often not able to recall or name familiar objects (anomia). Research targeting anomia initially focused solely on nouns (Bayles and Tomoeda 1983; Bayles et al. 1985; Bonilla and Johnson 1995).1 More recent studies have demonstrated a comparable deficit with verbs (Bird et al. 2000; Yi et al. 2007). Less attention has been paid to the effects of age and age-related pathologies at the clausal level. Where such studies exist, they are typically not based on conversational data. For example, Snowdon et al. (1996), in their famous ‘nun study’, looked at the relationship between linguistic abilities in early and in late life based on samples from autobiographical texts written by members of a religious order. They calculated ‘idea density’ (measured by number of propositions per 10 words) and grammatical complexity (measured, mainly, by number of complex sentences and embedded structures). One of their findings was that the nuns who showed signs of Alzheimer’s disease tended to be among those whose early essays scored low in idea density and grammatical complexity early on. (Of the two, low idea density seems to be associated more strongly with low cognitive functioning in later life than low grammatical complexity.) Kemper et al. (2001) in their longitudinal study on the syntactic structures produced by aging adults with and without Alzheimer’s disease apply Snowdon et al.’s (1996) methods of measuring complexity to spoken language. Their data come from elicited monologous speech samples (‘Describe an unexpected event that happened to you’). One of their findings was that both idea intensity and grammatical complexity are adversely affected by Alzheimer’s disease (Kemper et al. 2001: 611). However, even adults with advanced dementia were still ‘capable of producing grammatical sentences’ and ‘to convey much basic information despite their word finding and memory problems’ (Kemper et al. 2001: 612). In another study on syntactic complexity in the language of persons diagnosed with Alzheimer’s disease, Bates et al. (1995) looked at the use of particular grammatical constructions, such as the passive construction and topicalization, under highly constrained conditions. They asked aging speakers with and without a diagnosis of Alzheimer’s disease to describe a series of events presented to them in short black-and-white animated films. The events were chosen for their linguistic properties, such as transitivity and aspect (there were actions with and without a logical endpoint, such as running versus someone giving someone a flower), and they did not necessarily depict events from participants’ everyday lives (such as a snake biting a horse). Bates et al. (1995: 490) showed that under these very specific conditions, persons with Alzheimer’s disease exhibited ‘“structure finding problems” analogous (at an abstract level) to their well documented word finding problems’. This finding supports their assumption that there is no strict distinction between accessing the lexicon and accessing syntax. The measures applied in these studies cannot be transferred to our study, since we are interested in how language is used in conversation, that is in an interactive register. It would be inappropriate to apply criteria developed for the assessment of written language, such as the use of complex sentences with multiple levels of embedding (Snowdon et al. 1996) or for retelling events with a required degree of precision, such as use of passivization or topicalization (Bates et al. 1995), to measure a speaker’s linguistic ability to co-construct a conversation, a register that shows an overall avoidance of ‘elaboration or specification of meaning’ (Biber et al. 1999: 1044), and in which over one-third of units are typically non-clausal (Biber et al. 1999: 1071). We therefore decided to focus on a more basic syntactic concept, one that brings together word-level and sentence-level syntactic knowledge in every sentence. Transitivity is a grammatical construct that links semantic information and syntax at the clausal level. Transitivity—technically, the ability of a verb to take an object or the presence of an object in a clause (Huddleston and Pullum 2002: 2016)—allows a speaker to convey particular types of experiential knowledge through grammar. Specifically, transitivity patterns allow us to say who did what to whom. To convey (or interpret) the meaning of a verb, a speaker needs to know how to use the verb correctly in a syntactic frame. We follow a projectional approach to argument realization (Levin and Rappaport 2005), which assumes that a verb’s meaning is encoded in the mental lexicon and that open positions in its lexical representation are projected onto syntactic argument positions.2 In other words, whether a verb is transitive depends on the verb’s meaning, it is not an arbitrary syntactic feature. For example, the meaning of the verb break would be represented as in (1), with x being the external argument (the causer, linked to the subject position) and y being the internal argument (the argument affected by the action, linked to the object position). (1) [x cause [y become broken]] Additionally, in the case of break, a lexical rule can apply that allows for the deletion of the causing sub-event, resulting a sentence with no explicit causer (‘The mirror broke into a million pieces’). Adding an object where none is licensed through argument structure (‘John frowned his sister’) or leaving one out where it is required (‘Harry broke’) makes an utterance ungrammatical and difficult to comprehend. If a speaker’s ability to access the mental lexicon is compromised, one might expect that selection errors occur or that speakers resort to all-purpose verbs, like get. On a more practical level, if a speaker is unable to assign the correct syntactic frame to a specific verb, communication about even very basic events and experiences becomes quite difficult. Experimental studies have documented the effects of age and morbidity on persons’ verbal semantic abilities, primarily in their ability to comprehend argument structures during elicitation, not interactional, tasks. Price and Grossman (2005) tested persons with dementia of the probable Alzheimer’s type and frontal lobe dementia (FLD) on their sensitivity to transitivity errors. In this word detection procedure, the participants were asked to respond when they heard a particular word (target word). Within the target word sentences, 33 per cent contained a transitive/intransitive violation (e.g. the child limps/limps grass through the mud and leaves), 33 per cent contained a thematic argument violation (e.g. The lizards/feathers cling next to the cage), and 33 per cent of the remaining sentences served as fillers and contained other kinds of violations (Price and Grossman 2005: 229). Results showed that persons with dementia of the Alzheimer type were insensitive to transitive/intransitive violations, while persons with FLD were insensitive to both types of errors. Price and Grossman (2005) queried the