top lang disorders vol. 36, no. 1, pp. 65–79 prefrontal ... · the prefrontal cortex (pfc) during...

Top Lang DisordersVol. 36, No. 1, pp. 65–79Copyright c© 2016 Wolters Kluwer Health, Inc. All rights reserved.

Prefrontal Cortical ActivityDuring Discourse ProcessingAn Observational fNIRS Study

Michael S. Cannizzaro, Shaun R. Stephens,Max Breidenstein, and Cori Crovo

Discourse is a commonly occurring and cognitively demanding form of naturalistic communication(e.g., conversation, event narration, personal and fictional narratives, text reading/generation). Be-cause of the prevalence of these communication acts in daily routines (e.g., educational, vocational,and social), disrupted discourse is an important target for treatment of persons with cognitive-communication (CC) disorders (American Speech-Language Hearing Association, 2005). However,there is a paucity of information with regard to the underlying cognitive architecture and process-ing demands associated with the various forms of discourse in non-brain-injured adults. This is, inpart, due to a number of methodological constraints of functional neuroimaging technologies suchas functional magnetic resonance imaging that severely limit ecologically typical communicationacts such as listening and speaking during scanning. To circumvent these issues, this pilot inves-tigation used functional near infrared spectroscopy (fNIRS) to monitor hemodynamic activity inthe prefrontal cortex (PFC) during natural discourse-processing tasks in 13 neurologically healthyadult participants. Task demands were manipulated across a variety of discourse types to eluci-date the associated neural and cognitive resources. Results indicate that the comprehension ofwell-organized discourse text is minimally demanding on the PFC. However, discourse productionplaces a significant burden on the PFC and these processing demands generally reflect the relativecomplexity of the discourse task. These findings are discussed in terms of potential clinically rel-evant implications with regard to the elicitation, assessment, and remediation of CC impairmentsin clinical populations. Key words: brain injury, cognitive-communicative disorders, corticalactivity, discourse, functional near infrared spectroscopy, speech–language pathology

DISCOURSE AND NEUROIMAGING

Discourse is a cognitively demandingform of naturalistic communication, com-monly utilized in daily activities across ed-

Author Affiliations: Department of CommunicationSciences and Disorders (Dr Cannizzaro, Mr Stephens,and Ms Crovo), Center for Clinical and TranslationalScience (Mr Stephens), and UndergraduateNeuroscience Program (Dr Cannizzaro and MrBreidenstein), University of Vermont, Burlington.

The authors have indicated that they have no financialand no nonfinancial relationships to disclose.

Corresponding Author: Michael S. Cannizzaro, PhD,Department of Communication Sciences and Disorders,University of Vermont, 489 Main St, 401 Pomeroy Hall,Burlington, VT 05405 ([email protected]).

DOI: 10.1097/TLD.0000000000000082

ucational, vocational, and social circum-stances (Ash et al., 2006; Basso Moro, Cu-tini, Ursini, Ferrari, & Quaresima, 2013;Ylvisaker, Szekeres, & Feeney, 2001). Com-plex goal-directed communication manifestsin various forms of receptive and expres-sive discourse, such as conversation, pro-cedural and script narratives, and the cre-ation and comprehension of personal andfictional story narratives (Coelho, Ylvisaker,& Turkstra, 2005). Typical discourse pro-cessing engages classical language regionsof the perisylvian cortex mediated by theprefrontal cortex (PFC) and executive func-tion abilities that integrate cognitive infor-mation by organizing and combining succes-sive sentences or utterances (Biddle, McCabe,& Bliss, 1996; Cannizzaro, Coelho, & Youse,2002; Chapman et al., 2001; Coelho et al.,

Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

65

mailto:[email protected]

66 TOPICS IN LANGUAGE DISORDERS/JANUARY–MARCH 2016

2005; Snow, Douglas, & Ponsford, 1998;Tucker & Hanlon, 1998).

Most scientific studies of language over thepast 50 years have taken the form of investi-gations at the levels of words and single sen-tences (AbdulSabur et al., 2014). However,it is at the level of discourse where the prag-matic properties of language naturally emergeand where language systems and cognitiveprocesses, such as organizational schemas, arelikely to interact (AbdulSabur et al., 2014).The present study is an exploratory, quanti-tative, observational study of brain functionduring naturalistic discourse processing andproduction in non-brain-injured adults. It isinnovative in the use of a neuroimaging tech-nology, functional near infrared spectroscopy(fNIRS), that allows for high-resolution mon-itoring of brain function during naturalisticacts of receptive and expressive spoken dis-course (e.g., storytelling). Studying realisticspoken discourse communication is typicallynot possible using other neuroimaging tech-nologies. This represents a leap forward inecological validity, as typical discourse situa-tions can be used to reveal underlying brainfunction. This study sets the stage for fu-ture investigations of the neural underpin-nings of discourse processing in adults withand without traumatic brain injury (TBI) toinvestigate cognitive-communication impair-ments that are common in the brain-injuredpopulation. In persons with brain injury, im-pairments in discourse have been noted tooccur despite intact linguistic abilities at theword and sentence levels and are thought toreflect a disruption of the interplay betweenlanguage skills and cognitive processes, mostnotably disrupted executive functions and im-paired function of the PFC (Coelho, 2005;Galetto, Andreetta, Zettin, & Marini, 2013;Marini, Zettin, & Galetto, 2014; Ylvisaker,Feeney, & Capo, 2007; Ylvisaker et al., 2001).

