Memory and Afetamine

ORIGINAL INVESTIGATION

A triazolam/amphetamine dose–effect interaction study:dissociation of effects on memory versus arousal

Miriam Z. Mintzer & Roland R. Griffiths

Received: 5 December 2006 /Accepted: 24 January 2007 / Published online: 7 March 2007# Springer-Verlag 2007

AbstractRationale In addition to producing robust memory impair-ment, benzodiazepines also induce marked sedation. Thus,it is possible that the observed amnestic effects aresecondary to more global sedative effects and do not reflecta specific primary benzodiazepine effect on memorymechanisms.Objective The objective was to use the nonspecific stimu-lant d-amphetamine to dissociate the sedative and memory-impairing effects of the benzodiazepine triazolam.Materials and methods Single oral doses of placebo,triazolam alone (0.25, 0.50 mg/70 kg), d-amphetaminesulfate alone (20, 30 mg/70 kg), and triazolam (0.25,0.50 mg/70 kg) and d-amphetamine sulfate (20, 30 mg/70 kg) conjointly (at all dose combinations) were admin-istered to 18 healthy adult participants across nine sessionsin a double-blind, staggered-dosing, crossover design. Inaddition to standard data analyses, analyses were alsoconducted on z-score standardized data, enabling effects tobe directly compared across measures.Results Relative to the sedative measures, the memorymeasures generally exhibited a pattern of less reversal oftriazolam’s effects by d-amphetamine. The memory mea-sures ranged in degree of reversal such that the most

reversal was observed for reaction time on the n-backworking memory task, and the least reversal was observedfor accuracy on the Sternberg working memory task, withmost measures showing an overall pattern of partialreversal.Conclusions Benzodiazepines have specific effects onmemory that are not merely a by-product of the drugs’sedative effects, and the degree to which sedative effectscontribute to the amnestic effects varies as a function of theparticular memory process being assessed.

Keywords Triazolam . Amphetamine .Memory . Arousal .

Sedative

It is well documented that benzodiazepines (e.g., diazepam,Valium®; lorazepam, Ativan®; and triazolam, Halcion®)induce temporary amnesia when administered acutely tohealthy volunteers (for reviews, see Curran 1991, 2000;Ghoneim 2004a,b). Like neuropsychological studies ofbrain-damaged patients, which have played a critical rolein advancing the understanding of normal and abnormalmemory mechanisms, the investigation of benzodiazepine-induced amnesia can also be a powerful tool for elucidatingmemory mechanisms. However, in addition to amnesia,benzodiazepines also induce marked sedation as reflected inchanges in observer and subjective ratings of arousal andimpaired psychomotor performance (for reviews, seeHollister et al. 1993; Woods et al. 1992). Thus, in orderfor benzodiazepines to be used to advance theoreticalunderstanding of memory mechanisms, it is important tobe able to separate the amnestic effects from the sedativeeffects.

A variety of strategies can be adopted to differentiate theamnestic effects of benzodiazepines (or other sedativedrugs) from their sedative effects (cf. Curran 1991, 2000;

Psychopharmacology (2007) 192:425–440DOI 10.1007/s00213-007-0726-y

DO00726; No of Pages

M. Z. Mintzer (*) : R. R. GriffithsDepartment of Psychiatry and Behavioral Sciences,Behavioral Biology Research Center,Johns Hopkins University School of Medicine,5510 Nathan Shock Drive,Baltimore, MD 21224, USAe-mail: [email protected]

R. R. GriffithsDepartment of Neuroscience,Johns Hopkins University School of Medicine,Baltimore, MD 21224, USA

Ghoneim 2004b). First, the effects can be dissociatedstatistically by covarying measures of sedation frommeasures of memory; results suggest that drug effects onmemory typically are somewhat reduced but remainsignificant once effects on sedation are partialled out (cf.Curran 1991). Second, it can be demonstrated that differentdrugs produce different patterns of effects on sedation vsmemory. For example, it has been shown that at doses oflorazepam and the neuroleptic chlorpromazine (Green et al.1996) or the antihistamine diphenhydramine (Curran et al.1998) that produced similar effects on measures ofsedation, only lorazepam affected measures of memory.Likewise, using relative potency analyses, Kirk et al. (1990)demonstrated that triazolam showed relatively greaterpotency than pentobarbital on measures of memory thanon measures of sedation. Using pharmacokinetic–pharma-codynamic (PK–PD) modeling techniques, Veselis et al.(1997) showed that sedative drugs (midazolam [a benzodi-azepine], propofol, thiopental, and fentanyl) varied in theirrelative effects on sedation vs memory. Third, it can bedemonstrated that tolerance over the course of repeatedbenzodiazepine administration develops differentially toeffects on sedation vs memory. The general conclusionfrom these studies is that tolerance develops to the sedativeeffects before the amnestic effects, and that completetolerance may not develop to the amnestic effects evenafter years of chronic use (Curran et al. 1994; Ghoneim etal. 1981; Tata et al. 1994). Fourth, it can be demonstratedthat the dose–response and/or timecourse functions of theeffects on sedation vs memory have different characteristics(e.g., Blin et al. 2001; Smirne et al. 1989; Weingartner et al.1995). Using PK–PD modeling, Blin et al. (2001)dissociated the sedative and amnestic effects of lorazepambased on the median effective concentration (EC50) neededfor each effect. Fifth, sedative and amnestic effects can bedissociated through electrophysiological or neuroimagingtechniques. Curran et al. (1998) and Veselis et al. (2001)both provided evidence that the sedative and amnesticeffects of benzodiazepines are associated with differentevent-related potential components. To our knowledge, nostudies have attempted to differentiate the sedative andamnestic effects of a benzodiazepine using neuroimaging.Finally, it can be demonstrated that another compoundselectively reverses benzodiazepines’ effects on one func-tion but not the other. Several studies have attempted todifferentiate effects of benzodiazepines using the benzodi-azepine-receptor specific antagonist flumazenil (Curran andBirch 1991; Hommer et al. 1993; Preston et al. 1989).Results of these studies are mixed, and interpretation iscomplicated by the fact that flumazenil acts at benzodiaze-pine receptors, which are presumed to at least partiallymediate both the sedative and amnestic effects of benzodiaz-epines, and by evidence that flumazenil produces its own

weak agonist effects including significant memory impair-ment (e.g., Bishop and Curran 1995; Neave et al. 2000).

In the present study, we used this final strategy todissociate the sedative and amnestic effects of the benzo-diazepine hypnotic triazolam by administering the nonspe-cific stimulant dextroamphetamine (d-amphetamine).d-Amphetamine acts as an indirect catecholaminergicagonist by facilitating the action of dopamine and norepi-nephrine (for a review, see Solanto 1998). As reviewedabove, the results of previous studies provide evidence thatthe amnestic and sedative effects of benzodiazepines can bedissociated. However, they do not allow conclusions to bedrawn about possible differences among memory processesin the relative contribution of sedative effects to benzodi-azepine-induced memory impairment. In the present study,multiple memory tasks were included to investigate differ-ences among memory processes (working memory, episod-ic memory, and metamemory) in the degree to whichsedative effects contribute to amnestic effects. Results of aprevious interaction study in our laboratory with d-amphetamine and triazolam at single-dose levels indicatedthat d-amphetamine significantly reversed triazolam’seffects on all measures of sedative effects but failed toreverse TRZ-induced impairment on select measures ofmemory (Mintzer and Griffiths 2003). These results suggestthat benzodiazepines do have specific effects on memorythat are not merely a by-product of the drugs’ sedativeeffects, and that the degree to which sedative effectscontribute to the amnestic effects may vary as a function ofthe particular memory process being assessed. The presentstudy followed up on our previous single-dose study, using adose–effect design to dissociate the sedative and amnesticeffects of triazolam and to test whether there are reliabledifferences between memory processes in the degree towhich sedative effects contribute to the amnestic effects;doses of both triazolam and d-amphetamine were manipu-lated, enabling the relationship between the amnestic andsedative effects of triazolam to be examined at multiplelevels of each effect for each memory process.

