This item is only available as the following downloads:
1 DUAL TASK EFFECTS ON LANGUAGE PRODUCTION IN SENTEN CE AND DISCOURSE CONTEXTS I E By JONATHAN PAUL WILSON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 J onathan P aul W ilson
3 To my family for their support and encouragement
4 ACKNOWLEDGMENTS I thank the chair and members of my supervisor y committee for their careful mentoring, Dr. Chris Hass for project design, analysis and recruitment, the staff and members at the UF Libraries for their informed and professional research assistance, the participants in my projects for their honest and op en participation, the student members of the Language Over the Lifespan lab for their invaluable assistance with coding and transcription and the National Institute of Health for its generous support. In particular, I thank my colleagues, Audrey Hazamy an d Dr. Elizabeth Stegemller for their invaluable contribution to recruitment, testing, analysis, student supervision and coding. I also thank my parents and Kostas for their loving encouragement, which motivated me to complete my program of study.
5 TABLE OF CONTENT S page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTORY OVERVIEW ................................ ................................ ............... 12 2 LITERATURE REVIEW ................................ ................................ .......................... 16 ................................ ................................ .. 16 ................................ ................................ ............ 20 Attention Allocation ................................ ................................ ................................ 23 Effects of Healthy Aging ................................ ................................ ................... 24 ................................ ................................ ......... 26 La ................................ ......................... 31 General Summary ................................ ................................ ................................ ... 40 3 METHODS ................................ ................................ ................................ .............. 46 Overview ................................ ................................ ................................ ................. 46 Participants ................................ ................................ ................................ ............. 47 Cognitive Assessments ................................ ................................ ........................... 48 Exper imental Tasks ................................ ................................ ................................ 51 General Procedure ................................ ................................ ........................... 51 Task 1: Picture Description Experiment ................................ ............................ 56 Materials ................................ ................................ ................................ .... 56 Procedure ................................ ................................ ................................ .. 56 Scoring ................................ ................................ ................................ ....... 57 Task 2: Discourse Production Experiment ................................ ........................ 57 Materials ................................ ................................ ................................ .... 57 Procedure ................................ ................................ ................................ .. 58 Scoring ................................ ................................ ................................ ....... 58 Design and Analyses ................................ ................................ ........................ 60 4 RESULTS ................................ ................................ ................................ ............... 64 Sentence Production ................................ ................................ ............................... 64 Quantitative Measures ................................ ................................ ..................... 64 Overall word count ................................ ................................ ..................... 64
6 Sentence length ................................ ................................ ......................... 64 Number and proportion of nouns ................................ ............................... 65 Number and proportion of verbs ................................ ................................ 65 Number and p roportion of adjectives ................................ ......................... 66 Number and proportion of pronouns ................................ .......................... 67 Number of verbs per sentence ................................ ................................ ... 67 Number of prepositions ................................ ................................ .............. 68 Number of words before main verb ................................ ............................ 68 Number of modifiers for each noun phras e ................................ ................ 68 Summary ................................ ................................ ................................ .... 68 Qualitative Measures ................................ ................................ ........................ 69 Propositional density ................................ ................................ .................. 69 Concreteness and word frequency ................................ ............................ 69 Syntactic complexity ................................ ................................ .................. 69 Type token ratio ................................ ................................ ......................... 70 Action Verb Content ................................ ................................ ................... 70 Content word hypernymy ................................ ................................ ........... 71 Informat ion completeness ................................ ................................ .......... 71 Grammaticality ................................ ................................ ........................... 72 Summary ................................ ................................ ................................ .... 72 Discourse Producti on ................................ ................................ .............................. 73 Quantitative Measures ................................ ................................ ..................... 73 Overall word count ................................ ................................ ..................... 73 Sentence lengt h ................................ ................................ ......................... 73 Number and proportion of nouns ................................ ............................... 74 Number and proportion of verbs ................................ ................................ 74 Number and proportion of adjectives ................................ ......................... 75 Number and proportion of pronouns ................................ .......................... 75 Number of verbs per sentence ................................ ................................ ... 76 Number of prepositions ................................ ................................ .............. 76 Number of words before main verb ................................ ............................ 76 Number of modifiers f or each noun phrase ................................ ................ 76 Summary ................................ ................................ ................................ .... 76 Qualitative Measures ................................ ................................ ........................ 77 Propositi onal density ................................ ................................ .................. 77 Concreteness and imageability ................................ ................................ .. 77 Syntactic complexity ................................ ................................ .................. 78 Type token ratio ................................ ................................ ......................... 78 Verb frequency ................................ ................................ ........................... 78 Action Verb Content ................................ ................................ ................... 78 Content word hypernymy ................................ ................................ ........... 79 Cohesion ................................ ................................ ................................ .... 79 Coherence ................................ ................................ ................................ 80 Summary ................................ ................................ ................................ .... 80 Regression Analyses for Sentence Production ................................ ....................... 81 Quantitative Measures ................................ ................................ ..................... 81
7 Qualitative Measures ................................ ................................ ........................ 83 Regression Analyses for Discourse Production ................................ ...................... 85 Quantitative Measures ................................ ................................ ..................... 85 Qualitative Measures ................................ ................................ ........................ 88 5 DISCUSSION ................................ ................................ ................................ ....... 100 Lexical Selection ................................ ................................ ................................ ... 1 01 Information Integration ................................ ................................ .......................... 103 Information Monitoring ................................ ................................ .......................... 106 Implications ................................ ................................ ................................ ........... 109 Future Directions ................................ ................................ ................................ .. 112 Conclusion ................................ ................................ ................................ ............ 113 APPENDIX A PARTICIPANT WITH PARKINSON S DISEASE DISCOURSE SAMPLE ............. 114 B HEALTHY OLDER ADULT CONTROL DISCOURSE SAMPLE ........................... 115 LIST OF REFERENCES ................................ ................................ ............................. 117 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 129
8 LIST OF TABLES Table page 3 1 Rotated solution for f actor analysis. ................................ ................................ .... 61 3 2 Depen dent variables collected in language tasks. ................................ .............. 62 4 1 Simple effects & interaction for quantity measures of sentence production ....... 92 4 2 Simple effects & interaction for quality measures of sentence production ......... 92 4 3 Simple effects & interaction for quantit y measures of discourse ........................ 93 4 4 Simple effects & interaction for qualit y measures of discourse ......................... 93 4 5 Predictors for effects (p=<.055) of quantity measures of sentence production. .. 94 4 6 Predictors for effects (p=<.055) of qualit y measures of sentence production ... 94 4 7 Predictors for effects (p=<.055) of quantity mea sures for discourse .................. 95 4 8 Predictors for effects (p=<.055) of qualit y measures of discourse ..................... 96
9 LIST OF FIGURES Figure page 3 1 Sample stimuli used in the 0 Back task. ................................ ............................. 63 4 1 Task by Group interaction for use of passive verbs in sentence production. ...... 97 4 2 Task by Group interaction for use of participle verbs in sentence production .... 97 4 3 Task by Group interaction for completeness during sentence production ......... 98 4 4 Task by Group by Complexity for grammaticality during sentence production .. 98 4 5 Task by Group interaction f or local coherence during discourse ....................... 99
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philos ophy DUAL TASK EFFECTS ON LANGUAGE PRODUCTION IN SENTEN CE AND DISCOURSE CONTEXTS I E By Jonathan Paul Wilson August 2013 Chair: Lori J.P. Altmann Major: Communication Sciences and Disorders Language production is impaired in people w although controversy exists over which aspects of language are affected. The language impairments in PPD are attributed to disease related changes within prefrontal cortex. T his study investigated language production during s entence production and discourse during single and dual task (stationery cycling) conditions battery of cognitive tasks and two language tasks while riding a stationary exer cise bicycle and again as a single task. Participants described simple and complex picture events using single sentences and, in a separate task, provided three minutes of discourse. We predicted that both the quality and quantity of language production would be impaired in PPD, and that cognitive factors, including executive function, working memory capacity a nd slower general processing speed would account for group differences. While the sentence production task revealed evidence of dual task effects, the discourse task was more sensitive in highlighting language impairments in PPD. More
11 diverse and concrete nouns were produced in both language tasks under dual task conditions. Nouns became less specific and more imageable for PPD during discourse. Th e number of action verbs also increased in both production tasks during the dual task for both groups. Syntactic complexity differences between groups were evident in both language tasks. PPD used shorter less complex sentences during sentence production and discourse, and also conveyed less propositional information during discourse. D ifferences in processing speed and information updating successfully predicted group differences in language performance in both tasks. In conclusion, discourse is affecte d affecting syntactic and information complexity due to poorer executive function and slower processing speed and word choice is relatively well preserved. cts of language production were also observed. F inally language elicitation tasks are differently sensitive to dual task effects with greater dual task effects observed during a highly constrained language production task.
12 CHAPTER 1 INTRODUCTORY OVERVI EW Parkinson disease is a highly prevalent neurodegenerative movement disorder whose motor characteristics are well described. Disease pathogenesis is associated with slow deterioration of nigro striatal pathways limiting dopamine production and leading t o cardinal motor symptoms; tremor, rigidity and bradykinesia. Mild cognitive symptoms often appear early in the disease course associated with reduced dopamine uptake within prefrontal cortex. Cognitive characteristics include early executive, and workin g memory impairments and bradyphrenia. In contrast, while motor speech and language comprehension deficits are well documented, less is known about how cognitive and task demands support language production in this clinical population. Similarly, the eff ects of a dual motor task on language production in PPD are unknown. The dual task literature however, would predict that language production should suffer due to divided processing resources between the two tasks. However, activation of motor cortices m ay actually facilitate / improve some aspects of language use (Rodriguez, 2010) due to the association of areas of somatotopic motor cortex with action word production. In this explorative investigation, PPD were contrasted with a healthy control populati on. Both groups completed a battery of twelve cognitive tasks and two language production tasks, single sentence picture description (with 20 trials per session), and extended discourse production (for 3 minutes). Participants completed both tasks as a s ingle task and as a dual task (i.e. while concurrently riding on a stationery exercise bicycle at a self selected rate) on separate dates. The order of tasks within sessions was fixed while session order (i.e. single or dual task first) was
13 counterbalance d. Production responses were transcribed for off line coding. Picture descriptions were then manually coded for information completeness and grammaticality, and online software was used to code for quantitative and qualitative characteristics of language production for each participant and in each task. disease were impaired in quantitative characteristics of language production during two types of language production tasks durin g simple and dual task conditions. A second aim was to determine whether these individuals were also impaired in qualitative characteristics of language production during the same tasks and conditions. A final aim was to determine the extent to which a r ange of cognitive measures accounted for group level differences in language production. A general prediction was that, due to the condition of motor cortex in people with impa ired in both the qualitative and quantitative aspects of sentence and discourse production. In particular, it was predicted that action verbs would be most impaired and oth er word types. Additionally, it was predicted that critical aspects of language production would improve both quantitatively and qualitatively in the group of individuals at the cycling dual task would facilitate production of action verbs due to priming of motor cortex by the cycling task. Finally, it was predicted that most group level differences would be attributable to individual differences in cognition.
14 The cycling dual task had broader effects on picture description for both groups than expected, while the discourse production task was more sensitive in highlighting group level differences. We attribute these different patterns to cueing effects. The picture descr iption provides external cues for high frequency words and actions, while discourse production requires extended internal cueing, a documented deficit in people topic cohere nce and cohesion. In contrast, picture description is a more constrained behavior and entails further attentional demands relative to discourse production leading to dual task effects. ase used fewer verbs than controls in both tasks. Additionally also used fewer adjectives and pronouns than controls in both tasks and conditions. proper nouns during discourse. Noun use also increased for both groups under dual task conditions, but did not differ between tasks, while overall word counts were lower in the dual task. These finding s isease with relative preservation of noun use. Finally, while the cycling dual task increased the production of action verbs, as predicted, this effect was not greater for people with improved aspects of verb production and noun production in both groups G roup effects were most prevalent in discourse production and were associated with a simplified syntax which conveyed less propositional content with impaired information
15 structure ( e.g., coherence and cohesion). By contrast, dual task effects were most prevalent in the sentence production task in which participants used fewer words overall, produc ed sho rter less complex sentences containing fewer verbs per sentence, and produced fewer passive sentences. Contrary to predictions, the dual task did not improve these qualitative characteristics and had mixed effects on other quantitative and qualitative mea sures of production for both groups. For example, in dual task conditions sentence lengths and syntactic complexity decreased as words became less frequent during sentence production while lexical diversity tended to decrease in both tasks and the speci ficity of verbs increased but only during discourse. Finally, as predicted, both quantitative and qualitative measures in both picture description and discourse production tasks were sensitive to differences in cognitive impairment, particularly processing speed and poorer information updating. However, unexpectedly group level differences in a number of measures of syntactic complexity, information completeness during picture description and aspects of cohesion and coherence during discourse production we re impervious to cognitive factors and were exclusively predicted by group membership. The implications of these findings will be discussed in detail.
16 CHAPTER 2 LITERATURE REVIEW Physical Effects in form is a highly prevalent neurodegenerative illness of idiopathic origin associated with slow deterioration of the substantia nigra pars compacta and chronic depletion of dopamine production levels in the brain affecting the nigro striatal system (Bartel s and Leenders, 2009). Prevalence is reported at between 0.5 to 1% of people between the ages 65 to 69, rising to 1 to 3% among people 80 years of age and older (Nussbaum, R.L., Ellis, C.E., 2003) and incidence is 13.4 per 100,000 persons (Stephen K. Van D en Eeden, S.K, Tanner, C.M., Bernstein, A.L., Fross, R.D., Leimpeter, A., Bloch, D.A., Nelson, L.M., 2003). While pathogenesis is disease are often highlighted (Mayeux, De naro, Hemenegildo, Marder, Tang, Cote & Stern, 1992). During initial clinical presentation, cardinal symptoms, bradykinesia, rigidity and tremor at rest often present asymmetrically (Braak, Tredici, de Vos, Jansen Steur, & Braak, 2003; Bartels et al., 200 9) and motor symptoms appear to respond well to dopaminergic therapies, a differential characteristic (Braak, Tredici, de Vos, Jansen Steur, & Braak, 2003). Bradykinesia describes slowing of voluntary movement (Berardelli, Rothwell, Thompson, & Hallett, 2 001), rigidity is defined as increased passive resistance to stretch movement (Berardelli, Sabra, & Hallett, 1983), and tremor refers to involuntary extraneous movement at rest (Carr, 2001). Other physical features associated with PD include slowed initia tion of movements (Warabi, Fukushima, Olley, Chiba, & Yanagisawa, 2011) and failing motor automaticity (Wu and Hallett, 2005).
17 The subcortical basis for these physical impairments appears to be deterioration of the substantia nigra due to Lewy body patholo gy (Bartels et al., 2009). Healthy dopamine regulation along nigro striatal pathways becomes disrupted due to the disease process leading to volumetric reductions of the striatum (caudate nucleus and putamen). PD is also shown to be associated with thala mic dysregulation which increases longitudinally with disease duration (Peran, Cherubini et al., 2010; Strupp, 2010). Dopaminergic therapies which aim to restore dopamine levels have palliative effects on the motor system at early stages, and the effectiv eness of such treatments for movement disorders diminishes late in the disease course (Marsden, & Parkes, 1977; Rascol, Payoux, Ory, & Ferreira, 2003). In particular, there is evidence that dopaminergic restriction appears to affect functioning of the nig ro striatal system sub regionally accounting for the prominence of motor symptoms early in the disease course. For example, longitudinal studies in PD and healthy adults using positron emission tomography (PET) indicate that posterior striatal structures (posterior putamen) are more vulnerable to dopamine depletion than anterior structures (anterior putamen and head of the caudate nucleus) (Strupp, 2010; Jueptner, Firth, Brooks, Frackowiak, & Passingham, 1996; Lehericy, van der Moortele et al., 1998; Lehr icy, Ducros, Van De Moortele, Francois, Thivard, Poupon,. Kim, 2004). An influential theory of basal ganglia function asserts that deterioration of subcortical nuclei is associated with white matter degeneration of cortico striato pallidal thalamocorti cal circuits which impair cortically mediated functions regionally (Crosson, 1999, Alexander, DeLong, & Strick, 1986). In particular, it is asserted that sub cortical nuclei are connected to regions of motor, oculomotor, dorsolateral prefrontal, lateral
18 o rbitofrontal and anterior cingulate cortex along functionally and anatomically segregated white matter tracts (Alexander, DeLong, & Strick, 1986). Specifically, with respect to motor symptoms, deterioration of the putamen due to reduced levels of dopamine uptake leads to deleterious cortical effects primarily involving pre motor, supplementary motor and dorsolateral pre frontal cortex. Further, fMRI data indicate a patt ern of reduced activation of primary motor cortex together with reduced activation of supplementary motor planning cortex according to disease severity and staging (Tessa, Lucetti, Diciotti, Paoli, Cecchi et al., 2011). The pathophysiological basis of mot or control impairments is at present unclear. However an influential theory is that motor impairment in PD reflects the condition of prefrontal cortex (Rodriguez Oroz et al., 2009; Middleton et al., 2002). In particular, it is asserted that functioning o f the pre motor, supplementary motor, and dorsolateral prefrontal cortex are hypoactive in PD due to denervation (of the striatum) affecting cortico striato pallidal thalamocortical control circuits (Rodriguez Oroz et al., 2009; Middleton et al., 2002). In support, a number of imaging studies of healthy controls and PPD have shown that both the caudate nucleus and the dorsolateral prefrontal cortex are preferentially activated during the performance of executive tests requiring endogenous generation of nove l cognitive sets (category fluency) (Grahn, Parkinson, & Owen, 2008; Monchi, Petrides, Meja Constain, & Starfella, 2007). It is hypothesized that the caudate nucleus and dorsolateral prefrontal cortex modulate cognitive control while the putamen and suppl ementary motor cortex modulate habituated stimulus response learning (Grahn et al., 2008). An alternative account is proposed by Nishio and colleagues (2010) and by
19 Watson and colleagues (2010) who present imaging data that prefrontal cortical volumes are pathologically reduced early in the disease process and posit that motor impairments in PD are thus due to underlying cortical impairments rather than due to an underlying disconnection syndrome. Empirical evidence also suggests that the difficulty of the motor task can influence performance when it is used in a dual (or concurrent) task. It is self evident that some motor tasks, like walking, are more difficult than others and require more planning before movement is initiated (Shumway Cook, Guralnik, Ph Thus, simpler motor tasks, such as cycling, tend to be initiated faster than more complex motor tasks. With a rhythmic task like cycling, planning demands are low because the same movements are repeated. Furthermore, the motion of one leg pushing a pedal will cue the timing and magnitude of the movement of the other leg. In effect, in cycling the same areas of motor planning cortex are activating at a fixed interval, priming the cortex and releasing motor plans to primary motor cortex. In PPD the problem is slow initiation of movements which may potentially be reversed by support characteristics of the cycling task. In support, case series and survey evidence has been presented by Snijders and colleagues (2011; 2010) showing tha t cycling behaviors are strikingly better preserved than walking in a group of individuals with PD, a phenomena which may be related to kinesia paradoxica, a phenomena in which normal motor function is restored due to exogenous sensorimotor cueing (Snijder s et al., 2011). With specific reference to this dissertation, I therefore hypothesize that a cycling task will have beneficial effects on motor control in PPD and will generalize to language performance during dual (multi
20 tasking) conditions by increasin g the stability of motor performance and by reducing concurrent planning demands compared to for example walking. Cognition in Investigations into cognitively impaired populations and healthy aging describe language production as a comp lex behavior supported by limited cognitive resources including attention, executive function, working memory and processing speed (Thornton & Light, 2006; Altmann, & Troche, 2011; Ullman, Corkin, Coppola, Hickok, Growdon, Koroshetz, & Pinker, 1997; Ullman 2004). Critically, some aspects of cognition are subtly impaired even at early stages of PD over and above effects associated with healthy aging while cognitive effects on language production have not yet been fully described. A further interest, direc tly relevant to the current dissertation is the putative relationship between planning of actions and cognitive control, tested under dual task conditions. I hypothesize that the proximal cause of degraded dual task performance impairment in PPD is the co rtical and subcortical degeneration associated with PD. In support, dual task studies in healthy adults suggest that both tasks compete for limited cognitive resources, and show that interference effects between tasks are further exaggerated by PD (Marche se, Bove, & Abbruzzese, 2003; Oliveira, Gurd, performance in PD is of direct relevance to this dissertation since language production appears moderated by both cognition and action planning in PD, and the potential interactions of these tasks remain largely unreported. While motor symptoms tend to characterize initial clinical presentation, cognitive deficits in PD are often reported without clear evidence of dementia or e ven mild cognitive impairment (Watson and Leverenz, 2010; Nishio, Hirayama, Takeda et al.,
21 2010). Indeed, it is estimated that cognitive deficits in early stage PD are present in up to 40% of PPD (Rodriguez Oroz, Jahanshahai, Krack, Litvan, Macias, Bezard & Obeso, 2009; Aarsland, Andersen, Larsen, Lolk, & Kragh Sorensen, 2003; Aarsland et al., 2004; Cahn et al., 1998; Cummings, 1988; Pirozzolo, Hansch, Mortimer, Webster, & Kuskowski, 1982b). For example, at early clinical stages, executive function, memo ry and visuoperceptual deficits have often been observed (Watson et al., 2010) including evidence of bradyphrenia, defined as generalized psychomotor slowing (Rogers, Lees, Smith, Trimble, & Stern, 1986, 1987). As the disease progresses, the effects and de gree of cognitive impairments increases and many PPD convert to dementia late in the disease course due to an emergent cortical pathology (Aarsland, Andersen, Larsen, Lolk, Nielsen & Kragh Sorensen, 2001). Reported cognitive impairments in this popul ation include specific deficits in planning behaviors, problem solving abilities, difficulty with categorizing and clustering information into meaningful semantic sets, as well as problems with set shifting, inhibition, visuospatial processing and more globally with aspects of attention processing (Rodriguez Oroz et al., 2009; Altgassen, Phillips, Kopp, & Kliegel, 2007; Koerts, Leenders, & Brouwer, 2009; Muslimovic, Post, Speelman, & Schmand, 2005; Cooper et al., 2009; Uc et al., 2005). Thus, tasks which require active manipulation of domain general information including shifting of cognitive sets, rather than serial updating or simple recall, are most impaired in PPD (Altgassen et al., 2007; Bublak, Muller, Gron, Reuter, & von Cramon, 2002; Hoppe, Mueller, Werhe id, Thoene, & von Cramon, 2000; Muslimovic et al., 2005; Werheid et al., 2002). For example, Altgassen and colleagues (2007) observed no impairments in immediate, verbatim verbal or visual
22 memory tasks in PPD but revealed impairments in more complex cogni tive behaviors involving storage, updating and set shifting. Similarly, storage and updating problems were also evident during digit and word ordering tasks across a number of studies examining cognitive performance in PPD (Hoppe et al., 2000; Richards, C ote, & Stern, 1993; Werheid et al., 2002). Muslimovic and colleagues (2005) have also reported that individuals with PD were most impaired in tasks assessing either attention or executive function, and that over 60% of PPD were impaired in tasks assessing psychomotor speed and/or controlled processing. Thus, the basis for these cognitive impairments in PD appear to be due to hypometabolism in areas of pre frontal cortex involved in cognitive control (dorsolateral pre frontal cortex) as outlined above. The implications for the current dissertation are that tasks requiring executive control processes and processing speed are impaired in people with PD. In contrast, it appears that immediate verbatim recall is relatively well preserved. The practical signifi cance for this dissertation is that short term memory tasks which include storage, updating and set shifting processes such as the N Back task, will be impaired in PD, potentially due to recruitment of dorsolateral prefrontal cortex which is hypoactive in PPD. It is also suggested that activation of prefrontal cortex (involved in cognitive control and planning behaviors) will interact with motor planning cortex under dual task conditions, and that the direction and magnitude of such effects is determined by the complexity of the concurrent motor task. Consequently, it is predicted that aspects of language production will be impaired in PD due to reduced cognitive capacity. In particular, that constrained sentence production will be slower and less accurat e with respect to information content relative to healthy older adults (HOA) performance, due
23 to increased planning demands requiring executive control and set shifting under timed constraints. I further predict that discourse production will be impaired in PD due to psychomotor slowing and the reduced ability to incorporate new information into an organized grammatical and semantic message structure requiring high levels of cognitive control and memory updating to maintain structural cohesion, information complexity, and topic coherence. Attention Allocation The relevance of attention allocation for PPD is that language production in everyday life does not always occur as a single task. People commonly are expected to communicate while engaged in complex concurrent sensorimotor tasks like walking, shopping, or driving. People are required to allocate attention or other domain general resources dynamically according to contextual circumstances to maintain performance levels on both tasks. As will be revie wed below, evidence suggests that multi tasking or dual tasking becomes increasingly difficult for people with PD and older adults. However, it is currently unknown how language production is affected in PD relative to HOA when a motor task and a language production task are performed concurrently. A widely accepted theory of dual task effects is the capacity sharing model (Pasher, 1994; Yogev Seligmann, Hausdorff, & Giladi, 2008) which asserts that interference or task slowing due to changes in attention allocation occurs when cognitive and motor tasks compete for planning resources. Of particular relevance to this project is the question whether documented cognitive and motor planning problems in people with PD interact at a resource level impairing sent ence and discourse levels of language performance.