Analysis of discourse is useful in illustratinghow context and structure influence neuralactivity and how neural activity can changeand evolve over the course of a lengthier com-munication task (Ferstl, Neumann, Bogler, &von Cramon, 2008; Mason & Just, 2013; Xu,

Kemeny, Park, Frattali, & Braun, 2005). Aslanguage is understood in context, increasingneural activity and spreading activation are ob-served beyond the left hemisphere perisylvianareas as part of a larger “extended languagenetwork (ELN)” (Ferstl et al., 2008, p. 581).Ferstl and colleagues identified a number of re-gions in the ELN thought to be uniquely usedin discourse to support coherence-buildingprocesses, including the posterior cingulatecortex, inferior precuneus (BA 23/31), andaspects of the dorsomedial PFC (BA 9/10;Ferstl et al., 2008). Xu et al. (2005) observedsimilar activation patterns during story narra-tive processing: Bilateral activity in the dor-sal and ventral medial prefrontal regions (BA8–10) became increasingly engaged, as lan-guage tasks required higher levels of integra-tion (i.e., words < sentences < narratives).These prefrontal regions also respond to vari-ations in discourse processing as an indicationof task demand. During the reading of sim-ple narratives, as the organizational structureof the discourse increases story predictability,fewer processing demands are placed on thePFC (e.g., decreased activation; Cannizzaro,Dumas, Prelock, & Newhouse, 2012). Theseresults suggest a distinct bilateral processingnetwork comprising medial prefrontal cor-tical areas as well as the precuneus, impli-cated in binding subcomponents of discourseinto a cohesive whole (Bottini et al., 1994;Cannizzaro et al., 2012; Ferstl & von Cra-mon, 2001, 2002; Xu et al., 2005). The PFCappears to support organizational and coher-ence functions in cognition and, more specifi-cally, in communication as successive utter-ances are woven together to create a nar-rative (Silbert, Honey, Simony, Poeppel, &Hasson, 2014; Troiani et al., 2008). Similarly,mentalizing (perceiving and interpreting oth-ers’ behavior as part of their intentional men-tal states) and the construction of situationalmodels in discourse (e.g., context, chronol-ogy) are reliant on PFC activity (AbdulSaburet al., 2014). In addition, PFC networksare involved in coarse semantic processing(predominantly right temporal), coherencemonitoring (dorsolateral prefrontal), text


PFC in Discourse: An fNIRS Study 67

integration (left-frontotemporal), characterperspective monitoring (e.g., theory of mind;medial frontal), and spatial imagery (interpari-etal sulcus; Mason & Just, 2006).

Approaches unifying functional neuroimag-ing and functional task performance haveincreased awareness of component-specificneural networks involved in discourse pro-cessing (Mason & Just, 2006, 2013). Whereasthese findings contribute to our understand-ing of neural activity during discourse process-ing, the tasks used to produce functional brainimaging results in most discourse investiga-tions are not representative of typical every-day communication. That is, listening to pre-recorded narratives or reading narrative textsin a functional magnetic resonance imaging(fMRI) scanner is quite different from commu-nicating in naturalistic contexts (e.g., recount-ing life events with personal narratives in aface-to-face interaction). Until recently, neu-roimaging during discourse production wasimpossible due to head motion artifacts in-troduced into the imaging data during speak-ing. Therefore, nearly all research on the cog-nitive architecture underlying discourse wasderived from studies of receptive commu-nication in the form of silent text readingor expressive communication thinking aboutoverlearned and rehearsed stories while lyingsilently in the scanner.

In most functional neuroimaging investiga-tions of discourse, typically using fMRI, eco-logical validity is severely limited because ofthe restrictive nature of the equipment thatdoes not allow for speaking or direct inter-active communication (Quaresima, Bisconti,& Ferrari, 2012). In the past 5 years, how-ever, a relatively new and noninvasive func-tional neuroimaging technology has seen in-creasing use in language studies, in the form offNIRS. Functional near infrared spectroscopyis an optical neuroimaging modality that cancircumvent many of the most problematicissues in neuroimaging communication be-cause it is robust to the effects of motionartifacts from head and muscle movementsand is silent and therefore not conducted ina loud and isolating scanning environment

(see Quaresima, Bisconti, & Ferrari, 2012 fora review). In fact, fNIRS can easily be em-ployed in clinical, office, or bedside testingsituations, as it is portable and flexibly de-ployed for a variety of experimental condi-tions. Utilizing a matrix of light source andlight detector pairs, fNIRS passes specified fre-quencies of infrared light through the biolog-ical tissue of the cortex, providing calculatedabsorption data, indicating changes in con-centrations of oxygenated and deoxygenatedhemoglobin (Quaresima, Bisconti, & Ferrari,2012). As such, fNIRS provides noninvasive,nonionizing hemodynamic response data thatcan be obtained in a wide variety of ecologi-cally valid circumstances (Irani, Platek, Bunce,Ruocco, & Chute, 2007). The use of fNIRS isnot without limitations, as it has poorer spa-tial resolution than fMRI (∼1 cm vs. 1 mm)and is restricted to monitoring activity onlyin cortical areas under the optode pairs. How-ever, it offers researchers real-time monitoringof functional brain activity during naturalisticcommunication situations not possible withany other neuroimaging technologies.

The purpose of the present study was to usefNIRS to observe and characterize prefrontalhemodynamic activity during clinically rele-vant discourse tasks: a receptive task involv-ing reading sentences and short narratives, anexpressive task of formulating discourse, andan expressive narrative generation task. Threehypotheses were investigated:

1. During receptive tasks, reading cohesivenarratives will elicit greater PFC activa-tion than reading single sentences. Thishypothesis is based on the majority ofprevious fMRI research, suggesting thatthe integration of successive informationand structure-building increase process-ing demands in the PFC (Ferstl et al.,2008; Xu et al., 2005; contrasting withCannizzaro et al., 2012).

2. Expressive discourse production will besignificantly more demanding on the PFCthan receptive discourse comprehension(AbdulSabur et al., 2014; Coelho, 2002;Troiani et al., 2008; Van Leer & Turkstra,1999).