Materials and methods

Subjects

Eighteen adult volunteers (seven women) completed thisstudy. They ranged in age from 18 to 39 years (mean=23)and in weight from 55 to 103 kg (mean=74) and reportedhaving completed 12–21 years of education (mean=15).Fourteen participants reported consuming caffeinated bev-erages delivering 9–118 mg of caffeine/day (mean=55),whereas the other three did not report regular consumptionof caffeinated beverages. Twelve participants reported

426 Psychopharmacology (2007) 192:425–440

drinking alcohol socially. One participant reported smokingtobacco cigarettes (seven cigarettes/day). No participantsreported histories of drug or alcohol abuse, and all had anegative drug urine screen and breathalyzer test during theinitial screening interview. All participants were in goodhealth (as determined by medical history and personalinterview) with no contraindications to sedative or stimu-lant drugs. Individuals with current or past histories ofpsychiatric disorders were excluded. In the female partic-ipants, urine pregnancy tests conducted before the firstsession were negative. This study was approved by theInstitutional Review Board of the Johns Hopkins BayviewMedical Center. Participants gave their written informedconsent before beginning the study and were paid for theirparticipation. Participants were requested to refrain fromusing all psychoactive drugs (with the exception of tobaccoand caffeinated products) during the time they wereparticipants in the study. In each session, before drugadministration, participants were tested for the presence ofvarious drugs in their urine (benzodiazepines, opioids,methadone, and cocaine) using an EMIT system (Syva,Palo Alto, CA) and the presence of alcohol in expired airusing a breathalyzer test. Results of the urinalysis andbreathalyzer tests were negative for all participants.

General procedures

Participants completed a total of nine sessions as outpatientsat the Behavioral Pharmacology Research Unit. Participantswere informed that during their participation in the study,they would receive various drugs, and that these couldinclude placebo, various sedatives, anxiolytics, stimulantsand weight loss medications. Other than receiving thisgeneral information, participants were blind to the type ofdrug administered. Experimental measures (see descriptionsbelow) were administered before drug administration andrepeatedly after drug administration (except the episodicmemory study and test phases, which were administered atsingle timepoints after drug administration). The physiolog-ical measures and noncomputerized psychomotor measures(i.e., circular lights, balance) were administered approxi-mately 55, 75, 110, 175, 235, and 285 min (285 min, circularlights and balance only) after the first capsule administrationtimepoint. The participant ratings and DSST were adminis-tered approximately 60, 95, 135, 195, and 265 min after thefirst capsule administration timepoint. The n-back workingmemory task was administered approximately 140, 200, and270 min after the first capsule administration timepoint, andthe Sternberg working memory task was administeredapproximately 80, 120, 180, and 250 min after the firstcapsule administration timepoint. All computerized mea-sures were administered on an Apple Macintosh microcom-puter (Apple, Cupertino, CA).

Drug administration

Single doses of placebo (PL), triazolam alone (TRZ; 0.25,0.50 mg/70 kg), d-amphetamine sulfate alone (AMP; 20,30 mg/70 kg), and TRZ (0.25, 0.50 mg/70 kg) and AMP(20, 30 mg/70 kg) conjointly (at all dose combinations) wereadministered across nine sessions in a double-blind, within-subject, crossover design. All drug conditions were admin-istered orally in capsules. To synchronize the peak effects oftriazolam and d-amphetamine based on previous pharmaco-kinetic and behavioral data, a staggered dosing regimen wasused in which d-amphetamine was administered 30 minbefore triazolam. To maintain the double-blind design,during each session, the participant swallowed capsules attwo separate timepoints 30 min apart; at each timepoint, theparticipant received a total of three capsules. Capsules weretaken orally with approximately 150 ml of water. The orderof drug conditions was determined by two Latin squaresusing the Williams (1949) method to achieve balance in thepresentation order and in the order of drug conditionsrelative to one another. Triazolam (Halcion; Pharmacia andUpJohn, Kalamazoo, MI) and d-amphetamine (DextroStat;Shire US, Newport, KY) doses were prepared fromcommercially available 0.25-and 10-mg tablets, respectively.Tablets were crushed, and doses were adjusted by theparticipant’s body weight. All doses were dispensed in size0 capsules. Lactose was used to fill the remainder of thecapsules. PL capsules contained only lactose.

Experimental measures

Memory tasks

Episodic memory/metamemory After an initial study phase,episodic memory (conscious long-term memory for apersonally experienced event that is associated with aspecific spatial and temporal context; Tulving 1972, 1983)was tested via recognition and free recall tests. Stimuli foreach session consisted of a unique set (i.e., no stimulus wasrepeated across sessions) of 140 common concrete nounsselected from the Thorndike and Lorge (1944) word corpus(i.e., a total of 1,260 words across nine sessions); the setswere equated across the nine sessions for mean wordfrequency in the language (Thorndike and Lorge 1944) andword length. The 140-word stimulus set assigned to eachsession was randomly divided into two 70-word subsetswith the constraint that the subsets were equated for meanword frequency in the language and word length. In eachsubset, 35 words represented artificial (i.e., man made)objects, and 35 words represented natural objects. Onesubset was assigned to the old condition, and one wasassigned to the new condition (see below); the subsets

Psychopharmacology (2007) 192:425–440 427

assigned to the two conditions were counterbalanced suchthat, across participants, each subset appeared equally oftenin the old and new condition.

During the study phase, which was conducted 120 minafter the first capsule administration timepoint (to coincidewith the anticipated time of peak effects for both triazolamand d-amphetamine), participants were presented with a setof 70 words (those assigned to the old condition) thatappeared on the computer screen one at a time. Each wordappeared on the screen for 2 s. To ensure that participantswould attend to and encode the words presented during thestudy phase, participants were asked to perform a concep-tual categorization task on each word (to categorize eachword as representing an “artificial” or “natural” object; e.g.,Mintzer et al. 2001) and to make their responses by usingthe computer mouse to click on the appropriately labeledbutton on the screen.

Two hours after completing the study phase, participantswere administered recognition memory and free recall tests.The order of the two tasks was counterbalanced acrossparticipants. In the recognition memory test, participantswere presented with a set of 140 words that appeared on thescreen one at a time in random order; the set included the70 words that had been presented during the study phase(old) randomly mixed with the 70 words that had not beenpresented during the study phase (new). Participants wereinstructed to make judgments about the degree to whichthey recognized (old) or did not recognize (new) the wordfrom the study phase using a six-point confidence scale(definitely old, probably old, maybe old, maybe new,probably new, and definitely new) and to make theirresponse by clicking on the appropriately labeled buttonusing the computer mouse. Participants were encouraged todistribute their responses across the entire range from 1through 6. Each test word remained on the screen until theparticipant responded. In addition to measuring episodicmemory, this version of the recognition memory test alsoprovides a measure of the participants’ metamemory(awareness and knowledge of their own memory; Flavell1971; Metcalfe and Shimamura 1994) by comparingconfidence ratings given to correct vs incorrect recognitionresponses; presumably, if you are aware of the state of yourown memory, you will be more confident in your correct vsincorrect responses. In the free recall test, participants weregiven 5 min to write down all the words they rememberedseeing during the initial study phase.

Performance on the conceptual categorization task (studyphase) was assessed by analyzing the proportion of correctcategorization responses. Two sets of analyses were con-ducted on the data from the recognition memory test: one inwhich data were collapsed across confidence ratings (i.e.,definitely old, probably old, and maybe old were counted asold responses; definitely new, probably new, and maybe

new were counted as new responses) to provide measures ofepisodic memory, and one in which the confidence ratingsthemselves were analyzed to provide a measure of meta-memory. Analyses on the collapsed data included thefollowing measures: proportion of old responses made toold words (hit rate), proportion of old responses to newwords (false alarm rate), and signal detection measures ofsensitivity in distinguishing between old and new words (d’)and response bias (C; Snodgrass and Corwin 1988). Forboth old and new responses, confidence ratings were codedsuch that maybe=1, probably=2, and definitely=3. Meta-memory was measured by calculating the Goodman–Kruskal gamma correlation (between confidence ratingsand recognition memory accuracy computed for eachparticipant; Goodman and Kruskal 1954). The Goodman–Kruskal gamma correlation has been used by otherinvestigators to measure metamemory (cf. Nelson 1984).Values for gamma can range from 1 (complete concordancebetween confidence ratings and recognition memory accu-racy as reflected in high confidence ratings given to correctresponses/low confidence ratings given to incorrectresponses) to −1 (complete discordance between confi-dence ratings and recognition memory accuracy). To ensurethe availability of the data from a sufficient number ofitems per participant to enable the performance of theseanalyses, data were collapsed across old and new words.Dependent measures for the free recall test were number ofcorrect responses (i.e., number of old words; out of a totalof 70 possible) and number of (incorrect) intrusions (i.e.,number of “nonold” words) provided by participants.