24 Effects of Healthy Aging Experimental evidence suggests that dual task interference between cognitive and motor tasks (for example walking) increases with normal aging (Kemper et al., 2006; Li, Lindenberge r, Freund, & Baltes, 2001) and according to the degree of cognitive differences (Holtzer, Burright, & Donovick, 2004). In general however, healthy adults tend to balance task priorities well during dual tasks and allocate resources dynamically to accommoda te a more attention demanding task. In particular, dual task studies which have investigated effects of aging on walking speed, such as the study by Shumway Cook and colleagues (2007), report that as the difficulty of the concurrent distractor task increa ses (walking around obstacles, walking with weights, walking and talking) dual task effects in older adults (over 65 years) increase, leading to slower walking speed, in contrast to younger controls. A further study by Plummer and colleagues (2011 ) also reports increased dual task interference in older adults (slower gait speed) during concurrent cognitive tasks (an auditory Stroop task and a speech task) compared to younger controls. However, there is evidence that effects of a motor dual task on language production in older adults may be prioritized differently by healthy older adults. In particular, it is hypothesized that during concurrent language production younger and older speakers can accommodate to dual task demands in many different ways. In particular, Kemper and colleagues (2003) demonstrated during concurrent walking, production was more affected by dual task conditions than older adults, affecting sen tence length and the grammatical complexity of sentences produced. However, a follow up study by the same group in 2005, indicated that older adults allocate cognitive
25 resources differently from younger adults. In particular, the researchers showed that there were baseline differences in grammatical complexity and that as the difficulty of the motor tasks increased (walking, walking with weights, walking and climbing steps) young adults produced simplified sentences, while older adults, who were already u sing simpler sentences, became less fluent. More typical dual task effects have also been demonstrated with respect to interference from language tasks on motor performance for both young and older healthy adults. In particular, performing language and mo tor tasks simultaneously can lead to bidirectional interactions, as reported by Kemper and colleagues (2005) (discourse production while walking or finger tapping). In a further study by Kemper, Schmalzried, Hoffman, & Herman (2010) using a digital pursuit rotor task it was reported that when the complexity of the motor task increased, both groups changed comparably in speech fluency, information content and syntactic complexity. Dual task effects have also been investigated with respect to effects of a mot or task on cognitive performance in young and older adults. In contrast to other motor tasks, strenuous cycling improves performance in some cognitive abilities and not others, specifically tests of executive functioning, working memory, tone discriminati on, and visual search (Tomporowski, 2003; Sjoeberg, 1975; Lucas et al., 2001; Audifren et al., 2008; Audifren et al., 2009). These facilitatory dual task effects are typically attributed to increases in physiological state arousal associated with strenuou s cycling affecting prefrontal functioning in healthy adults (Audifren et al., 2008). Plummer D'Amato et al., 2008) have also reported that discourse production in particular has significant effects on gait performance in
26 individual s recovering from stroke, and interference due to a discourse production task is greater than interference produced by either concurrent working memory (WM) or executive function (EF) tasks. These studies demonstrate that language and motor tasks interfere with each other in dual task experiments, particularly in populations with already impaired cognitive resources, such as stroke survivors and older adults. These separate studies suggest that language performance during a motor dual task can be protected by using a reduced language register which only becomes vulnerable when the demands of the concurrent task increase. In particular, when the motor task becomes sufficiently demanding, the motor task will result in dual task interference affecting multipl e language measures. While speech rate and fluency are most vulnerable in older populations, semantic and syntactic processing become impaired as attention is taxed well beyond capacity. Thus, motor effects on language production are not limited to motor interference and may include cognitive interference on language performance. In support, a number of studies have shown that both propositional content (Power, 1985) and grammaticality of sentences (Hartsuiker & Barkhuysen, 2006) are impaired when perfor ming a simultaneous memory task. In summary, it has been shown that older individuals with reduced cognitive resources are most vulnerable to attention allocation affecting motor, cognitive and language task performance under dual task conditions. Effects of Disease It has been widely observed that cognitive capacity is reduced in PD and that PPD are increasingly vulnerable to dual task interference effects (Yogev Seligmann et al., 2008). These effects are not domain specific and evidence sugg ests that both cognitive and motor (Ho, Iansek, & Bradshaw, 2002) and motor and motor (Brown, & Jahanshahi,
27 1998) interference during dual task experiments increase in PPD. Specifically, motor performance decrements in PD occur during concurrent motor tas ks (Benecke, Rothwell, Dick, Day, & Marsden, 1986), and during concurrent cognitive tasks (Marchese, Bove, & Abbruzzese, 2003; Oliveira, Gurd, Nixon, Marshall, & Passingham, 1998). As Marchese an d colleagues (2003) and O'Shea and colleagues (2002) have de monstrated, concurrent cognitive tasks impair both gait and balance in PD. More recent studies by Kemps and colleagues (2005) and Yogev and colleagues (2005) also report exaggerated differences in motor performance between healthy older adults (HOA) and pe ople with PD when distracter tasks are more cognitively demanding. motor interference from language or speech tasks. For example, Ho and colleagues (2002) examine walking whi le talking in a task that involves repeating sentences and reported increased reductions in both stride length and walking speed as the complexity of sentences increases. Critically however, it is noted that none of these studies report dual task effects o n language production (Brown & Marsden, 1991; Morris, Iansek, Smithson, & Huxham, 2000; Rochester et al., 2004). Indeed performance on cognitive or language tasks used as distracter tasks are rarely reported in any of these studies. For example, Ho, Iansek and Bradshaw (2002) examined the effects of performing a visuo manual tracking task which involved continuously monitoring the position of a randomly moving needle, and making fine motor adjustments using a joystick to counter the direction of movement, w hile counting or producing general discourse. The researchers reported that discourse, in particular, causes significantly greater dual task effects on motor speech measures for individuals with PD than for healthy controls, but
28 do not report any dual task effects on discourse. In particular, the researchers argue that vulnerabilities in PD and healthy adults are observed for complex task behaviors in favor of less complex task behaviors during dual tasking. Indeed, the researchers report that the speech o f PD patients declined both in volume and fluency and in many cases speech production stopped altogether as motor task difficulty increased. Thus, deficits in attention in PD were further attested by better performance on conversational tasks than continu ous speech tasks and by an observed strategy of task switching from the speech task to the motor task rather than multi tasking to maintain performance levels on the motor task (Ho et al., 2002). Yogev and colleagues (2005) may also be the only study that reports bidirectional dual task effects in PD. The researchers found significant dual task effects on cognitive and language tasks in both the PD and control groups, but that only the PD group experienced a dual task effect on gait, particularly in the lan guage comprehension task. These studies suggest that language based tasks may be particularly demanding of cognitive resources. However, the degree to which language production is vulnerable to dual task interference in PD remains an unanswered question. T he hypothesized vulnerability of automaticity (the ability to perform habitual tasks and behaviors without requiring high levels of attention) in PD which appears independent of behavioral domain or stimulus modality and may be graded to the extent that si mpler tasks are more vulnerable to effects associated with PD than more complex tasks. In support, Holmes and colleagues (2010) performed a dual task paradigm in PD in which postural control and balance, a simple motor task, was shown to be more vulnerabl e to dual task interference from generating a monologue than rote repetition, a
29 language task requiring less cognitive control which was prioritized by individuals with PD (Holmes, Jenkins, Johnson, Adams, & Spaulding, 2010). Further, as Bialystock and c olleagues (2008) have demonstrated under conditions in which task switching is required by the dual task experiment, PD patients give preference to the exogenously cued task over another task and tend to avoid required switch costs. For this experiment the dual task was to prepare a virtual breakfast, setting a table while foods were being cooked for specific times (food preparation). In particular, the researchers found that during a complex visuo motor task requiring cognitive control including planning, updating, working memory, and monitoring, priority was given to the cued task over another less complex related visuo motor scenario requiring less cognitive control. Thus, it appears that people with PD may prioritize tasks differently than healthy adult s favoring more complex tasks irrespective of behavioral domain or stimulus modality, perhaps by allocating cognitive resources exclusively to one task to avoid switch costs (Bialystock et al., 2008). This appears to be the case even when the less comple x motor task is critical (e.g. balance) to task performance and safety (Holmes et al., 2010). This pattern of behavior in PPD contrasts with older adults who maintain cognitive flexibility between tasks, as demonstrated by Doumas and colleagues (2009) and Li and colleagues (2001). The reason for this difference in PPD is unclear and may reflect a general shift in attention allocation away from more automatic behaviors due to difficulties with shifting of cognitive sets in PD, an effect which may be revers ed by exogenous verbal cues demonstrated in the study by Bialystock and colleagues (2008). The neural substrates of dual task effects in PPD to our knowledge have only been explored in one fMRI study to date. In this study it was found that motor to mot or
30 and cognitive to motor interference showed similar patterns of activation (Wu & Hallett, 2008) suggesting that dual task ing activates circumscribed regions of the brain independently of behavioral domain or stimulus modality. Overlapping activation of dual task areas compared to single tasks was under additive, indicating that process overlap regionally was not driving dual task interference effects (Wu et al., 2008) in contradiction of bottleneck processing accounts of dual task interference ( e.g. Ruth ruff, Pashler, & Klaassen, 2001). Bottleneck processing accounts assert that dual task effects are the result of two tasks competing for activation of the same area of cortex resulting in a process bottleneck leading to bidirectional slowing (Ruthruff, Pa shler, & Klaassen, 2001). The study by Wu and colleagues (2008) also showed that PD patients demonstrated increased activation in the cerebellum, prefrontal, middle frontal, parietal, and temporal cortex during dual task s compared to age matched controls and that increased activity in these regions was positively correlated with increased dual task interference (Wu et al., 2008). This finding of increased prefrontal activation in particular is consistent with the findings of Holtzer and colleagues (2005) who determined that executive functioning is the most important single predictor of dual task performance in healthy adults. In the fMIRI study by Wu and colleagues (2008) the only area exclusively and consistently activated (compared to single task) duri ng all dual tasks was the precuneus which was more active for PD patients than controls (Wu et al., 2008). The precuneus is a region of parietal cortex involved in the processing of self related mental imagery at rest (Cavanna & Trimble, 2005). The implic ations of this literature for this dissertation is that dual task effects are exaggerated in PD and that when the distracter motor task is complex, for example
31 during walking or visuo motor tracking, interference effects may become exaggerated in PPD. Ho wever, it is noted that effects of a motor task on language performance in PD have not been investigated. There is some evidence from studies of healthy older adults that a motor dual task interferes with sentence and discourse levels of production and th disease. However, it should also be noted that the choice of dual task will also likely effect language performance, as has been demonstrated in healthy adult populations. In par ticular, there is evidence that concurrent cycling can improve some aspects of cognition but not others. Moreover, due to the low cognitive demands of cycling and repetitive activation of the motor system, some asp ects of language production require activ ation of the motor system, such as action verbs which may be generally facilitated by cycling. Indeed, conceptual content may improve generally under dual task conditions in PD. I predict these improvements will be specific to people with PD due to docume nted impairments in motor performance typically involving hypometabolism of the frontal lobes affecting dual task language production. The status of language processing, especially in non demented PPD is less well documented (Bartels & Leenders, 2009, Altmann & Troche, 2011, Wilson & Altmann, submitted). However, there is a growing body of empirical research which indicates that language performance is subtly i mpaired (Altmann & Troche, 2011; Grossman, et al., 199 1; Grossman, et al., 2000 ; Troche & Altmann, 2012 ). These reported effects for language also appear to reflect cognitive demands during select language processing tasks (Altmann & Troche, 2011 ; Angwin, Chenery, Copland, Murdoch, & Silburn, 2005, 2006; Gro ssman, et al., 1991; Grossman, Carvell, Stern, Gollomp, & Hurtig, 1992;
32 Grossman, et al., 2000; Hochstadt, Nakano, Lieberman, & Friedman, 2006; Lieberman, et al., 1992 ; Troche & Altmann, 2012 ). In effect, for people with PD, as processing demands increase language performance deteriorates. Only a small number of studies have expressly investigated effects of cognitive impairment on language use in PD. These studies have primarily focused on investigating the relationship between cognition and sentence com prehension. Indeed, a relatively consistent finding is that sentence comprehension is impaired in individuals with PD only when presented sentences are syntactically complex (Angwin, Chenery, Copland, Murdoch, & Silburn, 2005, 2006; Grossman, et al., 1991 ; Grossman, Carvell, Stern, Gollomp, & Hurtig, 1992; Grossman, et al., 2000; Hochstadt, Nakano, Lieberman, & Friedman, 2006; Lieberman, et al., 1992). Syntactic complexity effects during comprehension in PD have also been attributed to various impairments in cognitive resources (Grossman, et al., 1991; Grossman, et al., 2000) including: reduced processing speed (Angwin, et al., 2005, 2006), impairments in sequencing and task switching behaviors (Lieberman, 2001), reductions in verbal working memory (Caplan & Waters, 1999), and phonological memory (Lieberman, 2001). These apparently inconsistent findings across studies may be attributed to differences in sentence stimuli, presentation methodology, and the cognitive abilities assessed. As Hochsatdt, Naknao, Lieberman, and Friedman (2006), point out it is likely that multiple cognitive resources interact during sentence comprehension in PD. Very few studies have investigated language production in PD. Indeed, in a recent review, Altmann and Troche (2011) dra w specific attention to several dimensions of language production that have been found to be impaired in PD: information content,
33 grammaticality and fluency. A general finding for people with PD relative to their health peers is that information content i s significantly reduced during language production tasks, including object and picture description, responses to rhetorical prompts, and story generation (Bayles, 1990; Cummings, Darkins, Mendez, Hill, & Benson, 1988; Illes, Metter, Hanson, & Iritani, 1988 ; Murray, 2000). For example, in an early study Iles and colleagues (1988) elicited several minutes of discourse from individuals with PD using leading questions. Subsequent analyses revealed that the proportion of referential comments and interjections is significantly reduced compared to healthy older adults (Illes, et al., 1988). This study indicates that PPD produce less elaborated speech and that information structure is simplified without significant evidence of normal monitoring to engage the list ener. A further study by Small and colleagues (1997) investigated written productions of healthy older adults and individuals with PD from sentences provided on the Mini Mental Status Exam ( MMSE ) which revealed that information content was only reduced in sentences by individuals with increased dementia severity. One of the few studies to explicitly investigate effects of message complexity on information content during sentence level production is a study by Troche and Altmann (2012). In this experiment adults with and without PD describe simple (two actor) and complex (three actor) pictures that are presented offline. While the group with PD produced a lower proportion of sentences with complete information overall, the magnitude of the effects of the complexity manipulation on completeness of information was equivalent between groups. These findings suggest that, while information content may be reduced PPD, increasing message generation demands does not have a disproportionate effect on information c ontent in PD during sentence
34 level production. Wilson and Altmann (submitted) reported a similar study, using the same stimuli that were presented by computer in a single task and during a cycling dual task. Results confirmed that PPD were vulnerable to e ffects of information complexity particularly during dual task conditions which affected response times, grammaticality, and information completeness but not fluency. Further, this study also demonstrated that processing speed predicted decreased performa nce on all sentence measures in PD and that grammaticality scores were also influenced by working memory conditions and vocabulary knowledge. Similarly, there are also inconsistent findings in the literature with respect to grammaticality. Ullman and coll eagues (1997) tested regular and irregular verb production in patient groups (including PPD) by asking participants to read and eported that PPD had particular difficulty forming regular verb endings (rush/rushed) in contrast to relatively well preserved use of irregular verbs (drive/drove). The researchers account for this finding according to a Declarative/Procedural model of ve rb production in which syntactic rules (procedures) are disrupted due to the condition of basal ganglia and frontal cortex while declarative memory stores (supporting use of irregular verb forms) remain relatively well preserved due to the condition of tem poral cortex in PPD. A number of further studies also report that grammaticality is impaired during a range of language production tasks and that syntactic complexity of responses tends to be further reduced in PPD (Holtgraves, McNamara, Cappaert, & Durso 2010; Murray, 2000; Troche & Altmann, 2012). In contrast, a number of other studies report the opposite finding, that grammaticality is
35 unimpaired in PD (Illes, et al., 1988; Murray & Lenz, 2001; Small, et al., 1997). This difference across studies can perhaps be accounted for by task demands. For example, Murray (2000) finds that individuals with PD produce fewer grammatical sentences during extended picture description, while, in a separate study analyzing conversational discourse, Murray and Lenz (2 grammatical as healthy older adults in their productions across a broad range of syntactic measures (Murray & Lenz, 2001). Thus, Murray and Lenz point out that there is an important difference in elicitation task s: Picture description may be a more resource intensive behavior than conversation due to the constraints it places on content, and that effects of grammaticality in PD may be only apparent when the production task is sufficiently demanding. This hypothes is is further supported by the previously mentioned study by Troche and Altmann (2012) using picture description. This study reports that participants with PD produced fewer grammatical sentences overall than healthy older adults, and that these effects we re exaggerated when message complexity increased. However, in a sentence repetition task, the participants with PD did not differ in grammaticality from the healthy older adults. In summary, there are only inconsistent findings with respect to grammatical sentence production in persons with PD, although a careful reading of literature suggests that grammatical impairments are only revealed during more constrained language production tasks, requiring generation of specific information under specific conditio ns. We are aware of only three language production studies that measure fluency effects in PD. This lack of reporting may be due to a wide spread assumption that all dysfluency in PD is due to motor speech deficits (Illes, 1989). However, in an early
36 stu dy of spontaneous discourse, Iles and colleagues (1988) report that the proportion of silent hesitations increases in the speech of PD participants. Further support for fluency impairment in persons with PD during sentence production is provided by the st udy by Walsh and colleagues (2011) who found that both response times and fluency of responses are generally impaired during a repetitive reading task (Walsh & Smith, 2011). However, Walsh and colleagues (2011) found no effects of syntactic complexity aff ecting fluency during their repeated sentence reading task. Similarly, Troche and Altmann (2012) report no effects of syntactic complexity on sentence repetition fluency, although the group of adults with PD was also less fluent than healthy older adults. However, in their picture description task, they find that fluency is disproportionately impaired in PD when message complexity increases. Troche and Altmann (2012) attribute the difference between tasks to the increased demands of message generation and lexical access in the picture description task. In summary, relative to healthy older adults, fluency during language generation tasks is generally impaired in PD, and the difference from healthy older adults is exacerbated when message complexity increas es. Ve rb p rocessing during s ingle word p roduction Of direct relevance to this dissertation, there is compelling evidence that action word processing is impaired in people with PD. In particular, it is documented that action naming to pictures is impaired in contrast to relatively well preserved object naming (Cotellia, Borronib, Manentic, Zanettib, Arevalo, Cappa & Padovani 2007; Rodrguez Ferreiroa, Menndez, Ribacoba & Cuetos, 2009). Additionally, studies have shown that verb fluency (a category fluenc y task) is impaired in people with PD (McDowd, Hoffman, Rozek, Lyons,
37 Pahwa, Burns, & Kemper, 2011; Signorini,& Volpato, 2006; Piatt, Fields, Paolo, Koller, Troster, 1999; Raskin, Sliwinski, & Borod, 1992; Hanley, Dewick, Davies, Playfer, & Turnbull, 1990; Bayles, Trosset, Tomoeda, Montgomery, & Wilson, 1993; Downes, Sharp, Costall, Sagar, & Howe, 1993; Peran, Rascol, DeMonet, Celsis, Nespoulous, Dubois, & Cardebat, 2003). It is hypothesized that action word impairments in people with PD are due to hypome tabolic functioning of motor planning cortex which supports verb selection to a greater extent than noun selection (Peran, Rascol, DeMonet, Celsis, Nespoulous, Dubois, & Cardebat, 2003). With respect to verbal fluency both Signori and colleagues (2006) a nd Peran and colleagues (2002) have shown that action word fluency is an area of greater impairment in PD than other category fluency tasks and is positively associated with executive function impairments by Piatt and colleagues (2006). However, McDowd an d colleagues (2011) used a more extensive neuropsychological battery and determined that action word fluency impairments in PD are due to speed of processing factors rather than difficulties accessing semantic information, or problems with executive contro l. In support, neuroimaging studies of neurologically healthy adults using fMRI have revealed that words associated with body parts activate the same somatotopic areas of motor planning and motor execution cortex as the same physical actions performed in n eurologically healthy adults (Hauk, Johnsrude, & Pulvermller, 2004; Tettamanti, Buccino, Saccuman, Gallese, Danna, Scifo, Fazio, Rizzolatti, Cappa, & Perani, 2006). These findings are also supported by EEG (Shtyrov, Hauk, & Pulvermller, 2004) and MEG st udies (Ntnen, Tervaniemi, Sussman, & Paavilainen, & Winkler, 2001) which confirm that temporal activation patterns in hemodynamic responding to action words
38 are not due to a post lexical response behaviors. Rather, activation patterns are directly rela ted to lexico semantic stages of speech production. Extending this body of work further, Kemmerer and colleagues (2008) used fMRI and a semantic similarity judgment task to identify specific activation patterns associated with semantic features using diff erent verb classes. The researchers discovered that action features of verbs were associated with activation of primary motor and premotor cortices, and revealed that motion features were associated with activation within posterolateral temporal cortices, that contact features were associated with activation across the intraparietal sulcus and the inferior parietal lobule, and that change of state features were associated with activation of ventral temporal cortex. Further research by Kemmerer and colleag ues (2012) has demonstrated that individuals with traumatic brain injuries involving the left hemisphere were impaired in action word comprehension and verb production and localized effects using lesion imaging to inferior frontal, pre central and motor an d pre motor cortices. These same areas are shown to be hypometabolic in PPD. In summary, both action verb naming and category fluency are impaired in PPD, potentially due to reduced activation of motor planning cortex. The implications of this finding fo r the current dissertation are that verb production may also be impaired in language production tasks for people with PD at sentence and discourse levels. This could potentially affect both the quality and quantity of information provided during language production. This conjecture is due to the central syntactic and semantic role that verbs appear to play in sentence and discourse levels of production. While effects of verb impairment have not been explicitly investigated in individuals with PD I make t he following predictions based on the literature outlined above.
