####################### PVTdemoAGEPIQ.py Description ######################### # PVTdemoAGEPIQ.py 10-09-06 replicating Psychomotor Vigilance Task (PVT) model # by Gunzelmann, G.Gluck, K., Van Dongen, H., & Dinges, D.F. in press (2006) # Decreased Arousal as a Result of Sleep Deprivation: The Unraveling of # Cognitive Control (chapter in Modeling Integrated Cognitive Systems) # reaction time model for letter X response X with trial interval 2-10 sec # experimental data collected 10 minutes every 2 hours for 88 hours # (missing latency feedback to the display) # simulating total sleep deprivation (TSD) for 0 1 2 3 days # PythonActor code by Terry Stewart and Bruce Landon #note: description from Loh,Laland,Dorrian,Roach and Dawson (2004) describing #their replication of Dinges and Powell (1985) which is not easily avialable. #PVT ran for a period of 10min. Subjects were required to respond to a visual #stimulus presented at a variable interval (2000-10000ms) by pressing a button #with the thumb on the dominant hand. The visual stimulus was a four-digit #LED counter turning on and incrementing from 0 to 60s at 1ms. Button press #stopped the LED incrementing allowing the subject 1s to read their reaction #time before the counter restarted. If a response was not made within 60s #the clock was reset and the counter restarted. If a response was made prior #to the presentaiton of the stimulus, a "False Start" message was displayed #If the button was not released after 3s a reminder message ("Release Button") #was displayed.(there is now a Palm verison of the PVT Saxena & George (2005) # # J Gerontol. 1994 Jul;49(4):P179-89. # Age differences and changes in reaction time: the Baltimore Longitudinal # Study of Aging by Fozard JL, Vercryssen M, Reynolds SL, Hancock PA, Quilter # RE. National Institutes of Health, National Institute on Aging, Nathan W. # Shock Laboratories of the Gerontology Research Center, Baltimore. # This study analyzed auditory reaction time (RT) data from 1,265 # community-dwelling volunteers (833 males and 432 females) who ranged in age # from 17 to 96. Cross-sectional analyses revealed slowing of simple (SRT) # and relatively greater slowing of disjunctive (DRT; aka "go-no-go") reaction # time across decades for both males and females. Repeated testing within # participants (longitudinal analyses) over eight years showed consistent # slowing and increased variability with age. Males were faster than females # cross age groups, RT tasks, and visits. Beginning at about age 20, RTs # increased at a rate of approximately 0.5 msec/yr for SRT and 1.6 msec/yr # for DRT. Errors also increased, making unlikely a tradeoff of accuracy # for faster responses. The findings are consistent with the hypotheses that # slowing of behavior is: (a) a continuous process over the adult life span; # (b) characterized by age-associated increases in within-participant # variability; (c) a direct function of task complexity and, presumably, the # degree of mediation by higher regions in the central nervous system; and # (d) greater in women than men. # # A unifieded computational theory for # answering 20 questions (and more) about cognitive ageing # David E. Meyer, Jennifer M. Glass, Shane T. Mueller, # Travis L. Seymour, and David E. Kieras University of Michigan, USA # EUROPEAN JOURNAL OF COGNITIVE PSYCHOLOGY, 2001, 13 (1/2), 123–164 # Quoted from p. 130 Cognitive-processor parameters. # The rate of processing in EPIC is constrained # by the cognitive processor’s mean cycle time. For young adults, # we assume that this parameter equals 50 ms. Our assumption is supported # by both behavioural and psychophysiological data (e.g., KristoVerson, # 1967). It conforms well with the mean period between zero crossings in # the brain’s alpha rhythm, which equals about 50 ms for young adults and # correlates positively with their mean simple RTs (Callaway & Yeager, # 1960; Surwillo, 1963; WoodruV, 1975). It appears