ability of persons with dementia to aquire the argument structure, or theta grid, for a nonce word (‘wamble’). Evaluation for the acquisition of the thematic matrix of the new word was tested using grammaticality judgments of sentences containing the new word. The first study showed persons with dementia of the probable Alzheimer’s type had limited ability in acquiring a new verb’s thematic grid, with a 64.3 per cent accuracy rate for the new word versus an 87.8 per centaccuracy rate for known words (Grossman et al. 2007). These results suggest that for persons with dementia of the probable Alzheimer’s type, neurocognitive insult may impair the neuropathways that link semantic–syntactic modules. As a result, this decrement may create difficulties in comprehension and, possibly, production of syntactically and semantically fitted argument structures during everyday interactions. A few interactional studies suggest that breaks in the semantic–syntactic system due to dementia result in difficulties comprehending the semantic and pragmatic import of co-participants’ utterances (Murdoch et al. 1987; Mendez and Cummings 2003; Forbes-McKay and Venneri 2005). Yet, an understanding of how changes in the linguistic competency of persons with dementia create difficulties in real time has not yet been explored satisfatorily from a linguistic point of view (Davis 2005; de Lira et al. 2011). Further research is needed to gain a nuanced appreciation of both degeneration and resilience of linguistic abilities in productive and receptive skills concomitant with various cognitive pathologies.3 We hope to contribute to the linguistic portrait of persons with dementia by providing an analysis of transitivity patterns produced (and processed) by persons with dementia during a communicative task that many may, and we hope do, experience, despite research that reports of such few such interaction: conversations on mundane topics. METHODOLOGY Our multidimensional approach In our study, we pursue a mixed analytic approach that combines insights from corpus-based syntactic analysis with interactional or discourse methods, which we believe best capitalizes on the the online, performance data of our conversation data. We start with a close syntactic analysis of the verbal patterns during actual conversations produced by the persons residing in assisted living facilities who have a diagnoses of dementia. This allows us to describe two levels of linguistic production: First, we characterize their linguistic behavior in terms of grammatical structures. In our analysis, the transitivity pattern types and frequencies are compared to patterns of usage recorded in benchmark corpus studies that capture the syntactic patterns and errors of (presumably) non-impaired persons (Biber et al. 1999). On a second level, we describe how the utterances emerge within a given conversation as each participant both uses a particular verbal structure and responds to the linguistic resources used by the other. In tracking the grammaticality of the utterances, we also note whether particular ‘errors’ or usages seem to create difficulty or trouble during the talk as evidenced by the participants themselves (e.g. overt clarification questions such as ‘what?’ or ‘huh?’). Additionally, we closely follow both the fittedness (i.e. correctness) and productivity (i.e. original verbal forms versus reuse of a co-participant’s linguistic resources) selected by each participant to better understand the use of linguistic resources in relation to the progression of talk, in terms of both topic development and communication actions (e.g. question–answer and compliment–response). Thus, we quantitatively and qualitatively track the production of verb-argument structures within actual conversations. These two-level analyses allow us to see how a person diagnosed with dementia makes use of linguistic resources that so far have mainly been analyzed through studies of grammaticality judgement, error recognition, and sentence comprehension. Data and participants The data for our study do not come from psycholinguistic experiments or clinical tests. The primary data for this study consist of audio-recorded conversations of face-to-face talk between 20 dyads, each comprised a person diagnosed with dementia of the probable Alzheimer’s type and a non-impaired co-participant. The data are part of the Carolina Conversation Collection (CCC), a specialized Web portal housed at the Medical University of South Carolina (Pope and Davis 2011). The CCC contains collections of conversational interviews with people who are 65 or older, some of whom have been institutionalized due to a diagnosis of dementia. Full disclosure of the participants’ medical records, including tests and comorbidities, is not required to be part of the CCC research database. While we would prefer to have complete disclosure of each person’s demographic information, co-morbidities, and the results of diagnostic exams, these conversational interviews do provide a window into the daily interactions that persons institutionalized with a dementia diagnosis engage in as they interact with non-intimate, non-impaired persons. The 20 dyads in our data set consist of an elderly person with an Alzheimer’s disease diagnosis, who resides in an assisted living facility, and who is engaged in a conversation with an undergraduate or graduate student volunteer, a mere acquaintance or relative stranger to the person institutionalized due to dementia. The duration of conversations was held at a fairly constant 10 min of talk; the median duration of the segments is 9 min and 50 s. Total duration for the 20 audio recordings in our corpus is 205 min. Our corpus of talk by participants with dementia contains 11,482 words and 1,976 clausal units.4 Normalized transcripts provided through the CCC portal were transcribed using Transcriber software (http://trans.sourceforge.net). The transcripts obtained through the CCC Web portal were subsequently converted to meet conversational analytic standards (Atkinson and Heritage 1984; Jefferson 2004). Names and identifying information were permanently removed from the audio, and a tone of the exact duration was substituted to maintain the length of the audio recordings. Pseudonyms were ascribed to the participants in the transcripts to protect privacy. In the transcript excerpts provided, the three-letter code used in the title is associated with a single institutionalized participant’s identity (e.g. WKD indicates a particular institutionalized person with a dementia diagnosis). In the line-by-line transcript excerpts, the institutionalized participant is referred to as Mr or Ms plus an initial (e.g. Mr K); the volunteer co-participant is identified using a given name (e.g. Dell). Syntactic analysis We manually coded the transitivity patterns used by all persons with dementia. We based our coding on the nine transitivity (or ‘valency’) patterns described in the Longman Grammar of Spoken and Written English (henceforth LGSWE, Biber