39 Specifically, I hypothesize that a breakdown in the selection and use of verbs will impair the overall quality of sentences produced by people with PD. In particular, I predict that the effe ct of verb impairment in PD during the generation of single sentences and discourse would lead to a mean reduction in the amount of information produced during speech, and would further reduce coherence during discourse. Furthermore, I predict that propos itional density would decrease as compensation to maintain information density and fluency. Specifically, I hypothesiz e that reductions in action verb content would be compensated by increased production of nominal and descriptive words in individuals wit h PD. Thus, the number of nominals and other supporting word classes would increase in single sentences to compensate for slowed access to action verbs. In particular, I predict that both the number of modifiers per noun phrase and the number of words be fore a main verb would increase due to slower access to verbs during sentence planning. With respect to discourse production, I predict that both cohesion and coherence would be reduced due to decreased working memory capacity in PPD. I further predict v erb frequency effects, such that more frequent verbs and more abstract verbs are routinely selected by people with PD compared to HOA and that type token ratio for verbs will also be reduced. In effect, more familiar and less imageable light verbs will ch aracterize PD language performance together with reduced diversity of different verb types. Specific predictions for language effects associated with concurrent cycling are more challenging. However, based on the findings that cycling is generally unimpai red in PD due to low cognitive demands on balance and leg movements and that concurrent cycling improves aspects of cognition in healthy adults, I predict that action verb
40 production will be generally facilitated during both language tasks under dual task conditions due to increased activation of motor cortex without competing attention demands. In effect, both quantitative and qualitative measures of language production would normalize during dual task conditions in PD due to motor priming effects. Genera l Summary The theory of embodied cognition asserts that concept knowledge supporting language production is grounded in sensorimotor cortex (Lakoff, & Johnson, 1980). In particular, there is functional magnetic resonance imaging (fMRI) evidence that actio n words associated with body parts activate the same somatotopic areas of motor planning and motor execution cortex as the same physical actions in neurologically healthy adults (Hauk, Johnsrude, & Pulvermller, 2004; Tettamanti, Buccino, Saccuman, Gallese Danna, Scifo, Fazio, Rizzolatti, Cappa, & Perani, 2006). These findings are also supported by both electroencephalographic (Shtyrov, Hauk, & Pulvermller, 2004) and magnetoencephalographic (Ntnen, Tervaniemi, Sussman, & Paavilainen, & Winkler, 2001) approaches. A further fMRI study by Kemmerer and colleagues (2008) in healthy adults provides evidence that action verbs, in particular action features of verbs are associated with activation of primary motor and premotor cortices, in contrast to verbs as sociated with motion, contact and change of state features or nouns. These findings in healthy adults are important in the context of documented differences in action word processing in PPD. In particular, both verb naming (Cotelli, Borroni, Manenti, Zanet ti, Arevalo, Cappa & Padovani 2007; Rodrguez Ferreiroa, Menndez, Ribacoba & Cuetos, 2009) and verb fluency (a category fluency task) are shown to be impaired in PPD (Signorini,& Volpato, 2006; Piatt, Fields, Paolo, Koller, Troster, 1999; Raskin, Sliwins ki, & Borod, 1992;
41 Hanley, Dewick, Davies, Playfer, & Turnbull, 1990; Bayles, Trosset, Tomoeda, Montgomery, & Wilson, 1993; Downes, Sharp, Costall, Sagar, & Howe, 1993; Peran, Rascol, DeMonet, Celsis, Nespoulous, Dubois, & Cardebat, 2003). It has been hy pothesized that the neural substrate for verb impairment is impairment of motor planning cortex resulting from degeneration of the basal ganglia in PPD affecting connecting tracts (Peran, Rascol, DeMonet, Celsis, Nespoulous, Dubois, & Cardebat, 2003) in p articular, due to white matter degeneration of a cortico striato pallidal thalamocortical circuit which connect regions of basal ganglia with motor planning cortex (Alexander, DeLong & Strick, 1986). However, it remains untested whether verb impairments in PD during single word production are also apparent during sentence and discourse production and whether such effects vary due to differences in cognitive abilities, and task difficulty, and whether these effects are exaggerated by multi tasking conditions Thus, the purpose of this dissertation was to investigate whether task demands and cognitive factors increase language production difficulties in people with and mak e the following group level predictions about sentence and discourse level production during single tasks and effects of concurrent cycling. My first specific aim is to determine whether people with PD and healthy adults differ in quantitative characterist ics of language production during single task and dual task conditions and whether group differences can be detected using online coding programs. Specific measures will include basic word counts, and word counts for each word class, together with the num ber of words before main verbs and modifiers per noun phrase. Specifically, I predict that the PD group will produce fewer verbs and an
42 increased number of other words (nouns, adjectives and adverbs) as compensation. I predict that PD participants will i nclude more nominals before main verbs and more modifiers per noun phrase. I further predict that these effects will be exaggerated during single sentence production for pictures with more entities to describe. With respect to cycling effects, I predict an interaction between group and dual task, such that quantification of language produced will improve during dual task for adults with PD but not for healthy older adults. Specific predictions are that cycling will increase the number of action verbs pro duced by people with PD and reduce the number of other word classes. Additionally, due to facilitation of verb access, I predict that there will be fewer words before main verbs and fewer modifiers per noun phrase. The basis for this prediction is the hy pothesis that a cycling task facilitates motor preparation in PD and increases excitation of primary motor cortex facilitating the production of verbs during sentence and discourse production. My second specific aim is to determine whether people with PD a nd healthy adults differ in quantitative characteristics of language production during single task and dual task conditions and whether group differences can be detected using online coding programs. Specific measures will include propositional density, t ype token ratio for verbs, lexical frequencies, concreteness, imag eability, hypernymy, action verb content and information completeness and grammaticality for sentences and cohesion and coherence metrics for discourse production. With respect to group lev el differences, as above we predict that verb production will be impaired in the PD group. This hypothesis is consistent with predictions of the single word production literature which indicates that people with PD have greater difficulty producing verbs than naming objects. Due to
43 reduced use of verbs we predict that the PD group will show reduced propositional content, reduced variety of verbs indexed by type token ratio, reduced action verb content, concreteness indexed by verb hypernymy and an increas ed reliance on high frequency verbs compared to healthy older adults. On more global measures of information content, I predict that information content will be reduced for single sentences and that coherence but not cohesion will be reduced at a group le vel in PD compared to healthy adult controls. With respect to cycling, for the reasons outlined in the previous section I predict that cycling will generally facilitate information completeness for sentences and will reduce verb effects associated with dis course production, in particular affecting coherence. Additionally, I predict that propositional density will increase; a greater variety of verbs will be selected during production behaviors and that action verb content will increase due to priming effec ts of the motor task affecting action verb semantic fields. My final specific aim is to determine whether cognitive abilities in the single task will predict performance during sentence and discourse production in people with PD and healthy older adults du ring both single and dual task conditions. Specific measures to be predicted are information completeness for sentences, propositional density and cohesion and coherence for discourse. Further measures will include any dependent variable showing main eff ects of group or interactions including quantitative measures of verb use, and frequency and qualitative measures of action verb content and type token ratio. With respect to predictors I cannot determine what factors will be extracted during factor analy sis for entry into regression but based on previous literature using similar tasks I predict the inclusion of executive function, working memory, and
44 processing speed factors. My specific predictions with respect to performance are that processing speed, executive function and group will together predict information completeness for sentences. Alternatively I predict that cohesion and coherence metrics will be predicted by a model which contains group, working memory, and processing speed for discourse. With respect to propositional density which is a shared measure between sentence and discourse level tasks I predict an effect of group and working memory. My rationale for making these predictions is that people with PD will be slower to process informat ion, will exhibit problems with cognitive control and will have a reduced ability to manipulate information structures which include updating and shifting of cognitive sets. These problems will manifest differently in different language tasks which place different levels task demands on the language production system. A sentence production task which involves describing simple and complex pictures under time constraints will require the ability to quickly inhibit information during lexical selection and c ontinuously switching between pictures. Simultaneously, the speaker must, maintain other information active in short term storage to support sentence planning, which requires intact executive functioning and speed of processing. In contrast, discourse pro duction is less constrained with respect to lexical selection and more subject to information demands during sentence construction requiring intact working memory. For discourse each utterance is placed within the context of the goals of discourse and the topic selected. As such, the discourse task places high resource demands on working memory, which may be impaired in PD, associated with a reduced ability to actively store and manipulate information. Additionally, there are fewer time
45 constraints on lan guage production in discourse than in a timed sentence production task.
46 CHAPTER 3 METHODS Overview This dissertation is an investigation of language production in people with this study and an additional 19 healthy older adults served as experimental controls. The experimental protocol consisted of a battery of twelve neuropsychological tests and two language tasks. These cognitive tests included measures of basic processing speed, controlled processing, domain specific working memory, and executive function. T wo language production tasks were included a sentence production task and a task in which participants respond to open ended questions. Both the cognitive battery and experimental language tasks were presented to participants as a single task, seated in front of a computer monitor, and as a dual task, riding a stationary exercise bicycle while performing cognitive and language tasks. Discourse production consisted of 3 minutes of connected speech on a given topic; the sentence production task consisted of describing pictured events at two levels of information complexity (simple and complex pictures). Quantitative and qualitative measures of language production were c ollected and analyzed to determine effects of group, dual task and cognitive variability on language performance. Thus, the experimental design consisted of two language tasks tested with respect to two groups (healthy older adults and people with Parkins coded for multiple measures of both quantitative and qualitative indicators of language production performance using manual coding and online corpus analysis software described below.
47 Participants adults (age 72.4, 9.76) participating in this study. Healthy older adults were older than t (1,57)= 2.477, p =.016, while there w as no significant difference 4.04) and healthy older adults (18.4 years, 2.00), p = .260. All individuals with scored between 2 and 3 on the modified Ho ehn and Yahr scale (indicating mild to moderate bilateral disability) and were recruited from the Center for Movement Disorders and Neurorestoration at the University of Florida, in Gainesville, Florida. Healthy older adults were separately recruited from the Speech, Language and Hearing Sciences participant pool at the University of Florida and through contact lists provided by the Department of Physiology and Kinesiology, through word of mouth and flyers distributed at local church and exercise venues. I individuals between the ages of 30 disease. Further criteria included a modified Hoehn and Yahr scale score between 1 and 3 parkinsonian and/or psychotropic medication. Exclusion criteria were signs of secondary or atypical Parkinsonism, or symptoms of cognitive impairment. Any individual with a history of falls, dep ression or generalized anxiety was excluded from participating. Healthy older adults were separately recruited from the Speech, Language and Hearing Sciences participant pool at the University of Florida and through the contact lists described above. The inclusion criteria for controls were monolingual, neurologically healthy individuals (i.e. MMSE > 24) between the ages of 30 85 years. Exclusion criteria were
48 relevant recent history of speech language therapy, major psychiatric disorder, falls, or the u se of medication later shown to alter cognitive functioning. All participants had normal vision or corrected to normal vision and signed an Informed Consent approved by the University of Florida Healthy Science Center Institutional Review Board prior to e nrollment in this project. Cognitive Assessments Each participant completed a battery of cognitive tests during each session. These tests included measures of basic processing speed, controlled processing, domain specific working memory, and executive fun ction (Colcombe & Kramer, 2003) Processing speed was assessed using t hree inter times as they could in 10 seconds. In an adapted version of the digit symbol substitution test, participants viewed an array of non verbal symbols (letters in the Korean alphabet) paired with digits and one larger sym bol displayed below. Participants were required to say the number of the digit associated with the featured symbol. Dependent variables were response times for the star and digit symbol tasks and accuracy in the adapted digit symbol task. Controlled processing was assessed using two tasks: In the visual 0 Back task, participants viewed a continuous series of non verbally encodable tic tac toe figures containing two based on whether the current figure matched a pre specified target figure introduced at the beginning of the task as shown in Figure 3 1 In the Stroop color naming task, participants named colors as quickly as
49 a black screen in either a red, blue or green font. The dependent variable for these two tasks was response time. Working memory was assessed using two visual and two verbal working memory tasks. Visual working memory was assessed with a visual 1 back ta sk and a novel visual working memory task. In the visual 1 back task, participants viewed a continuous series of the same tic tac based on whether the current figure matched the one shown immediately befor e it. In the visual working memory task, participants viewed a series of the tic tac toe figures one at a time, followed by an array showing the same number of symbols. Analysis ranged from 1 4 figures, but the 1 figure condition was excluded from analysis due to its similarity with the 1 Back task. Participants were asked to determine whether the array presented the same figures in the same order as the sequence just viewed. The verbal working memory task s consisted of two digit span tasks in which parti cipants repeated increasingly long lists of single digits either forwards (verbatim) or backwards (in reverse order). The dependent variable for working memory tasks was the percentage of correct trials. Executive function was assessed using three tasks. In a version of the operation span task, participants were shown and instructed to remember 6 consonant strings, and were asked to repeat these back in presentation order following verification responses of 0 4 single step arithmetic solutions (e.g. 3+ 2= 5 3+ 2= recalled in correct order per trial. In the 2 Back task, participants viewed a continuous series of the tic tac toe figures and responded yes or no based on whether the
50 current figure matched the figure shown two screens before it. The dependent variable required participants to name the color font of a series of color words displayed in incongruent color fonts (e.g. the word RED presented in a green font).The dependent variable was the response time for this task and the response time for this task less the response time for the Stroop color naming task, (i.e., the Interference score). Since one of the specific aims of this project was to assess the influence of individ ual differences in healthy older adults on measures of language production I performed principal components analysis on cognitive measures to identify orthogonal factors suitable for use in linear regression. T he following cognitive measures were included in the factor analysis : digit spans forward and backward accuracy, visual memory task accuracy, overall operation span accuracy, Stroop task color naming and color word naming response times, 0 Back response times, 1 Back response times, 2 Back accuracy for digit symbol substitution response times, star task respo nse times, and the The results of this analysis using Varimax procedures successfully identified 3 separate orthogonal factors. Based on the factor loading in Table 3 1, t he first factor which accounted for 36 .8% of the variance, was named the processing speed factor, because the largest factor loadings included Stroop response times, digit symbol substitution response times, the star task response times, 0 and 1 Back response times and articulation speed The second factor which accounted for 16.5% of the variance, was named the working memory factor since the strongest factor loadings w ere to 2 Back accuracy, visual memory task accuracy and
51 operation span accuracy. Finally, the third factor which accounted for 8.8% of the variance, was named the updating factor since the strongest factor loadings were to 2 Back accuracy visual memory and operation span accuracy Information updating is a sub process of executive function ( Miyake, Friedman, Emerson, Witzki & Howerter, 2000 ) which requires monitoring and coding new information for task relevance and revising stored information appropriately. These three orthogonal cognitive factors were used in the regression analyse s described below Experimental Tasks Gen eral Procedure Each participant was randomly assigned to one of two groups. One group undertook a single task session first followed by a dual task session within 7 days and the remaining participants performed the dual task session first followed by the single task session within 7 days. This procedure was employed to control for practice effects associated with task demands. Single task sessions were conducted with participants seated in a comfortable, quiet room, free from other distractions, with all stimuli presented on a 15.6 inch widescreen laptop computer. The dual task session was conducted with participants concurrently riding a stationary exercise bicycle at a self selected comfortable rate in a laboratory under quiet conditions projected onto a back lit projection screen located within 4 meters of the seated participant. Participant safety and comfort was continuously monitored at baseline and during the dual tasks using a heart rate monitor and paired sensor. Prior to testing, the seat on t he exercise bike was adjusted for each participant. Participants were instructed not to cycle during information screens or between tasks. During both single and dual task sessions participants wore a wireless headset microphone for voice recording. Reco rding
52 sensitivity was calibrated prior to testing for each participant. Each session was recorded for transcription, scoring, and reliability analysis. Voice responses were collected as mp3 files. Additionally, a digital master audio recording of the ent ire session was collected as a backup procedure. Following the session participant responses were transcribed verbatim. General s coring Sentences and discourse were transcribed verbatim. Coding procedures consisted of qualitative and quantitative measur es of language per formance as itemized in Table 3 2 Quantitative measures were defined as measures of specific information content while qualitative measures were defined as information measures describing information characteristics. Quantitative codin g consisted of global word counts and proportions obtained using CPIDR (Brown, Snodgrass, & Covington, 2007a). The researchers report high levels of agreement between human raters and CPIDR for both propositional counts and density rating s Similarly, hig h levels of reliability have also been reported with respect to discourse and picture description samples collected from individuals with stroke aphasia (Altmann, Hazamy, Carvajal, Benjamin, Rosenbek and Crosson, in press). Sentence transcriptions were ent ered into CPIDR to obtain sentence by sentence proportion counts and word class tagging. Word class tags from CPIDR were then exported to an Excel spreadsheet to obtain word counts by word class, word class proportions and to extract lists of verbs produc ed to calculate number of verbs per sentence. Responses were subsequently entered into Coh Metrix software (Graesser, McNamara, Louwerse, & Cai, 2004) to obtain counts for number of words before main verbs, number of modifiers per noun phrase, and action verb content. Reliability (Duran, McCarthy, Graesser, & McNamara, 2007) and
53 c onstruct validity (McNamara, Ozuru, Graesser & Louwerse, 2006) of coherence and cohesion indices calculated by Coh Metrix has been demonstrated using published text samples. Th e software has also been used to investigate language proficiency in second language learners (Crossley, Salsbury, & McNamara, 2011), effects of healthy aging (Kemper, Bontempo, Schmalzried, McKedy, Tagaliaferri, & King, 2013) and stroke aphasia (Levy, Hoo ver, Waters, Kiran, Caplan, Berardino, & Snadberg, 2012) Qualitative coding consisted of measures of information content including propositional density, type token ratios for nouns and verbs (based on lemma types), verb frequency, action verb content, no un and verb hypernymy and measures of concreteness and frequency. Three additional measures were collected, syntactic similarity, syntactic complexity and Flesch Kincaid grade reading level. All of these scores were obtained using Coh Metrix. Propositiona l density is the amount of propositional information provided controlling for total number of words produced. Propositional density and number of propositions was extracted using CPIDR and was based on trimmed discourse and sentence transcriptions Trimm ing involved the removal of filler words, repetitions, task commentary, false starts and abandoned utterances. T rimming is a standard practice in the calculation of propositional information including density scores (Turner & Green, 1977; Nicholas & Broo kshire, 1988; Kemper & Sumner, 2001) the purpose of which is to control for calculation of false estimates. False estimates arise due to analysis of the same propositional content (inflated density scores) or by increasing word counts without producing m ore propositions (reduced density scores). All other corpus based analysis for coding used untrimmed sentence and discourse samples. Type token ratio is the number of unique words (called types) divided by the