Working memory Working memory (a type of short-termmemory that enables the temporary maintenance and on-line manipulation of information in the service of behav-ioral goals; Baddeley 1992) was assessed via the n-backtask and a modified Sternberg task.

The n-back task (Jonides et al. 1997) assessed theparticipant’s ability to recall letters presented n back (0, 1,2, and 3 back) in a continuous string of letters. For each nback, 60 consonant letters (excluding L, W, and Y) werepresented consecutively on the screen for 0.5 s each(interstimulus interval of 3.0 s), and participants wereinstructed to click the mouse on ‘yes’ whenever the currentletter on the screen matched (target) the letter ‘n’ positionsback in the sequence and to click on ‘no’ when there wasno match (nontarget). For example, in the 2-back task,participants would respond ‘yes’ to the final M in thesequence mFM; to prevent simple perceptual matching,upper and lower case letters were randomly intermixed.Memory load increases as a function of n. The 0-back is acontrol condition that involves minimal memory andprovides a measure of focused attention only; participants


were told to click ‘yes’ whenever the letter on the screenmatched a predetermined target letter and to click ‘no’whenever there was no match. The probability of a targetwas approximately 33% in each condition. The order of thefour n-back conditions (0, 1, 2, and 3 back) was randomlydetermined for each participant but was consistent acrosssessions within a participant. Dependent measures are theproportion of yes responses made to target letters (hit rate),proportion of yes responses made to nontarget letters (falsealarm rate), signal detection measures of sensitivity indistinguishing between target and nontarget letters (d’) andresponse bias (C), and median reaction time (RT) on correcttrials. In addition, before the start of each n back,participants were asked to rate how well they thought theywould perform on the task compared to normal (preperfor-mance estimate). Likewise, after each n back, they wereasked to rate how well they thought they had performedcompared to normal (postperformance estimate). Partici-pants made these ratings by clicking along a line on thescreen labeled much worse at the left extreme and muchbetter at the right extreme, with normal labeled in themiddle. Scores range from −50 (much worse) to 50 (muchbetter). These ratings measured the participants’ cognitiveawareness of their performance. The participant estimatesof performance were compared to the actual task scores bycalculating difference scores as the percentage of thepredrug estimate minus the percentage of the predrugactual task score. Thus, a positive score represents anunderestimation of performance impairment.

The second working memory task was a modifiedversion of the classic Sternberg task (Sternberg 1969) andused procedures similar to those described by Postle et al.(1999) and D’Esposito et al. (1999). A memory setconsisting of seven randomly selected and randomlyordered consonant letters (excluding L and Y; e.g.,PCZMSTF) was presented on the screen followed by aprobe consisting of a letter-digit pair (e.g., c-2), andparticipants were asked to decide whether the probed letterhad appeared in the memory set in the ordinal positionrepresented by the digit (e.g., 2=second position in memoryset). The probed letter was always a letter that had appearedin the memory set. To prevent simple perceptual matchingto the memory set stimuli (which were presented in uppercase), the probe was always presented in lower case. Effectson rehearsal processes were tested by varying the delaybetween memory set presentation and testing (i.e., probepresentation; 0, 12 s). There were 12 trials in each of thetwo delay conditions and 12 trials in a control conditionin which the memory set remained on the screen duringprobe presentation (i.e., 36 trials total); trials in allconditions were randomly intermixed. The control condi-tion was designed to control for drug effects onnonmemory processes (e.g., motor coordination, percep-

tion, attention, and other cognitive processes). Withineach condition, the probed digit represented the correctposition of the probed letter on half of the trials. Thememory set appeared on the screen for 4 s followed bythe predetermined delay and presentation of the probe fora period of 7 s during which the participant respondedby using the computer mouse to click on a buttonlabeled ‘yes’ or ‘no’. RT from onset of the probe wasrecorded. After each response, participants were requiredto return the mouse cursor to a screen position that wasequidistant between the two response buttons; thisposition was indicated by a circle on the screen thatwas illuminated when the cursor touched it. There was a2-s intertrial interval. Dependent measures are the propor-tion of correct responses and median RT on correct trials.

Psychomotor performance tasks

Psychomotor performance was measured via circular lights(Griffiths et al. 1983), a standing balance task, and acomputerized version of the digit symbol substitution test(DSST; McLeod et al. 1982). Circular lights involved rapidhand-eye coordinated movements in which the participantpressed a series of 16 buttons (circularly arranged around a54-cm diameter) as rapidly as possible in response to therandomly sequenced illumination of their associated lights.The score was the number of correct button presses duringa 60-s trial. The balance task assessed the participant’sability to stand upright on one foot with his/her eyes closedand arms extended to the side at shoulder height (for amaximum of 30 s on each foot). For the DSST, in responseto randomly selected digits (1–9) appearing on the screen,participants pressed button positions on a numeric keypadto reproduce the geometric symbol pattern associated withthat digit by using the digit-symbol code displayedcontinuously at the top of the screen. The scores were thenumber attempted and the proportion correct during a 90-strial. In addition, awareness of performance was assessedbefore (preperformance estimate) and after (postperform-ance estimate) the DSST with the same two questions asdescribed for the n-back task.

Participant Ratings

Participants rated the strength of drug effect on a five-pointscale (coded numerically from 0 to 4), their disliking/likingof the drug effect on a nine-point scale (coded numericallyfrom −4 [maximum disliking] to +4 [maximum liking],with 0 representing neutral), 34 items about their physicaland mental state (see Table 1 for specific items) on a five-point scale (coded numerically from 0 to 4), and theiralertness/sleepiness on a 100-mm visual analog scale


Table 1 Means and statistical results of comparisons among drug conditions for all measures for which there was a significant main effect of drugcondition in the overall ANOVA (ANCOVA for the working memory tasks)

Measure PL TRZ AMP TRZ/AMP

0.25 0.50 20 30 0.25/20 0.25/30 0.50/20 0.50/30

Participant ratings: sedationSleepiness (visual analog scale) 1.17 19.47 23.97 −10.03 −17.32 −5.18*c −10.10*c −1.31*c −2.94*c

Sedating/depressant 0.35 1.19 1.85 0.32 0.34 0.68*c 0.50*c 0.87*p 0.83*p

Sleepy −0.14 0.93 1.37 −0.37 −0.89 −0.12*c −0.28*c −0.24*c −0.56*c

Fatigued 0.21 0.39 0.42 0.08 −0.03 0.05* 0.16 0.26 0.14Tired/lazy 0.08 0.72 0.84 −0.15 −0.28 0.08*c −0.01*c 0.11*c 0.19*c

Psychomotor performance tasksBalance: s balanced 98.39 70.16 45.50 106.26 104.50 93.48*c 86.41*c 58.93n 60.54n

Circular lights: no. of responses 97.63 89.84 75.91 104.45 104.75 101.31*c 101.44*c 91.02*c 93.85*c

DSST: number attempted 101.75 87.59 78.77 106.69 108.49 96.70*c 96.48*c 89.65*p 88.78*p

DSST: proportion correct 100.33 96.95 91.40 101.41 102.06 100.19 100.42 98.51*c 97.14*c

DSST: estimate (pre) minus actual −3.28 7.99 14.06 −2.72 −3.03 2.76p 2.62p 9.51n 11.53n

Working memory tasksN-back: hit rate 102.04 97.63 88.18 99.73 100.07 98.67 99.66 91.82n 95.57*p

N-back: false alarm rate 196.84 161.48 309.11 157.14 173.12 189.61 167.39 196.01*c 298.31n

N-back: d’ (sensitivity) 101.11 99.18 79.58 101.16 100.43 96.33 97.93 86.32n 90.01*p

N-back: RT (ms) 96.31 105.58 111.51 93.64 95.99 104.44n 107.71n 118.89*n 119.24*n

Sternberg: proportion correct 99.73 94.52 89.27 101.76 102.58 101.68* 101.90* 94.65p 95.70*c

Sternberg: RT (ms) 103.83 114.62 126.17 98.69 97.74 113.51n 107.22p 125.02n 116.03n

Episodic memory tasksStudy: proportion correct 0.97 0.95 0.92 0.97 0.97 0.96 0.96 0.94n 0.92n