et al. 1999: 384.): SV (intransitive, ‘He is sleeping’), SVA (copular plus adverbial, ‘I didn’t get through completely’), SV + A (intransitive plus adverbial, ‘I just stood there’), SVO (monotransitive, ‘I don’t like peas’), SVP (copular verb with subject predicate, ‘She is my sister’), SVOO (ditransitive pattern, ‘They gave me a book’), SVOP (complex transitive pattern, ‘They called me a liar’), SV + clause (‘He said he would leave’), and SVO + clause (‘He told me I could get up’). It should be noted that most common verbs in English allow more than one pattern. For example, ‘see’ occurs most often in the SVO pattern (‘I can’t see you’) but is also often use intransitively in conversations (‘I see’) and with a clausal complement in academic prose (‘We will see that…’). Therefore, it is important to consider the verb in the context in which it is used. We next compared the findings of the transitivity patterns in the talk of persons with dementia to benchmark patterns described and reported within the LGSWE corpus to see if there is an adherence to specific patterns. We chose the LGSWE corpus rather than the speech produced by the interlocuters not diagnosed with dementia in our sample as a comparison because the LGSWE corpus is much bigger and designed to be representative (the corpus as a whole contains over 40 million words of text, with about 4 million words in the British English conversation subcorpus and about 2.5 million words in the American English conversation subcorpus).5 These patterns included both the types and frequency of verb-argument patterns (e.g. copula, monotransitive, and ditransitive) and the frequency of lexical verb choices as compared to frequency data reported for conversation. Comparing verb usage reported within different corpora gives us insight into the question of whether persons diagnosed with dementia limit themselves to ‘safe’ choices, that is verbs that are appropriate in a variety of semantic contexts (e.g. go or get), or whether they rely on common collocations (‘I don’t know’, ‘I don’t think so’) or multi-word lexical bundles (Biber et al. 1999), as a strategy that would minimize the cognitive load of word retrieval and syntactic processing while carrying a conversation. We acknowledge that the LGSWE corpus does not and cannot supply the detailed interactional information about the conversations from which the data are derived and that data come from both British and American conversations. Still, it represented the most comprehensive information on specific verbal or transitive patterns within conversation available to us. In addition to our record of transitivity patterns, we coded interactional data of the conversations: progression of conversation measured in time increments, topic development, and speech action. This information provides interactional evidence of possible breaks within the syntactic and semantic system potentially occurring due to contributing interactional stressors. Some previous studies have found that accessing the lexicon and assessing syntactic structures are not impaired in the same way as persons diagnosed with dementia. Rather than looking at how purely syntactic processes, such as gapping, might be affected, in the manner of Bates et al. (1995), we decided to focus on an aspect of syntax that has its roots in the lexicon. Our analysis is designed to bring into relief differences of transitivity usage—frequencies or error—possibly due to dementia during the online task of face-to-face interaction as compared to patterns reflected in English conversation corpus studies involving non-impaired speakers. Using conversational data Our close analysis of the conversations allows us to better hear, through repeated listening to the audio recordings, and see, through our detailed transcripts, how the choices of verbal structures emerge in the conversation as a collaborative process of joint sense-making between the participants as the conversation unfolded moment by moment. This approach allows us to investigate transitivity patterns in the talk of persons with dementia as choices in response to or independent of the verbal patterns of the co-participants and within the topic contexts of each conversation. Specifically, we look at how initial choices of verbs and their argument structures were initiated—in full or partial articulation—and whether they were resused, repeated, replaced with either full lexical replacements or through repetition of the initial lexical choice. We note the length of pauses that occurred around the verb phrase—before, after, or between constituents of the verbal phrase during articulation—and at clausal junctures. This allowed us to better understand the syncopation of syntactic and semantic resources for persons with dementia as they were responding to contingencies such as their own thought-language processing or as adjustments made respective of participant responses—verbal or gestural—during the online production of language. In this way, we identify areas of trouble, trouble experienced by the participants, and evidenced by a participant’s use of repair initiators to include self-repairs (full or partial repeats or replacements of lexical resources) or through other-initiated repairs (e.g. ‘you mean x’), or through open class repair initiators (e.g. ‘what?’, ‘huh?’) (Drew 1997). Additionally, through this close attention to the participants’ linguistic choices, we became increasingly attuned to the impact topic selection and development had on the complexity of syntactic production of both participants, particularly topics that were or were not engaged with by the participant with dementia. From our observations, we present an analysis of how verbal choices fit the progression of topics during the moment-to-moment unfolding of talk, and how these choices may impact co-participants’ evaluations of the linguistic and interactional abilities of persons with dementia. FINDINGS AND DISCUSSION Our syntactic analyses show that persons with dementia can exhibit syntactic resilience, even under the cognitive pressure of interaction. In relation to our focus on transitivity, we noted only minor infractions within the execution of verb-argument selection, as well as little evidence of misunderstanding. Our analysis of both topic selection and progression highlights the importance of collaboratively selected topics on the syntactic complexity produced by persons with dementia. Finally, we found that persons who are institutionalized for dementia did display considerable interactional competency in selectively reusing linguistic material put into play by their co-participants and producing interactionally and grammatically well-fitted verbal constructions. We will discuss these results in greater detail below. Syntactic patterns The production of clausal units for the participants with dementia fell within a normal range, which is in line with reports that persons diagnosed with Alzheimer’s disease retain the ability to produce complete sentences until late stages of the disease (Kempler et al. 1987). In total, 40.9 per cent of all units were non-clausal, which compares closely with the LGSWE report of 38.6 per cent for conversation (Biber et al. 1999: 1071). Non-clausal units are typically either inserts, such as ‘hi’, or syntactic non-clausal units, such as ‘Did you?’ (Biber et al. 1999: 1082). Of these two types of non-clausal units, the first type was more frequent in our data than in the LGSWE corpus: mean length for non-clausal units averages 1.31 words as compared with 1.91 words reported in the LGSWE corpus. Clausal units contain 2,043 fully produced verbs, which include verbs that were abandoned and/or replaced by a different lexical choice. This depicts a high density of one verb per every 5.6 words in the talk of participants with dementia compared to one verb per 10 words reported in the LGSWE, an indicator of shorter turns and less elaborate noun phrases. The 10 most common lexical6 verbs in the LGSWE conversation corpus are say, get, go, know, think, see, make, come, take, and want (Biber et al. 1999: 375). These verbs reflect the communicative purpose of the register of communication: ‘talking about what people have done (activity verbs), what they think or feel (mental verbs), or what they said (communication verbs)’ (Biber et al. 1999: 371). These verbs (almost all of which, incidentally, have irregular past tense forms and we did not see any morphological errors in our corpus) belong to different syntactic classes and many of them allow a variety of syntactic frames. Intransitive patterns are represented by come and go. Say, know, think, and want are transitive mental verbs or verbs of communication that are often followed by a clause. Make and get are syntactically and semantically very versatile verbs. In addition to their uses with concrete objects (make (someone) a cake, get (someone) a dog), they can be used as causative verbs with more complex complementation (make someone go to the store, get someone to go to the store). The most common lexical verbs in our corpus look very similar: They are have (145 tokens), know (129), get (83), come (59), go (55), think (52), do (44), take (32), want (31), make/give (29), and tell (26). As in the LGSWE corpus, most clauses exhibit transitive verb patterns (verbs followed by a noun phrase or a clause). They make up about a quarter (27.4 per cent, n = 539) of the verbs that are used in clausal units. However, when considering all verbs, that is also verbs in non-clausal units, copular verbs, the least common verb class in the LGSWE conversation subcorpus (1999: 359), are almost as frequent (26.0 per cent, n = 512) as transitive verbs in our corpus. As to lexical variety, there is no great difference either. In total, 197 different verbs were used, and each speaker used a variety of transitivity patterns with hardly any subcategorization errors. The number of grammatical errors that we found was low (n = 37), and they were mostly agreement errors resulting from complex noun phrases (‘Some of the kids … wants to look’), not argument selection errors. Nine persons with dementia used transitive verbs without objects, but these object deletions (n = 38) were always context-appropriate and did not result in comprehension problems. Thus, we found no evidence that the omission of these objects created trouble for the participants as long as the referent was recoverable in the preceding turns of talk. Excerpt 1 provides an illustration. Excerpt 1: WKD 001b 250 Mr. K: now, angelfish you don't eat those. 251 you never kill them. but grouper, yes. you eat them. 252 (3.8) 253 Dell: so: this was your dinner one night. 254 Mr. K: yeah. that was- that was a large- that made a couple- 255 (1.4) 256 Mr. K: well, my fa- well, his family made it and he had us over. 257 uh, yeah. my son in law did that one. 258 Dell: that is your um- daughter's son- daughter's husband. 259: Mr. K: yeah, my s- son-in-[law.] 260: Dell: [#yeah(.h)] 261: Mr. K: he's the underwater photographer. 262. (4.5) Mr K omits the complement (i.e. subject predicate) in the line 254, but there is no indication in the talk that his omission created trouble for the co-participant, whose reference to dinner in line 253 provides adequate context for the earlier referent ‘angelfish’ in Mr K’s utterance in line 250. Mr K’s absent noun phrase in tandem with the long pause of 1.4 min of silence (line 255) suggests he may be experiencing lexical retrieval difficulties. His utterance at line 256 confirms he is referencing a meal that was prepared from the fish. We conjecture that such omissions are simply part of the economy of conversation and that this practice mirrors that found in the speech patterns of non-impaired interlocutors involved in mundane conversation. As such, we liken Mr K’s omission to other elliptic constructions typical of conversations due to the shared time/space environment (Biber et al. 1999: 156–8). Overall, with regard to both grammaticality and variety, the two populations do not display great differences in the use of transitivity patterns. Together, these facts seem to indicate that knowledge of the link between verb meaning and syntactic behavior is still intact for the dementia patients whose speech we analyzed. However, the rather high number of long pauses preceding verb phrases (n = 64, median duration = 1.19 s) as well as repairs to initially selected verbs (n = 98) and the abandonment of initially selected verbs (n = 44) suggest verb selection/retrieval difficulties. In the next section, we will discuss strategies that speakers in our corpus employed to compensate for these difficulties. Interactional findings: the collaboration of competence During coding, it became apparent that participants with dementia, across all 20 cases, produced a variety of syntactic structures and transitivity patterns. However, fluctuations in the complexity of their syntactic production during the course of conversation seemed, in many cases, to follow similar patterns. During initiation of the conversation, single words or phrases with minimal syntax were often the patterns produced by the participants with dementia. Additionally, these utterances often displayed comprehension problems, some of which could be due to hearing loss and the calibration of both a new participant, particularly one with whom the person with dementia is not intimate. Nonetheless, in those first few minutes of talk, co-participants could be left with impressions that the linguistic abilities of the person with dementia were severely impaired. Most often, however, the participants with dementia go on to produce greater syntactic complexity during the course of the 10-min conversation. Additionally, we noticed that after about 7–9 min of talk, minimal utterances reemerged in which the utterances could lack syntactic structure, most relevant to this study, the absence of verb phrases. What clearly emerged from our analysis of the conversations between persons with dementia and their