54 number of tokens (occurrences) of these words Each unique word in a text is considered a word type. Verb type token ratio calculates specific ratios based on verb lemmas as types. Lower ratios indicate more repetition of verbs, thus less diversity of verb choice. Action verb content is referred t Metrix and represents the number of verbs that are action related performed by animate entities. The higher the incidence of intentional actions in a text, the more the text is assumed to convey goal driven co ntent (Landa uer, Foltz, & Laham, 1998). Verb hypernymy is a measure of the mean number of levels in a conceptual taxonomic hierarchy superordinate to a main verb. A word having more hypernym levels is more concrete and a word with fewer hypernym levels is more abstra ct (e.g. broil would have 3 levels: do, make, cook). Propositional density was coded by CPIDR; Type token ratio for verbs (based on lemmas), action verb content, noun and verb hypernymy, concreteness and word frequency for content words and for all words t ogether were collected from Coh Metrix. Concreteness rating are values indexed from the MRC Psycholinguistic database (Coltheart, 1981) that provide a measure of how concrete or abstract a word is with lower scores indicating a les s concrete word (e.g. ha ppiness ) and higher scores indicating a more concrete word (e.g. ball). Written word frequencies for content words collected from Coh Metrix were frequency values abstracted from the CELEX online database (Max Planck Institute for Psycholinguistics, 2001). Lists of verbs were also entered into the Corpus of Contemporary American English (Davis, 2010) to obtain spoken word verb frequencies using a larger corpus of 450 million words. Syntactic similarity scores collected from Coh Metrix quantify the proport ion of intersecting tree nodes between adjacent sentences (local syntactic similarity) and all combination of
55 sentences (global syntactic similarity). Higher values indicate use of a greater proportion of syntactically similar sentences and lower values i ndicate a lower of 1, indicating a high level of local syntactic similarity, while the 0.211, indicating a low level of local syntactic similarity. Syntactic complexity scores Metrix) are z scores representing the mean number of words per sentence controll ing for the frequency of simple ( familiar ) and complex ( unfamiliar ) score of 1.353) whil e lower scores indicate more score of 0.523). Finally, the Flesch Kincaid grade reading level collected in Coh Metrix calculates the Reading Ease Score and converts this to a U.S. grade school level score (grades 0 through 12). Higher scores indicate more complex vocabulary and syntactic structure and, thus, less readable texts while lower scores indicate simpler more readable texts. Counts of specific subtypes of verbs were inclu ded in our analysis to determine specificity and complexity effects in verb use These included casual or causative verbs which are a type of action verb that identify a highly specific cause effect relationship between animate entities. In contrast, use of agentless passive verbs omit agents from sentences and thus provide a low level of information specificity. Syntactic complexity effects were further examined at a lexical level using simple verb s such as 3 rd person
56 singular verbs, and more complex v erb constructions such as verbs appearing as past and present participles Task 1: Picture Description Experiment M aterials Stimulus materials for the picture description experiment consisted of 40 black and white line drawings selected from previous stu dies investigating constrained sentence level production demands (Bock, Loebell, & Morey, 1992; Troche & Altmann, 2012) and the Kempler Comprehension Test (2003). The final selection of stimulus materials was determined by pilot testing on groups of 30 hea lthy older and 30 younger adults who described the 40 pictures included in this project (Wilson & Altmann, submitted). actors [animate entities] in the picture togethe r with an appropriate action) were deemed accurate overall. Using this procedure, 2 separate lists were created, each containing 20 pictures matched for accuracy and response time. Each list consisted of 10 two actor pictures in the simple condition, and 10 three actor pictures in the complex condition. Participants were randomly assigned to different list orders. Pictures were displayed on a standard computer monitor in the single task and on a projector screen in the dual task at a screen resolution s et to 640 x 480 pixels using MediaLab (Jarvis, 2006b) running DirectRT (Jarvis, 2006a) programs. Each picture remains visible for 5500ms after the initiation of a verbal response to reduce working memory demands and record the response. Procedure At the b eginning of the sentence generation task, participants were instructed to produce one sentence that describes the event in the picture and were reminded to
57 include all actors in the picture while avoiding the use of pronouns. Participants were also inform ed that each picture would remain visible for a short while after they began speaking, and were cautioned that they should continue their sentence even if the picture disappeared while they were still speaking. Pictures remained visible until the participa nt responded and for 5500ms after they began speaking. Complex 3 actor and simple 2 actor pictures were presented randomly for each participan t. Each picture trial was initiated manually by the experimenter using a wireless mouse click to ensure the part trial. Scoring Scoring measures are outlined under general scoring above. Additionally, information completeness and grammaticality were coded only for sentences produced in t he Picture Description experiment. These information completeness and grammaticality measures consisted of manual coding for each sentence, as binary measures of whether a response was complete or not and whether the response was grammatical or not. A com plete response was defi ned as a response which included all items of critical information, including all animate entities together with an appropriate action. A grammatical response was defined as a syntactically acceptable response, i.e. a response which did not contain any grammatical error. Task 2: Discourse Production Experiment Materials Stimulus materials for the discourse production experiment consisted of four open ended questions selected from previous studies investigating discourse level producti on demands among younger and older adults (Kemper, Herman, & Lian, 2003; Kemper,
58 McDowd, Pohl, Herman, & Jackson, 2006; Kemper et al., 2009; Kemper & Sumner, the last 100 remember very w Questions were centered on a white background in black Arial font at 24 points. Procedure At the beginning of the discourse task, participants were instructed that they would be provided with a topic which they should talk about for 3 minutes and that they should take time to prepare their response and then indicate when they were ready to begin. The question prompt was then presented to participants and the trial was initiated when the participant signaled they were ready to begin. The prompt remained visible for a further 3 minutes while their response was recorded at which point the screen changed r ecorded as mp3 sound file s Scoring Scoring measures are outlined under the general scoring section above. Q ualitative coding unique to the discourse task included measures of cohesion and coherence. Cohesion refers to the structural unity of discourse p roduction indexed by the use of syntax and lexical structures in text which provide for cohesive ordering of information content (Graesser, McNamara, Louwerse, & Cai 2004) Local cohesion refers to the extent to which lexical structures do or do not (bina ry) reference adjacent sentences and global cohesion refers to the extent to which lexical structures do or do
59 not (binary) reference all sentence combinations. Local and global measures collected from Coh Metrix include noun, argument, stem, content word and anaphor overlap. Higher numbers for all measures of cohesion indicate greater cohesion. Noun overlap refers to the mean number of sentences which contain nouns which match each other ot overlap) locally or global ly. A rgument overlap refers to the mean number of sentences which contain nouns and pronouns (arguments) which match each other locally or globally. S tem overlap refers to the mean tch other content words (nouns, verbs, adjectives, and drawn sente nces either locally or globally. C ontent word overlap refers to the proportion of explicit content words which match each other locally or globally. A naphor overlap refers to the mean number of sentences which contain an anaphor (noun or pronoun) which references another noun or pronoun used in an earlier sentence either locally or globally. Verb cohesion is a global measure of cohesion which references the de gree to which there are overlapping verbs in the text. Coherence refers to the semantic unity of discourse production (calculated using latent semantic analysis), and describes the degree to which sets of words in sentences overlap conceptually within a s hared semantic space (a statistical construct based on co occurrences within a large written word corpus). Scores represent conceptual similarity measured as cosine similarity between adjacent sentences (local coherence) and all other sentence combination s (global coherence) on a scale of 0 to 1. Larger cosine values represent more conceptually similar sentences, hence more coherent discourse and smaller cosine values represent more conceptually distinct sentences, hence less coherent discourse
60 Local c oherence refers to the degree to which adjacent sentences overlap with each other and global coherence refers to the degree to which all sentence combinations overlap with each other within shared semantic space. For example two identical sentences have a cosine value of 1.0, indicating perfect local coherence; two ball bat conceptually distinct boy chased the ball dog chased the car Design and A nalyses Specific Aims 1 and 2 were addressed using a series of two way ANOVAs for each dependent varia ble to assess group level differences and dual task effects in sentence and discourse production. Dependent variables are detailed in Table 3 2. Specific Aim 3 was addressed using a series of hierarchical linear regressions. For each dependent variable sho wing a group main effect or interactions, I performed hierarchical stepwise linear regression analysis using the three extracted cognitive factor scores defined above in the first step, and the group predictor group in the second step. This approach analy zed the relative contribution of cognitive abilities to language performance and sought to determine whether group differences in cognitive abilities were due solely to cognitive abilities or whether the disease also contributed independently to performanc e. Bonferroni adjustment was applied for multiple comparisons showing effects below a critical alpha of .05.
61 Table 3 1. Rotated solution for Factor analysis Measure Processing Speed Working Memory Information Updating Digit Span Forward Accuracy 885 Digit Span Backward Accuracy .872 Stroop XXX Response Time .525 .437 Stroop Color Word Response Time .915 0 Back Response Time .734 1 Back Response Time .756 .218 Digit Symbol Substitution Response Time .811 Star Task Response Time .780 2 Back Accuracy .803 Visual Memory (levels 2 4) Accuracy .299 .716 Operation Span Accuracy .466 .296 .348 Pa Task Number .475 .349 Note: Extraction method Principal Component Analysis
62 Table 3 2 Dependent vari ables collected in language tasks and programs used to extract data. Measure Program Used Quantitative Word Count Coh Metrix Mean Length Of Utterance Coh Metrix Number of Nouns CPIDR & Excel Number of Verbs CPIDR & Excel Number of Adjectives CPIDR & Excel Number of Pronouns CPIDR & Excel Number of Verbs per sentence Coh Metrix Number of words before main Verb Coh Metrix Number of modifiers per noun phrase Coh Metrix Qualitative Propositional Density CPIDR Type Token ratio for verbs Coh Met rix Verb frequency MRC Psycholinguistic database Action Verb Content Coh Metrix Verb hypernymy Coh Metrix Information Completeness* Manual Coding Grammaticality* Manual Coding Cohesion** Coh Metrix Coherence** Coh Metrix P icture description exper iment only ; ** D iscourse production experiment only
63 Reference Stimulus Individual T rials Verbal Response Figure 3 1. Sample stimuli used in the 0 Back task. NO NO YES
64 CHAPTER 4 RESULTS Sentence Production T he purpose of this section is to address specific aims 1 and 2 of this project with respect to sentence production Repeated measures ANOVA s were performed for each quantitative and qualitative dependent variable to assess group level effects and interacti ons wi th task conditions. Significance was set at a critical alpha of .05 and explorations of significant interaction below .05 were Bonferroni corrected for multiple comparisons as shown in Table 4 1 and Table 4 2 Means and standard error s have been r eported for each simple effect and interaction. The dependent measures of grammaticality and information completeness include d an additional level of analysis, testing for effects of picture complexity. Quantitative Measures Overall word count There was a simple effect of task, F ( 1,56)=12.294, p = .001; all participants produced fewer words during dual task (192.582, 7.6) compared to single task (207.517, 7.85). However, t here was no simple effect of group ( p = .759) or interaction ( p = .482). There were no group level differences in word counts resulting from sentence trimming in either single ( p = .741) or dual tasks ( p = .506). S entence length All participants produced fewer words per sentence, F ( 1,58)=9.965, p = .003, during dual task (10.124, .400) compared to single task (10.963, .435). Variability in sentence length (i.e. Standard Deviation) tended to be less during dual task conditions (3.398, .216) compared to single task (3.682, .213) but this task level contrast was
65 not significant ( p = .061). There w as also no simple effect of group ( p = .513; p = .382) or interaction ( p = .382; p = .911) for either measure. N umber and proportion of nouns The number of noun s for all participants, F ( 1,58)=4.741, p = .034, was larger in the dual task (328.333, 8.102) compared to the single task (312.836, 6.625). There was, however, no simple effect of group ( p = .537) or interaction ( p = .249) for this measure. For all participants the proportion of nouns of all types produced tended to increase in the dual task (.401, .00 4) com pared to the single task (.321, .00 7). However this trend did not re ach significance F ( 1,56)=3.286, p = .075 and there was no simple effect of group ( p = .118) or interaction ( p = .131). N umber and proportion of verbs There was no simple effect of group ( p = 167), task ( p = .514) or interaction ( p = .111) for the total number of verbs of all types produced; however three verb forms showed interesting patterns. There was a significant interaction between group and task for the number of verbs using the agentless p assive voice, F ( 1,58)=7.872, p = .007, as shown in Figure 4 1. Post hoc analysis with Bonferroni correction indicated that people with .346) than controls (2.084, .509) only in dual task co nditions, p = .025, and equivalent numbers during single task conditions (PD 1.287, .365; Controls .450, .536), p = .202. However, there were no simple effects of group ( p = .559) or task (.209) for this measure. There was also a tendency for the dual task t o affect the production of verbs in the past participle form differently for each group, such that production of this verb type was more frequent in dual task conditions (3.632, .928) compared to single task conditions (1.684, .500) for controls, p = 034, .647; 1.949,
66 .349), p = .775. However, this interaction only approached significance F ( 1,56)=3.795, p = .056, and there were also no significant simple effects of group ( p = .276) or task ( p = .111). Analysis further revealed that control subjects tended to use a greater proportion of verbs of all types in the dual task (31.007, 5.076) compared to the single task (22.132, between the number of verbs produced in dual (21.279, 3.543) and single tasks (23.115, .742). However, this interaction only approached significance F ( 1,56)=3.194, p = .079, and there was no simple effect of group ( p = .193) or task ( p = .245). With respect to verb sub .008) used a smaller proportion of third person singular verbs than controls (.097, .00 1), F ( 1,56)=4.012, p = .050. There was, however, no simple effect of task ( p = .307) or interaction ( p = .115) for this measu re. There was also a significant three way interaction affecting the proportion of verbs in past participle used in single and dual tasks between groups, F ( 1,56)=5.342, p = .025, as shown in Figure 4 2. Controls used a greater proportion of past participle verbs in dual task (.020, .005) compared to single task (.008, .00 3), p = of past participle verbs used between dual (.009, .00 3) and single task (.010, .00 2), p = .731. In contrast, there were no simple effects for either group ( p = .354) or task ( p = .060) for this measure. N umber and proportion of adjectives As predicted, the number of adjective s produced disease (16.410, .243) was lower overa ll than for controls (25.735, 3.559), F ( 1,58)=4.690, p = .034. There was, however, no simple effect of task ( p = .258) or
67 interaction ( p = produce a smaller proportion of adjectives of all type s (.015, .003) than controls (.022, .004). However this trend was not significant, F ( 1,56)=3.188, p = .080, and there was no simple effect of task ( p = .605) or interaction ( p = .541). N umber and proportion of pronouns There w ere no simple effect s of group, t ask or interaction for the number and proportion of pronouns of all types produced; however, particular subtypes of pronouns .330) produced fewer possessive pronouns than controls (2.895, 0.472) overall, F ( 1,56)=5.336, p = .025. There was, however, no simple effect of task ( p = .961) or interaction ( p = .326) also tended to produce fewer wh pronouns than controls (2.526, .467), but this trend was not quite significant, F ( 1,56)=3.838, p = .055, and there was no effect of task ( p = .825) or interaction ( p = .162) also tended to produce a smaller proportion of Wh pronouns tha n controls (1.223, .232). However this trend again only approached significance F ( 1,56)=3.934, p = .052, and there was no simple effect of task ( p = .734) or interaction ( p = .301). N umber of verbs per sentence All participants produced fewer verbs per senten ce in the dual task (2.156, .099) compared to the single task (2.333, .094), F ( 1,56)=5.024, p = .029. However, t here was no simple effect of group ( p = .464) or interaction ( p = .300).
68 N umber of prepositions There was no simple effect of group ( p = .170; p = .269 ), task ( p = .494; p = .930) or interaction ( p = .610; p = .298) for the number of prepositions or prepositional density produced by participants. N umber of words before main verb There was no simple effect of group ( p = .166), task ( p = .501), or interaction ( p = .356) on the mean number of words produced before main verbs. N umber of modifiers for each noun phrase There was no simple effect of group ( p = .416), task ( p = .560), or interaction ( p = .109) on the number of number of modifiers for each noun phrase. Summary The pr eceding analyses have revealed pervasive simple effects of task associated with reduced word counts, fewer words per sentence, and fewer verbs used in sentences in dual task conditions for all participants. At the level of word class, the number of nouns increased in dual task conditions. In contrast, on the dual task did not affect the number or proportion of verbs, adjectives, pronouns or prepositions produced. singu lar verbs, adjectives and possessive pronouns overall than controls. People with use of participle verbs in dual task by control subjects and produced fewer passive ve rbs than controls, but only in dual task conditions. There were, however no group level effects on the number or proportion of nouns, adjectives or prepositions produced.
69 Qualitative Measures Propositional density All participants produced fewer proposit ions during the dual task (65.355, 3.832) compared to the single task (72.804, 3.972), F ( 1,56)=10.078, p = .002. In contrast, there was no simple effect of group ( p = .768) or interaction ( p = .136) for this measure. There was also no simple effect of group ( p = .562), task ( p = .119) or interaction ( p = .277) for propositional density. Concreteness and word frequency All participants used more concrete words of all types during dual task (5.065, .168) compared to single task conditions (4.774, .169), F ( 1,58)=6.5 51, p = .013. There was, however, no simple effect of group ( p = .991) or interaction ( p = .396) for this measure. Similarly, concreteness of content words (i.e. nouns, verbs and adjectives), F ( 1,58)=6.061, p = .017, increased in dual task (537.185, 5.370) comp ared to single task conditions (529.541, 5.509) for all participants. There was, however, no simple effect of group ( p = .797) or interaction ( p = .509) for this measure. Similarly, there was also a trend for increased imageability in dual task conditions ( 554.634, 4.789) compared to the single task (549.540, 5.016) but this was only marginally significant, F (1,58)=3.254, p =.076. Both groups also used lower frequency content words in dual task (2.145, .020) compared to single task conditions (2.194, .01 9), F ( 1,58)=5.565, p = .022, and there was no significant simple effect of group ( p = .676) or interaction ( p = .373). Syntactic complexity All participants produced syntactically less complex sentences during dual task (1.337, .159) compared to single task con ditions (1.094, .147), F ( 1,58)=5.787,
70 p = .019. There was, however, no simple effect of group ( p = .645) or interaction ( p = .487). Similarly was lower in dual task conditions (2 .940, .202) compared to single task conditions (3.314, .194), F ( 1,58)=8.184, p = .006. There was, however, no simple effect of group ( p = .394) or interaction ( p = .877). Type token ratio All participants tended to produce a greater number of lexically divers e words overall (higher type token ratios) in dual task (.381, .010) compared to single task conditions (.363, .010). However, this trend was not significant, F ( 1,58)=3.787, p = .056, and neither the simple effect of group ( p = .832) nor the interaction ter m ( p = .662) was significant for this measure. There was also no simple effect of group ( p = .685), task ( p = .135), or interaction ( p = .341) when type token ratio was limited to content word lemmas. Action Verb Content As predicted, all participants produced a higher number of action verbs in dual task (74.910, 4.302) compared to single task conditions (67.989, 3.905), F ( 1,58)=4.342, p = .042. There were also marginal group differences in the effects of dual task on the number of action verb s produced. Control s tended to increase the number of action verb content under dual task conditions (76.980, 7.112) compared to the single task (63.674, 6.456), p = to be impervious to dual task effects (72.841, 4.841; 72. 304, 4.395), p = .886. However, this interaction only showed a trend toward significance F ( 1,58)=3.695, p = .059, and there was no simple effect of group ( p = .766) for this measure.
