THE CONTRIBUTION OF LOCATION CONTEXT ON VERB AND THEMATIC ROLE ASSOCIATION: EVIDENCE FROM EYE TRACKING IN YOUNGER AND OLDER ADULTS By HYEJIN PARK 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 2014
Â© 2014 Hyejin Park
To My Husband
4 ACKNOWLEDGMENTS M y adventure in a doctora l program has b een wonderful because of the many people who have supported me. Studying in a foreign country and starting a new family with my husband would have been challenging without them. I would first like to thank my two advisors Dr. Lisa A. Edmonds and Dr. Jamie Reilly for their support and guidance i n my academic career. They have always believed in me and encouraged me to challenge myself so that I can advance my abilities in research and teaching. I would also like to thank my committee members, Dr. H. Wind Co wles and Dr. Lise Abrams for their valuable advice on my research. My work could not have been successfully finished without their help. I thank assistants and colleagues in the Aphasia Bilingualism Lab , Catherine Bradley, Becca Wolf, Jacki Rubin, Brayleah Kernan for their help in developing stimuli, running participants, and collecting data for my dissertation . I would also like to thank my friends in the doctoral program who have supported, encouraged, and consoled me the last five years. I wish them the best in their new journeys and send my love. There are also several people who I would like to express my thanks to , Dr. HyangHee Kim and Dr. Jungtak Lee are the reason that I came into this field. They introduced me to this field and inspired me to cont inue my work in a Ph.D. Their warmest consideration and advice motivated me. Mrs. Becky Howard is my friend, tutor, and mother in the U.S. She is always interested in my work and willing to help with everything. Lastly I would like to thank my family and family in law for their loving support. S pecial thanks go to my husband, Joonki Yoon, who is always on my side and understanding. I ap preciate his enormously support and love with all my heart.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIS T OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 LIST OF ABBREVIATIONS ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................................ ... 10 CHAPTER 1 LITERATURE REVIEW ................................ ................................ ................................ ....... 12 Verbs ................................ ................................ ................................ ................................ ....... 12 Verb and Thematic Role Associa tion: Focusing on Patient and Instrument Roles ................ 14 Argument and Adjunct/Modifier ................................ ................................ ..................... 15 Semantic Features ................................ ................................ ................................ ............ 18 Behavioral Evidence Comparing Patient to Instrument Activations by Verbs ............... 22 Context Effects in Typical Language Processing ................................ ................................ ... 28 Semantic Processing and Context Effects in Aging ................................ ............................... 30 2 CONTEXT EFFECTS IN TYPICAL LANGAUGE PROCESSING ................................ .... 37 Introduction ................................ ................................ ................................ ............................. 37 Methods ................................ ................................ ................................ ................................ .. 40 Participants ................................ ................................ ................................ ...................... 40 Stimuli D evelopment ................................ ................................ ................................ ....... 40 Materials ................................ ................................ ................................ .......................... 41 Procedures ................................ ................................ ................................ ....................... 42 Statistical A nalysis ................................ ................................ ................................ .......... 44 Results ................................ ................................ ................................ ................................ ..... 45 Total Fixation Time ................................ ................................ ................................ ......... 45 Time Course Analysis ................................ ................................ ................................ ..... 46 Interim Discussion ................................ ................................ ................................ .................. 49 3 CONTEXT EFECTS IN AGING ................................ ................................ ........................... 53 Introduction ................................ ................................ ................................ ............................. 53 Methods ................................ ................................ ................................ ................................ .. 55 Participants ................................ ................................ ................................ ...................... 55 Language and Cognitive T ests ................................ ................................ ........................ 57 Materials and P rocedures ................................ ................................ ................................ 59 Statistical A nalysis ................................ ................................ ................................ .......... 60
6 Results ................................ ................................ ................................ ................................ ..... 60 Total Fixation Time ................................ ................................ ................................ ......... 60 Time Course Analysis ................................ ................................ ................................ ..... 61 Interim Discussion ................................ ................................ ................................ .................. 64 4 VERB THEMATIC ROLE ASSOCIATION AND CONTEXT EFFECT COMPARISONS BETWEEN THE TWO AGE GROUPS ................................ .................. 68 Introduction ................................ ................................ ................................ ............................. 68 Methods ................................ ................................ ................................ ................................ .. 68 Results ................................ ................................ ................................ ................................ ..... 69 Mean Difference Comparisons ................................ ................................ ........................ 69 Total Fixation Time Comparisons ................................ ................................ ................... 70 Fixation Count Average Comparisons ................................ ................................ ............ 71 Entry Time Average Compari sons ................................ ................................ .................. 72 Interim Discussion ................................ ................................ ................................ .................. 73 5 GENERAL DISCUSSION ................................ ................................ ................................ ..... 76 Verb and The matic Role Comparison ................................ ................................ .................... 77 Context Effects ................................ ................................ ................................ ....................... 78 Methodological Differences ................................ ................................ ............................ 78 Evidence of Context Effects ................................ ................................ ............................ 80 Semantic Processing Changes in Aging ................................ ................................ ................. 82 Conclusion ................................ ................................ ................................ .............................. 88 Significance of Study ................................ ................................ ................................ ...... 88 Future Directions ................................ ................................ ................................ ............. 89 STIMULUS LIST ................................ ................................ ................................ .......................... 91 LIST OF REFERENCES ................................ ................................ ................................ ............... 94 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 102
7 LIST OF TABLES Table page 2 1 Mean (sd) proportions of total fixation time proportion (%) in younger participants. ...... 46 3 1 Summary of Research Questions and Hypotheses Based on Models of Language Proc essing in Aging ................................ ................................ ................................ ........... 56 3 2 Demographic information and language and cognitive test scores ................................ ... 58 3 3 Mean (sd) proportions of total fixa tion time in HOAs ................................ ...................... 61 4 1 Total fixation time proportion differences between AOIs in conditions between groups ................................ ................................ ................................ ................................ . 70 4 2 Total fixat ion time proportion comparisons in conditions between groups ...................... 71 A 1. S timulus List ................................ ................................ ................................ ...................... 92
8 LIST OF FIGURES Figure page 1 1 based representation of lexical information. ............................... 16 2 1 Example set of cutting & kitchen context in each condition in experimental prese ntation ................................ ................................ ................................ ........................ 44 2 2 Fixation proportions of each AOI in each time window (100ms) in younger participants . ................................ ................................ ................................ ........................ 47 3 1 Fixation proportio ns of each AOI in each time window (100ms) in HOAs ...................... 62 4 1 Fixation count averages in each AOI between groups. ................................ ...................... 72 4 2 Entry tim e in each AOI between groups. ................................ ................................ ........... 73
9 LIST OF ABBREVIATIONS AOI Area Of Interest CC condition Congruent Context condition CLQT Cognitive Linguistic Quick Test CP & II condition Congruent Patient and Incongruent Instrument c ondition HOA(s) Healthy Older Adult(s) IP & CI condition Incongruent Patient and Congruent Instrument condition KDT Kissing and Dancing Test NC condition Neutral condition PP Prepositional Phrase PPT test Pyramid and Palm Tree test sd Standard Devi ation
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 Philosophy THE CONTRIBUTION OF LOCATION CONTEXT ON VERB AND TH EMATIC ROLE ASSOCIATION: EVIDENCE FROM EYE TRACKING IN YOUNGER AND OLDER ADULTS By Hyejin Park August 2014 Chair: Lisa A. Edmonds Major: Communication Sciences and Disorders Verbs spontaneously activate thematic roles in an event schema. However, it i s unknown how strongly verbs are associated with each type of thematic role, and whether th e s e association s are affected by previously shown context information. The aims of th is eye tracking study were to investigate ts fixate longer on patients ( o nion ) or instruments ( knife ) (1) with simultaneous presentation of a semantically associated verb ( slice ) , and (2) whe n the verb and thematic roles are primed by a context (location scene) . We investigated these questions wit hin and across a younger and older adult group . E xperiment 1 included 35 young er adults (18 30 years) , and experiment 2 included 17 healthy older adults (60 80 years) . Using the visual world eye tracking paradigm, we manipulated congruency of the locati on scene as context and presented it prior to simultaneous presentation of a spoken verb and semantically associated patient and instrument pictures. We compared the fixation time of patients to instruments during the total picture presentation time (3 sec onds) and at every 100ms time bin across conditions. W e observed significantly longer fixations o n instruments than patients in younger adults ; t he same pattern was also observed in older adults , but the comparison was not significant . These
11 findings sugg est a stronger semantic relationship between verb s and instrument s compared to verb s and patient s, though both are reduce d in aging . A lthough we did not find significant context effects, we found evidence that context influence d eye movements differently for the two age groups. Congruent c ontext facilitated fixations on thematic roles in younger adults . However, congruent context suppressed instrument fixations and did not affect patients in older adults. S tronger association s between verbs and instrument s may be related to their intertwined meanings (i.e., the concepts of slice and knife are somewhat interdependent) . The lack of context effects might be due to the implicit nature of the task . Also, the reduced thematic role priming and suppressed instrume nt activations within congruent context s in older adults might be explained by the transmission deficit theory and neuronal synaptic depression. Further studies are needed to investigate association s between different types of verbs and thematic roles in v arious populations.
12 CHAPTER 1 LITERATURE REVIEW Verbs A verb is an essential element in language; it not only delivers a main purpose of the to express th e action or status of the nouns. Because of its importance, investigating verb processing has been an important topic of research . A line of literature investigated, for example, relates to how listeners/readers parse a sentence based on verb information ( e.g., Rayner, Carlson, & Frazier, 1983) and how verbs generate expectation about upcoming nouns (e.g., Altmann & Kamide, 1999) . It is notable that many research questions with verb s have focused on how verb s cooperate with nouns rather than verb concept s b y themselves. Th is might be due to verbs distinctive characteristics. First, verbs convey both semantic and syntactic information (Druks, 2000) . That is, when a verb is used in a sentence, it delivers the semantic meaning as well as syntactic information of how many noun phrases are required following the type of verb. For example, the verb break requires one noun phrase (e.g., The boy breaks the window ), whereas the verb give requires two noun phrases (e.g., The boy gives the girl the book ). Second, the s emantic concept s of verb s are relatively flexible compared to noun s . Th is is because the mental representation of a verb is highly influenced by other elements such as the person who performs the action, the object that receives or is being used to do the action, the location, etc. (Jiang et al., 2014) . For example, the concept of a pencil is approximately the same across people in terms of its shape (e.g., has a long shape, has a sharp edge, has an eraser at the end, etc.), its function (e.g., to write), a nd its tactile information (e.g., is relatively hard) with subtle differences (e.g., color or length). On the other hand, the mental representation of a verb, cut , can be different depending
13 hold the knife. However, in the situation in a hair salon or barber shop, a hand motion is not the main focus of action but f inger motions manipulating scissors might considered the main concept of the action. Jiang et al. (2014) proposed that nouns are organized hierarchically in domains, while verbs form a more matrix like organization. That is, (1) verb concepts necessitate dependent categories of noun components that are directly connected to the verb; also (2) verbs have many elements of meaning across semantic field (e.g., manner, intention). With respect to verb concepts being highly associated with noun components, Tanen haus, Carlson, and Trueswell (1989) also described the assumption of information available during verb recognition associated with the verb, (3) the types of consti tuents that can serve as complements of the verb, and (4) how roles and syntactic constituents are connected to one another (Tanenhaus et al., 1989, p. 213) in conjunction with noun components , and how verb and noun concepts are intertwined and influence each other. Thematic Role and Schema. The concept s of thematic roles and schema theory explain verb noun association s . An event s chema is a mental structure representing our knowledge of conce pts including objects, social situations, actions, events, or sequential events (Barsalou, 1988; Ferretti, McRae, & Hatherell, 2001; Rumelhart, 1980) . According to schema theory, situations consist of structured lists of slots (thematic role) that are fill ed by specific fillers. W hen a verb describes an action (e.g., slicing), it activates thematic roles of a doer of action (agent), a receiver of action (patient), an object being used to perform the action (instrument), a place where the
14 action occurs (loca tion), and so on. Each slot for the agent, patient, instrument, and location is filled by a specific filler word of a chef (agent filler), an onion (patient filler), a knife (instrument filler), and a kitchen (location filler), for example. Therefore, if a thematic noun follows its associated verb, processing speed of the noun increases compared to when the noun following an unassociated verb (Edmonds & Mizrahi, 2011; Ferretti et al., 2001) . Similarly during sentence processing , verb s generate expectancy o f upcoming noun s. Research has shown that participants more quickly look at the next item (e.g., cake ) given picture options when a verb is semantically enriched (e.g., eat ) compared to less informative (e.g., move ) (Altmann & Kamide, 1999; Kamide, Altmann , & Haywood, 2003) . McRae, Hare, Elman, and Ferretti (2005b) also claimed that verbs are powerful expectancy generators of thematic role nouns processing because verbs increase activations toward thematic role nouns. Ferretti et al. (2001) reported that ve rbs facilitate agent, patient, and instrument nouns compared to unrelated nouns (also see Edmonds & Mizrahi, 2011 for similar findings with agents and patients) . However, we do not have enough knowledge yet about how different verbs are associated with eac h thematic role , what factors contribute to verb thematic role associations, and the mechanisms of relative contribution for different thematic roles. Verb and Thematic Role Association: Focusing on Patient and Instrument Roles The main focus of the curre nt study is therefore to investigate whether verbs are more strongly associated with a particular thematic role and what might contribute to any differences that are found because t he two roles are filled by non animate object entities, which allow compari son of their associations to verbs. In addition, patient and instrument roles possess different semantic and syntactic manifestations, which allow investigation of what elements of the thematic roles significantly contribute to the verb thematic associatio ns. Also, verbs are
15 grammatically subcategorized to intransitive, transitive, and ditransitive depending on how many noun phrases syntactically follow the verb. Also, Levin (1993) and VerbNet (Kipper, 2005) subcategorize verb s with the aspect of semantic c haracteristics (e.g., manner of motion class). Each subcategory of verbs possess es different argumenthood and/or semantic relationship with thematic roles; therefore, it should be noted that the current study limits our interest to transitive action verbs and how much they activate a patient role compared to an instrument role. I n this chapter, the literature describing mechanism s of transitive action verb patient and/or instrument associations is reviewed. Argument and Adjunct/Modifier There are two class es of dependents of verbs called arguments and adjuncts/modifiers (Sussman, 2006) . In sentences, a rguments are considered to be syntactically required by verbs , and they significantly contribute to specific semantic component of a verb (e.g., who does what ); adjuncts or modifiers are syntactically optional and minimally contribute to semantic information. I n a typical English subject verb object (SVO) sentence with a transitive verb, arguments include subject s and direct object s , and typically subject and d irect objects serve agent and patient role s , respectively , as thematic role s . A djunct s include a ll prepositional phrases (PPs) in a sentence, which can serve instrument and location roles. Koenig, Mauner, and Bienvenue (2003) illustrated the organization o f the mental lexicon in a multidimensional hierarchy of categories, and each category include d a combination of syntactic, semantic, and morphological information. For example , as seen in Figure 1, the transitive verb of break semantically represents the n ode labeled break caus , which includes a cause ( cause relation ) and affected entity ( affected relation ). Break also inherits from the node, transitive verb , which summarizes syntactic category information. S yntactic information also
16 activates the semantic r epresentation of cause relation and affected relation . In fact, the salience of activations of the two relations might partially be caused by the information of syntactic processes. The semantic arguments, cause relation and affected relation in the themat ic role view are considered as agent and patient roles to a transitive verb. Therefore, a transitive verb syntactically highly activates arguments (i.e., agent and patient roles) but not adjuncts (e.g., instrument and location roles) (see Koenig et al., 20 03 for full explanation) . Figure 1 based representation of lexical information . Note: Adopted from Koenig et al . (2003, p. 70) with permission Consequently , listeners/speakers read arguments faster than adjuncts during sentence processing (Gardner et al., 2012) and tend to resolve lexical ambiguity in favor of arguments (Munoz, Broughton, Goldring, & Armstrong, 1998) . That is because listeners/speakers prefer to parse a sentence using the fewest syntactic nodes, called the Minima l Attachment principle (Frazier & Rayner, 1982) . Returning to our interest in patient and instrument comparisons, therefore, it is likely that patient nouns would have advantage over an instrument when a verb is processed at a single word level. Further, because verbs do not stand alone (McRae et al., 2005b) , readers/listeners spontaneously activate the words that directly follow the verb , which are most likely patients due
17 to the strong subject verb object(patient) word order in English. If this is the ca se , facilitation effects in patients would be significantly greater than instruments . There is an interesting controversy about whether instruments are arguments or adjuncts. SchÃ¼tze (1995) suggested that there are two perspectives from which to define ar gumenthood : 1. An argument fills a role in the relation described by its associated head, whose presence may be implied by the head. In contrast, a modifier predicates a separate property of its associated head or phrase. 2. A phrase P is an argument of a head H if the semantic contribution of P to the meaning of a sentence in which P is associated with H depends on the particular identity of H. Conversely, P is a modifier if its semantic contribution is relatively constant across a range of sentences in which i t combines with different heads. (SchÃ¼tze, 1995, p. 100) The first definition focuses on how semantic information of the phrase fills syntactic requirements (syntactic perspectives), and the second definition focuses on how the phrase contributes to sema ntic completion (semantic perspectives). In supporting of the first view, Spivey Knowlton (1994; 1995) occur with VP pro forms, 2) they can be ordered after modifiers, 3) they allow emphatic refl exive insertion, 4) they are not restricted in their occurrence, and 5) their relationship to an On the other hand, SchÃ¼tze (1995) claimed that instruments ( with PPs) are a rguments because first and foremost , they fill a role in the relation described by the verb. For example, semantic properties (examples are adopted from SchÃ¼tze, 1995, p. 126) . Similarly, Brunson (1992) suggested that both agent s and instrument s are the fundamental role Cause, which allows an instrument role to have a potent ial argument relation that is expressed in a sentence. Secondly,