Free recall: no. correct 16.89 7.56 4.24 18.56 18.89 11.78*p 11.67*p 6.33n 6.28n

Recognition: hit rate 0.85 0.72 0.67 0.90 0.90 0.81*c 0.81*c 0.75n 0.68n

Recognition: false alarm rate 0.19 0.29 0.39 0.22 0.16 0.22*c 0.23p 0.33n 0.32n

Recognition: d’ (sensitivity) 2.12 1.26 0.87 2.21 2.56 1.88*c 1.83*c 1.32*p 1.09n

Gamma (metamemory) 0.71 0.44 0.38 0.77 0.73 0.62*c 0.61*c 0.42n 0.34n

Participant ratings: overallStrength of drug effect 0.46 1.70 2.28 0.93 1.41 1.52 1.56 1.87 1.97Drug effect 0.43 1.30 1.82 0.76 1.05 1.18 1.33 1.55 1.60Liking 0.00 −0.21 0.18 0.44 0.75 0.95* 1.32* 0.80* 1.04*Like effects 0.15 0.37 0.62 0.71 0.98 1.04* 1.39* 1.18* 1.24*Good effects 0.14 0.47 0.61 0.82 1.01 0.96* 1.30* 1.12* 1.20*Participant ratings: arousalArousing/stimulant 0.05 0.02 0.14 0.42 0.83 0.51* 0.78* 0.70* 0.69*Excited 0.00 0.04 0.19 0.23 0.46 0.23 0.62* 0.47 0.67*Energetic −0.06 −0.16 −0.14 0.33 0.74 0.26 0.61* 0.36* 0.46*Participant ratings: other specificConfused 0.11 0.27 0.48 0.05 0.07 0.05* 0.16 0.19*c 0.21*c

Blurred vision 0.02 0.40 0.48 0.06 0.00 0.22p 0.05*c 0.27n 0.33n

Slurred speech 0.00 0.09 0.10 0.00 0.03 0.04 0.02 0.02 0.19Limp −0.02 0.21 0.50 0.16 0.10 0.29 0.22 0.50n 0.33n

Lightheaded/dizzy 0.14 0.35 0.65 0.25 0.41 0.46 0.54 0.78n 0.77n

Unsteady 0.10 0.26 0.71 0.08 0.21 0.29 0.28 0.55n 0.60n

Difficulty concentrating 0.11 0.79 1.21 0.12 −0.06 0.24*c 0.31*c 0.61*p 0.60*p

Mentally slow 0.19 0.67 1.04 0.22 0.05 0.20*c 0.24*c 0.43*c 0.53*c

Limbs heavy 0.11 0.33 0.66 0.19 0.18 0.46 0.51 0.35p 0.46n

Forgetful 0.09 0.42 0.47 0.12 0.08 0.09*c 0.08*c 0.21p 0.15*c

Talkative −0.06 0.02 0.17 0.37 0.59 0.34 0.88* 0.67* 0.81*Comfortable −0.09 0.08 −0.07 0.37 0.22 0.41 0.44 0.23 0.24Relaxed −0.07 0.10 −0.11 0.18 0.07 0.66* 0.46 0.29 0.61*Restless/keep moving 0.06 0.06 0.06 0.16 0.34 0.15 0.42* 0.33* 0.37*Shakey/jittery 0.00 0.00 0.00 0.12 0.41 0.20 0.25* 0.27* 0.23


(labeled very alert and sleepy at the left and right extremes,respectively).

Physiological measures

Blood pressure (systolic and diastolic) and heart rate weremeasured using the Sentry II system (NBS Medical, CostaMesa, CA). Respiration rate was measured manually (byobservation over a 15-s duration).

Data analysis

Data from two participants were missing for severalmeasures in the TRZ 0.50 condition at two timepointsbecause of strong drug effects that prevented them fromcompleting assessments at those timepoints, and data fromone participant were missing for the DSST in the AMP 30condition at one timepoint because of a technical problem.Gamma correlations could not be computed for twoparticipants in at least one drug condition because theyfailed to use the entire confidence rating scale. Therefore,Proc mixed (a procedure in SAS that models missing data)was used for all analyses. For participant ratings andphysiological measures, the analyzed data were differencescores from predrug values. For working memory andpsychomotor performance tasks, the analyzed data werepercentages of predrug values. Area under the timecoursecurve (AUC) values were calculated for each participant bythe method of trapezoids.

Standard analyses Data were analyzed by repeated mea-sures analysis of variance (ANOVA) with drug condition(PL; TRZ 0.25, 0.50; AMP 20, 30; and TRZ/AMP 0.25/20,0.25/30, 0.50/20, and 0.50/30) and time (see above forspecific timepoints) as factors. A second set of analyses wasconducted using repeated measures one factor ANOVAwith drug condition as the factor; data analyzed were data

from the episodic memory/metamemory tasks administeredonly at a single timepoint and AUC data. Data from theworking memory tasks were analyzed by analysis ofcovariance (ANCOVA) with the corresponding nonmemorycontrol condition as covariate and drug condition andmemory load (1, 2, and 3 back; n-back task) or delay (0,12 s; Sternberg task) as factors. Data from the free recalland recognition memory tests initially were analyzed withthe order of the two tasks (which was counterbalancedacross participants; see “Materials and methods”) as abetween subjects factor. However, given that there wereno significant interactions between drug condition andtask order, analyses in which data were collapsed acrossthe two task orders are reported. For all statistical tests,p≤0.05 was considered significant. Significant maineffects and interactions were followed up with simpleeffects tests as appropriate and modified Bonferronicorrections were used (cf. Keppel 1991). Given that thefocus of the study was AMP’s reversal of TRZ’s effects,comparisons among doses were limited to the following:

1. TRZ vs PL (i.e., TRZ 0.25 vs PL; TRZ 0.50 vs PL)2. AMP vs PL (i.e., AMP 20 vs PL; AMP 30 vs PL)3. TRZ/AMP vs TRZ (i.e., TRZ/AMP 0.25/20 vs TRZ

0.25; TRZ/AMP 0.25/30 vs TRZ 0.25; TRZ/AMP 0.50/20 vs TRZ 0.50; TRZ/AMP 0.50/30 vs TRZ 0.50)

Measures for which the TRZ vs PL comparison wassignificant were categorized in terms of the degree ofreversal of the TRZ effect (complete, partial, and none) ineach of the relevant TRZ/AMP combination conditionsusing the following definitions (measures with an AMP vsPL comparison in the same direction as the significant TRZvs PL comparison were excluded from categorization):

1. Complete=both of the following are true:

a. TRZ/AMP vs TRZ was significantb. TRZ/AMP vs PL was not significant

Table 1 (continued)

Measure PL TRZ AMP TRZ/AMP0.25 0.50 20 30 0.25/20 0.25/30 0.50/20 0.50/30

PhysiologicalSystolic blood pressure 1.17 −0.96 0.15 11.67 19.81 15.38* 18.68* 14.79* 19.54*Diastolic blood pressure −1.55 −0.98 −0.10 7.72 13.22 4.68* 11.72* 9.87* 13.90*Heart rate −5.05 −2.39 2.05 0.94 0.39 1.18 0.40 4.98 6.40

For participant ratings and physiological measures, the analyzed data were difference scores from predrug values. For working memory andpsychomotor performance tasks, the analyzed data were percentages of predrug values. AUC values are shown for all measures except theepisodic memory tasks, which were administered at a single timepoint. Bold indicates a significant difference between active drug and PL. Anasterisk (*) indicates a significant difference between TRZ/AMP and the corresponding TRZ dose. For TRZ/AMP combination doses, c indicatescomplete reversal by AMP of TRZ’s effects, p indicates partial reversal, n indicates no reversal (see text for definitions), and the absence of c, p,or n indicates that the combination dose was excluded from categorization for that measure (if the corresponding TRZ vs PL comparison was notsignificant, the combination doses were excluded from categorization; measures with an AMP vs PL comparison in the same direction as thesignificant TRZ vs PL comparison were also excluded from categorization; see text for details).


2. Partial=only one of the following is true:

a. TRZ/AMP vs TRZ was significantb. TRZ/AMP vs PL was not significant

3. None=both of the following are true:

a. TRZ/AMP vs TRZ was not significantb. TRZ/AMP vs PL was significant

z-Score standardized analyses To facilitate the comparisonof AMP’s reversal of TRZ’s effects across measures, anadditional set of analyses was conducted on data from thesedative and memory measures (see Table 2 for specificmeasures). For purposes of this analysis, sedative measuresincluded participant ratings of sedative effects and psycho-motor performance tasks. To limit the number of measures,a single measure was selected for assessments with multipledependent variables (e.g., for the DSST, only number oftrials attempted was included). The analyzed data for theassessments administered repeatedly were AUC. Analyzeddata for the working memory tasks were the control-taskadjusted values from the ANCOVA collapsed acrossmemory load and delay as appropriate.