71 Content word hypernymy All participants produced more specific nouns and verb s (larger noun and verb hypernymy) in dual task conditions (2.503, .053) compared to single task (2.397, .045), F ( 1,58)=5.540, p = .022. There was, however, no simple effect of group ( p = .466) and no significant interaction ( p = .344). In contrast, hypernym measures for nouns and verbs separately were not significant. There was also no simple effect of group ( p = .864; p = .554), task ( p = .605; p = .192), or interaction ( p = .246; p = .602) for either measure of verb or noun hypernymy. Information completeness There w as a simple effect of picture complexity on information completeness, F ( 1,56)=56.389, p = .0001. All participants provided more complete information describing simple (.737, .022) compared to complex pictures (.567, .737). Unexpectedly, people with Parki more informative sentences (.713, .030) than controls (.592, .043), F ( 1,56)=5.460, p = .023. These simple effects were secondary to a significant group interaction with task and picture complexity, F ( 1,56)=4.871, p = .031, as sh own in Figure 4 3. Post hoc analysis using Bonferroni correction revealed that group level performance advantages for people with p = .020) and complex pictures ( p = .007) were significant in the single task, and eliminate d in the dual task for simple ( p = .068) and complex pictures ( p = produced more informative descriptions than controls, both for simple (PD .800, .029; Controls .679, .041) and complex pictures (PD .646, .043; Controls .437, .061) In contrast, there were no group level differences in dual task conditions, either for simple
72 (PD .787, .032; Controls .684, .045) or complex pictures (PD .619, .044; Controls .567, .063). Grammaticality There was a simple effect of picture complexity for grammaticality, F ( 1,56)=4.727, p = .034. All participants produced fewer grammatical errors (larger accuracy scores) describing simple (.783, .031) compared to complex pictures (.755, .032). This simple effect was secondary to a significant group interaction with picture complexity, F ( 1,56)=5.322, p = .025, as shown in Figure 4 4. Post hoc analysis using Bonferroni errors describing c omplex pictures (.730, .037) compared to simple pictures (.790, .035), p = .0001, in contrast, there were no differences in control performance between complex (.779, .053) and simple picture descriptions (.777, .050) for this measure, p = .936. There was however, no simple effect of group ( p = .774), task ( p = .817), or interaction between group and task ( p = .992), between task and complexity ( p = .472) or between group, task and complexity ( p = .855) for grammaticality. Summary Pervasive simple effects of task w ere revealed that were associated with reduced propositional density, decreased syntactic complexity and lower grade reading level scores. Action verb use increased in dual task for all participants and both nouns and verbs were more specific under dual t ask conditions. Type token ratios also tended to be lower in the dual task, although this effect of task was not significant ( p = .056). In contrast, there was no task level effect for verb frequency and group level differences reported were limited to inf ormation completeness only. All participants produced more
73 disease provided more complete information than controls in single task conditions. There were, however no reported group level differences for propositional density, syntactic complexity, type token ratio, verb frequency, action verb content, or for noun or verb hypernymy. Discourse Production The purpose of this analysis is to address specific aims 1 and 2 of this dissertation project. The same procedure detailed above was performed for each quantitative and qualitative dependent variable collected in the discourse production task as shown in Table 4 3 and Table 4 4 In contrast to the sentence producti on task, separate analysis and measures for information completeness and grammaticality were not collected or performed. However, measures of coherence and cohesion were added. Quantitative Measures Overall word count There was no simple effect of group ( p = .113), task ( p = .609), and no interaction ( p = .251) for overall word counts. There was also no group level difference in word counts resulting from trimming discourse ( p = .161) in either task ( p = .868). S entence length e (14.201, .428) produced fewer words per sentence, F ( 1,54)=11.344, p = .001, with decreased variability (8.236, .260) in sentence length (i.e. Standard Deviation), F ( 1,54)=7.497, p = .008, compared to controls (16.675, .597; 9.457, .362). There was howev er no effect of task ( p = .297; p = .103) and no interaction ( p = .621; p = .842) for either mean sentence length or sentence variability.
74 N umber and proportion of nouns The number of noun phrases produced by all participants increased significantly in the dual ta sk (378.761, 4.772) compared to the single task (364.296, 4.316), F ( 1,54)=5.588, p = .022. However, there was no simple effect of group ( p = .673) and no interaction with task ( p = disease produced fewer plural nouns (16.014, 1.019), F ( 1,54)=6.043, p = .017, and more proper nouns (13.216, .954) than controls (20.316, 1.423; 9.342, 1.331), F ( 1,54)=5.596, p = .022. However, there was no simple effect of task ( p = .155; p = .387) and no interaction ( p = .91 8; p = .171) for either measure. There was also no simple effect of group ( p = .517), task ( p = .604), and no interaction ( p = .374) for the number of nouns of all types produced. The total proportion of noun phrases produced showed no significant simple effect of group ( p = .241), task ( p = .977) or interaction ( p = .705). .246) produced a higher proportion of proper nouns than controls (2.495, .344) F ( 1,54)=9.547, p = .003. However there was no simple effect of tas k ( p = .569) and no interaction ( p = .167) for this measure. N umber and proportion of verbs The number of verbs produced showed no simple effects of group ( p = .179), task ( p = .924), and no interaction ( p = .396). In contrast, both participant groups produced fewe r verbs using past and present participles in dual task (16.885, .945) compared to the single task (20.170, 1.122), F ( 1,54)=6.029, p = .017. However, there was no simple effect of group ( p = .238) and no interaction ( p = .651) for this measure. Additionally, 1.303) used a higher number of ca us al verbs than controls (25.119, 1.819), F ( 1,54)=4.430, p = .040. There was however no simple effect of task ( p = .164) and no interaction ( p = .331) for this measure. Both particip ant
75 groups produced a smaller proportion of verbs in present and past participle during dual task conditions ( .035 003 ) compared to single task conditions ( .042 00 2), F ( 1,54)=4.667, p = .035. However there was no simple effect of group ( p = 983) and no interaction ( p = .802) for this measure. There was also no simple effect of group ( p = .676), task ( p = .497) and no interaction ( p = .800) for the proportion of all verbs produced. N umber and proportion of adjectives As predicted, the number of adjective s for pe (16.410, .243) was lower overall than for controls (25.735, 3.559), F ( 1,58)=4.690, p = .034. There was, however, no simple effect of task ( p = .258) or interaction ( p = .838) ded to produce a smaller proportion of adjectives of all types (.015, .003) than controls (.022, .004). However this trend was not significant, F ( 1,56)=3.188, p = .080, and there was no simple effect of task ( p = .605) or interaction ( p = .541). There was no simple effect of group ( p = .245), task ( p = .920) and no interaction ( p = .317) for the proportion of all adjectives produced. N umber and proportion of pronouns The total number of pronouns showed no significant simple effects of group ( p = .159), task ( p = .258), and no interaction ( p = .282). Similarly, there was no simple effect of group ( p =.732), task ( p =.193) and no interaction ( p =.294) for the proportion of pronouns of all types produced by participants. .201) produced fewer wh pronouns than controls (2.711, .280), F ( 1,54)=6.831, p = .012. However, there was no simple effect of task ( p = .653) and no interaction ( p = .055) tended to produce a sma ller proportion of Wh pronouns than controls (.700, .076).
76 However, this trend was not significant, F ( 1,54)=3.250, p = .077, and there was no simple effect of task ( p = .469) and no interaction ( p = .554). N umber of verbs per sentence isease (2.762, .078) produced fewer verbs per sentence than controls (3.248, .109), F ( 1,54)=13.025, p = .001. There was however no simple effect of task ( p = .257) and no interaction ( p = .488) for this measure. N umber of prepositions disease (39.932, 1.911) produced fewer prepositions than controls (46.947, 2.666), F ( 1,54)=4.574, p = .037. However, there was no simple effect of task ( p = .870) and no interaction ( p = .499) for this measure. The number of prepositional phrases produced wa (97.368, 2.682) compared to controls (107.236, 3.743), F ( 1,54)=4.592, p = .037, while there was no simple effect of task ( p = .742) and no interaction ( p = .527). N umber of words before main verb There was no s imple effect of group ( p = .460), task ( p = .777), and no interaction ( p = .975) on the number of words produced before main verbs. N umber of modifiers for each noun phrase There was no simple effect of group ( p = .410), task ( p = .420), and no interaction ( p = .221) on the number of number of modifiers for each noun phrase. Summary In contrast to findings for sentence production pervasive group level effects were revealed for discourse. Simple effects of group were associated with reduced sentence length and fewer v compared to controls. At a lexical level
77 more proper nouns, more caus al verbs but fewer plural nouns, fewer adjectives overall, fewer wh pronouns, fe wer prepositions and fewer prepositional phrases than controls. In contrast, there were no group level differences in word counts, or the number and proportion of verbs and pronouns produced overall. Additionally the groups did not differ significantly in the number of words produced before main verbs, and the number of modifiers produced for each noun phrase. Task level differences are also reported. The number of noun phrases produced decreased in dual task conditions for all participants. Both parti cipant groups also used fewer verbs in participle form in dual task conditions. In contrast, task level effects were not reported for word counts, sentence length, number and proportion of nouns, verbs, pronouns, adjectives overall. There was also no sig nificant effect of task for either the number of words used before main verbs or the number of modifiers used for each noun phrase. Qualitative Measures Propositional density (.465, .005) than controls (.492, .007), F ( 1,54)=10.414, p = .003. However, there was no simple effect of task ( p = .542) or interaction ( p = .374) for this measures. Similarly, the e (160.135, 7.020) than controls (185.921, 9.796), F ( 1,54)=4.578, p = .037. However, there was no simple effect of task ( p = .857) or interaction ( p = .113) and for this measure. Concreteness and imageability Words of all types became less concrete in dual ta sk conditions (.085, .109) compared to single task (.385, .133), F ( 1,54)=4.191, p = .046. However, there was no simple effect of group ( p = .572) or interaction ( p = .658) for this measure. In contrast for
78 all participants, under dual task conditions conten t words became more concrete (370.301, 3.603) compared to single task conditions (360.229, 3.597) F ( 1,54)=5.722, p = .020, and more imageable (404.987, 3.299) compared to single task conditions (394.941, 3.644) F ( 1,54)=5.481 p = .023 There were no oth er significant effects in these analyses. Syntactic complexity .075) compared to controls ( .301, .104), F ( 1,54)=7.871, p = .007. There was however no simple effect of task ( p = .623) and no significant interaction with group ( p = .694). Similarly, grade reading level of discourse produced by people with .256) was lower than controls (7.247, .357), F ( 1,54)=5.545, p = .022. There was, however, no simple effect of task ( p = .394) or interaction (.999) for grade reading level. Type token ratio Neither the analysis of type token ratio for all words or for content word lemmas revealed any significant effects. There was no simple effect of group ( p = .200; p = .32 8, respectively), task ( p = .246; p = .057) or interaction ( p = .611; p = .280). Verb frequency There was no simple effect of group ( p = .463), task ( p = .637), and no interaction ( p = .999) on spoken word frequency for verbs. Action Verb Content There was no simple eff ect of group ( p = .118), task ( p = .077), and no interaction ( p = .827) on the number of action verbs produced.
79 Content word hypernymy Both participant groups produced more specific verbs under dual task conditions (1.479, .021) compared to single task (1.426, .022), F ( 1,54)=4.005, p = .050, but there was no simple effect of group ( p = .449) or interaction ( p = .620) for verb hypernymy. In .057) produced less specific nouns than controls (6.525, .080), F ( 1,54)=4.71 7, p = .034. There was, however, no simple effect of task ( p = .633) and no interaction ( p = .515) for noun hypernymy. Cohesion .014) than controls (.473, .020), F ( 1,54)=8.282, p = .006, u sed fewer referring anaphors both locally (.533, .020) than controls (.619, .027 ), F ( 1,54)=6.558, p = .013, and globally (.260, .015) than controls ( .336, .021 ), F ( 91,54)=8.830, p = .004, and used fewer temporal connective words (15.405, .915) than contro ls ( 18.614, 1.277 ), F ( 1,54)=4.170, p = .046. disease produced sentences with more syntactic similarity locally (PD=.123, .005; HOA=.103, .007), F (1,54)=5.537, p =.022, and globally (PD = .108, .003; HOA=.095, .005), F (1,54)=5.043, p =.029. produced less variability in their sentence structures than controls. There was however no simple effect of task and no interaction with group in either analysi s For both groups, the dual task (.932, .109) resulted in reduced verb cohesion compared to the single task (1.274, .105), F ( 1,54)=5.805, p = .019, while there was no simple effect of group ( p = .216) and no interaction ( p = .098). No other measure s of cohe sion were significant.
80 Coherence There was a significant group interaction with task for local coherence, F ( 1,54)=4.145, p = .047, as shown in Figure 4 5. Post hoc analysis using Bonferroni correction revealed that local coherence was equivalent between sing le (.142, .008) p = .418. In contrast, controls tended to be more coherent in single task (.177, .012) than in dual task (.150, .011) conditions however, this contrast was not significant, p = .060. Thus, people with .008) were less locally coherent than controls (.177, .012), p = .020, only in the single task but there were no group level differences between .008) and contr ols (.150, .011) in dual task conditions, p = .974. There was also no simple effect of either group ( p = .137) or task ( p = .283) on this single measure of local coherence. Global coherence, by contrast, was unaffected by group ( p = .115), task ( p = .564), and th ere was no significant interaction ( p = .261). Summary Pervasive simple effects of group on discourse were revealed that were associated with a reduced proportion and number of propositions produced by people S yntactic complexity and grade reading and less locally and globally cohesive ties which were also less locally coherent. whose syntax was more similar between sentences both locally and globally. disease also produced less specific nouns than controls. In contrast, no group level effects were reported for global coherence, type t oken ratio, verb frequency, action verb
81 content, and hypernymy. There were also some effects of the dual task reported B oth groups of participants used more specific verbs under dual conditions and verb cohesion was generally reduced in the dual task com pared to the single task. In contrast, there were no reported task level effects for propositional density, syntactic complexity, type token ratio, verb frequency, action verb content and coherence. Regression Analyses for Sentence Production The purpose of this analysis is to address specific aim 3 of this dissertation project. Analysis consisted of a series of hierarchical linear regressions for each dependent variable showing a simple effect of group or interaction with group Cognitive factors were en tered stepwise in the first level of the regression followed by forced entry of group in the subsequent step. The predictors of performance for sentence production are displayed in Table 4 5 and predictors of performance for discourse are presented in Ta ble 4 6. Quantitative Measures There was no significant predictor of the number of agentless passive voice verbs used in single task conditions. In contrast, individual differences in the number of agentless passive voice verbs used in dual task condition s was predicted only by the group factor accounting for 8.3% of variance in scores ( R 2 = .083, F ( 1,59)=5.263, p = .025). Individual differences in the number of past participle verbs used in single task conditions was predicted only by the working memo ry factor accounting for 7.7% of score variance ( R 2 = .077, F ( 1,57)=4.787, p = .033). The influence of the group factor was not significant in improving the model fit, p = .557. Individual differences in the number of past participle verbs used in dual task conditions was predicted by two
82 factors accounting for 14.0% of score variance (R 2 = .140, F ( 1,56)=5.195, p = .026). The working memory factor ( R 2 = .093, .305, p = .019), and the processing speed factor (R 2 = .277, p = .026) predicted the number of past participles produced in dual task conditions. The influence of the group factor was not significant in improving the model fit, p = .729. There was no significant predictor of number of adjective s in single task conditions. In contrast, individual dif ferences in the number of adjectives produced in dual task conditions was predicted only by the group factor accounting for 7.5% of score variance ( R 2 = .075, F ( 1,58)=4.670, p = .035). There was no significant predictor of the number of wh pronouns produced in single task conditions. However, there was a marginal effect of group ( R 2 = .061, F ( 1,57)=3.681, p = .060) accounting for 6.1% of sco re variance in single task conditions. In contrast, individual differences in the number of wh pronouns produced in dual task conditions was predicted only by the updating factor accounting for 7.8% of score variance ( R 2 = .078, F ( 1,57)=4.818, p = .03 2) while t he influence of the group factor was not significant in improving the model fit, p = .828. There was no significant predictor of the proportion of wh pronouns produced in single task conditions. In contrast, individual differences in the proporti on of wh pronouns produced in dual task conditions was predicted only by the updating factor accounting for 9.3% of score variance ( R 2 = .093, F ( 1,57)=5.817, p = .019) while t he influence of the group factor was not significant in improving the model fit, p = .584.
83 T here were no significant cognitive or group predictors of the proportion all verbs, the proportion of 3 rd person singul ar verbs, or the number of possessive pronouns produced by participants. Summary Processing speed, working memory and information updating predicted group level differences in quantitative sentence production. Slower processing and poorer working memory predicted reduced use of past participle verbs in dual task conditions, while only poorer working memory predicted differences in single task conditions. B oth the number and proportion of wh pronouns under dual task conditions were predicted by the updati ng factor. Group differences not accounted for by our measures of cognitive processing included fewer passive verbs and fewer adjectives and wh pronouns in single task conditions. In contrast, there were no signi ficant cognitive or group predictors for o verall word count, mean sentence length, mean sentence variability, number of noun s noun proportion, verb proportion, singular verb proportions, number of possessive pronouns, or verbs per sentence in either single of dual task conditions. There were als o no significant predictors for number of adjective s or wh pronoun proportions in single task conditions. Qualitative Measures There was no significant cognitive predictor of information completeness for either single or dual task conditions and simple and complex pictures. However, in single task conditions group differences accounted for 8.3% of score variance ( .314, R 2 = .083, F ( 1,58)=6.237, p = .015) for simple pictures and 13.0% of score variance ( .360, R 2 = .130, F ( 1,58)=8.494, p = .005) for complex pictures. In contrast, there were no cognitive predictors or group level effects affecting information completen ess in dual task conditions for either simple or complex pictures.
84 Individual differences for grammaticality in single task conditions were predicted by two factors for simple pictures accounting for 29.3% of score variance, (R 2 = .293, F ( 1,56)=11.583, p = .00 01), and by three factors for complex pictures accounting for 33.5% of score variance, (R 2 = .335, F ( 1,55)=9.231, p = .017). The processing speed factor ( R 2 = .230 .480 p = .0001), and the working memory factor (R 2 = p = .030) predicted the grammaticality of sentences describing simple pictures in single task conditions. The influence of the group factor again was not significant in improving the model f it, p = .187. The processing speed factor ( R 2 = .149, .386, p = .003), the working memory factor (R 2 = p = .005), and the updating factor (R 2 = .074, p = .005) predicted the grammaticality of sentences describing complex pictures in single task conditions. The influence of the group factor was not significant in improving the model fit, p = .433. Individual differences for grammaticality in dual task conditions were predicted only by the processing speed factor which accounted for 6.8% of score variance for simple pictures ( R 2 =.068, .260, F ( 1,57)=4.132, p = .047) and 10.8% of score variance for complex pictures ( R 2 =.108, .329, F ( 1,57)=6.917, p = .011). The influence of the group factor was not significant in improving the model fit for either simple ( p = .513) or complex pictures ( p = .956) in dual task conditions. T here were no significant cognitive or group level predictors of the number of action verb s produced Summary There were no cognitive or group level predictors for number of prepositions, concreteness overall, concreteness for content words, written word frequency for content words, syntactic complexity, grade reading level, type token ratio, number of action verb s and hypernymy for either single or dual task conditions.
85 However, information completeness in single task con ditions for both simple and com plex pictures was predicted by the group factor There were, however no cognitive or group predictors in dual task conditions for this measure. Slower processing speed predicted fewer grammatical sentences at all levels of picture complexity both during single and dual task conditions. However, during the single task the production of fewer grammatical sentences was also associated with poorer working memory both when participants described simple and complex pictures. P oorer updating performance also predicted use of fewer grammatical sentences but only when participants described complex pictures in dual task conditions. Regression Analyses for Discourse Production The same regression procedure reported above for sente nce production was applied to discourse measures to address specific aim 3 of this project as shown in Table 4 7 and Table 4 8 Quantitative Measures Individual differences in mean sentence length used in single task conditions was predicted only by the g roup factor accounting for 14.4% of score variance ( R 2 =.144, F ( 1,55)=9.260, p = .004). Similarly, individual differences in mean sentence length used in dual task conditions was predicted only by the group factor accounting for 9.4% of score varianc e ( R 2 =.094, F ( 1,56)=5.838, p = .019). There was no significant predictor of sentence length variability used in single task conditions. However, individual differences in sentence length variability in dual task conditions was predicted only by the group factor accounting for 8.3% of score variance ( R 2 =.083, F ( 1,56)=5.092, p = .028).