18 iterativity is to argue whether a phrase can iterate (adjuncts) as in sentence (a) or not (arguments) not iterate with other with phrases, therefore they should be treated as arguments (examples are adopted from SchÃ¼tze, 1995, pp. 103, 129) . a. She knew a member in Northampton, in Massachusetts. b. *She knew a member of Congress, of the House. c. *Mary c ut the meat with a tool with a saw. d. *John cut the meat with a knife with the sharp end. (Note: *Incorrect sentence) Lastly, relative to the syntactic view, instruments can be replaced by a pro form (a type of function word that expresses another word or p hrase that is stated previously in a sentence, e.g., him, her, that, it, etc.) which is the same patterns of arguments as in sentence (e) and (f) (examples are adopted from SchÃ¼tze, 1995, p. 131) . e. I know that several construction workers were planning to repair the house with various tools. With which tool do you wonder who will actually do so? f. I know that Bill fixed the fan with a hammer. Tell me which tool Mary did so with. In sum, SchÃ¼tze (1995) summarized that both syntactic and semantic tests support that instruments should be considered arguments rather than adjuncts. I nstruments are strongly attach ed to verbs semantically, and therefore they may be considered arguments . Consequently, this may indicate a strong verb instrument association. While Sch Ã¼ tze argued for the argumenthood of instrument s in sentence processing, the strong semantic attachment of instruments to verbs may also be present at the word level. Semantic Features W e have discussed the association between verbs and their thematic roles at the macro level of s entence processing . N ow we will switch to a more micro level analysis by decomposing semantic features of each thematic role concept . Investigating semantic features is one of the
19 major approaches of exploring semantic concepts and s emantic relationships. Semantic features are smaller units of meaning that are often used to categorize or arrange concept hierarchy (Harley, 2008) . For example, the concept of father includes semantic features are +human, +older, and female, while mother features include +human, +older, and +female. Thus, father and mother are categorized under and older group but differentiated by sex. One of the linguistic models claims that shared features (e.g., human and older in the example of father a nd mother ) play an important role in the activations between concepts. The Distributed Network Model (Roelofs, 1997) posits that concepts are presented as patterns of activation across a network of interconnected units which represent various aspects of an object or event. Units are organized into modules representing a particular kind of information (e.g., verbal or visual features) or accomplishing a particular information processing goal (e.g., input or output system). Units within a modules are richly interconnected each other; units in different modules may or may not be connected. When a concept is processed, it causes patterns of activations across units. T he particular patterns are determined by the weights o f the connections between the units. S im ilar concepts are represented by similar patterns of activation. Because the patterns are determined by shared semantic features, concepts that share more semantic features are closely connected and more likely to be co activated. McRae, Cree, Seidenberg, and McNorgan (2005a) reported semantic feature norms of objects. In their database, functional features of tools were usually described in action verb forms. For example, a semantic feature of an oven used for baking something ball was used by throwing . According to Vinson and Vigliocco (2008) functional features are considered one of the salient features of tools , clothing, other artifacts, and body parts. Because common objects
20 serve as instruments , sources of action verb activations in the semantic network would strongly activate these concepts. On the other hand, a common patient noun for baking is a pie , whose semantic features are usually described with perceptual features (e.g., looks round , is hot , is sweet , etc.). These perceptual fe atures are not related to verb concepts, and therefore verb activations in the semantic network would not necessarily activate related patient nouns (in terms of shared features). Further, it is also possible that an oven serves a patient role for a verb, cleaning or opening . However, because its semantic features are associated with the action of baking , the concept of an oven may be most likely activated by an instrument related verb (e.g., baking) rather than patient related verb (e.g., cleaning) . Kipper , Dang, and Palmer (2000) supported this assumption and claimed that each verb possess specific selectional restrictions (e.g., an instrument of kicking must be of type foot ) and semantic characteristics (e.g., a particular manner of directed motion). Intu itively, selectional restrictions are more engaged for instrument s than patient s because instruments (e.g., oven ) are primarily created and developed to the specific action (e.g., baking ) , whereas the action can be done many similar kinds of entities (e.g. , pie , bread , pizza , cookie , etc.). Nagy and Gentner (1990) and Talmy (1975) also argued for a strong dependency between verbs and instruments and N ot only does behavioral evidence in the psycholinguistic literature support co activated patterns of action verb s and instrumental tools, but there is neuronal activation evidence as well . Embodied cognition theory claims that concepts and linguistic information are pr esented in modality specific areas in the brain. Certain concepts are associated with relevant modalities and neuronal networks which are represented by different neuronal assemblies involved in semantic
21 knowledge (Lawrence W Barsalou, 2008; PulvermÃ¼ller, 2005) . Physiologically when learning a concept, nerve cells associated with the motor and sensory semantic features of the concept fir e together . In neuroplasticity, connected neurons that frequently fire together in synchrony are coerced together more clo sely at a functional level (Bhatnagar, 2008; Kipper et al., 2000) . For example, in the brain, the primary motor cortex in the precentral gyrus of the frontal lobe is involved in planning of skilled motor movements (Bhatnagar, 2008) . The primary motor cort ext is organized to control each body part (motor homunculus); the lower motor cortex is responsible for the face, speech muscles, and head; the upper region is responsible for the arms and trunk; and mid sagittal area is responsible for legs and toes (Bha tnagar, 2008) . Therefore, according to the embodied cognition, during processing of a concept such as grasping , mirror neuron activations of the specific motion activate the related neurological areas (e.g., hand movements in the upper area of the primary motor cortex ) (Aggujaro, Crepaldi, Pistarini, Taricco, & Luzzatti, 2006) . Carota, Moseley, and PulvermÃ¼ller (2012) also found supportive evidence for embodied cognition theory. They presented three different types of words (animal, food, and tool) to part icipants and asked them to read the words silently while their brain activations were scanned. The authors found activation in the general language area in the inferior frontal, middle superior temporal, and inferior temporal fusiform areas as well as addi tional activations in areas specific to each type of word category strong dorsolateral precentral cortex activations near the somatotopic motor finger representation for tool words and inferolateral precentral area activations near the tongue representat ion for food words. Not only have neuroimaging studies supported these patterns of activation, but research with clinical populations has also reported language characteristics supporting embodied cognition theory. A study reported action verb
22 processing d ifficult ies disorders (Fernandino et al., 2013) . Also, individuals with autism spectrum with pervasive motor problems show specific deficit s in the semantic processing of action related words (Moseley et al., 2013) . We discussed embodied cognition theory and supportive findings in the literature, because this theory can also explain strong behavioral and neurological activation s and associations between verb s and instrument s . Action verb s generally poses more functional and motoric features than visual or perceptual features, and those features are usually the main features of tools and artifacts compared to animals or fruits (see Vinson & Vigliocco, 2008 for details) . Therefore, greater verb instrument semantic features overlap may resul t in greater co activations as compared to verb patient activation. Behavioral Evidence Comparing Patient to Instrument Activations by Verbs A review of linguistic and neuropsychological literature provide s us with some reasonable suppositions of how action verbs are associated with patient or instrument role s. The psycholinguistic literature with behavioral evidence show s how verbs activate thematic roles and how that association affects language processin g. Altmann and Kamide (1999) argued that verb s generate expectancy of upcoming thematic nouns. Using a semi realistic visual scene with pictures of a target (e.g., cake ) and distractors (e.g., toys movements while the parti cipants hear d a sentence containing the verb eat or move . Because eat requir es an edible patient, participants looked at the target object significantly earlier than the move sentence Therefore, the study support s the position that verb s restrict the seman tic domain of expected items . Kamide and colleagues further showed that the verb based information not only generates expectancy of the immediately following object (i.e., patients) but also affect s later occurring objects (e.g., The woman will spread the butter on the bread ) (Kamide et al.,
23 2003) . Their first study showed the evidence of patient expectation from a verb. The second study did not use instrument nouns to support the same evidence. However, because instruments are syntactically located in simi lar distance of the PP phrased in their second study, it is likely assumed that verbs generate expectations about instruments as well . Sussman (2006) also used eye tracking methods to investigate how verbs infl uence thematic roles during sentence processing. However, unlike Altmann and Kamide, Sussman presented both patient and instrument pictures to explore which thematic role is more expected by verbs. S he also categorized verbs into two groups based on the de gree to which verbs are instrument biased ( e.g., smash is an instrument biased verb; break is a non instrument biased verb) . In E xperiment 1 and 2, she presented pictures of a target instrument (e.g., mallet ) with an upcoming patient (e.g., house ) along wi th distractors and asked participants to manipulate an item as they hear d a sentence stating Smash the house or Break the house. The results revealed more looks to instruments compared to patients at the verb + the phrases when participants hear d instr ument biased verb sentences , whereas there was no difference in looks between instrument and patient in the non instrument biased verb sentences. These results are the opposite of what Altmann and colleagues found. Sussman suggested that th ese unexpected f indings may be in part due to differences in task type. Whereas participants were asked to listen to sentences while watching the pictures in Altmann and Kamide studies, Sussman asked participants to carry out a specific action, which may have introduced a ction planning using instruments. Therefore, Sussman conducted Experiment 3 using the same task as in Altmann and Kamide, but also found the similar results in Experiment 1 and 2. Thus , she confirmed immediate effects of verb information on instrument proc essing .
24 toward instruments over patients during sentence processing. Earlier when we reviewed argument and adjunct concepts, although the argumenthood of instruments wa s still arguable, patients were clearly arguments in sentences. Therefore, patients should take advantage of there seem to be factors that tightly link verbs w ith instruments, which are even strong enough to violate syntactic restrictions. We hypothesize that their strong association is led from semantic coupling. Therefore, to strip off other factors such as syntactic components and to focus on semantic aspects , we now concentrate on single word processing. Ferretti et al. (2001) designed a priming paradigm with verb primes and four thematic role noun targets (agent, patient, instrument, and location). Noun targets in written words were presented following verb primes with 250ms delays (250ms stimulus onset asynchrony, SOA). Participants were asked to determine whether the targets were real words or not (lexical decision task) or whether the targets are animate or inanimate (animacy decision task). The results s howed that verb primes facilitate agent, patient, and instrument target words (but not locations), which supports the idea that verb s generate activations on their related concepts based on the schema. Their work is significant because it supports verbs th ematic role activations at a single word level. That is, even without enforcement of integrating information to comprehend a sentence, listeners/readers make verb thematic role association automatically. However, mainly because it was not intended in their work, they did not compare the effect size of each association and therefore, which thematic role is more strongly associated with verbs is unknown. They reported facilitation effect sizes and not all thematic role had the same degree of facilitation (ver b agent: 27ms, verb patient: 32ms, verb instrument: 32ms). Although we do not
25 know whether the differences are significant, the study alludes to different strengths of association between a verb and different types of thematic roles. On e study (Park, Cowl es, & Edmonds, 2013) attempted to answer this question directly . The authors presented highly related patient (e.g., bathtub ) and instrument (e.g., sponge ) pictures y simultaneously heard a verb (e.g., scrubbing ). Park et al. (2013) explored which pictures participant s looked at when they heard an associated verb. When they calculated total fixation time during the picture presentation time (3 seconds), there was a te ndency toward longer fixation on instruments than patients, but the difference was not significant. However, additional analyses on fixation time in each 50ms time window across the 3 seconds revealed significantly longer fixation times on instruments com pared to patients in between the 1000ms ( rough offset time of verb presentation) and 2200ms time windows. Therefore, the results revealed verb preference on highly related instruments over patients. This preference became greater with manipulation of the semantic relatedness of patients and instruments to be either highly related or less related to a verb. When highly related instruments ( sponge ) were paired with less related patients ( mirror ), the difference of fixation time between patient and instrument was significantly greater than when less related instruments ( towel ) are paired with highly related patients ( bathtub ). Therefore, they suggested that instruments are more sensitive to manipulating semantic strength than patients. Park et al. (2013) init ially hypothesized that highly related patients ( bathtub ) would be looked at longer than highly related instrument s ( sponge ). That is because the patients were obligatory from a syntactic perspective and thus must co occur with the verb in sentences. As a result, there would be an additional advantage of lexical co occurrence compared to instruments,
26 whose co occurrence is very low (0.8%) with verb presentation in sentences (Koenig et al., 2003) . With the advantage of lexical co occurrence in addition to be ing highly related to verbs, it was expected that patients would be looked at longer than instruments . However, against their hypothesis lexical co occurrence did not contribute to greater fixation time s for highly related patients . Therefore, Park et al. (2013) suggested that lexical co occurrence effects were minimal when examining the relationship between the verb and highly related patients and instruments; rather semantic relatedness effects overrode lexical co occurrence effects. One explanation for t he strong verb instrument association observed in Park et al. (2013) was thematic role filler competition . When a presented verb evokes patient and instrument fillers (i.e., specific items), individual people activate their own default filler depending on a specific schema he/she is more familiar with based on their personal experience. Although Park et al. selected very common patient and instrument fillers based on normative questionnaire data ( bathtub and sponge for scrubbing ), there are other possible f illers commonly associated with verbs (e.g., floor or toilet for patients, and brush for instruments). Intuitively, there are a lot of possible patient fillers (e.g., bathtub, floor, toilet, table for scrubbing ) but a relatively small number o f instrument fillers (e.g., sponge, brush for scrubbing ). Because an instrument role has few possible fillers, it is more likely that the target filler in experiments match individual s default filler. On the other hand, because a patient role includes many possible fi llers, there is a lower chance to activate the target filler that matches their default filler. A s imilar hypothesis was also proposed in a priming study by Park and Edmonds (2012) . The primary purpose of the study wa s to investigate whether a verb prime facilitates thematic role noun naming (patient s and instrument s ). A written word of a verb primed a picture of related patient or instrument with 50ms and 300ms SOA, and the reaction time of the naming speed was
27 compared to their corresponding unrelated p airs. Contrary to their hypothesis that verb prime s would facilitate thematic role noun naming, they reported null effects in all verb prime patient/instrument target pairs and even interference effect in the verb prime patient target pairs at 300ms SOA. H owever, interestingly they observed different trends between in verb patient pairs and in verb instrument pairs. They found a facilitation trend in verb instrument pairs at 50 and 300ms SOAs but an interference trend in verb patient pairs at 50ms SOA, and this interference trend became significant at the longer SOA (300ms). The authors argued that the different trends may be due to lexical competition that is involved in a naming task. Most previous studies that reported facilitation effects of verb prime s to noun target s implemented a perceptual task such as a lexical or animacy decision task. Those tasks necessitate the semantic level activations for their responses to the targets, whereas a naming task requires the lexical level activations in addition to the semantic activations. During the lexical level activations, the target has to inhibit competitors to be named properly, and the more competitors are activated, the greater inhibition is required (Roelofs, 1997) . Intuitively, there are more possible co mpetitors for patient fillers (e.g., tomato , onion , meat , pizza , paper , etc. for slicing ) than for instrument fillers (e.g., knife and slicer for slicing ). Because of a greater number of competitors for patients, the facilitation effects from the semantic level were removed and resulted in the interference trend; with longer SOA the interference trends increased and became significant. On the other hand, due to fewer possible competitors for instrument nouns, the facilitation effects from the semantic level were not entirely washed out and remained, albeit insignificantly. Therefore, the purpose of the current study is to investigate these potential filler competition effects and how much this competition influences eye movements toward patients and instru f the
28 verb thematic role association is affected by specific schema activat ions, then it is hypothesized that restricting to a particular schema would narrow down the number of possible thematic role fillers. C onsequently, reduc ing the number of thematic role fillers within the particular schema would promote increased expectancy of the target filler due to a decrease in the number of competitors. I n other words, providing a sp ecific schema will allow us to increase the likelihood of expectancy of the particular thematic role filler . To test this hypothesis in the current study , a specific context was present ed in advance of verb thematic role presentation to activate target sch emas. Further, w e added an experimental group and investigated this context effect on verb thematic role processing in different age groups (healthy younger and older adults) to determine potential differences across age groups. Context Effects in Typical Language Processing C ontext effect s have been shown to influence the expectancy process during language comprehension (Federmeier, Wlotko, De Ochoa Dewald, & Kutas, 2007) . Early studies found the context effect in lexical ambiguity resolution. The reordere d access model (Duffy, Morris, & Rayner, 1988) suggests that when a word has more than one meaning (e.g., bat, bank, mint ), the more dominant meaning is activated first, but later the prior context information influences the meaning activations. Therefore, when the prior context information does not match the dominant meaning and rather is congruent to the secondary meaning of the word, readers look at the ambiguous word longer than when the context information is neutral or congruent to the dominant meanin g (Binder, 2003; Binder & Morris, 1995; Binder & Rayner, 1998; Rayner, Binder, & Duffy, 1999) . Federmeier et al. (2007) more specifically investigated the context effect by manipulating the degree of constrain t on the context He bought her a pearl necklace for her birthday/collection He looked worried because he might
29 have broken his arm/collection increased N400 effect for the unexpected word in both the strongly and weakly constraining conditions but could not find the difference between the two conditions. However, instead they reported increased positivity from 500 to 900ms post stimulus onset for the unexpected word only in the strongly co nstraining condition. The context effect in their study was not observed with N400 effect; however, the P500 900 effect indicated a possible cost involved in the context effect on sentence comprehension. The context effect is greater for verbs than nouns (Gentner & Boroditsky, 2001; Kersten, 1998; Kersten & Earles, 2004; Kintsch, 2001) . According to the natural partitions/relational relativity theory (Gentner, 1981, 1982) , children learn noun concepts easily because they use nonlinguistic perceptional cues such as proximity, common fate, and boundedness. Therefore, the meaning of nouns is stable in various linguistic contexts. On the other hand, the meaning of verbs is relational, so it is not directly mapped onto the nonlinguistic perceptual categories. Th erefore, the meaning of a verb is less stable and changeable depending on various contexts. Kersten and Earles (2004) supported this theory with two memory tasks (recognition and retrieval tasks). They reported that memory for verbs was significantly bette r when the verbs were tested in the same sentences where the verbs were presented previously (e.g., The ball bounced. ) compared to when the verbs were tested in the new sentences (e.g., The quarter bounced. ). However for nouns , participants scored statisti cally the same regardless of whether the nouns were tested in the previously presented sentences (e.g., The ball bounced. ) or in the new sentences (e.g., The ball rolled. ). Kersten and Earles schema/thematic role v iews. If different sentences in their study represent the different schema for the verb, their results suggest that the verb compared to noun is more likely memorized with