non-impaired co-participants is that competence or, at least, the impression of competence is, largely, achieved in interactions. That is to say, aspects of talk such as fluctuations in syntactic complexity, requests for clarification or confirmations, absence of the second part of a two-sided action (e.g. question–answer and topic proffers), absent grammatical structures (e.g. referents and objects), and long gaps of silence between turns can be oriented to by co-participants in ways that take on heightened importance. From minimal to complex Keeping the interactional duration consistent, with most conversations occurring over the course of about 10 minu (median duration of the interactions is 9 min and 50 s), we noticed minimal word production with absent verb arguments occurring during the initial 1–3 min of talk. Our data support and contribute to Guendouzi and Müller’s (2001) observation of a warm-up period in their analysis of the conversations with an individual with dementia. Indeed, of the participants with dementia we observed, 12 of the 20 used single word or minimal word utterances strung together by simple syntactic structures (e.g. phrases; clauses with ellipsis; and omission of subjects, pronoun referents, and objects) during this initial period, as illustrated by Excerpt 2. Excerpt 2: TYR 001a 001 Pam: Okay. 002 (0.9) 003 Pam: .h tsk. Alright. 004 (3.2) 005 Pam: So, uhm. tsk. did you have breakfast today? 006 (0.5) 007 Mr. T: Of course I did. 008 Pam: Okay, what did you eat. ha ((laughter)) 009 (1.1) 010 Mr. T: The usual. 011 (0.4) 012 Pam: The usual? 013 Mr. T: Scrambled eggs. 014 (0.6) 015 Mr. T: Sausage. 016 (0.4) 017 Mr. T: Toast. 018 (0.9) 019 Pam: Ok[ay:]. 020 Mr. T: [Cof]fee. 021 (0.9) 022 Pam: Okay. so, uhm, I know the last time I talked to you 023 it was before Thanksgiving. so, how was your Thanksgiving? 025 (0.5) 026 Mr. T: Very good. 027 (0.3) 028 Pam: It wa:s? oka:y? Mr T’s response to Pam’s question is a series of minimal utterances, with only line 007 including a verb form. Mr T’s list of breakfast items and his minimal assessment of the holiday is not oriented to by Pam to be in any way insufficient. Of the 12 participants whose initial production displayed similar minimal syntactic complexity, 7 later produced complex syntactic constructions—full verb phrases with correct and complete argument structures—during one or more segments of their conversations. Five participants produced few complete syntactic patterns during their conversations and instead relied heavily on minimal word–phrase utterances that employed simple copula, monotransitive, or ditransitive constructions. Eight of the participants with dementia displayed complex syntactic patterns throughout their conversations, although the complexity of the syntactic patterns waxed and waned during their interactions. These patterns are comparable to those produced by non-impaired participants engaged in mundane conversations as illustrated by Excerpt 3. Excerpt 3: TTR 002 045 Ms. R: And we had a three story house. 046 (0.4) 047 Bob: mm 048 Ms. R: Very well equipped and very well to look- very good to look at. 049 (1.0) 050 Bob: Nice. 051 Ms. R: Yeah, I- en- I- I- how many times that I have said 052 that I've been very fortunate and I have been. 053 Bob: Nice. 054 Ms. R: So, 055 Bob: Wow. In this segment of talk, Bob, the non-impaired co-participant, produces one-word utterances in response to Ms R’s telling about her childhood life and home (lines 047, 050, 053, and 055). These minimal utterances perform several functions including showing active recipiency, making evaluative comments on her story, and encouraging her to continue with the narrative. As such, they seem to be common aspects of talk and appear quite fitted to interaction. We noted that after several minutes of talk (around the 7- to 9-min mark), participants with dementia seemed to revert to minimal utterances and simplistic syntactic constructions, despite previous engagement with one or more topics. This may suggest that after even short durations of talk, participant resources—be they cognitive or linguistic—are taxed in such a way that continuing with the conversation may be burdensome. Excerpt 6 presents the conversation at about 1:30 min into the talk; Excerpt 7 represents the same participants 4 min later into the conversation. Excerpt 6: WEA 001a 053 Sara: What’s your favorite meal to eat here, (0.2) 054 breakfast, lunch, or dinner. 055 (1.3) 056 Mr. W: Seems like lunch is the best. 057 (0.4) 058 Sara: Lunch is the best. 059 (0.5) 060 Mr. W: Yeah. 061 Sara: Why is lunch the best. 062 (0.6) 063 Mr. W: It’s got the best, best place food.(.) they don’t screw it 064 up as much as- ((noise)) 065 Sara: Well, that’s good. I’m glad they don’t mess it up as much. 066 So what’s on the agenda for today? 067 Mr. W: Uh, I didn’t even look at it. we can take some time. 068 Sara: Okay. At this point in the talk, the joint engagement seems to be shared and enjoyed by the two participants as they converse about food choices provided by the assisted living facility. Mr W produces multiunit turns with moderate complexity. Additionally, at line 067, Mr W expresses a request that demonstrates interactional competence by his suggesting that the two ‘can take some time’ together to engage in this conversation. However, about 4 min later, Mr W is producing single-word utterances with gaps of silence between his responses as shown in Excerpt 7. Excerpt 7: WEA 001b 234 Sara: It’s almost Christmas soon = Christmas and Thanksgivin’. 235 (0.2) 236 Mr. W: Yeah. 237 (2.4) 238 Mr. W: Yeah. 239 (2.2) 240 Mr. W: Did you get enough for the s’material for your— sh- 241 (1.2) 242 Mr. W: Yeah.= 243 Sara: =I just need a little bit longer 244 (0.8) 245 Mr. W: hmm 246 Sara: But uhm the rest of it’s good. uhm, 247 (2.6) 248 Sara: You been outside lately to see how cold it is? 249 (0.5) 250 Mr. W: he ((laughter)) Y(h)esterday I was. 251 (0.5) 252 Sara: Yeah? 253 (0.5) At line 240, Mr W makes an overt bid to end the talk by asking his co-participant if she has enough material—recorded material—for the course project that this recorded conversation contributes. Sara is able, however, to reengage Mr W in the talk, and the conversation continues for another 6 min. The reemergence of minimal syntax, long pauses, and displays of trouble in understanding can contribute to the ending of the interaction; however, as noted here, a co-participant can do interactional work to reengage the person with dementia (line 246ff). We believe the initial moments of interaction may contribute to co-participants’ overall impression that the person with dementia is displaying features of linguistic or interactional decrement. Likewise, winding down periods may also be interpreted by co-participants as evidence that persons with dementia are not able to engage in lengthy conversations, and, unlike our co-participant in Excerpt 7, they may not attempt to reengage the person with dementia. Topic selection Topics that appeared to contribute to disfluency in the grammatical production of persons with dementia often include quiz-like question–answer sequences that were, for example, focused on what the person had to eat that day, names, and/or ages of relatives, or other types of inquiries that required details about personal or family history for which the participant with dementia struggled with memory difficulties. We noted that topics that did not engage the participant with dementia resulted in the continuation of or return to minimal utterances, simple syntax including the absence of verb phrases, and arguments or utterances that could be interprested as error (‘can’t think myself’ (line 234, below), and increased gaps of silence, as evidenced in Excerpt 8. Excerpt 8 TGG 001 232 Lisa: What's your brothers’ names. 