86 There was no significant predictor of plural noun counts used in single task conditions. However, individual differences in plural noun counts used in dual task c onditions was predicted by two factors accounting for 27.7% of score variance (R 2 = .277, F ( 1,55)=10.820, p = .0001). The processing speed factor (R 2 = .418, p = .001), and the updating factor (R 2 = p = .007) predicted the number of plural noun s produced in dual task conditions. The influence of the group factor was not significant in improving the model fit, p = .507. Individual differences in proper noun counts used in single task conditions was predicted only by the updating factor accounting for 7.7% of score variance ( R 2 =.077, .278, F ( 1,55)=4.620, p = .036). The influence of the group factor was not significant in improving the model fit, p = .081. While there was no significant predictor of proper noun counts used in dual task conditions. Individual differences in proper noun proportions used in single task conditions was predicted only by the updating factor .406, R 2 = .164, F ( 1,55)=10.828, p = .002). The influence of the group factor was not significant in improving the model fit, p = .334 There were no significant predictor s of proper noun proportions used in dual task conditions. Similarly, t here were no significant cognitive or group predictors for number of causal verb s or number of adjective s in either the single or the dual task. Th ere was no significant predictor of wh pronoun counts u sed in single task conditions. In contrast, i ndividual differences in wh pronoun counts in dual task conditions was predicted by the group factor accounting for 8.7% of score variance ( R 2 =.087, F ( 1,56)=5.305, p = .025). There were no significant cognitive or group level predictors for wh pronoun proportions in single task conditions. However,
87 although 2 = .056) in dual task cond itions this was not significant F ( 1,56)=3.316, p = .074. Individual differences in number of prepositions us ed in single task conditions were predicted only by the updating factor accounting for 10.2% of score variance ( R 2 =.102, F ( 1,55)=6.251, p = .01 5). The influence of the group factor was not significant in improving the model fit, p = .249. Individual differences in the number of prepositions used in dual task conditions was predicted by two factors accounting for 21.4% of score variance (R 2 = .214, F ( 1,55)=7.507, p = .0001). The updating factor (R 2 = .382, p = .003), and the processing speed factor (R 2 = .261, p = .033) predicted the number of prepositions produced in dual task conditions. The influence of the group factor was not significant in improving the model fit, p = .382. Individual differences in number of verbs per sentence used in single task conditions was predicted only by the group factor accounting for 15.0% of score variance ( R 2 =.150, F ( 1,55)=9.686, p = .003). Similarly, individual differences in number of verbs per sen tence used in dual task conditions was predicted only by the group factor accounting for 7.0% of score variance ( R 2 =.070, F ( 1,56)=4.200, p = .045). There were no significant predictor s of syntactic complexity used in single task conditions. Summary Processing speed and information updating but not working memory predicted group level differences in quantitative discourse production. Slower processing speed and poorer updating scores predicted a reduced number of plural nouns and prepositions in dual conditions only. Those individuals with poorer updating skills produced fewer prepositions in both single and dual task conditions and more
88 proper nouns i n single task conditions only. Group differences not accounted for by measures of cognitive process ing included: sentence length in single task, and sentence length, variability and number of verbs produced per sentence in both single and dual task conditions. The number of wh pronouns produced was also predicted by group membership but only in single task conditions. Qualitative Measures Individual differences in proposition density used in single task conditions was predicted only by the processing speed factor accounting for 12.4% of score variance ( R 2 =.124, .352, F ( 1,55)=7.784, p = .007). The influence of the group factor was not significant in improving the model fit, p = .489. Individual differences in proposition density used in dual task conditions was predicted only by the processing speed factor account ing for 19.3% of score variance ( R 2 =.193, F ( 1,56)=13.404, p = .001). The influence of the group factor ( R 2 = .054) was not significant in improving the model fit, p = .052. However, individual differences in the number of propositions used in si ngle task conditions was predicted by three factors accounting for 38.2% of score variance (R 2 = .382, F ( 1,53)=10.913, p = .0001). The processing speed factor ( R 2 = .151, .389, p = .003), and the working memory factor (R 2 = p = .001), and the updatin g factor ( R 2 = .072, .269, p = .016) predicted the number of propositions produced in single task conditions. The influence of the group factor was not significant in improving the model fit, p = .695. Similarly individual differences in the number of propo sitions used in dual task conditions was predicted by two factors accounting for 26.7% of score variance (R 2 = .267, F ( 1,55)=10.041, p = .0001). The updating factor ( R 2 = .146, p = .003), and the processing speed factor (R 2 = .348, p = .004)
89 predicte d the number of number of propositions produced in dual task conditions. The influence of the group factor was not significant in improving the model fit, p = .188. Individual differences in syntactic complexity used in dual task conditions was predicted on ly by the group factor accounting for 10.5% of score variance ( R 2 =.324, .324, F ( 1,56)=6.572, p = .013). There were no significant predictor of reading grade level in single task conditions although the influence of group ( R 2 =.058, ) approached significance in dual task conditions F ( 1,56)=3.479, p = .067. There was no si gnificant predictor of noun hypernymy used in single task, however the influence of group ( R 2 = .059) was marginally significant, F ( 1,55)=3.476, p = .068). There was however, no significant predictor of noun hypernymy used in dual task conditions. I n dividual differences in argument cohesion in single task conditions was predicted by two factors accounting for 20.4% of score variance (R 2 = .204, F ( 1,54)=6.905, p = .002). The processing speed factor ( R 2 = .140, .374 p = .004), and the updating factor (R 2 = p = .042) predicted argument cohesion in single task conditions. The influence of the group factor was not significant in improving the model fit, p = .129. There was however, no significant predictor of argument cohesion in dual task conditions Individual differences in local anaphor cohesion used in single task conditions was predicted only by the updating factor accounting for 8.2% of score variance ( R 2 =.082, F ( 1,55)=4.881, p = .031). The influence of the group factor was not significa nt in improving the model fit, p = .550. Individual differences in local anaphor cohesion used in
90 dual task conditions was predicted only by the group factor accounting for 9.2% of score variance ( R 2 =.092, F ( 1,56)=5.706, p = .020). Individual differenc es in global anaphor cohesion used in single task conditions was predicted only by the updating factor accounting for 10.9% of score variance ( R 2 =.109, F ( 1,55)=6.734, p = .012). The influence of the group factor was not significant in improving the model fit, p = .289. Individual differences in global anaphor cohesion used in dual task conditions was predicted only by the group factor accounting for 7.5% of score variance ( R 2 =.075, F ( 1,56)=4.516, p = .038). There was no significant cognitive or group level predictor of number of temporal connective word s used in single task conditions Individual differences in the number of temporal connective word s produced in dual task conditions was predicted only by the updating factor accounting for 7.7% of score variance ( R 2 =.077, F ( 1,56)=4.691, p = .035). The influence of the group factor was not significant in improving the model fit, p = .716. There was no significant cognitive or group level predictor of local syntactic similarity in single task. Individual differences in l ocal syntactic similarity in dual task conditions was predicted only by the group factor accounting for 11.7% of score variance ( R 2 =.117, .343, F ( 1,56)=7.455, p = .008). There was no significant cognitive or group level predictor of global syntactic simil arity in single task. Individual differences in global syntactic similarity in dual task conditions was predicted only by the group factor accounting for 7.5% of score variance ( R 2 =.075, .275, F ( 1,56)=4.572, p = .037). Individual differences in local cohe rence in single task conditions was predicted only by the group factor accounting for 9.9% of score variance ( R 2 =.099,
91 F ( 1,55)=6.045, p = .017). In contrast, there was no significant cognitive or group level predictor of local coherence in dual task conditions. Summary Processing speed, working memory and information updating predicted several group level differences in qualitative discourse production. Slower processing speeds predicted reductions in the number and density of propositions, and redu ced argument cohesion in the single task Slower processing speed was also associated with individuals producing less propositional information in dual task conditions. Poorer working memory in contrast only predicted the production of fewer propositio ns only under single task conditions. Poorer updating skills in contrast, had pervasive effects on qualitative measures. In particular poorer updating was associated with reductions in propositional density during dual task, number of propositions both in single and dual task, argument cohesion in single task, local anaphor cohesion in both single and dual task conditions, global anaphor cohesion only in single task conditions and fewer connective words during dual task conditions only. Group level memb ership predicted propositional density, syntactic complexity, local anaphor cohesion, global anaphor cohesion, and both global and local syntactic similarity only under dual task conditions. Group membership also predi cted local coherence during single task conditions.
92 Table 4 1 Simple effects and interaction for quantitative measures of sentence production showing significance and marginal significance. Measure Group Effect Dual Task Effe ct Interaction Total word count p=.001^ Sentence length p=.003^ Sentence variability p=.061^ Noun count p=.034 Noun proportion p=.075 Agentless passive verb count p=.007 Past participle verb count p=.056 Verb proportion p=.079 Verb 3rd person sing. proportion p=.050* p=.025 Adjective count p=.034* Possessive pronoun count p=.025* Wh pronoun count p=.055* Wh pronoun proportion p=.052* Verbs per sentence p=.029^ Note: indicates PD scored lower on t his measure than controls; ^ denotes dual task interference Table 4 2. Simple effects and interaction for qualitative measures of sentence production showing significance and marginal significance. Measure Group Effect Dual Task Effect Interaction Prep osition count p=.002^ Concreteness all words p=.013 Concreteness content words p=.017 Word frequency content words p=.022^ Syntactic complexity p=.019 ^ Grade reading level p=.006^ Type token ratio all words p=.056 Action verb count p=.042 p=.059 Hypernymy nouns and verbs p=.022 Information completeness ** p=.031 Grammaticality ** p=.025 Note: indicates PD scored lower on this measure than controls; ^ denotes dual task interference. ** Analysis of this dependen t variable included the independent variable picture complexity.
93 Table 4 3. Simple effects and interaction for quantitative measures of discourse production showing significance and marginal significance. Measure Group Effect Dual Task Effect Interactio n Sentence length p=.001* Sentence length variability p=.008* Noun phrase count p=.022 Plural noun count p=.017* Proper noun count p=.022 Proper noun proportion p=.003 Participle verb count p=.017^ Participle verb proporti on p=.035^ Causal verb count p=.040 Adjective count p=.034* Wh pronoun count p=.012* Wh pronoun proportion p=.077* Prepositions count p=.037* Verbs per sentence p=.001* Note: indicates PD scored lower on this measure than controls; ^ denotes dual task interference. Table 4 4. Simple effects and interaction for qualitative measures of discourse production showing significance and marginal significance. Measure Group Effect Dual Task Effect Interaction Proposition density p=.003* Proposition count p=.037* Concreteness all words p=.046^ Concreteness content words p=.020 Imageability content words p=.023 Syntactic complexity p=.007* Grade reading level p=.022* Type token ratio content words p=. 057 Action verb count p=.077 Hypernymy verb p=.050 Hypernymy noun p=.034* Argument cohesion p=.006* Local anaphor cohesion p=.013* Global anaphor cohesion p=.004* Temporal connective word count p=.046* Local syntactic similar ity p=.022 Global syntactic similarity p=.029 Verb cohesion p=.019 ^ Local coherence p=.047 Note: indicates PD scored lower on this measure than controls; ^ denotes dual task interference.
94 Table 4 5. Predictors for significant and marginal (p=<.055) effects for quantitative measures of sentence production Measure Processing Speed Working Memory Updating Group Beta R 2 Beta R 2 Beta R 2 Beta R 2 Agentless passive voice verb dual task 0.288 0.083 Past participle verb count single task 0.278 0.077 Past participle verb count dual task 0.277 0.077 0.305 0.093 Adjective count dual task 0.273 0.075 Wh pronoun count single task 0.246 0.061 Wh pronoun count dual task 0.279 0.078 Wh pronoun proport ion dual task 0.304 0.093 Table 4 6. Predictors for significant and marginal (p=<.055) effects for qualitative measures of sentence production Measure Processing Speed Working Memory Updating Group Beta R 2 Beta R 2 Beta R 2 Beta R 2 Info rmation completeness simple single task 0.314 0.099 Information completeness complex single task 0.360 0.130 Grammaticality simple single task 0.480 0.230 0.250 0.063 Grammaticality complex single task 0.386 0.149 0.334 0.112 0.272 0.074 Grammaticality simple dual task 0.260 0.068 Grammaticality complex dual task 0.329 0.108
95 Table 4 7. Predictors for significant and marginal (p=<.055) effects for quantitative measures of discourse production M easure Processing Speed Working Memory Updating Group Beta R 2 Beta R 2 Beta R 2 Beta R 2 Sentence length single task 0.380 0.144 Sentence length dual task 0.307 0.094 Sentence length variability dual task 0.289 0.083 Plural noun count dual task 0.418 0.175 0.319 0.102 Proper noun count single task 0.278 0.077 Proper noun proportion single task 0.406 0.164 Wh pronoun count dual task 0.294 0.087 Prepositions count single task 0.319 0.102 Prepositions count dual task 0.261 0.068 0.382 0.146 Verbs per sentence single task 0.387 0.150 Verbs per sentence dual task 0.264 0.070
96 Table 4 8. Predictors for significant and marginal (p=<.055) effects for qualitat ive measures of discourse production Measure Processing Speed Working Memory Updating Group Beta R 2 Beta R 2 Beta R 2 Beta R 2 Proposition density single task 0.352 0.124 Proposition density dual task 0.439 0.193 0.243 0.054 Proposition count single task 0.389 0.151 0.398 0.158 0.269 0.072 Proposition count dual task 0.348 0.121 0.383 0.146 Syntactic complexity dual task 0.324 0.105 Argument cohesion single task 0.374 0.140 0.253 0.064 Local anaphor cohesio n single task 0.285 0.082 Local anaphor cohesion dual task 0.304 0.092 Global anaphor cohesion single task 0.330 0.109 Global anaphor cohesion dual task 0.273 0.075 Temporal connective word count dual task 0.278 0. 077 Local syntactic similarity dual task 0.343 0.117 Global syntactic similarity dual task 0.275 0.075 Local coherence single task 0.315 0.099
97 Figure 4 1 P D produced fewer passive verb s than controls only in dual task. Figure 4 2 Controls use d more participle verbs in dual task in contrast to P PD. 0.00 0.50 1.00 1.50 2.00 2.50 3.00 Passive Verbs (Density) p <.05 Dual Task PD Control PD Control Single Task 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Past Participle Verb (Percentage) p <.05 Dual Task PD Control PD Control Single Task
98 Figure 4 3. PD provided more critical information than controls only in single task. Figure 4 4. PD wer e less grammatical in the dual task compared to the single task 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Information Completeness (Proportion) Complex PD Simple Simple Simple Simple Complex Complex Control PD Control p <.01 p <.05 Complex Dual Task Single Task 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 Grammaticality (proportion) p <.001 Complex Picture PD Control PD Control Simple Picture
99 Figure 4 5. Controls bec ame more locally coherent than PD only in the single task 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 Local Coherence (cosine) p <.05 Dual Task PD Control PD Control Single Task
100 CHAPTER 5 DISCUSSION have not been investigated in the literature ; thus, findings from this project extend the literature on language production in this population In this exploratory study, our findings revealed broad thematic differences between language tasks. While the effects of a cycling dual task were prominent during picture description, the discourse production task was more sensitive in capturing differences in language use between people with and healthy controls. As has been observed by Murray and Lenz (2001) picture description is a more constrained behavior which can also be more resource intensive due to the increased demands it puts on lexical selection. This study revealed that while dual task effects are prominent in the sentence production task, these effects were not greater in people with It is noted however, that the picture description task also provide d participants with external cues for high frequency words and actions. Thus, while at a resource level the picture des cription task may have been more demanding than the dual task, word choice is externally supported. In contrast, discourse production is associated with reduced resource demands but requires extended internal cueing to support sustained message generation Internal cueing, as in the discourse production task, is difficult for people group differences in quantitative and qualitative language impairments were evident in discourse produ ction but not in the picture description task. In particular, this study finds evidence of impairments in lexical selection, information integration at syntactic and information levels, and information monitoring. These findings will be discussed below.