Data from these measures were first converted toz scores for each participant. Next, difference scores were

computed for each participant for each measure between thePL value for that measure and the value of each of thefollowing doses for that measure: TRZ 0.25, 0.50 (indicesof the TRZ effect for that measure) and TRZ/AMP 0.25/20,0.25/30, 0.50/20, 0.50/30 (indices of the combination effectfor that measure). To ensure that the direction of the effectwas consistent across measures, difference scores formeasures in which a TRZ effect is indicated by a highervalue for TRZ vs PL (e.g., RT measures; participant ratingsof sedative effects), were computed by subtracting the PLvalue from the active drug value, whereas difference scoresfor measures in which a TRZ effect is indicated by a lowervalue for TRZ vs PL (e.g., accuracy measures) werecomputed by subtracting the active drug value from thePL value. The following sets of planned comparisons wereconducted on these difference scores as follows:

1. TRZ/AMP vs TRZ for each measure (i.e., TRZ/AMP0.25/20 vs TRZ 0.25; TRZ/AMP 0.25/30 vs TRZ 0.25;TRZ/AMP 0.50/20 vs TRZ 0.50; TRZ/AMP 0.50/30 vsTRZ 0.50). A significant difference suggests significantreversal by AMP of the TRZ effect for that measure.

2. Comparisons between pairs of specific memory mea-sures at each of the six doses. A significant differencebetween measures at the combination doses suggestsdifferential reversal by AMP of the TRZ effect

Table 2 Mean difference scores for TRZ minus PL (columns 2–3) and TRZ/AMP minus PL (columns 4–7) calculated from z-score standardizeddata for the sedative and memory measures and results of planned comparisons conducted on these data

Measure TRZ minus PL TRZ/AMP minus PL

0.25 0.50 0.25/20 0.25/30 0.50/20 0.50/30

SedativeParticipant ratings: sedationSleepiness (visual analog scale) 0.82 1.06 −0.28 −0.50 −0.11 −0.18Sedating/depressant 0.96 1.70 0.38 0.17 0.59 0.55Sleepy 0.88 1.29 0.01 −0.11 −0.09 −0.35Tired/lazy 0.73 0.90 0.00 −0.10 0.04 0.12Psychomotor performance tasksBalance: s balanced 0.91 1.74 0.16 0.39 1.28 1.22Circular lights: no. responses 0.58 1.62 −0.27 −0.28 0.49 0.28DSST: no. attempted 1.15 1.91 0.41 0.43 0.98 1.05

MemoryWorking memory tasksN-back: d’ (sensitivity) 0.53a 1.84a 0.40a 0.29a 1.15a 0.91ab

N-back: RT (ms) 0.90a 1.52ab 0.69a 0.73a 1.79c 1.93c

Sternberg: proportion correct 0.62a 1.54ab −0.32b −0.33b 0.49b 0.47b

Sternberg: RT (ms) 0.87a 1.77a 0.50a 0.22ab 1.34ac 0.99ab

Episodic memory tasksFree recall: no. correct 1.07a 1.50ab 0.59a 0.60a 1.21ac 1.22a

Recognition: d’ (sensitivity) 0.95a 1.38ab 0.27ab 0.32a 0.89ab 1.13a

Gamma (metamemory) 0.87a 1.05b 0.29ab 0.33a 0.90ab 0.94ab

Bold indicates a significant difference between TRZ/AMP and the corresponding TRZ dose (i.e., significant reversal by AMP of the TRZ effectfor that measure). In the memory section of the table, letters a, b, and c are used to indicate comparisons between pairs of the seven memorymeasures at each dose; within the same column, any two means designated with the same letter are not significantly different from each other.


(assuming there is no significant difference betweenthese measures at the corresponding TRZ alone dose).

3. Comparisons betweenmemorymeasures (collapsed acrossall memory measures) and sedative measures (collapsedacross all sedative measures) at each of the six doses.These comparisons test the hypothesis that the combina-tion effect is significantly greater (i.e., less reversal) formemory vs sedative measures (whereas no difference ishypothesized between the measures for the TRZ effect).

Results

Timecourse of effects

Before presenting the dose effect data, timecourse data arebriefly summarized here, and representative timecoursefunctions are shown in Fig. 1. As expected, both TRZ andAMP produced orderly time-related increases relative to PLin participant ratings of overall drug effect (e.g., strength of

Fig. 1 Timecourse functions(TRZ and AMP, alone and incombination) for participant rat-ing of strength of drug effect,performance on the balancetask, and systolic blood pres-sure. x-Axis, time in minutesafter the first capsule adminis-tration timepoint; d-amphet-amine was administered at thefirst capsule administrationtimepoint, and triazolam wasadministered 30 min later; 0indicates predrug. Unadjustedmeans + 1 S.E.M are shown(excluding missing data). Filledsymbols indicate active drugvalues that are significantly dif-ferent from the correspondingPL value at the same timepoint


drug effect). TRZ also produced time-related decrements inpsychomotor performance (e.g., balance) and AMP pro-duced increases in physiological measures (e.g., systolicblood pressure). Given that various assessments wereadministered at different times, it is difficult to determinethe exact time of peak drug effects; however, for both TRZand AMP, effects appeared to peak somewhere betweenapproximately 75 and 140 min after the first capsuleadministration timepoint.

Dose effects

Results across all drug conditions are shown in Table 1.

Effects of TRZ alone TRZ produced significant dose-relatedincreases relative to PL in participant ratings of sedativeeffects and significant decrements in psychomotor perfor-mance (Fig. 2). TRZ also significantly impaired DSSTperformance estimates (preperformance only) such thatbefore performing the DSST, participants underestimatedtheir degree of impairment. The ANCOVAs on the workingmemory tasks revealed no significant interaction betweendose and memory load (n-back) or delay (Sternberg).Therefore, data are shown in Table 1 and Fig. 3 collapsedacross memory load and delay. As expected, TRZ producedsignificant dose-related decrements in working memoryaccuracy (TRZ 0.50 only) and increases in RT on bothtasks. During the study phase of the episodic memory task,TRZ (0.50 only) significantly decreased the proportion ofcorrect categorization responses relative to PL. On the freerecall test, TRZ produced a significant dose-related de-crease in the number of correct responses (Fig. 4) but didnot significantly affect the number of intrusion responsesrelative to PL (data not shown). On the recognition memorytest, TRZ produced significant dose-related decreases in hitrate and d’ (Fig. 4) and a significant increase in false alarmrate relative to PL; TRZ did not affect the response biasmeasure C (data not shown). TRZ also produced asignificant dose-related decrease in the Gamma correlation,indicating impaired metamemory.

TRZ also produced significant dose-related increases inparticipant ratings of overall drug effects and a number ofspecific items. TRZ produced a significant increase in heartrate relative to PL (TRZ 0.50 only) but did not significantlyaffect blood pressure or respiration rate.

Effects of AMP alone AMP (30 only) produced significantincreases in all participant ratings of arousal and significantdecreases in some participant ratings of sedative effectsrelative to PL (Fig. 2). AMP also produced significant dose-related increases in participant ratings of overall drugeffects (including liking and good effects) and of a numberof specific items. AMP (30 only) significantly enhanced

performance on the DSST (number attempted) and circularlights relative to PL but did not significantly affect balanceperformance. AMP did not significantly enhance workingmemory, performance during the study phase of theepisodic memory task, free recall, or metamemory but didproduce a significant increase in d’ on the recognitionmemory test (AMP 30 only). AMP produced significantdose-related increases relative to PL in both systolic anddiastolic blood pressure but did not significantly affect heartrate or respiration.