30 the associated schema, including thematic roles. Therefore, when some information do es not match and thus the particular schema is not activated, it is difficult for people to retrieve the verb. A verb in many cases is associated with various schemas and each schema includes its own thematic role fillers. I f given thematic role filler s d o not fit in the schema that participants expect , it would be difficult for them to integrate the given fillers with ones they expected . Based on concrete evidence of the context effect reviewed above , we hypothesize that previously given context informati on would activate target schemas which would restrict untargeted thematic role fillers and increase likelihood of target fillers. Also, context effect s would be greater for a thematic role that has greater number of possible fillers (i.e., patient) than a role that has a smaller number of possible fillers (i.e., instrument). That is because the more possible fillers are restricted, the greater increas e of a chance activating target filler would be given. Semantic Processing and Context Effects in Aging The semantic memory system in general does not greatly decline with ag e as healthy older adults are able to understand the concepts of words or semantic features of concept s . However, age related changes may occur depending on the type of task (e.g., automati c vs. attentional process) or the degree of task demand (e.g., single word vs. complex sentence comprehension) (Burke & Shafto, 2008) . There is evidence that vocabulary knowledge increases with aging (Verhaeghen, 2003) because of more language experience, suggesting that semantic representation in older adults is richer compared to young adults (Ackerman & Rolfhus, 1999) . This would affect semantic lexical processing in older adults because it increases relative frequency for low frequency words (Burke & S hafto, 2008) . Gomez (2002) showed this frequency effect between age groups. Using a homophone priming paradigm with word fragment completion task that primed high and low
31 frequency words, they found that young adults had greater priming effects for low fre quency words whereas older adults did not show difference between high and low frequency words. Meanwhile, semantic organization has been tested using the semantic priming paradigm. A great deal of studies demonstrated no age related changes; as with you ng adults, older adults show facilitation priming effect to semantically related target (Balota, Watson, Duchek, & Ferraro, 1999; Howard, McAndrews, & Lasaga, 1981; Tree & Hirsh, 2003) . However, it is also notable that there is evidence of slow rate in pri ming effects in healthy older adults compared to young adults (Bowles & Poon, 1988; Burke, White, & Diaz, 1987; Chiarello, Chur, & Hoyer, 1985; Howard et al., 1981; Linnville, 1995) . The result of slow speed in priming supports the general slowing theories which argue that age related declines in cognitive process are due to slowing of component process (Madden, 2001; Myerson, Hale, Chen, & Lawrence, 1997; Salthouse, 1996, 2000) . Additional evidence of age related change in semantic process from the priming paradigm literature is called hyperpriming effects (Laver, 2000; Laver & Burke, 1993; Myerson et al., 1997) . Hyperpriming is a phenomenon in older adults whose facilitation effect size is greater than in younger adults. T here are a number of views to exp lain language processing in aging. The first view , which has been used to explain hyperpriming, is by Laver and Burke (1993) who argued that greater facilitation is due to more numerous direct and indirect connections between concepts for older adults due to their more extensive language experience as compared to young adults . The s econd view is explained by resource limitations such as inhibition, working memory, or reduced speed (Hasher & Zacks, 1988). Reduced i nhibition model explains poorer performance of reading or listening in older adults compared to younger adults when they are distracted (Connelly, Hasher, & Zacks, 1991; Tun, O Kane, & Wingfield, 2002) . It also has been used to
32 explain off topic conversation in older adults (e.g., Arbuckle, Nohara L eClair, & Pushkar, 2000). Working memory difficulty model in aging postulates that reduced storage or processing capacity of working memory constrains complex semantic and syntactic comprehension or production (Kemper & Kemptes, 1999). T he model of slow ed processing (Myerson et al., 1997) states that because older adults process prime s more slowly than young adults, they have a longer processing time , and this exp a nded time results in increased priming effects. Transmission deficit theory is developed by connectionists, who argue that connections between concepts in the semantic network increase with frequency of use but decrease in aging. As connection strength decreases, there is a reduc ed transmission of priming which weakens activations to connected co ncepts. This transmission deficit therefore causes greater limitation in general (connected) processing rather than single processing . T his view is also supported by neurobiological characteristics of aging. Brain atrophy in aging includes white matter atr ophy and reduced total length of the myelinated fibers of white matter, which results in reduced neural connectivity (e.g., Marner, Nyengaard, Tang, & Parkkenberg, 2003; Tang, Nyengaard, Pakkenberg, & Gundersen, 1997; Raz, 2000). Most semantic priming s tudies have focused on noun prime noun target pairs, whereas noun verb pairs have not received much attention. Hagoort (1993) is one of a few studies that used verb primes. In his experiments, two primes were presented in sequence for noun targets. The fir st prime was a noun and the second prime was a homographic homophone (words that have more than two meanings) in Dutch which has either two noun noun meanings (e.g., ball ) or noun verb meanings (e.g., bark ). Four conditions were manipulated by the prime ta rget relationship: concordant pairs (the second prime is related to the first prime and target by same meaning), discordant pairs (the second prime is related to the first prime and target in different
33 ways), neutral pairs (the second prime is related to t he target but not to the first prime), and unrelated pairs (all three words are unrelated). The results in young adults showed priming effects in all four conditions for both noun noun pairs and noun verb pairs; on the other hand, in healthy older adults, priming effects were found in all four conditions for noun noun pairs but only in concordant and neutral conditions for noun verb pairs. The author suggested that older adults necessitate both form (grammatical class match) and context (appropriate meanin g) information to inhibit inappropriate meaning whereas young adults e ffectively suppress the inappropriate meaning without additional information. Edmonds and Mizrahi (2011) also showed different priming effects in healthy older adults compared to youn g er adults. They investigated bidirectional verb thematic role noun (agent and patient) priming effects. In a young adult group, they found facilitation effects in all pairs (verb prime agent target, verb prime patient target, agent prime verb target, and patient prime verb target). However, in an older adult group, the facilitation effect in verb prime agent target disappeared. They argued that violating the canonical order decreased expectancy generated from a verb to agent in older adults. Similar to Hag oort additional information (syntactic rule) to boost expectancy process. In sum, it is still arguable whether age related changes in semantic process clearly exist. However, the evidence of slow rate and hyperpriming effect s indicate that at least in some instances there are potential differences in processing systems in older adults compared to young adult. Context Effects in Aging . If older adults process semantic information differently than young adults, the context effect we addressed above may also be affected somehow. Affected context effects in older adults have been shown with ERP studies (Federmeier & Kutas, 2005; Federmeier, McLennan, Ochoa, & Kutas, 2002) . Federmeier et al. (2002) designed high and low
34 constraint se ntences and manipulated the ending word to be an expected word, an unexpected word from the same semantic category, or unexpected word from a different category. While listening to a sentence, both young er and older adult groups elicited the greatest N400 effect in the unexpected word from the different category and least N400 effect in the expected word although overall N400 responses were smaller and delayed in older adults than in young adults. Constraint context effect s w ere different between age groups . The N400 responses between the two types of unexpected words (unexpected word from the same category vs. unexpected word from the different category) in high constraint context sentences were significant in young adults but not in older adults. Instead t he slight difference between the two types of unexpected words was found in low constraint context in older adults. That is, when the unexpected word shared semantic features with the expected word, young adults but not older adults received some benefits from it in high constraint context compared to when the unexpected word is not related to the expected word. Therefore, the authors explained that young adults actively use the context information to predict semantic features of upcoming words. However, ol der adults use context differently and their expectancy process is directly related to plausibility (whether the target is expected or not). Federmeier and Kutas (2005) further focused on constraint context effect and found similar trends. Using strongly and weakly constrained context sentences rendering the target word predictable, ERP responses were recorded while participants read the sentences . The results were compared between age groups (young vs. older adults). Overall N400 effects were delayed and smaller in older adults compared to young adults. Also, whereas young adults were sensitive to violating expected target words in strongly constraint context sentences than in weakly in both strongly and
35 weakly constraint context sentences, indicating older adults are not sensitive to the constraint context effect. They concluded that older adults are less influence d by the context constraint effect. In the two studies (Federmeier & K utas, 2005; Federmeier et al., 2002) the authors further correlated ERP responses (amplitude and timing) with language and cognitive measures such as reading span, verbal and nonverbal fluency and reported significant correlations with age (Federmeier & Ku tas, 2005) , reading span (Federmeier & Kutas, 2005) , verbal fluency (Federmeier et al., 2002) , and vocabulary (Peabody Picture Vocabulary Test (Dunn & Dunn, 1997) ) (Federmeier et al., 2002) . Reading span and verbal fluency are related to executive function especially with working memory , thus suggesting that working memory decline in aging may attribute to difficulty in holding and actively using context information. This is nicely in line with the resource limitation view that listeners/readers need to int egrate context information requiring working memory during global sentence processing. Thus far, w e have discussed semantic language processing , including context effects in aging. T here is evidence of difference s in older participants compared to youn ger participants , which might be explained by cognitive decline such as working memory deficits or slower processing speed. We are interested in the semantic system in aging because it is related not only with cognitive processing but also with lexical ret rieval. Research with individuals with dementia lose semantic concepts and therefore are unable to name the concepts . This is considered a semantic storage deficit. Meanwhile, individuals with aphasia cause d by a left hemisphere stroke show semantic acc ess impairments, which cause inability of strateg ically manipulate the semantic links in the network. These individuals show inconsistent naming errors (Jefferies, Baker, Doran, & Lambon Ralph, 2007; Jefferies & Lambon Ralph, 2006) . This phenomenon reflec ts how semantic impairments cause lexical retrieval difficulties which
36 therefore disrupt communication. In a linguistic model, the semantic system is interrelated with the lexical system and therefore affects naming ability. Therefore, understanding semant ic processing difference in healthy older adults (HOAs) compared to younger adults is important to investigate communication in HOAs. The current study investigates verb thematic role association as well as semantic context effects. The former primarily a ims to reveal how strongly concepts are connected to each other (i.e., whether patients or instruments are strongly associated with verbs) and the latter aims to demonstrate how fast or effectively context information is addressed and activates target sche mas and corresponding thematic role fillers. T here is evidence that the semantic memory system as well as the constraint context effects are affected in aging due to probably slow process speed or reduced executive function. Also unlike previous studies us ing sentence processing tasks, this study evaluates single words and concepts to minimize confounds such as syntactic manifestation in sentences while emphasizing semantic networks between concepts.
37 CHAPTER 2 CONTEXT EFFECTS IN TYPICAL LANGAUGE PROCESSIN G Introduction To investigate context effects on the expectancy process from a verb to thematic role fillers (patient and instrument), we implement ed the visual world paradigm used in Park et al. (2013) . T he visual world paradigm is a relatively recently developed method investigating language processing using eye tracking, which presents auditory input with a visual array to participants and follow s their natural eye movements. Therefore, this methodology combines aspects of scene perception, visual sear ch, and language processing (Rayner, 2009). Calculating participant eye fixations or saccadic movements on target objects is rationalized by the linking hypothesis, which assumes the link between word recognition and eye movements. This hypothesis sugges ts that auditory input shift participants attention to target pictures to make a saccadic eye movement to fixat e them (Tanenhaus & Tureswell, 2005), and eye movements reflect the direction of visual attention depending on vi sual and auditory input (Huetti g, Rommers, & Meyer, 2011). An advantage of the visual world paradig m compared to other psycholinguistic paradigms such as lexical decision or grammaticality/animacy judgments in priming paradigms, is that participants do not need to make meta linguistic j udgments, which might cause compounding in measuring the language processing investigators target to because interpretation relies on eye movement data (Huetti g, Rommers, & Meyer, 2011). Also, visual world paradigm allows researchers to present several obj ects including target, competitor and foil and to compare eye movements directly between objects in a circumstance where objects compete with each other to be looked at. I n the visual world paradigm with the same protocol in Park et al. (2013) we additiona lly prime d the verb with a visual context. As a visual context, we use d a picture of location scene. Location is one of thematic roles that reliably facilitates
38 activations of a semantically related verb (McRae et al., 2005b) . Among the four major thematic roles (agent, patient, instrument, and location), patient and instrument were excluded as context information because they are the experimental targets. Agent was excluded because not many agent fillers are appropriately presentable as a picture. Although some agent fillers are clear with a uniform (e.g., policeman ), some agent fillers such as teacher, secretary are difficult represent without environmental information (e.g., blackboard, desk ). Using the visual modality to prime context is beneficial compa red to using linguistic information because (1) visual information passes the phonological/lexical level and directly activates semantic concepts; therefore, it is faster than linguistic context effect (Carr, McCauley, Sperber, & Parmelee, 1982) ; (2) visua l context information is highly structured and stable over time (Chun, 2000) ; (3) a mental image of a concept varies in each individual based on their own personal experiences; however, visual information causes less personal bias because it is straightfor wardly given to the participants (Chun, 2000) ; and therefore, (4) the invariant information reduces the task complexity of representing the information (Gibson, 1963) . The current study include d four conditions presenting different types of context to prim e verb and patient/instrument relationships in order to investigate the context effect on the expectancy process. The neutral (no context, NC) condition does not constrain to a specific context; the congruent context (CC) condition present s a location (e.g ., kitchen ) that is congruent to the verb action (e.g., cutting ), patient (e.g., onion ), and instrument (e.g., knife ). I n the congruent patient & incongruent instrument ( CP& I I ) condition, the location (e.g., kitchen ) is congruent to the verb action (e.g., cutting ) and patient (e.g., onion ) but not to the instrument (e.g., scissors ) ; i n the incongruent patient & congruent instrument (IP & C I ) condition, the location (e.g., kitchen ) is congruent to the verb action (e.g., cutting ) and instrument (e.g., knife )
39 b ut not to the patient (e.g., hair );. In each condition, two foil pictures are also included as distractors. While four pictures were presented on a computer screen , we monitored how long participants look ed at the patient and instrument pictures. Research Questions (RQs) and Hypotheses. Specific research questions and hypotheses are addressed below. RQ1 : Are verbs are more strongly associated with either patients or instruments without context information? I.e., will fixation proportion in either patients o r instruments is significantly greater than the other in the NC condition? Hypothesis : Based on the findings of Park et al. (2013) , fixation proportion to instruments w ill be significantly greater than patients. RQ2 : Will constraining context informatio n increase the fixation proportion for patients and instruments compared to when no context is provided ? I.e., will fixation proportion in patients and instruments in the CC condition be significantly greater than in the NC condition? Hypothesis : Because restricting context narrows down the possible schema s, and thus reduces the number of possible fillers , the likelihood for the target filler activation would be increase d . Therefore, fixation proportion in both instruments and patients in the CC condition would be greater than in the NC condition. RQ2 1 : If RQ2 is true, will the constraining context effect be greater for either patient s or instrument s? I .e., w ill the increase of fixation proportion for patient s from the NC to CC condition be significantly greater /smaller for instruments? Hypothesis : Because patients retain many possible fillers compared to instruments, the fixation proportion increase would be greater for patients than instrument. RQ3 : Does incongruent context information generate untarg eted patient and instrument fillers and therefore reduce expectations of target p atient s and instrument s? I .e., Will fixation proportion in patient s decrease from the CC to the I P & C I conditions and will fixation proportion in instruments decrease from CC to the CP & II conditions? Hypothesis : Because activating a non target schema with an incongruent context generates non target patient and instrument fillers, it would decrease the likelihood for the target filler activation. Therefore, fixation proporti on in both patients and instruments in the CC condition would be greater than in the IP & CI and CP & II conditions, respectively. RQ3 1 : If RQ3 is true, will the incongruent context interference be greater for either patient s or instrument s? I .e., will the decrease of fixation proportion in patient s from the C C to
40 I P & CI conditions be significantly greater/smaller than in instruments from the CC to the CP & II conditions ? Hypothesis : With the same reasoning of the RQ2 1 hypothesis, the fixation proport ion decrease would be greater for patients than instruments. Methods Participants Thirty five healthy young adult participants from University of Florida participate d in the study. Inclusion criteria were : (1) age 18 to 30 years , (2) no history of brain injury or neurological disorder, (3) no history of learning disability, (4) no history of drug or alcohol addiction, (5) a native monolingual English speaker, and (6) right handed. All participants provide d informed consent and were compensated with eithe r a $5 Starbucks gift card or course credit. Six males and twenty nine females participated in the study. The mean age was 20.9 (sd=1.9) years old and education was 14.5 (sd=1.4) years. Stimuli Development In order to create a list of semantically relat ed patients and instruments to verbs and location, we conducted a commonness rating questionnaire adopted by Ferretti et al. (2001) . For cut in a kitchen patient relationship with four options (e.g., onion, hair, computer, paper a(n) _____ to be used to cut something in kitchen instrument relationship with four options (e.g., knife, scissors, axe, clipper ). Participants rate d the commonness using a 7 poi nt scale (1=not common, 5=very common), and the pairs whose average ratings were over 5.5 were included.
41 All verb object noun pairs were matched for written word frequency using the MRC database (Wilson, 1988) . In addition, lexical association was contro lled using BSG (backward strength) from the South Florida Association Norms (Nelson, McEvoy, & Schreiber, 1998) . We use d backward strength because the cue in the norms was a noun whereas the cue in our study is a verb and the target is a noun. Verb noun pa irs whose BSG value was less than 0.15 were included. Materials Four conditions were included: (1) neutral (no context, NC) condition with a neutral picture ( XXXXX ) and verb, semantically related patient, instrument, and two foils (2) congruent context (C C) condition with location, verb, and semantically related patient, instrument, two foils (3) congruent patient & incongruent instrument ( CP & I I ) condition with location, verb, semantically related patient, unrelated instrument, and two foils, and (4)inc ongruent patient & congruent instrument (IP & C I ) condition with location, verb, semantically unrelated patient, related instrument, and two foils . I n each condition twenty transitive action verbs were included and each verb was paired with two locations (A and B) where the action of the verb commonly occurs. Semantically related patient s and instrument s for each location were also included in the congruent condition. For the two incongruent ( CP & II and IP & C I ) conditions, the patient and instrument for location A were included as unrelated patient and instrument for location B and vice versa. Therefore, the stimuli in congruent and incongruent conditions were cross matched. R egarding the foils, one (e.g., ladle ) was semantically related to the locations ( kitchen ) but was un related to the verb ( cutting ); the other foil (e.g., towel ) was not semantically r elated to the location or verb. Inclu sion of the semantically related f oil was important to rul e out the possibility that the eye
42 movements toward the re lated patients/instruments w ere due only to the presentation of the location picture . All stimuli are listed in the Appendix . The verb stimuli were digitally recorded in present progressive form by a female native speaker of American English. Then, each ve rb was trimmed using Goldwave (v.5.53). The mean length of auditory verb stimuli was 862.8ms with the range between 650 and 1082ms. Location scene pictures and patient and instrument object pictures were black and white line drawings that were chosen from normative data (Druks, 2000; Snodgrass & Vanderwart, 1980) or Internet searches. Visual complexity across location scene pictures and across object pictures w as matched by ratings from an additional 10 participants (age and years of education matched to th e experimental group ; mean age=20.8 (sd=1.1) years and mean education=14.4 (sd=1.1) years ) . The definition of complexity the amount of detail in each picture with regard to the lines that make up each picture using a 5 point scale (1=very simple , 5=very complex) . F our object pictures were located in each corner of the screen and the distribution of the location will be counter balanced across trials. Procedures Each participant was seated in front of a 46 The SensoMotoric Inst ruments (SMI) remote RED eyetracker with a sampling rate of 120Hz and a spatial resolution of less than 0.03Â° was was between 20 25 inches. After eye movement calibration, participants read instructions on the monitor and asked if they have questions. Instructions were as below: In this experiment, you will see cross hair fixation followed by pictures on the screen and hear a word from the headphon es. You are allowed to look at anything you want on the screen. After the picture and auditory presentations, you will be randomly asked whether the word you heard corresponds to the word on the screen. Please press a keyboard button to answer the question as soon as possible.