101 L exical Selection Measures of word choice in both language production tasks were subject to the effects of the cycling dual task and provided little evidence that word choice is impaired ed more nouns in both language tasks during cycling. These nouns tended to be lower frequency words which were also more concrete in both language tasks and more specific in the picture description task. This finding, that noun use is relatively well pre served across two language tasks, extends the findings from the single word production literature that Manentic, Zanettib, Arevalo, Cappa & Padovani 2007; Rodrguez Ferreiro a, Menndez, Ribacoba & Cuetos, 2009). Thus, broad effects of the dual task on noun choice affecting both the quantity and quality (frequency and concreteness) of nouns produced were evident in both language production tasks but were not more prominent in We attribute this effect to two factors. First, we attribute dual task facilitation for nouns in dual task conditions to increases in cortical excitation bilaterally. We speculate that this pattern of cortical excitation is associa ted with sustained activation of distributed semantic memory stores which facilitate noun selection. Second, we to the broader finding in the literature that noun use is relatively well preserved. Thus, the dual task would elicit no additional resource demands above those experiences by healthy controls. There was however a single group difference in the quality of nouns produced nouns used in the dual task discour se were less specific for people with
102 discourse to a general deterioration of information structure at a lexical level due to the increased task demands for message gen eration in the discourse task As predicted, the cycling dual task also tended to facilitate the production of action verbs in both language tasks. However, there was no evidence that the dual task facilitated the production of action verbs more for people than for healthy controls. The quality of verbs also changed in the dual task in both language tasks. Action verbs were more specific under dual task conditions in both language tasks for both groups We attribute these finding s to predictions of the theory of embodied cognition (Lakoff, & Johnson, 1980) that since verb production is associated with activation of motor cortices, activation of these cortices by a motor task (in this case concurrent stationery cycling) will facil itate production of word types associated with movement. In support, it is also noted that an action naming impairment has been Manentic, Zanettib, Arevalo, Cappa & Padova ni 2007; Rodrguez Ferreiroa, Menndez, Ribacoba & Cuetos, 2009). However, we found no evidence that action verbs were Importantly, while the production of action verbs increased in dua l task conditions, verb cohesion in discourse was generally reduced for all participants in dual task conditions, indicating that the production of additional action verbs did not improve the structure of information content. As predicted, there was some e vidence of group differences in the production of significantly more causative verbs overall than healthy controls, in the discourse
103 processing tasks. The increased use of ca usative verbs during discourse is difficult to explain, however as with action verbs which specify a relational action by a single entity or between entities, causative verbs are used to indicate a directional relationship between entities and increased us may indicate increased specificity in the use of verbs. However, it is noted that these differences, while significant represented only a small number of verbs of this type. It may also be important that the comprehension of causative verbs in sentences is Information Integration There were pervasive differences in information integration at both syntactic and information levels disease are widely reported in the literature (Altmann & Troche, 2011; Angwin, Chenery, Copland, Murdoch, & Sil burn, 2005, 2006; Grossman, et al., 1991; Grossman, Carvell, Stern, Gollomp, & Hurtig, 1992; Grossman, et al., 2000; Hochstadt, Nakano, Lieberman, & Friedman, 2006; Lieberman, et al., 1992; Troche & Altmann, 2012) which contrasts with a relative preserva tion of lexical production noted by Illes and Meter (1988) and others (Cotellia, Borronib, Manentic, Zanettib, Arevalo, Cappa & Padovani 2007; Rodrguez Ferreiroa, Menndez, Ribacoba & Cuetos, 2009). pronouns in both language p redicted by poorer information upd ating during discourse. In contrast, in the picture description task, there were no significant cognitive predictors, althou gh people with pronouns in dual task conditions. Thus, the
104 elaborate sentences, as evidenced by fewer embedded clauses. Similarly, it was also during picture description, an effect predicted only by the presence of disease. Effects on measures of information complexity were more pervasive during discou produced fewer adjectives in both language tasks, they produced fewer prepositions and conveyed less propositional information only during discourse. The communication of less propo sitional information and production of fewer prepositions in dual task conditions was predicted by slower processing speed and poorer information updating In contrast, adjective but only durin were also evident for measures of syntactic or structural complexity during discourse production. person sing ular verbs but only in the sentence processing task and not for discourse. The production of fewer present tense third person singular verbs in the sentence production task is attributed to task demands associated with the picture description task. In part icular, the picture description task constrains the choice of this verb type for re
105 ph enomenon will require more detailed analysis of exactly what syntactic forms were produced. complex sentences with less variability which contained fewer prepositions, fewer adject ives and fewer verbs per sentence. Reductions in syntactic complexity, sentence length and variability and the number of verbs per sentence were predicted only by people with disea se showed differences in lexical choice which could result in less elaborate noun phrases and sentences; however, these lexical effects were not all due to the same underlying impairments. Some were traceable to cognitive deficits, some were attributable o nly to the disease itself, and another was not associated with any of the predictors used in the current study. Thus the current study found evidence of general affecting with the amount of information produced and syntactic complexity for both quantitative and qualitative measures of language production. In support, regression analysis indicates that these differences were predicted by disease state and, for some measure s, by slower processing speeds and a decreased ability to manipulate and integrate new information with old (i.e. updating ). Until now, there has been considerable disagreement over whether grammatical complexity is impaired in n & Troche, 2011), but previous studies were more limited in scope and included fewer participants. The current study extends these findings by explicitly contrasting two language tasks, and incorporating a broader of measures of information and syntactic complexity at both quantitative and qualitative levels in a
106 larger group than has reported previously. In so doing, we have identified patterns of impairments that strongly support the conclusion of impaired syntactic complexity for people with Parkinson Syntactic complexity was also affected by the cycling dual task, but these effects were more evident in the sentence production task. During the dual task, participants described pictures using shorter and less structurally variable sentences w hich used fewer verbs per sentence and contained fewer prepositions. Consistent with these findings, sentences produced in the dual task also used simpler syntax at a lower grade reading level. Evidence from group by task interactions suggests that the cy cling dual task particularly impaired sentence complexity. For example under dual task conditions sentence constructions). In contrast, healthy controls tended to use a g reater proportion of verbs and more past participle verbs under dual task conditions suggesting increased use of passive or perfective constructions. The decreased use of past participle verbs was predicted by slower processing speed and poorer informatio n updating Moreover, in the dual task all participants tended to use fewer verb participle s during the discourse task. Information Monitoring There was evidence that information tracking was generally impaired in G roup effects were pe rvasive in the discourse production task while there was only limited evidence that information tracking was impaired in between tasks are due to different task demands be tween language tasks. The picture
107 description involved serial presentation of picture stimuli and was supported by external picture cues which remained on screen until a response was initiated. Thus, there was no requirement to maintain consistency, cohes ion or coherence with previous utterances. While in contrast, the discourse production task involved extended production of new information that needed to be fully integrated with previous utterances. Thus, information monitoring constraints are greatest in the discourse production task, an interpretation which may explain why group effects were most apparent in this language tasks. Information monitoring deficits in people with were apparent in increased grammatical errors in sentence production and in impaired discourse cohesion and coherence. For example, during describing complex pictures than simple pictures. Considering that maintaining grammat icality entails tracking details of previous parts of a sentence, (e.g., the number and person of the subject, the tense of previously used verbs, whether a word needs a definite or indefinite article), it necessarily places a premium on information tracki ng. Indeed, information monitoring demands increased with picture complexity for people sentences, while controls had no additional difficulty as picture complexity increased either under single or dual task conditions. These picture complexity effects on grammaticality in both single and dual task conditions that primarily affected people with The rate of grammatical errors in single task conditions was also predicted by poorer working memory and reduced information updating ability, but, there were no additional
108 predictors for grammaticality under dual task conditions. The finding that use of common finding in the literature (Holtgraves, McNamara, Cappaert, & Durso, 2010; Murray, 2000; Troche & Altmann, 2012), however it also possible that grammatical errors increased when describing complex pictures because the message generation demands increased (Troche and Altmann, 2012) or because there were more opportunities for errors in multi clause sentences. During sentence production, it was revealed that people wi provided more critical information content describing both simple and complex pictures during single task conditions. There were no cognitive predictors to explain this phenomenon in single or dual task conditions and the production of more complete descriptions in the single task was only predicted by group. Thus interpreting this effect Since, t his finding contrasts with the literature whic h indicates that information completeness is impaired in people with the disease across a wide range of language production tasks (Bayles, 1990; Cummings, Darkins, Mendez, Hill, & Benson, 1988; Illes, Metter, Hanson, & Iritani, 1988; Murray, 2000). More pe rvasive effects on separate measures of information monitoring were revealed for discourse production compared to picture description, and there was no produced a grea ter number of proper nouns, which was associated with increased problems with information updating and slower processing speeds suggesting slower more impaired people may have had difficulty updating whom they were actually talking
109 about Additionally fo were impaired. Separate measures of cohesion including argument and anaphor cohesion at both local and global levels, showed similar patterns. Increased difficulty maintaining cohesion betwee n adjacent sentences and within the discourse as a whole (i.e. global cohesion) was predicted by poorer information updating skills and by the presence of the disease. Thus, it appears that the ability to use connective words referring to previous senten reduced ability to monitor previous utterances and to updating these to provide well structured communication during discourse. This is the first study to specifically track measures of cohesion i n discourse production within this clinical population. Similarly, impairment as at the structural level (cohesion). In particular, it was revealed that people with Park local level during discourse production. That is, sentences tended to introduce new information rather than referring to and expanding upon previous utterances. This group effect was only p redicted by the presence of the disease. There were however, no group differences for global coherence. We attribute this finding to task design. During the discourse production task, the rhetorical prompt remains on the screen throughout the task and t hus, provides a continuous external cue enabling participants to remain on topic throughout. Implications The broad implications of this study are that language tasks are not equivalent either in their difficulty for the groups tested or in how the dual ta sk operates on language performance at both qualitative and quantitative levels of production. The
110 differences between these language tasks can be traced to differences in task and resource demands and cueing effects. This has implications for what concl usions may be drawn from this study. Thus, it appears that the cycling dual task only elicit ed effects during constrained sentence production and that dual task effects were generally absent during discourse. Dual task paradigms are traditionally associa ted with a pattern of interference for cognitive or motor performance and not with a pattern of facilitation observed for many of our language measures used in this project Thus, a general prediction is that a dual task will interfere with performance in a cognitive task, or in this case, language production. However, we revealed evidence that the cycling task, a disease, elicited facilitative effects on some quantit ative and qualitative measures of picture description. Since the general effects of a dual task on language production in people with another type of dual task on language production might be. We attributed facilitation effects of the cycling task on language measures and the relative scarcity of dual task effects to the low attention demands of the cycling task and bilateral activation of motor cortices Hence, we would p redict that, as the attention dem ands of the motor task increase a pattern of dual task interference would increase for multiple measures of A further implication of these findings is that disc ourse production is generally these group differences primarily impacted use of syntactic structures to convey
111 messages and information tracking. In contrast, there i s relatively little indication that controls. There is very limited evidence of lexical impairment affecting verb use, only for the use of causative verbs which may not be impaired while all other verb effects were grammatical. As predicted the use of action verbs in particular increased during cycling impairments in verb use were most evident within sentence complexity contexts. There were significant reductions in number of verbs used in each sentence during discourse during message generation reduced elaboration at lexical, syntactic and informational levels. In future studies, it might be fruitful to specifically contrast production of types of discourse with different syntactic structures i n this population. For example, procedural discourse with high intentional content and a simplified syntactic structure might be contrasted with a discourse production task which emphasizes descriptive content and elaboration (i.e. increased use of adject ives, pronouns and nouns) using more complex syntactic structures to express information. A final implication is that the pattern of our findings indicates that information disease were not impaired in global coherence, this conclusion seems suspect and as has been detailed may be an artifact of task design, in which a stimulus prompt externa lly cued topic coherence. In terms of planning an intervention, the current study
112 found that a constrained language task leads to more normal patterns of language use (i.e., more similar to healthy adults) affecting both the quantity and quality of infor mation produced. Since, both syntactic complexity and cohesion are reduced in constrains both lexical and syntactic production closely may generalize well to everyday disc life and reducing the burden on the speech motor production system. Future Directio ns This project highlights the effect of a motor task on language performance. A pattern of facilitation and interference was produced and was not sensitive in identifying group differences across measures. This finding is not consistent with the dual tas k literature (e.g., Kemper et al., 2003; Plummer attentional demands of the cycling task. Thus, a future direction for this line of research would be to vary the concurrent demands of the distractor task. This could be done at two levels, either by increasing the rate of stationery cycling during performance of the language tasks or by using a different motor task altogether. It may also be worthwhile to investigate dual task effects of cycling using an upper b ody ergometer. The benefit of this task is that somatotopic representation of the upper extremities is closer cortically to areas of cortex associated with language production. If effects observed in this project are maintained using the upper body ergom eter then this has potential as a home The project also highlights information complexity effects. The next step will be to analyze these complexity effects by manipulating the discourse task. This can be done
113 at a number of different levels. For example, we can vary the choice of production task (i.e. procedural versus descriptive versus autobiographical versus narrative discourse), or we can vary task demands by presenting a rhetorical prompt versus a p icture prompt versus a picture sequence. Finally, analyzing the effects of picture complexity within this data set on individual measures of quantity and quality during picture description might be a fruitful future line of investigation. Additionally, an alysis of the pattern of production errors made during picture description and discourse production could also be useful. Given the pervasive effects of information and syntactic complexity it seems warranted to investigate these questions by carefully ex amining the different effects of complexity within the existing data set. Conclusion In summary and with respect of our specific aims and predictions, we revealed discourse across multiple measures of quantity and quality. A pattern was revealed indicating that while the production of individual words was largely unaltered in and on how that information was updating and maintained. Importantly, it was revealed that dual task effects, as predicted, affected the information structure by increasing action verbs production and also production of nouns generally. However, the dual task did n ot improve production of individual word types, which instead, showed evidence of differences in language production in the quantity and quality of information and syntact ic structure and monitoring were predicted largely by differences in information processing speed and the ability to update and maintain new information efficiently.
114 APPENDIX A PARTICIPANT WITH PAR DISCOURSE SAMPLE life. But I think my father was one who probably had, would be the one that I would pick part of the twelfth grade. training and and later on worked in a factory. But whatever job he was in, he did his very best. He gave all he was expected to give and more. He was totally honest. He was conscious all the time honest. He could go into the bank and shake hands with a guy and come out with s a quiet man. But when he spoke, he knew it would be best to listen. And we learned that lesson early on in life. He was he was kind to other people. He was charitable. And later on, it was how about the time I was in tenth or eleventh grade, my mother ha medications then that we have now, I believe. And my father looked after her basically for the rest of her life and then died about a year or two. Before she did, so it was actually, they called it an unkn own disease. My best guess would have been gangrene. But they did some surgery and he passed away the same, the next day. But I think what out of his children. And he got that because we had the example that he presented for us. He he was a Christian man. And he had a high ethical standard, he never, to my knowledge, never mistreated anyone.
115 APPENDIX B HEALTHY OLDER ADULT CONTROL DISCOURSE SA MPLE My mother had a very impor tant impact on my life. First, she was my mother. She gave me life. She was the one who nursed me when I was sick, played with me when I then also as an adult. She helped me through times of trial in my life. And when my brother died, she was the person I turned to for comfort. She was also the person who taught me how to be a woman. She learnt, taught me that family was important and as a mother you were the one who had t do what their tasks were in their life and their tasks were as jobs within the household. She taught me how to be a giving person to the people around me. She taught me my life was more important when I ga ve it to others that I helped other people in ways that may not be tangible but would help them. She gave me ideas about the way to treat people, the way to treat animals. She helped me with advice about what marriage, after I was married, would be, what f amily life should be. So that I would raise my children in the way they should be raised. She also gave me an idea about what it was like to grow old. She showed me that how once your family left home, you were still important that you had an important job to do within your family with your husband within the world around you to help others in whatever ways that you could that they needed. She taught me what it was like to be without a husband, how to be a widow and how to still feel important in living and doing. She also taught me how to face life when you could no longer do the things that you needed to be doing. How to accept the fact that you are older that there are things you cannot do and how to gradually accept the fact you needed constant care and were not afraid about what life was giving you. And about
116 what death would mean. I probably would not have been able to give to others the way that I can. I, she was a Christian and she taught me how to be a Christian. This is an important part of my life.
117 LIST OF REFERENCES Aarsland, D., Andersen, K., Larsen, J. P., Perry, R., Wentzel Larsen, T., Lolk, A., & Kragh Sorensen, P. (2004). The rate of cognitive decline in Parkinson disease. Archives of neurology, 61 (12), 1906 1911. Alexander, G. E., DeLong, M R., & Strick, P. L. (1986). Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annual review of neuroscience, 9 (1), 357 381. Altgassen, M., Phillips, L., Kopp, U., & Kliegel, M. (2007). Role of working memory com ponents in planning performance of individuals with Parkinson's disease. Neuropsychologia, 45 (10), 2393 2397. Altmann, L., Hazamy, A., Carvajal, P., Benjamin, M., Rosenbek, J., & Crosson, B. (in press). Delayed stimulus specific imporvements in discourse following anomia treatment using an intentional gesture. Journal of Speech, Language and Hearing Research. Altmann, L. J. P., & Troche, M. S. (2011). High Level Language Production in Parkinson's Disease: A Review. Parkinson's Disease, 2011 12 pages. Retr ieved from doi:10.4061/2011/238956 Angwin, A. J., Chenery, H. J., Copland, D. A., Murdoch, B. E., & Silburn, P. A. (2005). Summation of semantic priming and complex sentence comprehension in Parkinson's disease. Cognitive brain research, 25 (1), 78 89. An gwin, A. J., Chenery, H. J., Copland, D. A., Murdoch, B. E., & Silburn, P. A. (2006). Self paced reading and sentence comprehension in Parkinson's disease. Journal of Neurolinguistics, 19 (3), 239 252. Audiffren, M., Tomporowski, P. D., & Zagrodnik, J. (20 08). Acute aerobic exercise and information processing: Energizing motor processes during a choice reaction time task. Acta Psychologica, 129 (3), 410 419. Audiffren, M., Tomporowski, P. D., & Zagrodnik, J. (2009). Acute aerobic exercise and information pr ocessing: modulation of executive control in a random number generation task. Acta Psychologica, 132 (1), 85 95. Bartels, A. L., & Leenders, K. L. (2009). Parkinson's disease: The syndrome, the pathogenesis and pathophysiology. Cortex, 45 (8), 915 921. Bay les, K. A. (1990). Language and Parkinson disease. Alzheimer Disease & Associated Disorders, 4 (3), 171 180. Bayles, K. A., Trosset, M. W., Tomoeda, C. K., Montgomery Jr, E. B., & Wilson, J. (1993). Generative naming in Parkinson disease patients. Journal of Clinical and Experimental Neuropsychology, 15 (4), 547 562.
118 Benecke, R., Rothwell, J., Dick, J., Day, B., & Marsden, C. (1986). Performance of simultaneous movements in patients with Parkinson's disease. Brain, 109 (4), 739 757. Berardelli, A., Rothwell J., Thompson, P. D., & Hallett, M. (2001). Pathophysiology of bradykinesia in Parkinson's disease. Brain, 124 (11), 2131 2146. Berardelli, A., Sabra, A. F., & Hallett, M. (1983). Physiological mechanisms of rigidity in Parkinson's disease. Journal of Neu rology, Neurosurgery, and Psychiatry, 46 45 53. Bialystok, E., Craik, F. I. M., Stefurak, T., Braver, T., Barch, D., Broadbent, D., Barker, R. (2008). Planning and task management in Parkinson's disease: Differential emphasis in dual task performan ce. Journal of the International Neuropsychological Society, 14 (2), 257 265. Bock, K., Loebell, H., & Morey, R. (1992). From conceptual roles to structural relations: Bridging the syntactic cleft. Psychological review, 99 (1), 150 171. Braak, H., Tredici, K. D., de Vos, R. A. I., Jansen Steur, E. N. H., & Braak, E. (2003). Staging of brain pathology related to sporadic Parkinson's disease. Neurobiology of aging, 24 (2), 197 211. Brown, C., Snodgrass, T., & Covington, M. A. (2007a). CPIDR 3.2 (Computerized Propositional Idea Density Rater). Open source software available from http://www.ai.uga.edu/caspr Brown, R., & Marsden, C. (1991). Dual task performance and processing resources in normal subjects and patients with Parkinson's disease. Brain, 114 (1), 215 231. Brown, R. G., & Jahanshahi, M. (1998). An unusual enhancement of motor performance Journal of Neurology, Neurosurgery & Psychiatry, 64 (6), 813 816. Brown, R., & Marsden, C. (1988). Internal versus external cues and the control of Brain, 111 (2), 323 345. Bublak, P., Mller, U., Grn, G., Reuter, M., & Von Cramon, D. Y. (2002). Manipulation of working memory information is imp aired in Parkinson's disease and related to working memory capacity. Neuropsychology; Neuropsychology, 16 (4), 577 590. Cahn, D. A., Sullivan, E. V., Shear, P. K., Pfefferbaum, A., Heit, G., & Silverberg, G. (1998). Differential contributions of cognitive and motor component processes to Archives of clinical neuropsychology, 13 (7), 575 583.
119 Caplan, D., & Waters, G. (2003). The relationship between age, processing speed, working me mory capacity, and language comprehension. Memory, 13 (3 4), 403 413. Carr, J. (2002). Tremor in Parkinson's disease. Parkinsonism & related disorders, 8 (4), 223 234. Cavanna, A. E., & Trimble, M. R. (2006). The precuneus: a review of its functional anato my and behavioural correlates. Brain, 129 (3), 564 583. Colcombe, S., & Kramer, A. F. (2003). Fitness effects on the cognitive function of older adults: A meta analytic study. Psychological Science, 14 (2), 125 130. Coltheart, M. (1981). The MRC Psycholing uistic Database quarterly. Journal of Experimental Psychology, 33 497 505. Cooper, C., Mikos, A., Wood, M., Kirsch Darrow, L., Jacobson, C., Okun, M., Fernandez, H. (2009). Does laterality of motor impairment tell us something about cognition in Pa rkinson disease? Parkinsonism & related disorders, 15 (4), 315 317. Cotelli, M., Borroni, B., Manenti, R., Zanetti, M., Arevalo, A., Cappa, S., & Padovani, A. (2007). Action and object naming in Parkinson's disease without dementia. European Journal of Neu rology, 14 (6), 632 637. Crossley, S., Salsbury, T., & McNamara, D. (2011). Predicting the proficiency level of language learners using lexical indices. Language Testing, 29 (2), 243 263. Crosson, B. (1999). Subcortical Mechanisms in Language: Lexical Sema ntic Mechanisms and the Thalamus. Brain and cognition, 40 (2), 414 438. Cummings, J. L. (1988). Intellectual impairment in Parkinson's disease: clinical, pathologic, and biochemical correlates. Journal of geriatric psychiatry and neurology, 1 (1), 24 36. D a vies, M. (2010). The Corpus of Contemporary A merican English as the first reliable monitor corpus of English. Literary and Linguistic Computing, 25 (4), 447 464. Doumas, M., Rapp, M. A., & Krampe, R. T. (2009). Working memory and postural control: adult ag e differences in potential for improvement, task priority, and dual tasking. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 64 (2), 193 201. Downes, J. J., Sharp, H. M., Costall, B. M., Sagar, H. J., & Howe, J. (1993). Al ternating fluency in Parkinson's disease. Brain, 116 (4), 887 902. Duran, N.D., McCarthy, P.M., Graesser, A.C., & McNamara, D S. ( 2007 ). Using temporal cohesion to predict temporal coherence in narrative and expository texts Behavior Research Methods 39 ( 2 ), 212 223
120 Edmonds, L. A., & Babb, M. (2011). Effect of Verb Network Strengthening Treatment in Moderate to Severe Aphasia. American Journal of Speech Language Pathology, 20 (2), 131. Edmonds, L. A., & Mizrahi, S. (2011). Online priming of agent and patie nt thematic roles and related verbs in younger and older adults. Aphasiology, 25 (12), 1488 1506. El Sharkawi, A., Ramig, L., Logemann, J., Pauloski, B., Rademaker, A., Smith, C., Werner, C. (2002). Swallowing and voice effects of Lee Silverman Voice Treatment (LSVT): a pilot study. Journal of Neurology, Neurosurgery & Psychiatry, 72 (1), 31 36. Ferretti, T.R., McRae, K., & Hatherell, A. (2001). Integrating verbs, situation schemas, and thematic role concepts. Journal of Memory and Language, 44 516 54 7. Geyer, H.L., & Grossman, M. (1994). Investigating the basis for the sentence Journal of Neurolinguistics, 8 (3), 191 205. Graesser, A. C., McNamara, D. S., Louwerse, M. M., & Cai, Z. (2004). Coh Metrix: Ana lysis of text on cohesion and language. Behavior Research Methods, Instruments, and Computers, 36 193 202. Grahn, J. A., Parkinson, J. A., & Owen, A. M. (2008). The cognitive functions of the caudate nucleus. Progress in neurobiology, 86 (3), 141 155. Gr ossman, M., Carvell, S., Gollomp, S., Stern, M. B., Vernon, G., & Hurtig, H. I. (1991). Neurology, 41 (10), 1620 1626. Grossman, M., Carvell, S., Stern, M. B., Gollomp, S., & Hurtig, H. I. (1992). Sentence comprehension in Parkinson's disease: The role of attention and memory. Brain and Language, 42 (4), 347 384. Grossman, M., Kalmanson, J., Bernhardt, N., Morris, J., Stern, M. B., & Hurtig, H. I. (2000). Cognitive Resource Limitations durin g Sentence Comprehension in Parkinson's Disease. Brain and Language, 73 (1), 1 16. Hanlly, J., Dewick, H., Davies, A., Playeer, J., & Turnbull, C. (1990). Verbal fluency in Parkinson's disease. Neuropsychologia, 28 (7), 737 741. Hartsuiker, R. J., & Barkhu ysen, P. N. (2006). Language production and working memory: The case of subject verb agreement. Language and Cognitive Processes, 21 (1 3), 181 204. Hauk, O., Johnsrude, I., & Pulvermller, F. (2004). Somatotopic representation of action words in human mot or and premotor cortex. Neuron, 41 (2), 301 307.