Fig. 2 AUC dose–effect functions (TRZ, AMP, and TRZ/AMP) forparticipant rating of sleepiness (visual analog scale) and performanceon the circular lights task. Unadjusted means ± 1 S.E.M are shown(excluding missing data). Filled symbols indicate active drug valuesthat are significantly different from PL; an asterisk indicates asignificant difference between TRZ/AMP and the correspondingTRZ dose


Reversal by AMP of TRZ’s effects AMP produced completereversal (as defined above) of TRZ’s effects at all four dosecombinations on all participant ratings of sedative effects(Fig. 2) except sedating/depressant that showed completereversal at 0.25/20 and 0.25/30 but only partial reversal at0.50/20 and 0.50/30. AMP produced complete reversal at allfour combinations for circular lights (Fig. 2) but only at0.25/20 and 0.25/30 for balance. For the DSST number oftrials attempted, AMP produced complete reversal at 0.25/20 and 0.25/30 but only partial reversal at 0.50/20 and 0.50/30. For DSST proportion correct, AMP produced completereversal at 0.50/20 and 0.50/30 (the only relevant combi-nations because TRZ 0.25 was not significantly differentfrom PL). For DSST preperformance estimates, there wasonly partial reversal at 0.25/20 and 0.25/30 and no reversalat 0.50/20 and 0.50/30. Participant ratings of other specific

effects varied in the degree of reversal of TRZ’s effects byAMP, with ratings of mental state (confused, difficultyconcentrating, mentally slow, and forgetful) tending toproduce more reversal than ratings of somatic state (blurredvision, limp, lightheaded/dizzy, unsteady, and limbs heavy).

AMP did not produce complete reversal of TRZ’s effectsat all four dose combinations on any of the memorymeasures. The memory measures ranged in degree ofreversal from none across all relevant combinations (i.e.,combinations for which the corresponding TRZ dose wassignificantly different from PL; n-back working memorytask, RT) to none across none of the relevant combinations(Sternberg working memory, proportion correct; only 0.50/20 and 0.50/30 relevant because TRZ 0.25 was notsignificantly different from PL) with most of the memorymeasures showing an overall pattern of partial reversal

Fig. 3 AUC dose–effect functions (TRZ, AMP, and TRZ/AMP) formeasures of performance on the working memory tasks (n-back andSternberg). Control-task adjusted means from the ANCOVA collapsed

across memory load (n-back) and delay (Sternberg) are shown.Symbols are identical to those described for Fig. 2


(Figs. 3 and 4). It should be noted that for the n-backworking memory task, RT was actually significantly longer(i.e., worse performance) for TRZ/AMP 0.50/20 and 0.50/30 relative to TRZ 0.50.

Results of the planned comparisons conducted on thedifference scores calculated for each participant from z-score standardized data for the sedative and memorymeasures are shown in Table 2. The TRZ/AMP vs TRZcomparisons for each measure revealed an overall pattern ofreversal similar to that of the standard analyses describedabove. Thus, participant ratings of sedative effects andcircular lights showed significant reversal (i.e., a significantTRZ/AMP vs TRZ difference) at all four combinations(except sedating/depressant that showed significant reversalat all but the 0.25/20 combination), whereas balance showedsignificant reversal only at 0.25/20. Again, the memorymeasures ranged in degree of reversal from no significantreversal at any combination (n-back working memory task,RT; free recall, number of correct responses; gammacorrelation) to significant reversal at all four combinations(Sternberg working memory, proportion correct).

Comparisons between pairs of memory measures at eachof the four combinations (see the memory section ofTable 2) revealed that the combination effect was signifi-cantly greater (i.e., less reversal) for RT on the n-backworking memory task relative to all other measures for oneor both of the 0.50 combinations (i.e. 0.50/20, 0.50/30), andthat the combination effect was significantly smaller (i.e.,more reversal) for proportion correct on the Sternbergworking memory task relative to all other measures for atleast one of the four combinations.

Results of the comparisons between memory measures(collapsed across all seven memory measures in Table 2)and sedative measures (collapsed across all seven sedativemeasures in Table 2) at each of the six doses are shown inFig. 5. As predicted, the combination effect was signifi-cantly greater (i.e., less reversal) for memory vs sedativemeasures at each of the four combinations, whereas therewas no significant difference between the measures for theTRZ alone effect at either TRZ dose.

Discussion

Effects of triazolam and d-amphetamine alone

The observed effects of triazolam alone (increases inparticipant ratings of sedative effects, impaired psychomo-tor performance, working memory, episodic memory, andmetamemory) replicate results of previous studies withtriazolam and other benzodiazepines (for reviews, seeCurran 1991, 2000; Hollister et al. 1993; Woods et al.1992). Likewise, effects of d-amphetamine alone (increasesin participant ratings of arousal and blood pressure,enhanced psychomotor performance) replicate previousfindings with amphetamine (for reviews, see Solanto1998; Weiss and Laties 1962) including those of ourprevious interaction study (Mintzer and Griffiths 2003).Consistent with previous demonstrations that amphetamineenhances the speed but not the accuracy of psychomotorperformance (Fleming et al. 1995; Mintzer and Griffiths2003; Rapoport et al. 1980; Robbins and Everitt 1982), d-amphetamine’s enhancement of DSST performance wasreflected in an increase in the number of digits attemptedbut not in the proportion of digits correct (Table 1).

Previous studies examining amphetamine’s acute effectson memory in healthy adults do not yield a consistentpattern of results. In early studies focused on the acquisitionof stimulus–response relationships, there are reports ofenhancement by amphetamine (i.e., fewer trials to learningcriterion; Weitzner 1965), no effect (Kornetsky 1958; Hurstet al. 1969), and even impairment (Burns et al. 1967).Although results of one study suggest no effect on short-term/working memory (Soetens et al. 1995), results of

Fig. 4 Dose–effect functions for measures of performance on the freerecall and recognition memory tests. Unadjusted means ± 1 S.E.M areshown (excluding missing data). Symbols are identical to thosedescribed for Fig. 2


another study (Kennedy et al. 1990) suggest enhancement.Results of recent neuroimaging and genetic studies suggestthat individual differences in amphetamine’s effects onworking memory may be related to differences in dopaminefunction, which has been shown to vary across genotypes(Mattay et al. 2000, 2003). Specifically, an inverted U-shaped function seems to characterize the relationshipbetween dopamine function and working memory perfor-mance such that performance is optimal at moderate levelsof dopamine and worse at lower and higher dopaminelevels. Thus, working memory is enhanced (as evidencedby enhanced performance on the n-back task and/orenhanced efficiency of relevant prefrontal cortical net-works) after acute d-amphetamine administration in indi-viduals with lower baseline prefrontal dopamine functionand unchanged or reduced in individuals with higherbaseline prefrontal dopamine function. In the present study,d-amphetamine did not enhance working memory perfor-mance on either the n-back or Sternberg task. Although thelack of enhancement of the accuracy measures may be anartifact of the near-ceiling accuracy on both tasks in theplacebo condition, the lack of enhancement of RT isunlikely due to ceiling effects. To test the hypothesis thatthe degree of enhancement by d-amphetamine of workingmemory performance is inversely related to baseline levelof working memory performance as suggested by previousstudies, we performed secondary analyses on data from then-back (d’; control task-adjusted values, collapsed across

memory load) and Sternberg (proportion correct; control-task adjusted values, collapsed across delay) workingmemory tasks. Specifically, for each measure, we calculatedPearson correlations between the value at the predrugtimepoint (collapsed across all nine drug conditions) andan ‘enhancement score’ (calculated for each participant asthe AUC value for d-amphetamine minus the AUC valuefor placebo, divided by the AUC value for placebo as abaseline); Bonferroni corrections were used. For the n-backtask (similar to the task used by Mattay et al. 2000), therewas a significant negative correlation between baselineperformance and the d-amphetamine enhancement score forthe 20-mg d-amphetamine dose (r=−0.67, p=0.002) aspredicted. The correlation was not significant for the 30-mgd-amphetamine dose. There were no significant correlationsfor the Sternberg task.