43 If the answer is yes, please press the x key. If the answer is no, please press the z key. The pictures and words may be presented more than once. Now, please wear the headphones, and place your left hand on the keyboard and be r eady to press the button. Six practice trials were given before the experimental trials. In total, each participant was presented with a total of 320 trials ( 4 0 experimental trials x 4 conditions with 160 foils). The experimental and foil trials were pseu do randomly divided into 4 blocks. The number of experimental and foil trials were counter balanced across blocks, and all trials in each block were presented in a random order. During each trial, a cross hair fixation appear ed in order to begin the trial . P articipants were asked to look at the fixation for 1 second to trigger the trial . Then, a picture of the location or a symbol screen appear ed for 2 5 0ms (Biederman, Mezzanotte, & Rabinowitz, 1982; Biederman, Teitelbaum, & Mezzanotte, 1983; Chun, 2000) . T hen there was a black screen with a second fixation for 100ms. After that the auditory verb stimulus was presented simultaneously with four object pictures , which remained for 3 seconds on the screen. Lastly, a yes/no question was presented only after expe rimental trials. Example presentation in each condition is illustrated in Figure 2 1. The entire experimental duration was approximately an hour and t hree 5 minutes breaks were given between blocks.
44 Figure 2 1. Example set of cutting & kitchen context in each condition in experimental presentation . Note: NC=Neutral context, CC=Congruent context, CP & II=Congruent patient and incongruent instrument , IP & CI=Incongruent patient and congruent instrument Statistical A nalysis Total fixation time analysis: Eac h box at the corner of the screen was designated as an area of interest (AOI) to calculate fixation duration. Mean percentage of the total fixation time in each AOI was calculated across participants for each trial. Then, the mean percentage s of each patie nt , instrument , foil1 and foil2 AOI were compared using a 4 (conditions) by 4 (pictures) repeated measure ANOVA with Bonferroni post hoc corrections .
45 Time course analysis: The mean percentage s of fixation in each AOI were calculated across participant s for each trial in each 10 0ms time window. A 4 (pictures) by 30 (time windows) repeated measure ANOVA was conducted in each condition with Sidak post hoc tests . Results Total Fixation Time Results showed that there was a main effect of picture items [ F (3, 37) =102.62, p =.000], but no main effect of condition [ F (3, 37)=2.02, p =.128]. And i nteraction between condition and picture was not significant [ F (3, 37)=.798, p =.621]. Across all conditions, instruments were looked at significantly longer than patients and f oils regardless of whether the instruments were congruent to the location or not. The mean percentages of fixation time in each AOI in each condition are shown in Table 2 1. With comparisons between picture items in each condition, the NC conditions showe d that instruments were looked at significantly longer than patients [ t (39)=4.12, p =.001 ] , foil 1 [ t (39)=9.66, p =.000 ] and foil 2 [ t (39)=8.93, p =.000 ] . P atients were looked significantly longer than foil 1 [ t (39)=3.48, p =.008 ] and foil 2 [ t (39)=3.74, p =.00 4 ] . In the CC condition s , instruments were looked at significantly longer than patients [ t (39)=4.07, p =.001 ] , foil 1 [ t (39)=8.49, p =.000 ] , and foil 2 [ t (39)=7.73, p =.000 ] . P atients were looked at significantly longer than foil 1 [ t (40)=4.14, p =.001 ] and fo il 2 [ t (40)=5.30, p =.000 ] . In the CP & II condition s , instruments were looked at significantly longer than patients [ t (39)=3.05, p =.025 ] , foil 1 [ t (39)=6.29, p =.000 ] , and foil 2 [ t (39)=5.91, p =.000 ] . P atients were looked at significantly longer than foil 1 [ t (39)=3.99, p =.002 ] and foil 2 [ t (39)=4.10, p =.001 ] . In the IP & CI con dition s , instruments were looked at significantly longer than patients [ t (39)=4.10, p =.001 ] , foil 1 [ t (39)=8.36, p =.000 ] and foil 2 [ t (39)=7.78, p =.000 ] . P atients were looked signific antly longer than foil 2 [ t (39)=3.36, p =.010 ] but not foil 1 [ t (39)=2.69, p =.063 ] .
46 With comparisons of each picture item across conditions, no AOI was significantly different across all four conditions ( p >.05). Table 2 1. Mean (sd) proportions of total f ixation time proportion (%) in younger participants. Mean (sd) p value vs. I vs. F oil 1 vs. F oil 2 Condition 1: Neutral condition (NC) Patient Instrument Foil 1 Foil 2 15.03 (6.14) 22.21 (6.24) 10.72 (3.26) 10.08 (4.22) .001 .008 .000 .004 .000 Condition 2: Congruent context condition (CC) Patient Instrument Foil 1 Foil 2 15.06 (5.24) 22.12 (7.25) 10.13 (3.70) 9.61 (4.02) .001 .001 .000 .000 .000 Condition 3: C ongruent patient & Inc ongruent instrument ( C P & I I) Patient Instrument Foil 1 Fo il 2 15.38 (4.62) 19.67 (6.11) 11.05 (3.67) 10.90 (4.16) .025 .002 .000 .001 .000 Condition 4: Inc ongruent patient & C ongruent instrument ( I P & C I) Patient Instrument Foil 1 Foil 2 14.58 (6.53) 22.19 (7.00) 10.87 (3.54) 10.14 (3.98) .001 .0 63 .000 .0 1 0 .000 Note: I=Instrument, F oil 1= semantically related to context , F oil 2= unrelated Time Course Analysis In the NC condition s , fixations on instruments became significantly greater than patients after 1500ms , foil 1 after 90 0ms, and foil 2 after 10 00ms. Fixations on patients were significantly greater than foil 1 at 1200 1400ms and 2600 3000ms ; and greater than foil 2 after 1 200 ms ( p <.05 for all). In the CC condition s , fixations on instruments became significantly greater than patients after 1 30 0ms, foil 1 after 1000ms , and foil 2 after 1000 ms. Fixations on patients were significantly greater than foil 1 after 1400ms ; and greater than foil 2 after 1000 ms ( p <.05 for all).
47 In the CP & II condition s , fixations on instruments became significantly greater than patients at 1300ms, 1800ms, 2100ms, and 2500 3000ms ; greater than foil 1 after 1200ms , and foil 2 after 1100ms . Fixations on patients were significantly greater than foil 1 after 1800ms ; and greater than foil 2 at 1400 1500ms and 1900 3000ms ( p <.05 for al l). In the IP & CI condition s , fixations on instruments became significantly greater than patients after 1500 ms, foil 1 after 900 ms, and foil 2 after 90 0ms. Fixations on patients were significantly greater than foil 1 at 1900 3000ms, 1100ms, 1400ms, and af ter 2400ms ( p <.05 for all). Figure 2 1 show s AOI fixations at every 50ms time window in each condition. A. Figure 2 2 . Fixation proportion s of each AOI in each time window ( 10 0ms) in younger participants . A) Condition1: NC condition. B) Condition2: CC co ndition. C) Condition 3: CP & II condition. D) Condition 4: IC & CP condition. Note: I=Instrument, P=Patient, * p <.05
48 B. C. D. Figure 2 2. Continued .
49 Interim Discussion With regard to RQ1 , which compar ed verb instrument and verb patient associatio n s, our hypothesis of stronger verb instrument compared to verb patient associations was supported by significantly longer fixations on instruments than patients in the NC and CC conditions. This finding supports Park and collegues results ( Park et al., 2 013) . In their study, when highly related patients and instruments were preceded by verbs, longer fixations were observed in instruments as compared to patients . However, since the difference between patients and instruments in their study was not signific ant in total fixation , the significant results in the current study suggest a stronger association between verb s and instrument s than verbs and patients . The difference between studies may be due to more sensitive AOIs in the current study. T hat is, in Par k et al (2013) AOIs of pictures were located in each quadrant in a 2x2 array, which covered the entire computer screen. Therefore, it i s possible that fixations in the center of the screen were included in the fixation calculations . In the current study th e AOIs of the four pictures were located at the corner of the screen in a 3x3 array, and only the fixations intentionally fixated on in the AOIs were calculated . A re analysis of Park et al. (2013) data and f urther studies may be able to support this hypot hesis. For RQ2 and RQ3 , we hypothesized that congruent context information would enhance activations of target instruments and /or patients compared to when no context or incongruent context was presented. Thus, w e expected significantly longer fixations on both instruments and patients in the CC condition compared to the NC , and longer instrument fixations on the IP & CI condition and longer patient fixations on the CP & II condition. However, i n all four conditions the fixation time proportion of patients and instruments was statistically the same across conditions, with significantly longer looking times for instruments in all conditions.
50 Although we did not find a total fixation time difference between patients and instruments across conditions, we found some subtle evidence of context effects in the time course analysis. First, the time when significantly different fixations between patients and instruments were facilitated in the CC condition was earlier (1200ms) than in the NC condition ( 1400ms ). Consid ering the range of auditory verb presentation time was 650 to 1082ms in the CC condition, instruments were looked at significantly longer than patients shortly after the verb disappeared. This suggests that the congruent context sped up the selection of th e target schema and guide d participants to look at corresponding instruments. Thus, congruent context in younger adults may not largely affect the overall time of fixations on instruments , but it may influence speed of fixations toward instruments . Evide nce of incongruent context effects was suggested by the results . Table 2 1 shows that overall fixation in each of the four AOIs w as almost identical across the four conditions. However, there was some difference noted in the CP & II conditions, where the p roportion of fixations in instruments (19.67%) was significantly smaller than the mean fixations of instruments in the three conditions that presented no context or congruent context to instruments (mean=22.18% of 22.21% in the NC, 22.12% in the CC and 22. 19% in the IP & CI condition) [ t (39)= 2.200, p =.034]. This difference was not significant in our ANOVA analysis may be because of Bonferroni corrections . This indicates that the instrument fixation was significantly reduced in condition 3 due to some degre e of incongruent context effects on instruments. Time course analysis data also show ed slightly facilitated patient fixations and decreased instrument fixations in the CP & II condition. The number of time windows that instruments were significantly longe r fixated than patients was smaller in the CP & II compared to the CC condition. Also, the patient fixation slopes in other condition s seemed flat across 3 seconds,
51 whereas the patient fixations in the CP & II conditions slightly increased with time. Becau se the context was not congruent to instruments but congruent to patients in the CP & II condition s , it is possible that the incongruent context interfer es with target instruments and reduced fixations, but the congruent context to patients somewhat facili tated target patients and increased fixations slightly. In addition, subtle evidence of incongruent context effects also appeared in the IP & CI condition s . Regarding total fixation time s in the IP & CI conditions , instruments were looked at significantly longer than foil 1 and 2, but patients were looked at significantly longer than only foil 2 but not foil 1. R ecall that our foil 1 items were semantically associated with location contexts but not with verbs . T herefore, it is not surprising that foil 1 co uld be looked at longer than foil 2 because it is primed by the context. However, because foil 1 is not primed by verbs or the context, it should be looked at less than patients and instruments. The fact that patients were looked at as much as foil 1 in th e IP & CI condition indicate s reduced patient association with verbs by incongruent context presentation . The time course analysis also shows a similar pattern. In sum, w e found evidence to support stronger verb instrument associations compared to verb pa tient associations, similar to what was observed in Park et al. (2013) . Although between condition comparisons did not support the proposition that context effects would highly activate target patients and instrument s corresponding to the congruent context schema, there was evidence of sped time at which instruments were significantly longer fixated than patients. We also observed evidence of incongruent context effects with reduc ed fixations on the thematic role that is not congruent to the context a s well as subtle increas es in fixations on the thematic role congruent to the context. While the patterns observed in the time course analyses were subtle, it
52 is important to remember that the only difference across conditions was the context picture. Thus, the different patterns observed are likely due to the context. The n ext chapters investigate the same research questions we addressed here but with healthy older adults to investigate whether their semantic processing of verb thematic role activations is diff erent from younger participants.
53 CHAPTER 3 CONTEXT EFECTS IN AGING Introduction We implement ed the same described in Chapter 2 to a healthy older population to investigate whether constraining the context influences verb to thematic role filler activat ions in the same m an ner with an older group as compared to the younger group . There is no consensus on how to group older adults into different a ge classification s . Researchers commonly use the off age varies (Ferraro, 1980; Rosenwaike, 1985; Suzm an & Riley, 1985) . In the current study, we include d e se groups have shown significant language and cognitive declines compared to young adults (Faustmann, Murdoch, Finnigan, & Copland, 2007; Federmeier & Kutas, 2005; L ee & Federmeier, 2011) . However, due to difficulty recruiting HOA participants in th is limited age range, we expanded the age range to between 60 and 84 years. In Chapter 1, we reviewed three explanations for semantic processing differences observed in older adults: (1) greater language experience in older adults compared to young adults, (2) recourse limitation of working memory or processing speed , and (3) transmission deficit . Because there is no clear consensus that any of the se hypotheses explains s emantic processing in aging, it is our goal to find out whether there are differences across age groups and if so, which hypothesis may explain any observed difference . We have the same three research questions as in Chapter 2, and our hypotheses on each r esearch question were driven by the aging models.
54 Research Questions (RQs) and Hypotheses. Specific research questions and hypotheses are addressed below. RQ1 : Are verbs more strongly associated with either patients or instruments without context informat ion? I.e., w ill fixation proportion in either patients or instruments be significantly greater in the NC condition? Hypothesis 1 1 : If the greater language experience model or slower processing speed model is supported, as in Chapter 2 with young partici pants, fixation proportion for instruments will be significantly greater than for patients. Hypothesis 1 2 : If the transmission deficit theory or working memory limitation model is supported, the fixation proportion in both patients and instruments will be reduced, and might even lose significant fixation differences between patients and instruments. RQ2 : Will constraining context information increase the fixation proportion for patients and instruments compared to when no context is provided ? I.e., w il l fixation proportion in patients and instruments in the CC condition be significantly greater than in the NC condition? Hypothesis 2 1 : If the greater language experience model is supported, connections between verb and target fillers will be strongly ac tivated by context information. Therefore as in Chapter 2 with younger participants, the target filler activations would be increase d by the target schema . Therefore, fixation proportion in both instruments and patients in the CC condition would be signifi cantly greater than in the NC condition. Hypothesis 2 2 : If the slower speed model is supported, verb target filler connections would be affected by context information as in Hypothesis 2 1 . However, due to slower speed in processing this information, ove rall fixation time would be slowed in the CC compared to NC condition. Hypothesis 2 3 : If the transmission deficit theory or working memory limitation model is supported, connections between verb, target fillers, and context would be weak or unable to be integrated . Therefore, context effects would be minimal or not observed. Therefore, fixation proportion in both instruments and patients in the CC condition would not be significantly greater than in the NC condition. RQ3 : Will incongruent context inform ation activate untargeted patient and instrument fillers and therefore reduce expectations of target p atient s and instrument s that are presented? I .e., w ill fixation proportion in patient s decrease from the CC to the I P & C I conditions and will fixation pr oportion in instruments decrease from CC to the CP & II conditions? Hypothesis 3 1 : If the greater language experience model is supported, with the same reason in Hypothesis 2 1 , the fixation proportion in both instruments and patients in the CC condition would be significantly greater than in the IP & CI and CP & II conditions, respectively.
55 Hypothesis 3 2 : If the slower speed model is supported, verb target filler connections would be affected by context information as in Hypothesis 3 1 . However, due to slower speed in processing th is information, significantly increased fixation time in the CC compared to the IP & CI and CP & II conditions would be delayed. Hypothesis 3 3 : If the transmission deficit theory or working memory limitation model is suppor ted, with the same reason in Hypothesis 2 3 , fixation proportion in both instruments and patients in the CC condition would not be significantly greater than in the IP & CI and the CP & II conditions, respectively. Research questions and hypotheses depend ing on each model are summarized in Table 3 1. Methods Participants Seventeen healthy older adult participants were recruited from the Gainesville, Florida community. Inclusion criteria were : (1) ages 6 0 to 84, (2) no history of brain injury or neurologic al disorder, (3) no history of learning disability, (4) no history of drug or alcohol addiction, (5) a native monolingual English speaker, (6) right handed, and (7) normal vision with corrective lenses and hearing with hearing aids. All participants provid e d informed consent (UF IRB #201300670) and were compensated with a $5 Starbucks gift card. We included 8 male participants and 9 female participants. The mean of age was 67. 35 (sd=5. 59 ) years old and education was 17. 06 (SD=2.8 4 ) years. Individual partici demographic information and language/cognitive test scores are shown in Table 3 2 .
56 Table 3 1. Summary of Research Questions and Hypotheses Based on Models of Language Processing in Aging Greater Experience Model Slower Processing Model Transmission Deficit Theory Working Memory Limitation RQ1. Patients vs. Instruments Instruments will be looke d at significantly longer than patients . Instruments will be looked at significantly longer than patients . Instruments will be looked at longer but not significantly RQ2. Congruent context vs. No context Both patient and instrument fixation time will be longer in the CC than NC condition. Both patient and instrument fixation time will be longer in the CC than NC condition, but the patterns will appear later time. Patient and instrument fixation time in the CC will be similar to the NC condition. RQ3. Co ngruent context vs. Incongruent context Patient fixation time will be longer in the CC than IP & CI condition. Instrument fixation time will be longer in the CC than CP & II condition. Patient fixation time will be longer in the CC than IP & CI conditio n, but the patterns will appear later time. Instrument fixation time will be longer in the CC than CP & II condition, but the patterns will appear later time. Patient fixation time in the CC will be similar to the IP & CI condition. Instrument fixatio n time in the CC will be similar to the CP & II condition.
57 Language and Cognitive T ests We use d the Cognitive Linguistic Quick Test (CLQT) (Helm Esta brooks, 2001) as a cognitive and linguistic deficit screening test. The CLQT consists of 10 subtests which assess five cognitive domains: attention (e.g., symbol cancellation and symbol trails tests), memory (e.g., story retelling and design memory tests), executive functions (e.g., mazes and symbol trails tests), language (e.g., confrontation naming, story retelling, and generative naming tests), and visuospatial skills (e.g., symbol cancellation, clock drawing, and mazes). In addition to the screening te sts, we measure d language and cognitive processing more in depth with separate tests. We included two domains of language and cognitive tests: (1) semantic association tests and (2) short term or working memory tests. This is, because (1) our study measure s semantic association between verbs and thematic roles and therefore we needed p reserved, and (2) our study present sequential stimuli including a location context, verb, and four pictures, which r equires short term and working memory support to integrate and process all information. Semantic association tests include the Pyramids and Palm Trees (PPT) (D. Howard & Patterson, 1992) and Kissing and Dancing tests (KD T ) (Bak & Hodges, 2003) . The two te sts require participants to choose semantic associations between words and pictures of nouns (PPT) and verbs (KD T ). In both tests, three words or pictures are presented (one on top and two on the bottom), and the participants must choose which word/picture presented on the bottom (e.g., tongue and heart) is semantically related to the word/picture presented on top (e.g., stethoscope). We test ed the two tests in the eye tracking setting.