3. Different genres of expressive discoursetasks will induce differing profiles of cor-tical oxygen consumption based on fluc-tuating task demands, as is seen in be-havioral data (Coelho, 2002; Van Leer& Turkstra, 1999). See Figure 1 for aschematic of the study design.

MATERIALS AND METHODS

Participants

Thirteen healthy adult female subjects(mean = 22.8 years; range = 20–38 years)with 3–5 years of college-level education par-ticipated in this study. Informed consent wasobtained in accordance with the protocol-approved process of the Research ProtectionsOffice at the University of Vermont. All par-ticipants were primary speakers of AmericanEnglish, without personal history of brain in-jury, neurological impairment, psychiatric dis-order, or learning disability.

fNIRS instrumentation and signalprocessing

A 16-channel continuous wave fNIRS sys-tem (Biopac fNIRS 100; Biopac Systems, Go-leta, CA) was used to monitor baseline-relativechanges in the concentration of oxygenated(HbO2) and deoxygenated (Hb) blood flowin the PFC. This device uses light-emittingdiodes (LED optode sources) to transmit lightwaves through biological tissue. Two detec-tors are paired with each of eight LED sourcesat a distance of 2.5 cm, measuring changes

in optical density (OD) as the signal passesthrough tissue and cerebral blood sources.The fNIRS headband consists of 16 sensor de-tector pairings that are created from four LEDsources and 10 OD detectors and was fastenedon participants according to optimal position-ing guidelines following the manufacturer’sinstructions. Using the international 10–20system for standardized electrode placementin electroencephalography, the lower row ofchannels is placed on the forehead so that thehorizontal axis (the central y-axis) coincideswith the symmetry axis of the head (i.e., be-tween the eyes). On the vertical z-axis, thesensor was positioned at Fpz, with Fp1 andFp2 marker locations positioned on the bot-tom voxel row (Ayaz et al., 2006; Chatrian,Lettich, & Nelson, 1985; Homan, Herman, &Purdy, 1987). The horizontal axis of the lowerrow of channels is then aligned with Fp1, Fpz,and Fp2. The detectors cover PFC areas BA 44,45, 9, and 10 (fNIR Devices LLC, 2013). Softfoam tape was used to cover the outer sur-face of the device and prevent the intrusionof ambient light to the detectors.

Experimental procedures

Participants were given instructions regard-ing task participation and asked to refrainfrom making large body movements or fore-head movement. Subjects were seated at acomfortable distance from a 20-in. computermonitor that reiterated instructions in text.Participants were asked to relax and look pas-sively at a fixation cross in the middle of thescreen as a baseline measurement was taken.

Figure 1. Schematic of the study design. NYC = New York City.



This fixation cross is a task-free rest periodbetween communication tasks and lasted for15 s each appearance before and after all tasks.Once the software achieved a stable baselinemeasurement, participants were prepared fortheir first task and were asked to indicatewhen they were ready to begin. Then the ex-aminer initiated the timed stimuli sequence,which ran as an automated slide-show presen-tation delivered via PowerPoint. The fNIRSOD data recording was initiated simultane-ously with the stimuli. Four experimental con-ditions were chosen, two receptive tasks andtwo expressive tasks, which included read-ing of noncohesive sentences, reading well-constructed simple narratives, generating anarrative based on a pictured scene, and gen-erating discourse given a situational prompt(described later).

For the receptive condition, six narrativeswere constructed, each identical in length(five sentences, 60 words), based on the orig-inal work of Hickmann and Schneider (2000)and Schneider and Winship (2002). Thesestories were designed to conform strictly tostory grammar conventions by creating a sim-ple story narrative, each a single completeepisode. Following the model of simple chil-dren’s narratives, the stories included con-crete high-frequency words and simple syn-tactic structures. Each narrative included in-troductory and setting information, followedby an initiating event (problem introduced),an attempt (to solve the problem), a directconsequence (success or failure at attempt),and an internal response of the character(reflective thought following episode con-clusion). Half of these narratives were thenpseudo-randomly intermixed at the sentencelevel. This created the three “sentence” con-dition tasks that resembled the true narrativesin length, sentence structure, and contentbut were not narratives because successivesentences did not cohere. Participants werepresented with three blocks of stimuli thatincluded all six receptive constructs (threesets of unconnected sentences, three narra-tives). Stimuli were presented in a pseudo-randomized order. Each receptive task lasted

30 s, followed by 15 s of rest (i.e., visual fixa-tion of “+” on the screen). Participants wereinstructed to pay attention to each item thatappeared on the computer screen and try toread each word of each narrative silently andattentively without moving their mouth. Par-ticipants were not required to respond in anyway.

For the first expressive discourse task, par-ticipants were instructed to verbally describethe steps involved in planning a trip to NewYork City (NYC). This procedure is based onthe work of Kiran, Harris, Marquardt, andPalmieri (2005) and has been found to be asensitive measure of complex communicationuseful in differentiating healthy participantsfrom persons with mild cognitive impairment(Fleming & Harris, 2008). The following in-structions were given to the participants:

Imagine that you are going on a vacation a weekfrom now. You are traveling to NYC for a 2-weekstay. Think about all you will have to do to getready to go, such as how you will get there, whatyou will bring, and what you will do. When theslide changes, describe in detail all the activitiesassociated with the trip.

Participants were given 60 s to completethis task.

The second expressive task was a narrativetask that included a visual of Norman Rock-well’s painting The Runaway. The followinginstructions were given to participants:

Tell me a story about the picture on the screen.The scene in this picture represents a momentin time. Something happened to cause the pic-tured event and something is going to happen af-terwards. When the slide changes, please tell methe whole story from what happened before thepictured event through what will happen after thisscene.