Few studies have examined amphetamine’s effects onepisodic memory. Rapoport et al. (1980) reported enhance-ment of free recall but not cued recall performance, whereasHurst et al. (1969) reported enhancement of cued recallperformance. Soetens et al. (1995) reported enhancement ofboth free recall and recognition memory performance;however, the timing of stimulus presentation relative todrug administration was designed to measure effects onconsolidation processes rather than on encoding or retriev-al. In the present study, d-amphetamine did not affect freerecall or metamemory but significantly enhanced d’ on therecognition memory test (only at the 30-mg dose). The

Fig. 5 Difference scores (TRZminus PL; TRZ/AMP minusPL) calculated from z-scorestandardized data for the seda-tive measures (collapsed acrossall seven sedative measures inTable 2) and memory measures(collapsed across all sevenmemory measures in Table 2).Unadjusted means + 1 S.E.Mare shown (excluding missingdata). An asterisk indicates asignificant difference betweenthe sedative and memory vari-ables in that condition


study phase of the episodic memory task was timed tocoincide with the anticipated time of peak effects for bothtriazolam and amphetamine to measure effects on encoding.Inconsistencies in effects of amphetamine on memoryacross studies may be related to individual differences inresponse to amphetamine. In addition to the effects ofdifferences in dopamine function discussed above, there isevidence that the response to amphetamine varies as afunction of gender and hormone levels (Justice and de Wit2000a,b; White et al. 2002) although the effects of thesevariables have not been tested on memory specifically.Consistent with the results of our previous interaction study(Mintzer and Griffiths 2003), d-amphetamine did not affectmetamemory in the present study. To our knowledge, otherstudies have not examined effects of amphetamine onmetamemory.

Reversal by d-amphetamine of triazolam’s effects

In the standard analyses, d-amphetamine produced acomplete reversal of triazolam’s effects at all four dosecombinations on all but one participant rating of sedativeeffects and complete or partial reversal of triazolam’seffects on all psychomotor performance measures exceptthe balance task. The planned comparisons conducted on z-score standardized data revealed a similar pattern of resultsfor the sedative measures and support the overall conclu-sion that the sedative effects of triazolam were reversed byd-amphetamine. The finding of complete or partial reversalfor ratings of specific aspects of mental state associatedwith sedation (confused, difficulty concentrating, mentallyslow, and forgetful) is also consistent with this conclusion.Interestingly, ratings of somatic state (blurred vision, limp,lightheaded/dizzy, unsteady, and limbs heavy) showed lessreversal than ratings of mental state. This pattern isconsistent with the finding of less reversal for the balancetask (a measure of basic motor coordination) and suggeststhat the physical aspects of sedation may be dissociatedfrom the mental aspects.

As discussed in detail below, the memory measuresvaried in the degree of reversal by d-amphetamine oftriazolam’s effects. However, it is important to note that,overall, the memory measures exhibited a pattern of lessreversal than the sedative measures. This conclusion issupported by both the standard and z-score standardizedanalyses. Specifically, in the standard analyses, d-amphet-amine did not produce complete reversal of triazolam’seffects at all four dose combinations on any of the memorymeasures, and in the z-score standardized analyses, thecombination effect was significantly greater (i.e., lessreversal) for memory vs sedative measures at each of thefour combinations (whereas there was no significantdifference between the measures for the triazolam alone

effect at either triazolam dose; Fig. 5). Thus, results of thepresent dose–response study are consistent with those of ourprevious single-dose interaction study (Mintzer and Griffiths2003) and provide further evidence that the amnestic andsedative effects of benzodiazepines can be dissociated.

In the standard analyses, the memory measures ranged indegree of reversal from none across all relevant combina-tions (n-back working memory task, RT) to none acrossnone of the relevant combinations (Sternberg workingmemory task, proportion correct) with most of the memorymeasures showing an overall pattern of partial reversal(Table 1; Figs. 3 and 4). In general, there was relativelymore reversal by d-amphetamine at the 0.25-mg triazolamdose than at the 0.50-mg dose, but there were differencesamong relevant measures at the 0.25-mg dose such thatsome measures showed complete reversal (recognition, hitrate, d’; gamma correlation), whereas others showed partialreversal (Sternberg, RT; free recall, number correct;recognition, false alarm rate) or none (n-back, RT). Thus,the dose–effect interaction design enables differences toemerge between measures in the relative contribution oftriazolam-induced sedative effects to triazolam-inducedmemory effects.

The present study was designed to focus on therelationship between the sedative and amnestic effects ofbenzodiazepines. However, it is important to note that inaddition to sedation, benzodiazepines produce other effectsthat may also contribute to their memory-impairing effects.For example, benzodiazepines have been shown to impairfocused and selective attention (for a review, see Buffett-Jerrott and Stewart 2002). The nonmemory control con-ditions used in the two working memory tasks weredesigned to control for effects on attention. However, therewere no analogous controls for the episodic memory tasks.Thus, it is possible that benzodiazepine-induced effects onattention may have contributed to the observed memoryimpairment in these tasks. Given that sedation may also playa role in attentional effects, it is also possible that thecontribution of sedation to memory impairment was actuallymediated by attentional mechanisms. Further adding to thecomplexity, d-amphetamine has been shown to enhanceattention (for reviews, see Solanto 1998; Weiss and Laties1962). The present study was not designed to disentangle thecomplex interrelationships among sedation, attention, mem-ory, and other processes. However, these issues are intrigu-ing and warrant exploration in future studies.

From a methodological perspective, the z-score stan-dardized analysis technique is interesting and enableseffects (both the triazolam effect and the triazolam/amphetamine combination effect in this study) to bedirectly compared statistically across different measuresusing the same metric. For example, direct comparison ofthe combination effect for the collapsed sedative measures


vs collapsed memory measures (shown in Fig. 5) enabledthe sedative and memory measures to be dissociatedstatistically, thus strengthening the conclusion that thesedative and amnestic effects of benzodiazepines can bedissociated. The z-score standardized analysis techniquealso enabled statistical comparisons to be made betweenpairs of memory measures. The conclusions of relativelyless reversal for RT on the n-back working memory taskand relatively more reversal for proportion correct on theSternberg working memory task were supported both by thestandard analyses and by the comparisons conducted onthe z-score standardized data and appear to be reliable.Given that we did not specifically predict these differencesbetween memory measures, we will not speculate on pos-sible underlying mechanisms. However, these differencesare intriguing and warrant exploration in future studies.

In summary, using both standard analyses and a novel z-score standardized analysis technique, this dose–effectinteraction study provides strong evidence that benzodi-azepines have specific effects on memory that are notmerely a by-product of the drugs’ sedative effects, and thatthe degree to which sedative effects contribute to theamnestic effects varies as a function of the particularmemory process being assessed.

Acknowledgment This project was supported by the NationalInstitute on Drug Abuse Research Grant DA-11936. The authorsthank Crystal Barnhouser, Kristina Burns, and Koty Nadeau forprotocol management, John Yingling for computer programmingassistance and technical support, and Linda Felch and Paul Nuzzofor assistance with data analysis.

References

Baddeley AD (1992) Working memory: the interface betweenmemory and cognition. J Cogn Neurosci 4:281–288

Bishop KI, Curran HV (1995) Psychopharmacological analysis ofimplicit and explicit memory: a study with lorazepam and thebenzodiazepine antagonist flumazenil. Psychopharmacology121:267–278

Blin O, Simon N, Jouve E, Habib M, Gayraud D, Durand A,Bruguerolle B, Pisano P (2001) Pharmacokinetic and pharmaco-dynamic analysis of sedative and amnesic effects of lorazepam inhealthy volunteers. Clin Neuropharmacol 24:71–81

Buffett-Jerrott SE, Stewart SH (2002) Cognitive and sedative effectsof benzodiazepine use. Curr Pharm Des 8:45–58

Burns JT, House RF, Fensch FC, Miller JG (1967) Effects ofmagnesium pemoline and dextroamphetamine on human learn-ing. Science 155:849–851

Curran HV (1991) Benzodiazepines, memory and mood: a review.Psychopharmacology 105:1–8

Curran HV (2000) Psychopharmacological approaches to humanmemory. In: Gazzaniga MS (ed) The new cognitive neuro-sciences, 2nd edn. Boston, MIT, pp 797–804

Curran HV, Birch B (1991) Differentiating the sedative and amnesticeffects of benzodiazepines: a study with midazolam and the

benzodiazepine antagonist, flumazenil. Psychopharmacology103:519–523

Curran HV, Bond A, O’Sullivan G, Bruce M, Marks I, Lelliot P, ShineP, Lader M (1994) Memory functions, alprazolam and exposuretherapy: a controlled longitudinal study of patients with agora-phobia and panic disorder. Psychol Med 24:969–976

Curran HV, Pooviboonsuk P, Dalton JA, Lader MH (1998) Differen-tiating the effects of centrally acting drugs on arousal and memory:an event-related potential study of scopolamine, lorazepam anddiphenhydramine. Psychopharmacology 135:27–36

D’Esposito M, Postle BR, Ballard D, Lease J (1999) Maintenanceversus manipulation of information held in working memory: anevent-related fMRI study. Brain Cogn 41:66–86