58 Table 3 2 . Demographic information and language and cognitive test sc ores Sex Age (yrs) Edu (yrs) CLQT Semantic Association STM / WM A M EF L VS O PPT KDT Semantic STM % Digit spans F / B Symbol spans HOA01 F 71 16 4 4 4 4 4 4 . . 7 7 / 6 . HOA02 F 70 16 4 4 4 4 4 4 51 49 . 7 / 6 32 HOA03 M 60 16 4 4 4 4 4 4 48* 51 7 7 / 6 24 HOA04 F 65 18 4 4 4 4 4 4 48* 49 7 7 / 6 18 HOA05 M 77 18 4 4 4 4 4 4 51 48 7 8 / 4 24 HOA06 F 64 12 4 4 4 4 4 4 48* 48 7 6 / 4 23 HOA07 M 72 18 4 4 4 4 4 4 45* 45* 7 10 / 8 20 HOA08 M 70 13 4 4 4 4 4 4 51 46* 7 12 / 9 22 H OA09 M 69 19 4 3 4 4 4 3.8 50 49 6.4 7 / 4 18 HOA10 M 60 17 4 2 4 4 4 3.6 47* 48 7 8 / 6 17 HOA11 M 61 25 4 4 4 4 4 4 50 48 7 7 / 7 27 HOA12 F 65 16 4 4 4 4 4 4 50 50 7 10 / 9 24 HOA13 F 64 16 4 4 4 4 4 4 49* 49 7 8 / 9 22 HOA14 M 62 16 4 4 4 4 4 4 47* 44* 7 6 / 8 25 HOA15 F 64 20 4 4 4 4 4 4 . . 7 12 / 9 33 HOA16 F 75 18 4 4 4 4 4 4 51 49 7 14 / 9 17 HOA17 F 76 16 4 4 4 4 4 4 48* 46* 7 12 / 6 21 Average M=8 F=9 67.35 (5.59) 17.06 (2.84) 4.0 (0) 3.82 (.53) 4.0 (0) 4.0 (0) 4.0 (0) 3.96 (. 11) 48.93 (1.83) 47.93 (1.91) 7 (0) 8.71/6.82 (2.47/1.85) 22.94 (4.77) A=Attention, M=Memory, EF=Executive Function, L=Language, VS=Visuospatial, O = Overall PPT=Pyramid and Palm Tree test, KDT=Kissing and Dancing Test, STM=Short term memory, WM=Working mem ory *asterisk indicates scores below normal range. CLQT score: 4 .0 3.5 = within normal limit , 3 .4 2.5 = mild , 2 .4 1.5 = moderate , 1 .4 1.0 = severe . Normal range cutoff: PPT=50, KDT=48, Semantic STM=4.64, Digit span forward=5, Digit span backward=3, Symbol span= 11 1 1 Control norm for dementia population (aged 55 75) in the subset of symbol span in Wechsler (1997)
59 Short term and working memory tests included the forward and backward digit spans and the symbol spans from Wechsler Memory Scale (WMS III) (Wechsler, 1997) . Forward and backward digit spans tasks are commonly used to test working memory function. Pa rticipants were asked to repeat the several numbers with the same order (forward) or opposite order (backward). The symbol spans is similar to forward digit spans except it uses symbol deigns rather than numbers. The Semantic Short term Memory Test (seman tic probe subtest) (Kalinyak Fliszar, Kohen, & Martin, 2011) was also included. In this test, participants listen to series of words and determine whether a probe word is in the same category of any of previously presented words for that series of words. L ike span tests, the number of words increases from 1 word to 7 words. This test measures not only short term memory ability but also semantic association processing between concepts. Materials and P rocedures A fter receiving consent, all HOA participants were tested with vision and hearing screening tests. We showed participants several letters that are approximately the same font size of our stimuli in the experiment with the similar distance of participant to the computer screen, and asked them to read t he letters with their glasses or contact lenses, if necessar y. For the hearing screening, we played five recorded words with the similar volume of auditory stimuli in the experiment and semantic short term memory test and ask ed participants to repeat the w ords they heard. P articipants continued in the study if they were able to read and repeat all the letters and words . After these screening tests, we administered the CLQT. We only included participants whose overall CLQT score was within the normal range. All 17 participants passed the vision and hearing screening tests and the CLQT (see Table 3 1).
60 Next, participants were administered the experiment al task . The m aterials and experimental procedures were the same as those used for the younger adults . Afte r participants completed the experimental task, they finished the other language and cognitive tests. Participants received breaks between tests as needed or requested . T he experiment and language / cognitive tests were completed in a pproximate ly 2.5 hours i n one session . Statistical A nalysis Statistical analysis was the same as in Chapter 2. Results Total Fixation Time Results showed that there was a main effect of picture items [ F (3, 37)=29.37, p =.000], but no main effect of condition [ F (3, 37)=.315, p =.815]. And interaction between condition and picture was not significant [ F (3, 37)=.581, p =.803]. The mean percentages of fixation time in each AOI in each condition are shown in Table 3 3 . In the NC condition , the AOIs with the instruments were looked a t significantly longer than foil 1 [ t (39)=4.94, p =.000 ] and foil 2 [ t (39)=6.10, p =.000 ] , but not patients [ ( t (39)=2.10, p =.277). P atients were looked at significantly longer than foil 2 [ t (39)=2.90, p =.034 ] but not foil 1 [ t (39)=2.47, p =.125 ] . In the CC co ndition, instruments were looked at significantly longer than foil 1 [ t (39)=6.29, p =.000 ] and foil 2 [ t (39)=5.79, p =.000 ] , but not patients [ t (39)=2.86, p =.119 ] . P atients were looked significantly longer than foil 2 [ t (39)=4.33, p =.007 ] but not foil 1 [ t (3 9)=3.31, p =.106 ] . In the CP & II conditions, instruments were looked at significantly longer than patients [ t (39)=3.25, p =.015 ] , foil 1 [ t (39)=3.86, p =.008 ] , and foil 2 [ t (39)=4.22, p =.001 ] . P atients were not looked at significantly longer than foil 1 [ t (3 9)=1.43, p =1.000 ] and foil 2 [ t (39)=1.42, p =.653 ] . In the IP & CI conditions, instruments were looked at significantly longer than foil 1 [ t (39)=5.72, p =.000 ] , foil 2 [ t (39)=5.56, p =.000 ] , but not than patients [ t (39)=3.12,
61 p =.100 ] . P atients were not looke d at significantly longer than foil 1 [ t (39)=1.58, p =.607 ] and foil 2 [ t (39)=1.92, p =.261 ] . With comparisons of each picture across conditions, no picture was significantly different across all four conditions ( p >.05). Table 3 3 . Mean (sd) proportion s of total fixation time in HOAs Mean (sd) p value vs. I vs. F oil 1 vs. F oil 2 Condition 1: Neutral condition (NC) Patient Instrument Foil 1 Foil 2 10.12 (4.32) 12.38 (4.28) 8.06 (3.01) 7.17 (3.25) .277 .125 .000 .034 .000 Condition 2: Cong ruent context condition (CC) Patient Instrument Foil 1 Foil 2 9.76(3.17) 12.18 (4.00) 7.97 (2.66) 7.30 (3.11) .119 .106 .000 .007 .000 Condition 3: C ongruent patient & Inc ongruent instrument ( C P & I I) Patient Instrument Foil 1 Foil 2 9.38 (3.35) 12.20 (4.60) 8.43 (0.40) 7.92 (3.12) .015 1.000 .008 .653 .001 Condition 4: Inc ongruent patient & C ongruent instrument ( I P & C I) Patient Instrument Foil 1 Foil 2 9.62 (3.88) 12.16 (3.70) 8.04 (2.92) 7.95 (2.73) .100 .607 .000 .2 61 .000 Note: I=Instrument, F oil 1= semantically related to context , F oil 2= unrelated Time Course Analysis In the NC condition, fixations on instruments became significantly greater than patients at 2000 2100ms , greater than foil 1 at 1000 2200ms , and gre ater than foil 2 at 1300ms, and 1500 2500ms . Fixations on patients were not significantly greater than foil 1 at none of time windows; and foil 2 at 2200 2400ms ( p <.05 for all). In the CC condition, fixations on instruments became significantly greater tha n patients at 1900ms and 2100 2400ms ; foil 1 at 1300 2800ms ; and foil 2 after 1400ms . Fixations on patients
62 were significantly greater than foil 1 at 1300ms, and 1600 1700ms ; and greater than foil 2 at 1800 2100ms ( p <.05 for all). In the CP & II condition, fixations on instruments became significantly greater than patients at 1100 1700ms ; greater than foil 1 at 1100 1800ms and 2800ms ; and foil 2 at 1000 2200ms and 3000ms . Fixations on patients were not significantly greater than foil 1 neither foil 2 in any time windows ( p >.05 for all ) . In the IP & CI condition, fixations on instruments became significantly greater than patients at 1800ms ; foil 1 at 1300 1400ms, 1800 2000ms, and 2300 2800ms ; and foil 2 at 1200 2800ms (except at 1400ms and 1700ms) ( p < .05 for all). Fixations on patients were not significantly greater than foil 1 or foil 2 ( p >.05 for all) . Figure 2 1 showed AOI fixations in every 50ms time windows in each condition. A. Figure 3 1. Fixation proportion s of each AOI in each time window ( 10 0ms) in HOAs . A) Condition1: NC condition. B) Condition2: CC condition. C) Condition 3: CP & II condition. D) Condition 4: IC & CP condition. Note: I=Instrument, P=Patient, * p <.05
63 B. C. D. Figure 3 1. C ontinued .
64 Interim Discussion In regard to RQ1 , althou gh we did not find significant differences in the patient and instrument fixation comparisons, the fact that instrument fixations but not patient fixations were significantly longer than foils may indirectly suggest stronger verb instrument associations co mpared to verb patient associations. Also, overall reduced thematic role fixations may indicate weakened semantic association s between verbs and thematic roles. That is, instruments were looked longer than patients but the difference was not significant . A lso, instruments exhibited significantly longer fixations than the two types of foils . However, patients were not looked at longer than foil 1 (semantically related to context) , suggesting that the semantic association between verb s and patient s is not bei ng responded to. Our hypotheses about congruent context effects compared to no context (RQ2) and incongruent context (RQ3) were not supported, as we did not find any significant difference s in between condition comparisons in the total fixation time analys es. However, the time course analyses provided conceivable evidence of context influencing eye movements . To discuss fixation patterns in each condition, we designated three time zones based on what we observed in approximately the time course results in H OAs. Unlike the time course results in younger adults where time windows that show instrument fixations were significantly greater than patients throughout the picture presentation after verb offset, those time windows in HOAs only appeared partially tho se time windows in one condition appeared very early after verb offset and in some conditions appeared later time after verb offset. Therefore, to make interpretation simpler and more understandable , we divided the time course into three (700ms duration af ter verb offset). The first zone (zone 1) is from after the mean of verb presentation offset at 900ms to 1600ms; the second zone (zone 2) is from 1600ms to 2300ms; and the third zone (zone 3) is from 2300ms to 3000ms when the picture presentation ends. We will use these time zones to explain the
65 fixation patterns in this interim discussion and also to explain the mechanism of the patterns in the general discussion in Chapter 5. We acknowledge that using these time zones is somewhat arbitrary and is only for convenience of data interpretation. Our observation based on these time zones could be preliminary evidence for future studies that investigate the timeline of cognitive processing in language comprehension . For the context vs. no context effects (RQ2), the number of time windows that showed significant differences between patients and instruments increased with congruent context presentation (i.e., 200ms duration in the NC condition vs. 500ms duration in the CC condition). This suggests that the congrue nt context increased instrument fixation time. One interesting observation, however, between the NC and CC conditions is that the initial time when instrument fixations were significantly longer than the two types of foils was slower in the CC condition ( 1 300 ms with f oil 1 and 1400 ms with f oil 2) than the NC condition ( 1000 ms with f oil 1 and 1300 ms with f oil 2). Looking at graphs ( a ) and ( b ) in Figure 3 1, it is notable that the line indicating fixation on instruments in the NC condition started diverging f rom the patient and two foils in zone 1 (900 1600ms) and then it progressively increased . O n the other hand, the line indicating instrument fixation in the CC condition was integrated with patient and foil fixations in the beginning in zone 1 and then it s uddenly increased in the middle of zone 1 ( at around 1 30 0ms ) . The pattern in the CC condition may be due to the additional processes integrating the location context into verb and thematic role concepts within the target schema. Th ese additional processes might have slowed down fixations on the instruments . However, once the participants processed the context information , facilitation occurred from verbs and contexts to instruments. We will discuss more about the proposed mechanisms for this in the general discussion in Chapter 5.
66 For the incongruent context effects (RQ3), we interestingly found that context facilitated incongruent instrument fixations and decreased congruent instruments fixations, which is opposite to our hypothesis . In the total fixation t ime analysis , the patient instrument difference was significant only in the CP & II conditions. That is, when context was congruent to patients, the verb instrument association appeared to be boosted as evidenced by increased fixation time to instruments. A similar pattern was seen in Figure 3 1. Comparing graph b (CC condition) with c (CP & II conditions), instrument fixations were significantly longer than patients in most of zone 1 in in the CP & II conditions whereas no time window showed significant di fference between instrument and patients in zone 1 in the CC condition. . However, in the beginning of zone 2 (at 1800ms) in the CP & II condition, instrument fixations decreased and became not significantly different from patients. Thus, when the location context was incongruent to instruments , it rather increased instrument fixations in the beginning , but decreased after certain amount of time . In this circumstance, patients did not receive the benefits of a congruent context . P atient fixations were almos t identical to foil fixations, and none of the time windows showed significant differences between patients and foils (both foil 1 and 2). Similar to patients in the CP & II condition, instruments also showed a disadvantage of being congruent to context in the IP & CI conditions. Instruments were not looked at significantly longer than patients in zone 1 , and they were looked at longer at only 100ms duration in zone 2 . On the other hand, i ncongruent patients did not seem affected by context . To summari ze, we found a tendency of stronger association between verbs and related instruments compared to verbs and related patients. Congruent contexts increased fixations on instruments compared to when no context was given, but the initial time when instrument fixations became significantly greater than foils was slowed . Incongruent contexts showed a
67 dissociated influence between patients and instruments. That is, context increased incongruent instrument fixations and decreased congruent instrument fixations , wh ereas patients were not affected by any type of context. Between older adult and younger adult groups, there were common and distinctive aspects in regard to verb thematic role association and context effects. Both groups showed the pattern of stronger ve rb instrument associations compared to verb patients; however, th is association was significant ly shown in younger adults, but not in older adults. Context influence d eye movements in both the groups, but the aspects were different between the groups. The c ongruent context for both patients and instruments facilitated the initial time at which instrument fixations were significantly longer than patients or foils in younger adults, but it slowed facilitation in instruments older adults. Also, in younger adul ts , context decreased incongruent patient and instrument fixations, but in older adults it increased incongruent instruments with no affect either congruent or incongruent patients. The overall strength of verb thematic role association appeared to be red uced in older adults compared to younger adults, because the significant difference between patients and instruments in younger adults disappeared in older adults. We will come back to discuss the possible mechanisms of verb thematic role association reduc tion in Chapter 5. However, before such a discussion, we need to ensure that this difference is statistically significant. Therefore, in the next chapter, we evaluate mean difference comparisons between groups. In addition, we noticed that the total time f ixation in each AOI was smaller in older adults as compared to younger adults. We also explore whether the total fixation proportion in older adults was significantly less than younger adults, and if so, why. Therefore, the number of fixations in each AOI (fixation counts) and the initial fixation time in each AOI (entry time) were analyzed.
68 CHAPTER 4 VERB THEMATIC ROLE ASSOCIATION AND CONTEXT EFFECT COMPARISONS BETWEEN THE TWO AGE GROUPS Introduction In Chapter 2 and 3, we investigated the strength of verb thematic role association and how context affects processing for patient and instruments in a younger and older group. Although in Chapter 3 we described different fixation time patterns between the groups, we could not statistically support the diffe rence. Therefore, the purpose of Chapter 4 is to investigate whether the difference between patient and instrument fixation proportion is significantly different between the groups. Additionally, we reported that older adults looked at AOIs less than young er adults. Therefore, another purpose is to find out whether the reduced fixations are significant, and if so, what caused this. With this purpose, we conducted additional analyses with the data we obtained in Chapter 2 and 3: (1) mean total fixation time difference between AOIs between groups, (2) total fixation time proportion comparisons between groups, (3) total fixation count comparisons between groups, and (4) entry time comparisons between groups. The last two analyses were conducted to understand t he reason for shorter fixation times in AOIs in older adults compared to younger adults, if any. That is, if a significantly reduced fixation is found in older adults, we will try to understand if their search pattern was different (e.g., whether they freq uently moved fixations on each AOI or stayed in a few spots). An entry time analysis will show how fast eye gazes were fixated in AOIs, which would be related to speed of eye movements. Methods Statistical Ana l ysis . For the mean difference (effect size) comparisons, we subtracted the total fixation time of one AOI from another AOI. For example, for patient vs. instrument comparisons, we subtracted fixation time of patients from fixation time of instruments. We
69 therefore calculated five pair comparisons: i nstrument vs. patient, instrument vs. foil 1, instrument vs. foil 2, patient vs. foil 1, and patient vs. foil 2. After calculations, we conducted a three way ANOVA (age group x condition x pair comparison) with Bonferroni post hoc tests. Total fixation ti me proportions in Chapter 2 and 3 were used for between group comparisons. A mixed ANOVA test was also conducted for three way comparisons (age group x condition x AOI) with Bonferroni post hoc tests. Averages of total fixation count and e ntry time in eac h AOI were calculated by items (item analysis). A three way mixed ANOVA (age group x condition x picture item) was conducted with Bonferroni post hoc tests. Results Mean Difference Comparisons The overall mean difference in younger adults was significant ly greater than older adults [F(1, 1560)=150.394, p =.000]. The mean difference in each condition was significantly greater in younger participants than HOAs; and the mean difference of each AOI comparison was significantly greater in younger than HOAs. Tab le 4 1 shows the means (sd) and p values. Specifically, in all conditions except the CP & II, the mean difference was significant in instrument vs. patient, instrument vs. foil 1, and instrument vs. foil 2, but not in patient vs. foil 1 and patient vs. fo il 2. In the CP & II condition, the mean difference was significant in instrument vs. foil 1, instrument vs. foil 2, and patient vs. foil 1, but not in instrument vs. patient and patient vs. foil 2. See Table 4 1.
70 Table 4 1. Total fixation time proporti on differences between AOIs in conditions between groups Younger HOA t ( df ), p value Condition 1: Neutral I vs. P 7.18 (11.02) 2.26 (6.94) t (39)=2.95, p = .003 I vs. Foil 1 11.50 (7.52) 4.32 (5.80) t (39)=4.30, p = .000 I vs. Foil 2 12.15 (8.59) 5 .21 (5.70) t (39)=4.15, p = .000 P vs. F oil 1 4.32 (7.86) 2.06 (5.41) t (39)=1.35, p = .177 P vs. F oil 2 4.97 (8.39) 2.95 (6.37) t (39)=1.21, p = .228 Mean 8.02 (9.27) 3.36 (6.13) t (39)=3.48, p = .000 Condition 2: Congruent Context I vs. P 7. 04 (10.91) 2.42 (6.30) t (39)=2.77, p = .006 I vs. F oil 1 12.02 (8.89) 4.21 (5.12) t (39)=4.67, p = .000 I vs. F oil 2 12.54 (10.20) 4.88 (6.18) t (39)=4.59, p = .000 P vs. F oil 1 4.98 (7.55) 1.79 (4.57) t (39)=1.91, p = .057 P vs. F oil 2 5.50 (6.49 ) 2.46 (4.45) t (39)=1.82, p = .069 Mean 8.42 (9.44) 3.15 (5.44) t (39)=7.05, p = .000 Condition 3: Congruent Patient & Incongruent Instrument I vs. P 4.26 (8.84) 2.82 (5.49) t (39)=0.86, p = .389 I vs. F oil 1 8.60 (8.60) 3.77 (6.93) t (39)=2.89, p = .004 I vs. F oil 2 8.74 (9.31) 4.28 (6.66) t (39)=2.67, p = .008 P vs. F oil 1 4.35 (6.86) .96 (5.17) t (39)=2.03, p = .042 P vs. F oil 2 4.49 (6.90) 1.47 (5.64) t (39)=1.81, p = .071 Mean 6.09 (8.36) 2.66 (6.09) t (39)=4.59, p = .000 Cond ition 4: Incongruent Patient & Congruent Instrument I vs. P 7.65 (11.68) 2.54 (6.43) t (39)=3.06, p = .002 I vs. F oil 1 11.28 (8.60) 4.12 (4.87) t (39)=4.29, p = .000 I vs. F oil 2 12.02 (9.80) 4.21 (5.18) t (39)=4.67, p = .000 P vs. F oil 1 3.63 (8. 70) 1.58 (5.94) t (39)=1.23, p = .219 P vs. F oil 2 4.37 (8.30) 1.67 (5.05) t (39)=1.62, p = .106 Mean 7.79 (10.01) 2.82 (5.59) t (39)=6.65, p = .000 Across Conditions I vs. P 6.53 (10.65) 2.51 (6.26) t (39)=4.82, p = .000 I vs. F oil 1 10.85 (8.44) 4.11 (5.66) t (39)=8.08, p = .000 I vs. F oil 2 11.36 (9.53) 4.65 (5.91) t (39)=8.04, p = .000 P vs. F oil 1 4.32 (7.71) 1.60 (5.26) t (39)=3.26, p = .001 P vs. F oil 2 4.83 (7.51) 2.14 (5.41) t (39)=3.23, p = .001 Mean 7.58 (9.31) 3.00 (5.82) t (39)=12.28 , p = .000 Note: I=Instrument, P=Patient, F oil 1=semantically related to context, F oil 2 = unrelated Total Fixation Time Comparisons There was a main effect of group [F(1, 1248)=403.997, p =.000]; overall total fixation time in younger adults 14.36% (6.85) was significantly longer than in HOAs 9.41% (3.92). In all conditions, fixation time was significantly longer in younger than in older adults. Also, in all
71 AOIs, fixation time was significantly longer in younger compared to older adults. Table 4 2 shows total fixation times in each condition and each picture item. In three way comparisons, total fixation times in younger participants compared to HOAs were significantly longer in all pictures in all conditions ( p <.05 for all). Table 4 2. Total fixation time pro portion comparisons in conditions between groups Younger HOA t ( df ), p value Condition 1: Neutral Patient 15.04 (6.16) 10.12 (4.32) t (39)=5.28, p = .000 Instrument 22.22 (6.23) 12.38 (4.28) t (39)=9.92, p = .000 Foil 1 10.72 (3.26) 8.06 (3.01) t (39) =3.24, p = .007 Foil 2 10.07 (4.22) 7.17 (3.25) t (39)=3.51, p = .003 Mean 14.51 (7.03) 9.44 (4.24) t (39)=10.98, p = .000 Condition 2: Congruent Context Patient 15.11 (3.69) 9.76 (3.17) t (39)=5.62, p = .000 Instrument 22.15 (4.23) 12.18 (4.00) t (39)=9.37, p = .000 Foil 1 10.14 (3.69) 7.97 (2.66) t (39)=3.05, p = .028 Foil 2 9.61 (4.03) 7.30 (3.11) t (39)=2.99, p = .019 Mean 14.25 (7.23) 9.31 (3.76) t (39)=10.52, p = .000 Condition 3: Congruent Patient & Incongruent Instrument Patient 15.39 (4.61) 9.38 (3.35) t (39)=6.61, p = .000 Instrument 19.65 (6.04) 12.20 (4.60) t (39)=7.65, p = .000 Foil 1 11.04 (3.67) 8.43 (3.40) t (39)=3.43, p = .008 Foil 2 10.91 (4.15) 7.92 (3.12) t (39)=3.49, p = .002 Mean 14.25 (5.89) 9.48 (3.99 ) t (39)=10.59, p = .000 Condition 4: Incongruent Patient & Congruent Instrument Patient 14.53 (6.48) 9.62 (3.88) t (39)=5.77, p = .000 Instrument 22.17 (7.00) 12.16 (3.70) t (39)=9.92, p = .000 Foil 1 10.89 (3.57) 8.04 (2.92) t (39)=3.50, p = .004 F oil 2 10.16 (3.96) 7.95 (2.73) t (39)=2.86, p = .025 Mean 14.44 (7.22) 9.44 (3.73) t (39)=11.03, p = .000 Across Conditions Patient 15.02 (5.62) 9.72 (3.68) t (39)=11.64, p = .000 Instrument 21.55 (6.66) 12.23 (4.12) t (39)=18.44, p = .000 Foil 1 10.70 (3.54) 8.13 (2.99) t (39)=6.61, p = .000 Foil 2 10.19 (4.08) 7.59 (3.05) t (39)=6.42, p = .000 Mean 14.36 (6.85) 9.42 (3.92) t (39)=21.56, p = .000 Note: Foil 1=semantically related to context, Foil 2=unrelated Fixation Count Average Comparisons T he overall fixation count average was significantly greater in younger adults than in older adults [ t ( 39) = 19.09 , p =.000]. The mean (sd) in younger participants was 2.13 (.88) and in
72 HOAs was 1.52 (.55). Figure 4 1 illustrates total fixation counts in each AOI in each age group. Fixation counts on all AOIs were significantly greater in younger compared to older adults ( p =.000 for all). Further in three way comparisons between groups in each AOI in each condition, fixation counts in younger participants were significantly greater than HOAs in all comparisons ( p <.05 for all) except for foil 1 in the CC condition ( p =.144) and in the CP & II condition ( p =.057), and foil 2 in the CC condition ( p =.059) and in the IP & CI condition ( p =.087). Figure 4 1. Fixation count averages in each AOI between groups. *** p <.001 Entry Time Average Comparisons The overall entry time between groups was significantly faster in younger adults than in older adults [ t ( 39 )= 3.15 , p =.002]. The overall mean (sd) entry time in younger pa rticipants was 1097.63ms (373.96) and in HOAs was 1158.30ms (308.07). Between AOIs, entry time of t ( 39 )= 2.00 , p =.046]. The significance of difference was marginal in patients bet ween the groups [ t ( 39 )= 1.93 , p =.054].
73 Further three way comparisons between groups in each picture item in each condition were not significant in any comparisons ( p >.05 for all). Figure 4 2 shows entry times of each AOI in both age groups. Figure 4 2. E ntry time in each AOI between groups. * p <.05 Interim Discussion The mean difference comparisons showed significant group differences. First, mean differences of instruments vs. patients in all four conditions were significantly longer in younger adults co mpared to older adults. Therefore, although both younger and older participants fixated on instruments more than patients when related verbs are presented, this preference toward instruments was significantly greater in younger compared to older adults. M ean difference between instruments and foils was reduced with aging, but the mean difference between patients and foils was not. That is, the mean difference between instruments and foil1 or 2 was significantly greater in younger than older adults in all c onditions. On the other hand, the mean difference between patients vs. foil1 or 2 in younger adults was not significantly different from older adults across all conditions except in the CP & II condition.
74 This suggests that the reduced difference size in i nstruments vs. patients in older adults compared to younger adults was due to a decline of verb instrument association in older adults. Decreased strength in the verb instrument connection in aging may indicate some degree of semantic changes with aging. It might be due to semantic representation decline (storage deficits) or semantic access difficulty (access deficits). And whatever the cause of the semantic changes, it might be related to the unexpected incongruent context effects observed in older adult s. We will discuss more about this possibility in Chapter 5. Second, we confirmed that reduced total fixation time proportions on AOIs in older adults were significant compared to younger adults. The additional analysis of fixation counts and entry time a lso showed that older adults fixated in AOIs fewer times and more slowly than younger adults. T hese results indicat e that older adults looked at AOIs for a shorter time than younger adults, because they fixated on the AOIs fewer times. To determine what ol der adults were looking at, we calculated fixation time proportion and fixation counts in non AOI areas. Fixation time in non AOIs in older adult was 17.12% (3.26), which was significantly longer than in younger adults (mean=14.99% , sd=2.22). Fixation cou nt in non AOI in older adults was 2.85 (.58), which was also significantly greater than in younger adults (mean=2.31, sd=.31). Spending more time looking at non AOI areas might be due, in part, to slower eye movements toward AOIs, which is supported by the ir delayed entry time data. Munoz et al. (1998) reported that the time to onset of eye movement s in healthy older adults (60 79 years of age) was slower than other age groups (young children and young adults), which may reflect nervous system degeneration , which might explain our shorter fixations and slower entry time. A nother purpose of our additional eye movement analysis was to rule out atypical eye movement patterns such as constant moving around AOIs and fixating very shortly,
75 which might indicate cog nitive impairment (MÃ¼ller, Richter, Weisbrod, & Klingberg, 1991) . However, w e did not find any evidence of th e s e atypical patterns in our participants.
76 CHAPTER 5 GENERAL DISCUSSION The current study investigated whether there is a difference in how pati ent and instrument thematic roles are semantically associated with verbs. In order to investigate this research question, we used an eye tracking methodology and compared which thematic role pictures participants looked at longer. Further, we also included location pictures as context information to activate a specific schema and constrain non target thematic role fillers. We manipulated congruency of context to measure the context effects as well. We conducted the same experiment with two different age gro ups (younger adults and healthy older adults) to explore how eye movements in the two groups may be different and what they might reveal about the semantic relationships between verbs and related instruments and patients with aging . Our results suggeste d significantly stronger verb instrument associations compared to verb patient associations in younger adults. This pattern was similarly shown in older adults, but the tendency was significantly reduced in older adults. Context did not successfully activa te the target schema and contribute to increasing the likelihood of looking at target fillers, because total fixation times in patients and instruments were statistically the same regardless of the type of context in both age groups. However, we observed t hat context somewhat facilitated congruent patients and instruments in younger adults, but suppressed instruments in older adults. On the other hand, patients were not in older adults affected because of their weak association with verbs. In the current study, we have three main topics (1) verb and thematic role comparison, (2) context effects, and (3) semantic processing changes in aging. In this chapter, we will discuss each topic and attempt to describe possible mechanisms of semantic processing in rel ation to our questions.
77 Verb and Thematic Role Comparison Our findings suggest that there is a stronger association between verbs and related instruments as compared to verbs and related patients, and this pattern was similar in both age groups. In younge r adults, instrument fixations were significantly longer than patient fixations not only when there was no context or a congruent instrument context but also when there was an incongruent instrument context presented prior to verb thematic role presentatio n. In older adults, the pattern of longer fixations on instruments than patients was observed, but the difference between instruments and patients was not significant. Even though there was not a significant difference between patients and instruments in t he older group, instrument fixations were significantly longer than the two types of foils regardless of context manipulation, whereas patient fixations were not significantly longer than foil 1, which was semantically related to the context (though patien ts were significantly longer than unrelated foil 2). These findings suggest that the patient association with verbs is weakened as much as foil association with verbs, whereas the verb instrument association is stronger than the verb foil association. The finding of a stronger verb instrument association compared to verb patient association may be explained by a difference in the nature of relationships between verbs and the different thematic roles. We discussed in Chapter 1 about how instruments were att ached to verbs by their meanings such that concepts of instruments fill the meaning of verbs in sentences, and semantic features of instruments were overlapped with concepts of verbs. In addition, a smaller number of possible instruments for individual ver bs increase the likelihood of co activation so that repeated activation between the specific instrument and a verb makes their connections strong. That is, although both patients and instruments are related with verbs in frequency usages and schema based a ctivations, instruments are additionally conceptually
78 attached to the meaning of related verbs. These characteristics privilege instruments to be highly activated by verb. In the current study, we selected both patients and instrument to be highly associ ated with verbs based on commonness questionnaire norms. However, in addition to high degree of association, instruments were conceptually tightly interconnected with verbs. Further, the current study only included transitive actions verbs. O ur instruments were not limited to tool objects but also included non manipulable objects (e.g., for covering ) . We did not measure semantic features of our instruments and verbs; however , intuitively most instruments were specified to perform the purpose o f verb actions (e.g., is to cover a bat is to hit a baseball). Therefore, semantic feature activation might also be one of the factors increasing instrument fixations in the current study. Context Effects Context effects in t he current study were investigated in two ways: (1) the congruent context condition was compared to the no context condition, and (2) the congruent context condition was compared to the incongruent context conditions. However, by comparing between conditio ns, we did not find a significant main effect of context on eye movements toward thematic roles , although we did f in d subtle evidence of context influence mostly in the time course analyses . Methodological Differences W e may have only found a small degree of context effects compared to previous studies (Binder, 2003; Binder & Morris, 1995; Binder & Rayner, 1998; Federmeier & Kutas, 2005; Federmeier et al., 2002; Rayner et al., 1999) due to differences in methodologies : (1) picture presentation of context v s. language context , (2) single word level concept comprehension vs. sentences or passages, and (3) implicit vs. explicit task.
79 One of the differences of our study compared to other studies is that we used location scenes to prime to activate target schem as whereas the other studies used contents of sentences or passages . Therefore, it might be argued that participants did not process the context pictures sufficiently for priming. However, we presented the location scene for 250ms with a 100ms mask. Accord ing to Biederman et al. (1982) , people can recognize a visual scene that is presented as quickly as 150ms, and our scene pictures were not as complex as Biederman and colleagues pictures . Further, we tested a few students in our lab to determine if the lo cation scenes were recognizable during presentation , and the students named all locations correctly. In the current study, we included two primes: location scenes and verbs. Therefore, to activate the target schema and target thematic role fillers , the tw o types of primes need ed to build on each other. Additive priming effects with multiple primes has been suggested by Khader, Scherag, Streb, and RÃ¶sler (2003) who reported faster response times when two nouns prime related verbs compare to when a single no un primes verbs (see also Balota & Paul, 1996 for similar findings) . In contrast, Herlofsky and Edmonds (2012) did not find evidence for additive priming when they presented two types of thematic roles (agent and patient) as primes for verb targets. On lex ical decision for the target verb, participants were not faster with two related noun primes as compared to one related noun prime directly preceding the verb. The difference between the two studies was that Herlofsky & Edmonds used a lexical decision task in order to reduce attentional processes rather than a relatedness decision task, which is considered more attentional task. Balota and Paul (1996) suggested that attentional processing benefits additive priming effects; therefore, this might explain the lack of additive effect in Herlofsky & Edmonds. The current study wa s aimed to measure automatic semantic activations between verbs and thematic roles. Therefore, we did not use an explicit task but asked participants to answer
80 whether the currently presen ted verb was the same as the previously presented verbs. I t is possible that participants did not necessarily integrate the two primes to activate corresponding schemas to perform the implicit task . Similarly, previous studies investigating context effect s used tasks such as short passages (Binder, 2003; Binder & Morris, 1995) or sentences (Binder & Rayner, 1998) with a comprehension task which also recruited attentional processes. In the current study, we used single word level concepts with no explicit t ask , which does not ask participants to integrate semantic information of the sequentially presented stimuli (i.e., location, verb, and thematic noun pictures). Without integration enforcement, the semantic information related to the location may decay rap idly and therefore does not significantly influence thematic role activations. Sanford and Garrod (1981) claimed that once a schema is activated, default items fill each thematic role, and these initially activated defaults always fill the thematic roles u nless other strongly biased contextual information is driven. Our experiment was aimed at switch ing the default items to our targeted patients and instrument s by presenting a location context . However, our findings suggest that locations were very weakly e ffective, perhaps because there was not an explicit task to activate targets at a single word level. This may indicate that verbs and their default items are very tightly associated and not switched to other items unless a strong attentional task is provid ed such as passage comprehension task or semantic relatedness judgment task. Evidence of Context Effects Although we did not find significant context effects, we reported a number of fixation patterns that might be conceivable evidence of context effe cts. Also, we found evidence that context effects impact the age group s differently .