Participants were given 60 s to completethis task. Standard measures of narrative mi-crostructure and macrostructure were ob-tained. Discourse samples were recorded andtranscribed verbatim.

Transcripts for both expressive discoursetasks were then distributed into T-units,which are more reliably identified than



sentences in spoken language (Hughes,McGillivray, & Schmidek, 1997). A T-unitconsists of an independent clause plus anyassociated dependent clauses (Hunt, 1970).Verbal output and efficiency were measuredas the total number of words, number ofedited words, total number of T-units, totalnumber of T-units within episode structure.For the story narrative generation (i.e., Run-away condition), story grammar knowledgewas assessed. Story grammar knowledgereflects the regularities in the internal struc-ture of stories that guide an individual’scomprehension and production of the logicalrelationships (i.e., causal and temporal)between people and events (Stein & Glenn,1979). The number of complete episodes,incomplete episodes, and the proportionof T-units within episode structure weremeasured. An episode is judged as completeif it contains all three, logically related com-ponents: (a) an initiating event that prompts acharacter to formulate a goal-directed behav-ioral sequence, (b) an action or an attemptat achieving the goal, and (c) a direct conse-quence marking attainment or nonattainmentof the goal. Incomplete episodes containonly two of the three related story grammarelements. For the procedural NYC discoursegeneration, an inventory of core elementswas tabulated on the basis of the work ofFleming and Harris (2008). These discourseelements were used to measure discourseefficiency.

Brain activation maps were derived fromHbO2 data calculated using a modified Beer–Lambert law from the 16 midpoint locationsbetween the source and detector pairs, at asampling rate of 2 Hz. All recordings weremade following a short device-testing periodto ensure adequate signal characteristicsfor the OD data (i.e., registered steady-statevalues between 800 and 4,000 mV during ano-task condition (fNIR Devices LLC, 2013).All fNIRS OD signal data were captured andrecorded through the Cognitive Optical BrainImaging Studio (COBI) software package(fNIR Devices LLC, 2013). Real-time mon-itoring of the raw and calculated percentchange signal values was used for initial

quality assurance for detection of motionartifacts and potential loss or oversaturationof signal due to poor headband contactwith the participant’s forehead. Data analysiswas conducted using Statistical ParametricMapping for Near Infrared Spectroscopysoftware package (NIRS-SPM Version 4.1;Ye, Tak, Jang, Jung, & Jang, 2009) accordingto manual specifications for best practice(Tak, 2011). Data from each channel for eachparticipant were inspected in the NIRS-SPMtime-series window for signal integrity andcorrespondence to the experimental designtimeline. NIRS-SPM wavelet minimum de-scription length (wavelet-MDL) detrendingwas applied to each data set to remove signalfluctuations related to nonrelevant biologicalsignals (e.g., breathing, cardiac, vasomotion;Jang et al., 2009). To correct for serial corre-lations in the data, precoloring of the data setwas used with low-pass filter, hemodynamicresponse function shape, and full width athalf maximum (fwhm) = 4 s (Ye et al., 2009).Individual activation maps and significant βweights are all reported with significancelevels set at p < .05, using Lipschitz–Killingcurvature with Euler characteristics to ac-count for expected correlations (Tak, 2011).

RESULTS

Summary microstructural linguistic data onthe two expressive discourse tasks are pre-sented in Table 1. There is relative consistencyon task performance: Participants produceda similar number of T-units and number oftotal words, T-units: t(12) = 1.27, p > .05,words: t(12) = 1.56, p > .05, under both con-ditions. Scores for word efficiency (i.e., totalwords/edited words: due to false starts, rep-etitions, restatements, etc.) were significantlydifferent between tasks, t(12) = 2.76, p =.008, indicating that participants were moreefficient in their word choice when produc-ing the Runaway story. For both tasks, effi-ciency scores are quite high at approximately95% or better. Overall, the Runaway narrativeelicited greater word productivity.

Macrostructural analysis of the expressivediscourse forms revealed that for the NYC



Tab

le1

.Ex

pre

ssiv

ed

isco

urs

ean

dn

arra

tive

pro

du

ctio

nsu

mm

ary

dat

a

Dis

cou

rse:

Tri

pto

NY

CSt

ory

T-U

nit

sT

ota

lW

ord

sE

dit

edW

ord

sW

ord

Pro

du

ctiv

ity

Sto

ryE

lem

ents

Sto

ryE

ffici

ency

Ave

rage

(SD

)8.

75(3

.95)

114.

46(5

2.28

)10

7.46

(47.

56)

94.7

4%(0

.03)

6.77

(3.1

1)59

.31%

(0.2

6)N

arra

tive

:W

ord

Co

mp

lete

Inco

mp

lete

%C

om

ple

teSt

ory

tell

ing

Ru

naw

aySt

ory

T-U

nit

sT

ota

lW

ord

sE

dit

edW

ord

sP

rod

uct

ivit

yE

pis

od

esE

pis

od

esE

pis

od

esE

ffici

ency

Ave

rage

(SD

)10

.77

(3.9

4)13

4.61

(49.

35)

131.

54(4

9.47

)97

.36%

(0.0

1)1.

46(0

.66)

1.77

(1.1

7)52

.15%

(0.2

5)79

.77%

(0.0

9)

Note

.NY

C=

New

Yo

rkC

ity.

discourse, participants produced an averageof 6.77 core discourse elements with a relativeefficiency (T-units that include core or sup-plemental elements/total T-units) of roughly60%. For the Runaway story, participants pro-duced on average 1.46 complete episodes,1.77 incomplete episodes per story, and hada storytelling efficiency (number of T-units inepisode structure/total number of T-units) ofapproximately 80%. As with word-level mea-sures of efficiency, discourse generation ef-ficiency was significantly better for the Run-away story task, t(12) = 3.12, p = .004.

Hypothesis 1 was that reading unconnectedsentences would place a greater processingburden on the PFC than cohesive discourse,as indicated by increased cortical oxygenconsumption (HbO2). Data were analyzed toassess for greater cortical activation in theconnected discourse than in the sentence con-dition (i.e., Discourse minus Sentence). Signif-icant cortical activations measured via fNIRSwere observed in only 3 of the 13 participants(Table 2). Across these three participants,no clear cortical activation pattern emerged.Among individuals who did show activations,activation patterns were in the medial ante-rior PFC (two left medial activation, and oneright medial activation; Figure 2).

Hypothesis 2 was that expressive dis-course production would be significantlymore demanding on the PFC than receptivenarrative tasks. Both expressive discoursetasks were contrasted against the receptivenarrative reading task. Eight of 13 participants(62%) showed significantly greater corticalactivation in the PFC during expressive tasks.In the Expressive: NYC versus receptivenarrative contrast, two demonstrated pre-dominantly right-sided significant activation,three demonstrated predominantly left-sidedsignificant activation, and three demonstratedbilateral activation patterns (Figure 3). Inthe Expressive: Runaway versus receptivenarrative contrast, eight participants demon-strated greater activation during expressivetasks than during receptive tasks and five hadno significant difference: three participantsdemonstrated predominantly right-sided



Table 2. Participants with significant activation by condition

Participant

ReceptiveNarrative −

Sentence

ExpressiveDiscourse (NYC)

− Receptive(Narrative)

ExpressiveDiscourse (NYC)

− ExpressiveNarrative

(Runaway)

ExpressiveNarrative

(Runaway) −Receptive

(Narrative)

1 L R2 B3 B B4 B B L5 L L B L6 L7 R R R R8 R9 L B L10 B11 L B12 B B13 L R RTotal N (%) 3 (23%) 8 (62%) 10 (77%) 8 (62%)

Note. Filled cells indicate participants demonstrating significant cortical activation during the indicated task compar-isons. The order of the comparisons is indicated in the column header. For example, Narrative-Sentences is the relativecortical activation (i.e.,% HbO2 change from baseline) for the narrative comprehension condition minus the relativecortical activation for the sentence condition. B = bilateral activation pattern; Hb = hemoglobin; L = left-predominantactivation pattern; NYC = New York City; R = right-predominant activation pattern.

significant activation, four demonstrated pre-dominantly left-sided significant activation,and one demonstrated bilateral activationpatterns (Figure 4).

Subsequent analysis of the HbO2 data wasexplored to characterize the relationship be-tween discourse efficiency measures (derivedfrom micro- and macrostructure data) and pro-cessing load on the PFC. Bivariate correla-tions between the β weights (e.g., standard-ized measure of the strength of response)of task-relative HbO2 reactions were corre-lated with discourse efficiency percentages(i.e., story grammar components/T-unit forthe Rockwell narrative and core discourseelements/T-unit for the NYC discourse task).Because of the exploratory nature of this anal-ysis, a conservative significance value of 0.006was adopted on the basis of the standard Bon-ferroni correction (i.e., 16 channels of data ×0.10). The bivariate fit of these variables sug-gests that discourse production for the NYC

task was not significantly correlated to sto-rytelling efficiency (Table 3). The data sug-gest that in the narrative (Runaway) task, leftlateralized activation is significantly positivelycorrelated to storytelling efficiency. That is,good storytelling efficiency on the Runawaytask, in contrast to the nonnarrative NYCtask, was related to higher processing de-mands in left PFC, as seen in left upper lateralchannels and left lower medial channels (seeTable 3).

Hypothesis 3 was that varying expressivediscourse tasks will induce differing profilesof cortical oxygen consumption based onfluctuating task demands, as is seen in behav-ioral data (Coelho, 2002; Van Leer & Turkstra,1999). The two expressive discourse pro-duction tasks were contrasted to investigatetask-related processing demands on the PFC.Ten of 13 participants (77%) demonstratedsignificant activation patterns in the storynarrative generation (Runaway) versus the



Figure 2. Example of varied activation patterns observed in the receptive narratives versus sentences con-trast. Note. Participant 5 with left lateralized significant activation, and Participant 7 with right lateralizedsignificant activation. Color map representation indicates level of significance in t-test statistic.

Figure 3. Example of varied activation patterns observed in the expressive discourse (NYC) versus recep-tive narrative contrasts (NYC–Receptive Narrative). Note. Participant 11 with predominantly left lateralizedsignificant activation; Participant 4 with significant bilateral activation; and Participant 7 with predomi-nantly right lateralized significant activation. Color map representation indicates level of significance int-test statistic. NYC = New York City.



Figure 4. Example of varied activation patterns observed in the expressive narrative (Runaway) versusreceptive narrative contrasts (Runaway–Receptive Narrative). Note. Participant 4 with predominantlyleft lateralized significant activation; Participant 10 with significant bilateral activation; and Participant7 with predominantly right lateralized significant activation. Color map representation indicates level ofsignificance in t-test statistic.

procedural discourse (NYC) versus contrast.Seven of these activation patterns werebilateral, one was predominantly left-sidedactivation and two were predominantly right-sided activation (see Table 2 and Figure 5).This pattern of predominantly medial and gen-erally strong and extensive activation was themost pronounced of any contrast condition.

DISCUSSION AND CLINICALIMPLICATIONS

The results of this study suggest a com-plex and varied picture of the involvement ofthe PFC during naturalistic discourse process-ing. Single-participant analysis and descrip-tive comparisons suggest that there are some

general trends (e.g., discourse production re-quires greater resource allocation to the PFCthan receptive discourse tasks); however, ef-fects across conditions and participants indi-cate that these demands are somewhat indi-vidualistic. Having the ability to carry out anecologically valid study of complex cognitive-communicative behavior has numerous ad-vantages over the limitations found in fMRI.However, a limitation of using fNIRS is thatthe present equipment could only monitor arelatively limited swath of cortical surface andtherefore could miss contributions of otherbrain structures known to be involved in dis-course processing. It is through this lens thatwe discuss the results and clinical implica-tions of the current data set.



Table 3. Bivariate correlations of story efficiency and cortical activation

Channel #Channel

LateralizationChannelLocation

Discourse(NYC)

Narrative(Runaway)

1 Left lateral Upper n.s. r = .87, p = .0052 Left lateral Lower n.s. n.s.3 Left lateral Upper n.s. r = .88, p = .0044 Left lateral Lower n.s. n.s.5 Left medial Upper n.s. n.s.6 Left medial Lower n.s. r = .90, p = .0027 Left medial Upper n.s. n.s.8 Left medial Lower n.s. r = .86, p = .0069 Right medial Upper n.s. n.s.10 Right medial Lower n.s. n.s.11 Right medial Upper n.s. n.s.12 Right medial Lower n.s. n.s.13 Right lateral Upper n.s. n.s.14 Right lateral Lower n.s. n.s.15 Right lateral Upper n.s. n.s.16 Right lateral Lower n.s. n.s.

Note. Significant correlations reported at the Bonferroni corrected value of 0.006 (16 channels × p = .1). n.s. = notsignificant; NYC = New York City.

In our first hypothesis, we predicted that inreceptive language tasks, cohesive discoursewould instantiate increased cortical activity inthe PFC compared with unrelated sentences.This hypothesis was based on a number offMRI studies outside of our laboratory thatsuggest narrative processing increased activ-ity over sentence-level processing. However,the current fNIRS data do not support thesefindings. In the present study, cohesive dis-course elicited increased activation of the PFCin only 3 of the 13 participants. Previous fMRIwork in our laboratory indicated that cohesivediscourse (e.g., simple narratives organizedaccording to story grammar principles) can ac-tually reduce processing demands of the PFCand other posterior language areas comparedwith unconnected sentences (Cannizzaroet al., 2012). Nevertheless, the comparisonsin this study indicate that there were no sig-nificant differences in PFC activation in themajority of participants (10/13) during read-ing of coherent discourse compared with un-related sentences. However, the similar ac-tivation patterns of the PFC for both condi-

tions suggest that the simplistic constructionof the cohesive discourse does not increaseactivity in the PFC, potentially due to the pre-dictable nature of the discourse constructions(Cannizzaro et al., 2012). Similarly, the uncon-nected sentences do not increase demand onthe PFC because there is no anticipation ofcoherence or discourse construction duringthat condition (e.g., Xu et al., 2005).

Our second hypothesis predicted that dis-course production would be significantlymore demanding on the PFC than receptivenarrative tasks, as indicated by increased corti-cal oxygen consumption. This hypothesis wassupported by the majority of participants andseen in both types of discourse generationtasks employed. For the NYC procedural dis-course, executive and planning skills are nec-essarily incorporated with prior knowledge oftravel, NYC, and environmental factors. Thisnaturalistic and spontaneous task was foundto place significant demand on the PFC in62% of individuals in this study. Participantswith significant activation in this comparisondemonstrated individually diverse activation



Figure 5. Example of activation patterns observed in the expressive discourse (NYC) versus expressivenarrative (Runaway) contrasts (NYC–Runaway). Note. Participants 9 and 12 with significant bilateralactivation, the dominant pattern for this contrast; and Participant 7 with predominantly right lateralizedsignificant activation. Color map representation indicates level of significance in t-test statistic. NYC =New York City.

patterns in the PFC, with relatively equal num-bers of right, left, and bilateral activities. Theproduction of the NYC procedural discoursedid not significantly relate to discourse effi-ciency either, further indicating that when thePFC is engaged, the weight of activation isnot a direct reflection of narrative productionquality. These findings are generally in sup-port of previous fMRI work that indicates in-creased processing demand in the PFC duringnarrative generation (AbdulSabur et al., 2014;Troiani et al., 2008). However, this finding isunique in that the narrative generation taskswere naturalistic, eliciting a novel responseon demand, similar to everyday communica-tion interactions.

Our third hypothesis predicted that differ-ent discourse production tasks would inducevarying profiles of cortical oxygen consump-tion based on fluctuating task demands, asthis pattern has been seen in behavioral data(Coelho, 2002; Van Leer & Turkstra, 1999).This hypothesis was well supported by thedata, as the majority of participants (10/13)demonstrated significantly greater activationprofiles when generating the NYC proceduraldiscourse as compared with the Runaway nar-rative. The observed task-related differencesin PFC activity indicate fluctuating cortical in-volvement based on discourse elicitation task.It is notable that the results from this contrastwere the strongest patterns of activation in



terms of consistency (e.g., bilateral-medial ac-tivation and number of participants with sig-nificant activation) and to the greatest extent(e.g., largest β weights and t-test scores). Thebehavioral data on narrative efficiency alsosupport this pattern, as participants demon-strated significantly less efficiency in produc-ing the NYC procedural discourse. As faras we are aware, this is the only investiga-tion directly contrasting clinical behavioralmeasures of discourse production and task-relevant PFC function. These findings supportprevious clinical literature suggesting a varietyof discourse elicitation procedures are neces-sary to adequately assess the varying demandsrelated to generating different discourse gen-res (Coelho et al., 2005).

This study had several limitations. The num-ber of participants was relatively small. Thespatial resolution of fNIRS is less than forfMRI (although temporal resolution is better),and given the person-to-person variability intask-related PFC activation, within-person taskcomparisons and neuroimaging to biobehav-ioral comparisons may be more valid thanperson-to-person comparisons. Finally, onlya small number of discourse behaviors weresampled. This study is a preliminary attempt ataddressing the lack of information on neuro-logical correlates of discourse task demands inthe adult population. There are as yet no pub-lished norms for discourse generation skillsin adults, with or without TBI. Understanding

how discourse structure influences discourse-processing abilities and the role of PFC inthat process can inform assessment and inter-vention practices. This is important becausedisrupted discourse and communication abil-ities following TBI are significantly debilitat-ing, reducing the quality of interpersonal re-lationships for sufferers of TBI and acting asa barrier to independent and productive em-ployment in this population (Coelho, 2005,2007; Snow et al., 1998; Ylvisaker et al., 2001,2007).

The current findings indicate that fluctuat-ing demands in discourse are observable byfunctional neuroimaging in common ecolog-ically valid measures of communication andthat task-dependent activity of the PFC ishighly individualistic across tasks. This vari-ability, demonstrated here in healthy youngparticipants, suggests that the PFC is involvedacross a number of discourse acts but is notnecessarily related to discourse proficiency inall tasks. Future investigations could comparethe neural architecture activated by differentorganizational structures in discourse (e.g.,personal narratives vs. story narratives vs. pro-cedural narratives) or a variety of organiza-tional patterns (e.g., chain narratives vs. heapnarratives vs. classic narratives). Monitoringwider areas of the cortex using fNIRS wouldalso provide a more complete picture ofthe neural demands of naturalistic discourseprocessing.

REFERENCES

AbdulSabur, N. Y., Xu, Y., Liu, S., Chow, H. M., Bax-ter, M., Carson, J., et al. (2014). Neural correlatesand network connectivity underlying narrative pro-duction and comprehension: A combined fMRI andPET study. Cortex, 57, 107–127.

American Speech-Language-Hearing Association. (2005).Roles of speech–language pathologists in the identifi-cation, diagnosis, and treatment of individuals withcognitive-communication disorders: Position state-ment. Rockville, MD: Author.

Ash, S., Moore, P., Antani, S., McCawley, G., Work, M., &Grossman, M. (2006). Trying to tell a tale: Discourseimpairments in progressive aphasia and frontotempo-ral dementia. Neurology, 66(9), 1405–1413.

Ayaz, H., Izzetoglu, M., Platek, S. M., Bunce, S., Izzetoglu,K., Pourrezaei, K., et al. (2006). Registering fNIR datato brain surface image using MRI templates. Confer-ence Proceedings IEEE Engineering in Medicine andBiology Society, 1, 2671–2674.

Basso Moro, S. B., Cutini, S., Ursini, M. L., Ferrari, M.,& Quaresima, V. (2013). Prefrontal cortex activationduring story encoding/retrieval: A multi-channel func-tional near-infrared spectroscopy study. Frontiers inHuman Neuroscience, 7(925), 1–11.

Biddle, K. R., McCabe, A., & Bliss, L. S. (1996). Narra-tive skills following traumatic brain injury in childrenand adults. Journal of Communication Disorders,29, 447–469.



Bottini, G., Corcoran, R., Sterzi, R., Paulesu, E., Schenone,P., Scarpa, P., et al. (1994). The role of the right hemi-sphere in the interpretation of figurative aspects oflanguage: A positron emission tomography activationstudy. Brain, 117, 1241–1253.

Cannizzaro, M. S., Coelho, C. A., & Youse, K. (2002).Treatment of discourse deficits following TBI. Perspec-tives on Neurophysiology and Neurogenic Speechand Language Disorders, 12, 14–19.

Cannizzaro, M. S., Dumas, J., Prelock, P., & Newhouse, P.(2012). Organizational structure reduces processingload in the prefrontal cortex during discourse process-ing of written text: Implications for high-level readingissues after TBI. SIG 2 Perspectives on Neurophysi-ology and Neurogenic Speech and Language Disor-ders, 22(2), 67–78.

Chapman, S. B., McKinnon, L., Levin, H. S., Song, J., Meier,M. C., & Chiu, S. (2001). Longitudinal outcome of ver-bal discourse in children with traumatic brain injury:Three-year follow-up. The Journal of Head TraumaRehabilitation, 16, 441–455.

Chatrian, G. E., Lettich, E., & Nelson, P. L. (1985). Ten per-cent electrode system for topographic studies of spon-taneous and evoked EEG activities. American Journalof EEG Technology, 25(2), 83–92.

Coelho, C. (2005). Direct attention training as a treatmentfor reading impairment in mild aphasia. Aphasiology,19(3–5), 275–283.

Coelho, C. (2007). Management of discourse deficits fol-lowing traumatic brain injury: Progress, caveats, andneeds. Seminars in Speech and Language, 28(2),122–135.

Coelho, C. A. (2002). Story narratives of adults withclosed head injury and non-brain-injured adults: In-fluence of socioeconomic status, elicitation task, andexecutive functioning. Journal of Speech, Language,and Hearing Research, 45, 1232–1248.

Coelho, C. A., Ylvisaker, M., & Turkstra, L. S. (2005).Nonstandardized assessment approaches for individu-als with traumatic brain injuries. Seminars in Speechand Language, 26, 223–241.

Fleming, V. B., & Harris, J. L. (2008). Complex dis-course production in mild cognitive impairment:Detecting subtle changes. Aphasiology, 22(7–8),729–740.

Ferstl, E. C., Neumann, J., Bogler, C., & von Cramon, D.Y. (2008). The extended language network: A meta-analysis of neuroimaging studies on text comprehen-sion. Human Brain Mapping, 29(5), 581–593.

Ferstl, E. C., & von Cramon, D. Y. (2001). The role ofcoherence and cohesion in text comprehension: Anevent-related fMRI study. Brain Research. CognitiveBrain Research, 11, 325–340.

Ferstl, E. C., & von Cramon, D. Y. (2002). What does thefronto-medial cortex contribute to language process-ing: Coherence or theory of mind? Neuroimage, 17,1599–1612.

fNIR Devices LLC. (2013). fNIR imager & COBI Studiomanual. Potomac, MD: Author.

Galetto, V., Andreetta, S., Zettin, M., & Marini, A. (2013).Patterns of impairment of narrative language in mildtraumatic brain injury. Journal of Neurolinguistics,26(6), 649–661.

Hickmann, M., & Schneider, P. (2000). Coherence andcohesion anomalies and their effects on children’s ref-erent introduction in narrative retell. In M. Perkins, &S. Howard (Eds.), New directions in language devel-opment and disorders (pp. 251–260). New York, NY:Kluwer Academic/Plenum Publishers.

Homan, R. W., Herman, J., & Purdy, P. (1987). Cerebrallocation of international 10-20 system electrode place-ment. Electroencephalography and Clinical Neuro-physiology, 66(4), 376–382.

Hughes, D., McGillivray, L., & Schmidek, M. (1997).Guide to narrative language. Eau Claire, WI: Think-ing Publications.

Hunt, K. (1970). Syntactic maturity in school childrenand adults. Monographs of the Society for Researchin Child Development, 35(Serial No. 134).

Irani, F., Platek, S. M., Bunce, S., Ruocco, A. C., & Chute,D. (2007). Functional near infrared spectroscopy(fNIRS): An emerging neuroimaging technology withimportant applications for the study of brain disor-ders. The Clinical Neuropsychologist, 21(1), 9–37.

Jang, K. E., Tak, S., Jung, J., Jang, J., Jeong, Y, & Ye, J.C. (2009). Wavelet minimum description length de-trending for near-infrared spectroscopy. Journal ofBiomedical Optics, 14, 034004.

Kiran, S., Harris, J. L., Marquardt, T. P., & Palmieri,M. (2005). Communication Wellness Checkups: Age-related changes in communication. Miami, FL: ASHANational Convention.

Marini, A., Zettin, M., & Galetto, V. (2014). Cognitive cor-relates of narrative impairment in moderate traumaticbrain injury. Neuropsychologia, 64, 282–288.

Mason, R. A., & Just, M. A. (2006). Neuroimaging contri-butions to the understanding of discourse processes.In Handbook of psycholinguistics (p. 799). Amster-dam: Elsevier.

Mason, R. A., & Just, M. A. (2013). Identifying componentdiscourse processes from their fMRI time course sig-natures. In A. M. Britt, S. R. Goldman, & J.-F. Rouet(Eds.), Reading—From words to multiple texts (pp.147–162). New York, NY: Routledge.

Quaresima, V., Bisconti, S., & Ferrari, M. (2012). A briefreview on the use of functional near-infrared spec-troscopy (fNIRS) for language imaging studies in hu-man newborns and adults. Brain and Language,121(2), 79–89.

Schneider, P., & Winship, S. (2002) Adults’ judgments offictional story quality. Journal of Speech, Language,and Hearing Research, 45(2), 372–383.

Silbert, L. J., Honey, C. J., Simony, E., Poeppel, D., & Has-son, U. (2014). Coupled neural systems underlie theproduction and comprehension of naturalistic narra-tive speech. Proceedings of the National Academyof Sciences of the United States of America, 111(43),E4687–E4696.



Snow, P., Douglas, J., & Ponsford, J. (1998). Conversa-tional discourse abilities following severe traumaticbrain injury: A follow up study. Brain Injury, 12,911–935.

Stein, N. L., & Glenn, C. G. (1979). An analysis of storycomprehension in elementary school children. InR. O. Freedle (Ed.), New directions in discourse pro-cessing (pp. 53–120). Norwood, NJ: Ablex.

Tak, S. (2011). NIRS-SPM: Statistical parametric map-ping for near infrared spectroscopy user’s guide. Re-trieved from http://bispl.weebly.com/nirs-spm.html#

Troiani, V., Fernández-Seara, M. A., Wang, Z., Detre,J. A., Ash, S., & Grossman, M. (2008). Narrativespeech production: An fMRI study using continu-ous arterial spin labeling. Neuroimage, 40(2), 932–939.

Tucker, F. M., & Hanlon, R. E. (1998). Effects of mild trau-matic brain injury on narrative discourse production.Brain Injury, 12, 783–792.

Van Leer, E., & Turkstra, L. (1999). The effect of elicitationtask on discourse coherence and cohesion in adoles-

cents with brain injury. Journal of CommunicationDisorders, 32(5), 327–349.

Ye, J. C., Tak, S., Jang, K. E., Jung, J., & Jang, J.(2009). NIRS-SPM: Statistical parametric mapping fornear-infrared spectroscopy. Neuroimage, 44(2), 428–447.

Ylvisaker, M., Feeney, T., & Capo, M. (2007). Long-termcommunity supports for individuals with co-occurringdisabilities after traumatic brain injury: Cost effective-ness and project-based intervention. Brain Impair-ment, 8(03), 276–292.

Ylvisaker, M., Szekeres, S., & Feeney, T. (2001). Com-munication disorders associated with traumatic braininjury. In R. Chapey (Ed.), Language interven-tion strategies in aphasia and related neurogeniccommunication disorders (4th ed., pp. 745–807).Philadelphia, PA: Lippincott Williams & Wilkins.

Xu, J., Kemeny, S., Park, G., Frattali, C., & Braun, A.(2005). Language in context: Emergent features ofword, sentence, and narrative comprehension. Neu-roimage, 25(3), 1002–1015.


http://bispl.weebly.com/nirs-spm.html#

top lang disorders vol. 36, no. 1, pp. 65–79 prefrontal ... · the prefrontal cortex (pfc) during...

Documents