121 Ho, A. K., Iansek, R., & Bradshaw, J. L. (2002). The effect of a concurrent task on Parkinsonian speech. Journal of Clinical and Experimental Neuropsychology, 24 (1), 36 47. Hochstadt, J., Nakano, H., Lieber man, P., & Friedman, J. (2006). The roles of sequencing and verbal working memory in sentence comprehension deficits in Parkinson's disease. Brain and Language, 97 (3), 243 257. Holmes, J., Jenkins, M., Johnson, A., Adams, S., & Spaulding, S. (2010). Dual task interference: The effects of verbal cognitive tasks on upright postural stability in Parkinson's disease. Parkinson's Disease, 2010 1 5. Holtgraves, T., McNamara, P., Cappaert, K., & Durso, R. (2010). Linguistic correlates of asymmetric motor sympto m severity in Parkinson's disease. Brain and cognition, 72 (2), 189 196. Holtzer, R., Burright, R. G., & Donovick, P. J. (2004). The sensitivity of dual task performance to cognitive status in aging. Journal of the International Neuropsychological Society, 10 ( 2), 230 238. Holtzer, R., Stern, Y., & Rakitin, B. C. (2005). Predicting age related dual task effects with individual differences on neuropsychological tests. Neuropsychology; Neuropsychology, 19 (1), 18 27. Hoppe, C., Mller, U., Werheid, K., Thne, A., & von Cramon, Y. D. (2000). Digit Ordering Test: clinical, psychometric, and experimental evaluation of a verbal working memory test. The Clinical Neuropsychologist, 14 (1), 38 55. Illes, J. (1989). Neurolinguistic features of spontaneous language pro duction dissociate three forms of neurodegenerative disease: Alzheimer's, Huntington's, and Parkinson's. Brain and Language, 37 (4), 628 642. Illes, J., Metter, E., Hanson, W., & Iritani, S. (1988). Language production in Parkinson's disease: Acoustic and linguistic considerations. Brain and Language, 33 146 160. Jarvis, B. G. (2006a). DirectRT (Version 2006.2.028). New York, NY: Empirisoft. Jarvis, B. G. (2006b). MediaLab (Version 2006.2.40). New York, NY: Empirisoft Corporation. Jueptner, M., Frith, C ., Brooks, D., Frackowiak, R., & Passingham, R. (1997). Anatomy of motor learning. II. Subcortical structures and learning by trial and error. Journal of Neurophysiology, 77 (3), 1325 1337. Kelly, V. E., Janke, A. A., & Shumway Cook, A. (2010). Effects of instructed focus and task difficulty on concurrent walking and cognitive task performance in healthy young adults. Experimental brain research 1 9.
122 Kemmerer, D., Castillo, J. G., Talavage, T., Patterson, S., & Wiley, C. (2008). Neuroanatomical distributi on of five semantic components of verbs: Evidence from fMRI. Brain and Language, 107 (1), 16 43. Kemper, S., Bontempo, D., Svhmalzried, R., McKedy, W., Tagaliaferri, B., & Kieweg, D. (2013). Tracking reading: Dual task costs of oral reading for young versu s older adults. Journal of Psycholinguistic Research, 1 22 Kemper, S., Herman, R. E., & Lian, C. H. T. (2003). The costs of doing two things at once for young and older adults: Talking while walking, finger tapping, and ignoring speech of noise. Psychol ogy and Aging, 18 (2), 181 192. Kemper, S., Herman, R. E., & Lian, C. H. T. (2003). The costs of doing two things at once for young and older adults: Talking while walking, finger tapping, and ignoring speech of noise. Psychology and Aging, 18 (2), 181 192. Kemper, S., Herman, R. E., & Nartowicz, J. (2005). Different effects of dual task demands on the speech of young and older adults. Aging, Neuropsychology, and Cognition, 12 (4), 340 358. Kemper, S., McDowd, J., Pohl, P., Herman, R., & Jackson, S. (2006). Revealing language deficits following stroke: the cost of doing two things at once. Aging, Neuropsychology, and Cognition, 13 (1), 115 139. Kemper, S., Schmalzried, R., Herman, R., Leedahl, S., & Mohankumar, D. (2009). The effects of aging and dual task de mands on language production. Aging, Neuropsychology, and Cognition, 16 (3), 241 259. Kemper, S., Schmalzried, R. L., Hoffman, L., & Herman, R. (2010). Aging and the vulnerability of speech to dual task demands. Psychology and Aging, 25 (4), 949 962. Kempe r, S., & Sumner, A. (2001). The structure of verbal abilities in young and older adults. Psychology and Aging, 16 (2), 312 322 Kemps, E., Szmalec, A., Vandierendonck, A., & Crevits, L. (2005). Visuo spatial processing in Parkinson's disease: evidence for diminished visuo spatial sketch pad and central executive resources. Parkinsonism & related disorders, 11 (3), 181 186. Koerts, J., Leenders, K. L., & Brouwer, W. H. (2009). Cognitive dysfunction in non demented Parkinson's disease patients: Controlled and automatic behavior. Cortex, 45 (8), 922 929. Lakoff, G., & Johnson, M. (1980). The metaphorical structure of the human conceptual system. Cognitive Science, 4 (2), 195 208.
123 Landauer, T. K., Foltz, P. W., & Laham, D. (1998). An introduction to latent seman tic analysis. Discourse processes, 25 (2 3), 259 284. Lehricy, S., Ducros, M., Van De Moortele, P. F., Francois, C., Thivard, L., Poupon, C., Kim, D. S. (2004). Diffusion tensor fiber tracking shows distinct corticostriatal circuits in humans. Annal s of neurology, 55 (4), 522 529. Lehricy, S., van de Moortele, P. F., Lobel, E., Paradis, A. L., Vidailhet, M., Frouin, V., Le Bihan, D. (1998). Somatotopical organization of striatal activation during finger and toe movement: a 3 T functional magne tic resonance imaging study. Annals of neurology, 44 (3), 398 404. Levy, J., Hoover, E., Waters, G., Kiran, S., Caplan, D., Berardino, A., & Sandberg, C. ( 2012 ). Effects of syntactic complexity, semantic reversibility, and explicitness on discourse comprehe nsion in persons with aphasia and in healthy controls American Journal of Speech Language Pathology 21 S154 S165 Li, K. Z. H., Lindenberger, U., Freund, A. M., & Baltes, P. B. (2001). Walking While Memorizing: Age Related Differences in Compensatory B ehavior. Psychological Science, 12 (3), 230 237. Lieberman, P. (2001). Human Language and Our Reptilian Brain. Perspectives in Biology and Medicine, 44 (1), 32 51. Lieberman, P., Kako, E., Friedman, J., Tajchman, G., Feldman, L. S., & Jiminez, E. B. (1992) Speech production, syntax comprehension, and cognitive deficits in Parkinson's disease. Brain and Language, 43 (2), 169 189. Mahon, B.Z., & Caramazza, A. (2009). Concepts and categories: A cognitive neuropsychological perspective. Annual Review of Psycho logy, (60), 27 51. Marchese, R., Bove, M., & Abbruzzese, G. (2003). Effect of cognitive and motor tasks on postural stability in Parkinson's disease: a posturographic study. Movement disorders, 18 (6), 652 658. Marsden, C., & Parkes, J. (1977). Success an d problems of long term levodopa therapy in Parkinson's disease. The Lancet, 309 (8007), 345 349. Mayeux, R., Denaro, J., Hemenegildo, N., Marder, K., Tang, M. X., Cote, L. J., & Stern, Y. (1992). A population based investigation of Parkinson's disease wit h and without dementia: relationship to age and gender. Archives of neurology, 49 (5), 492 497. McDowd, J., Hoffman, L., Rozek, E., Lyons, K. E., Pahwa, R., Burns, J., & Kemper, S. (2011). Understanding verbal fluency in healthy aging, Alzheimer's disease, and Parkinson's disease. Neuropsychology, 25 (2), 210 225.
124 McNamara, D.S., Ozuru, Y., Graesser, A.C., & Louwerse, M. (2006) Validating Coh Metrix. In R. Sun & N. Miyake(Eds), Proceedings of the 28th Annual Conference of the Cognitive Science Scociety (pp. 573). Mahwah, NJ: Eribaum. McRae, K., Hare, M., Elman, J.L., & Ferretti, T. (2005). A basis for generating expectancies for verbs from nouns. Memory and Cognition, 33 (7), 1174 1184. Middleton, F. A., & Strick, P. L. (2002). Basal he prefrontal cortex of the primate. Cerebral Cortex, 12 (9), 926 935. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., & Howerter, A. (2000). The unity and diversity of executive functions and their contributions to complex A latent variable analysis. Cognitive Psychology, 41 49 100. Monchi, O., Petrides, M., Mejia Constain, B., & Strafella, A. P. (2007). Cortical activity in Parkinson's disease during executive processing depends on striatal involvement. Brain, 130 (1), 233 244. Morris, M., Iansek, R., Smithson, F., & Huxham, F. (2000). Postural instability in Parkinson's disease: a comparison with and without a concurrent task. Gait & posture, 12 (3), 205 216. Murray, L. L. (2000). Spoken language production in Huntington' s and Parkinson's diseases. Journal of Speech, Language, and Hearing Research, 43 (6), 1350 1366. Murray, L. L., & Lenz, L. P. (2001). Productive syntax abilities in Huntington's and Parkinson's diseases. Brain and cognition, 46 (1 2), 213 219. patients with newly diagnosed Parkinson disease. Neurology, 65 (8), 1239 1245. Ntnen, R., Tervaniemi, M., Sussman, E., Paavilainen, P., & Winkler, I. (2001). Primitive intelligence'in the auditory cortex. Trends in Neurosciences, 24 (5), 283 288. Nicholas, L.E., Brookshire, R.H. (1992). A system for scoring main concepts in the discourse of non brain damaged and aphasic speakers. Clinical Aphasiology, 21 87 99. Nishio, Y., Hirayama, K., Takeda, A., Hosokai, Y., Ishioka, T., Suzuki, K., Mori, E. European Journal of Neurology, 17 (8), 1090 1097. Nussbaum, R.L., & Ellis, C.E. (2003). Alzhe New England Journal of Medicine, 348 1356 1364.
125 O'Shea, S., Morris, M. E., & Iansek, R. (2002). Dual task interference during gait in people with Parkinson disease: effects of motor versus cognitive secondary tasks Physical Therapy, 82 (9), 888 897. Oliverira, R. M., Gurd, J. M., Nixon, P., Marshall, J. C., & Passingham, R. E. (1998). Hypometria in Parkinson's disease: automatic versus controlled processing. Movement disorders, 13 (3), 422 427. Pashler, H. (1994). Dual task interference in simple tasks: data and theory. Psychological Bulletin; Psychological Bulletin, 116 (2), 220 244. Pran, P., Cherubini, A., Assogna, F., Piras, F., Quattrocchi, C., Peppe, A., Stefani, A. (2010). Magnetic resonance imaging ma nigrostriatal signature. Brain, 133 (11), 3423 3433. Pran, P., Rascol, O., Dmonet, J. F., Celsis, P., Nespoulous, J. L., Dubois, B., & Cardebat, D. (2003). Deficit of verb generation in nondemented patients with Parkinson's d isease. Movement disorders, 18 (2), 150 156. Piatt, A. L., Fields, J. A., Paolo, A. M., Koller, W. C., & Trster, A. I. (1999). Lexical, semantic, and action verbal fluency in Parkinson's disease with and without dementia. Journal of Clinical and Experimen tal Neuropsychology, 21 (4), 435 443. Pirozzolo, F. J., Hansch, E. C., Mortimer, J. A., Webster, D. D., & Kuskowski, M. A. (1982). Dementia in Parkinson disease: a neuropsychological analysis. Brain and cognition, 1 (1), 71 83. Plummer D'Amato, P., Altmann L. J. P., & Reilly, K. (2011). Dual task effects of spontaneous speech and executive function on gait in aging: Exaggerated effects in slow walkers. Gait & posture 33(2), 233 237. Plummer D'Amato, P., Altmann, L. J. P., Saracino, D., Fox, E., Behrman, A L., & Marsiske, M. (2008). Interactions between cognitive tasks and gait after stroke: a dual task study. Gait & posture, 27 (4), 683 688. Power, M. (1986). A technique for measuring processing load during speech production. Journal of psycholinguistic r esearch, 15 (5), 371 382. Rascol, O., Payoux, P., Ory, F., Ferreira, J. J., Brefel Courbon, C., & Montastruc, J. L. (2003). Limitations of current Parkinson's disease therapy. Annals of Neurology, 53 (S3), S3 S15. Raskin, S. A., Sliwinski, M., & Borod, J. C (1992). Clustering strategies on tasks of verbal fluency in Parkinson's disease. Neuropsychologia, 30 (1), 95 99. Richards, M., Cote, L. J., & Stern, Y. (1993). Executive function in Parkinson's disease: set shifting or set maintenance? Journal of Clinic al and Experimental Neuropsychology, 15 (2), 266 279.
126 Rochester, L., Hetherington, V., Jones, D., Nieuwboer, A., Willems, A. M., Kwakkel, G., & Van Wegen, E. (2004). Attending to the task: Interference effects of functional tasks on walking in Parkinson's disease and the roles of cognition, depression, fatigue, and balance. Archives of physical medicine and rehabilitation, 85 (10), 1578 1585. Rodriguez, A. (2010). Semantic motor representations : effects on language and motor production (Unpublished doctor al dissertation). University of Florida, Florida. Rodrguez Ferreiro, J., Menndez, M., Ribacoba, R., & Cuetos, F. (2009). Action naming is impaired in Parkinson disease patients. Neuropsychologia, 47 (14), 3271 3274. Rodriguez Oroz, M. C., Jahanshahi, M., Krack, P., Litvan, I., Macias, R., Bezard, E., & Obeso, J. A. (2009). Initial clinical manifestations of Parkinson's disease: features and pathophysiological mechanisms. The Lancet Neurology, 8 (12), 1128 1139. Rogers, D., Lees, A., Smith, E., Trimble, M. & Stern, G. (1987). Bradyphrenia in Parkinson's disease and psychomotor retardation in depressive illness. Brain, 110 (3), 761 776. Ruthruff, E., Pashler, H. E., & Klaassen, A. (2001). Processing bottlenecks in dual task performance: Structural limitatio n or strategic postponement? Psychonomic Bulletin & Review, 8 (1), 73 80. Shumway Cook, A., Guralnik, J. M., Phillips, C. L., Coppin, A. K., Ciol, M. A., Bandinelli, S., & Ferrucci, L. (2007). Age Associated Declines in Complex Walking Task Performance: Th e Walking InCHIANTI Toolkit. Journal of the American Geriatrics Society, 55(1), 58 65. Shtyrov, Y., Hauk, O., & Pulvermller, F. (2004). Distributed neuronal networks for encoding category specific semantic information: the mismatch negativity to action wo rds. European Journal of Neuroscience, 19 (4), 1083 1092. Signorini, M., & Volpato, C. (2006). Action fluency in Parkinson's disease: A follow up study. Movement disorders, 21 (4), 467 472. Sjberg, L., Svensson, E., & Persson, L. O. (1979). The measuremen t of mood. Scandinavian Journal of Psychology, 20 (1), 1 18. Small, J. A., Lyons, K., & Kemper, S. (1997). Grammatical abilities in Parkinson's disease: evidence from written sentences. Neuropsychologia, 35 (12), 1571 1576. Snijders, A. H., van Kesteren, M ., & Bloem, B. R. (2011). Cycling is less affected than walking in freezers of gait. Journal of Neurology, Neurosurgery & Psychiatry, 83 (5), 575 576.
127 dementia and new gene therapy. Journal of neurology, 257 (1), 154 157. Tessa, C., Lucetti, C., Diciotti, S., Paoli, L., Cecchi, P., Giannelli, M., Mascalchi, M. disease. Neuroradiology 1 8. Tettamanti, M., Buccino, G., Saccuman, M. C., Gallese, V., Danna, M., Scifo, P., Perani, D. (2005). Listening to action related sentences activates fronto parietal motor circuits. Journal of cognitive neuroscience, 17 (2), 273 281. Thornton, R., & Light, L. L. (2006). Language Comprehension and Production in Normal Aging. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the Psychology of Aging (6th ed., Vol. 2, pp. 253 288). Burlington MA: Elsevier Academic Press. Tomporowski, P. D. (2003). Effect s of acute bouts of exercise on cognition. Acta Psychologica, 112 (3), 297 324. Troche, M. S., & Altmann, L. J. P. (2012). Sentence production in Parkinson disease: Effects of conceptual and task complexity. Applied Psycholinguistics, 1 (1), 1 27. Uc, E., Rizzo, M., Anderson, S., Qian, S., Rodnitzky, R., & Dawson, J. (2005). Visual dysfunction in Parkinson disease without dementia. Neurology, 65 (12), 1907 1913. Ullman, M. T. (2004). Contributions of memory circuits to language: The declarative/procedural m odel. Cognition, 92 (1 2), 231 270. Ullman, M. T., Corkin, S., Coppola, M., Hickok, G., Growdon, J. H., Koroshetz, W. J., & Pinker, S. (1997). A neural dissociation within language: Evidence that the mental dictionary is part of declarative memory, and tha t grammatical rules are processed by the procedural system. Journal of cognitive neuroscience, 9 (2), 266 276. Van Den Eeden, S.K., Tanner, C.M., Bernstein, A.L., Fross, R.D., Leimpeter, A., Bloch, ase: variation by age, gender, and race/ethnicity. American Journal of Epidemiology, 157 (11), 1015 1022. Walsh, B., & Smith, A. (2011). Linguistic Complexity, Speech Production, and Comprehension in Parkinson's Disease: Behavioral and Physiological Indices Journal of Speech, Language, and Hearing Research, 54 (3), 787 802. Warabi, T., Fukushima, K., Olley, P. M., Chiba, S., & Yanagisawa, N. (2011). Difficulty in terminating the preceding movement/posture explains the impaired initiation of new movements in Parkinson's disease. Neuroscience letters, 496 (2), 84 89. Watson, G., & Leverenz, J. B. (2010). Profile of Cognitive Impairment in Parkinson's Disease. Brain Pathology, 20 (3), 640 645.
128 Werheid, K., Hoppe, C., Thne, A., Mller, U., Mngersdorf, M., & Vo n Cramon, D. Y. (2002). The Adaptive Digit Ordering Test: Clinical application, reliability, and validity of a verbal working memory test. Archives of clinical neuropsychology, 17 (6), 547 565. Wilson, J. P., & Altmann, L. J. P. (in revision). Aging, cogni tion and sentence production: A test of two theories. Wu, T., & Hallett, M. (2005). A functional MRI study of automatic movements in patients with Parkinson's disease. Brain, 128 (10), 2250 2259. Wu, T., & Hallett, M. (2008). Neural correlates of dual tas k performance in patients with Journal of Neurology, Neurosurgery & Psychiatry, 79 (7), 760 766. Yogev, G., Giladi, N., Peretz, C., Springer, S., Simon, E. S., & Hausdorff, J. M. (2005). Dual tasking, gait rhythmicity, and Parkinson's disease: which aspects of gait are attention demanding? European Journal of Neuroscience, 22 (5), 1248 1256. Yogev Seligmann, G., Hausdorff, J. M., & Giladi, N. (2008). The role of executive function and attention in gait. Movement disorders, 23 (3), 329 34 2.
129 BIOGRAPHICAL SKETCH Jonathan Paul Wilson was born in Belfas t, United Kingdom The younger of two siblings he grew up in the seaside town of Bangor and later Belfast, Northern Ireland. Jonathan graduated from his high school, the Royal Belfast Acade mical Institution in 1989. Jonathan continued into third level education and was awarded a B.A. (Hons ) in English language and l iterature in 1992. Following graduation Jonathan worked in the information technology se ct or within the financial services industry and enrolled pa rt time in graduate studies in computing and information s ystems at the University of Ulster, Northern Ireland Jonathan subsequently graduated with an advanced graduate diplom a in computing and inf ormation s ystems from the University of Ulster United Kingdom Jonathan elected to continue further academic studies in the U S on a full time basis and was subsequently awarded an M.A. in communication s ciences and d isorders from the University of Flori da in August 2008 Following graduation Jonathan continued in to the doctoral program in communication sciences and d isorders combining his program of study with a 2 year clinical fellowship in speech l anguage p athology Jonathan is recognized as a board certified speech language p athologist by the American Speech Language and Hearing Association as of August 2010 and is expected to complete his Ph D program in August 2013. Upon graduation, Jonathan will take a tenure track appointment with Midwestern U niversity, IL, as an assistant professor where he will work as a n independent research scienti st and teacher in the field of speech language p athology specializing in the assessment and treatment of adult neurogenic disorders.