233 (1.4) 234 Ms. T: Charlie:: and (0.9) eh m’see:, (1.8) can't think myself. 235 he he .h [ah:,] ((laughter)) 236 Lisa: [It's] okay. 237 (1.1) Lisa’s question draws attention to Ms T’s memory difficulties as seen in her palliative response at line 236: [It's] okay. Such moments were common when a participant with dementia could not answer questions. In Excerpt 9, we first present an interaction that demonstrates how topic selection can impair the interaction and obscure the linguistic competency of the participant with dementia and lead to a bid by the person with dementia to end the conversation. Excerpt 9: TYR 001c 152 Pam: Yep, So h(h)ow's your gr(h)andson? (.) Have you talked to 153 him lately? Hh 154 (1.0) 155 Mr. Y: He- (.) I guess, (1.8) he doing fine. 156 (2.4) 157 Pam: He came for Thanksgiving too. 158 (0.7) 159 Mr. Y: Pardon? 160 (0.3) 161 Pam: He came for Thanksgiving- on Thanksgiving Day. 162 (0.3) 163 Mr. Y: No. Pam has proceeded to ask Mr Y about his family, and this topic is met with some resistance. Mr Y’s utterances have reduced to single words and simplistic syntactic utterances. As Pam continues to pursue this topic, Mr Y asks when she will be done with recording him, see Excerpt 10. Excerpt 10: TYR 001d 176 Mr. Y: When will you be through with this. 177 (0.8) 178 Pam: Uhm. 179 (0.3) 180 Pam: .h Today is my last day a:nd (1.1) h::m, (0.8) h::m, 181 a few more minutes and, (1.0) we'll be do:ne. Changing interactional roles When Mr Y reengages in the conversation by switching roles and initiating his own set of questions for Pam to answer, the conversation proceeds and Mr Y’s deployment of complex syntactic structures that include a variety of transitivity patterns emerges. Mr Y’s use of questions displays resilient syntactic complexity, see Excerpt 11. Excerpt 11: TYR 001e 223 Mr. Y: When will you be graduating? 224 (1.1) 225 Pam: I gradua:te (.) next year. 226 (0.7) 227 Pam: Ma:y two thousand and ten. 228 (0.3) 229 Mr. Y: Um’hmm. 230 (0.8) 231 Pam: Yeah, so I have about a year left. 232 (5.4) 233 Mr. Y: And what- what will you be in- what are y- (1.0) 234 going to school for. 235 (0.5) 236 Pam: I'm going to school to be a social worker. By taking the role of questioner, Mr Y resumes interest and the talk continues for 3.5 min. When the participant with dementia shows engagement with the topic, we found the ability to construct well-fitted, syntactically complex utterances followed. In sum, both topic development and interactional roles contributed to displays of complex grammatical structures by persons with dementia. In contrast to topics that did not engage participants with dementia, topics that were jointly pursued resulted in displays of complex syntax, more equally distributed talk, and, sometimes, longer conversations than expected by the participants as evidenced in the their comments. Topics that were centered on locally present objects (e.g. activity outside or seen through a window, décor, and television programs), cultural or imminent events (e.g. Thanksgiving, a fashion show), and participant-experienced events (e.g. a birthday party or a family visit) proved to generate the most complex utterances and maximize an equitable volley of turns. Additionally, by allowing the person with dementia time and opportunity to take control of the conversation through speech actions such as questions also led to more robust displays of grammatical complexity in transitivity patterns and in sentence structure. SUMMARY AND APPLICATIONS Looking at data from naturally occurring conversations, we investigated the language of persons with dementia from a syntactic and interactional perspective. Starting with the close syntactic analysis of naturally elicited talk of 20 persons diagnosed with participants with dementia who were engaged in conversations with non-impaired co-participants who were non-intimates or, relative, strangers, we examined the linguistic structure of transitivity as our object of study to document both grammatical accuracy and range of linguistic complexity represented in the talk as a possible indicator of decrement due to disease. What we discovered was one slice of one interaction may highlight single utterances with little import to the topic at hand. Another moment, however, may display great syntactic complexity. While this is true for all of us across the many conversations we have, for persons with dementia who are residing in an assisted living facility, there are fewer opportunities to engage in meaningful conversations (Brooker 2011), and their linguistic competence is often judged based on these very small ‘slices’ of conversation. Our syntactic analysis provides evidence that the grammatical competence of participants diagnosed with dementia remains largely intact during the production of talk. The high density of verbs, however, suggests that persons with dementia may, indeed, select a verb and then choose a different verb or reissue the same verb, a process referred to in the interactional literature as ‘repair’. This finding suggests that persons with dementia may be uncertain of the appropriateness of the syntactic–semantic fit upon their initial selection. Nonetheless, second reissues of the same or different verbs proved overwhelmingly felicitous. Additionally, the use of all nine transitivity patterns across the speech of participants with dementia serves as further evidence of the semantic and syntactic resilience despite the impediment of neurocognitive disease and despite difficulties noted in the experimental research that reports problems with recognizing transitivity errors and/or difficulty in learning transitivity patterns for new or nonce verbs (Price and Grossman 2005). Participants with dementia, overall, produced shorter strings of words. The mean length utterance for this population proved 31 per cent lower than the rate for non-impaired persons.7 That is, persons with dementia may be seen to rely heavily on minimal syntactic formulations in their conversations. However, after conducting a close analysis of the collaborative production of the talk between co-participants, the relevance of syntactic complexity and average length of utterance were better understood as the consequence of