that alpha-rhythm zero # crossings perhaps correspond to the onsets of cognitive-processor cycles. # In contrast, for older adults around 70 years of age, the mean zero # crossing period of the alpha rhythm is about 10–15% longer than for # young adults (Marsh & Thompson, 1977; Obrist & Busse, 1965; Surwillo, # 1963), and older adults’ mean simple RTs are also about 10–15% longer # (Cerella, 1985; Somberg & Salthouse, 1982). These results lead us to # make the following strong but at least somewhat plausible assumption: In # EPIC models for older adults (circa 70 years of age), the mean cognitive # processor cycle time should be set to 56.5 ms, about 13% longer than the # mean cycle time in models for young adults (circa 20 years of age). This # assumption is consistent with Kail and Salthouse (1994), who claimed # that ageing affects ‘‘a fundamental component of the architecture of # human cognition’’ which functions like the CPU clock of a microcomputer # whose ‘‘clock speed’’ determines the rate of information processing. # Our assumption also agrees with Madden (1992), who claimed that in # visual word-recognition tasks, mean RTs increase by roughly 4–10 ms per # decade of age. Under EPIC, such tasks usually require about 3–8 cognitive- # processor cycles to be completed. Moreover, ageing would involve # roughly a 1.3 ms increase per decade in the mean cycle time. The latter # factor, multiplied by 3–8 cycles, yields a range close to Madden’s 4–10ms # per decade. ... Furthermore, another key parameter that presumably # mediates RTs and age-related slowing under EPIC is the number of cycles # taken to complete a task, which depends on what production rules are used # for task performance (Kieras & Meyer, 2000). Insofar as older adults use # less effcient rules and so take more cognitive-processor cycles than young # adults, age-related RT diVerences may be larger than if only the mean # 130 MEYER ET AL. # # Alpha EEG predicts visual reaction time. # Jin, Yi, Department of Psychiatry and Human Behavior, University of # California Irvine College of Medicine, Irvine, CA, US, yjin@uci.edu # O'Halloran, James P.,Plon,Lawrence,Sandman, Curt A.,Potkin, Steven G., # International Journal of Neuroscience, Vol 116(9), 2006. pp. 1035-1044. # Abstract: Studies have suggested that consciousness is encoded # discretely in time and synchronously in space of the brain. The present # study was to model the alpha EEG as a brain clock to carry out the # functions and to test whether the quality and rate of the oscillation # could predict behavioral timing. Results showed that the alpha peak # frequency was correlated with the conflict reaction time, and the # selectivity was associated with the simple reaction time. These findings # are consistent with previous reports and support the hypothesis that # alpha EEG represents excitability cycles and may serves as a brain # clock for spatial synchronization. # ######################## Define task setup ################################## repetitions=100 # the number of stimuli to show in simulation run goal_threshold = 1.98 #enable ccm simulation 1.98 1.80 1.66 1.58 TSD 0 1 2 3 utility_threshold = 1.84 #enable simulation 1.84 1.78 1.70 1.64 TSD 0 1 2 3 cycle_noise = 0.0125 # variability in fundamental cycle time of 50ms adjust_utility_threshold = 0.035 # reduce utility threshold after idle cycle ######################## Define personal parameters ######################### Agefix = 0.0 # ((age -20)*0.00013)lengthen fundamental cycle for aging # (IQz * -0.31 = RTz) assume like IQ decline with aging -1 sd = 30yr older # from Salthouse (1992) Anderson text 2004 p.437 (PIQ85 at 20=50 with PIQ100) PIQfix = -0.0 # PIQ100=0 PIQ085=+0.0039 PIQ115=-0.0039, PIQ130=-0.0078 ####################### Editable parameters above ########################### from ccm.env.objects import *#http://www.carleton.ca/ics/ccmlab/ccmsuite.html import random import math class MatchEnvironment(ObjectEnvironment): def start(self): # define internal variables self.count=0 self.falseAlarm=0 self.lapse=0 self.sleepAttack=0 self.letter=Object(isa='letter',x=0.5,y=0.5,value=None) # initialize self.numrt=0 # for calculating okrtaverage self.sumrt=0 # for calculating okrtaverage self.rtsumX=0 # for calculating self.okrtsd self.rtsumXsqr=0 # for calculating self.okrtsd self.rtnum=0 # for calculating self.okrtsd yield random.uniform(2,10) # random wait 2-10 seconds self.letter.value='X' self.target=self.letter.value self.cueTime=self.now() # get the current time @ccm.parallelize def press(self,key): rt=self.now()-self.cueTime # calculate reaction time log.rt=rt # record the reaction time to the log if rt<0.150 or self.letter.value==None: # response before stimulus self.falseAlarm+=1 elif rt>30: self.sleepAttack+=1 elif rt>0.5: self.lapse+=1 else: self.numrt+=1 # ok rt between >0.15 (falseAlarm) <0.5 (lapse) self.sumrt+=rt self.rtsumX+=rt self.rtsumXsqr+=(rt * rt) self.rtnum+=1 self.letter.value=None self.count+=1 if self.count==repetitions: # report results in log # convert to proportions scale=1.0/repetitions log.falseAlarm=self.falseAlarm*scale log.lapse=self.lapse*scale log.sleepAttack=self.sleepAttack*scale log.oktraverage=self.sumrt/self.numrt # over 150ms under 500ms log.self.okrtsd= math.sqrt((self.rtnum * self.rtsumXsqr -self.rtsumX * self.rtsumX)/ (self.rtnum * (self.rtnum-1))) print 'Age20fix=' + str(Agefix) # via production cycle time print 'PIQ100fix=' + str(PIQfix) # via production cycle time print 'alphaEEG=' + str((0.05 + Agefix + PIQfix)*200) #normal self.stop() else: self.model.pgc.goal=goal_threshold # reset goal at start of trial yield random.uniform(2,10) # random inter-trial interval 2-10s self.letter.value='X' # the cue stimuls (for rt counter digits) self.target=self.letter.value self.cueTime=self.now() # start trial reaction timer from ccm.lib.actr import * # http://www.carleton.ca/ics/ccmlab/ccmsuite.html class PMpgcGoalReducer(PMpgc): # enables falling asleep or micro sleeps def below_threshold(self): if env.letter.value==None: return # don't reduce G if no stimuli self.goal-=adjust_utility_threshold # reduce threshold after idle cycle if self.goal < 0 : # self.goal is G which is motivational goal value self.goal=0.0001 #keep G positive even during sleep attack ################# Define the ACT-R Model Components ########################## class MyModel: # minimal model (with no memory,mood,feeling,pain or alertness) goal=Buffer() visual=Buffer() env=DirectEnvironment() vision=SOSVision(visual,delay=0.085, delay_sd=0.01) procedural=Procedural(threshold=utility_threshold, delay=(0.05 + Agefix + PIQfix),delay_sd=cycle_noise) pgc=PMpgcGoalReducer(procedural,goal=goal_threshold) #sleepless adjust G pnoise=PMNoise(procedural,0.25) # add in the noise part of the utility ################# Define ACT-R Model Production Memories ##################### def attend(goal='attend',vision='busy:False'): # looking for stimulus X vision.request('isa:letter') goal.set('attend') def respond(goal='attend',visual='isa:letter value:?letter'): #see X env.press(letter) # model presses key X to stop the reaction time clock visual.clear() def just_click(goal='attend',vision='busy:False'): # enables false alarms env.press('X') # press X before stimulus appears due to model noise visual.clear() ####################### Define Environment setup ############################ log=ccm.log() # enable log to print out results env=MatchEnvironment() # don't automatically try to log things env.model=ACTR(MyModel,logging=False) #turn off logging internal cycle states env.model.goal.set('attend') # set starting goal buffer # needed to set the P value for 'just_click' to zero env.model.pgc.set('just_click',successes=0,failures=1,lock=True) # turn on visual display or just run env.run() #import ccm.display #ccm.display.Display(env) #env.run(rate=1) # to run at real-time env.run() # run simulation with model