Flavell JH (1971) What is memory development the development of ?Hum Dev 14:272–278

Fleming K, Bigelow LB, Weinberger DR, Goldberg TE (1995)Neuropsychological effects of amphetamine may correlate withpersonality characteristics. Psychopharmacol Bull 31:357–362

Ghoneim MM (2004a) Drugs and human memory (part 1): clinical,theoretical, and methodologic issues. Anesthesiology 100:987–1002

Ghoneim MM (2004b) Drugs and human memory (part 2): clinical,theoretical, and methodologic issues. Anesthesiology 100:1277–1297

Ghoneim MM, Mewaldt SP, Berie JL, Hinrichs V (1981) Memory andperformance effects of single and 3 week administration ofdiazepam. Psychopharmacology 73:147–151

Goodman LA, Kruskal WH (1954) Measures of association for cross-classifications. J Am Stat Assoc 49:732–764

Green JF, McElholm A, King DJ (1996) A comparison of thesedative and amnestic effects of chlorpromazine and lorazepam.Psychopharmacology 128:67–73

Griffiths RR, Bigelow GE, Liebson I (1983) Differential effects ofdiazepam and pentobarbital on mood and behavior. Arch GenPsychiatry 40:865–873

Hollister LE, Müller-Oerlinghausen B, Rickels K, Shader RI (1993)Clinical uses of benzodiazepines. J Clin Psychopharmacol13:1S–169S

Hommer D, Weingartner H, Breier A (1993) Dissociation ofbenzodiazepine-induced amnesia from sedation by flumazenil.Psychopharmacology 112:455–460

Hurst PM, Radlow R, Chubb NC, Bagley SK (1969) Effects of D-amphetamine on acquisition, persistence, and recall. Am JPsychol 82:307–319

Jonides J, Schumacher EH, Smith EE, Lauber EJ, Awh E, MinoshimaS, Koeppe RA (1997) Verbal working memory load affectsregional brain activation as measured by PET. J Cogn Neurosci9:462–475

Justice AJH, de Wit H (2000a) Acute effects of d-amphetamine duringthe early and late follicular phases of the menstrual cycle inwomen. Pharmacol Biochem Behav 66:509–515

Justice AJH, de Wit H (2000b) Acute effects of estradiol pretreatmenton the response to d-amphetamine in women. Neuroendocrinology71:51–59

Kennedy RS, Odenheimer RC, Baltzley DR, Dunlap WP, Wood CD(1990) Differential effects of scopolamine and amphetamine onmicrocomputer-based performance tests. Aviat Space EnvironMed 61:615–621

Keppel G (1991) Design and analysis: a research’s handbook, 3rd edn.Prentice Hall, Englewood Cliffs

Kirk T, Roache JD, Griffiths RR (1990) Dose–response evaluation ofthe amnestic effects of triazolam and pentobarbital in normalsubjects. J Clin Psychopharmacol 10:160–167

Kornetsky C (1958) Effects of meprobamate, phenobarbital anddextroamphetamine on reaction time and learning in man. JPharmacol Exp Ther 123:216–219

Mattay VS, Callicott JH, Bertolino A, Heaton I, Frank JA, Coppola R,Berman KF, Goldberg TE, Weinberger DR (2000) Effects of


dextroamphetamine on cognitive performance and corticalactivation. NeuroImage 12:268–275

Mattay VS, Goldberg TE, Fera F, Hariri AR, Tessitore A, Egan MF,Kolachana B, Callicott JH, Weinberger DR (2003) Catechol O-methyltransferase val158-met genotype and individual variationin the brain response to amphetamine. Proc Natl Acad Sci USA100:6186–6191

McLeod DR, Griffiths RR, Bigelow GE, Yingling J (1982) Anautomated version of the digit symbol substitution test (DSST).Behav Res Meth Instrum 14:463–466

Metcalfe J, Shimamura AP (1994) Metacognition: knowing aboutknowing. Cambridge, MIT

Mintzer MZ, Griffiths RR (2003) Lorazepam and scopolamine: asingle-dose comparison of effects on human memory andattentional processes. Exp Clin Psychopharmacol 11:56–72

Mintzer MZ, Griffiths RR, Contoreggi C, Kimes AS, London ED, ErnstM (2001) Effects of triazolam on brain activity during episodicmemory encoding: a PET study. Neuropsychopharmacology25:744–756

Neave N, Reid C, Scholey AB, Thompson JM, Moss M, Ayre G,Wesnes K, Girdler NM (2000) Dose-dependent effects offlumazenil on cognition, mood, and cardio-respiratory physiolo-gy in healthy volunteers. Br Dent J 189:668–674

Nelson TO (1984) A comparison of current measures of theaccuracy of feeling of knowing predictions. Psychol Bull 95:109–133

Postle BR, Berger JS, D’Esposito M (1999) Functional neuroanatom-ical double dissociation of mnemonic and executive controlprocesses contributing to working memory performance. ProcNatl Acad Sci USA 96:12959–12964

Preston KL, Bigelow GE, Liebson IA (1989) Antagonist effects ofnalbuphine in opioid-dependent human volunteers. J PharmacolExp Ther 248:929–937

Rapoport JL, Buchsbaum MS, Weingartner H, Zahn TP, Ludlow C,Mikkelsen EJ (1980) Dextroamphetamine. Its cognitive andbehavioral effects in normal and hyperactive boys and normalmen. Arch Gen Psychiatry 37:933–943

Robbins TW, Everitt BJ (1982) Functional studies of the centralcatecholamines. Int Rev Neurobiol 23:303–365

Smirne S, Ferini-Strambi L, Pirola R, Tancredi O, Franceschi M, PintoP, Bareggi SR (1989) Effects of flunitrazepam on cognitivefunctions. Psychopharmacology 98:251–256

Snodgrass JG, Corwin J (1988) Pragmatics of measuring recognitionmemory: applications to dementia and amnesia. J Exp PsycholGen 117:34–50

Soetens S, Casaer S, D’Hooge R, Hueting JE (1995) Effect ofamphetamine on long-term retention of verbal material.Psychopharmacology 119:155–162

Solanto MV (1998) Neuropsychopharmacological mechanisms ofstimulant drug action in attention-deficit hyperactivity disorder:a review and integration. Behav Brain Res 94:127–152

Sternberg S (1969) The discovery of processing stages: extensions ofDonder’s method. Acta Psychol 30:276–315

Tata PR, Rollings J, Collins M, Pickering A, Jacobson RR (1994)Lack of cognitive recovery following withdrawal from long-termbenzodiazepine use. Psychol Med 24:203–213

Thorndike EL, Lorge I (1944) The teacher’s word book of 30,000words. Teacher’s College, Columbia University, New York

Tulving E (1972) Episodic and semantic memory. In: Tulving E,Donaldson W (eds) Organization of memory. New York,Academic, pp 381–403

Tulving E (1983) Elements of episodic memory. Oxford University,New York

Veselis RA, Reinsel RA, Feshchenko VA (2001) Drug-inducedamnesia is a separate phenomenon from sedation: electrophysi-ologic evidence. Anesthesiology 95:896–907

Veselis RA, Reinsel RA, Feshchenko VA, Wronski M (1997) Thecomparative amnestic effects of midazolam, propofol, thiopental,and fentanyl at equisedative concentrations. Anesthesiology87:749–764

Weingartner HJ, Sirocco K, Rawlings R, Joyce E, Hommer D (1995)Dissociations in the expression of the sedative effects oftriazolam. Psychopharmacology 119:27–33

Weiss B, Laties VG (1962) Enhancement of human performance bycaffeine and the amphetamines. Pharmacol Rev 14:1–36

Weitzner M (1965) Manifest anxiety, amphetamine, and performance.J Psychol 60:71–79

White TL, Justice AJH, de Wit H (2002) Differential subjective effectsof D-amphetamine by gender, hormone levels and menstrualcycle phase. Pharmacol Biochem Behav 73:729–741

Williams EJ (1949) Experimental designs balanced for the estimationof residual effects of treatments. Aust J Sci Res 2:149–168

Woods JH, Katz JL, Winger G (1992) Benzodiazepines: use, abuse,and consequences. Pharmacol Rev 44:151–347


Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Memory and Afetamine

Documents

visual analog

false alarm

appropriately

capsule administration

signal detection

balanced circular

impaired psychomotor

score standardized