other, contextual factors. The close analysis of the interactional context in which the verbal choices are produced complements the discrete linguistic analysis patterns; together a more complete portrait of the linguistic and interactional abilities of persons with dementia is provided. We found that shorter mean length utterances that include absent verbs or incomplete argument structures typically occurred at the peripheries of the conversation—during the initial first few minutes of the interaction and the minute or two preceding the end of the conversation. We speculate that there may, indeed, be both a warm-up and wind down period. Buidling upon Guendouzi and Müller’s (2001) findings, we suggest that the participants with dementia need time to calibrate their own receptive and productive abilities with multiple aspects of the interaction to include the linguistic features of their co-participant. One unintended but potentially beneficial outcome of analyzing conversations involving a person with dementia engaged with a non-intimate, or relative stranger, is the potential insights these conversations may provide for interactions that occur in the later course of the disease. That is to say, when interactions with family members become more like those involving strangers, which can be frustrating to both participants, our data suggest the following: (i) family members should recognize that this warm-up time is needed and is not necessarily evidence of linguistic or cognitive inability to converse but rather an interactional process affected by the presence of disease and dislocation of place (i.e. the assisted living facility); (ii) topic selections that include those located in the moment or that involve subjects that are omnipresent to all participants, such as weather, holidays, items in the room, rather than personal history or relationships, may be more productive; and (iii) when possible, non-imparied participants should allow or encourage the person with dementia to control the flow of talk. For these reasons, we believe assessing the linguistic ability of a person with dementia after only a few minutes of talk seems inappropriate, let alone after a battery of lexical retrival tests during a stressful evaluation. In addition to the need for a warm-up period, we suggest there is also a wind down period in which participants with dementia seem to be taxed by interaction, particularly by the demands of questions posed to them. The features we have identified may help co-participants watch for possible bids by the person with dementia to end the interactions include a return to minimal token responses and repetitions of co-participatnt utterances. As noted above, one strategy that seems to reengage the person with dementia is redirecting the topic to a more locally available subject, such as some item in the room, in view through a window, an upcoming holiday or event. As noted, we found topics that arise from objects or local events that are more or less equally accessible to all the participants are less affected by memory impairment and allow for smoother interactions. In contrast, topics that require past or recent memories, particularly those that specifically request discrete information such as names, dates, time-related information (e.g. ages, anniversaries, and names of grandchildren), create disturbances in the progression of talk, as participants with dementia may display memory loss or cognitive difficulties. Avoiding such topics could eliminate many of the overt challenges to a fluent discourse. These challenges can, indeed, shut down the responsiveness of persons with dementia. Overall, our findings support a multidimensional approach to assessing the linguistic abilities of persons diagnosed with dimentia. Assessments based on a single method, such as psycholinguistic tests, may overlook significant aspects of the communication abilities of persons with dementia, resulting in a focus on deficits. We therefore propose developing diagnostic tools that would better capture an individual’s linguistic and interactional competency. Additionally, we hope that our findings will contribute to developing best communication practices between practitioners or caregivers and persons with dementia by raising awareness of factors such as the warm-up and wind down periods, the import of interactive roles, and the selection of immediate and local topics of conversation. Trini Stickle is assistant professor of English at Western Kentucky University, Bowling Green and Glasgow. She has published on the linguistic and interactional strategies of persons in meeting settings and children with autism during clinical evaluations. Her main research interest is the linguistic and interactional competency of persons of special populations (e.g. dementia and autism). She has also published on the Dictionary of American Regional English (DARE) and is currently working on a secondary curricula using the DARE materials for high school English, literature, and history courses. Address for correspondence: Department of English, Western Kentucky University, 1906 College Heights Blvd #11086, Bowling Green, KY 42101-1086, USA. <firstname.lastname@example.org> Anja Wanner is professor of English Linguistics at the University of Wisconsin-Madison. She has published on theoretical and applied aspects of transitivity in English, the relationship between syntax and genre, and the role of prescription in language change. NOTES 1 The lexical retrieval of noun also plays a prominent role in standardized assessments of dementia, such as the Addenbrooke’s Cognitive Examination. Interactional linguistic abilities are not assessed. 2 There is a lively debate in linguistics on how exactly children acquire this kind of knowledge—whether through a general-purpose learning mechanism (Tomasello 2003) or through a domain-specific language acquisition device (Lidz and Gleitman 2004). Whichever way it is, information about a verb’s arguments constitutes knowledge of a speaker’s mental grammar that must be accessed every time a verb is used. 3 The effects on verb-semantic comprehension in persons with other types of dementia (e.g. vascular, Lewy body, Parkinson’s, or Huntington’s) have yet to be investigated. 4 While this seems very small compared to the size of the LGSWE conversation corpus (3.9 mio words), this corpus is actually on the larger size for studies that use the methodology of conversation analysis. For example, Kärkkäinen’s (2003: 6) study of expistemic stance markers in American conversation relies on five convesations, the total duration of which is 1 h and 43 min—no exact word count is provided. 5 According to Biber et al. (1999: 27), a 1 million word corpus of speech corresponds to 140–150 h of conversation. 