81 I n younger adults, a congruent context enhances verb association with the congruent thematic role. When congruent context both to patients and instruments was presented compared to no context was given, the initial time at which instrument fixations were significantly longer than patient s was faster. This may suggest that the congruent context evoked the target schema , and instruments within the schema were quickly determ ined. On the other hand, when context was in congruent to one of the thematic roles (i.e., incongruent context conditions ), the amount of time of thematic role fixations was different . When context was only congruent to patients but incongruent to instrum ents, the total fixation time of instruments was reduced compared to other conditions, where context was congruent to instruments. Also, there were fewer time windows at which instrument fixations were significantly longer than patients. B ecause context wa s congruent to patients, patient fixations seemed somewhat increased , but this increase was very subtle. When context was congruent to instruments but not to patients, the total fixation time of patients was reduced to be statistically same as foils that w ere semantically related to context. These pieces of evidence may support our hypotheses that context information subtly activates the target schema and its corresponding patients and instruments. O lder adults, however, showed that context influences eye movements , but in a pattern that contrast ed with the pattern in younger adults. When the context was congruent to both patients and instruments as compared to no context, the number of time windows at which instrument fixations were significantly longer th an patients increased; however, the initial time at which instrument fixations were significantly longer than foils was slowed. On the other hand, when context was congruent to patients but incongruent to instruments, patient fixations were reduced and ins trument fixations increased in the beginning of time course. C ontext congruent to
82 instruments but not patients decreased instrument fixations but did not influence patient fixations . The evidence that context decreases the congruent thematic role and incre ases the incongruent thematic role is the opposite pattern of the more expected results seen in the younger adults. Semantic Processing Changes in Aging In addition to the pattern differences related to context manipulation between the two age groups desc ribed above, we also showed that the fixation differences between patients and instruments as well as the difference s between patients and foils were not significant in older adults, which were significantly different in younger adults. Further, we calcula ted a mean difference analysis to determine which picture differences were significantly reduced in aging. The significant reduction in older adults compared to younger adults was shown between instruments vs. patients and between instruments vs. foils, bu t not between patients vs. foils. This indicates that the reduced difference between instrument s and patients in aging was due to a greater reduction in instrument fixations than patient s. However, it may be that the strength of the verb patient associatio n was already weak in younger adults; thus , even a small degree of decrease in the verb patient association caused patients to be statistically the same as foils (i.e., floor effects) in older adults. Also, more importantly compared to foils in HOAs, instr uments were looked at significantly more than two types of foils, but patients were looked at statistically the same as the foil that are semantically related to context. This indicates that patients were only activated by context but not by verbs. B oth r educed semantic strength of the verb instrument relationship, and the opposite fixation patterns responding to context effects in older adults compared to younger adults may indicate some degree of semantic processing changes in aging. In Chapter 1 and 3, we introduced three semantic models in aging and hypotheses based on each model: (1) greater language
83 experience , (2) resource limitation including working memory and speed, and (3) transmission deficit. With our current results , we rule out the greater l anguage experience model because there was no indicator of patients or instruments being hyper related with verbs and/or location contexts. We rather found overall reduced verb thematic role association in older adults. The resource limitation theory may p artially explain the results . Slow speed might also explain overall reduced fixations on thematic roles because it might be due to slow language/cognitive processing speed in HOAs. This theory is also related with their slowed entry fixation time in AOIs, as well as the slower initial time at which instrument fixations were significantly longer than patient s in the congruent context condition compared to the neutral condition , because this slowed processing in fact increased a number of time windows where i nstrument fixations were significantly longer than patients. However, we did not find any other delayed fixation patterns in other conditions or overall slowness. This model also does not explain the fixation patterns in the incongruent context conditions in older adults. Both the w orking memory limitation model and transmission deficit theory mode l explain no significant context effects and reduced fixations on thematic roles compared to younger adult s because of difficulty with integration of the informat ion to activate the target schema for 3 seconds in the working memory limitation model but due to reduced connections between concepts in the transmission deficit theory. However, the working memory deficit model does not explain the disadvantage in fixati ons by congruent vs. incongruent context. Moreover, working memory ability in all HOA participants were within the normal range based on the digit spans and symbol spans and the semantic short term memory tests (see Table 3 2). On the other hand, the tran smission deficit theory could explain our results , b ecause this theory argues that age weakens strength of linkage between concepts (MacKay & Burke, 1990) .
84 This explains reduced fixations on thematic roles because it suggests loosened connections between v erbs and thematic roles. In addition, a lthough this theory does not specifically compare verb patient association reduction to verb instrument association reduction in aging, it helps us to explain the reduced fixations of patients with simultaneous verb p resentation. In Chapter 1, we discussed that individual instruments are conceptually intertwined with a related verb meaning in addition to instruments hav ing schema based activations with their related verbs, while patients only have schemas based acti vations to related verbs. Therefore, we speculate that their different semantic characteristics may have impacted the degree of semantic association reduction between verbs and patients . It has been known that semantic concept knowledge itself doe s not dec line in aging but rather increase s with richer experience s compared to younger adults (Verhaeghen, 2003; Ackerman & Rolftus, 1999). Thus, semantic concept based association s (i.e., semantic features ) between verb s and related instrument s is relatively pres erved compared to the purely activation linkage based association (i.e., frequency of usage) in verb s and related patients. As previously discussed , the transmission deficit theory is also supported by n eurobiological changes in healthy ag ed brain s , and s ynaptic depression and refractory effects caused by neuronal changes may explain the negative incongruent context effects. A healthy brain is vulnerable with aging for volume loss (atrophy) (Jiang et al., 2014) and decreased neuronal connections and neurot ransmission (Samorajski, 1977) . Reduced neurotransmission not only affect s information transferability and activation s between concepts but also reduces synaptic depression. Synaptic depression is a reduced responsiveness of neurons shortly after they fir ed, and neurotransmitters such as acetylcholine improve ongoing neural activities and reduce synaptic depression. Therefore, a decrease in the neurotransmitters would increase synaptic depression across semantically related items, which is sensitive to pre sentation rate and
85 repetition . This synaptic depression may cause r efractory effects that are inconsistent with lexical retrieval errors when stimuli are repeatedly presented as targets and distracters over a number of cycles (Jefferies et al., 2007) . Howe ver, Gardner et al. (2012) reported that refractory effects were also present in various non verbal tasks in stoke patients with aphasia . Although this refractory effect caused by synaptic depression has been investigated with clinical population s , the pr efrontal lobe is one of the areas that Lambon Ralph and colleagues reported to be responsible for semantic control (Gardner et al., 2012; Jefferies & Lambon Ralph, 2006) , which is also one of the most vulnerable area s for the changes (Jiang et al., 2014) . Therefore, although older adults behavioral outcomes in cognitive tests were mostly within a normal range, it may be possible that semantic processing might have been impacted by normally occurring neuronal changes. In the research questions about contex t effects we reported (1) slowed fixation in instruments and (2) negative incongruent context effects on instruments. First, the collapsed instrument fixations in the beginning of time course in the congruent context condition may be explained by synaptic depression. When no context is presented, verbs are only activated with thematic roles (patients and instruments). Therefore, once the connections fire, synaptic depression may occur but quickly resolves between trials. On the other hand, when a congruent context is presented prior to verb and thematic roles , the location concept activates along with the related verbs along with the corresponding thematic role item s in the schema. Due to th e activation of many semantically related concepts , there is a slowe d fixation response to the target thematic role pictures, as the thematic roles are temporary suppressed due to synaptic depression. The instrument fixation collapse in the beginning of the time course in the congruent context condition may reflect th at sy naptic depression. Once synaptic depression is resolv ed ( after
86 around 1300ms) , instruments re obtain connections with verbs and location context. Therefore, instrument fixations become significantly longer than patients, and more time windows showed signif icantly greater fixations than patients (500ms duration) compared to no context was given (200ms duration) . We are aware that the initial time at which instrument fixations were significantly longer than patients was similar in the congruent condition (1 9 0 0ms) to the neutral condition ( 2000 ms). However, the insignificant differences between patients and instruments in the beginning of the time course in the neutral condition were insignificant mainly due to a large number of Bonferroni corrections (4 AOIs b y 61 time points). Considering the error bars in the Figure 3 1 (a), the patients and instruments became separated at around 900ms. On the other hand, the patient and instrument fixations in the beginning time course in the congruent context condition were almost identical until around 1300ms. In addition, in the neutral condition, instrument fixations became significantly greater than foils in earlier time (at around 1000ms), whereas in the congruent context condition, instrument fixations were not even st atistically different from foils until they start separating from patient fixations (at around 1300ms). Thus, this demonstrates collapsed instrument fixations in the beginning of time course and their slowed time of instrument fixation increase, supporting the synaptic depression effects. R epetitive presentation and fast rate are the key factors causing refractory effects (Jefferies et al., 2007) . In the current study, t he stimulus pictures were presented in four conditions , and target and distractors wer e interchanged between conditions (e.g., onion is a target with a kitchen context in the CP & II condition, but is a distractor with a barber shop context in the IP & CI condition) . Also, we have very fast presentation time for each stimulus (250ms for loc ation context, 800ms for verb, 3000ms for patients and instruments).
87 Refractory effects by synaptic depression could also explain the fixation patterns in the incongruent context conditions. When context is congruent to patients but not instruments, becaus e i nstruments are not related with context but related with verbs, instruments are not suppressed by context and remain connected with verbs. After the synaptic depression is resolved (at around 1800ms) and context properly corresponds with thematic roles, instrument fixations decrease because they are not the target instruments in the location schema. On the other hand, when context is congruent to instruments but not patients, instrument fixations decrease in the beginning time course with the same mechan isms we described above, and therefore the difference between instruments and patients in those time windows are not significant (until at 1750ms). In later time windows after the depression resolved, instruments were fixated longer than patients. However , interestingly , this negative incongruent context effects do es not seem to impact patient fixations. Although the patient fixations increase slightly later in the time course analysis when context was congruent to patients, their fixations were still shor ter than instrument fixations and not significantly longer than the two types of foils. When context was incongruent to patients, we also did not find a particular pattern of increased or decreased patient fixations. This dissociation could be due to gen erally weakened a ssociation between verbs and patients. We know from our findings with younger adults that there is facilitation to patients and instruments, with more fixations for both thematic roles as compared to foils. Further, both the patients and i nstruments should have be en looked at longer than foils regardless of context manipulation. However, t otal fixation time of patients was not significantly longer than foil 1 (semantically related to context) in all conditions , suggesting that verb primes d id not activate
88 patient . This observation is also in contrast to our hypothesis of greater context effects on patients than instruments due to a greater number of possible patients than instruments. Conclusion In the General Discussion , we suggested t hat there is a stronger verb instrument compared to verb patients association in both age groups. With r espect to the lack of hypothesized main effects of context, we offered that this finding might be explained by task effects, which did not require our p articipants to consciously integrate semantic information in our experiment . The reduced fixations on patients and instruments, and the subtle evidence of context effects toward instruments in the time course analyses in older adults w as explained by seman tic control difficulty that might be a part of normal aging, though more research is needed to examine this proposition. Significance of Study Previous literature has studied verb and thematic role association s ; however, the current study is novel to use verb and thematic role concepts at a single word level and to directly compare the two types of thematic roles (patient and instrument) . This study provided important insight to understand pure semantic association between verbs and thematic roles by mini mizing confounds such as syntax , an d most previous studies investigated verb attachment to instruments using sentences. However, b ecause sentence processing requires not only semantic but also syntactic components, it can be difficult to understand pure se mantic relationships between concepts. The findings in the current study using single concepts , therefore , are important to understanding pure semantic processing stripped of syntactic components. In addition, as we used picture stimuli (except verbs) rath er than words, we tried to minimize lexical activation and associations as well.
89 T o our knowledge , this study is the first to investigate spontaneous context effects in semantic processing using location visual scenes in a non explicit task. Therefore, t his study was very conservative in measuring automatic brain activations without forcing participants to consciously activate concepts for tasks such as sentence comprehension. Also, this study showed how younger and older adults use context information di fferently during automatic semantic processing, which allowed us to understand aspects of semantic changes in older adults. Future Directions The current study suggested stronger verb activations to instruments compared to patients; however, the current fi ndings were only suggested with transitive action verbs. Different types of verbs depending on syntactic or semantic requirements (e.g., intransitive verbs or abstract verbs) would have different relationships with patient and instrument roles. Also, the c urrent study only compared patient and instrument activations with verbs. Agent vs. patient comparison or instrument vs. location comparison would also be interesting with regard to controlling syntactic argumenthood of the two thematic roles and comparing semantic contributions. Therefore, future studies are needed to investigate more specific mechanisms of verb thematic role activations. Although it was our purpose to use an implicit task in the experiment to measure spontaneous verb and thematic role ac tivation as well as context effects, it would also be valuable to conduct the same experiment with an explicit task to understand better about task effects on verb thematic role activations and context effects. A lso, we found interesting sematic processin g patterns in healthy older adults compared to younger adults. We explained the patterns with semantic processing decline in both semantic storage and access decline. To expand our understanding and support our semantic processing
90 hypothesis , a future stud y may include healthy older older adults (aged 80 to 100) to investigate their semantic processing. Furthermore, a s researchers in the field of speech language pathology, our ultimate goal of research is to understand language processing in healthy adul ts so that we can understand language processing in individuals with language disorders (aphasia caused by stroke or dementia). Therefore, our next step is to expand this study to clinical populations to investigate how their semantic processing of verb s a nd thematic role s are impaired or preserved. In general discussion, we briefly discussed semantic aphasia with semantic cognitive deficits. Because these patients have difficulties controlling semantic activations in selecting target concepts, their eye mo vements in the current experiment c ould also provide valuable information. Given the knowledge , it is our ultimate goal to develop assessments and/or treatments for their language rehabilitation.
91 APPENDIX STIMULUS LIST
92 Table A 1. S timulus List Verb Loca tion Congruent Patient Congruent Instrument Incongruent Patient Incongruent Instrument Foil1 Foil2 Catching Baseball stadium Baseball Glove Wave Surfing board Cap Coral Catching Ocean Wave Surfing board Baseball Glove Coral Cap Covering Op erating room Mouth Manhole Manhole cover Scalpel Traffic light Covering Street Manhole Manhole cover Mouth Traffic light Scalpel Cutting Kitchen Onion Knife Beard Scissors Ladle Towel Cutting Barber shop Beard Scissors Onion Knife Towel Ladle Drying Laundry room Shirt Drier Hair Hair drier Iron Soap Drying Bathroom Hair Hair drier Shirt Drier Soap Iron Examining Eye clinic Eye Eye exam chart Diamond Magnifying glass Eye drop Ring box Examining Jewe lry shop Diamond Magnifying glass Eye Eye exam chart Ring box Eye drop Hanging Closet Dress Hanger Curtain Curtain rod Socks Balcony Hanging Window Curtain Curtain rod Dress Hanger Balcony Socks Hitting Boxing ring Punching bag Boxing gloves Nail Hamm er Bell Bicycle Hitting Garage Nail Hammer Punching bag Boxing gloves Bicycle Bell Inflating Bike shop Tire Air pump Balloon Helium tank Lock Table Inflating Park Balloon Helium tank Tire Air pump Table Lock Lifting Gym Barbell Hand Bricks Crane Wa ter fountain Wagon Lifting Construction area Bricks Carne Barbell Hand Wagon Water fountain Lighting Church Candle Match Grill Lighter Bench Fire place Lighting Camping area Grill Lighter Candle Match Fire place Bench A 1. C ontinued.
93 Verb Location C ongruent Patient Congruent Instrument Incongruent Patient Incongruent Instrument Foil1 Foil2 Observing Observatory Planet Telescope Cells Microscope Satellite Beaker Observing Science lab Cells Microscope Planet Telescope Beaker Satellite Measuring Car Lumber Tape measure Fabric roll Yard stick Chisel Pins cushion Measuring Sewing room Fabric roll Yard stick Lumber Tape measure Pin cushion Chisel Opening Bar Beer bottle Bottle opener Envelope Letter opener Pitcher Messenger bag Opening Office Envelope Letter opener Beer bottle Bottle opener Messenger bag Pitcher Painting Art studio Picture Brush Finger nail Nail polish Chimney Clipper Painting Beauty shop Finger nail Nail polish Picture Brush Clipper Chimney Scooping Ice cream cart Ice cream Scoop Rocks Bulldozer Syrup Hard hat Scooping Mining area Rocks Bulldozer Ice cream Scoop Hard hat Syrup Serving Restaurant Steak Tray Shuttlecock Racquet Chef hat Wristband Serving Badminton court Shuttlecock Racquet Steak Tray Wristband Chef hat Shooting Hockey rink Puck Hockey stick Arrow Bow Skate Olympic symbol Shooting Archer field Arrow Bow Puck Hockey stick Olympic symbol Skate Tightening Mechanic shop Nut Wrench Shoes Strings Ladder Whistle Tightening Running track Shoes Str ings Nut Wrench Whistle Ladder Watering Yard Flower pot Watering can Corn Sprinkler Rake Scarecrow Watering Farm Corn Sprinkler Flower pot Watering can Scarecrow Rake Wiping Classroom Blackboard Eraser Windshield Wiper blades Podium Steering wheel Wip ing Car Windshield Wiper blades Blackboard Eraser Steering wheel Podium
94 LIST OF REFERENCES Ackerman, P. L., & Rolfhus, E. L. (1999). The locus of adult intelligence: Knowledge, abilities, and nonability traits. Psychology and Aging, 14 , 314 330. Agguj aro, S., Crepaldi, D., Pistarini, C., Taricco, M., & Luzzatti, C. (2006). Neuro anatomical correlates of impaired retrieval of verbs and nouns: Interaction of grammatical class, imageability and actionality. Journal of Neurolinguistics, 19 (3), 175 194. doi : 10.1016/j.jneuroling.2005.07.004 Altmann, G., & Kamide, Y. (1999). Incremental interpretation at verbs: Restricting the domain of subsequent reference. Cognition, 73 (3), 247 264. Arbuckle, T. Y., Nohara LeClair, M., & Pushkar, D. (2000). Effect of off t arget verbosity on communication efficiency in a referential. Psychology and Aging , 15(1), 65 77. Bak, T. H., & Hodges, J. R. (2003). Kissing and dancing a test to distinguish the lexical and conceptual contributions to noun/verb and action/object dissoci ation. Preliminary results in patients with frontotemporal dementia. Journal of Neurolinguistics, 16 (2), 169 181. Balota, D. A., & Paul, S. T. (1996). Summation of activation: evidence from multiple primes that converge and diverge within semantic memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22 (4), 827 845. Balota, D. A., Watson, J. M., Duchek, J. M., & Ferraro, F. R. (1999). Cross modal semantic and homograph priming in healthy young, healthy old, and in Alzheimer's diseas e individuals. Journal of the International Neuropsychological Society, 5 (7), 626 640. Barsalou, L. W. (1988). The content and organization of autobiographical memories. In U. Neisser & E. Winograd (Eds.), Remembering reconsidered: Ecological and traditio nal approaches to the study of memory (pp. 193 243). New York: Cambridge Univ. Press. Barsalou, L. W. (2008). Grounded cognition. Annu. Rev. Psychol., 59 , 617 645. doi: 10.1146/annurev.psych.59.103006.093639 Bhatnagar, S. C. (2008). Neuroscience for the st udy of communicative disorders (3rd ed.). Baltimore, MD: Lippincott Williams & Wilkins. Biederman, I., Mezzanotte, R. J., & Rabinowitz, J. C. (1982). Scene perception: Detecting and judging objects undergoing relational violations. Cognitive psychology, 14 (2), 143 177. Biederman, I., Teitelbaum, R. C., & Mezzanotte, R. J. (1983). Scene perception: A failure to find a benefit from prior expectancy or familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9 (3), 411 429. Binder, K. S. (2003). Sentential and discourse topic effects on lexical ambiguity processing: An eye movement examination. Memory & cognition, 31 (5), 690 702.
95 Binder, K. S., & Morris, R. K. (1995). Eye movements and lexical ambiguity resolution: effects of prior en counter and discourse topic. Journal of Experimental Psychology. Learning, Memory, and Cognition, 21 (5), 1186 1196. Binder, K. S., & Rayner, K. (1998). Contextual strength does not modulate the subordinate bias effect: Evidence from eye fixations and self paced reading. Psychonomic Bulletin & Review, 5 (2), 271 276. Bowles, N. L., & Poon, L. W. (1988). Age and context effects in lexical decision: An age by context interaction. Experimental Aging Research, 14 (4), 201 205. Brunson, B. A. (1992). Thematic di scontinuity. (Ph.D. Dissertation), University of Toronto. Burke, D. M., & Shafto, M. A. (2008). Language and aging. In F. I. M. Craik & T. A. Salthouse (Eds.), The handbook of aging and cognition (3rd ed., pp. 373 443). New York, NY: Psychology Press. B urke, D. M., White, H., & Diaz, D. L. (1987). Semantic priming in young and older adults: Evidence for age constancy in automatic and attentional processes. Journal of experimental psychology: Human Perception and Performance, 13 (1), 79 88. Carota, F., Mo seley, R., & PulvermÃ¼ller, F. (2012). Body part specific representations of semantic noun categories. Journal of Cognitive Neuroscience, 24 (6), 1492 1509. doi: 10.1162/jocn_a_00219 Chiarello, C., Chur, K. L., & Hoyer, W. J. (1985). Automatic and controlled semantic priming: Accuracy, response bias, and aging. Journal of Gerontology, 40 (5), 593 600. Chun, M. M. (2000). Contextual cueing of visual attention. Trends in cognitive sciences, 4 (5), 170 178. Connelly, S. L., Hasher, L., & Zacks, R. T. (1991). Age and reading: The impact of distraction. Psychology and Aging , 6(4), 533 541. Druks, J. (2000). Object and action naming battery . London, UK: Psychology Press. Duffy, S. A., Morris, R. K., & Rayner, K. (1988). Lexical ambiguity and fixation times in readin g. Journal of memory and language, 27 (4), 429 446. Dunn, D. M., & Dunn, L. M. (1997). Peabody picture vocabulary test (3rd ed.). Circle Pines, MN: American Guidance Service. Edmonds, L. A., & Mizrahi, S. (2011). Online priming of agent and patient themati c roles and related verbs in younger and older adults. Aphasiology, 25 (12), 1488 1506. Faustmann, A., Murdoch, B. E., Finnigan, S. P., & Copland, D. A. (2007). Effects of advancing age on the processing of semantic anomalies in adults: Evidence from event related brain
96 potentials. Experimental Aging Research, 33 (4), 439 460. doi: 10.1080/03610730701525378 Federmeier, K. D., & Kutas, M. (2005). Aging in context: Age related changes in context use during language comprehension. Psychophysiology, 42 (2), 133 141. Federmeier, K. D., McLennan, D. B., Ochoa, E., & Kutas, M. (2002). The impact of semantic memory organization and sentence context information on spoken langu age processing by younger and older adults: An ERP study. Psychophysiology, 39 (2), 133 146. Federmeier, K. D., Wlotko, E. W., De Ochoa Dewald, E., & Kutas, M. (2007). Multiple effects of sentential constraint on word processing. Brain Research, 1146 , 75 8 4. Fernandino, L., Conant, L. L., Binder, J. R., Blindauer, K., Hiner, B., Spangler, K., & Desai, R. action verbs. Brain and Language, 127 (1), 65 74. doi: 10.1016/j.bandl. 2012.07.008 Ferraro, K. F. (1980). Self ratings of health among the old and the old old. Journal of Health and Social Behavior, 21 , 377 383. Ferretti, T. R., McRae, K., & Hatherell, A. (2001). Integrating verbs, situation schemas, and thematic role concep ts. Journal of Memory and Language, 44 (4), 516 547. Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14 (2), 178 210. doi: 10.1016/0010 0285(82)90008 1 Gardner, H. E., Ralph, M. A. L., Dodds, N., Jones, T., Ehsan, S., & Jefferies, E. (2012). The differential contributions of pFC and temporo parietal cortex to multimodal semantic control: exploring refractory effects in semant ic aphasia. Journal of cognitive neuroscience, 24 (4), 778 793. Gentner, D. (1981). Some interesting differences between verbs and nouns. Cognition and Brain Theory, 4 (2), 161 178. Gentner, D. (1982). Why nouns are learned before verbs: Linguistic relativ ity versus natural partitioning. In S. A. Kuczaj (Ed.), Language development: Vol. 2. Language, thought, and culture (pp. 301 334). Hillsdale, NJ: Erlbaum. Gentner, D., & Boroditsky, L. (2001). Individuation, relativity, and early word learning. In M. Bowe rman & S. C. Levinson (Eds.), Language acquisition and conceptual development (Vol. 3, pp. 215 256). Cambridge: Cambridge: Cambridge University Press. Gibson, E. J. (1963). Perceptual learning. Annual Review of psychology, 14 (1), 29 56. Gomez, R. (2002). Word frequency effects in priming performance in young and older adults. The Journals of Gerontology: Psychological Sciences and Social Sciences, 57 (3), 233 240.
97 Hagoort, P. (1993). Impairments of lexical semantic processing in aphasia: Evidence from the processing of lexical ambiguities. Brain and Language, 45 , l89 232. Harley, T. A. (2008). The psychology of language from data to theory (3rd ed.). New York, NY: Psychology Press: Taylor & Francis Group. Hasher, L. H., & Zacks, R. T. (1988). Working memor y, comprehension, and aging: A review and a new view. Psychology of Learning and Motivation , 22, 193 225. Helm Estabrooks, N. (2001). Cognitive Linguistic Quick Test . New York, NY: The Psychological Corporation. Herlofsky, S. M., & Edmonds, L. A. (2012). A ctivating Situation Schemas: The Effects of Multiple Thematic Roles on Related Verbs in a Continuous Priming Paradigm. Journal of Psycholinguistic Research , 1 19. Howard, D., & Patterson, K. E. (1992). The Pyramids and Palm Trees Test: A test of semantic access from words and pictures . London, UK: Harcourt Assessment. Howard, D. V., McAndrews, M. P., & Lasaga, M. I. (1981). Semantic priming of lexical decisions in young and old adults. Journal of Gerontology, 36 (6), 707 714. Huettig, F., Rommers, J., & Me yer, A. S. (2011). Using the visual world paradigm to study language processing: A review and critical evaluation. Acta Psychologica. doi: 10.1016/j.actpsy.2010.11.003. Jefferies, E., Baker, S. S., Doran, M., & Lambon Ralph, M. A. (2007). Refractory effect s in stroke aphasia: A consequence of poor semantic control. Neuropsychologia, 45 (5), 1065 1079. doi: 10.1016/j.neuropsychologia.2006.09.009 Jefferies, E., & Lambon Ralph, M. A. (2006). Semantic impairment in stroke aphasia versus semantic dementia: a case series comparison. Brain, 129 (8), 2132 2147. Jiang, J., Sachdev, P., Lipnicki, D. M., Zhang, H., Liu, T., Zhu, W., . . . Reppermund, S. (2014). A longitudinal study of brain atrophy over two years in community dwelling older individuals. NeuroImage, 86 , 203 211. Kalinyak Fliszar, M., Kohen, F., & Martin, N. (2011). Remediation of language processing in aphasia: Improving activation and maintenance of linguistic representations in (verbal) short term memory. Aphasiology, 25 (10), 1095 1131. Kamide, Y., Al tmann, G., & Haywood, S. L. (2003). The time course of prediction in incremental sentence processing: Evidence from anticipatory eye movements. Journal of Memory and Language, 49 (1), 133 156. Kemper, S., & Kempter, K. A. (1999). Limitations on syntactic p rocessing. In Kemper, S., Kliegl, R., (eds). Constraints on lanauge: Aging, grammar and memory. Boston: Kluwer Academic Publishers. pp. 79 106.
98 Kersten, A. W. (1998). A division of labor between nouns and verbs in the representation of motion. Journal of Ex perimental Psychology: General, 127 , 34 54. Kersten, A. W., & Earles, J. L. (2004). Semantic context influences memory for verbs more than memory for nouns. Memory & cognition, 32 (2), 198 211. Khader, P., Scherag, A., Streb, J., & RÃ¶sler, F. (2003). Diff erences between noun and verb processing in a minimal phrase context: a semantic priming study using event related brain potentials. Cognitive Brain Research, 17 (2), 293 313. doi: http://dx.doi.org/10.1016/S0926 6410(03)00130 7 Kintsch, W. (2001). Predicat ion. Cognitive Science, 25 (2), 173 202. Kipper, K. (2005). VerbNet: A broad coverage, comprehensive verb lexicon. (PhD Dissertation). Kipper, K., Dang, H. T., & Palmer, M. (2000). Class based construction of a verb lexicon. Paper presented at the American Association from Artificial Intelligence. Koenig, J. P., Mauner, G., & Bienvenue, B. (2003). Arguments for adjuncts. Cognition, 89 (2), 67 103. doi: 10.1016/S0010 0277(03)00082 9 Laver, G. D. (2000). A speed accuracy analysis of word recognition in young a nd older adults. Psychology and Aging, 15 (4), 705 709. Laver, G. D., & Burke, D. M. (1993). Why do semantic priming effects increase in old age? A meta analysis. Psychology and Aging, 8 (1), 34 43. Lee, C. L., & Federmeier, K. D. (2011). Differential age effects on lexical ambiguity resolution mechanisms. Psychophysiology, 48 (7), 960 972. doi: 10.1111/j.1469 8986.2010.01158.x. Levin, B. (1993). English verb classes and alternation: A preliminary investigation. The University of Chicago Press. Linnville, S. E. (1995). Are there age related processing differences in semantic priming? Perceptual and Motor Skills, 80 (2), 585 586. MacKay, D., & Burke, D. M. (1990). Chapter five cognition and aging: A theory of new learning and the use of old connections. Aging and Cognition , 71, 213 263. Madden, D. J. (2001). Speed and timing of behavioral processes. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (5th ed., pp. 288 312). San Diego, CA: Academic Press. Marner, L., Nyengaard, J. P., Tang , Y., & Pakkenberg, B. (2003). Marked loss of myelinated nerve fibers in the human brain with age. Journal of Comparative Neurology, 462(2), 144 152.
99 McRae, K., Cree, G. S., Seidenberg, M. S., & McNorgan, C. (2005). Semantic feature production norms for a large set of living and nonliving things. Behavior Research Methods, 37 (4), 547 559. McRae, K., Hare, M., Elman, J. L., & Ferretti, T. (2005). A basis for generating expectancies for verbs from nouns. Memory & Cognition, 33 (7), 1174 1184. Moseley, R. L. , Mohr, B., Lombardo, M. V., Baron Cohen, S., Hauk, O., & PulvermÃ¼ller, F. (2013). Brain and behavioral correlates of action semantic deficits in autism. Frontiers in Human Neuroscience, 7 , 1 10. doi: 10.3389/fnhum.2013.00725 MÃ¼ller, G., Richter, R. A., We isbrod, S., & Klingberg, F. (1991). Impaired eye tracking performance in patients with presenile onset dementia. International Journal of Psychophysiology, 11 (2), 167 177. doi: 10.1016/0167 8760(91)90009 M Munoz, D., Broughton, J., Goldring, J., & Armstron g, I. (1998). Age related performance of human subjects on saccadic eye movement tasks. Experimental Brain Research, 121 (4), 391 400. Myerson, J., Hale, S., Chen, J., & Lawrence, B. (1997). General lexical slowing and the semantic priming effect: The role s of age and ability. Acta Psychologica, 96 (1), 83 101. Nagy, W., & Gentner, D. (1990). Semantic constraints on lexical categories. Language and Cognitive Processes, 5 (3), 169 201. Nelson, D., McEvoy, C., & Schreiber, T. (1998). The University of South F lorida word association, rhyme, and word fragment norms. . http://www.usf.edu/FreeAssociation/ . Park, H., Cowles, W. H., & Edmonds, L. A. (2013). Comparing Semantic and Syntactic Expectation between Verbs and Thematic Roles: Evidence from Eyetracking . Pap er presented at the Clinical Aphasiology Conference, Tucsan, AZ. Park, H., & Edmonds, L. A. (2012). Priming effects on semantic relatedness: Comparing verb and noun primes in picture naming . Paper presented at the Academy of Aphasia, San Francisco, CA. P ulvermÃ¼ller, F. (2005). Brain mechanisms linking language and action. Nature Reviews Neuroscience, 6 (7), 576 582. doi: 10.1038/nrn1706 Rayner, K. (2009). Eye movements and attention in reading, scene perception, and visual search. The Quarterly Journal of Esperimental Psychology, 62(8), 1457 1506. Rayner, K., Binder, K. S., & Duffy, S. A. (1999). Contextual strength and the subordinate bias effect: Comment on Martin, Vu, Kellas, and Metcalf. The Quarterly Journal of Experimental Psychology Section A : Human Experimental Psychology, 52 (4), 841 852.
100 Rayner, K., Carlson, M., & Frazier, L. (1983). The interaction of syntax and semantics during sentence processing: Eye movements in the analysis of semantically biased sentences. Journal of Verbal Learning and Ver bal Behavior, 22 (3), 358 374. Raz, N. (2000). Aging of the brain and its impact on cognitive performance: Integration of structural and functional findings. In Craik, F. I. M. & Salthouse, T. A. (Eds). The handbook of aging and cognition (2 nd ed.). Lawren ce Erlbaum Associates Publishers. pp.1 90. Roelofs, A. (1997). The WEAVER model of word form encoding in speech production. Cognition, 64 (3), 249 284. Rosenwaike, I. (1985). A demographic portrait of the oldest old. The Milbank Memorial Fund Quarterly. He alth and Society, 63 (2), 187 205. Rumelhart, D. E. (1980). Schemata: The building blocks of cognition . Hillsdale, NJ: Erlbaum. Salthouse, T. A. (1996). The processing speed theory of adult age differences in cognition. Psychological review, 103 (3), 403 42 8. Salthouse, T. A. (2000). A theory of cognitive aging . Amsterdam: North Holland. Samorajski, T. (1977). Central neurotransmitter substances and aging: a review. Journal of the American Geriatrics Society, 25 (8), 337 348. Sanford, A. J., & Garrod, S. C. (1981). Understanding written language: Explorations of comprehension beyond the sentence . New York, NY: Wiley SchÃ¼tze, C. T. (1995). PP attachment and argumenthood. MIT Working Papers in Linguistics, 26 , 95 151. Sedivy, J., & Spivey Knowlton, M. (1994) . The use of structural, lexical, and pragmatic information in parsing attachment ambiguities. In C. Clifton, L. Franzier & K. Rayner (Eds.), Perspectives on sentence processing (pp. 389 413). Hillsdale, NJ: Lawrence Erlbaum. Snodgrass, J. G., & Vanderwart , M. (1980). A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of experimental psychology: Human learning and memory, 6 (2), 174 215. Spivey Knowlton, M., & Sedivy, J. C. (1995). Reso lving attachment ambiguities with multiple constraints. Cognition, 55 (3), 227 267. Sussman, R. S. (2006). Verb Instrument Information during on line processing. (Ph.D.), Unversity of Rochester. Suzman, R., & Riley, M. W. (1985). Introducing the" oldest old". The Milbank Memorial Fund quarterly. Health and society, 63 (2), 177 186.
101 Talmy, L. (1975). Semantics and syntax of motion. In J. Kimball (Ed.), Syntax and semantics (Vol. 4). London: Academic Press. Tanenhaus, M. K., & Trueswell, J. C. (2005). Eye movements as a tool for bridging the language as product and language as action traditions. In Trueswell, J. C. & Tanenhaus, M. K. Approaches to studying world situated language use: Bridging the language as product and language as action traditions. Cambr idge, MA: MIT Press. pp.3 37. Tanenhaus, M. K., Carlson, G., & Trueswell, J. C. (1989). The role of thematic structures in interpretation and parsing. Language and Cognitive Processes, 4 (3 4), 211 234. Tree, J. J., & Hirsh, K. W. (2003). Sometimes faster, sometimes slower: Associative and competitor priming in picture naming with young and elderly participants. Journal of Neurolinguistics, 16 (6), 489 514. Tun, P., O Kane, G., & Wingfield, A. (2002). Distraction by competing speech in young and older adult listeners. Psychology and Aging , 17(3), 453 467. Verhaeghen, P. (2003). Aging and vocabulary score: A meta analysis. Psychology and Aging, 18 (2), 332 339. Vinson, D. P., & Vigliocco, G. (2008). Semantic feature production norms for a large set of objects and events. Behavior Research Methods, 40 (1), 183 190. doi: 10.3758/BRM.40.1.183 Wechsler, D. (1997). Wechsler memory scale (WMS III) . San Antonio, TX: Psychological Corporation. Wilson, M. (1988). MRC Psycholinguistic Database: Machine usable dictionary, version 2.00. Behavior Research Methods, 20 (1), 6 10.
102 BIOGRAPHICAL SKETCH Hyejin Park received her Bachelor of Arts degree in Korean Language and Literature from Seoul Women s University in Seoul, South Korea in 2003. Later, she became interested in t he field of communication disorders and was determined to continue her education in Speech Language Pathology at Yonsei University in Seoul, South Korea where she obtained her Master of Science in 2007. Hyejin desired to e xperience and study the field at a n advance level and therefore came to the U.S. for her doctoral degree. She enrolled in a doctoral program in Speech, Language, and Hearing Sci ences at University of Florida and is expected to complete her Ph.D. in August 2014. During her doctoral program , her research interests included neurogenic language disorders and semantic processing related to word retrieval difficulty , especially verb and noun concept relationships . Hyejin s research during the doctoral program investigated th e semantic processing of healthy younger and older adults , and she would like to expand her research with clinical populations with language disorders. Upon graduation, she will work as an assistant professor at Radford University, VA and continue her rese arch and teaching in the field.