6 Auxiliaries were tabulated separately and are not considered here. 7 A similar observation was noted in a words per turn comparison of persons with dementia and normal elderly by Ripich et al. (1991). Acknowledgements The authors would like to thank the audience at the 3rd Conference of the International Society for the Linguistics of English for valuable feedback and the Carolina Conversation Collection’s team for giving us access to their corpus. Support for this research was provided by the Office of the Vice Chancellor for Research and Graduate Education at the University of Wisconsin-Madison with funding from the Wisconsin Alumni Research Foundation (#140192). References American Psychological Association. 2013. Diagnostic and Statistical Manual of Mental Disorders , 5th edn AMA. Atkinson J. M., Heritage J.. 1984. Structures of Social Action . Cambridge University Press. Bates E., Harris C., Marchman V., Wulfeck B., Kritchevsky M.. 1995. ‘Production of complex sentences in normal ageing and Alzheimer’s disease,’ Language and Cognitive Processes 10: 487– 539. Google Scholar CrossRef Search ADS Bayles K. A., Tomoeda C. K.. 1983. ‘Confrontation naming impairment in dementia,’ Brain and Language 19: 98– 114. Google Scholar CrossRef Search ADS PubMed Bayles K. A., Tomoeda C. K., Kaszniak A. W., Stern L. Z., Eagans K. K.. 1985. ‘Verbal perseveration of dementia patients,’ Brain and Language 25: 102– 16. Google Scholar CrossRef Search ADS PubMed Biber D., Johansson S., Leech G., Conrad S., Finegan E.. 1999. Longman Grammar of Spoken and Written English . Longman. Bird H. M., Lambon R. A., Patterson K., Hodges J. R.. 2000. ‘The rise and fall of frequency and imageability: Noun and verb production in semantic dementia,’ Brain and Language 73: 17– 49. Google Scholar CrossRef Search ADS PubMed Bonilla J. L., Johnson M. K.. 1995. ‘Semantic space in Alzheimer's disease patients,’ Neuropsychology 9: 345. Google Scholar CrossRef Search ADS Bourgeois M., Hickey E.. 2009. Dementia: From Diagnosis to Management—A Functional Approach . Taylor and Francis. Brooker D. 2011. ‘The enriched opportunities program for people with dementia: A cluster-randomised controlled trial in 10 extra care housing schemes,’ Aging & Mental Health 1– 10. Burgio L. D., Scilley K., Hardin J. M., Janosky J., Bonino P., Slater S. C., Engberg R.. 1994. ‘Studying disruptive vocalization and contextual factors in the nursing home using computer-assisted real-time observation,’ Journal of Gerontology 49: 230– 9. Google Scholar CrossRef Search ADS Cohen-Mansfield J., Marx M. S., Werner P.. 1992. ‘Agitation in elderly persons: An integrative report of findings in a nursing home,’ International Psychogeriatrics 4/ 4: 221– 40. Google Scholar CrossRef Search ADS PubMed Davis B. 2005. Alzheimer Talk, Text and Context: Identifying Communication Enhancement . Palgrave McMillan Ltd. de Lira J. O., Ortiz K. Z., Campanha A. C., Bertolucci P. H. F., Minett T. S. C.. 2011. ‘Microlinguistic aspects of the oral narrative in patients with Alzheimer's disease,’ International Psychogeriatrics 23: 404– 412. Google Scholar CrossRef Search ADS PubMed Drew P. 1997. ‘Open’class repair initiators in response to sequential sources of troubles in conversation,’ Journal of Pragmatics 28: 69– 101. Google Scholar CrossRef Search ADS Forbes-McKay K. E., Venneri A.. 2005. ‘Detecting subtle spontaneous language decline in early Alzheimer’s disease with a picture description task,’ Neurological Sciences 26: 243– 54. Google Scholar CrossRef Search ADS PubMed Grossman M., Muray R., Koenig P., Ash S., Cross K., Moore P., Troiani V.. 2007. ‘Verb acquisition and representation in Alzheimer’s disease,’ Neuropsychologia 45: 2508– 18. Google Scholar CrossRef Search ADS PubMed Guendouzi J. A., Müller N.. 2001. ‘Intelligibility and rehearsed sequences in interactions with a DAT patient,’ Clinical Linguistics and Phonetics 15: 91– 5. Google Scholar CrossRef Search ADS PubMed Huddleston R. D., Pullum G. K.. 2002. The Cambridge Grammar of the English Language . Cambridge University Press. Jefferson G. 2004. ‘Glossary of transcript symbols with an introduction’ in Lerner G. H. (ed.): Conversation Analysis: Studies from the First Generation . John Benjamins, pp. 13– 23. Kärkkäinen E. 2003. Epistemic Stance in English Conversation: A Description of its Interactional Functions, with a Focus on I think , vol. 115. John Benjamins. Kemper S., Marquis J., Thompson M.. 2001. ‘Longitudinal change in language production: Effects of aging and dementia on grammatical complexity and propositional content,’ Psychology and Aging 16: 600– 14. Google Scholar CrossRef Search ADS PubMed Kempler D., Curtiss S., Jackson C.. 1987. ‘Syntactic preservation in Alzheimer’s disease,’ Journal of Speech and Hearing Research 30: 343– 50. Google Scholar CrossRef Search ADS PubMed Levin B., Rappaport H. M.. 2005. Argument Realization, Research Surveys in Linguistics Series . Cambridge University Press. Lidz J., Gleitman L.. 2004. ‘Yes, we still need Universal grammar,’ Cognition 94: 85– 93. Google Scholar CrossRef Search ADS Logsdon R. G., McCurry S. M., Teri L.. 2007. ‘Evidence-based interventions to improve quality of life for individuals with dementia,’ Alzheimer's Care Today 8: 309. Google Scholar PubMed Mendez M. F., Cummings J. L.. 2003. Dementia: A Clinical Approach . Butterworth-Heinemann. Murdoch B. E., Chenery H. J., Wilks V., Boyle R. S.. 1987. ‘Language disorders in dementia of the Alzheimer type,’ Brain and Language 31: 122– 37. Google Scholar CrossRef Search ADS PubMed Nawate Y., Kaneko F., Hanaoka H., Okamura H.. 2008. ‘Efficacy of group reminiscence therapy for elderly dementia patients residing at home: A preliminary report,’ Physical and Occupational Therapy in Geriatrics 26: 57– 68. Google Scholar CrossRef Search ADS Pope C., Davis B. H.. 2011. ‘Finding a balance: The Carolinas conversation collection,’ Corpus Linguistics and Linguistic Theory 7: 143– 61. Google Scholar CrossRef Search ADS Price C. C., Grossman M.. 2005. ‘Verb agreement during on-line sentence processing in Alzheimer’s disease and frontotemporal dementia,’ Brain and Language 94: 217– 32. Google Scholar CrossRef Search ADS PubMed Ripich D., Vertes D., Whitehouse P. J., Fulton S., Ekelman B.. 1991. ‘Turn-taking and speech act patterns in the discourse of senile dementia of the Alzheimer's Type patients,’ Brain and Language 40: 330– 343. Google Scholar CrossRef Search ADS PubMed Snowdon D., Kemper S., Mortimer J., Greiner L., Wekstein D., Markesbery W.. 1996. ‘Linguistic ability in early life and cognitive function and Alzheimer's disease in late ife. Findings from the nun study,’ Journal of the American Medical Association 275: 528– 32. Google Scholar CrossRef Search ADS PubMed Sugatani Y. 2008. ‘Gerontolinguistics and care communication,’ Medical Communications 3: 91– 100. Tomasello M. 2003. Constructing a Language: A Usage-Based Theory of Language Acquisition , Harvard University Press. Yi H.-A., Moore P., Grossman M.. 2007. ‘Reversal of the concreteness effect for verbs in semantic dementia,’ Neuropsychologia 21: 9– 19. Google Scholar CrossRef Search ADS © Oxford University Press 2017
Applied Linguistics – Oxford University Press
Published: Mar 16, 2017
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera